Behavioral finance.ppt

Post on 01-Dec-2014

3.422 views 2 download

Tags:

description

 

Transcript of Behavioral finance.ppt

An Introduction to Behavioral Finance

SIP Course on “Stock Market Anomalies and Asset Management”

Professors S.P. Kothari and Jon Lewellen

March 15, 2004

2

An Introduction to Behavioral Finance

Efficient markets hypothesis Large number of market participants Incentives to gather and process information about

securities and trade on the basis of their analysis until individual participant’s valuation is similar to the observed market price

Prices in such markets reflect information available to the participants, which means opportunities to earn above-normal rates of return on a consistent basis are limited

Prediction: Stock returns are (almost) impossible to predict Except that riskier securities on average earn higher rates of

returns compared to less risky firms

3

An Introduction to Behavioral Finance

Behavioral finance Widespread evidence of anomalies is inconsistent with the

efficient markets theory Bad models, data mining, and results by chance Alternatively, invalid theory

Anomalies as a pre-cursor to behavioral finance Challenge in developing a behavioral finance theory of

markets Evidence of both over- and under-reaction to events

Event-dependent over- and under-reaction, e.g., IPOs, dividend initiations, seasoned equity issues, earnings announcements, accounting accruals

Horizon dependent phenomenon: short-term overreaction, medium-term momentum, and long-run overreaction

4

An Introduction to Behavioral Finance

Behavioral finance theory rests on the following three assumptions/characteristics Investors exhibit information processing biases that

cause them to over- and under-react Individual investors’ errors/biases in processing

information must be correlated across investors so that they are not averaged out

Limited arbitrage: Existence of rational investors should not be sufficient to make markets efficient

5

Behavioral finance theories

Human information processing biases Information processing biases are generally

relative to the Bayes rule for updating our priors on the basis of new information

Two biases are central to behavioral finance theories

Representativeness bias (Kahneman and Tversky, 1982) Conservatism bias (Edwards, 1968). Other biases: Over confidence and biased self-attribution

6

Behavioral finance theories

Human information processing biases Representativeness bias causes people to over-

weight recent information and deemphasize base rates or priors

E.g., conclude too quickly that a yellow object found on the street is gold (i.e., ignore the low base rate of finding gold)

People over-infer the properties of the underlying distribution on the basis of sample information

For example, investors might extrapolate a firm’s recent high sales growth and thus overreact to news in sales growth

Representativeness bias underlies many recent behavioral finance models of market inefficiency

7

Behavioral finance theories

Human information processing biases Conservatism bias: Investors are slow to update their

beliefs, i.e., they underweight sample information which contributes to investor under-reaction to news

Conservatism bias implies investor underreaction to new information

Conservatism bias can generate short-term momentum in stock returns The post-earnings announcement drift, i.e., the tendency of

stock prices to drift in the direction of earnings news for three-to-twelve months following an earnings announcement also entails investor under-reaction

8

Behavioral finance theories

Human information processing biases Investor overconfidence

Overconfident investors place too much faith in their ability to process information

Investors overreact to their private information about the company’s prospects

Biased self-attribution Overreact to public information that confirms an

investor’s private information Underreact to public signals that disconfirm an investor’s

private information Contradictory evidence is viewed as due to chance Genrate underreaction to public signals

9

Behavioral finance theories

Human information processing biases Investor overconfidence and biased self-attribution

In the short run, overconfidence and biased self-attribution together result in a continuing overreaction that induces momentum.

Subsequent earnings outcomes eventually reveal the investor overconfidence, however, resulting in predictable price reversals over long horizons.

Since biased self-attribution causes investors to down play the importance of some publicly disseminated information, information releases like earnings announcements generate incomplete price adjustments.

10

Behavioral finance theories

In addition to exhibiting information-processing biases, the biases must be correlated across investors so that they are not averaged out

People share similar heuristics Focus on those that worked well in our evolutionary past Therefore, people are subject to similar biases Experimental psychology literature confirms systematic

biases among people

11

Behavioral finance theories

Limited arbitrage Efficient markets theory is predicated on the

assumption that market participants with incentives to gather, process, and trade on information will arbitrage away systematic mispricing of securities caused by investors’ information processing biases

Arbitrageurs will earn only a normal rate of return on their information-gathering activities

Market efficiency and arbitrage: EMH assumes arbitrage forces are constantly at work

Economic incentive to arbitrageurs exists only if there is mispricing, i.e., mispricing exists in equilibrium

12

Behavioral finance theories

Behavioral finance assumes arbitrage is limited. What would cause limited arbitrage? Economic incentive to arbitrageurs exists only if there

is mispricing Therefore, mispricing must exist in equilibrium Existence of rational investors must not be sufficient Notwithstanding arbitrageurs, inefficiency can persist

for long periods because arbitrage is costly Trading costs: Brokerage, B-A spreads, price impact/slippage Holding costs: Duration of the arbitrage and cost of short

selling Information costs: Information acquisition, analysis and

monitoring

13

Behavioral finance theories

Why can’t large firms end limited arbitrage? Arbitrage requires gathering of information about a firm’s

prospects, spotting of mispriced securities, and trading in the securities until the mispricing is eliminated

Analysts with the information typically do not have the capital needed for trading

Firms (principals) supply the capital, but they must also delegate decision making (i.e., trading) authority to those who possess the information (agents)

Agents cannot transfer their information to the principal, so decisions must be made by those who possess information

Agents are compensated on the basis of outcomes, but the principal sets limits on the amount of capital at the agent’s disposal (the book)

Limited capital means arbitrage can be limited

14

Behavioral finance theories

Like the efficient markets theory, behavioral finance makes predictions about pricing behavior that must be tested Need for additional careful work in this respect

Only then can we embrace behavioral finance as an adequate descriptor of the stock market behavior

Recent research in finance is in this spirit just as the anomalies literature documents inconsistencies with the efficient markets hypothesis

15

Stock Returns, Aggregate Earnings Surprises, and Behavioral Finance

 S.P. Kothari, Jonathan Lewellen,

Jerold B. Warner

 

 

SIP Course on “Stock Market Anomalies and Asset Management”

March 15, 2004

16

Objective of the study

We study the relation between market index returns and aggregate earnings surprises We focus on concurrent and lagged

surprises Do prices react slowly? Is there discount rate information in

aggregate earnings changes?

17

At the firm level, post-earnings announcement drift is well-known

The slow adjustment to public information is inconsistent with market efficiency

Slow adjustment is consistent with behavioral finance Barberis/Shleifer/Vishny (BSV, 1998) Daniel/Hirshleifer/Subrahmanyam (DHS, 1998) Hong/Stein (HS, 1999)

Aggregate return-earnings relation serves as an out-of-sample test of the behavioral hypothesis of investor underreaction

Literature concentrates on cross-sectional return predictability

We provide time-series evidence

Motivation

18

Main findings Aggregate relation does not mimic the firm-level

relation Market returns do not depend on past earnings surprises Inconsistent with underreaction (or overreaction)

Market returns are negatively (not positively) related to concurrent earnings news

#s seem economically significant Earnings and interest/ discount rate shocks are positively

correlated Good aggregate earnings news can be bad news

Decomposing earnings changes does not fully eliminate the negative correlation between earnings news and returns, a troubling result

19

Firm level drift and behavioral models

Drift could occur if investors systematically ignore the time-series properties of earnings.

Bernard/Thomas (1990) show that quarterly earnings changes have positive serial dependence (.34,.19,.06 at the first 3 lags)

If investors underestimate the dependence, prices will respond slowly and they will be surprised by predictable changes in earnings.

Consistent with this, the pattern of trading profits at subsequent earnings announcements matches the autocorrelation pattern.

20

Evidence

Time-series properties of earnings Stock returns and aggregate earnings

surprises Returns, earnings, and discount rates

21

Earnings series Compustat Quarterly database, 1970 – 2000 NYSE, Amex, and NASDAQ stocks with …

Earnings before ext. items, quarter t and t – 4 Price, quarter t – 4 Book value, quarter t – 4

 Plus … December fiscal year end Price > $1 Exclude top and bottom 0.5% based on dE/P

22

Sample

Quarterly returns (%), 1970 – 2000

Returns N VW EW

CRSP avg. 6,062 3.34 3.82 std. deviation -- 8.79 12.60 Sample avg. 2,423 3.26 3.42 std. deviation -- 8.38 11.40

23

E/P, 1970 – 2000

-0.04

-0.02

0.00

0.02

0.04

0.06

1970.1 1974.1 1978.1 1982.1 1986.1 1990.1 1994.1 1998.1

E/P-agg

E/P-ew

24

Firms w/ positive earnings, 1970 – 2000

0

1000

2000

3000

4000

5000

1970.1 1974.1 1978.1 1982.1 1986.1 1990.1 1994.1 1998.1

0.0

0.2

0.4

0.6

0.8

1.0

Number of firms (left scale)

Fraction E > 0 (right scale)

25

Quarterly earnings changes (%),

1970 – 2000 Aggregate VW EW

dE/P dE/B dE/E dE/P dE/P Full sample avg 0.15 0.25 8.26 0.10 0.30 stdev 0.39 0.59 18.58 0.36 0.55 Small stocks avg 0.42 0.39 -- 0.56 0.86 stdev 1.18 1.14 -- 0.90 1.13 Large stocks avg 0.14 0.25 7.90 0.10 0.08 stdev 0.37 0.58 17.60 0.35 0.38 Low B/M avg 0.17 0.54 12.11 0.16 0.60 stdev 0.23 0.73 16.69 0.22 0.69 High B/M avg 0.19 0.11 -- 0.09 0.22 stdev 1.13 0.81 -- 1.02 1.21

26

Aggregate earnings growth, 1970 – 2000

-0.60

-0.40

-0.20

0.00

0.20

0.40

0.60

1970.1 1974.1 1978.1 1982.1 1986.1 1990.1 1994.1 1998.1

dE/E-AGG

27

dE scaled by lagged price, 1970 – 2000

-.015

-.010

-.005

.000

.005

.010

.015

1970.1 1974.1 1978.1 1982.1 1986.1 1990.1 1994.1 1998.1

dE/P-VWdE/P-EW

28

Autocorrelations

Seasonally-differenced earnings (dE = Et – Et-4)

Estimation  dE/St = 0 + k dE/St-k + t 

dE/St = 0 + 1 dE/St-1 + 2 dE/St-2 + ….. +

5 dE/St-5 + t

Market: Time-series regressions Firms: Fama-MacBeth cross-sectional

regressions

29

Autocorrelations, dE/P, 1970 – 2000

Simple regressions Multiple regressions

Lag Slope T-stat Adj. R2 Slope T-stat Adj. R2

Firms 1 0.38 18.48 -- 0.40 18.39 -- 2 0.22 14.58 -- 0.14 11.20 3 0.08 5.67 -- 0.06 6.47 4 -0.28 -16.82 -- -0.42 -22.83 5 -0.11 -7.03 -- 0.16 12.93

EW 1 0.64 8.81 0.39 0.61 6.33 0.43 2 0.40 4.62 0.14 0.11 1.05 3 0.14 1.49 0.01 0.00 0.01 4 -0.15 -1.62 0.01 -0.30 -2.76 5 -0.21 -2.26 0.03 0.04 0.40 VW 1 0.73 11.54 0.52 0.73 7.75 0.57 2 0.52 6.65 0.26 0.22 1.93 3 0.23 2.55 0.04 -0.22 -1.92 4 -0.00 -0.03 -0.01 -0.18 -1.62 5 -0.12 -1.30 0.01 0.07 0.80

30

Implications

Basic message Pattern similar for firms and market Persistence stronger for market – good for tests

Specifics Transitory, idiosyncratic component in firm

earnings Aggregate earnings changes are permanent Earnings changes predictable but volatile ( = 18.6%)

AR1 similar to AR5

31

Returns and earnings surprises

Rt+k = + dE/Pt + et+k

 k = 0, …, 4  Changes and surprises  Market: Time-series regressions  Firms: Fama-MacBeth cross-sectional

regressions

32

Returns and earnings, 1970 – 2000

Earnings change Earnings surprise

k Slope T-stat Adj. R2 Slope T-stat Adj. R2

Firms 0 0.53 26.94 -- 1 0.58 28.70 -- 2 0.20 10.66 -- 3 0.09 5.24 -- 4 0.00 0.03 --

EW 0 -1.30 -0.90 0.00 1.54 0.85 0.04 1 -3.75 -2.60 0.05 -3.70 -2.04 0.05 2 -2.81 -1.97 0.02 -3.03 -1.65 0.01 3 -1.36 -0.95 0.00 1.15 0.63 0.03 4 -3.14 -2.23 0.03 -4.48 -2.43 0.03 VW 0 -4.98 -2.31 0.03 -2.59 -0.83 0.04 1 -5.23 -2.41 0.04 -10.10 -3.34 0.07 2 -0.80 -0.37 -0.01 0.51 0.16 -0.01 3 -1.34 -0.63 -0.01 -1.41 -0.45 -0.01 4 -0.90 -0.42 -0.01 -3.05 -0.97 -0.01

33

Contemporaneous relation

Explanatory power: 4 – 8%  Fitted values: dE/P-vw

Std. dev. of earnings surprises = 0.25% Slope = –10.10 Two std. deviation shock –5% drop in prices

 Historical Earnings change in top 25%: return 1% (s.e. =

1.7%) Earnings change in bottom 25%: return 7% (s.e. =

1.6%)

34

Contemporaneous relation

Early overreaction No theory Not in firm returns 

Movements in discount rates 

Rt = d,t – r,t

Cash flow news vs. expected-return news

35

Returns and past earnings Zero to negative No evidence of under-reaction Inconsistent with behavioral theories Results are robust

Alternative definitions of earnings Subperiods Annual returns and earnings Subsets of stocks (size, B/M terciles)

36

Summary observations

Large portfolio Earnings more persistent Initial market reaction more negative Puzzling from a cashflow-news perspective

Small portfolio Reversal at lag 4 Negatively related to CRSP, but not own

returns

37

Earnings and discount rates

Rt = d,t – r,td,t = cashflow newsr,t = expected-return news = discount-rate news

Returns and earningscov(dEt, Rt) = cov(dEt, d,t) – cov(dEt, r,t)cov(dEt, r,t)?

inflation and interest rates (+) consumption smoothing (–) changes in aggregate risk aversion (–)

38

Earnings and the macroeconomy, 1970 – 2000: Correlations

Nominal dE Real dE

EW VW EW VW

TBILL 0.35 0.60 0.27 0.50 TERM -0.35 -0.52 -0.33 -0.52 DEF -0.59 -0.37 -0.66 -0.49

SENT 0.37 0.13 0.39 0.20

GDP 0.40 0.54 0.61 0.67 IPROD 0.67 0.65 0.72 0.74 CONS 0.29 0.42 0.53 0.52

dE = seasonally-differenced earnings Macro = annual changes or growth rates, ending in qtr t

39

Earnings and the macroeconomy, 1970 – 2000

dEt = + TBILLt + TERMt + DEFt + dEt-1 + Nominal dE Real dE

EW VW EW VW

TBILL 0.04 0.04 0.02 0.03 1.39 2.72 0.73 1.78

TERM 0.00 -0.01 -0.01 -0.02 0.09 -0.29 -0.23 -0.69

DEF -0.55 -0.22 -0.64 -0.26 -4.95 -3.96 -5.70 -4.79

dEt-1 0.39 0.53 0.35 0.53 4.62 7.53 4.29 7.71

Adj. R2 0.49 0.62 0.53 0.62

Adj. R2 w/o AR1 0.41 0.44 0.46 0.43

40

Controlling for discount rates

Two-stage approach

dEt = + TBILLt + TERMt +

DEFt + dEt-1 + Rt+k = + Fitted(dEt) + Residual(dEt) + et+k

Timing?

Rt Rt+1 Rt+2 Rt+3 Rt+4

dEt

41

Returns and earnings, 1970 – 2000

Rt+k = + Fitted(dEt) + Residual(dEt) + et+k, Fitted dE Residual dE

k Slope T-stat Slope T-stat Adj. R2

EW 0 -6.86 -3.44 3.57 1.89 0.10 1 -5.01 -2.51 -3.02 -1.55 0.05 2 -2.93 -1.45 -2.44 -1.23 0.01 3 -4.20 -2.09 1.47 0.75 0.02 4 -1.55 -0.76 -4.53 -2.28 0.03 VW 0 -9.08 -3.27 0.76 0.23 0.07 1 -2.58 -0.95 -9.27 -2.84 0.05 2 -2.84 -1.02 2.30 0.69 0.00 3 -1.09 -0.39 -1.65 -0.49 -0.01 4 0.29 0.10 -2.53 -0.75 -0.01

42

Annual dE/P, 1970 – 2000 Rt+k = + Fitted(dEt) + Residual(dEt) + et+k, Fitted dE Residual dE

k Slope T-stat Slope T-stat Adj. R2

EW 0 -4.49 -2.03 -2.30 -1.15 0.11 1 -0.64 -0.26 1.29 0.58 -0.06 2 2.19 0.88 0.71 0.32 -0.04 3 1.11 0.45 -0.27 -0.13 -0.07 VW 0 -5.86 -2.04 -3.97 -1.23 0.11 1 -1.19 -0.40 7.74 2.29 0.11 2 2.95 0.91 -1.75 -0.48 -0.04 3 1.41 0.44 0.71 0.20 -0.07

43

How big are the effects?

Over the last 30 years, CRSP VWT portfolio Increased 6.5% in value in the quarters with

negative earnings growth Increased 1.9% in value in quarters with

positive earnings growth

44

Conclusions

Market’s reaction to earnings surprises much different at the aggregate level Negative reaction to good earnings news Past earnings contain little (inconsistent) information

about future returns Investment strategy: Long in quarters when aggregate

earnings changes are negative Open questions

Do earnings proxy for discount rates? Is there a coherent behavioral story for the patterns?

45

Richardson and Sloan (2003): External

Financing and Future Stock Returns Prior evidence: Market is sluggish in rationally

incorporating information in managers’ market timing motivation for external financing

Market timing: Raise funds when the firm is overvalued and repurchase shares when the firm is undervalued.

Slow assimilation of the information can be because of investors’ information processing biases

Sluggish reaction means opportunities for abnormal returns

How large are the returns to a trading strategy? What is the source of the abnormal returns? Is it related to the

use of proceeds from external financing? Richardson and Sloan: Examine returns to a trading rule

based on net external financing (not individual decisions like share repurchasing)

46

Returns following external financing

Prior evidence Low returns following equity offerings, debt offerings,

and bank borrowings High returns following share repurchases Managers seem to time external financing

transactions to exploit mispricing Market’s immediate reaction to the financing decisions

is incomplete (underreaction to public announcements of voluntary decisions)

Market gradually reacts over the following one-to-three years – inconsistent with market efficiency and consistent with some of the information-processing biases

47

Returns following external financing

Richardson and Sloan show that Net external financing generates a 12-month

abnormal return of about 16% (Table 5) The return is on long-minus-short position that has

a zero initial investment Long position is in firms that raise the least external

financing (i.e., repurchase shares or retire debt) Short position is in firms that raise the most

external financing – issue equity or debt or borrow from a bank

48

Returns following external financing

Richardson and Sloan show that Use of the proceeds from external financing

matters (Table 6) Investment in operating assets generates highest

return on the zero-investment portfolio Suggests managers over-invest in assets Market fails to fully assimilate information in accruals

What are accruals? Earnings (X) = CF + Accruals (A) When you sell on credit, earnings increase, cash flow

does not, but accruals in the form of accounts receivables increase

Investment in operating assets is a form of accrual

49

Returns following external financing

Acrobat Document

50

Returns following external financing

External financing decisions as well as exceptional corporate performance (high sales growth or extreme decline) are all associated with large accruals A large increase in sales translates into a large

increase in receivables, so an accrual increase is associated with increased sales

Accruals also present opportunities to the management to manipulate them and/or create them fictitiously A fictitious dollar of sales and receivables accruals

contributes dollar for dollar to earnings before taxes and also enhances profit margin (because the cost of goods sold is not increased with fictitious sales)

51

Returns following external financing

Since extreme performance or financing activities or fictitious sales are typically not sustainable, accruals revert

If investors suffer from information processing biases, do they recognize the time-series properties of accruals and its implications for future earnings?

In particular, does the market recognize that “The persistence of current earnings is decreasing in the magnitude of accruals and increasing in cash flows?”

Market overvalues accruals (i.e., fails to recognize that accruals-based earnings are not permanent)

Trading strategy implication: Long in low accrual stocks and short in high accrual stocks to generate above-normal performance.

Trading strategy based on external financing is based on accruals – raise capital means high accruals means go short

52

Conclusions

Investors exhibit many behavioral biases If the biases are similar across individuals and arbitrage

forces are limited, then the behavioral biases can cause prices to deviate systematically from economic fundamentals

Recent attempts to test the effects of behavioral biases in stock price data

Aggregate earnings data and stock returns Individual firms’ financial data and stock returns

Stock returns associated with external financing decisions Stock returns due to investors’ alleged inability to process

information in accounting accruals Next set of issues

How large is the mispricing? Can it be exploited? What are the barriers to implementation and what are the implications for asset management?