BOYSWILLBEBOYS:GENDER,OVERCONFIDENCE...
Transcript of BOYSWILLBEBOYS:GENDER,OVERCONFIDENCE...
BOYS WILL BE BOYS GENDER OVERCONFIDENCEAND COMMON STOCK INVESTMENT
BRAD M BARBER AND TERRANCE ODEAN
Theoretical models predict that overcondent investors trade excessively Wetest this prediction by partitioning investors on gender Psychological researchdemonstrates that in areas such as nance men are more overcondent thanwomen Thus theory predicts that men will trade more excessively than womenUsing account data for over 35000 households from a large discount brokeragewe analyze the common stock investments of men and women from February 1991through January 1997 We document that men trade 45 percent more thanwomen Trading reduces menrsquos net returns by 265 percentage points a year asopposed to 172 percentage points for women
Itrsquos not what a man donrsquot know that makes him a fool but what he does knowthat ainrsquot so
Josh Billings nineteenth century American humorist
It is difcult to reconcile the volume of trading observed inequity markets with the trading needs of rational investors Ra-tional investors make periodic contributions and withdrawalsfrom their investment portfolios rebalance their portfolios andtrade to minimize their taxes Those possessed of superior infor-mation may trade speculatively although rational speculativetraders will generally not choose to trade with each other It isunlikely that rational trading needs account for a turnover rate of76 percent on the New York Stock Exchange in 19981
We believe there is a simple and powerful explanation forhigh levels of trading on nancial markets overcondence Hu-man beings are overcondent about their abilities their knowl-edge and their future prospects Odean [1998] shows that over-condent investorsmdashwho believe that the precision of theirknowledge about the value of a security is greater than it actually
We are grateful to the discount brokerage rm that provided us with thedata for this study and grateful to Paul Thomas David Moore Paine Webber andthe Gallup Organization for providing survey data We appreciate the commentsof Diane Del Guercio David Hirshleifer Andrew Karolyi Timothy LoughranEdward Opton Jr Sylvester Schieber Andrei Shleifer Martha Starr-McCluerRichard Thaler Luis Viceira and participants at the University of AlbertaArizona State University INSEAD the London Business School the University ofMichigan the University of Vienna the Institute on Psychology and Markets theConference on Household Portfolio Decision-making and Asset Holdings at theUniversity of Pennsylvania and the Western Finance Association Meetings Allerrors are our own Terrance Odean can be reached at (530) 752-5332 orodean] ucdavisedu Brad Barber can be reached at (530) 752-0512 orbmbarber] ucdavisedu
1 NYSE Fact Book for the Year 1999
copy 2001 by the President and Fellows of Harvard College and the Massachusetts Institute ofTechnologyThe Quarterly Journal of Economics February 2001
261
ismdashtrade more than rational investors and that doing so lowerstheir expected utilities Greater overcondence leads to greatertrading and to lower expected utility
A direct test of whether overcondence contributes to exces-sive market trading is to separate investors into those more andthose less prone to overcondence One can then test whethermore overcondence leads to more trading and to lower returnsSuch a test is the primary contribution of this paper
Psychologists nd that in areas such as nance men are moreovercondent than women This difference in overcondenceyields two predictions men will trade more than women and theperformance of men will be hurt more by excessive trading thanthe performance of women To test these hypotheses we partitiona data set of position and trading records for over 35000 house-holds at a large discount brokerage rm into accounts opened bymen and accounts opened by women Consistent with the predic-tions of the overcondence models we nd that the averageturnover rate of common stocks for men is nearly one and a halftimes that for women While both men and women reduce theirnet returns through trading men do so by 094 percentage pointsmore a year than do women
The differences in turnover and return performance are evenmore pronounced between single men and single women Singlemen trade 67 percent more than single women thereby reducingtheir returns by 144 percentage points per year more than dosingle women
The remainder of this paper is organized as follows Wemotivate our test of overcondence in Section I We discuss ourdata and empirical methods in Section II Our main results arepresented in Section III We discuss competing explanationsfor our results in Section IV and make concluding remarks inSection V
I A TEST OF OVERCONFIDENCE
IA Overcondence and Trading on Financial Markets
Studies of the calibration of subjective probabilities nd thatpeople tend to overestimate the precision of their knowledge[Alpert and Raiffa 1982 Fischhoff Slovic and Lichtenstein1977] see Lichtenstein Fischhoff and Phillips [1982] for a re-view of the calibration literature Such overcondence has been
262 QUARTERLY JOURNAL OF ECONOMICS
observed in many professional elds Clinical psychologists [Os-kamp 1965] physicians and nurses [Christensen-Szalanski andBushyhead 1981 Baumann Deber and Thompson 1991] invest-ment bankers [Stael von Holstein 1972] engineers [Kidd 1970]entrepreneurs [Cooper Woo and Dunkelberg 1988] lawyers[Wagenaar and Keren 1986] negotiators [Neale and Bazerman1990] and managers [Russo and Schoemaker 1992] have all beenobserved to exhibit overcondence in their judgments (For fur-ther discussion see Lichtenstein Fischhoff and Phillips [1982]and Yates [1990])
Overcondence is greatest for difcult tasks for forecastswith low predictability and for undertakings lacking fast clearfeedback [Fischhoff Slovic and Lichtenstein 1977 LichtensteinFischhoff and Phillips 1982 Yates 1990 Grifn and Tversky1992] Selecting common stocks that will outperform the marketis a difcult task Predictability is low feedback is noisy Thusstock selection is the type of task for which people are mostovercondent
Odean [1998] develops models in which overcondent inves-tors overestimate the precision of their knowledge about thevalue of a nancial security2 They overestimate the probabilitythat their personal assessments of the securityrsquos value are moreaccurate than the assessments of others Thus overcondentinvestors believe more strongly in their own valuations andconcern themselves less about the beliefs of others This intensi-es differences of opinion And differences of opinion cause trad-ing [Varian 1989 Harris and Raviv 1993] Rational investors onlytrade and only purchase information when doing so increasestheir expected utility (eg Grossman and Stiglitz [1980]) Over-condent investors on the other hand lower their expected util-ity by trading too much they hold unrealistic beliefs about howhigh their returns will be and how precisely these can be esti-mated and they expend too many resources (eg time andmoney) on investment information [Odean 1998] Overcondent
2 Other models of overcondent investors include De Long Shleifer Sum-mers and Waldmann [1991] Benos [1998] Kyle and Wang [1997] Daniel Hirsh-leifer and Subramanyam [1998] Gervais and Odean [1998] and Caballe andSakovics [1998] Kyle and Wang argue that when traders compete for duopolyprots overcondent traders may reap greater prots However this prediction isbased on several assumptions that do not apply to individuals trading commonstocks
Odean [1998] points out that overcondence may result from investors over-estimating the precision of their private signals or alternatively overestimatingtheir abilities to correctly interpret public signals
263BOYS WILL BE BOYS
investors also hold riskier portfolios than do rational investorswith the same degree of risk aversion [Odean 1998]
Barber and Odean [2000] and Odean [1999] test whetherinvestors decrease their expected utility by trading too muchUsing the same data analyzed in this paper Barber and Odeanshow that after accounting for trading costs individual investorsunderperform relevant benchmarks Those who trade the mostrealize by far the worst performance This is what the models ofovercondent investors predict With a different data set Odean[1999] nds that the securities individual investors buy subse-quently underperform those they sell When he controls for li-quidity demands tax-loss selling rebalancing and changes inrisk aversion investorsrsquo timing of trades is even worse Thisresult suggests that not only are investors too willing to act on toolittle information but they are too willing to act when they arewrong
These studies demonstrate that investors trade too much andto their detriment The ndings are inconsistent with rationalityand not easily explained in the absence of overcondence Nev-ertheless overcondence is neither directly observed nor manip-ulated in these studies A yet sharper test of the models thatincorporate overcondence is to partition investors into thosemore and those less prone to overcondence The models predictthat the more overcondent investors will trade more and realizelower average utilities To test these predictions we partition ourdata on gender
IB Gender and Overcondence
While both men and women exhibit overcondence men aregenerally more overcondent than women [Lundeberg Fox andPuncochar 1994]3 Gender differences in overcondence arehighly task dependent [Lundeberg Fox and Puncochar 1994]Deaux and Farris [1977] write ldquoOverall men claim more abilitythan do women but this difference emerges most strongly on masculine task[s]rdquo Several studies conrm that differences incondence are greatest for tasks perceived to be in the masculinedomain [Deaux and Emswiller 1974 Lenney 1977 Beyer andBowden 1997] Men are inclined to feel more competent than
3 While Lichtenstein and Fishhoff [1981] do not nd gender differences incalibration of general knowledge Lundeberg Fox and Puncochar [1994] arguethat this is because gender differences in calibration are strongest for topics in themasculine domain
264 QUARTERLY JOURNAL OF ECONOMICS
women do in nancial matters [Prince 1993] Indeed casual ob-servation reveals that men are disproportionately represented inthe nancial industry We expect therefore that men will gen-erally be more overcondent about their ability to make nancialdecisions than women
Additionally Lenney [1977] reports that gender differencesin self-condence depend on the lack of clear and unambiguousfeedback When feedback is ldquounequivocal and immediately avail-able women do not make lower ability estimates than menHowever when such feedback is absent or ambiguous womenseem to have lower opinions of their abilities and often do under-estimate relative to menrdquo Feedback in the stock market is am-biguous All the more reason to expect men to be more condentthan women about their ability to make common stockinvestments
Gervais and Odean [1998] develop a model in which investorovercondence results from self-serving attribution bias Inves-tors in this model infer their own abilities from their successesand failures Due to their tendency to take too much credit fortheir successes they become overcondent Deaux and Farris[1977] Meehan and Overton [1986] and Beyer [1990] nd thatthe self-serving attribution bias is greater for men than forwomen And so men are likely to become more overcondent thanwomen
The previous study most like our own is Lewellen Lease andSchlarbaumrsquos [1977] analysis of survey answers and brokeragerecords (from 1964 through 1970) of 972 individual investorsLewellen Lease and Schlarbaumrsquos report that men spend moretime and money on security analysis rely less on their brokersmake more transactions believe that returns are more highlypredictable and anticipate higher possible returns than dowomen In all these ways men behave more like overcondentinvestors than do women
Additional evidence that men are more overcondent inves-tors than women comes from surveys conducted by the GallupOrganization for PaineWebber Gallup conducted the survey f-teen times between June 1998 and January 2000 There wereapproximately 1000 respondents per survey In addition to otherquestions respondents were asked ldquoWhat overall rate of returndo you expect to get on your portfolio in the NEXT twelvemonthsrdquo and ldquoThinking about the stock market more generallywhat overall rate of return do you think the stock market will
265BOYS WILL BE BOYS
provide investors during the coming twelve monthsrdquo On averageboth men and women expected their own portfolios to outperformthe market However men expected to outperform by a greatermargin (28 percent) than did women (21 percent) The differencein the average anticipated outperformance of men and women isstatistically signicant (t = 33)4
In summary we have a natural experiment to (almost) di-rectly test theoretical models of investor overcondence A ratio-nal investor only trades if the expected gain exceeds the transac-tions costs An overcondent investor overestimates the precisionof his information and thereby the expected gains of trading Hemay even trade when the true expected net gain is negative Sincemen are more overcondent than women this gives us two test-able hypotheses
H1 Men trade more than women
H2 By trading more men hurt their performance more than dowomen
It is these two hypotheses that are the focus of our inquiry5
II DATA AND METHODS
IIA Household Account and Demographic Data
Our main results focus on the common stock investments of37664 households for which we are able to identify the gender ofthe person who opened the householdrsquos rst brokerage accountThis sample is compiled from two data sets
Our primary data set is information from a large discountbrokerage rm on the investments of 78000 households for the sixyears ending in December 1996 For this period we have end-of-month position statements and trades that allow us to reasonablyestimate monthly returns from February 1991 through January
4 Some respondents answered that they expected market returns as high as900 We suspect that these respondents were thinking of index point moves ratherthan percentage returns Therefore we have dropped from our calculations re-spondents who gave answers of more than 100 to this question If alternativelywe Windsorize answers over 900 at 100 there is no signicant change in ourresults
5 Overcondence models also imply that more overcondent investors willhold riskier portfolios In Section III we present evidence that men hold riskiercommon stock portfolios than women However gender differences in portfoliorisk may be due to differences in risk tolerance rather than (or in addition to)differences in overcondence
266 QUARTERLY JOURNAL OF ECONOMICS
1997 The data set includes all accounts opened by the 78000 house-holds at this discount brokerage rm Sampled households wererequired to have an open account with the discount brokerage rmduring 1991 Roughly half of the accounts in our analysis wereopened prior to 1987 while half were opened between 1987 and1991 On average men opened their rst account at this brokerage47 years before the beginning of our sample period while womenopened theirs 43 years before During the sample period menrsquosaccounts held common stocks for 58 months on average and wom-enrsquos for 59 months The median number of months men held com-mon stocks is 70 For women it is 71
In this research we focus on the common stock investmentsof households We exclude investments in mutual funds (bothopen- and closed-end) American depository receipts (ADRs) war-rants and options Of the 78000 sampled households 66465 hadpositions in common stocks during at least one month the re-maining accounts either held cash or investments in other thanindividual common stocks The average household had approxi-mately two accounts and roughly 60 percent of the market valuein these accounts was held in common stocks These householdsmade over 3 million trades in all securities during our sampleperiod common stocks accounted for slightly more than 60 per-cent of all trades The average household held four stocks worth$47000 during our sample period although each of these guresis positively skewed6 The median household held 26 stocksworth $16000 In aggregate these households held more than$45 billion in common stocks in December 1996
Our secondary data set is demographic information compiled byInfobase Inc (as of June 8 1997) and provided to us by the broker-age house These data identify the gender of the person who openeda householdrsquos rst account for 37664 households of which 29659(79 percent) had accounts opened by men and 8005 (21 percent) hadaccounts opened by women In addition to gender Infobase providesdata on marital status the presence of children age and householdincome We present descriptive statistics in Table I Panel A Thesedata reveal that the women in our sample are less likely to bemarried and to have children than men The mean and median agesof the men and women in our sample are roughly equal The womenreport slightly lower household income although the difference isnot economically large
6 Throughout the paper portfolio values are reported in current dollars
267BOYS WILL BE BOYS
TA
BL
EI
DE
SC
RIP
TIV
ES
TA
TIS
TIC
SF
OR
DE
MO
GR
AP
HIC
SO
FF
EM
AL
EA
ND
MA
LE
HO
US
EH
OL
DS
Var
iabl
e
All
hous
ehol
dsM
arri
edho
use
hold
sS
ingl
eho
useh
olds
Wom
enM
en
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enmdash
men
)
Pan
elA
In
foba
seda
ta
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Per
cen
tage
mar
ried
680
757
27
7P
erce
nta
gew
ith
chil
dren
252
322
27
033
640
42
68
106
105
01
Mea
nag
e50
950
30
649
951
12
12
530
482
48
Med
ian
age
480
480
00
480
480
00
500
460
40
Mea
nin
com
e($
000)
730
756
22
681
279
61
656
762
82
61
w
ith
inco
me
gt$1
250
0011
211
72
05
142
130
12
59
74
21
5
Pan
elB
Sel
f-re
port
edda
ta
Nu
mbe
rof
hou
seho
lds
263
711
226
170
77
700
652
218
4N
etw
orth
($00
0)90
thP
erce
ntil
e50
00
500
00
050
00
500
00
035
00
450
02
100
075
thP
erce
ntil
e20
00
250
02
500
250
025
00
00
175
020
00
225
0M
edia
n10
00
100
00
010
00
100
00
010
00
100
00
025
thP
erce
ntil
e60
074
52
145
625
745
212
040
062
02
220
10th
Per
cent
ile
270
370
210
035
037
02
20
200
350
215
0E
quit
yto
net
wor
th(
)M
ean
133
132
01
129
129
00
144
143
01
Med
ian
67
67
00
63
66
20
37
97
40
5In
vest
men
tex
peri
ence
()
Non
e5
43
42
04
73
41
37
43
04
4L
imit
ed46
834
112
744
934
210
752
633
319
3G
ood
391
485
29
440
848
52
77
333
488
215
5E
xten
sive
87
140
25
39
613
92
43
67
149
28
2
The
sam
ple
con
sist
sof
hous
ehol
dsw
ith
com
mon
stoc
kin
vest
men
tat
ala
rge
disc
oun
tbr
oker
age
rm
for
whi
chw
ear
eab
leto
iden
tify
the
gend
erof
the
pers
onw
hoop
ened
the
hous
ehol
drsquos
rs
tac
coun
tD
ata
onm
arit
alst
atus
ch
ildr
en
age
and
inco
me
are
from
Info
base
Inc
asof
June
1997
S
elf-
repo
rted
data
are
info
rmat
ion
supp
lied
toth
edi
scou
ntbr
oker
age
rm
atth
eti
me
the
acco
unt
isop
ened
byth
epe
rson
onop
enin
gth
eac
coun
tIn
com
eis
repo
rted
wit
hin
eigh
tra
nges
whe
reth
eto
pra
nge
isgr
eate
rth
an$1
250
00W
eca
lcul
ate
mea
nsus
ing
the
mid
poin
tof
each
ran
gean
d$1
250
00fo
rth
eto
pra
nge
Equ
ity
toN
etW
orth
()i
sth
epr
opor
tion
ofth
em
arke
tva
lue
ofco
mm
onst
ock
inve
stm
ent
atth
isdi
scou
ntbr
oker
age
rm
asof
Janu
ary
1991
toto
tals
elf-
repo
rted
net
wor
thw
hen
the
hou
seho
ldop
ened
its
rst
acco
unt
atth
isbr
oker
age
Tho
seh
ouse
hold
sw
ith
apr
opor
tion
equ
ity
tone
tw
orth
grea
ter
than
100
perc
ent
are
dele
ted
wh
enca
lcul
atin
gm
ean
san
dm
edia
ns
Nu
mbe
rof
obse
rvat
ion
sfo
rea
chva
riab
leis
slig
htl
yle
ssth
anth
en
umbe
rof
repo
rted
hous
ehol
ds
268 QUARTERLY JOURNAL OF ECONOMICS
In addition to the data from Infobase we also have a limitedamount of self-reported data collected at the time each householdrst opened an account at the brokerage (and not subsequentlyupdated) which we summarize in Table I Panel B Of particularinterest to us are two variables net worth and investment expe-rience For this limited sample (about one-third of our total sam-ple) the distribution of net worth for women is slightly less thanthat for men although the difference is not economically largeFor this limited sample we also calculate the ratio of the marketvalue of equity (as of the rst month that the account appears inour data set) to self-reported net worth (which is reported at thetime the account is opened) This provides a measure albeitcrude of the proportion of a householdrsquos net worth that is in-vested in the common stocks that we analyze (If this ratio isgreater than one we delete the observation from our analysis)The mean household holds about 13 percent of its net worth in thecommon stocks we analyze and there is little difference in thisratio between men and women
The differences in self-reported experience by gender arequite large In general women report having less investmentexperience than men For example 478 percent of women reporthaving good or extensive investment experience while 625 per-cent of men report the same level of experience
Married couples may inuence each otherrsquos investment deci-sions In some cases the spouse making investment decisions maynot be the spouse who originally opened a brokerage accountThus we anticipate that observable differences in the investmentactivities of men and women will be greatest for single men andsingle women To investigate this possibility we partition ourdata on the basis of marital status The descriptive statistics fromthis partition are presented in the last six columns of Table I Formarried households we observe very small differences in ageincome the distribution of net worth and the ratio of net worthto equity Married women in our sample are less likely to havechildren than married men and they report having less invest-ment experience than men
For single households some differences in demographics be-come larger The average age of the single women in our sample isve years older than that of the single men the median is four yearsolder The average income of single women is $6100 less than thatof single men and fewer report having incomes in excess of$125000 Similarly the distribution of net worth for single women
269BOYS WILL BE BOYS
is lower than that of single men Finally single women reporthaving less investment experience than single men
IIB Return Calculations
To evaluate the investment performance of men and womenwe calculate the gross and net return performance of each house-hold The net return performance is calculated after a reasonableaccounting for the market impact commissions and bid-askspread of each trade
For each trade we estimate the bid-ask spread component oftransaction costs for purchases (sprdb
) or sales (sprds) as
sprds 5 S Pds
cl
Pds
s 2 1 D and sprdb 5 2 S Pdb
cl
Pdb
b 2 1 D Pds
cl and Pdb
cl are the reported closing prices from the Center forResearch in Security Prices (CRSP) daily stock return les on theday of a sale and purchase respectively Pds
s and Pdbb are the
actual sale and purchase price from our account database Ourestimate of the bid-ask spread component of transaction costsincludes any market impact that might result from a trade It alsoincludes an intraday return on the day of the trade The commis-sion component of transaction costs is calculated to be the dollarvalue of the commission paid scaled by the total principal value ofthe transaction both of which are reported in our account data
The average purchase costs an investor 031 percent whilethe average sale costs an investor 069 percent in bid-ask spreadOur estimate of the bid-ask spread is very close to the trading costof 021 percent for purchases and 063 percent for sales paid byopen-end mutual funds from 1966 to 1993 [Carhart 1997]7 Theaverage purchase in excess of $1000 cost 158 percent in commis-sions while the average sale in excess of $1000 cost 145 percent8
We calculate trade-weighted (weighted by trade size) spreadsand commissions These gures can be thought of as the total cost
7 Odean [1999] nds that individual investors are more likely to both buyand sell particular stocks when the prices of those stocks are rising This tendencycan partially explain the asymmetry in buy and sell spreads Any intraday pricerises following transactions subtract from our estimate of the spread for buys andadd to our estimate of the spread for sells
8 To provide more representative descriptive statistics on percentage com-missions we exclude trades less than $1000 The inclusion of these trades resultsin a round-trip commission cost of 5 percent on average (21 percent for purchasesand 31 percent for sales)
270 QUARTERLY JOURNAL OF ECONOMICS
of conducting the $24 billion in common stock trades (approxi-mately $12 billion each in purchases and sales) Trade-sizeweighting has little effect on spread costs (027 percent for pur-chases and 069 percent for sales) but substantially reduces thecommission costs (077 percent for purchases and 066 percent forsales)
In sum the average trade in excess of $1000 incurs a round-trip transaction cost of about 1 percent for the bid-ask spread andabout 3 percent in commissions In aggregate round-trip tradescost about 1 percent for the bid-ask spread and about 14 percentin commissions
We estimate the gross monthly return on each common stockinvestment using the beginning-of-month position statementsfrom our household data and the CRSP monthly returns le In sodoing we make two simplifying assumptions First we assumethat all securities are bought or sold on the last day of the monthThus we ignore the returns earned on stocks purchased from thepurchase date to the end of the month and include the returnsearned on stocks sold from the sale date to the end of the monthSecond we ignore intramonth trading (eg a purchase on March6 and a sale of the same security on March 20) although we doinclude in our analysis short-term trades that yield a position atthe end of a calendar month Barber and Odean [2000] provide acareful analysis of both of these issues and document that thesesimplifying assumptions yield trivial differences in our returncalculations
Consider the common stock portfolio for a particular house-hold The gross monthly return on the householdrsquos portfolio (Rht
gr)is calculated as
Rhtgr 5 O
i= 1
sht
pitRitgr
where pit is the beginning-of-month market value for the holdingof stock i by household h in month t divided by the beginning-of-month market value of all stocks held by household h Rit
gr is thegross monthly return for that stock and sht is the number ofstocks held by household h in month t
For security i in month t we calculate a monthly return netof transaction costs (Rit
net) as
(1 1 Ritnet) 5 (1 1 Rit
gr) (1 2 cits )(1 1 c it 2 1
b )
271BOYS WILL BE BOYS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
ismdashtrade more than rational investors and that doing so lowerstheir expected utilities Greater overcondence leads to greatertrading and to lower expected utility
A direct test of whether overcondence contributes to exces-sive market trading is to separate investors into those more andthose less prone to overcondence One can then test whethermore overcondence leads to more trading and to lower returnsSuch a test is the primary contribution of this paper
Psychologists nd that in areas such as nance men are moreovercondent than women This difference in overcondenceyields two predictions men will trade more than women and theperformance of men will be hurt more by excessive trading thanthe performance of women To test these hypotheses we partitiona data set of position and trading records for over 35000 house-holds at a large discount brokerage rm into accounts opened bymen and accounts opened by women Consistent with the predic-tions of the overcondence models we nd that the averageturnover rate of common stocks for men is nearly one and a halftimes that for women While both men and women reduce theirnet returns through trading men do so by 094 percentage pointsmore a year than do women
The differences in turnover and return performance are evenmore pronounced between single men and single women Singlemen trade 67 percent more than single women thereby reducingtheir returns by 144 percentage points per year more than dosingle women
The remainder of this paper is organized as follows Wemotivate our test of overcondence in Section I We discuss ourdata and empirical methods in Section II Our main results arepresented in Section III We discuss competing explanationsfor our results in Section IV and make concluding remarks inSection V
I A TEST OF OVERCONFIDENCE
IA Overcondence and Trading on Financial Markets
Studies of the calibration of subjective probabilities nd thatpeople tend to overestimate the precision of their knowledge[Alpert and Raiffa 1982 Fischhoff Slovic and Lichtenstein1977] see Lichtenstein Fischhoff and Phillips [1982] for a re-view of the calibration literature Such overcondence has been
262 QUARTERLY JOURNAL OF ECONOMICS
observed in many professional elds Clinical psychologists [Os-kamp 1965] physicians and nurses [Christensen-Szalanski andBushyhead 1981 Baumann Deber and Thompson 1991] invest-ment bankers [Stael von Holstein 1972] engineers [Kidd 1970]entrepreneurs [Cooper Woo and Dunkelberg 1988] lawyers[Wagenaar and Keren 1986] negotiators [Neale and Bazerman1990] and managers [Russo and Schoemaker 1992] have all beenobserved to exhibit overcondence in their judgments (For fur-ther discussion see Lichtenstein Fischhoff and Phillips [1982]and Yates [1990])
Overcondence is greatest for difcult tasks for forecastswith low predictability and for undertakings lacking fast clearfeedback [Fischhoff Slovic and Lichtenstein 1977 LichtensteinFischhoff and Phillips 1982 Yates 1990 Grifn and Tversky1992] Selecting common stocks that will outperform the marketis a difcult task Predictability is low feedback is noisy Thusstock selection is the type of task for which people are mostovercondent
Odean [1998] develops models in which overcondent inves-tors overestimate the precision of their knowledge about thevalue of a nancial security2 They overestimate the probabilitythat their personal assessments of the securityrsquos value are moreaccurate than the assessments of others Thus overcondentinvestors believe more strongly in their own valuations andconcern themselves less about the beliefs of others This intensi-es differences of opinion And differences of opinion cause trad-ing [Varian 1989 Harris and Raviv 1993] Rational investors onlytrade and only purchase information when doing so increasestheir expected utility (eg Grossman and Stiglitz [1980]) Over-condent investors on the other hand lower their expected util-ity by trading too much they hold unrealistic beliefs about howhigh their returns will be and how precisely these can be esti-mated and they expend too many resources (eg time andmoney) on investment information [Odean 1998] Overcondent
2 Other models of overcondent investors include De Long Shleifer Sum-mers and Waldmann [1991] Benos [1998] Kyle and Wang [1997] Daniel Hirsh-leifer and Subramanyam [1998] Gervais and Odean [1998] and Caballe andSakovics [1998] Kyle and Wang argue that when traders compete for duopolyprots overcondent traders may reap greater prots However this prediction isbased on several assumptions that do not apply to individuals trading commonstocks
Odean [1998] points out that overcondence may result from investors over-estimating the precision of their private signals or alternatively overestimatingtheir abilities to correctly interpret public signals
263BOYS WILL BE BOYS
investors also hold riskier portfolios than do rational investorswith the same degree of risk aversion [Odean 1998]
Barber and Odean [2000] and Odean [1999] test whetherinvestors decrease their expected utility by trading too muchUsing the same data analyzed in this paper Barber and Odeanshow that after accounting for trading costs individual investorsunderperform relevant benchmarks Those who trade the mostrealize by far the worst performance This is what the models ofovercondent investors predict With a different data set Odean[1999] nds that the securities individual investors buy subse-quently underperform those they sell When he controls for li-quidity demands tax-loss selling rebalancing and changes inrisk aversion investorsrsquo timing of trades is even worse Thisresult suggests that not only are investors too willing to act on toolittle information but they are too willing to act when they arewrong
These studies demonstrate that investors trade too much andto their detriment The ndings are inconsistent with rationalityand not easily explained in the absence of overcondence Nev-ertheless overcondence is neither directly observed nor manip-ulated in these studies A yet sharper test of the models thatincorporate overcondence is to partition investors into thosemore and those less prone to overcondence The models predictthat the more overcondent investors will trade more and realizelower average utilities To test these predictions we partition ourdata on gender
IB Gender and Overcondence
While both men and women exhibit overcondence men aregenerally more overcondent than women [Lundeberg Fox andPuncochar 1994]3 Gender differences in overcondence arehighly task dependent [Lundeberg Fox and Puncochar 1994]Deaux and Farris [1977] write ldquoOverall men claim more abilitythan do women but this difference emerges most strongly on masculine task[s]rdquo Several studies conrm that differences incondence are greatest for tasks perceived to be in the masculinedomain [Deaux and Emswiller 1974 Lenney 1977 Beyer andBowden 1997] Men are inclined to feel more competent than
3 While Lichtenstein and Fishhoff [1981] do not nd gender differences incalibration of general knowledge Lundeberg Fox and Puncochar [1994] arguethat this is because gender differences in calibration are strongest for topics in themasculine domain
264 QUARTERLY JOURNAL OF ECONOMICS
women do in nancial matters [Prince 1993] Indeed casual ob-servation reveals that men are disproportionately represented inthe nancial industry We expect therefore that men will gen-erally be more overcondent about their ability to make nancialdecisions than women
Additionally Lenney [1977] reports that gender differencesin self-condence depend on the lack of clear and unambiguousfeedback When feedback is ldquounequivocal and immediately avail-able women do not make lower ability estimates than menHowever when such feedback is absent or ambiguous womenseem to have lower opinions of their abilities and often do under-estimate relative to menrdquo Feedback in the stock market is am-biguous All the more reason to expect men to be more condentthan women about their ability to make common stockinvestments
Gervais and Odean [1998] develop a model in which investorovercondence results from self-serving attribution bias Inves-tors in this model infer their own abilities from their successesand failures Due to their tendency to take too much credit fortheir successes they become overcondent Deaux and Farris[1977] Meehan and Overton [1986] and Beyer [1990] nd thatthe self-serving attribution bias is greater for men than forwomen And so men are likely to become more overcondent thanwomen
The previous study most like our own is Lewellen Lease andSchlarbaumrsquos [1977] analysis of survey answers and brokeragerecords (from 1964 through 1970) of 972 individual investorsLewellen Lease and Schlarbaumrsquos report that men spend moretime and money on security analysis rely less on their brokersmake more transactions believe that returns are more highlypredictable and anticipate higher possible returns than dowomen In all these ways men behave more like overcondentinvestors than do women
Additional evidence that men are more overcondent inves-tors than women comes from surveys conducted by the GallupOrganization for PaineWebber Gallup conducted the survey f-teen times between June 1998 and January 2000 There wereapproximately 1000 respondents per survey In addition to otherquestions respondents were asked ldquoWhat overall rate of returndo you expect to get on your portfolio in the NEXT twelvemonthsrdquo and ldquoThinking about the stock market more generallywhat overall rate of return do you think the stock market will
265BOYS WILL BE BOYS
provide investors during the coming twelve monthsrdquo On averageboth men and women expected their own portfolios to outperformthe market However men expected to outperform by a greatermargin (28 percent) than did women (21 percent) The differencein the average anticipated outperformance of men and women isstatistically signicant (t = 33)4
In summary we have a natural experiment to (almost) di-rectly test theoretical models of investor overcondence A ratio-nal investor only trades if the expected gain exceeds the transac-tions costs An overcondent investor overestimates the precisionof his information and thereby the expected gains of trading Hemay even trade when the true expected net gain is negative Sincemen are more overcondent than women this gives us two test-able hypotheses
H1 Men trade more than women
H2 By trading more men hurt their performance more than dowomen
It is these two hypotheses that are the focus of our inquiry5
II DATA AND METHODS
IIA Household Account and Demographic Data
Our main results focus on the common stock investments of37664 households for which we are able to identify the gender ofthe person who opened the householdrsquos rst brokerage accountThis sample is compiled from two data sets
Our primary data set is information from a large discountbrokerage rm on the investments of 78000 households for the sixyears ending in December 1996 For this period we have end-of-month position statements and trades that allow us to reasonablyestimate monthly returns from February 1991 through January
4 Some respondents answered that they expected market returns as high as900 We suspect that these respondents were thinking of index point moves ratherthan percentage returns Therefore we have dropped from our calculations re-spondents who gave answers of more than 100 to this question If alternativelywe Windsorize answers over 900 at 100 there is no signicant change in ourresults
5 Overcondence models also imply that more overcondent investors willhold riskier portfolios In Section III we present evidence that men hold riskiercommon stock portfolios than women However gender differences in portfoliorisk may be due to differences in risk tolerance rather than (or in addition to)differences in overcondence
266 QUARTERLY JOURNAL OF ECONOMICS
1997 The data set includes all accounts opened by the 78000 house-holds at this discount brokerage rm Sampled households wererequired to have an open account with the discount brokerage rmduring 1991 Roughly half of the accounts in our analysis wereopened prior to 1987 while half were opened between 1987 and1991 On average men opened their rst account at this brokerage47 years before the beginning of our sample period while womenopened theirs 43 years before During the sample period menrsquosaccounts held common stocks for 58 months on average and wom-enrsquos for 59 months The median number of months men held com-mon stocks is 70 For women it is 71
In this research we focus on the common stock investmentsof households We exclude investments in mutual funds (bothopen- and closed-end) American depository receipts (ADRs) war-rants and options Of the 78000 sampled households 66465 hadpositions in common stocks during at least one month the re-maining accounts either held cash or investments in other thanindividual common stocks The average household had approxi-mately two accounts and roughly 60 percent of the market valuein these accounts was held in common stocks These householdsmade over 3 million trades in all securities during our sampleperiod common stocks accounted for slightly more than 60 per-cent of all trades The average household held four stocks worth$47000 during our sample period although each of these guresis positively skewed6 The median household held 26 stocksworth $16000 In aggregate these households held more than$45 billion in common stocks in December 1996
Our secondary data set is demographic information compiled byInfobase Inc (as of June 8 1997) and provided to us by the broker-age house These data identify the gender of the person who openeda householdrsquos rst account for 37664 households of which 29659(79 percent) had accounts opened by men and 8005 (21 percent) hadaccounts opened by women In addition to gender Infobase providesdata on marital status the presence of children age and householdincome We present descriptive statistics in Table I Panel A Thesedata reveal that the women in our sample are less likely to bemarried and to have children than men The mean and median agesof the men and women in our sample are roughly equal The womenreport slightly lower household income although the difference isnot economically large
6 Throughout the paper portfolio values are reported in current dollars
267BOYS WILL BE BOYS
TA
BL
EI
DE
SC
RIP
TIV
ES
TA
TIS
TIC
SF
OR
DE
MO
GR
AP
HIC
SO
FF
EM
AL
EA
ND
MA
LE
HO
US
EH
OL
DS
Var
iabl
e
All
hous
ehol
dsM
arri
edho
use
hold
sS
ingl
eho
useh
olds
Wom
enM
en
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enmdash
men
)
Pan
elA
In
foba
seda
ta
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Per
cen
tage
mar
ried
680
757
27
7P
erce
nta
gew
ith
chil
dren
252
322
27
033
640
42
68
106
105
01
Mea
nag
e50
950
30
649
951
12
12
530
482
48
Med
ian
age
480
480
00
480
480
00
500
460
40
Mea
nin
com
e($
000)
730
756
22
681
279
61
656
762
82
61
w
ith
inco
me
gt$1
250
0011
211
72
05
142
130
12
59
74
21
5
Pan
elB
Sel
f-re
port
edda
ta
Nu
mbe
rof
hou
seho
lds
263
711
226
170
77
700
652
218
4N
etw
orth
($00
0)90
thP
erce
ntil
e50
00
500
00
050
00
500
00
035
00
450
02
100
075
thP
erce
ntil
e20
00
250
02
500
250
025
00
00
175
020
00
225
0M
edia
n10
00
100
00
010
00
100
00
010
00
100
00
025
thP
erce
ntil
e60
074
52
145
625
745
212
040
062
02
220
10th
Per
cent
ile
270
370
210
035
037
02
20
200
350
215
0E
quit
yto
net
wor
th(
)M
ean
133
132
01
129
129
00
144
143
01
Med
ian
67
67
00
63
66
20
37
97
40
5In
vest
men
tex
peri
ence
()
Non
e5
43
42
04
73
41
37
43
04
4L
imit
ed46
834
112
744
934
210
752
633
319
3G
ood
391
485
29
440
848
52
77
333
488
215
5E
xten
sive
87
140
25
39
613
92
43
67
149
28
2
The
sam
ple
con
sist
sof
hous
ehol
dsw
ith
com
mon
stoc
kin
vest
men
tat
ala
rge
disc
oun
tbr
oker
age
rm
for
whi
chw
ear
eab
leto
iden
tify
the
gend
erof
the
pers
onw
hoop
ened
the
hous
ehol
drsquos
rs
tac
coun
tD
ata
onm
arit
alst
atus
ch
ildr
en
age
and
inco
me
are
from
Info
base
Inc
asof
June
1997
S
elf-
repo
rted
data
are
info
rmat
ion
supp
lied
toth
edi
scou
ntbr
oker
age
rm
atth
eti
me
the
acco
unt
isop
ened
byth
epe
rson
onop
enin
gth
eac
coun
tIn
com
eis
repo
rted
wit
hin
eigh
tra
nges
whe
reth
eto
pra
nge
isgr
eate
rth
an$1
250
00W
eca
lcul
ate
mea
nsus
ing
the
mid
poin
tof
each
ran
gean
d$1
250
00fo
rth
eto
pra
nge
Equ
ity
toN
etW
orth
()i
sth
epr
opor
tion
ofth
em
arke
tva
lue
ofco
mm
onst
ock
inve
stm
ent
atth
isdi
scou
ntbr
oker
age
rm
asof
Janu
ary
1991
toto
tals
elf-
repo
rted
net
wor
thw
hen
the
hou
seho
ldop
ened
its
rst
acco
unt
atth
isbr
oker
age
Tho
seh
ouse
hold
sw
ith
apr
opor
tion
equ
ity
tone
tw
orth
grea
ter
than
100
perc
ent
are
dele
ted
wh
enca
lcul
atin
gm
ean
san
dm
edia
ns
Nu
mbe
rof
obse
rvat
ion
sfo
rea
chva
riab
leis
slig
htl
yle
ssth
anth
en
umbe
rof
repo
rted
hous
ehol
ds
268 QUARTERLY JOURNAL OF ECONOMICS
In addition to the data from Infobase we also have a limitedamount of self-reported data collected at the time each householdrst opened an account at the brokerage (and not subsequentlyupdated) which we summarize in Table I Panel B Of particularinterest to us are two variables net worth and investment expe-rience For this limited sample (about one-third of our total sam-ple) the distribution of net worth for women is slightly less thanthat for men although the difference is not economically largeFor this limited sample we also calculate the ratio of the marketvalue of equity (as of the rst month that the account appears inour data set) to self-reported net worth (which is reported at thetime the account is opened) This provides a measure albeitcrude of the proportion of a householdrsquos net worth that is in-vested in the common stocks that we analyze (If this ratio isgreater than one we delete the observation from our analysis)The mean household holds about 13 percent of its net worth in thecommon stocks we analyze and there is little difference in thisratio between men and women
The differences in self-reported experience by gender arequite large In general women report having less investmentexperience than men For example 478 percent of women reporthaving good or extensive investment experience while 625 per-cent of men report the same level of experience
Married couples may inuence each otherrsquos investment deci-sions In some cases the spouse making investment decisions maynot be the spouse who originally opened a brokerage accountThus we anticipate that observable differences in the investmentactivities of men and women will be greatest for single men andsingle women To investigate this possibility we partition ourdata on the basis of marital status The descriptive statistics fromthis partition are presented in the last six columns of Table I Formarried households we observe very small differences in ageincome the distribution of net worth and the ratio of net worthto equity Married women in our sample are less likely to havechildren than married men and they report having less invest-ment experience than men
For single households some differences in demographics be-come larger The average age of the single women in our sample isve years older than that of the single men the median is four yearsolder The average income of single women is $6100 less than thatof single men and fewer report having incomes in excess of$125000 Similarly the distribution of net worth for single women
269BOYS WILL BE BOYS
is lower than that of single men Finally single women reporthaving less investment experience than single men
IIB Return Calculations
To evaluate the investment performance of men and womenwe calculate the gross and net return performance of each house-hold The net return performance is calculated after a reasonableaccounting for the market impact commissions and bid-askspread of each trade
For each trade we estimate the bid-ask spread component oftransaction costs for purchases (sprdb
) or sales (sprds) as
sprds 5 S Pds
cl
Pds
s 2 1 D and sprdb 5 2 S Pdb
cl
Pdb
b 2 1 D Pds
cl and Pdb
cl are the reported closing prices from the Center forResearch in Security Prices (CRSP) daily stock return les on theday of a sale and purchase respectively Pds
s and Pdbb are the
actual sale and purchase price from our account database Ourestimate of the bid-ask spread component of transaction costsincludes any market impact that might result from a trade It alsoincludes an intraday return on the day of the trade The commis-sion component of transaction costs is calculated to be the dollarvalue of the commission paid scaled by the total principal value ofthe transaction both of which are reported in our account data
The average purchase costs an investor 031 percent whilethe average sale costs an investor 069 percent in bid-ask spreadOur estimate of the bid-ask spread is very close to the trading costof 021 percent for purchases and 063 percent for sales paid byopen-end mutual funds from 1966 to 1993 [Carhart 1997]7 Theaverage purchase in excess of $1000 cost 158 percent in commis-sions while the average sale in excess of $1000 cost 145 percent8
We calculate trade-weighted (weighted by trade size) spreadsand commissions These gures can be thought of as the total cost
7 Odean [1999] nds that individual investors are more likely to both buyand sell particular stocks when the prices of those stocks are rising This tendencycan partially explain the asymmetry in buy and sell spreads Any intraday pricerises following transactions subtract from our estimate of the spread for buys andadd to our estimate of the spread for sells
8 To provide more representative descriptive statistics on percentage com-missions we exclude trades less than $1000 The inclusion of these trades resultsin a round-trip commission cost of 5 percent on average (21 percent for purchasesand 31 percent for sales)
270 QUARTERLY JOURNAL OF ECONOMICS
of conducting the $24 billion in common stock trades (approxi-mately $12 billion each in purchases and sales) Trade-sizeweighting has little effect on spread costs (027 percent for pur-chases and 069 percent for sales) but substantially reduces thecommission costs (077 percent for purchases and 066 percent forsales)
In sum the average trade in excess of $1000 incurs a round-trip transaction cost of about 1 percent for the bid-ask spread andabout 3 percent in commissions In aggregate round-trip tradescost about 1 percent for the bid-ask spread and about 14 percentin commissions
We estimate the gross monthly return on each common stockinvestment using the beginning-of-month position statementsfrom our household data and the CRSP monthly returns le In sodoing we make two simplifying assumptions First we assumethat all securities are bought or sold on the last day of the monthThus we ignore the returns earned on stocks purchased from thepurchase date to the end of the month and include the returnsearned on stocks sold from the sale date to the end of the monthSecond we ignore intramonth trading (eg a purchase on March6 and a sale of the same security on March 20) although we doinclude in our analysis short-term trades that yield a position atthe end of a calendar month Barber and Odean [2000] provide acareful analysis of both of these issues and document that thesesimplifying assumptions yield trivial differences in our returncalculations
Consider the common stock portfolio for a particular house-hold The gross monthly return on the householdrsquos portfolio (Rht
gr)is calculated as
Rhtgr 5 O
i= 1
sht
pitRitgr
where pit is the beginning-of-month market value for the holdingof stock i by household h in month t divided by the beginning-of-month market value of all stocks held by household h Rit
gr is thegross monthly return for that stock and sht is the number ofstocks held by household h in month t
For security i in month t we calculate a monthly return netof transaction costs (Rit
net) as
(1 1 Ritnet) 5 (1 1 Rit
gr) (1 2 cits )(1 1 c it 2 1
b )
271BOYS WILL BE BOYS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
observed in many professional elds Clinical psychologists [Os-kamp 1965] physicians and nurses [Christensen-Szalanski andBushyhead 1981 Baumann Deber and Thompson 1991] invest-ment bankers [Stael von Holstein 1972] engineers [Kidd 1970]entrepreneurs [Cooper Woo and Dunkelberg 1988] lawyers[Wagenaar and Keren 1986] negotiators [Neale and Bazerman1990] and managers [Russo and Schoemaker 1992] have all beenobserved to exhibit overcondence in their judgments (For fur-ther discussion see Lichtenstein Fischhoff and Phillips [1982]and Yates [1990])
Overcondence is greatest for difcult tasks for forecastswith low predictability and for undertakings lacking fast clearfeedback [Fischhoff Slovic and Lichtenstein 1977 LichtensteinFischhoff and Phillips 1982 Yates 1990 Grifn and Tversky1992] Selecting common stocks that will outperform the marketis a difcult task Predictability is low feedback is noisy Thusstock selection is the type of task for which people are mostovercondent
Odean [1998] develops models in which overcondent inves-tors overestimate the precision of their knowledge about thevalue of a nancial security2 They overestimate the probabilitythat their personal assessments of the securityrsquos value are moreaccurate than the assessments of others Thus overcondentinvestors believe more strongly in their own valuations andconcern themselves less about the beliefs of others This intensi-es differences of opinion And differences of opinion cause trad-ing [Varian 1989 Harris and Raviv 1993] Rational investors onlytrade and only purchase information when doing so increasestheir expected utility (eg Grossman and Stiglitz [1980]) Over-condent investors on the other hand lower their expected util-ity by trading too much they hold unrealistic beliefs about howhigh their returns will be and how precisely these can be esti-mated and they expend too many resources (eg time andmoney) on investment information [Odean 1998] Overcondent
2 Other models of overcondent investors include De Long Shleifer Sum-mers and Waldmann [1991] Benos [1998] Kyle and Wang [1997] Daniel Hirsh-leifer and Subramanyam [1998] Gervais and Odean [1998] and Caballe andSakovics [1998] Kyle and Wang argue that when traders compete for duopolyprots overcondent traders may reap greater prots However this prediction isbased on several assumptions that do not apply to individuals trading commonstocks
Odean [1998] points out that overcondence may result from investors over-estimating the precision of their private signals or alternatively overestimatingtheir abilities to correctly interpret public signals
263BOYS WILL BE BOYS
investors also hold riskier portfolios than do rational investorswith the same degree of risk aversion [Odean 1998]
Barber and Odean [2000] and Odean [1999] test whetherinvestors decrease their expected utility by trading too muchUsing the same data analyzed in this paper Barber and Odeanshow that after accounting for trading costs individual investorsunderperform relevant benchmarks Those who trade the mostrealize by far the worst performance This is what the models ofovercondent investors predict With a different data set Odean[1999] nds that the securities individual investors buy subse-quently underperform those they sell When he controls for li-quidity demands tax-loss selling rebalancing and changes inrisk aversion investorsrsquo timing of trades is even worse Thisresult suggests that not only are investors too willing to act on toolittle information but they are too willing to act when they arewrong
These studies demonstrate that investors trade too much andto their detriment The ndings are inconsistent with rationalityand not easily explained in the absence of overcondence Nev-ertheless overcondence is neither directly observed nor manip-ulated in these studies A yet sharper test of the models thatincorporate overcondence is to partition investors into thosemore and those less prone to overcondence The models predictthat the more overcondent investors will trade more and realizelower average utilities To test these predictions we partition ourdata on gender
IB Gender and Overcondence
While both men and women exhibit overcondence men aregenerally more overcondent than women [Lundeberg Fox andPuncochar 1994]3 Gender differences in overcondence arehighly task dependent [Lundeberg Fox and Puncochar 1994]Deaux and Farris [1977] write ldquoOverall men claim more abilitythan do women but this difference emerges most strongly on masculine task[s]rdquo Several studies conrm that differences incondence are greatest for tasks perceived to be in the masculinedomain [Deaux and Emswiller 1974 Lenney 1977 Beyer andBowden 1997] Men are inclined to feel more competent than
3 While Lichtenstein and Fishhoff [1981] do not nd gender differences incalibration of general knowledge Lundeberg Fox and Puncochar [1994] arguethat this is because gender differences in calibration are strongest for topics in themasculine domain
264 QUARTERLY JOURNAL OF ECONOMICS
women do in nancial matters [Prince 1993] Indeed casual ob-servation reveals that men are disproportionately represented inthe nancial industry We expect therefore that men will gen-erally be more overcondent about their ability to make nancialdecisions than women
Additionally Lenney [1977] reports that gender differencesin self-condence depend on the lack of clear and unambiguousfeedback When feedback is ldquounequivocal and immediately avail-able women do not make lower ability estimates than menHowever when such feedback is absent or ambiguous womenseem to have lower opinions of their abilities and often do under-estimate relative to menrdquo Feedback in the stock market is am-biguous All the more reason to expect men to be more condentthan women about their ability to make common stockinvestments
Gervais and Odean [1998] develop a model in which investorovercondence results from self-serving attribution bias Inves-tors in this model infer their own abilities from their successesand failures Due to their tendency to take too much credit fortheir successes they become overcondent Deaux and Farris[1977] Meehan and Overton [1986] and Beyer [1990] nd thatthe self-serving attribution bias is greater for men than forwomen And so men are likely to become more overcondent thanwomen
The previous study most like our own is Lewellen Lease andSchlarbaumrsquos [1977] analysis of survey answers and brokeragerecords (from 1964 through 1970) of 972 individual investorsLewellen Lease and Schlarbaumrsquos report that men spend moretime and money on security analysis rely less on their brokersmake more transactions believe that returns are more highlypredictable and anticipate higher possible returns than dowomen In all these ways men behave more like overcondentinvestors than do women
Additional evidence that men are more overcondent inves-tors than women comes from surveys conducted by the GallupOrganization for PaineWebber Gallup conducted the survey f-teen times between June 1998 and January 2000 There wereapproximately 1000 respondents per survey In addition to otherquestions respondents were asked ldquoWhat overall rate of returndo you expect to get on your portfolio in the NEXT twelvemonthsrdquo and ldquoThinking about the stock market more generallywhat overall rate of return do you think the stock market will
265BOYS WILL BE BOYS
provide investors during the coming twelve monthsrdquo On averageboth men and women expected their own portfolios to outperformthe market However men expected to outperform by a greatermargin (28 percent) than did women (21 percent) The differencein the average anticipated outperformance of men and women isstatistically signicant (t = 33)4
In summary we have a natural experiment to (almost) di-rectly test theoretical models of investor overcondence A ratio-nal investor only trades if the expected gain exceeds the transac-tions costs An overcondent investor overestimates the precisionof his information and thereby the expected gains of trading Hemay even trade when the true expected net gain is negative Sincemen are more overcondent than women this gives us two test-able hypotheses
H1 Men trade more than women
H2 By trading more men hurt their performance more than dowomen
It is these two hypotheses that are the focus of our inquiry5
II DATA AND METHODS
IIA Household Account and Demographic Data
Our main results focus on the common stock investments of37664 households for which we are able to identify the gender ofthe person who opened the householdrsquos rst brokerage accountThis sample is compiled from two data sets
Our primary data set is information from a large discountbrokerage rm on the investments of 78000 households for the sixyears ending in December 1996 For this period we have end-of-month position statements and trades that allow us to reasonablyestimate monthly returns from February 1991 through January
4 Some respondents answered that they expected market returns as high as900 We suspect that these respondents were thinking of index point moves ratherthan percentage returns Therefore we have dropped from our calculations re-spondents who gave answers of more than 100 to this question If alternativelywe Windsorize answers over 900 at 100 there is no signicant change in ourresults
5 Overcondence models also imply that more overcondent investors willhold riskier portfolios In Section III we present evidence that men hold riskiercommon stock portfolios than women However gender differences in portfoliorisk may be due to differences in risk tolerance rather than (or in addition to)differences in overcondence
266 QUARTERLY JOURNAL OF ECONOMICS
1997 The data set includes all accounts opened by the 78000 house-holds at this discount brokerage rm Sampled households wererequired to have an open account with the discount brokerage rmduring 1991 Roughly half of the accounts in our analysis wereopened prior to 1987 while half were opened between 1987 and1991 On average men opened their rst account at this brokerage47 years before the beginning of our sample period while womenopened theirs 43 years before During the sample period menrsquosaccounts held common stocks for 58 months on average and wom-enrsquos for 59 months The median number of months men held com-mon stocks is 70 For women it is 71
In this research we focus on the common stock investmentsof households We exclude investments in mutual funds (bothopen- and closed-end) American depository receipts (ADRs) war-rants and options Of the 78000 sampled households 66465 hadpositions in common stocks during at least one month the re-maining accounts either held cash or investments in other thanindividual common stocks The average household had approxi-mately two accounts and roughly 60 percent of the market valuein these accounts was held in common stocks These householdsmade over 3 million trades in all securities during our sampleperiod common stocks accounted for slightly more than 60 per-cent of all trades The average household held four stocks worth$47000 during our sample period although each of these guresis positively skewed6 The median household held 26 stocksworth $16000 In aggregate these households held more than$45 billion in common stocks in December 1996
Our secondary data set is demographic information compiled byInfobase Inc (as of June 8 1997) and provided to us by the broker-age house These data identify the gender of the person who openeda householdrsquos rst account for 37664 households of which 29659(79 percent) had accounts opened by men and 8005 (21 percent) hadaccounts opened by women In addition to gender Infobase providesdata on marital status the presence of children age and householdincome We present descriptive statistics in Table I Panel A Thesedata reveal that the women in our sample are less likely to bemarried and to have children than men The mean and median agesof the men and women in our sample are roughly equal The womenreport slightly lower household income although the difference isnot economically large
6 Throughout the paper portfolio values are reported in current dollars
267BOYS WILL BE BOYS
TA
BL
EI
DE
SC
RIP
TIV
ES
TA
TIS
TIC
SF
OR
DE
MO
GR
AP
HIC
SO
FF
EM
AL
EA
ND
MA
LE
HO
US
EH
OL
DS
Var
iabl
e
All
hous
ehol
dsM
arri
edho
use
hold
sS
ingl
eho
useh
olds
Wom
enM
en
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enmdash
men
)
Pan
elA
In
foba
seda
ta
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Per
cen
tage
mar
ried
680
757
27
7P
erce
nta
gew
ith
chil
dren
252
322
27
033
640
42
68
106
105
01
Mea
nag
e50
950
30
649
951
12
12
530
482
48
Med
ian
age
480
480
00
480
480
00
500
460
40
Mea
nin
com
e($
000)
730
756
22
681
279
61
656
762
82
61
w
ith
inco
me
gt$1
250
0011
211
72
05
142
130
12
59
74
21
5
Pan
elB
Sel
f-re
port
edda
ta
Nu
mbe
rof
hou
seho
lds
263
711
226
170
77
700
652
218
4N
etw
orth
($00
0)90
thP
erce
ntil
e50
00
500
00
050
00
500
00
035
00
450
02
100
075
thP
erce
ntil
e20
00
250
02
500
250
025
00
00
175
020
00
225
0M
edia
n10
00
100
00
010
00
100
00
010
00
100
00
025
thP
erce
ntil
e60
074
52
145
625
745
212
040
062
02
220
10th
Per
cent
ile
270
370
210
035
037
02
20
200
350
215
0E
quit
yto
net
wor
th(
)M
ean
133
132
01
129
129
00
144
143
01
Med
ian
67
67
00
63
66
20
37
97
40
5In
vest
men
tex
peri
ence
()
Non
e5
43
42
04
73
41
37
43
04
4L
imit
ed46
834
112
744
934
210
752
633
319
3G
ood
391
485
29
440
848
52
77
333
488
215
5E
xten
sive
87
140
25
39
613
92
43
67
149
28
2
The
sam
ple
con
sist
sof
hous
ehol
dsw
ith
com
mon
stoc
kin
vest
men
tat
ala
rge
disc
oun
tbr
oker
age
rm
for
whi
chw
ear
eab
leto
iden
tify
the
gend
erof
the
pers
onw
hoop
ened
the
hous
ehol
drsquos
rs
tac
coun
tD
ata
onm
arit
alst
atus
ch
ildr
en
age
and
inco
me
are
from
Info
base
Inc
asof
June
1997
S
elf-
repo
rted
data
are
info
rmat
ion
supp
lied
toth
edi
scou
ntbr
oker
age
rm
atth
eti
me
the
acco
unt
isop
ened
byth
epe
rson
onop
enin
gth
eac
coun
tIn
com
eis
repo
rted
wit
hin
eigh
tra
nges
whe
reth
eto
pra
nge
isgr
eate
rth
an$1
250
00W
eca
lcul
ate
mea
nsus
ing
the
mid
poin
tof
each
ran
gean
d$1
250
00fo
rth
eto
pra
nge
Equ
ity
toN
etW
orth
()i
sth
epr
opor
tion
ofth
em
arke
tva
lue
ofco
mm
onst
ock
inve
stm
ent
atth
isdi
scou
ntbr
oker
age
rm
asof
Janu
ary
1991
toto
tals
elf-
repo
rted
net
wor
thw
hen
the
hou
seho
ldop
ened
its
rst
acco
unt
atth
isbr
oker
age
Tho
seh
ouse
hold
sw
ith
apr
opor
tion
equ
ity
tone
tw
orth
grea
ter
than
100
perc
ent
are
dele
ted
wh
enca
lcul
atin
gm
ean
san
dm
edia
ns
Nu
mbe
rof
obse
rvat
ion
sfo
rea
chva
riab
leis
slig
htl
yle
ssth
anth
en
umbe
rof
repo
rted
hous
ehol
ds
268 QUARTERLY JOURNAL OF ECONOMICS
In addition to the data from Infobase we also have a limitedamount of self-reported data collected at the time each householdrst opened an account at the brokerage (and not subsequentlyupdated) which we summarize in Table I Panel B Of particularinterest to us are two variables net worth and investment expe-rience For this limited sample (about one-third of our total sam-ple) the distribution of net worth for women is slightly less thanthat for men although the difference is not economically largeFor this limited sample we also calculate the ratio of the marketvalue of equity (as of the rst month that the account appears inour data set) to self-reported net worth (which is reported at thetime the account is opened) This provides a measure albeitcrude of the proportion of a householdrsquos net worth that is in-vested in the common stocks that we analyze (If this ratio isgreater than one we delete the observation from our analysis)The mean household holds about 13 percent of its net worth in thecommon stocks we analyze and there is little difference in thisratio between men and women
The differences in self-reported experience by gender arequite large In general women report having less investmentexperience than men For example 478 percent of women reporthaving good or extensive investment experience while 625 per-cent of men report the same level of experience
Married couples may inuence each otherrsquos investment deci-sions In some cases the spouse making investment decisions maynot be the spouse who originally opened a brokerage accountThus we anticipate that observable differences in the investmentactivities of men and women will be greatest for single men andsingle women To investigate this possibility we partition ourdata on the basis of marital status The descriptive statistics fromthis partition are presented in the last six columns of Table I Formarried households we observe very small differences in ageincome the distribution of net worth and the ratio of net worthto equity Married women in our sample are less likely to havechildren than married men and they report having less invest-ment experience than men
For single households some differences in demographics be-come larger The average age of the single women in our sample isve years older than that of the single men the median is four yearsolder The average income of single women is $6100 less than thatof single men and fewer report having incomes in excess of$125000 Similarly the distribution of net worth for single women
269BOYS WILL BE BOYS
is lower than that of single men Finally single women reporthaving less investment experience than single men
IIB Return Calculations
To evaluate the investment performance of men and womenwe calculate the gross and net return performance of each house-hold The net return performance is calculated after a reasonableaccounting for the market impact commissions and bid-askspread of each trade
For each trade we estimate the bid-ask spread component oftransaction costs for purchases (sprdb
) or sales (sprds) as
sprds 5 S Pds
cl
Pds
s 2 1 D and sprdb 5 2 S Pdb
cl
Pdb
b 2 1 D Pds
cl and Pdb
cl are the reported closing prices from the Center forResearch in Security Prices (CRSP) daily stock return les on theday of a sale and purchase respectively Pds
s and Pdbb are the
actual sale and purchase price from our account database Ourestimate of the bid-ask spread component of transaction costsincludes any market impact that might result from a trade It alsoincludes an intraday return on the day of the trade The commis-sion component of transaction costs is calculated to be the dollarvalue of the commission paid scaled by the total principal value ofthe transaction both of which are reported in our account data
The average purchase costs an investor 031 percent whilethe average sale costs an investor 069 percent in bid-ask spreadOur estimate of the bid-ask spread is very close to the trading costof 021 percent for purchases and 063 percent for sales paid byopen-end mutual funds from 1966 to 1993 [Carhart 1997]7 Theaverage purchase in excess of $1000 cost 158 percent in commis-sions while the average sale in excess of $1000 cost 145 percent8
We calculate trade-weighted (weighted by trade size) spreadsand commissions These gures can be thought of as the total cost
7 Odean [1999] nds that individual investors are more likely to both buyand sell particular stocks when the prices of those stocks are rising This tendencycan partially explain the asymmetry in buy and sell spreads Any intraday pricerises following transactions subtract from our estimate of the spread for buys andadd to our estimate of the spread for sells
8 To provide more representative descriptive statistics on percentage com-missions we exclude trades less than $1000 The inclusion of these trades resultsin a round-trip commission cost of 5 percent on average (21 percent for purchasesand 31 percent for sales)
270 QUARTERLY JOURNAL OF ECONOMICS
of conducting the $24 billion in common stock trades (approxi-mately $12 billion each in purchases and sales) Trade-sizeweighting has little effect on spread costs (027 percent for pur-chases and 069 percent for sales) but substantially reduces thecommission costs (077 percent for purchases and 066 percent forsales)
In sum the average trade in excess of $1000 incurs a round-trip transaction cost of about 1 percent for the bid-ask spread andabout 3 percent in commissions In aggregate round-trip tradescost about 1 percent for the bid-ask spread and about 14 percentin commissions
We estimate the gross monthly return on each common stockinvestment using the beginning-of-month position statementsfrom our household data and the CRSP monthly returns le In sodoing we make two simplifying assumptions First we assumethat all securities are bought or sold on the last day of the monthThus we ignore the returns earned on stocks purchased from thepurchase date to the end of the month and include the returnsearned on stocks sold from the sale date to the end of the monthSecond we ignore intramonth trading (eg a purchase on March6 and a sale of the same security on March 20) although we doinclude in our analysis short-term trades that yield a position atthe end of a calendar month Barber and Odean [2000] provide acareful analysis of both of these issues and document that thesesimplifying assumptions yield trivial differences in our returncalculations
Consider the common stock portfolio for a particular house-hold The gross monthly return on the householdrsquos portfolio (Rht
gr)is calculated as
Rhtgr 5 O
i= 1
sht
pitRitgr
where pit is the beginning-of-month market value for the holdingof stock i by household h in month t divided by the beginning-of-month market value of all stocks held by household h Rit
gr is thegross monthly return for that stock and sht is the number ofstocks held by household h in month t
For security i in month t we calculate a monthly return netof transaction costs (Rit
net) as
(1 1 Ritnet) 5 (1 1 Rit
gr) (1 2 cits )(1 1 c it 2 1
b )
271BOYS WILL BE BOYS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
investors also hold riskier portfolios than do rational investorswith the same degree of risk aversion [Odean 1998]
Barber and Odean [2000] and Odean [1999] test whetherinvestors decrease their expected utility by trading too muchUsing the same data analyzed in this paper Barber and Odeanshow that after accounting for trading costs individual investorsunderperform relevant benchmarks Those who trade the mostrealize by far the worst performance This is what the models ofovercondent investors predict With a different data set Odean[1999] nds that the securities individual investors buy subse-quently underperform those they sell When he controls for li-quidity demands tax-loss selling rebalancing and changes inrisk aversion investorsrsquo timing of trades is even worse Thisresult suggests that not only are investors too willing to act on toolittle information but they are too willing to act when they arewrong
These studies demonstrate that investors trade too much andto their detriment The ndings are inconsistent with rationalityand not easily explained in the absence of overcondence Nev-ertheless overcondence is neither directly observed nor manip-ulated in these studies A yet sharper test of the models thatincorporate overcondence is to partition investors into thosemore and those less prone to overcondence The models predictthat the more overcondent investors will trade more and realizelower average utilities To test these predictions we partition ourdata on gender
IB Gender and Overcondence
While both men and women exhibit overcondence men aregenerally more overcondent than women [Lundeberg Fox andPuncochar 1994]3 Gender differences in overcondence arehighly task dependent [Lundeberg Fox and Puncochar 1994]Deaux and Farris [1977] write ldquoOverall men claim more abilitythan do women but this difference emerges most strongly on masculine task[s]rdquo Several studies conrm that differences incondence are greatest for tasks perceived to be in the masculinedomain [Deaux and Emswiller 1974 Lenney 1977 Beyer andBowden 1997] Men are inclined to feel more competent than
3 While Lichtenstein and Fishhoff [1981] do not nd gender differences incalibration of general knowledge Lundeberg Fox and Puncochar [1994] arguethat this is because gender differences in calibration are strongest for topics in themasculine domain
264 QUARTERLY JOURNAL OF ECONOMICS
women do in nancial matters [Prince 1993] Indeed casual ob-servation reveals that men are disproportionately represented inthe nancial industry We expect therefore that men will gen-erally be more overcondent about their ability to make nancialdecisions than women
Additionally Lenney [1977] reports that gender differencesin self-condence depend on the lack of clear and unambiguousfeedback When feedback is ldquounequivocal and immediately avail-able women do not make lower ability estimates than menHowever when such feedback is absent or ambiguous womenseem to have lower opinions of their abilities and often do under-estimate relative to menrdquo Feedback in the stock market is am-biguous All the more reason to expect men to be more condentthan women about their ability to make common stockinvestments
Gervais and Odean [1998] develop a model in which investorovercondence results from self-serving attribution bias Inves-tors in this model infer their own abilities from their successesand failures Due to their tendency to take too much credit fortheir successes they become overcondent Deaux and Farris[1977] Meehan and Overton [1986] and Beyer [1990] nd thatthe self-serving attribution bias is greater for men than forwomen And so men are likely to become more overcondent thanwomen
The previous study most like our own is Lewellen Lease andSchlarbaumrsquos [1977] analysis of survey answers and brokeragerecords (from 1964 through 1970) of 972 individual investorsLewellen Lease and Schlarbaumrsquos report that men spend moretime and money on security analysis rely less on their brokersmake more transactions believe that returns are more highlypredictable and anticipate higher possible returns than dowomen In all these ways men behave more like overcondentinvestors than do women
Additional evidence that men are more overcondent inves-tors than women comes from surveys conducted by the GallupOrganization for PaineWebber Gallup conducted the survey f-teen times between June 1998 and January 2000 There wereapproximately 1000 respondents per survey In addition to otherquestions respondents were asked ldquoWhat overall rate of returndo you expect to get on your portfolio in the NEXT twelvemonthsrdquo and ldquoThinking about the stock market more generallywhat overall rate of return do you think the stock market will
265BOYS WILL BE BOYS
provide investors during the coming twelve monthsrdquo On averageboth men and women expected their own portfolios to outperformthe market However men expected to outperform by a greatermargin (28 percent) than did women (21 percent) The differencein the average anticipated outperformance of men and women isstatistically signicant (t = 33)4
In summary we have a natural experiment to (almost) di-rectly test theoretical models of investor overcondence A ratio-nal investor only trades if the expected gain exceeds the transac-tions costs An overcondent investor overestimates the precisionof his information and thereby the expected gains of trading Hemay even trade when the true expected net gain is negative Sincemen are more overcondent than women this gives us two test-able hypotheses
H1 Men trade more than women
H2 By trading more men hurt their performance more than dowomen
It is these two hypotheses that are the focus of our inquiry5
II DATA AND METHODS
IIA Household Account and Demographic Data
Our main results focus on the common stock investments of37664 households for which we are able to identify the gender ofthe person who opened the householdrsquos rst brokerage accountThis sample is compiled from two data sets
Our primary data set is information from a large discountbrokerage rm on the investments of 78000 households for the sixyears ending in December 1996 For this period we have end-of-month position statements and trades that allow us to reasonablyestimate monthly returns from February 1991 through January
4 Some respondents answered that they expected market returns as high as900 We suspect that these respondents were thinking of index point moves ratherthan percentage returns Therefore we have dropped from our calculations re-spondents who gave answers of more than 100 to this question If alternativelywe Windsorize answers over 900 at 100 there is no signicant change in ourresults
5 Overcondence models also imply that more overcondent investors willhold riskier portfolios In Section III we present evidence that men hold riskiercommon stock portfolios than women However gender differences in portfoliorisk may be due to differences in risk tolerance rather than (or in addition to)differences in overcondence
266 QUARTERLY JOURNAL OF ECONOMICS
1997 The data set includes all accounts opened by the 78000 house-holds at this discount brokerage rm Sampled households wererequired to have an open account with the discount brokerage rmduring 1991 Roughly half of the accounts in our analysis wereopened prior to 1987 while half were opened between 1987 and1991 On average men opened their rst account at this brokerage47 years before the beginning of our sample period while womenopened theirs 43 years before During the sample period menrsquosaccounts held common stocks for 58 months on average and wom-enrsquos for 59 months The median number of months men held com-mon stocks is 70 For women it is 71
In this research we focus on the common stock investmentsof households We exclude investments in mutual funds (bothopen- and closed-end) American depository receipts (ADRs) war-rants and options Of the 78000 sampled households 66465 hadpositions in common stocks during at least one month the re-maining accounts either held cash or investments in other thanindividual common stocks The average household had approxi-mately two accounts and roughly 60 percent of the market valuein these accounts was held in common stocks These householdsmade over 3 million trades in all securities during our sampleperiod common stocks accounted for slightly more than 60 per-cent of all trades The average household held four stocks worth$47000 during our sample period although each of these guresis positively skewed6 The median household held 26 stocksworth $16000 In aggregate these households held more than$45 billion in common stocks in December 1996
Our secondary data set is demographic information compiled byInfobase Inc (as of June 8 1997) and provided to us by the broker-age house These data identify the gender of the person who openeda householdrsquos rst account for 37664 households of which 29659(79 percent) had accounts opened by men and 8005 (21 percent) hadaccounts opened by women In addition to gender Infobase providesdata on marital status the presence of children age and householdincome We present descriptive statistics in Table I Panel A Thesedata reveal that the women in our sample are less likely to bemarried and to have children than men The mean and median agesof the men and women in our sample are roughly equal The womenreport slightly lower household income although the difference isnot economically large
6 Throughout the paper portfolio values are reported in current dollars
267BOYS WILL BE BOYS
TA
BL
EI
DE
SC
RIP
TIV
ES
TA
TIS
TIC
SF
OR
DE
MO
GR
AP
HIC
SO
FF
EM
AL
EA
ND
MA
LE
HO
US
EH
OL
DS
Var
iabl
e
All
hous
ehol
dsM
arri
edho
use
hold
sS
ingl
eho
useh
olds
Wom
enM
en
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enmdash
men
)
Pan
elA
In
foba
seda
ta
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Per
cen
tage
mar
ried
680
757
27
7P
erce
nta
gew
ith
chil
dren
252
322
27
033
640
42
68
106
105
01
Mea
nag
e50
950
30
649
951
12
12
530
482
48
Med
ian
age
480
480
00
480
480
00
500
460
40
Mea
nin
com
e($
000)
730
756
22
681
279
61
656
762
82
61
w
ith
inco
me
gt$1
250
0011
211
72
05
142
130
12
59
74
21
5
Pan
elB
Sel
f-re
port
edda
ta
Nu
mbe
rof
hou
seho
lds
263
711
226
170
77
700
652
218
4N
etw
orth
($00
0)90
thP
erce
ntil
e50
00
500
00
050
00
500
00
035
00
450
02
100
075
thP
erce
ntil
e20
00
250
02
500
250
025
00
00
175
020
00
225
0M
edia
n10
00
100
00
010
00
100
00
010
00
100
00
025
thP
erce
ntil
e60
074
52
145
625
745
212
040
062
02
220
10th
Per
cent
ile
270
370
210
035
037
02
20
200
350
215
0E
quit
yto
net
wor
th(
)M
ean
133
132
01
129
129
00
144
143
01
Med
ian
67
67
00
63
66
20
37
97
40
5In
vest
men
tex
peri
ence
()
Non
e5
43
42
04
73
41
37
43
04
4L
imit
ed46
834
112
744
934
210
752
633
319
3G
ood
391
485
29
440
848
52
77
333
488
215
5E
xten
sive
87
140
25
39
613
92
43
67
149
28
2
The
sam
ple
con
sist
sof
hous
ehol
dsw
ith
com
mon
stoc
kin
vest
men
tat
ala
rge
disc
oun
tbr
oker
age
rm
for
whi
chw
ear
eab
leto
iden
tify
the
gend
erof
the
pers
onw
hoop
ened
the
hous
ehol
drsquos
rs
tac
coun
tD
ata
onm
arit
alst
atus
ch
ildr
en
age
and
inco
me
are
from
Info
base
Inc
asof
June
1997
S
elf-
repo
rted
data
are
info
rmat
ion
supp
lied
toth
edi
scou
ntbr
oker
age
rm
atth
eti
me
the
acco
unt
isop
ened
byth
epe
rson
onop
enin
gth
eac
coun
tIn
com
eis
repo
rted
wit
hin
eigh
tra
nges
whe
reth
eto
pra
nge
isgr
eate
rth
an$1
250
00W
eca
lcul
ate
mea
nsus
ing
the
mid
poin
tof
each
ran
gean
d$1
250
00fo
rth
eto
pra
nge
Equ
ity
toN
etW
orth
()i
sth
epr
opor
tion
ofth
em
arke
tva
lue
ofco
mm
onst
ock
inve
stm
ent
atth
isdi
scou
ntbr
oker
age
rm
asof
Janu
ary
1991
toto
tals
elf-
repo
rted
net
wor
thw
hen
the
hou
seho
ldop
ened
its
rst
acco
unt
atth
isbr
oker
age
Tho
seh
ouse
hold
sw
ith
apr
opor
tion
equ
ity
tone
tw
orth
grea
ter
than
100
perc
ent
are
dele
ted
wh
enca
lcul
atin
gm
ean
san
dm
edia
ns
Nu
mbe
rof
obse
rvat
ion
sfo
rea
chva
riab
leis
slig
htl
yle
ssth
anth
en
umbe
rof
repo
rted
hous
ehol
ds
268 QUARTERLY JOURNAL OF ECONOMICS
In addition to the data from Infobase we also have a limitedamount of self-reported data collected at the time each householdrst opened an account at the brokerage (and not subsequentlyupdated) which we summarize in Table I Panel B Of particularinterest to us are two variables net worth and investment expe-rience For this limited sample (about one-third of our total sam-ple) the distribution of net worth for women is slightly less thanthat for men although the difference is not economically largeFor this limited sample we also calculate the ratio of the marketvalue of equity (as of the rst month that the account appears inour data set) to self-reported net worth (which is reported at thetime the account is opened) This provides a measure albeitcrude of the proportion of a householdrsquos net worth that is in-vested in the common stocks that we analyze (If this ratio isgreater than one we delete the observation from our analysis)The mean household holds about 13 percent of its net worth in thecommon stocks we analyze and there is little difference in thisratio between men and women
The differences in self-reported experience by gender arequite large In general women report having less investmentexperience than men For example 478 percent of women reporthaving good or extensive investment experience while 625 per-cent of men report the same level of experience
Married couples may inuence each otherrsquos investment deci-sions In some cases the spouse making investment decisions maynot be the spouse who originally opened a brokerage accountThus we anticipate that observable differences in the investmentactivities of men and women will be greatest for single men andsingle women To investigate this possibility we partition ourdata on the basis of marital status The descriptive statistics fromthis partition are presented in the last six columns of Table I Formarried households we observe very small differences in ageincome the distribution of net worth and the ratio of net worthto equity Married women in our sample are less likely to havechildren than married men and they report having less invest-ment experience than men
For single households some differences in demographics be-come larger The average age of the single women in our sample isve years older than that of the single men the median is four yearsolder The average income of single women is $6100 less than thatof single men and fewer report having incomes in excess of$125000 Similarly the distribution of net worth for single women
269BOYS WILL BE BOYS
is lower than that of single men Finally single women reporthaving less investment experience than single men
IIB Return Calculations
To evaluate the investment performance of men and womenwe calculate the gross and net return performance of each house-hold The net return performance is calculated after a reasonableaccounting for the market impact commissions and bid-askspread of each trade
For each trade we estimate the bid-ask spread component oftransaction costs for purchases (sprdb
) or sales (sprds) as
sprds 5 S Pds
cl
Pds
s 2 1 D and sprdb 5 2 S Pdb
cl
Pdb
b 2 1 D Pds
cl and Pdb
cl are the reported closing prices from the Center forResearch in Security Prices (CRSP) daily stock return les on theday of a sale and purchase respectively Pds
s and Pdbb are the
actual sale and purchase price from our account database Ourestimate of the bid-ask spread component of transaction costsincludes any market impact that might result from a trade It alsoincludes an intraday return on the day of the trade The commis-sion component of transaction costs is calculated to be the dollarvalue of the commission paid scaled by the total principal value ofthe transaction both of which are reported in our account data
The average purchase costs an investor 031 percent whilethe average sale costs an investor 069 percent in bid-ask spreadOur estimate of the bid-ask spread is very close to the trading costof 021 percent for purchases and 063 percent for sales paid byopen-end mutual funds from 1966 to 1993 [Carhart 1997]7 Theaverage purchase in excess of $1000 cost 158 percent in commis-sions while the average sale in excess of $1000 cost 145 percent8
We calculate trade-weighted (weighted by trade size) spreadsand commissions These gures can be thought of as the total cost
7 Odean [1999] nds that individual investors are more likely to both buyand sell particular stocks when the prices of those stocks are rising This tendencycan partially explain the asymmetry in buy and sell spreads Any intraday pricerises following transactions subtract from our estimate of the spread for buys andadd to our estimate of the spread for sells
8 To provide more representative descriptive statistics on percentage com-missions we exclude trades less than $1000 The inclusion of these trades resultsin a round-trip commission cost of 5 percent on average (21 percent for purchasesand 31 percent for sales)
270 QUARTERLY JOURNAL OF ECONOMICS
of conducting the $24 billion in common stock trades (approxi-mately $12 billion each in purchases and sales) Trade-sizeweighting has little effect on spread costs (027 percent for pur-chases and 069 percent for sales) but substantially reduces thecommission costs (077 percent for purchases and 066 percent forsales)
In sum the average trade in excess of $1000 incurs a round-trip transaction cost of about 1 percent for the bid-ask spread andabout 3 percent in commissions In aggregate round-trip tradescost about 1 percent for the bid-ask spread and about 14 percentin commissions
We estimate the gross monthly return on each common stockinvestment using the beginning-of-month position statementsfrom our household data and the CRSP monthly returns le In sodoing we make two simplifying assumptions First we assumethat all securities are bought or sold on the last day of the monthThus we ignore the returns earned on stocks purchased from thepurchase date to the end of the month and include the returnsearned on stocks sold from the sale date to the end of the monthSecond we ignore intramonth trading (eg a purchase on March6 and a sale of the same security on March 20) although we doinclude in our analysis short-term trades that yield a position atthe end of a calendar month Barber and Odean [2000] provide acareful analysis of both of these issues and document that thesesimplifying assumptions yield trivial differences in our returncalculations
Consider the common stock portfolio for a particular house-hold The gross monthly return on the householdrsquos portfolio (Rht
gr)is calculated as
Rhtgr 5 O
i= 1
sht
pitRitgr
where pit is the beginning-of-month market value for the holdingof stock i by household h in month t divided by the beginning-of-month market value of all stocks held by household h Rit
gr is thegross monthly return for that stock and sht is the number ofstocks held by household h in month t
For security i in month t we calculate a monthly return netof transaction costs (Rit
net) as
(1 1 Ritnet) 5 (1 1 Rit
gr) (1 2 cits )(1 1 c it 2 1
b )
271BOYS WILL BE BOYS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
women do in nancial matters [Prince 1993] Indeed casual ob-servation reveals that men are disproportionately represented inthe nancial industry We expect therefore that men will gen-erally be more overcondent about their ability to make nancialdecisions than women
Additionally Lenney [1977] reports that gender differencesin self-condence depend on the lack of clear and unambiguousfeedback When feedback is ldquounequivocal and immediately avail-able women do not make lower ability estimates than menHowever when such feedback is absent or ambiguous womenseem to have lower opinions of their abilities and often do under-estimate relative to menrdquo Feedback in the stock market is am-biguous All the more reason to expect men to be more condentthan women about their ability to make common stockinvestments
Gervais and Odean [1998] develop a model in which investorovercondence results from self-serving attribution bias Inves-tors in this model infer their own abilities from their successesand failures Due to their tendency to take too much credit fortheir successes they become overcondent Deaux and Farris[1977] Meehan and Overton [1986] and Beyer [1990] nd thatthe self-serving attribution bias is greater for men than forwomen And so men are likely to become more overcondent thanwomen
The previous study most like our own is Lewellen Lease andSchlarbaumrsquos [1977] analysis of survey answers and brokeragerecords (from 1964 through 1970) of 972 individual investorsLewellen Lease and Schlarbaumrsquos report that men spend moretime and money on security analysis rely less on their brokersmake more transactions believe that returns are more highlypredictable and anticipate higher possible returns than dowomen In all these ways men behave more like overcondentinvestors than do women
Additional evidence that men are more overcondent inves-tors than women comes from surveys conducted by the GallupOrganization for PaineWebber Gallup conducted the survey f-teen times between June 1998 and January 2000 There wereapproximately 1000 respondents per survey In addition to otherquestions respondents were asked ldquoWhat overall rate of returndo you expect to get on your portfolio in the NEXT twelvemonthsrdquo and ldquoThinking about the stock market more generallywhat overall rate of return do you think the stock market will
265BOYS WILL BE BOYS
provide investors during the coming twelve monthsrdquo On averageboth men and women expected their own portfolios to outperformthe market However men expected to outperform by a greatermargin (28 percent) than did women (21 percent) The differencein the average anticipated outperformance of men and women isstatistically signicant (t = 33)4
In summary we have a natural experiment to (almost) di-rectly test theoretical models of investor overcondence A ratio-nal investor only trades if the expected gain exceeds the transac-tions costs An overcondent investor overestimates the precisionof his information and thereby the expected gains of trading Hemay even trade when the true expected net gain is negative Sincemen are more overcondent than women this gives us two test-able hypotheses
H1 Men trade more than women
H2 By trading more men hurt their performance more than dowomen
It is these two hypotheses that are the focus of our inquiry5
II DATA AND METHODS
IIA Household Account and Demographic Data
Our main results focus on the common stock investments of37664 households for which we are able to identify the gender ofthe person who opened the householdrsquos rst brokerage accountThis sample is compiled from two data sets
Our primary data set is information from a large discountbrokerage rm on the investments of 78000 households for the sixyears ending in December 1996 For this period we have end-of-month position statements and trades that allow us to reasonablyestimate monthly returns from February 1991 through January
4 Some respondents answered that they expected market returns as high as900 We suspect that these respondents were thinking of index point moves ratherthan percentage returns Therefore we have dropped from our calculations re-spondents who gave answers of more than 100 to this question If alternativelywe Windsorize answers over 900 at 100 there is no signicant change in ourresults
5 Overcondence models also imply that more overcondent investors willhold riskier portfolios In Section III we present evidence that men hold riskiercommon stock portfolios than women However gender differences in portfoliorisk may be due to differences in risk tolerance rather than (or in addition to)differences in overcondence
266 QUARTERLY JOURNAL OF ECONOMICS
1997 The data set includes all accounts opened by the 78000 house-holds at this discount brokerage rm Sampled households wererequired to have an open account with the discount brokerage rmduring 1991 Roughly half of the accounts in our analysis wereopened prior to 1987 while half were opened between 1987 and1991 On average men opened their rst account at this brokerage47 years before the beginning of our sample period while womenopened theirs 43 years before During the sample period menrsquosaccounts held common stocks for 58 months on average and wom-enrsquos for 59 months The median number of months men held com-mon stocks is 70 For women it is 71
In this research we focus on the common stock investmentsof households We exclude investments in mutual funds (bothopen- and closed-end) American depository receipts (ADRs) war-rants and options Of the 78000 sampled households 66465 hadpositions in common stocks during at least one month the re-maining accounts either held cash or investments in other thanindividual common stocks The average household had approxi-mately two accounts and roughly 60 percent of the market valuein these accounts was held in common stocks These householdsmade over 3 million trades in all securities during our sampleperiod common stocks accounted for slightly more than 60 per-cent of all trades The average household held four stocks worth$47000 during our sample period although each of these guresis positively skewed6 The median household held 26 stocksworth $16000 In aggregate these households held more than$45 billion in common stocks in December 1996
Our secondary data set is demographic information compiled byInfobase Inc (as of June 8 1997) and provided to us by the broker-age house These data identify the gender of the person who openeda householdrsquos rst account for 37664 households of which 29659(79 percent) had accounts opened by men and 8005 (21 percent) hadaccounts opened by women In addition to gender Infobase providesdata on marital status the presence of children age and householdincome We present descriptive statistics in Table I Panel A Thesedata reveal that the women in our sample are less likely to bemarried and to have children than men The mean and median agesof the men and women in our sample are roughly equal The womenreport slightly lower household income although the difference isnot economically large
6 Throughout the paper portfolio values are reported in current dollars
267BOYS WILL BE BOYS
TA
BL
EI
DE
SC
RIP
TIV
ES
TA
TIS
TIC
SF
OR
DE
MO
GR
AP
HIC
SO
FF
EM
AL
EA
ND
MA
LE
HO
US
EH
OL
DS
Var
iabl
e
All
hous
ehol
dsM
arri
edho
use
hold
sS
ingl
eho
useh
olds
Wom
enM
en
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enmdash
men
)
Pan
elA
In
foba
seda
ta
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Per
cen
tage
mar
ried
680
757
27
7P
erce
nta
gew
ith
chil
dren
252
322
27
033
640
42
68
106
105
01
Mea
nag
e50
950
30
649
951
12
12
530
482
48
Med
ian
age
480
480
00
480
480
00
500
460
40
Mea
nin
com
e($
000)
730
756
22
681
279
61
656
762
82
61
w
ith
inco
me
gt$1
250
0011
211
72
05
142
130
12
59
74
21
5
Pan
elB
Sel
f-re
port
edda
ta
Nu
mbe
rof
hou
seho
lds
263
711
226
170
77
700
652
218
4N
etw
orth
($00
0)90
thP
erce
ntil
e50
00
500
00
050
00
500
00
035
00
450
02
100
075
thP
erce
ntil
e20
00
250
02
500
250
025
00
00
175
020
00
225
0M
edia
n10
00
100
00
010
00
100
00
010
00
100
00
025
thP
erce
ntil
e60
074
52
145
625
745
212
040
062
02
220
10th
Per
cent
ile
270
370
210
035
037
02
20
200
350
215
0E
quit
yto
net
wor
th(
)M
ean
133
132
01
129
129
00
144
143
01
Med
ian
67
67
00
63
66
20
37
97
40
5In
vest
men
tex
peri
ence
()
Non
e5
43
42
04
73
41
37
43
04
4L
imit
ed46
834
112
744
934
210
752
633
319
3G
ood
391
485
29
440
848
52
77
333
488
215
5E
xten
sive
87
140
25
39
613
92
43
67
149
28
2
The
sam
ple
con
sist
sof
hous
ehol
dsw
ith
com
mon
stoc
kin
vest
men
tat
ala
rge
disc
oun
tbr
oker
age
rm
for
whi
chw
ear
eab
leto
iden
tify
the
gend
erof
the
pers
onw
hoop
ened
the
hous
ehol
drsquos
rs
tac
coun
tD
ata
onm
arit
alst
atus
ch
ildr
en
age
and
inco
me
are
from
Info
base
Inc
asof
June
1997
S
elf-
repo
rted
data
are
info
rmat
ion
supp
lied
toth
edi
scou
ntbr
oker
age
rm
atth
eti
me
the
acco
unt
isop
ened
byth
epe
rson
onop
enin
gth
eac
coun
tIn
com
eis
repo
rted
wit
hin
eigh
tra
nges
whe
reth
eto
pra
nge
isgr
eate
rth
an$1
250
00W
eca
lcul
ate
mea
nsus
ing
the
mid
poin
tof
each
ran
gean
d$1
250
00fo
rth
eto
pra
nge
Equ
ity
toN
etW
orth
()i
sth
epr
opor
tion
ofth
em
arke
tva
lue
ofco
mm
onst
ock
inve
stm
ent
atth
isdi
scou
ntbr
oker
age
rm
asof
Janu
ary
1991
toto
tals
elf-
repo
rted
net
wor
thw
hen
the
hou
seho
ldop
ened
its
rst
acco
unt
atth
isbr
oker
age
Tho
seh
ouse
hold
sw
ith
apr
opor
tion
equ
ity
tone
tw
orth
grea
ter
than
100
perc
ent
are
dele
ted
wh
enca
lcul
atin
gm
ean
san
dm
edia
ns
Nu
mbe
rof
obse
rvat
ion
sfo
rea
chva
riab
leis
slig
htl
yle
ssth
anth
en
umbe
rof
repo
rted
hous
ehol
ds
268 QUARTERLY JOURNAL OF ECONOMICS
In addition to the data from Infobase we also have a limitedamount of self-reported data collected at the time each householdrst opened an account at the brokerage (and not subsequentlyupdated) which we summarize in Table I Panel B Of particularinterest to us are two variables net worth and investment expe-rience For this limited sample (about one-third of our total sam-ple) the distribution of net worth for women is slightly less thanthat for men although the difference is not economically largeFor this limited sample we also calculate the ratio of the marketvalue of equity (as of the rst month that the account appears inour data set) to self-reported net worth (which is reported at thetime the account is opened) This provides a measure albeitcrude of the proportion of a householdrsquos net worth that is in-vested in the common stocks that we analyze (If this ratio isgreater than one we delete the observation from our analysis)The mean household holds about 13 percent of its net worth in thecommon stocks we analyze and there is little difference in thisratio between men and women
The differences in self-reported experience by gender arequite large In general women report having less investmentexperience than men For example 478 percent of women reporthaving good or extensive investment experience while 625 per-cent of men report the same level of experience
Married couples may inuence each otherrsquos investment deci-sions In some cases the spouse making investment decisions maynot be the spouse who originally opened a brokerage accountThus we anticipate that observable differences in the investmentactivities of men and women will be greatest for single men andsingle women To investigate this possibility we partition ourdata on the basis of marital status The descriptive statistics fromthis partition are presented in the last six columns of Table I Formarried households we observe very small differences in ageincome the distribution of net worth and the ratio of net worthto equity Married women in our sample are less likely to havechildren than married men and they report having less invest-ment experience than men
For single households some differences in demographics be-come larger The average age of the single women in our sample isve years older than that of the single men the median is four yearsolder The average income of single women is $6100 less than thatof single men and fewer report having incomes in excess of$125000 Similarly the distribution of net worth for single women
269BOYS WILL BE BOYS
is lower than that of single men Finally single women reporthaving less investment experience than single men
IIB Return Calculations
To evaluate the investment performance of men and womenwe calculate the gross and net return performance of each house-hold The net return performance is calculated after a reasonableaccounting for the market impact commissions and bid-askspread of each trade
For each trade we estimate the bid-ask spread component oftransaction costs for purchases (sprdb
) or sales (sprds) as
sprds 5 S Pds
cl
Pds
s 2 1 D and sprdb 5 2 S Pdb
cl
Pdb
b 2 1 D Pds
cl and Pdb
cl are the reported closing prices from the Center forResearch in Security Prices (CRSP) daily stock return les on theday of a sale and purchase respectively Pds
s and Pdbb are the
actual sale and purchase price from our account database Ourestimate of the bid-ask spread component of transaction costsincludes any market impact that might result from a trade It alsoincludes an intraday return on the day of the trade The commis-sion component of transaction costs is calculated to be the dollarvalue of the commission paid scaled by the total principal value ofthe transaction both of which are reported in our account data
The average purchase costs an investor 031 percent whilethe average sale costs an investor 069 percent in bid-ask spreadOur estimate of the bid-ask spread is very close to the trading costof 021 percent for purchases and 063 percent for sales paid byopen-end mutual funds from 1966 to 1993 [Carhart 1997]7 Theaverage purchase in excess of $1000 cost 158 percent in commis-sions while the average sale in excess of $1000 cost 145 percent8
We calculate trade-weighted (weighted by trade size) spreadsand commissions These gures can be thought of as the total cost
7 Odean [1999] nds that individual investors are more likely to both buyand sell particular stocks when the prices of those stocks are rising This tendencycan partially explain the asymmetry in buy and sell spreads Any intraday pricerises following transactions subtract from our estimate of the spread for buys andadd to our estimate of the spread for sells
8 To provide more representative descriptive statistics on percentage com-missions we exclude trades less than $1000 The inclusion of these trades resultsin a round-trip commission cost of 5 percent on average (21 percent for purchasesand 31 percent for sales)
270 QUARTERLY JOURNAL OF ECONOMICS
of conducting the $24 billion in common stock trades (approxi-mately $12 billion each in purchases and sales) Trade-sizeweighting has little effect on spread costs (027 percent for pur-chases and 069 percent for sales) but substantially reduces thecommission costs (077 percent for purchases and 066 percent forsales)
In sum the average trade in excess of $1000 incurs a round-trip transaction cost of about 1 percent for the bid-ask spread andabout 3 percent in commissions In aggregate round-trip tradescost about 1 percent for the bid-ask spread and about 14 percentin commissions
We estimate the gross monthly return on each common stockinvestment using the beginning-of-month position statementsfrom our household data and the CRSP monthly returns le In sodoing we make two simplifying assumptions First we assumethat all securities are bought or sold on the last day of the monthThus we ignore the returns earned on stocks purchased from thepurchase date to the end of the month and include the returnsearned on stocks sold from the sale date to the end of the monthSecond we ignore intramonth trading (eg a purchase on March6 and a sale of the same security on March 20) although we doinclude in our analysis short-term trades that yield a position atthe end of a calendar month Barber and Odean [2000] provide acareful analysis of both of these issues and document that thesesimplifying assumptions yield trivial differences in our returncalculations
Consider the common stock portfolio for a particular house-hold The gross monthly return on the householdrsquos portfolio (Rht
gr)is calculated as
Rhtgr 5 O
i= 1
sht
pitRitgr
where pit is the beginning-of-month market value for the holdingof stock i by household h in month t divided by the beginning-of-month market value of all stocks held by household h Rit
gr is thegross monthly return for that stock and sht is the number ofstocks held by household h in month t
For security i in month t we calculate a monthly return netof transaction costs (Rit
net) as
(1 1 Ritnet) 5 (1 1 Rit
gr) (1 2 cits )(1 1 c it 2 1
b )
271BOYS WILL BE BOYS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
provide investors during the coming twelve monthsrdquo On averageboth men and women expected their own portfolios to outperformthe market However men expected to outperform by a greatermargin (28 percent) than did women (21 percent) The differencein the average anticipated outperformance of men and women isstatistically signicant (t = 33)4
In summary we have a natural experiment to (almost) di-rectly test theoretical models of investor overcondence A ratio-nal investor only trades if the expected gain exceeds the transac-tions costs An overcondent investor overestimates the precisionof his information and thereby the expected gains of trading Hemay even trade when the true expected net gain is negative Sincemen are more overcondent than women this gives us two test-able hypotheses
H1 Men trade more than women
H2 By trading more men hurt their performance more than dowomen
It is these two hypotheses that are the focus of our inquiry5
II DATA AND METHODS
IIA Household Account and Demographic Data
Our main results focus on the common stock investments of37664 households for which we are able to identify the gender ofthe person who opened the householdrsquos rst brokerage accountThis sample is compiled from two data sets
Our primary data set is information from a large discountbrokerage rm on the investments of 78000 households for the sixyears ending in December 1996 For this period we have end-of-month position statements and trades that allow us to reasonablyestimate monthly returns from February 1991 through January
4 Some respondents answered that they expected market returns as high as900 We suspect that these respondents were thinking of index point moves ratherthan percentage returns Therefore we have dropped from our calculations re-spondents who gave answers of more than 100 to this question If alternativelywe Windsorize answers over 900 at 100 there is no signicant change in ourresults
5 Overcondence models also imply that more overcondent investors willhold riskier portfolios In Section III we present evidence that men hold riskiercommon stock portfolios than women However gender differences in portfoliorisk may be due to differences in risk tolerance rather than (or in addition to)differences in overcondence
266 QUARTERLY JOURNAL OF ECONOMICS
1997 The data set includes all accounts opened by the 78000 house-holds at this discount brokerage rm Sampled households wererequired to have an open account with the discount brokerage rmduring 1991 Roughly half of the accounts in our analysis wereopened prior to 1987 while half were opened between 1987 and1991 On average men opened their rst account at this brokerage47 years before the beginning of our sample period while womenopened theirs 43 years before During the sample period menrsquosaccounts held common stocks for 58 months on average and wom-enrsquos for 59 months The median number of months men held com-mon stocks is 70 For women it is 71
In this research we focus on the common stock investmentsof households We exclude investments in mutual funds (bothopen- and closed-end) American depository receipts (ADRs) war-rants and options Of the 78000 sampled households 66465 hadpositions in common stocks during at least one month the re-maining accounts either held cash or investments in other thanindividual common stocks The average household had approxi-mately two accounts and roughly 60 percent of the market valuein these accounts was held in common stocks These householdsmade over 3 million trades in all securities during our sampleperiod common stocks accounted for slightly more than 60 per-cent of all trades The average household held four stocks worth$47000 during our sample period although each of these guresis positively skewed6 The median household held 26 stocksworth $16000 In aggregate these households held more than$45 billion in common stocks in December 1996
Our secondary data set is demographic information compiled byInfobase Inc (as of June 8 1997) and provided to us by the broker-age house These data identify the gender of the person who openeda householdrsquos rst account for 37664 households of which 29659(79 percent) had accounts opened by men and 8005 (21 percent) hadaccounts opened by women In addition to gender Infobase providesdata on marital status the presence of children age and householdincome We present descriptive statistics in Table I Panel A Thesedata reveal that the women in our sample are less likely to bemarried and to have children than men The mean and median agesof the men and women in our sample are roughly equal The womenreport slightly lower household income although the difference isnot economically large
6 Throughout the paper portfolio values are reported in current dollars
267BOYS WILL BE BOYS
TA
BL
EI
DE
SC
RIP
TIV
ES
TA
TIS
TIC
SF
OR
DE
MO
GR
AP
HIC
SO
FF
EM
AL
EA
ND
MA
LE
HO
US
EH
OL
DS
Var
iabl
e
All
hous
ehol
dsM
arri
edho
use
hold
sS
ingl
eho
useh
olds
Wom
enM
en
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enmdash
men
)
Pan
elA
In
foba
seda
ta
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Per
cen
tage
mar
ried
680
757
27
7P
erce
nta
gew
ith
chil
dren
252
322
27
033
640
42
68
106
105
01
Mea
nag
e50
950
30
649
951
12
12
530
482
48
Med
ian
age
480
480
00
480
480
00
500
460
40
Mea
nin
com
e($
000)
730
756
22
681
279
61
656
762
82
61
w
ith
inco
me
gt$1
250
0011
211
72
05
142
130
12
59
74
21
5
Pan
elB
Sel
f-re
port
edda
ta
Nu
mbe
rof
hou
seho
lds
263
711
226
170
77
700
652
218
4N
etw
orth
($00
0)90
thP
erce
ntil
e50
00
500
00
050
00
500
00
035
00
450
02
100
075
thP
erce
ntil
e20
00
250
02
500
250
025
00
00
175
020
00
225
0M
edia
n10
00
100
00
010
00
100
00
010
00
100
00
025
thP
erce
ntil
e60
074
52
145
625
745
212
040
062
02
220
10th
Per
cent
ile
270
370
210
035
037
02
20
200
350
215
0E
quit
yto
net
wor
th(
)M
ean
133
132
01
129
129
00
144
143
01
Med
ian
67
67
00
63
66
20
37
97
40
5In
vest
men
tex
peri
ence
()
Non
e5
43
42
04
73
41
37
43
04
4L
imit
ed46
834
112
744
934
210
752
633
319
3G
ood
391
485
29
440
848
52
77
333
488
215
5E
xten
sive
87
140
25
39
613
92
43
67
149
28
2
The
sam
ple
con
sist
sof
hous
ehol
dsw
ith
com
mon
stoc
kin
vest
men
tat
ala
rge
disc
oun
tbr
oker
age
rm
for
whi
chw
ear
eab
leto
iden
tify
the
gend
erof
the
pers
onw
hoop
ened
the
hous
ehol
drsquos
rs
tac
coun
tD
ata
onm
arit
alst
atus
ch
ildr
en
age
and
inco
me
are
from
Info
base
Inc
asof
June
1997
S
elf-
repo
rted
data
are
info
rmat
ion
supp
lied
toth
edi
scou
ntbr
oker
age
rm
atth
eti
me
the
acco
unt
isop
ened
byth
epe
rson
onop
enin
gth
eac
coun
tIn
com
eis
repo
rted
wit
hin
eigh
tra
nges
whe
reth
eto
pra
nge
isgr
eate
rth
an$1
250
00W
eca
lcul
ate
mea
nsus
ing
the
mid
poin
tof
each
ran
gean
d$1
250
00fo
rth
eto
pra
nge
Equ
ity
toN
etW
orth
()i
sth
epr
opor
tion
ofth
em
arke
tva
lue
ofco
mm
onst
ock
inve
stm
ent
atth
isdi
scou
ntbr
oker
age
rm
asof
Janu
ary
1991
toto
tals
elf-
repo
rted
net
wor
thw
hen
the
hou
seho
ldop
ened
its
rst
acco
unt
atth
isbr
oker
age
Tho
seh
ouse
hold
sw
ith
apr
opor
tion
equ
ity
tone
tw
orth
grea
ter
than
100
perc
ent
are
dele
ted
wh
enca
lcul
atin
gm
ean
san
dm
edia
ns
Nu
mbe
rof
obse
rvat
ion
sfo
rea
chva
riab
leis
slig
htl
yle
ssth
anth
en
umbe
rof
repo
rted
hous
ehol
ds
268 QUARTERLY JOURNAL OF ECONOMICS
In addition to the data from Infobase we also have a limitedamount of self-reported data collected at the time each householdrst opened an account at the brokerage (and not subsequentlyupdated) which we summarize in Table I Panel B Of particularinterest to us are two variables net worth and investment expe-rience For this limited sample (about one-third of our total sam-ple) the distribution of net worth for women is slightly less thanthat for men although the difference is not economically largeFor this limited sample we also calculate the ratio of the marketvalue of equity (as of the rst month that the account appears inour data set) to self-reported net worth (which is reported at thetime the account is opened) This provides a measure albeitcrude of the proportion of a householdrsquos net worth that is in-vested in the common stocks that we analyze (If this ratio isgreater than one we delete the observation from our analysis)The mean household holds about 13 percent of its net worth in thecommon stocks we analyze and there is little difference in thisratio between men and women
The differences in self-reported experience by gender arequite large In general women report having less investmentexperience than men For example 478 percent of women reporthaving good or extensive investment experience while 625 per-cent of men report the same level of experience
Married couples may inuence each otherrsquos investment deci-sions In some cases the spouse making investment decisions maynot be the spouse who originally opened a brokerage accountThus we anticipate that observable differences in the investmentactivities of men and women will be greatest for single men andsingle women To investigate this possibility we partition ourdata on the basis of marital status The descriptive statistics fromthis partition are presented in the last six columns of Table I Formarried households we observe very small differences in ageincome the distribution of net worth and the ratio of net worthto equity Married women in our sample are less likely to havechildren than married men and they report having less invest-ment experience than men
For single households some differences in demographics be-come larger The average age of the single women in our sample isve years older than that of the single men the median is four yearsolder The average income of single women is $6100 less than thatof single men and fewer report having incomes in excess of$125000 Similarly the distribution of net worth for single women
269BOYS WILL BE BOYS
is lower than that of single men Finally single women reporthaving less investment experience than single men
IIB Return Calculations
To evaluate the investment performance of men and womenwe calculate the gross and net return performance of each house-hold The net return performance is calculated after a reasonableaccounting for the market impact commissions and bid-askspread of each trade
For each trade we estimate the bid-ask spread component oftransaction costs for purchases (sprdb
) or sales (sprds) as
sprds 5 S Pds
cl
Pds
s 2 1 D and sprdb 5 2 S Pdb
cl
Pdb
b 2 1 D Pds
cl and Pdb
cl are the reported closing prices from the Center forResearch in Security Prices (CRSP) daily stock return les on theday of a sale and purchase respectively Pds
s and Pdbb are the
actual sale and purchase price from our account database Ourestimate of the bid-ask spread component of transaction costsincludes any market impact that might result from a trade It alsoincludes an intraday return on the day of the trade The commis-sion component of transaction costs is calculated to be the dollarvalue of the commission paid scaled by the total principal value ofthe transaction both of which are reported in our account data
The average purchase costs an investor 031 percent whilethe average sale costs an investor 069 percent in bid-ask spreadOur estimate of the bid-ask spread is very close to the trading costof 021 percent for purchases and 063 percent for sales paid byopen-end mutual funds from 1966 to 1993 [Carhart 1997]7 Theaverage purchase in excess of $1000 cost 158 percent in commis-sions while the average sale in excess of $1000 cost 145 percent8
We calculate trade-weighted (weighted by trade size) spreadsand commissions These gures can be thought of as the total cost
7 Odean [1999] nds that individual investors are more likely to both buyand sell particular stocks when the prices of those stocks are rising This tendencycan partially explain the asymmetry in buy and sell spreads Any intraday pricerises following transactions subtract from our estimate of the spread for buys andadd to our estimate of the spread for sells
8 To provide more representative descriptive statistics on percentage com-missions we exclude trades less than $1000 The inclusion of these trades resultsin a round-trip commission cost of 5 percent on average (21 percent for purchasesand 31 percent for sales)
270 QUARTERLY JOURNAL OF ECONOMICS
of conducting the $24 billion in common stock trades (approxi-mately $12 billion each in purchases and sales) Trade-sizeweighting has little effect on spread costs (027 percent for pur-chases and 069 percent for sales) but substantially reduces thecommission costs (077 percent for purchases and 066 percent forsales)
In sum the average trade in excess of $1000 incurs a round-trip transaction cost of about 1 percent for the bid-ask spread andabout 3 percent in commissions In aggregate round-trip tradescost about 1 percent for the bid-ask spread and about 14 percentin commissions
We estimate the gross monthly return on each common stockinvestment using the beginning-of-month position statementsfrom our household data and the CRSP monthly returns le In sodoing we make two simplifying assumptions First we assumethat all securities are bought or sold on the last day of the monthThus we ignore the returns earned on stocks purchased from thepurchase date to the end of the month and include the returnsearned on stocks sold from the sale date to the end of the monthSecond we ignore intramonth trading (eg a purchase on March6 and a sale of the same security on March 20) although we doinclude in our analysis short-term trades that yield a position atthe end of a calendar month Barber and Odean [2000] provide acareful analysis of both of these issues and document that thesesimplifying assumptions yield trivial differences in our returncalculations
Consider the common stock portfolio for a particular house-hold The gross monthly return on the householdrsquos portfolio (Rht
gr)is calculated as
Rhtgr 5 O
i= 1
sht
pitRitgr
where pit is the beginning-of-month market value for the holdingof stock i by household h in month t divided by the beginning-of-month market value of all stocks held by household h Rit
gr is thegross monthly return for that stock and sht is the number ofstocks held by household h in month t
For security i in month t we calculate a monthly return netof transaction costs (Rit
net) as
(1 1 Ritnet) 5 (1 1 Rit
gr) (1 2 cits )(1 1 c it 2 1
b )
271BOYS WILL BE BOYS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
1997 The data set includes all accounts opened by the 78000 house-holds at this discount brokerage rm Sampled households wererequired to have an open account with the discount brokerage rmduring 1991 Roughly half of the accounts in our analysis wereopened prior to 1987 while half were opened between 1987 and1991 On average men opened their rst account at this brokerage47 years before the beginning of our sample period while womenopened theirs 43 years before During the sample period menrsquosaccounts held common stocks for 58 months on average and wom-enrsquos for 59 months The median number of months men held com-mon stocks is 70 For women it is 71
In this research we focus on the common stock investmentsof households We exclude investments in mutual funds (bothopen- and closed-end) American depository receipts (ADRs) war-rants and options Of the 78000 sampled households 66465 hadpositions in common stocks during at least one month the re-maining accounts either held cash or investments in other thanindividual common stocks The average household had approxi-mately two accounts and roughly 60 percent of the market valuein these accounts was held in common stocks These householdsmade over 3 million trades in all securities during our sampleperiod common stocks accounted for slightly more than 60 per-cent of all trades The average household held four stocks worth$47000 during our sample period although each of these guresis positively skewed6 The median household held 26 stocksworth $16000 In aggregate these households held more than$45 billion in common stocks in December 1996
Our secondary data set is demographic information compiled byInfobase Inc (as of June 8 1997) and provided to us by the broker-age house These data identify the gender of the person who openeda householdrsquos rst account for 37664 households of which 29659(79 percent) had accounts opened by men and 8005 (21 percent) hadaccounts opened by women In addition to gender Infobase providesdata on marital status the presence of children age and householdincome We present descriptive statistics in Table I Panel A Thesedata reveal that the women in our sample are less likely to bemarried and to have children than men The mean and median agesof the men and women in our sample are roughly equal The womenreport slightly lower household income although the difference isnot economically large
6 Throughout the paper portfolio values are reported in current dollars
267BOYS WILL BE BOYS
TA
BL
EI
DE
SC
RIP
TIV
ES
TA
TIS
TIC
SF
OR
DE
MO
GR
AP
HIC
SO
FF
EM
AL
EA
ND
MA
LE
HO
US
EH
OL
DS
Var
iabl
e
All
hous
ehol
dsM
arri
edho
use
hold
sS
ingl
eho
useh
olds
Wom
enM
en
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enmdash
men
)
Pan
elA
In
foba
seda
ta
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Per
cen
tage
mar
ried
680
757
27
7P
erce
nta
gew
ith
chil
dren
252
322
27
033
640
42
68
106
105
01
Mea
nag
e50
950
30
649
951
12
12
530
482
48
Med
ian
age
480
480
00
480
480
00
500
460
40
Mea
nin
com
e($
000)
730
756
22
681
279
61
656
762
82
61
w
ith
inco
me
gt$1
250
0011
211
72
05
142
130
12
59
74
21
5
Pan
elB
Sel
f-re
port
edda
ta
Nu
mbe
rof
hou
seho
lds
263
711
226
170
77
700
652
218
4N
etw
orth
($00
0)90
thP
erce
ntil
e50
00
500
00
050
00
500
00
035
00
450
02
100
075
thP
erce
ntil
e20
00
250
02
500
250
025
00
00
175
020
00
225
0M
edia
n10
00
100
00
010
00
100
00
010
00
100
00
025
thP
erce
ntil
e60
074
52
145
625
745
212
040
062
02
220
10th
Per
cent
ile
270
370
210
035
037
02
20
200
350
215
0E
quit
yto
net
wor
th(
)M
ean
133
132
01
129
129
00
144
143
01
Med
ian
67
67
00
63
66
20
37
97
40
5In
vest
men
tex
peri
ence
()
Non
e5
43
42
04
73
41
37
43
04
4L
imit
ed46
834
112
744
934
210
752
633
319
3G
ood
391
485
29
440
848
52
77
333
488
215
5E
xten
sive
87
140
25
39
613
92
43
67
149
28
2
The
sam
ple
con
sist
sof
hous
ehol
dsw
ith
com
mon
stoc
kin
vest
men
tat
ala
rge
disc
oun
tbr
oker
age
rm
for
whi
chw
ear
eab
leto
iden
tify
the
gend
erof
the
pers
onw
hoop
ened
the
hous
ehol
drsquos
rs
tac
coun
tD
ata
onm
arit
alst
atus
ch
ildr
en
age
and
inco
me
are
from
Info
base
Inc
asof
June
1997
S
elf-
repo
rted
data
are
info
rmat
ion
supp
lied
toth
edi
scou
ntbr
oker
age
rm
atth
eti
me
the
acco
unt
isop
ened
byth
epe
rson
onop
enin
gth
eac
coun
tIn
com
eis
repo
rted
wit
hin
eigh
tra
nges
whe
reth
eto
pra
nge
isgr
eate
rth
an$1
250
00W
eca
lcul
ate
mea
nsus
ing
the
mid
poin
tof
each
ran
gean
d$1
250
00fo
rth
eto
pra
nge
Equ
ity
toN
etW
orth
()i
sth
epr
opor
tion
ofth
em
arke
tva
lue
ofco
mm
onst
ock
inve
stm
ent
atth
isdi
scou
ntbr
oker
age
rm
asof
Janu
ary
1991
toto
tals
elf-
repo
rted
net
wor
thw
hen
the
hou
seho
ldop
ened
its
rst
acco
unt
atth
isbr
oker
age
Tho
seh
ouse
hold
sw
ith
apr
opor
tion
equ
ity
tone
tw
orth
grea
ter
than
100
perc
ent
are
dele
ted
wh
enca
lcul
atin
gm
ean
san
dm
edia
ns
Nu
mbe
rof
obse
rvat
ion
sfo
rea
chva
riab
leis
slig
htl
yle
ssth
anth
en
umbe
rof
repo
rted
hous
ehol
ds
268 QUARTERLY JOURNAL OF ECONOMICS
In addition to the data from Infobase we also have a limitedamount of self-reported data collected at the time each householdrst opened an account at the brokerage (and not subsequentlyupdated) which we summarize in Table I Panel B Of particularinterest to us are two variables net worth and investment expe-rience For this limited sample (about one-third of our total sam-ple) the distribution of net worth for women is slightly less thanthat for men although the difference is not economically largeFor this limited sample we also calculate the ratio of the marketvalue of equity (as of the rst month that the account appears inour data set) to self-reported net worth (which is reported at thetime the account is opened) This provides a measure albeitcrude of the proportion of a householdrsquos net worth that is in-vested in the common stocks that we analyze (If this ratio isgreater than one we delete the observation from our analysis)The mean household holds about 13 percent of its net worth in thecommon stocks we analyze and there is little difference in thisratio between men and women
The differences in self-reported experience by gender arequite large In general women report having less investmentexperience than men For example 478 percent of women reporthaving good or extensive investment experience while 625 per-cent of men report the same level of experience
Married couples may inuence each otherrsquos investment deci-sions In some cases the spouse making investment decisions maynot be the spouse who originally opened a brokerage accountThus we anticipate that observable differences in the investmentactivities of men and women will be greatest for single men andsingle women To investigate this possibility we partition ourdata on the basis of marital status The descriptive statistics fromthis partition are presented in the last six columns of Table I Formarried households we observe very small differences in ageincome the distribution of net worth and the ratio of net worthto equity Married women in our sample are less likely to havechildren than married men and they report having less invest-ment experience than men
For single households some differences in demographics be-come larger The average age of the single women in our sample isve years older than that of the single men the median is four yearsolder The average income of single women is $6100 less than thatof single men and fewer report having incomes in excess of$125000 Similarly the distribution of net worth for single women
269BOYS WILL BE BOYS
is lower than that of single men Finally single women reporthaving less investment experience than single men
IIB Return Calculations
To evaluate the investment performance of men and womenwe calculate the gross and net return performance of each house-hold The net return performance is calculated after a reasonableaccounting for the market impact commissions and bid-askspread of each trade
For each trade we estimate the bid-ask spread component oftransaction costs for purchases (sprdb
) or sales (sprds) as
sprds 5 S Pds
cl
Pds
s 2 1 D and sprdb 5 2 S Pdb
cl
Pdb
b 2 1 D Pds
cl and Pdb
cl are the reported closing prices from the Center forResearch in Security Prices (CRSP) daily stock return les on theday of a sale and purchase respectively Pds
s and Pdbb are the
actual sale and purchase price from our account database Ourestimate of the bid-ask spread component of transaction costsincludes any market impact that might result from a trade It alsoincludes an intraday return on the day of the trade The commis-sion component of transaction costs is calculated to be the dollarvalue of the commission paid scaled by the total principal value ofthe transaction both of which are reported in our account data
The average purchase costs an investor 031 percent whilethe average sale costs an investor 069 percent in bid-ask spreadOur estimate of the bid-ask spread is very close to the trading costof 021 percent for purchases and 063 percent for sales paid byopen-end mutual funds from 1966 to 1993 [Carhart 1997]7 Theaverage purchase in excess of $1000 cost 158 percent in commis-sions while the average sale in excess of $1000 cost 145 percent8
We calculate trade-weighted (weighted by trade size) spreadsand commissions These gures can be thought of as the total cost
7 Odean [1999] nds that individual investors are more likely to both buyand sell particular stocks when the prices of those stocks are rising This tendencycan partially explain the asymmetry in buy and sell spreads Any intraday pricerises following transactions subtract from our estimate of the spread for buys andadd to our estimate of the spread for sells
8 To provide more representative descriptive statistics on percentage com-missions we exclude trades less than $1000 The inclusion of these trades resultsin a round-trip commission cost of 5 percent on average (21 percent for purchasesand 31 percent for sales)
270 QUARTERLY JOURNAL OF ECONOMICS
of conducting the $24 billion in common stock trades (approxi-mately $12 billion each in purchases and sales) Trade-sizeweighting has little effect on spread costs (027 percent for pur-chases and 069 percent for sales) but substantially reduces thecommission costs (077 percent for purchases and 066 percent forsales)
In sum the average trade in excess of $1000 incurs a round-trip transaction cost of about 1 percent for the bid-ask spread andabout 3 percent in commissions In aggregate round-trip tradescost about 1 percent for the bid-ask spread and about 14 percentin commissions
We estimate the gross monthly return on each common stockinvestment using the beginning-of-month position statementsfrom our household data and the CRSP monthly returns le In sodoing we make two simplifying assumptions First we assumethat all securities are bought or sold on the last day of the monthThus we ignore the returns earned on stocks purchased from thepurchase date to the end of the month and include the returnsearned on stocks sold from the sale date to the end of the monthSecond we ignore intramonth trading (eg a purchase on March6 and a sale of the same security on March 20) although we doinclude in our analysis short-term trades that yield a position atthe end of a calendar month Barber and Odean [2000] provide acareful analysis of both of these issues and document that thesesimplifying assumptions yield trivial differences in our returncalculations
Consider the common stock portfolio for a particular house-hold The gross monthly return on the householdrsquos portfolio (Rht
gr)is calculated as
Rhtgr 5 O
i= 1
sht
pitRitgr
where pit is the beginning-of-month market value for the holdingof stock i by household h in month t divided by the beginning-of-month market value of all stocks held by household h Rit
gr is thegross monthly return for that stock and sht is the number ofstocks held by household h in month t
For security i in month t we calculate a monthly return netof transaction costs (Rit
net) as
(1 1 Ritnet) 5 (1 1 Rit
gr) (1 2 cits )(1 1 c it 2 1
b )
271BOYS WILL BE BOYS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EI
DE
SC
RIP
TIV
ES
TA
TIS
TIC
SF
OR
DE
MO
GR
AP
HIC
SO
FF
EM
AL
EA
ND
MA
LE
HO
US
EH
OL
DS
Var
iabl
e
All
hous
ehol
dsM
arri
edho
use
hold
sS
ingl
eho
useh
olds
Wom
enM
en
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enndash
men
)W
omen
Men
Dif
fere
nce
(wom
enmdash
men
)
Pan
elA
In
foba
seda
ta
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Per
cen
tage
mar
ried
680
757
27
7P
erce
nta
gew
ith
chil
dren
252
322
27
033
640
42
68
106
105
01
Mea
nag
e50
950
30
649
951
12
12
530
482
48
Med
ian
age
480
480
00
480
480
00
500
460
40
Mea
nin
com
e($
000)
730
756
22
681
279
61
656
762
82
61
w
ith
inco
me
gt$1
250
0011
211
72
05
142
130
12
59
74
21
5
Pan
elB
Sel
f-re
port
edda
ta
Nu
mbe
rof
hou
seho
lds
263
711
226
170
77
700
652
218
4N
etw
orth
($00
0)90
thP
erce
ntil
e50
00
500
00
050
00
500
00
035
00
450
02
100
075
thP
erce
ntil
e20
00
250
02
500
250
025
00
00
175
020
00
225
0M
edia
n10
00
100
00
010
00
100
00
010
00
100
00
025
thP
erce
ntil
e60
074
52
145
625
745
212
040
062
02
220
10th
Per
cent
ile
270
370
210
035
037
02
20
200
350
215
0E
quit
yto
net
wor
th(
)M
ean
133
132
01
129
129
00
144
143
01
Med
ian
67
67
00
63
66
20
37
97
40
5In
vest
men
tex
peri
ence
()
Non
e5
43
42
04
73
41
37
43
04
4L
imit
ed46
834
112
744
934
210
752
633
319
3G
ood
391
485
29
440
848
52
77
333
488
215
5E
xten
sive
87
140
25
39
613
92
43
67
149
28
2
The
sam
ple
con
sist
sof
hous
ehol
dsw
ith
com
mon
stoc
kin
vest
men
tat
ala
rge
disc
oun
tbr
oker
age
rm
for
whi
chw
ear
eab
leto
iden
tify
the
gend
erof
the
pers
onw
hoop
ened
the
hous
ehol
drsquos
rs
tac
coun
tD
ata
onm
arit
alst
atus
ch
ildr
en
age
and
inco
me
are
from
Info
base
Inc
asof
June
1997
S
elf-
repo
rted
data
are
info
rmat
ion
supp
lied
toth
edi
scou
ntbr
oker
age
rm
atth
eti
me
the
acco
unt
isop
ened
byth
epe
rson
onop
enin
gth
eac
coun
tIn
com
eis
repo
rted
wit
hin
eigh
tra
nges
whe
reth
eto
pra
nge
isgr
eate
rth
an$1
250
00W
eca
lcul
ate
mea
nsus
ing
the
mid
poin
tof
each
ran
gean
d$1
250
00fo
rth
eto
pra
nge
Equ
ity
toN
etW
orth
()i
sth
epr
opor
tion
ofth
em
arke
tva
lue
ofco
mm
onst
ock
inve
stm
ent
atth
isdi
scou
ntbr
oker
age
rm
asof
Janu
ary
1991
toto
tals
elf-
repo
rted
net
wor
thw
hen
the
hou
seho
ldop
ened
its
rst
acco
unt
atth
isbr
oker
age
Tho
seh
ouse
hold
sw
ith
apr
opor
tion
equ
ity
tone
tw
orth
grea
ter
than
100
perc
ent
are
dele
ted
wh
enca
lcul
atin
gm
ean
san
dm
edia
ns
Nu
mbe
rof
obse
rvat
ion
sfo
rea
chva
riab
leis
slig
htl
yle
ssth
anth
en
umbe
rof
repo
rted
hous
ehol
ds
268 QUARTERLY JOURNAL OF ECONOMICS
In addition to the data from Infobase we also have a limitedamount of self-reported data collected at the time each householdrst opened an account at the brokerage (and not subsequentlyupdated) which we summarize in Table I Panel B Of particularinterest to us are two variables net worth and investment expe-rience For this limited sample (about one-third of our total sam-ple) the distribution of net worth for women is slightly less thanthat for men although the difference is not economically largeFor this limited sample we also calculate the ratio of the marketvalue of equity (as of the rst month that the account appears inour data set) to self-reported net worth (which is reported at thetime the account is opened) This provides a measure albeitcrude of the proportion of a householdrsquos net worth that is in-vested in the common stocks that we analyze (If this ratio isgreater than one we delete the observation from our analysis)The mean household holds about 13 percent of its net worth in thecommon stocks we analyze and there is little difference in thisratio between men and women
The differences in self-reported experience by gender arequite large In general women report having less investmentexperience than men For example 478 percent of women reporthaving good or extensive investment experience while 625 per-cent of men report the same level of experience
Married couples may inuence each otherrsquos investment deci-sions In some cases the spouse making investment decisions maynot be the spouse who originally opened a brokerage accountThus we anticipate that observable differences in the investmentactivities of men and women will be greatest for single men andsingle women To investigate this possibility we partition ourdata on the basis of marital status The descriptive statistics fromthis partition are presented in the last six columns of Table I Formarried households we observe very small differences in ageincome the distribution of net worth and the ratio of net worthto equity Married women in our sample are less likely to havechildren than married men and they report having less invest-ment experience than men
For single households some differences in demographics be-come larger The average age of the single women in our sample isve years older than that of the single men the median is four yearsolder The average income of single women is $6100 less than thatof single men and fewer report having incomes in excess of$125000 Similarly the distribution of net worth for single women
269BOYS WILL BE BOYS
is lower than that of single men Finally single women reporthaving less investment experience than single men
IIB Return Calculations
To evaluate the investment performance of men and womenwe calculate the gross and net return performance of each house-hold The net return performance is calculated after a reasonableaccounting for the market impact commissions and bid-askspread of each trade
For each trade we estimate the bid-ask spread component oftransaction costs for purchases (sprdb
) or sales (sprds) as
sprds 5 S Pds
cl
Pds
s 2 1 D and sprdb 5 2 S Pdb
cl
Pdb
b 2 1 D Pds
cl and Pdb
cl are the reported closing prices from the Center forResearch in Security Prices (CRSP) daily stock return les on theday of a sale and purchase respectively Pds
s and Pdbb are the
actual sale and purchase price from our account database Ourestimate of the bid-ask spread component of transaction costsincludes any market impact that might result from a trade It alsoincludes an intraday return on the day of the trade The commis-sion component of transaction costs is calculated to be the dollarvalue of the commission paid scaled by the total principal value ofthe transaction both of which are reported in our account data
The average purchase costs an investor 031 percent whilethe average sale costs an investor 069 percent in bid-ask spreadOur estimate of the bid-ask spread is very close to the trading costof 021 percent for purchases and 063 percent for sales paid byopen-end mutual funds from 1966 to 1993 [Carhart 1997]7 Theaverage purchase in excess of $1000 cost 158 percent in commis-sions while the average sale in excess of $1000 cost 145 percent8
We calculate trade-weighted (weighted by trade size) spreadsand commissions These gures can be thought of as the total cost
7 Odean [1999] nds that individual investors are more likely to both buyand sell particular stocks when the prices of those stocks are rising This tendencycan partially explain the asymmetry in buy and sell spreads Any intraday pricerises following transactions subtract from our estimate of the spread for buys andadd to our estimate of the spread for sells
8 To provide more representative descriptive statistics on percentage com-missions we exclude trades less than $1000 The inclusion of these trades resultsin a round-trip commission cost of 5 percent on average (21 percent for purchasesand 31 percent for sales)
270 QUARTERLY JOURNAL OF ECONOMICS
of conducting the $24 billion in common stock trades (approxi-mately $12 billion each in purchases and sales) Trade-sizeweighting has little effect on spread costs (027 percent for pur-chases and 069 percent for sales) but substantially reduces thecommission costs (077 percent for purchases and 066 percent forsales)
In sum the average trade in excess of $1000 incurs a round-trip transaction cost of about 1 percent for the bid-ask spread andabout 3 percent in commissions In aggregate round-trip tradescost about 1 percent for the bid-ask spread and about 14 percentin commissions
We estimate the gross monthly return on each common stockinvestment using the beginning-of-month position statementsfrom our household data and the CRSP monthly returns le In sodoing we make two simplifying assumptions First we assumethat all securities are bought or sold on the last day of the monthThus we ignore the returns earned on stocks purchased from thepurchase date to the end of the month and include the returnsearned on stocks sold from the sale date to the end of the monthSecond we ignore intramonth trading (eg a purchase on March6 and a sale of the same security on March 20) although we doinclude in our analysis short-term trades that yield a position atthe end of a calendar month Barber and Odean [2000] provide acareful analysis of both of these issues and document that thesesimplifying assumptions yield trivial differences in our returncalculations
Consider the common stock portfolio for a particular house-hold The gross monthly return on the householdrsquos portfolio (Rht
gr)is calculated as
Rhtgr 5 O
i= 1
sht
pitRitgr
where pit is the beginning-of-month market value for the holdingof stock i by household h in month t divided by the beginning-of-month market value of all stocks held by household h Rit
gr is thegross monthly return for that stock and sht is the number ofstocks held by household h in month t
For security i in month t we calculate a monthly return netof transaction costs (Rit
net) as
(1 1 Ritnet) 5 (1 1 Rit
gr) (1 2 cits )(1 1 c it 2 1
b )
271BOYS WILL BE BOYS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
In addition to the data from Infobase we also have a limitedamount of self-reported data collected at the time each householdrst opened an account at the brokerage (and not subsequentlyupdated) which we summarize in Table I Panel B Of particularinterest to us are two variables net worth and investment expe-rience For this limited sample (about one-third of our total sam-ple) the distribution of net worth for women is slightly less thanthat for men although the difference is not economically largeFor this limited sample we also calculate the ratio of the marketvalue of equity (as of the rst month that the account appears inour data set) to self-reported net worth (which is reported at thetime the account is opened) This provides a measure albeitcrude of the proportion of a householdrsquos net worth that is in-vested in the common stocks that we analyze (If this ratio isgreater than one we delete the observation from our analysis)The mean household holds about 13 percent of its net worth in thecommon stocks we analyze and there is little difference in thisratio between men and women
The differences in self-reported experience by gender arequite large In general women report having less investmentexperience than men For example 478 percent of women reporthaving good or extensive investment experience while 625 per-cent of men report the same level of experience
Married couples may inuence each otherrsquos investment deci-sions In some cases the spouse making investment decisions maynot be the spouse who originally opened a brokerage accountThus we anticipate that observable differences in the investmentactivities of men and women will be greatest for single men andsingle women To investigate this possibility we partition ourdata on the basis of marital status The descriptive statistics fromthis partition are presented in the last six columns of Table I Formarried households we observe very small differences in ageincome the distribution of net worth and the ratio of net worthto equity Married women in our sample are less likely to havechildren than married men and they report having less invest-ment experience than men
For single households some differences in demographics be-come larger The average age of the single women in our sample isve years older than that of the single men the median is four yearsolder The average income of single women is $6100 less than thatof single men and fewer report having incomes in excess of$125000 Similarly the distribution of net worth for single women
269BOYS WILL BE BOYS
is lower than that of single men Finally single women reporthaving less investment experience than single men
IIB Return Calculations
To evaluate the investment performance of men and womenwe calculate the gross and net return performance of each house-hold The net return performance is calculated after a reasonableaccounting for the market impact commissions and bid-askspread of each trade
For each trade we estimate the bid-ask spread component oftransaction costs for purchases (sprdb
) or sales (sprds) as
sprds 5 S Pds
cl
Pds
s 2 1 D and sprdb 5 2 S Pdb
cl
Pdb
b 2 1 D Pds
cl and Pdb
cl are the reported closing prices from the Center forResearch in Security Prices (CRSP) daily stock return les on theday of a sale and purchase respectively Pds
s and Pdbb are the
actual sale and purchase price from our account database Ourestimate of the bid-ask spread component of transaction costsincludes any market impact that might result from a trade It alsoincludes an intraday return on the day of the trade The commis-sion component of transaction costs is calculated to be the dollarvalue of the commission paid scaled by the total principal value ofthe transaction both of which are reported in our account data
The average purchase costs an investor 031 percent whilethe average sale costs an investor 069 percent in bid-ask spreadOur estimate of the bid-ask spread is very close to the trading costof 021 percent for purchases and 063 percent for sales paid byopen-end mutual funds from 1966 to 1993 [Carhart 1997]7 Theaverage purchase in excess of $1000 cost 158 percent in commis-sions while the average sale in excess of $1000 cost 145 percent8
We calculate trade-weighted (weighted by trade size) spreadsand commissions These gures can be thought of as the total cost
7 Odean [1999] nds that individual investors are more likely to both buyand sell particular stocks when the prices of those stocks are rising This tendencycan partially explain the asymmetry in buy and sell spreads Any intraday pricerises following transactions subtract from our estimate of the spread for buys andadd to our estimate of the spread for sells
8 To provide more representative descriptive statistics on percentage com-missions we exclude trades less than $1000 The inclusion of these trades resultsin a round-trip commission cost of 5 percent on average (21 percent for purchasesand 31 percent for sales)
270 QUARTERLY JOURNAL OF ECONOMICS
of conducting the $24 billion in common stock trades (approxi-mately $12 billion each in purchases and sales) Trade-sizeweighting has little effect on spread costs (027 percent for pur-chases and 069 percent for sales) but substantially reduces thecommission costs (077 percent for purchases and 066 percent forsales)
In sum the average trade in excess of $1000 incurs a round-trip transaction cost of about 1 percent for the bid-ask spread andabout 3 percent in commissions In aggregate round-trip tradescost about 1 percent for the bid-ask spread and about 14 percentin commissions
We estimate the gross monthly return on each common stockinvestment using the beginning-of-month position statementsfrom our household data and the CRSP monthly returns le In sodoing we make two simplifying assumptions First we assumethat all securities are bought or sold on the last day of the monthThus we ignore the returns earned on stocks purchased from thepurchase date to the end of the month and include the returnsearned on stocks sold from the sale date to the end of the monthSecond we ignore intramonth trading (eg a purchase on March6 and a sale of the same security on March 20) although we doinclude in our analysis short-term trades that yield a position atthe end of a calendar month Barber and Odean [2000] provide acareful analysis of both of these issues and document that thesesimplifying assumptions yield trivial differences in our returncalculations
Consider the common stock portfolio for a particular house-hold The gross monthly return on the householdrsquos portfolio (Rht
gr)is calculated as
Rhtgr 5 O
i= 1
sht
pitRitgr
where pit is the beginning-of-month market value for the holdingof stock i by household h in month t divided by the beginning-of-month market value of all stocks held by household h Rit
gr is thegross monthly return for that stock and sht is the number ofstocks held by household h in month t
For security i in month t we calculate a monthly return netof transaction costs (Rit
net) as
(1 1 Ritnet) 5 (1 1 Rit
gr) (1 2 cits )(1 1 c it 2 1
b )
271BOYS WILL BE BOYS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
is lower than that of single men Finally single women reporthaving less investment experience than single men
IIB Return Calculations
To evaluate the investment performance of men and womenwe calculate the gross and net return performance of each house-hold The net return performance is calculated after a reasonableaccounting for the market impact commissions and bid-askspread of each trade
For each trade we estimate the bid-ask spread component oftransaction costs for purchases (sprdb
) or sales (sprds) as
sprds 5 S Pds
cl
Pds
s 2 1 D and sprdb 5 2 S Pdb
cl
Pdb
b 2 1 D Pds
cl and Pdb
cl are the reported closing prices from the Center forResearch in Security Prices (CRSP) daily stock return les on theday of a sale and purchase respectively Pds
s and Pdbb are the
actual sale and purchase price from our account database Ourestimate of the bid-ask spread component of transaction costsincludes any market impact that might result from a trade It alsoincludes an intraday return on the day of the trade The commis-sion component of transaction costs is calculated to be the dollarvalue of the commission paid scaled by the total principal value ofthe transaction both of which are reported in our account data
The average purchase costs an investor 031 percent whilethe average sale costs an investor 069 percent in bid-ask spreadOur estimate of the bid-ask spread is very close to the trading costof 021 percent for purchases and 063 percent for sales paid byopen-end mutual funds from 1966 to 1993 [Carhart 1997]7 Theaverage purchase in excess of $1000 cost 158 percent in commis-sions while the average sale in excess of $1000 cost 145 percent8
We calculate trade-weighted (weighted by trade size) spreadsand commissions These gures can be thought of as the total cost
7 Odean [1999] nds that individual investors are more likely to both buyand sell particular stocks when the prices of those stocks are rising This tendencycan partially explain the asymmetry in buy and sell spreads Any intraday pricerises following transactions subtract from our estimate of the spread for buys andadd to our estimate of the spread for sells
8 To provide more representative descriptive statistics on percentage com-missions we exclude trades less than $1000 The inclusion of these trades resultsin a round-trip commission cost of 5 percent on average (21 percent for purchasesand 31 percent for sales)
270 QUARTERLY JOURNAL OF ECONOMICS
of conducting the $24 billion in common stock trades (approxi-mately $12 billion each in purchases and sales) Trade-sizeweighting has little effect on spread costs (027 percent for pur-chases and 069 percent for sales) but substantially reduces thecommission costs (077 percent for purchases and 066 percent forsales)
In sum the average trade in excess of $1000 incurs a round-trip transaction cost of about 1 percent for the bid-ask spread andabout 3 percent in commissions In aggregate round-trip tradescost about 1 percent for the bid-ask spread and about 14 percentin commissions
We estimate the gross monthly return on each common stockinvestment using the beginning-of-month position statementsfrom our household data and the CRSP monthly returns le In sodoing we make two simplifying assumptions First we assumethat all securities are bought or sold on the last day of the monthThus we ignore the returns earned on stocks purchased from thepurchase date to the end of the month and include the returnsearned on stocks sold from the sale date to the end of the monthSecond we ignore intramonth trading (eg a purchase on March6 and a sale of the same security on March 20) although we doinclude in our analysis short-term trades that yield a position atthe end of a calendar month Barber and Odean [2000] provide acareful analysis of both of these issues and document that thesesimplifying assumptions yield trivial differences in our returncalculations
Consider the common stock portfolio for a particular house-hold The gross monthly return on the householdrsquos portfolio (Rht
gr)is calculated as
Rhtgr 5 O
i= 1
sht
pitRitgr
where pit is the beginning-of-month market value for the holdingof stock i by household h in month t divided by the beginning-of-month market value of all stocks held by household h Rit
gr is thegross monthly return for that stock and sht is the number ofstocks held by household h in month t
For security i in month t we calculate a monthly return netof transaction costs (Rit
net) as
(1 1 Ritnet) 5 (1 1 Rit
gr) (1 2 cits )(1 1 c it 2 1
b )
271BOYS WILL BE BOYS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
of conducting the $24 billion in common stock trades (approxi-mately $12 billion each in purchases and sales) Trade-sizeweighting has little effect on spread costs (027 percent for pur-chases and 069 percent for sales) but substantially reduces thecommission costs (077 percent for purchases and 066 percent forsales)
In sum the average trade in excess of $1000 incurs a round-trip transaction cost of about 1 percent for the bid-ask spread andabout 3 percent in commissions In aggregate round-trip tradescost about 1 percent for the bid-ask spread and about 14 percentin commissions
We estimate the gross monthly return on each common stockinvestment using the beginning-of-month position statementsfrom our household data and the CRSP monthly returns le In sodoing we make two simplifying assumptions First we assumethat all securities are bought or sold on the last day of the monthThus we ignore the returns earned on stocks purchased from thepurchase date to the end of the month and include the returnsearned on stocks sold from the sale date to the end of the monthSecond we ignore intramonth trading (eg a purchase on March6 and a sale of the same security on March 20) although we doinclude in our analysis short-term trades that yield a position atthe end of a calendar month Barber and Odean [2000] provide acareful analysis of both of these issues and document that thesesimplifying assumptions yield trivial differences in our returncalculations
Consider the common stock portfolio for a particular house-hold The gross monthly return on the householdrsquos portfolio (Rht
gr)is calculated as
Rhtgr 5 O
i= 1
sht
pitRitgr
where pit is the beginning-of-month market value for the holdingof stock i by household h in month t divided by the beginning-of-month market value of all stocks held by household h Rit
gr is thegross monthly return for that stock and sht is the number ofstocks held by household h in month t
For security i in month t we calculate a monthly return netof transaction costs (Rit
net) as
(1 1 Ritnet) 5 (1 1 Rit
gr) (1 2 cits )(1 1 c it 2 1
b )
271BOYS WILL BE BOYS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
where cits is the cost of sales scaled by the sales price in month t
and c it 2 1b is the cost of purchases scaled by the purchase price in
month t 2 1 The cost of purchases and sales include the com-missions and bid-ask spread components which are estimatedindividually for each trade as previously described Thus for asecurity purchased in month t 2 1 and sold in month t both c it
s
and c it 2 1b are positive for a security that was neither purchased
in month t 2 1 nor sold in month t both cits and c it 2 1
b are zeroBecause the timing and cost of purchases and sales vary acrosshouseholds the net return for security i in month t will varyacross households The net monthly portfolio return for eachhousehold is
Rhtnet 5 O
i= 1
sht
pitRitnet
(If only a portion of the beginning-of-month position in stock i waspurchased or sold the transaction cost is only applied to theportion that was purchased or sold)
We estimate the average gross and net monthly returnsearned by men as
RMtgr 5
1nmt
Oh= 1
nmt
Rhtgr and RMnet 5
1nmt
Oh= 1
nmt
Rhtnet
where nmt is the number of male households with common stockinvestment in month t There are analogous calculations forwomen
IIC Turnover
We calculate the monthly portfolio turnover for each house-hold as one-half the monthly sales turnover plus one-half themonthly purchase turnover9 In each month during our sampleperiod we identify the common stocks held by each household atthe beginning of month t from their position statement To cal-culate monthly sales turnover we match these positions to sales
9 Sell turnover for household h in month t is calculated as S i= 1sht pit min (1
SitHit) where Sit is the number of shares in security i sold during the month pitis the value of stock i held at the beginning of month t scaled by the total value ofstock holdings and H it is the number of shares of security i held at the beginningof month t Buy turnover is calculated as S i= 1
sht pit+ 1 min (1 BitHit+ 1) where Bitis the number of shares of security i bought during the month
272 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
during month t The monthly sales turnover is calculated as theshares sold times the beginning-of-month price per share dividedby the total beginning-of-month market value of the householdrsquosportfolio To calculate monthly purchase turnover we matchthese positions to purchases during month t 2 1 The monthlypurchase turnover is calculated as the shares purchased timesthe beginning-of-month price per share divided by the total be-ginning-of-month market value of the portfolio10
IID The Effect of Trading on Return Performance
We calculate an ldquoown-benchmarkrdquo abnormal return for indi-vidual investors that is similar in spirit to those proposed byLakonishok Shleifer and Vishny [1992] and Grinblatt and Tit-man [1993] In this abnormal return calculation the benchmarkfor household h is the month t return of the beginning-of-yearportfolio held by household h11 denoted Rht
b It represents thereturn that the household would have earned if it had merely heldits beginning-of-year portfolio for the entire year The gross or netown-benchmark abnormal return is the return earned by house-hold h less the return of household hrsquos beginning-of-year portfolio( ARht
gr = Rhtgr 2 Rht
b or ARhtnet = Rht
net 2 Rhtb ) If the household did
not trade during the year the own-benchmark abnormal returnwould be zero for all twelve months during the year
In each month the abnormal returns across male householdsare averaged yielding a 72-month time-series of mean monthlyown-benchmark abnormal returns Statistical signicance is cal-culated using t-statistics based on this time-series ARt
gr[ s ( ARt
gr) Ouml 72] where
ARtgr 5
1nmt
Ot= 1
nmt
(Rhtgr 2 Rht
b )
10 If more shares were sold than were held at the beginning of the month(because for example an investor purchased additional shares after the begin-ning of the month) we assume the entire beginning-of-month position in thatsecurity was sold Similarly if more shares were purchased in the precedingmonth than were held in the position statement we assume that the entireposition was purchased in the preceding month Thus turnover as we havecalculated it cannot exceed 100 percent in a month
11 When calculating this benchmark we begin the year on February 1 Wedo so because our rst monthly position statements are from the month end ofJanuary 1991 If the stocks held by a household at the beginning of the year aremissing CRSP returns data during the year we assume that stock is invested inthe remainder of the householdrsquos portfolio
273BOYS WILL BE BOYS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
There is an analogous calculation of net abnormal returns formen gross abnormal returns for women and net abnormal re-turns for women12
The advantage of the own-benchmark abnormal return mea-sure is that it does not adjust returns according to a particularrisk model No model of risk is universally accepted furthermoreit may be inappropriate to adjust investorsrsquo returns for stockcharacteristics that they do not associate with risk The own-benchmark measure allows each household to self-select the in-vestment style and risk prole of its benchmark (ie the portfolioit held at the beginning of the year) thus emphasizing the effecttrading has on performance
IIE Security Selection
Our theory says that men will underperform women becausemen trade more and trading is costly An alternative cause ofunderperformance is inferior security selection Two investorswith similar initial portfolios and similar turnover will differ inperformance if one consistently makes poor security selectionsTo measure security selection ability we compare the returns ofstocks bought with those of stocks sold
In each month we construct a portfolio comprised of thosestocks purchased by men in the preceding twelve months Thereturns on this portfolio in month t are calculated as
R tpm 5 O
i= 1
npt
TitpmRit Y O
i= 1
npt
Titpm
where T itpm is the aggregate value of all purchases by men in
security i from month t 2 12 through t 2 1 Rit is the grossmonthly return of stock i in month t and npt is the number ofdifferent stocks purchased from month t 2 12 through t 2 1(Alternatively we weight by the number rather than the value oftrades) Four portfolios are constructed one for the purchases ofmen (Rt
pm) one for the purchases of women (Rtpw) one for the
sales of men (Rtsm) and one for the sales of women (Rt
sw)
12 Alternatively one can rst calculate the monthly time-series averageown-benchmark return for each household and then test the signicance of thecross-sectional average of these The t-statistics for the cross-sectional tests arelarger than those we report for the time-series tests
274 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
III RESULTS
IIIA Men versus Women
In Table II Panel A we present position values and turnoverrates for the portfolios held by men and women Women holdslightly but not dramatically smaller common stock portfolios($18371 versus $21975) Of greater interest is the difference inturnover between women and men Models of overcondencepredict that women who are generally less overcondent thanmen will trade less than men The empirical evidence is consis-tent with this prediction Women turn their portfolios over ap-proximately 53 percent annually (monthly turnover of 44 percenttimes twelve) while men turn their portfolios over approximately77 percent annually (monthly turnover of 64 percent timestwelve) We are able to comfortably reject the null hypothesis thatturnover rates are similar for men and women (at less than a 1percent level) Although the median turnover is substantially lessfor both men and women the differences in the median levels ofturnover are also reliably different between genders
In Table II Panel B we present the gross and net percentagemonthly own-benchmark abnormal returns for common stockportfolios held by women and men Women earn gross monthlyreturns that are 0041 percent lower than those earned by theportfolio they held at the beginning of the year while men earngross monthly returns that are 0069 percent lower than thoseearned by the portfolio they held at the beginning of the yearBoth shortfalls are statistically signicant at the 1 percent levelas is their 0028 difference (034 percent annually)
Turning to net own-benchmark returns we nd that womenearn net monthly returns that are 0143 percent lower than thoseearned by the portfolio they held at the beginning of the yearwhile men earn net monthly returns that are 0221 percent lowerthan those earned by the portfolio they held at the beginning ofthe year Again both shortfalls are statistically signicant at the1 percent level as is their difference of 0078 percent (094 percentannually)
Are the lower own-benchmark returns earned by men due tomore active trading or to poor security selection The calculationsdescribed in subsection IIE indicate that the stocks both men andwomen choose to sell earn reliably greater returns than the stocksthey choose to buy This is consistent with Odean [1999] whouses different data to show that the stocks individual investors
275BOYS WILL BE BOYS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EII
PO
SIT
ION
VA
LU
ET
UR
NO
VE
RA
ND
RE
TU
RN
PE
RF
OR
MA
NC
EO
FC
OM
MO
NS
TO
CK
INV
ES
TM
EN
TS
OF
FE
MA
LE
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eho
useh
olds
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
ren
ce(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
P
osit
ion
Val
ue
and
Tur
nove
rM
ean
[med
ian]
183
7121
975
23
604
17
754
222
932
453
9
196
5420
161
250
7
begi
nnin
gpo
siti
onva
lue
($)
[73
87]
[82
18]
[283
1]
[7
410
][8
175
][2
765]
[74
91]
[80
97]
[260
6]
Mea
n[m
edia
n]4
406
412
201
441
611
21
70
4
227
052
283
mon
thly
turn
over
()
[17
4][2
94]
[21
20]
[1
79]
[28
1][1
02]
[15
5][3
32]
[21
77]
Pan
elB
P
erfo
rman
ceO
wn-
benc
hmar
km
onth
ly2
004
1
20
069
0
028
2
005
0
20
068
0
018
20
029
20
074
0
045
abno
rmal
gros
sre
turn
()
(22
84)
(23
66)
(24
3)(2
289
)(2
367
)(1
28)
(21
64)
(23
60)
(25
3)
Ow
n-be
nchm
ark
mon
thly
20
143
2
022
1
007
8
20
154
2
021
4
006
0
20
121
2
024
2
012
0
abno
rmal
net
retu
rn(
)(2
970
)(2
108
3)(6
35)
(29
10)
(210
48)
(39
5)(2
668
)(2
111
5)(6
68)
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
T
ests
for
diff
eren
ces
inm
edia
ns
are
base
don
aW
ilcox
onsi
gn-r
ank
test
stat
isti
cH
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Beg
inni
ngpo
siti
onva
lue
isth
em
arke
tva
lue
ofco
mm
onst
ocks
held
inth
e
rst
mon
thth
atth
eho
use
hold
appe
ars
duri
ngou
rsa
mpl
epe
riod
M
ean
mon
thly
turn
over
isth
eav
erag
eof
sale
san
dpu
rch
ase
turn
over
[M
edia
nva
lues
are
inbr
acke
ts]
Ow
n-be
nch
mar
kab
norm
alre
turn
sar
eth
eav
erag
eh
ouse
hol
dpe
rcen
tage
mon
thly
abn
orm
alre
turn
calc
ula
ted
asth
ere
aliz
edm
onth
lyre
turn
for
ah
ouse
hol
dle
ssth
ere
turn
that
wou
ldha
vebe
enea
rned
ifth
eho
useh
old
had
held
the
begi
nnin
g-of
-yea
rpo
rtfo
lio
for
the
enti
reye
ar(i
e
the
twel
vem
onth
sbe
ginn
ing
Feb
ruar
y1)
T-s
tati
stic
sfo
rab
norm
alre
turn
sar
ein
pare
nthe
ses
and
are
calc
ulat
edu
sing
tim
e-se
ries
stan
dard
erro
rsac
ross
mon
ths
276 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
sell earn reliably greater returns than the stocks they buy Wend that the stocks men choose to purchase underperform thosethat they choose to sell by twenty basis points per month (t =2 279)13 The stocks women choose to purchase underperformthose they choose to sell by seventeen basis points per month (t =2 202) The difference in the underperformances of men andwomen is not statistically signicant (When we weight eachtrade equally rather than by its value menrsquos purchases under-perform their sales by 23 basis points per month and womenrsquospurchases underperform their sales by 22 basis points permonth) Both men and women detract from their returns (grossand net) by trading men simply do so more often
While not pertinent to our hypothesesmdashwhich predict thatovercondence leads to excessive trading and that this tradinghurts performancemdash one might want to compare the raw returnsof men with those of women During our sample period menearned average monthly gross and net returns of 1501 and 1325percent women earned average monthly gross and net returns of1482 and 1361 percent Menrsquos gross and net average monthlymarket-adjusted returns (the raw monthly return minus themonthly return on the CRSP value-weighted index) were 0081and 2 0095 percent womenrsquos gross and net average monthlymarket-adjusted returns were 0062 and 2 0059 percent14 Fornone of these returns are the differences between men andwomen statistically signicant The gross raw and market-ad-justed returns earned by men and women differed in part be-cause as we document in subsection IIID men tended to holdsmaller higher beta stocks than did women such stocks per-formed well in our sample period
In summary our main ndings are consistent with the twopredictions of the overcondence models First men who aremore overcondent than women trade more than women (asmeasured by monthly portfolio turnover) Second men lowertheir returns more through excessive trading than do women
13 This t-statistic is calculated as
t 5(Rt
pm 2 Rtsm)
s (Rtpm 2 Rt
sm) Icirc 72
14 The gross (net) annualized geometric mean returns earned by men andwomen were 187 (163) and 186 (169) percent respectively
277BOYS WILL BE BOYS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
Men lower their returns more than women because they trademore not because their security selections are worse
IIIB Single Men versus Single Women
If gender serves as a reasonable proxy for overcondence wewould expect the differences in portfolio turnover and net returnperformance to be larger between the accounts of single men andsingle women than between the accounts of married men andmarried women This is because as discussed above one spousemay make or inuence decisions for an account opened by theother To test this ancillary prediction we partition our sampleinto four groups married women married men single womenand single men Because we do not have marital status for allheads of households in our data set the total number of house-holds that we analyze here is less than that previously analyzedby about 4400
Position values and turnover rates of the portfolios held bythe four groups are presented in the last six columns of Table IIPanel A Married women tend to hold smaller common stockportfolios than married men these differences are smaller be-tween single men and single women Differences in turnover arelarger between single women and men than between marriedwomen and men thus conrming our ancillary prediction
In the last six columns of Table II Panel B we present thegross and net percentage monthly own-benchmark abnormal re-turns for common stock portfolios of the four groups The grossmonthly own-benchmark abnormal returns of single women( 2 0029) and of single men (2 0074) are statistically signicantat the 1 percent level as is their difference (0045mdashannually 054percent) We again stress that it is not the superior timing of thesecurity selections of women that leads to these gross returndifferences Men (and particularly single men) are simply morelikely to act (ie trade) despite their inferior ability
The net monthly own-benchmark abnormal returns of mar-ried women (2 0154) and married men ( 2 0214) are statisticallysignicant at the 1 percent level as is their difference (0060)The net monthly own-benchmark abnormal returns of singlewomen (2 0121) and of single men ( 2 0242) are statisticallysignicant at the 1 percent level as is their difference (0120mdashannually 14 percent) Single men underperform single women bysignicantly more than married men underperform marriedwomen (0120 2 060 = 060 t = 280)
278 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
In summary if married couples inuence each otherrsquos invest-ment decisions and thereby reduce the effects of gender differ-ences in overcondence then the results of this section are con-sistent with the predictions of the overcondence models Firstmen trade more than women and this difference is greatestbetween single men and women Second men lower their returnsmore through excessive trading than do women and this differ-ence is greatest between single men and women
IIIC Cross-Sectional Analysis of Turnover and Performance
Perhaps turnover and performance differ between men andwomen because gender correlates with other attributes that pre-dict turnover and performance We therefore consider severaldemographic characteristics known to affect nancial decision-making age marital status the presence of children in a house-hold and income
To assess whether the differences in turnover can be attrib-uted to these demographic characteristics we estimate a cross-sectional regression where the dependent variable is the observedaverage monthly turnover for each household The independentvariables in the regression include three dummy variables mari-tal status (one indicating single) gender (one indicating woman)and the presence of children (one indicating a household withchildren) In addition we estimate the interaction between mari-tal status and gender Finally we include the age of the personwho opened the account and household income Since our incomemeasure is truncated at $125000 we also include a dummyvariable if household income was greater than $12500015
We present the results of this analysis in column 2 of TableIII they support our earlier ndings The estimated dummyvariable on gender is highly signicant (t = 2 1276) and indi-cates that (ceteris paribus) the monthly turnover in marriedwomenrsquos accounts is 146 basis points less than in married menrsquosThe differences in turnover are signicantly more pronouncedbetween single women and single men ceteris paribus single
15 Average monthly turnover for each household is calculated for themonths during which common stock holdings are reported for that householdMarital status gender the presence of children age and income are from Info-basersquos records as of June 8 1997 Thus the dependent variable is observed beforethe independent variables This is also true for the cross-sectional tests reportedbelow
279BOYS WILL BE BOYS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
women trade 219 basis points (146 plus 73) less than single menOf the control variables we consider only age is signicantmonthly turnover declines by 31 basis points per decade thatwe age
We next consider whether our performance results can beexplained by other demographic characteristics To do so weestimate a cross-sectional regression in which the dependentvariable is the monthly own-benchmark abnormal net returnearned by each household The independent variables for the
TABLE IIICROSS-SECTIONAL REGRESSIONS OF TURNOVER OWN-BENCHMARK ABNORMAL
RETURN BETA AND SIZE FEBRUARY 1991 TO JANUARY 1997
Dependentvariable
Meanmonthlyturnover
()
Own-benchmarkabnormalnet return
Portfoliovolatility
Individualvolatility Beta
Sizecoefcient
Intercept 6269 2 0321 11466 11658 1226+ 0776(2 1147) (5885) (7098) (1144) (2216)
Single 0483 0002 0320 0330 0020 0079(424) (014) (340) (417) (212) (465)
Woman 2 1461 0058 2 0689 2 0682 2 0037 2 0136(2 1276) (427) (2 727) (2 854) (2 391) (2 800)
Single 3 2 0733 0027 2 0439 2 0540 2 0029 2 0138woman (2 338) (108) (2 245) (2 357) (2 160) (2 430)
Age10 2 0311 0002 2 0536 2 0393 2 0027 2 0055(2 926) (423) (2 1931) (2 1678) (2 955) ( 2 1100)
Children 2 0037 0008 2 0014 2 0051 2 0002 2 0008(2 040) (076) (2 019) (2 079) (2 022) (2 061)
Income 2 0002 00002 00003 0001 0003 00011000 (2 130) (133) (022) (138) (249) (031)
Income 2 00003 0027 0011 0012 2 0008 2 0018dummy (2 024) (154) (010) (011) (2 068) (2 082)
Adj R2 () 153 020 211 195 059 119
indicate signicantly different from zero at the 1 5 and 10 percent level respectively+ indicates signicantly different from one at the 1 percent levelEach regression is estimated using data from 26618 households The dependent variables are the mean
monthly percentage turnover for each household the mean monthly own-benchmark abnormal net return foreach household the portfolio volatility for each household the average volatility of the individual commonstocks held by each household estimated beta exposure for each household and estimated size exposure foreach household Own-benchmark abnormal net returns are calculated as the realized monthly return for ahousehold less the return that would have been earned if the household had held the beginning-of-yearportfolio for the entire year Portfolio volatility is the standard deviation of each householdrsquos monthly portfolioreturns Individual volatility is the average standard deviation of monthly returns over the previous threeyears for each stock in a householdrsquos portfolio The average is weighted equally across months and by positionsize within months The estimated exposures are the coefcient estimates on the independent variables fromtime-series regressions of the gross household excess return on the market excess return (Rmt 2 Rft) and azero-investment size portfolio (SMB t) Single is a dummy variable that takes a value of one if the primaryaccount holder (PAH) is single Woman is a dummy variable that takes a value of one if the primary accountholder is a woman Age is the age of the PAH Children is a dummy variable that takes a value of one if thehousehold has children Income is the income of the household and has a maximum value of $125000 WhenIncome is at this maximum Income dummy takes on a value of one (t-statistics are in parentheses)
280 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
regression are the same as those previously employed16 Theresults of this analysis presented in column 3 of Table III con-rm our earlier nding that men deduct more from their returnperformance by trading than do women The estimated dummyvariable on gender is highly signicant (t = 427) and indicatesthat (ceteris paribus) the monthly own-benchmark abnormal netreturn for married men is 58 basis points less than for marriedwomen The difference in the performance of single men andwomen 85 basis points a month (58 plus 27) is even greaterthan that of their married counterparts although not reliably soOf the control variables that we consider only age appears asstatistically signicant own-benchmark abnormal net returnsimprove by 02 basis points per decade that we age (The last fourcolumns of Table III are discussed in the following subsection)
IIID Portfolio Risk
In this subsection we estimate risk characteristics of thecommon stock investments of men and women Although not thecentral focus of our inquiry we believe the results that we presenthere are the rst to document that women tend to hold less riskypositions than men within their common stock portfolios Ouranalysis also provides additional evidence that men decreasetheir portfolio returns through trading more so than do women
We estimate market risk (beta) and the risk associated withsmall rms by estimating the following two-factor monthly time-series regression
(RMtgr 2 Rft) 5 a i 1 b i(Rmt 2 Rft) 1 s iSMBt 1 e it
where
Rft = the monthly return on T-Bills17
16 The statistical signicance of the results reported in this section shouldbe interpreted with caution On one hand the standard errors of the coefcientestimates are likely to be inated since the dependent variable is estimated witherror (Although we observe several months of each householdrsquos own-benchmarkreturns a householdrsquos observed own-benchmark returns may differ from itslong-run average) On the other hand these standard errors may be deatedsince different households may choose to trade in or out of the same securityresulting in cross-sectional dependence in the abnormal performance measureacross households We suspect that the problem of cross-sectional dependence isnot severe since the average household invests in only four securities
To reduce the undue inuence of a few households with position statementsfor only a few months we delete those households with less than three years ofpositions The tenor of our results is similar if we include these households
17 The return on T-bills is from Stocks Bonds Bills and Ination 1997Yearbook Ibbotson Associates Chicago IL
281BOYS WILL BE BOYS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
Rmt = the monthly return on a value-weighted market indexSMBt = the return on a value-weighted portfolio of small stocks
minus the return on a value-weighted portfolio of bigstocks18
a i = the interceptb i = the market betas i = coefcient of size risk ande it = the regression error term
The subscript i denotes parameter estimates and error termsfrom regression i where we estimate twelve regressions one eachfor the gross and net performances of the average man theaverage married man and the average single man and one eachfor the gross and net performance of the average woman theaverage married woman and the average single woman
In each regression the estimate of b i measures portfolio riskdue to covariance with the market portfolio The estimate of s i
measures risk associated with the size of the rms held in aportfolio a larger value of si denotes increased exposure to smallstocks Fama and French [1993] and Berk [1995] argue that rmsize is a proxy for risk19 Finally the intercept a i is an estimateof risk-adjusted return and thus provides an alternative perfor-mance measure to our own-benchmark abnormal return
The results of this analysis are presented in Table IV Thetime-series regressions of the gross average monthly excess re-turn earned by women and men on the market excess return anda size based zero-investment portfolio reveal that women holdless risky positions than men While relative to the total marketboth women and men tilt their portfolios toward high beta smallrms women do so less These regressions also conrm our nd-ing that men decrease their portfolio returns through tradingmore so than do women Men and women earn similar gross andnet returns however men do so by investing in smaller stocks
18 The construction of this portfolio is discussed in detail in Fama andFrench [1993] We thank Kenneth French for providing us with these data
19 Berk [1995] points out that systematic effects in returns are likely toappear in price since price is the value of future cash ows discounted by expectedreturn Thus size and the book-to-market ratio are likely to correlate withcross-sectional differences in expected returns Fama and French [1993] alsoclaim that size and the book-to-market ratio proxy for risk Not all authors agreethat book-to-market ratios are risk proxies (eg Lakonishok Shleifer and Vishny[1994]) Our qualitative results are unaffected by the inclusion of a book-to-market factor
282 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EIV
RIS
KE
XP
OS
UR
ES
AN
DR
ISK
-AD
JUS
TE
DR
ET
UR
NS
OF
CO
MM
ON
ST
OC
KIN
VE
ST
ME
NT
SO
FF
EM
AL
E
AN
DM
AL
EH
OU
SE
HO
LD
S F
EB
RU
AR
Y19
91T
OJA
NU
AR
Y19
97
All
hou
seho
lds
Mar
ried
hous
ehol
dsS
ingl
eh
ouse
hold
s
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)W
omen
Men
Dif
fere
nce
(wom
enndashm
en)
Wom
enM
enD
iffe
renc
e(w
omen
ndashmen
)
Nu
mbe
rof
hou
seho
lds
800
529
659
NA
489
419
741
NA
230
66
326
NA
Pan
elA
G
ross
aver
age
hou
seho
ldpe
rcen
tage
mon
thly
retu
rns
Tw
o-fa
ctor
mod
elin
terc
ept
20
044
20
083
003
92
005
12
008
20
031
20
036
20
099
006
3T
wo-
fact
orm
odel
coef
cie
ntes
tim
ate
on(R
mt
2R
ft)
105
0
108
1
20
031
1
053
1
075
0
022
103
5
108
8
005
3
Tw
o-fa
ctor
mod
elco
efc
ient
esti
mat
eon
SM
Bt
036
0
051
9
20
159
0
380
0
490
0
109
0
307
0
582
0
275
A
djus
ted
R2
938
920
653
934
921
528
944
915
706
Pan
elB
N
etav
erag
eho
useh
old
perc
enta
gem
onth
lyre
turn
s
Tw
o-fa
ctor
mod
elin
terc
ept
20
162
20
253
0
091
2
017
1
20
245
0
074
2
014
2
20
285
0
143
indi
cate
sign
ica
ntat
the
15
and
10pe
rcen
tle
vel
resp
ecti
vely
H
ouse
hold
sar
ecl
assi
ed
asfe
mal
eor
mal
eba
sed
onth
ege
nder
ofth
epe
rson
wh
oop
ened
the
acco
unt
Hou
seho
lds
are
clas
si
edas
mar
ried
orsi
ngle
base
don
the
mar
ital
stat
usof
the
head
ofho
use
hold
Coe
fci
ent
and
inte
rcep
tes
tim
ates
for
the
two-
fact
orm
odel
are
thos
efr
oma
tim
e-se
ries
regr
essi
onof
the
gros
s(n
et)
aver
age
hou
seho
ldex
cess
retu
rnon
the
mar
ket
exce
ssre
turn
(Rm
t2
Rf
t)
and
aze
ro-i
nve
stm
ent
size
port
foli
o(S
MB
t)
(RM
tgr
2R
ft)
=a
i+
bi(
Rm
t2
Rf
t)
+s i
SM
Bt
+e
it
283BOYS WILL BE BOYS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
with higher market risk20 The intercepts from the two-factorregressions of net returns (Table IV Panel B) suggest that aftera reasonable accounting for the higher market and size risks ofmenrsquos portfolios women earn net returns that are reliably higher(by nine basis points per month or 11 percent annually) thanthose earned by men21
Beta and size may not be the only two risk factors thatconcern individual investors These investors hold on averageonly four common stocks in their portfolios Those without othercommon stock or mutual fund holdings bear a great deal ofidiosyncratic risk To measure differences in the idiosyncraticrisk exposures of men and women we estimate the volatility oftheir common stock portfolios as well as the average volatility ofthe stocks they hold We calculate portfolio volatility as the stan-dard deviation of each householdrsquos monthly portfolio returns forthe months in which the household held common stocks Wecalculate the average volatility of the individual stocks they holdas the average standard deviation of monthly returns during theprevious three calendar years for each stock in a householdrsquosportfolio This average is weighted by position size within monthsand equally across months
To test whether men and women differ in the volatility oftheir portfolios and of the stocks they hold and to conrm thatthey differ in the market risk and size risk of their portfolios weestimate four additional cross-sectional regressions As in thecross-sectional regressions of turnover and own-benchmark re-turns the independent variables in these regressions includedummy variables for marital status gender the presence of chil-dren in the household and the interaction between marital statusand gender as well as variables for age and income and a dummy
20 During our sample period the mean monthly return on SMBt was sev-enteen basis points
21 Fama and French [1993] argue that the risk of common stock investmentscan be parsimoniously summarized as risk related to the market rm size and armrsquos book-to-market ratio When the return on a value-weighted portfolio of highbook-to-market stocks minus the return on a value-weighted portfolio of lowbook-to-market stocks (HMLt) is added as an independent variable to the two-factor monthly time-series regressions we nd that both men and women tilttheir portfolios toward high book-to-market stocks although men do more so Theintercepts from these regressions indicate that after accounting for the marketsize and book-to-market tilts of their portfolios women outperformed men bytwelve basis points a month (14 percent annually) Lyon Barber and Tsai [1999]document that intercept tests using the three-factor model are well specied inrandom samples and samples of large or small rms Thus the Fama-Frenchintercept tests account well for the small stock tilt of individual investors Duringour sample period the mean monthly return on HMLt was twenty basis points
284 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
variable for income over $125000 The dependent variable isalternately the volatility of each householdrsquos portfolio the aver-age volatility of the individual stocks held by each household thecoefcient on the market risk premium (ie beta) and the coef-cient on the size zero-investment portfolio Both coefcients areestimated from the two-factor model described above22
We present the results of these regressions in the last fourcolumns of Table III For all four risk measures (portfolio volatil-ity individual stock volatility beta and size) men invest inriskier positions than women Of the control variables that weconsider marital status age and income appear to be correlatedwith the riskiness of the stocks in which a household invests Theyoung and single hold more volatile portfolios composed of morevolatile stocks They are more willing to accept market risk and toinvest in small stocks Those with higher incomes are also morewilling to accept market risk These results are completely inkeeping with the common-sense notion that the young andwealthy with no dependents are willing to accept more invest-ment risk
The risk differences in the common stock portfolios held bymen and women are not surprising There is considerable evi-dence that men and women have different attitudes toward riskFrom survey responses of 5200 men and 6400 women BarskyJuster Kimball and Shapiro [1997] conclude that women aremore risk-averse than men Analyzing off-track betting slips for2000 men and 2000 women Bruce and Johnson [1994] nd thatmen take larger risks than women although they nd no evidenceof differences in performance Jianakoplos and Bernasek [1998]report that roughly 60 percent of the female respondents to the1989 Survey of Consumer Finances but only 40 percent of themen said they were not willing to take any nancial risks Kara-benick and Addy [1979] Sorrentino Hewitt and Raso-Knott[1992] and Zinkhan and Karande [1991] observe that men haveriskier preferences than women Flynn Slovic and Mertz [1994]Finucane Slovic Mertz Flynn and Sattereld [2000] and Finu-cane and Slovic [1999] nd that white men perceive a widevariety of risks as lower than do women and nonwhite menBajtelsmit and Bernasek [1996] Bajtelsmit and VanDerhei
22 These regressions are estimated for households with at least three yearsof available data Statistical inference might also be affected by measurementerror in the dependent variable and cross-sectional dependence in the risk mea-sures (see footnotes 11 and 12)
285BOYS WILL BE BOYS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
[1997] Hinz McCarthy and Turner [1997] and Sunden andSurette [1998] nd that men hold more of their retirement sav-ings in risky assets Jianakoplos and Bernasek [1998] report thesame for overall wealth Papke [1998] however nds in a sampleof near retirement women and their spouses that women do notinvest their pensions more conservatively than men
IV COMPETING EXPLANATIONS FOR DIFFERENCES
IN TURNOVER AND PERFORMANCE
IVA Risk Aversion
Since men and women differ in both overcondence and riskaversion it is natural to ask whether differences in risk aversionalone explain our ndings They do not While rational informedinvestors will trade more if they are less risk averse they will alsoimprove their performance by trading Thus if rational and in-formed men (and women) should improve their performance bytrading But both groups hurt their performance by trading Andmen do so more than women This outcome can be explained bydifferences in the overcondence of men and women and by dif-ferences in the risk aversion of overcondent men and women Itcannot be explained by differences in risk aversion alone
IVB Gambling
To what extent may gender differences in the propensity togamble explain the differences in turnover and returns that weobserve There are two aspects of gambling that we considerrisk-seeking and entertainment
Risk-seeking is when one demonstrates a preference for out-comes with greater variance but equal or lower expected returnIn equity markets the simplest way to increase variance withoutincreasing expected return is to underdiversify Excessive tradinghas a related but decidedly different effect it decreases expectedreturns without decreasing variance Thus risk-seeking may ac-count for underdiversication (although lack of diversicationcould also result from simple ignorance of its benets or fromovercondence) but it does not explain excessive trading
It may be that some men and to a lesser extent women tradefor entertainment They may enjoy placing trades that they ex-pect on average will lose money It is more likely that even thosewho enjoy trading believe overcondently that they have trading
286 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
ability This would be consistent with the Gallup Poll nding(reported in subsection IB) that both men and women expecttheir own portfolios to outperform the market but that menanticipate a greater outperformance
Some investors may set aside a small portion of their wealthto trade for entertainment while investing the majority moreprudently If ldquoentertainment accountsrdquo are driving our ndingswe would expect turnover and underperformance to decline as thecommon stocks in the accounts we observe represent a largerproportion of each householdrsquos total wealth We nd howeverthat this is not the case
Approximately one-third of our households reported their networth at the time they opened their accounts We calculate theproportion of net worth invested in the common stock portfolioswe observe as the beginning value of a householdrsquos common stockinvestments scaled by its self-reported net worth23 We then ana-lyze the turnover and investment performance of 2333 house-holds with at least 50 percent of their net worth invested incommon stock at this brokerage These households have similarturnover (625 percent per month24 75 percent annually) to ourfull sample (Table II) Furthermore these households earn grossand net returns that are very similar to the full sample
If households trade actively solely for entertainment then itis likely that the households that trade most actively nd tradingmost entertaining How much does active trading cost thesehouseholds compared with their other household expenses Toestimate the cost of trading for the 20 percent of households in oursample that trade most actively we calculate the monthly dollarloss as the own-benchmark abnormal return times the beginningof month position value This monthly dollar loss is averagedacross months for each household and multiplied by twelve toobtain an average annual cost of trading The average annual costof trading for the quintile of most active traders is $2849 TheDepartment of Labor Consumer Expenditure Survey reports onvarious categories of household expenditures The Survey parti-tions households into quintiles based on household income Theaverage annual income ($74000)25 of the quintile of most actively
23 This estimate is upwardly biased because the account opening dategenerally precedes our rst portfolio position observation and net worth is likelyto have increased in the interim
24 The standard error of the mean monthly turnover is 0225 This average annual income is somewhat understated since in calculating
287BOYS WILL BE BOYS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
trading households in our sample corresponds most closely to thatof the mean income ($92523) of the top quintile of the 1996Survey The active traders in our sample spent on average 39percent of their annual income on trading costs while comparablehouseholds in the Survey spent 33 percent of their income onutilities 26 percent on health care and 40 percent on all enter-tainment expenses including fees admissions television radioand sound equipment pets toys and playground equipment Ifthe active traders in our sample are trading solely for entertain-ment they must nd it very entertaining indeed
For mutual funds as for individuals turnover has a negativeimpact on returns [Carhart 1997]26 Some mutual fund managersmay actively trade and thereby knowingly reduce their fundsrsquoexpected returns simply to create the illusion that they areproviding a valuable service [Dow and Gorton 1994] If the ma-jority of active managers believe that they offer only disservice totheir clients this is a cynical industry indeed We propose thatmost mutual fund managers while aware that active manage-ment on average detracts value believe that their personal abil-ity to manage is above average Thus they are motivated to tradeby overcondence not cynicism
Individuals may trade for entertainment Mutual fund man-agers may trade to appear busy It is unlikely that most individ-uals churn their accounts to appear busy or that most fundmanagers trade for fun Overcondence offers a simple explana-tion for the high trading activity of both groups
V CONCLUSION
Modern nancial economics assumes that we behave withextreme rationality but we do not Furthermore our deviationsfrom rationality are often systematic Behavioral nance relaxesthe traditional assumptions of nancial economics by incorporat-ing these observable systematic and very human departuresfrom rationality into standard models of nancial markets Over-condence is one such departure Models that assume market
average household income we treat households in the over $125000 category asif their incomes were $125000 These account for less than 12 percent of thehouseholds for which we have income data
26 Lakonishok Shleifer and Vishny [1992] report a positive relation be-tween turnover and performance for 769 all-equity pension funds although thisnding puzzles the authors
288 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
participants are overcondent yield one central prediction over-condent investors will trade too much
We test this prediction by partitioning investors on the basisof a variable that provides a natural proxy for overcondencemdashgender Psychological research has established that men aremore prone to overcondence than women particularly so inmale-dominated realms such as nance Rational investors tradeonly if the expected gains exceed transactions costs Overcon-dent investors overestimate the precision of their information andthereby the expected gains of trading They may even trade whenthe true expected net gains are negative Models of investorovercondence predict that since men are more overcondentthan women men will trade more and perform worse thanwomen
Our empirical tests provide strong support for the behavioralnance model Men trade more than women and thereby reducetheir returns more so than do women Furthermore these differ-ences are most pronounced between single men and singlewomen
Individuals turn over their common stock investments about70 percent annually [Barber and Odean 2000] Mutual fundshave similar turnover rates [Carhart 1997] Yet those individu-als and mutual funds that trade most earn the lowest returns Webelieve that there is a simple and powerful explanation for thehigh levels of counterproductive trading in nancial marketsovercondence
UNIVERSITY OF CALIFORNIA DAVIS
REFERENCES
Alpert Marc and Howard Raiffa ldquoA Progress Report on the Training of Proba-bility Assessorsrdquo in Judgment Under Uncertainty Heuristics and BiasesDaniel Kahneman Paul Slovic and Amos Tversky eds (Cambridge and NewYork Cambridge University Press 1982) pp 294ndash305
Bajtelsmit Vickie L and Alexandra Bernasek ldquoWhy Do Women Invest Differ-ently than Menrdquo Financial Counseling and Planning VII (1996) 1ndash10
Bajtelsmit Vickie L and Jack L VanDerhie ldquoRisk Aversion and Pension Invest-ment Choicesrdquo in Positioning Pensions for the Twenty-rst Century MichaelS Gordon Olivia S Mitchell and Marc M Twinney eds (PhiladelphiaUniversity of Pennsylvania Press 1997) pp 45ndash65
Barber Brad M and Terrance Odean ldquoTrading Is Hazardous to Your WealthThe Common Stock Investment Performance of Individual Investorsrdquo Jour-nal of Finance LV (2000) 773ndash806
Barsky Robert B F Thomas Juster Miles S Kimball and Matthew D ShapiroldquoPreference Parameters and Behavioral Heterogeneity An Experimental Ap-proach in the Health and Retirement Studyrdquo Quarterly Journal of Econom-ics CXII (1997) 537ndash579
289BOYS WILL BE BOYS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
Baumann Andrea O Raisa B Deber and Gail G Thompson ldquoOvercondenceamong Physicians and Nurses The lsquoMicro-Certainty Macro-UncertaintyrsquoPhenomenonrdquo Social Science and Medicine XXXII (1991) 167ndash174
Benos Alexandros V ldquoOvercondent Speculators in Call Markets Trade Pat-terns and Survivalrdquo Journal of Financial Markets I (1998) 353ndash383
Berk Jonathan ldquoA Critique of Size Related Anomaliesrdquo Review of FinancialStudies VIII (1995) 275ndash286
Beyer Sylvia ldquoGender Differences in the Accuracy of Self-Evaluations of Perfor-mancerdquo Journal of Personality and Social Psychology LIX (1990) 960ndash970
Beyer Sylvia and Edward M Bowden ldquoGender Differences in Self-PerceptionsConvergent Evidence from Three Measures of Accuracy and Biasrdquo Personal-ity and Social Psychology Bulletin XXIII (1997) 157ndash172
Bruce Alistair C and Johnnie E Johnson ldquoMale and Female Betting BehaviourNew Perspectivesrdquo Journal of Gambling Studies X (1994) 183ndash198
Caballe Jordi and Jozsef Sakovics ldquoOvercondent Speculation with ImperfectCompetitionrdquo working paper Universitat Autonoma de Barcelona Spain1998
Carhart Mark M ldquoOn Persistence in Mutual Fund Performancerdquo Journal ofFinance LII (1997) 57ndash82
Christensen-Szalanski Jay J and James B Bushyhead ldquoPhysiciansrsquo Use ofProbabilistic Information in a Real Clinical Settingrdquo Journal of ExperimentalPsychology Human Perception and Performance VII (1981) 928ndash935
Cooper Arnold C Carolyn Y Woo and William C Dunkelberg ldquoEntrepreneursrsquoPerceived Chances for Successrdquo Journal of Business Venturing III (1988)97ndash108
Daniel Kent David Hirshleifer and Avanidar Subrahmanyam ldquoInvestor Psy-chology and Security Market Under- and Overreactionsrdquo Journal of FinanceLIII (1998) 1839ndash1885
Deaux Kay and Tim Emswiller ldquoExplanations of Successful Performance onSex-Linked Tasks What Is Skill for the Male is Luck for the Femalerdquo Journalof Personality and Social Psychology XXIX (1974) 80ndash85
Deaux Kay and Elizabeth Farris ldquoAttributing Causes for Onersquos Own Perfor-mance The Effects of Sex Norms and Outcomerdquo Journal of Research inPersonality XI (1977) 59ndash72
De Long J Bradford Andrei Shleifer Lawrence H Summers and Robert JWaldmann ldquoThe Survival of Noise Traders in Financial Marketsrdquo Journal ofBusiness LXIV (1991) 1ndash19
Dow James and Gary Gorton ldquoNoise Trading Delegated Portfolio Managementand Economic Welfarerdquo Journal of Political Economy CV (1994) 1024ndash1050
Fama Eugene F and Kenneth R French ldquoCommon Risk Factors in Returns onStocks and Bondsrdquo Journal of Financial Economics XXXIII (1993) 3ndash56
Finucane Melissa and Paul Slovic ldquoRisk and the White Male A Perspective onPerspectivesrdquo Framtider XVIII (1999) 24ndash29
Finucane Melissa Paul Slovic C K Mertz James Flynn and Theresa Satter-eld ldquoGender Race and Perceived Risk The lsquoWhite Malersquo Effectrdquo HealthRisk and Society II (2000) 159ndash179
Fischhoff Baruch Paul Slovic and Sarah Lichtenstein ldquoKnowing with CertaintyThe Appropriateness of Extreme Condencerdquo Journal of Experimental Psy-chology III (1977) 552ndash564
Flynn James Paul Slovic and C K Mertz ldquoGender Race and Perception ofEnvironmental Health Risksrdquo Risk Analysis XIV (1994) 1101ndash1108
Frank Jerome D ldquoSome Psychological Determinants of the Level of AspirationrdquoAmerican Journal of Psychology XLVII (1935) 285ndash293
Gervais Simon and Terrance Odean ldquoLearning to Be Overcondentrdquo workingpaper Wharton School University of Pennsylvania 1998
Grifn Dale and Amos Tversky ldquoThe Weighing of Evidence and the Determi-nants of Condencerdquo Cognitive Psychology XXIV (1992) 411ndash435
Grinblatt Mark and Sheridan Titman ldquoPerformance Measurement withoutBenchmarks An Examination of Mutual Fund Returnsrdquo Journal of BusinessLXVI (1993) 47ndash68
Grossman Sanford J and Joseph E Stiglitz ldquoOn the Impossibility of Informa-
290 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
tionally Efcient Marketsrdquo American Economic Review LXX (1980)393ndash408
Harris Milton and Artur Raviv ldquoDifferences of Opinion Make a Horse RacerdquoReview of Financial Studies VI (1993) 473ndash506
Hinz Richard P David D McCarthy and John A Turner ldquoAre Women Conser-vative Investors Gender Differences in Participant-Directed Pension Invest-mentsrdquo in Positioning Pensions for the Twenty-rst Century Michael SGordon Olivia S Mitchell and Marc M Twinney eds (Philadelphia Uni-versity of Pennsylvania Press 1997) pp 91ndash103
Jianakoplos Nancy A and Alexandra Bernasek ldquoAre Women More Risk AverserdquoEconomic Inquiry XXXVI (1998) 620ndash630
Karabenick Stuart A and Milton M Addy ldquoLocus of Control and Sex Differencesin Skill and Chance Risk-Taking Conditionsrdquo Journal of General PsychologyC (1979) 215ndash228
Kidd John B ldquoThe Utilization of Subjective Probabilities in Production Plan-ningrdquo Acta Psychologica XXXIV (1970) 338ndash347
Kyle Albert S and F Albert Wang ldquoSpeculation Duopoly with Agreement toDisagree Can Overcondence Survive the Market Testrdquo Journal of FinanceLII (1997) 2073ndash2090
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoThe Structure andPerformance of the Money Management Industryrdquo in Martin Neil Baily andClifford Winston eds Brookings Papers on Economic Activity Microeconom-ics (Washington DC Brookings Institution 1992)
Lakonishok Josef Andrei Shleifer and Robert W Vishny ldquoContrarian Invest-ment Extrapolation and Riskrdquo Journal of Finance LIX (1994) 1541ndash1578
Lenney Ellen ldquoWomenrsquos Self-Condence in Achievement Settingsrdquo PsychologicalBulletin LXXXIV (1977) 1ndash13
Lewellen Wilber G Ronald C Lease Gary G Schlarbaum ldquoPatterns of Invest-ment Strategy and Behavior among Individual Investorsrdquo Journal of Busi-ness L (1977) 296ndash333
Lichtenstein Sarah and Baruch Fischhoff ldquoThe Effects of Gender and Instruc-tions on Calibrationrdquo Decision Research Report 81-5 Eugene OR DecisionResearch 1981
Lichtenstein Sarah Baruch Fischhoff and Lawrence Phillips ldquoCalibration ofProbabilities The State of the Art to 1980rdquo in Judgment Under UncertaintyHeuristics and Biases Daniel Kahneman Paul Slovic and Amos Tverskyeds (Cambridge and New York Cambridge University Press 1982)
Lundeberg Mary A Paul W Fox Judith Punccohar ldquoHighly Condent butWrong Gender Differences and Similarities in Condence Judgmentsrdquo Jour-nal of Educational Psychology LXXXVI (1994) 114ndash121
Lyon John D Brad M Barber and Chih-Ling Tsai ldquoImproved Methods for Testsof Long-Run Abnormal Stock Returnsrdquo Journal of Finance LIV (1999)165ndash201
Meehan Anita M and Willis F Overton ldquoGender Differences in Expectancies forSuccess and Performance on Piagetian Spatial Tasksrdquo Merrill-Palmer Quar-terly XXXII (1986) 427ndash441
Neale Margaret A and Max H Bazerman Cognition and Rationality in Nego-tiation (New York The Free Press 1990)
Odean Terrance ldquoVolume Volatility Price and Prot When All Traders Areabove Averagerdquo Journal of Finance LIII (1998) 1887ndash1934
mdashmdash ldquoDo Investors Trade Too Muchrdquo American Economic Review LXXXIX(1999) 1279ndash1298
Oskamp Stuart ldquoOvercondence in Case-Study Judgmentsrdquo Journal of Consult-ing Psychology XXIX (1965) 261ndash265
Papke Leslie E ldquoIndividual Financial Decisions in Retirement Savings PlansThe Role of Participant-Directionrdquo working paper Michigan State Univer-sity 1998
Prince Melvin ldquoWomen Men and Money Stylesrdquo Journal of Economic Psychol-ogy XIV (1993) 175ndash182
Russo J Edward and Paul J H Schoemaker ldquoManaging Overcondencerdquo SloanManagement Review XXXIII (1992) 7ndash17
Sorrentino Richard M Erin C Hewitt and Patricia A Raso-Knott ldquoRisk-Taking
291BOYS WILL BE BOYS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS
in Games of Chance and Skill Information and Affective Inuences on ChoiceBehaviorrdquo Journal of Personality and Social Psychology LXII (1992)522ndash533
Stael von Holstein Carl-Axel S ldquoProbabilistic Forecasting An Experiment Re-lated to the Stock Marketrdquo Organizational Behavior and Human Perfor-mance VIII (1972) 139ndash158
Sunden Annika E and Brian J Surette ldquoGender Differences in the Allocation ofAssets in Retirement Savings Plansrdquo working paper Board of Governors ofthe Federal Reserve System 1998
Varian Hal R ldquoDifferences of Opinion in Financial Marketsrdquo Financial RiskTheory Evidence and Implications Courtenay C Stone ed (Proceedings ofthe Eleventh Annual Economic Policy conference of the Federal Reserve Bankof St Louis Boston 1989)
Wagenaar Willem and Gideon B Keren ldquoDoes the Expert Know The Reliabilityof Predictions and Condence Ratings of Expertsrdquo in Intelligent DecisionSupport in Process Environments Erik Hollnagel Giuseppe Mancini andDavid D Woods eds (Berlin Springer 1986)
Yates J Frank Judgment and Decision Making (Englewood Cliffs NJ PrenticeHall 1990)
Zinkhan George M and Kiran W Karande ldquoCultural and Gender Differences inRisk-Taking Behavior among American and Spanish Decision Makersrdquo Jour-nal of Social Psychology CXXXI (1991) 741ndash742
292 QUARTERLY JOURNAL OF ECONOMICS