VC

GU

a

ARAA

JGGGG

KBMV

1

Dppwtifbss

ecpttc

1d

The Quarterly Review of Economics and Finance 50 (2010) 527–537

Contents lists available at ScienceDirect

The Quarterly Review of Economics and Finance

journa l homepage: www.e lsev ier .com/ locate /qre f

alue investing anomalies in the European stock market: Multiple Value,onsistent Earner, and Recognized Value

regor Elze ∗

niversity of Graz, Department of Statistic and Operations Research, Universtaetsstr. 15, 8010 Graz, Austria

r t i c l e i n f o

rticle history:eceived 11 March 2010ccepted 21 June 2010vailable online 6 July 2010

EL classification:11

a b s t r a c t

Empirical academic studies have consistently found that value stocks outperform glamour stocks andthe market as a whole. This article extends prevailing research on existing value anomalies. It evaluatessimple value strategies for the European stock market (compared to many other studies that test marketdata on a country-by-country basis) as well as sophisticated multi-dimensional value strategies that alsoinclude capital return variables (Consistent Earner Strategy) and momentum factors (Recognized ValueStrategy), the latter reconciling intermediate horizon momentum and long-term reversals of behavioral

121419

eywords:ehavioural Finance

finance theories. It can be shown that these “enhanced” value strategies can produce superior returnscompared to returns of the whole market or “simple” value strategies without capturing higher risksapplying traditional risk measures.

© 2010 The Board of Trustees of the University of Illinois. Published by Elsevier B.V. All rights reserved.

(cbsfte

kSt1setv

arket Anomaliesalue Investing

. Introduction

In their 1934 book, Security Analysis1, Benjamin Graham andavid Dodd argued that out-of-favor stocks are sometimes under-riced in the marketplace, and that investors cognizant of thishenomenon could capture strong returns. This philosophy is nowidely known as value investing. Although value investing has

aken many forms since its inception, it generally involves buy-ng shares which appear underpriced based on some form(s) ofundamental analysis. Value shares typically feature low price-to-ook, price-to-earnings, or price-to-cash flow ratios, while glamourtocks generally are characterized by valuation metrics at the oppo-ite end of the spectrum.

As early as 1977, academic studies have used share price andarning per share data to classify stocks into the value or glamourategories and compare historical performance. Stocks with low

rice-to-earnings multiples (often called “value” stocks) appearo provide higher rates of return than stocks with high price-o-earnings ratios as first shown by Nicholson (1960) and lateronfirmed by Ball (1978), Basu (1977, 1983), and Fama and MacBeth

∗ Tel.: +49 17663198991.E-mail address: r elze@yahoo.com.

1 Graham and Dodd (2005).

m

Est

062-9769/$ – see front matter © 2010 The Board of Trustees of the University of Illinoisoi:10.1016/j.qref.2010.06.005

1973).2 De Bondt and Thaler (1985) obtain a similar result for theirontrarian strategy based on buying stocks with low past returnsecause of the behavioral hypothesis of investor overreaction. Atock’s price-to-book value ratio has also been found to be a use-ul predictor of future returns. Fama and French (1992) concludedhat size and price-to-book value together provide considerablexplanatory power for future returns in U.S. markets.

These results raised questions about the efficiency of the mar-et if one accepts the capital asset pricing model, as Lakonishok,chleifer and Vishny pointed out. In 1994, they published “Con-rarian Investment, Extrapolation, and Risk3”. Using data from968 to 1994, they grouped U.S. stocks into value and glamouregments based on price-to-book, price-to-cash flow, and price-to-arnings ratios, as well as sales growth. The researchers concludedhat, for a broad range of definitions of “value” and “glamour”,alue stocks consistently outperformed glamour stocks by wideargins.

In their 1998 study, “Value versus Growth: The International

vidence”, Fama and French tackled the question of whether valuetocks tended to outperform glamour stocks in markets outsidehe U.S. The researchers found that, from 1975 to 1995, value

2 Papers are cited in detail at the end of this article.3 An update was published in 2004: Chan, L., & Lakonishok, J. (2004).

. Published by Elsevier B.V. All rights reserved.

5 onomi

setVm

distas“gitasdot“

assEe

Svakahps

sVmv(dvaoictms

oehsr5

J

2

a2Jmiirb

syatEantohpfzand compute returns using a buy-and-hold strategy for years t + 1,t + 2 and t + 3 relative to the time of formation. If a stock is delistedfrom the stock exchange during a year, we continue with the sameportfolio using the return of that stock at the time it was last traded

6 Results for starting dates on January 1, April 1 and October 1 were also testedand results are comparable to conclusions drawn from yearly starting dates on July1.

7 If the 30th is not a weekday, then the last trading day of the month is used. Yearsin tables and graphs refer to a time period from July 1 that year until June 30 of thesubsequent year. Formation and reformation occur based on publicly available priceand accounting data from the previous year (t−1). Results for current year estimatesas accessible at formation and reformation dates were comparable. Reformation atthe beginning of the second quarter was chosen in order to ensure that fundamentalcompany information for the entire previous year published in annual reports, SECfilings or by the financial media was available to investors and incorporated intovaluation ratios. 1994 was chosen as the first formation year because the EuroStoxxwas created in 1999 and index constituents were recalculated back to this time.

8 Look-ahead and survivorship bias are common types of sample selection biases.The first is created by the use of information or data in a study or simulation thatwould not have been known or available during the period being analyzed. Thiswill usually lead to inaccurate results in the study or simulation. To avoid this biaswe calculated ratios based on data available at the time of portfolio formation andreformation, not from revisions published thereafter. The second bias occurs, forexample, when backtesting an investment strategy on a large group of stocks. Then itmay be convenient to look for securities that have data for the entire sample period.However, eliminating a stock that stopped trading, or shortly left the market, wouldinput a bias in data samples. To avoid this problem we used historical constituentlists for the EuroStoxx when we constructed our quantile portfolios.

9 Banz and Breen (1986), Kothari, Shanken, and Sloan (1992). La Porta (1993) alsopoints out that the selection bias is less serious among larger firms.

10 Rosenberg, Reid and Lanstein (1984) and Lee and Swaminathan (2000) pub-lished the first cited study that evaluated the performance of the book-to- marketstrategy.

11

28 G. Elze / The Quarterly Review of Ec

tocks outperformed glamour stocks in 12 of 13 major nationalquity markets. In their opinion, this laid to rest the possibilityhat the value outperformance seen by Lakonishok, Schleifer andishny was simply a sample-specific happenstance within the U.S.arket.4

While there is some agreement that value strategies have pro-uced superior returns, the interpretation of why they have done so

s more controversial. “Behavioralists” believe that investors con-istently tend to overpay for “growth” stocks that subsequently failo live up to expectations (for example, Kahneman & Riepe, 1998nd Gilovich, Griffin, and Kahneman (2002)). In their view valuetrategies produce higher returns because they are contrarian tonaive” strategies followed by other investors. These naive strate-ies might range from extrapolating past earnings growth too farnto the future, to assuming a trend in stock prices, to overreactingo good or bad news, or to simply equating a good investment withwell-run company irrespective of price. Regardless of the reason,

ome investors tend to get overly excited about stocks that haveone very well in the past and buy them up, so that these “glam-ur” stocks become overpriced. Similarly, they overreact to stockshat have done very poorly, oversell them, and these out-of-favorvalue” stocks become underpriced.

This article is based on prevailing research on existing valuenomalies.5 It evaluates simple value strategies for the Europeantock market as well as sophisticated multi-dimensional valuetrategies that also include capital return variables (Consistentarner Strategy) and momentum factors (Recognized Value Strat-gy).

In Section 2 of the article our methodology is briefly discussed.ection 3 (a) examines a variety of simple classification schemes foralue and glamour stocks based on dividend yield, price-to-booknd price-to-earnings ratio. Contrary to many studies that test mar-et data on a country-by-country basis, all strategies are appliednd modulated for the European stock market. The EuroStoxx indexas been selected as the market proxy. It can be shown that sim-le value strategies have produced superior returns motivating ourubsequent use of variable combinations.

Section 3.1 (b) evaluates strategies based on multi-dimensionalelection criteria. First, simple value measures are combined (Multialue Strategy). In a second step we combine more sophisticatedulti-dimensional value strategies that also include capital return

ariables (Consistent Earner Strategy) and momentum factorsRecognized Value Strategy). It can be shown that while multi-imensional value strategies based on a combination of simplealue variables do not further improve investment performancend statistical significance, strategies based on combinationsf value and capital return variables (e.g. Return on Equity)mprove the statistical significance of results (while generatingompatible investment returns). Strategies based on combina-ions of value and momentum variables improve both invest-

ent performance and significance compared to simple valuetrategies.

Finally in Section 4 the question of whether strategies basedn our investment selection criteria are fundamentally riskier isvaluated. Evidence is provided that, in general, value strategiesave outperformed glamour strategies quite consistently without

upport for the hypothesis that value strategies are fundamentallyiskier than glamour strategies. Conclusions are drawn in Section.

4 Simliar results had been shown by Lakonishok, Hamao, and Chan (1991) forapan.

5 We widely follow Lakonishok et al. (1994) in the structure of our analysis.

py

ed

v

cs and Finance 50 (2010) 527–537

. Methodology

The sample period covered in this study starts on July 1, 19946

nd ends June 30, 2009.7 For portfolio strategies that are tested over-year (3-year) performance horizons the last reformation date isuly 1, 2007 (July 1, 2006). As market proxy for the European stock

arket the EuroStoxx index has been selected. Results are also ver-fied for the EuroStoxx50 in order to verify that results still holdf only large capitalization equities are examined. Results for stocketurns of indices containing only large firms are less contaminatedy significant look-ahead or survivorship bias.8,9

Based on the index we form our model portfolios using as a firsttep one-dimensional (single) accounting ratios, such as dividendield (DY), price-to-book10 (P/B) and price-to-earnings11 (P/E). Inddition, corporate return (RoE) and momentum (Levy2712, Rela-ive Strength – 3 months) ratios13 are computed for a Consistentarner Strategy (trying to mimic investment styles of successfulnd well-known value investors who focus on “outstanding compa-ies at a sensible price”)14 and a Recognized Value Strategy (tryingo further improve performance by timing reversals better basedn the stock momentum life cycle hypothesis).15 Then ratios andistorical performance data are used to sort individual stocks intoortfolios.16 Based on the investment strategy chosen, deciles areormed for which performance is measured for 1–3-year time hori-ons. Within each of our portfolios, we equally weight all stocks

Basu (1977) first looked at the relationship between common stocks and theirrice-to-earnings ratios. He found that a low P/E ratio portfolio earned 6% more perear than a high P/E portfolio in the 14-year sample.12 Levy, R. (1967).13 RoE = Return on Equity, Levy27 = a stock’s price divided by its price 27 weeksarlier, Relative Strength – 3 months = performance of a stock compared to the indexuring the last 3 months (MO3m).14 Greenblatt (2006).15 Oyefeso (2004), Lee, C., & Swaminathan, B. (2000). Price momentum & tradingalue. Journal of Finance.16 As Lakonishok et al. (1994) we consider only positive ratios.

G. Elze / The Quarterly Review of Economics and Finance 50 (2010) 527–537 529

Table 1Yearly average decile returns for portfolio strategies based on one-dimensional classifications by various measures of valuea.

Panel A: DY (t0)

Return overview: Yearly Value premium

Glamour Value

1 2 3 4 5 6 7 8 9 10 10−1

R1 5.07% 9.89% 8.01% 10.42% 9.07% 9.42% 11.16% 12.86% 8.28% 13.66% 8.60%R2 6.81% 11.19% 5.69% 11.52% 9.71% 12.56% 13.02% 11.49% 11.59% 12.46% 5.64%R3 7.20% 5.56% 7.19% 9.56% 8.14% 9.39% 10.41% 7.67% 15.75% 15.48% 8.28%ANN 6.36% 8.85% 6.96% 10.50% 8.97% 10.45% 11.52% 10.65% 11.83% 13.86% 7.50%CR3 20.30% 28.97% 22.36% 34.91% 29.41% 34.74% 38.71% 35.48% 39.85% 47.61% 27.31%AR 6.36% 8.88% 6.96% 10.50% 8.98% 10.46% 11.53% 10.67% 11.87% 13.87% 7.51%

Panel B: P/B (t0)

Return overview: Yearly Value premium

Glamour Value

1 2 3 4 5 6 7 8 9 10 10−1

R1 7.23% 7.32% 8.04% 9.88% 10.39% 9.45% 8.60% 9.22% 9.01% 12.64% 5.40%R2 7.64% 8.70% 7.70% 10.44% 9.22% 9.88% 11.40% 12.88% 9.87% 18.99% 11.35%R3 5.09% 7.71% 6.20% 8.45% 7.64% 11.18% 10.81% 10.39% 12.95% 15.75% 10.67%ANN 6.65% 7.91% 7.31% 9.59% 9.08% 10.17% 10.26% 10.82% 10.60% 15.76% 9.12%CR3 21.29% 25.65% 23.57% 31.60% 29.78% 33.72% 34.05% 36.09% 35.29% 55.14% 33.84%AR 6.65% 7.91% 7.31% 9.59% 9.08% 10.17% 10.27% 10.83% 10.61% 15.79% 9.14%

Panel C: P/E (t0)

Return overview: Yearly Value premium

Glamour Value

1 2 3 4 5 6 7 8 9 10 10−1

R1 1.11% 5.87% 9.46% 11.03% 10.00% 9.49% 12.74% 9.04% 11.12% 13.76% 12.66%R2 7.51% 7.20% 8.92% 11.16% 11.58% 13.38% 9.31% 9.42% 11.97% 13.87% 6.35%R3 6.60% 9.26% 5.78% 8.72% 7.69% 5.94% 11.22% 10.12% 12.95% 13.39% 6.79%ANN 5.04% 7.44% 8.04% 10.30% 9.75% 9.56% 11.08% 9.53% 12.01% 13.68% 8.64%CR3 15.88% 24.01% 26.11% 34.18% 32.18% 31.51% 37.06% 31.39% 40.54% 46.89% 31.01%AR 5.07% 7.45% 8.05% 10.30% 9.76% 9.60% 11.09% 9.53% 12.02% 13.68% 8.60%

Panels A–C present the average percentage decile returns for one-dimensional value strategies formed in ascending order based on P/E and P/B; descending order based ondividend yield (DY). The value portfolio refers to the decile portfolio containing stocks ranking lowest on P/E or P/B, or highest on dividend yield (DY). The glamour portfoliocontains stocks with precisely the opposite set of rankings. Portfolio reformation occurs yearly at the beginning of July during the period from 1994/95 to 2008/09. Theright-most column contains the value premium based on the performance difference between decile 10 and 1.Panel A: percentage decile returns as described above for a one-dimensional value strategy based on dividend yield (DY).Panel B: percentage decile returns as described above for a one-dimensional value strategy based on P/E.P strateg

d 3-yet

utfsaoa

t(ii

oppt

c

41DLolgi

anel C: percentage decile returns as described above for a one-dimensional valuea R1 = 1st year return, R2 = 2nd year return, R3 = 3rd year return, ANN = annualize

o annotation 15.

ntil the end of the observation period.17 There is no reformation inhe portfolios during the performance measurement period. Port-olio formation occurs18 at the beginning of July each year and eachtock gets the same weight. Stock, dividend and index informationre derived from Thomson-Reuters Datastream and Factset basedn closing prices. Performance is measured by stock price changesnd dividend payouts. Transaction costs are not included.

In a second step we combine accounting ratios following

he selection procedure used by Lakonishok, Shleifer and Vishny1994). For two-dimensional value strategies, stocks are classifiednto nine groups by independently sorting them in ascend-ng/descending order into three arrays ((3) bottom 30%, (2) middle

17 The reasons for stocks to be delisted from the stock exchange during a ref-rmation period were mostly acquisitions by other corporations. The last tradedrice therefore was taken to approximate real performance of that stock in a giveneriod. Return on cash proceeds during the remaining period was not included inhe performance calculation.18 Performance data for year 2 and year 3 (assuming no reformation) is also indi-ated.

vLpofsa

ap

y based on P/B.ar return, CR3 = compounded 3 year return, AR = average 3 year return, (t0) = refer

0%, and (1) top 30%) based on each of two variables. The sorts are2 pairs of variables: DY and P/E, DY and P/B, P/E and P/B, RoE andY, RoE and P/E, RoE and P/B, DY and Levy27, DY and MO3m, P/E andevy27, P/E and MO3m, P/B and Levy27, P/B and MO3m. Dependingn the two variables being used for classification, the value portfo-io either refers to the portfolio containing stocks ranked in the toproup (1) on both variables from among P/E, P/B (sorted in ascend-ng order), or else the portfolio in the top group on one of thoseariables or/and in the top group (1) on reversely sorted DY, RoE,evy 27, and MO3m. The glamour portfolio19 contains stocks withrecisely the opposite set of rankings. Portfolio formation and ref-rmation occur yearly at the beginning of July during the period

rom 1994/95 to 2008/09. For each quartile, performance is mea-ured using the same procedure as for one-dimensional (single)ccounting ratios.

19 In the case of momentum variables, it is usually inappropriate to talk of valuend glamour criteria but for simplicity we refer to the value quantile for the toparameter characteristics and to glamour for the bottom.

530 G. Elze / The Quarterly Review of Economics and Finance 50 (2010) 527–537

G centaf ation

3s

ooTrTstIarcd

m5picib

tac

baEa

r1

bwwv

(

rribvda

3s

fEgo3a

raph 1. Yearly and 3-year annualized decile returns. Graph 1 shows the yearly peror a one-dimensional value strategy based on dividend yield (DY). Portfolio reform

. (a) Performance evaluation: one-dimensional valuetrategies

Table 1, Panels A–C present the yearly average decile returns forne-dimensional value strategies formed in ascending order basedn P/E and P/B;20 descending order based on dividend yield (DY).21

he value portfolio refers to the decile portfolio containing stocksanking lowest on P/E and P/B, or highest on dividend yield (DY).he glamour portfolio contains stocks with precisely the oppositeet of rankings. Portfolio formation and reformation occur yearly athe beginning of July during the period from 1994/95 to 2008/09.n addition, year 2 and year 3 returns (assuming no reformationfter year 1), the annualized 3-year return, the compounded 3-yeareturn and the average 3-year return are indicated. The right-mostolumn contains the value premium based on the performanceifference between decile 10 and 1.

The yearly average (R1) return differences (value (decile 10)inus glamour (decile 1)) presented above fall in a range between

.40% and 12.66% depending on the value variable chosen. The out-erformance is statistically significant based on a 5% significance

nterval except for price-to-book (P/B).22,23 Consequently, the con-lusion can be drawn that the value anomaly discovered for mostndexes worldwide also holds for the European market as proxiedy the EuroStoxx index.24

20 Ratios in this article are based on actual data available at formation and reforma-ion. Ratios in graphs and tables are therefore (for better differentiation) assigned(t0). Results for ratios based on estimated data were also tested. Outcomes are

omparable.21 Though RoE (Return in Equity), Levy27 (stock price divided by its price 27 weeksefore) and Relative Strength – 3 month (MO3m) constitute no value variable theyre used in combination with value variables for our two-dimensional Consistentarner Strategy and Recognized Value Strategy strategies. For their returns as stand-lone variable, refer to the appendix section.22 Results of the t-statistic for the test of the hypothesis that the differences ineturns between value and glamour are equal to zero are presented in Table 3, page4.23 For simple return and momentum variables, return differences fall in a rangeetween 2.50% and 11.44%. Results are presented in the appendix section. Though itas not part of the research question formulated in this study it nevertheless seemsorthwhile to mention that return differences for single capital and momentum

ariables are not statistically significant (page 14).24 Comparable results are obtained for the EuroStoxx50 index. Lakonishok et al.1994) divide the U.S. market into the 50% largest/smallest corporations instead

aMatooo(fPta3ypa

aa

t

ge decile portfolio returns for years 1–3 and annualized for the entire 3-year periodoccurs yearly at the beginning of July during the period from 1994/95 to 2008/09.

Graph 1 shows the yearly average percentage decile portfolioeturns for years 1–3 and annualized for the entire 3-year periodepresentatively for a one-dimensional value strategy based on div-dend yield (DY).25 In addition to the results obtained earlier it cane seen that returns rise considerably moving from glamour (1) toalue (10) deciles. It also can be observed that most of the valueifference (steeper curve) is captured in year 1 though differencesre still positive in years 2 and 3.

.1. (b) Performance evaluation: two-dimensional valuetrategies

Table 2 presents the average yearly percentage quantile returnsor two-dimensional value strategies (including our Consistentarner and Recognized Value Strategies) each classified into nineroups of stocks by independently sorting in ascending/descendingrder into three arrays ((1) bottom 30%, (2) middle 40%, and (3) top0%) each of two variables. The sorts are 12 pairs of variables: DYnd P/E, DY and P/B, P/E and P/B, RoE and DY, RoE and P/E, RoEnd P/B, DY and Levy27, DY and MO3m, P/E and Levy27, P/E andO3m, P/B and Levy27, P/B and MO3m. Depending on the two vari-

bles being used for classification, the value portfolio either referso the portfolio containing stocks ranked in the bottom group (1)n both variables from among P/E, P/B (sorted in ascending order),r else the portfolio in the bottom group on one of those variablesr/and in the bottom group (1) on reversely sorted dividend yieldDY), capital return (RoE), Levy27, and MO3m. The glamour port-olio contains stocks with precisely the opposite set of rankings.ortfolios reformation occurs yearly at the beginning of July duringhe period from 1994/95 to 2008/09. In addition, returns in year 2nd year 3 (assuming no reformation after year 1), the annualized

-year return, the compounded 3-year return and the average 3-ear return are indicated. The right-most column contains the valueremium based on the performance difference between groups 1/1nd 3/3.

nd likewise conclude (after eliminating the size-effect) that results still hold whennalyzing only large corporations.25 For decile performance graphs of different one and two-dimensional ratios refero the appendix section. Outcomes are comparable.

G. Elze / The Quarterly Review of Economics and Finance 50 (2010) 527–537 531

Table 2Yearly average quantile returns for portfolio strategies based on two-dimensional classifications by various value measures.

Return overview: Yearly Value premium

Glamour Value

3/3 3/2 2/3 3/1 2/2 1/3 2/1 1/2 1/1 Total 1/1−3/3

Multi valueDY&P/E 6.50% 9.79% 8.33% 13.20% 10.81% 3.22% 10.37% 12.97% 12.38% 9.88% 5.89%DY&P/B 8.91% 5.89% 10.74% 8.00% 9.84% 10.64% 10.62% 14.08% 11.85% 9.89% 2.94%P/E&P/B 6.77% 5.21% 10.09% 5.64% 10.60% 12.39% 12.21% 11.18% 10.98% 9.52% 4.20%

ConsistentDY&RoE 4.96% 8.68% 9.64% 10.62% 9.96% 11.10% 11.21% 12.12% 11.39% 9.89% 6.43%P/E&RoE 2.95% 12.18% 7.75% 12.22% 10.85% 10.60% 10.72% 9.25% 11.46% 9.50% 8.51%P/B&RoE −3.29% 5.44% 6.57% 9.28% 10.37% 9.73% 10.83% 11.59% 11.50% 9.19% 14.78%

Recognized valueDY&Levy27 2.42% 6.69% 5.85% 13.12% 8.91% 8.19% 14.79% 11.72% 18.86% 9.79% 16.44%DY&M03m 2.51% 8.03% 7.50% 12.24% 9.48% 9.57% 12.83% 10.67% 18.54% 9.78% 16.03%P/E&Levy27 −0.50% 8.25% 5.92% 7.16% 8.96% 10.39% 11.39% 15.59% 16.54% 9.33% 17.05%P/E&M0m3 0.84% 8.25% 4.32% 8.27% 10.45% 11.78% 10.66% 14.57% 14.97% 9.35% 14.14%P/B&Levy27 1.81% 5.29% 6.40% 8.96% 8.86% 13.43% 10.88% 13.99% 12.88% 9.22% 11.08%P/B&M03m 1.50% 7.32% 6.75% 9.26% 7.79% 14.72% 11.23% 13.63% 13.71% 9.20% 12.21%

Presents the yearly average percentage quantile returns for two-dimensional value strategies classified into nine groups of stocks by independently sorting in ascend-ing/descending order into three arrays ((1) bottom 30%, (2) middle 40%, and (1) top 30%) based on each of two variables. The sorts are 12 pairs of variables: DY and P/E, DYand P/B, P/E and P/B, RoE and DY, RoE and P/E, RoE and P/B, DY and Levy27, DY and MO3m, P/E and Levy27, P/E and MO3m, P/B and Levy27, P/B and MO3m. Depending on thetwo variables being used for classification, the value portfolio either refers to the portfolio containing stocks ranked in the bottom group (1) on both variables from amongP ne of( s withb n cona

a(bccis

vmfTkdogra

vmvtsaah

tPtc

4

vtedfikmabfiaaaaSt(tnas

/E, P/B (sorted in ascending order), or else the portfolio in the bottom group on oDY), capital return (RoE), Levy27, and MO3m. The glamour portfolio contains stockeginning of July during the period from 1994/95 to 2008/09. The right-most columnd 3/3.

For two-dimensional value strategies based on two value vari-bles (Multi Value Strategy) the yearly average return differencesquantile 1/1 minus quantile 3/3) presented above fall in a rangeetween 2.94% and 5.89% depending on the variable combinationhosen. These strategies do not improve investment performanceompared to simple value strategies. Furthermore statistical signif-cance is reduced26 (in fact, investment results get worse and aretatistically insignificant).27

Strategies based on combinations of value and capital returnariables28 (Consistent Earner Strategy) seem to result in invest-ent returns comparable to single variable value strategies ranging

rom 6.43% to 14.78%. Statistical significance however improves.29

he Consistent Earner Strategy mimics investment styles of well-now investors like Warren Buffett or Joel Greenblatt who furthereveloped the value investing concept by focusing on “finding anutstanding company at a sensible price” or buying “cheap andood companies with competitive advantages indicated by a higheturn on capital” rather than generic companies at a bargain prices originally promoted by Graham and Dodd.

Strategies based on combinations of value and momentumariables (Recognized Value Strategy) do improve both invest-ent performance and significance compared to single variable

alue strategies. Investment returns are in a range from 11.08%o 17.05% annually. The Recognized Value Strategy is based on thetock momentum life cycle hypothesis30 stating that stocks move

lternately through periods of relative “glamour” and “neglect”ttempting to reconcile intermediate horizon momentum and longorizon-reversals of behavioral finance theories.

26 Refer to Table 3, page 14 for a statistical result summary.27 An explanation may be that selection based on mixed value criteria does reducehe variable’s indication for value because of accounting features. For instance low/E’s are a value indicator for pharmaceutical companies, low P/B’s are not, sincehese companies have high amounts of intangible assets increasing this ratio artifi-ially.28 e.g. Return on Equity.29 Refer to Table 3, page 14 for a statistical result summary.30 Lee & Swaminathan (2000). Price momentum & trading value. Journal of Finance.

vvspef

sm

those variables or/and in the bottom group (1) on reversely sorted dividend yieldprecisely the opposite set of rankings. Portfolios reformation occurs yearly at the

tains the value premium based on the performance difference between group 1/1

. Risk evaluation

Two competing theories have been proposed to explain whyarious investment strategies have produced higher returns inhe past. The capital asset pricing model (CAPM) relating risk andxpected return is grounded on the theory that rational investorsemand higher returns for higher risks. Serving as a common modelor the pricing of risky securities in the financial industry, it takesnto account an asset’s sensitivity to non-diversifiable risk (alsonown as systematic risk or market risk), often referred to by theeasure beta (�), as well as the expected return of the market

nd the expected return of a theoretical risk-free asset. In contrast,ehavioral finance theories recognize a psychological element innancial decision-making, thus challenging traditional models thatssume investors will always weigh risk/return factors rationallynd act without bias.31 For example, the human tendency to avoiddmitting error, called “fear of regret” by psychologists, can causen investor to hold a losing stock too long or sell a winner too soon.imilarly, investment choices are influenced positively or nega-ively by attitudes toward wealth, by investor attention, mimicryherding instinct), etc. The premise of behavioral finance is thataking psychological factors into account can enhance the effective-ess of investment strategies and explain the success of contrariannd value strategies (described as anomalies or inefficiencies intandard financial literature).

In this section it will be examined whether superior returns byalue portfolios constructed based on one- and two-dimensionalariables (Multiple Value, Consistent Earner, Recognized Valuetrategies) formed using the EuroStoxx benchmark as the sam-le index (EuroStoxx Value Anomaly) can be explained by a higherxposure to systematic risk. In the subsequent analysis we widely

ollow the methodology used by Lakonishok et al. (1994).

Value stocks would be fundamentally riskier than glamourtocks if they underperform glamour stocks in periods of severearket declines in which the marginal utility of wealth is high,

31 Kahneman and Tversky (1979).

532 G. Elze / The Quarterly Review of Economics and Finance 50 (2010) 527–537

Graph 2. Yearly return differences (value minus glamour): EuroStoxx. Graph 2 shows the yearly percentage portfolio return differences (value quantile (1/1) minus glamourq ion schi y a “+b rs yea

mwoerv

ma2chs

c(–

estssbvS

b(

u

e

cop

tm

o&t3t

b(mii

bDP1tav

mm

uantile (3/3)) based on the EuroStoxx index for a two-dimensional value classificatndex (EuroStoxx) declined are marked by “N”, years in which it rose are indicated beginning of July and the end of June the following year. Portfolio reformation occu

aking value stocks unattractive to risk-averse investors. To begin,e look at the consistency of performance of each selected strategy

ver time and then we ask how each performs in different marketnvironments. In addition, we assess some traditional measures ofisk, such as beta and the standard deviation of returns, to comparealue and glamour strategies.32

Graph 2 shows the year-by-year performance differences (valueinus glamour) of a two-dimensional value strategy based on DY

nd P/E for the EuroStoxx index over the period from 1995/95 to008/09 (July 1, 1994 and June 30, 2009).33 The strategy has quiteonsistently generated a positive value difference. Using a 1-yearorizon, value outperformed glamour in 11 out of 15 years using atrategy based on DY& P/E.

While the number of years (N = 15) is too small to draw signifi-ant conclusions a more detailed examination of quarterly resultsN = 60) still reveals a comparable – while slightly less spectacularpicture.

Table 3 presents the quarter-by-quarter performance differ-nces (value minus glamour) of one- and two-dimensional valuetrategies for the EuroStoxx index over the period from July 1, 1994o June 30, 2009. The quarterly average return differences for eachtrategy are reported at the bottom of each column along with t-tatistic for the test of the hypothesis that the difference in returnsetween value and glamour is equal to zero. The corresponding p-alue, the standard deviation (quarterly and annualized) and the

harp Ratio34 are also presented at the bottom lines.

Over a 1-year time horizon, one-dimensional value strategiesased on DY, P/B, P/E generated negative performance differencesvalue minus glamour) in only 21, 26, and 18 instances respec-

32 Common portfolio measures are also indicated, however only to provide furtherseful information.33 Refer to the appendix section for a complete graphical overview of value differ-nce evolutions based on all variable combinations.34 The Sharp Ratio (SR) is a risk-adjusted measure developed by William F. Sharpe,alculated using standard deviation and excess return to determine reward per unitf risk. The higher the Sharpe ratio, the better a portfolio’s historical risk-adjustederformance.

it

dbgTd

ao

eme based on DY (t0) and P/E (t0). In addition, years in which the European market”. Yearly data in the 1994/95–2008/09 period refers to the 12 months between therly.

ively (35.0%, 43.3%, 30.0%). Results for single capital return andomentum variables are comparable.35

Multi Value Strategy: Two-dimensional value strategies basedn simple value variable combinations (e.g. DY & P/E, DY & P/B, P/EP/B) generated negative performance differences over a 1-year

ime horizon in 22, 22, and 23 instances respectively (36.7%, 36.7%,8.3%). Performance differences for these value variable combina-ions however are not statistically significant.

Consistent Earner Strategy: Two-dimensional value strategiesased on simple value and capital return variable combinationse.g. RoE & DY, RoE & P/E, RoE & P/B) generated negative perfor-

ance differences over a 1-year time horizon in 23, 20, and 20nstances respectively (38.3%, 33.3%, 33.3%). Statistical significancencreases compared to single value variables.

Recognized Value Strategy: Two-dimensional value strategiesased on simple value and momentum variable combinations (e.g.Y & Levy27, DY & MO3m, P/E & Levy27, P/E & MO3m, P/B & Levy27,/B & MO3m) generated negative performance differences over a-year time horizon in 21, 19, 19, 25, 18, and 18 instances respec-ively (35.0%, 31.7%, 31.7%, 41.7%, 30.0%, 30.0%). Return differencesnd statistical significance both increase compared to single valueariables.

In most cases presented above, the magnitude of underperfor-ance of value versus glamour was mostly small relative to theean outperformance, and return differences were negative dur-

ng market declines in even fewer instances. Thus we can concludehat downside risk is relatively low.

We proceed and examine the performance of one- and two-imensional strategies in extreme down markets on a monthly

asis thereby comparing the performance difference (value minuslamour) in the worst months for the stock market as a whole.able 4, Panels A–D, lists the performance of one- and two-imensional strategies based on DY and on combinations of DY

35 Note that value differences for single value variables are statistically significantt the 5.0% confidence interval (exception: P/B). Results for single variables basedn capital return and momentum ratios are generally not statistically significant.

G. Elze / The Quarterly Review of Economics and Finance 50 (2010) 527–537 533

Table 3Quarterly average portfolio return differences (value minus glamour) based on various value portfolio strategies for one- and two-dimensional classification schemes(EuroStoxx).

Value premiums 1/1−3/3 DY (tO) PB (tO) PE (tO) RoE (tO) Levy27 M03m DY&PE DY&PB PE&PB

Q3 94 + −4.22% −7.40% 5.40% 6.25% 1.50% 1.53% −1.75% −7.33% −5.94%Q4 94 + 3.11% 0.09% 4.70% 5.61% −2.88% −2.39% −0.02% 1.82% 2.49%Q1 95 N −8.86% −12.32% −3.63% 0.67% 1.07% 1.15% −8.38% −5.85% −6.07%Q2 95 + 10.63% 2.42% 9.94% 5.98% 0.08% 1.15% 8.77% 2.26% 0.75%Q3 95 + −6.15% −10.58% 4.00% 8.90% 5.19% 5.23% −2.52% −5.60% 0.11%Q4 95 + 11.56% −5.01% 9.23% 9.43% 3.45% 6.88% 7.53% 8.33% 3.78%Q1 96 + −1.65% −16.11% 3.25% 12.70% 12.16% −2.53% −3.03% −12.44% −0.21%Q2 96 + 3.05% −0.58% 4.94% 5.64% 6.58% −2.02% 0.50% −4.85% 1.94%Q3 96 + 0.32% −3.58% 0.05% 8.38% 0.35% 2.20% 0.22% −1.76% −4.39%Q4 96 + 0.25% −6.65% −2.72% −0.96% 8.62% 1.45% 0.58% −1.30% −4.01%Q1 97 + −2.88% 35.86% 8.70% −29.91% −6.26% −5.93% 1.90% 2.62% 6.69%Q2 97 + 1.70% 3.30% 10.96% 11.35% 2.77% 6.43% 2.08% 1.08% 1.16%Q3 97 + 2.53% 16.38% 1.72% −8.90% 10.95% 3.20% 4.04% 10.44% 11.45%Q4 97 N 0.16% −0.56% −1.53% 3.68% −4.15% −2.78% 2.21% 5.21% 5.03%Q1 98 + 10.92% 13.30% 6.03% −5.56% 12.35% 8.57% 3.35% 8.38% 5.75%Q2 98 + −14.15% −8.86% −4.04% 4.29% 2.74% −4.12% −4.53% −10.53% −6.93%Q3 98 N 1.69% −11.30% −8.97% 0.98% 5.92% 5.77% −8.39% −8.30% −13.76%Q4 98 + −8.11% −4.27% −7.42% −4.71% 16.19% 2.28% −12.20% −12.03% −10.73%Q1 99 + 14.73% 5.33% 12.45% −0.04% 3.22% 6.08% 11.52% 9.34% 8.59%Q2 99 + 8.14% 1.85% 6.72% −0.31% 0.53% 8.46% 3.11% 1.85% 1.82%Q3 99 N −7.09% 0.78% −10.13% −3.00% 41.98% 45.66% −5.15% 0.61% 2.75%Q4 99 + −18.40% −32.52% −17.44% 13.48% 1.31% −1.89% −15.85% −20.65% −30.90%Q1 00 + −6.69% −9.90% −11.61% 2.91% −11.49% −16.24% −4.92% −3.01% −12.15%Q2 00 N 7.47% 10.72% 5.12% −2.67% −11.11% −13.66% 5.25% 8.20% 10.71%Q3 00 N 7.88% 4.79% 6.92% 4.58% 3.75% 7.95% 5.51% 1.39% 3.03%Q4 00 N 27.42% 36.22% 25.93% 3.85% −6.57% 30.22% 19.06% 20.60% 25.50%Q1 01 N 17.79% 26.65% 25.98% −5.68% −15.84% 7.21% 15.63% 15.13% 18.92%Q2 01 + 7.12% 3.30% 8.23% 3.83% 2.19% 9.59% 4.14% 2.70% 4.23%Q3 01 N 8.88% 8.01% 7.58% 9.14% 18.30% 15.42% 6.47% 5.20% 4.30%Q4 01 + −1.28% −9.22% 7.72% −2.26% −33.65% −20.98% 2.40% −2.96% 0.21%Q1 02 N 4.19% 3.60% 14.64% 9.72% 5.55% −4.77% 6.18% 4.15% 6.34%Q2 02 N 11.89% 13.80% 16.84% 8.28% 21.09% 14.46% 11.29% 9.84% 12.55%Q3 02 N 8.13% 0.67% 2.57% 5.11% 28.21% 34.00% 4.17% 7.37% 2.68%Q4 02 + 2.11% 7.56% −3.79% −10.72% −16.23% −20.66% −1.44% 0.28% −5.66%Q1 03 N 10.57% 2.92% −4.06% −7.85% 0.21% 2.23% −0.87% 2.77% −1.32%Q2 03 + −5.13% −0.45% 3.36% −7.21% −18.58% −22.71% 3.44% −0.17% 5.66%Q3 03 N 5.66% 12.90% 10.86% −4.89% 0.94% −1.29% 3.84% 3.17% 4.88%Q4 03 + 2.52% −1.14% −4.43% −1.17% 8.49% 4.10% −0.72% −1.17% −0.31%Q1 04 + 5.68% −3.40% 3.98% −1.18% 6.17% 1.98% 4.60% 3.46% 3.14%Q2 04 + −4.46% 3.94% 2.27% 5.25% −0.30% −3.08% 2.52% 1.09% 3.70%Q3 04 N 5.29% −0.38% 6.94% 12.77% 10.47% 10.02% 6.09% 2.70% 4.05%Q4 04 + 7.02% 3.90% 1.98% 1.68% −0.62% 2.66% 3.61% 4.73% 4.30%Q1 05 + 0.48% 7.78% −0.10% −1.24% −1.47% −3.00% 1.11% 2.17% 2.87%Q2 05 + 2.74% 0.63% 7.11% 6.56% 4.89% 2.11% 4.09% 0.98% 2.87%Q3 05 + −1.08% 6.74% 3.90% −4.70% 3.28% 8.04% −0.88% 0.04% 1.27%Q4 05 + 1.50% 1.49% 4.71% −5.33% −3.50% 2.44% 4.53% 3.31% 7.84%Q1 06 + −1.32% 10.41% 4.65% −7.29% 7.95% 3.30% 1.49% 5.29% 5.09%Q2 06 N 3.21% −1.51% 4.40% 6.45% −0.91% −4.23% 2.76% 0.60% −0.82%Q3 06 + 3.44% 6.36% 4.22% −1.57% 2.71% −0.29% 1.45% 2.68% 1.13%Q4 06 + −0.56% 2.57% 2.46% −1.07% 4.49% 0.01% −0.32% −2.44% −0.70%Q1 07 + −0.14% 3.32% 2.30% −3.00% 0.92% −4.88% 0.34% 1.89% −1.11%Q2 07 + 0.43% 0.66% 9.94% 2.28% 2.67% 4.88% 1.75% 0.44% 3.14%Q3 07 N −0.77% −5.95% −3.40% 3.74% 5.11% 5.78% −1.47% −4.51% −7.87%Q4 07 + 0.18% −6.67% −7.57% 8.53% 7.86% 9.18% −1.84% −5.54% −8.20%Q1 08 N −1.04% −2.75% 4.37% 2.53% 0.39% −5.22% −1.13% −1.95% −0.04%Q2 08 N 2.15% −4.71% −3.41% 2.47% 13.28% 16.21% −3.68% −9.06% −5.58%Q3 08 N 5.35% 1.44% 0.86% 4.73% −12.80% −12.29% 2.37% 4.21% −0.52%Q4 08 N −1.51% −10.37% −4.79% 15.61% −0.11% 7.00% −5.91% −7.19% −8.23%Q1 09 N −7.37% −15.04% −6.27% 9.71% 14.24% 7.47% −9.16% −13.18% −10.56%Q2 09 + 6.64% 9.53% 10.08% −8.18% −5.51% −9.73% 10.98% 9.38% 12.55%

Mean 2.20% 1.30% 3.21%, 1.63% 2.64% 2.46% 1.45% 0.56% 0.95%t-statistic 2.2875 0.9034 3.1401 1.6815 1.8406 1.6572 1.8494 0.6185 0.8990p-value 0.0258 0.3700 0.0026 0.0979 0.0707 0.1028 0.0694 0.5386 0.3723st. dev. 7.43% 11.15% 7.92% 7.50% 11.09% 11.50% 6.05% 7.07% 8.22%Number 60 60 60 60 60 60 60 60 60st. dev. (ann.) 14.87% 22.30% 15.85% 15.00% 22.19% 23.00% 12.11% 14.15% 16.45%SR 0.59 0.23 0.81 0.43 0.48 0.43 0.48 0.16 0.23

534 G. Elze / The Quarterly Review of Economics and Finance 50 (2010) 527–537

Table 3 (Continued)

Valuepremiums1/1−3/3

RoE (tO) &DY (tO)

RoE (tO) &PB(tO)

RoE (tO) &PE (tO)

DY (tO) &Levy27

DY (tO) &M03m

PB (tO) &Levy27

PB (tO) &M03m

PE (tO) &Levy27

PE (t0)& M03m

Q3 94 + 4.77% 12.97% 3.22% −11.85% −0.69% −11.99% −4.82% −4.44% 7.83%Q4 94 + 4.64% 11.27% 2.78% 0.48% 7.50% −1.64% −3.51% −0.39% −7.93%Q1 95 N 4.38% −2.93% −0.82% −10.59% 2.73% −18.27% −7.96% −6.84% 9.71%Q2 95 + 5.49% 0.56% 2.81% 12.70% 10.10% 10.56% 0.26% 1.67% 0.26%Q3 95 + 2.84% 9.09% 8.76% −2.33% −1.65% 1.98% 3.26% 5.66% 9.02%Q4 95 + 13.72% 16.95% 8.03% 21.29% 14.61% 16.99% 4.53% 16.35% 0.64%Q1 96 + −0.02% 11.06% 4.02% −7.43% −2.66% -10.71% −9.53% −3.47% 9.60%Q2 96 + 6.98% 5.40% 2.82% 5.62% 4.84% 10.20% 4.58% 7.76% 2.86%Q3 96 + 5.73% −8.52% 3.46% −2.91% 7.28% −9.71% −3.00% 1.62% 6.10%Q4 96 + 8.00% −7.37% 1.80% 11.17% 13.35% 5.78% −7.72% 7.08% 7.39%Q1 97 + −6.45% 1.23% 1.48% −2.87% −7.85% 0.89% 52.23% −4.65% −2.50%Q2 97 + 7.97% 6.63% 8.32% −4.75% 13.05% −26.72% 7.75% −3.70% 13.83%Q3 97 + −4.90% −8.55% −3.98% 4.47% 6.66% 18.03% 16.98% 11.60% 6.40%Q4 97 N 0.31% 7.40% −0.62% −0.38% 0.20% −2.98% −6.86% −8.67% −4.18%Q1 98 + −5.97% −3.66% −3.51% 26.67% 16.22% 29.21% 26.69% 25.44% 20.01%Q2 98 + −0.58% 8.48% 2.76% −9.28% −13.24% −5.83% −5.80% 1.93% −4.09%Q3 98 N −6.80% −24.04% −11.70% 0.66% 7.69% −15.47% −12.76% −13.22% −9.19%Q4 98 + −0.93% −16.49% −1.52% 14.37% −3.92% 21.38% 4.73% 17.45% −9.10%Q1 99 + 3.62% −3.37% 6.09% 18.95% 14.72% 1.16% −0.80% 5.37% 8.96%Q2 99 + 0.51% 7.06% 2.84% −0.79% 1.15% 0.20% 1.24% −9.09% −1.55%Q3 99 N −4.03% 13.30% −4.44% 1.64% 0.57% 5.60% 7.77% 0.96% −5.22%Q4 99 + −16.14% −31.88% −5.22% 3.96% 5.77% 0.66% 4.61% 1.88% 7.20%Q1 00 + −9.74% −3.10% −1.00% −1.30% −11.57% 9.99% −11.53% −2.91% −7.79%Q2 00 N 5.50% 11.11% −3.18% −9.76% −9.79% 0.33% −4.72% −3.28% −5.40%Q3 00 N 8.90% 19.21% 8.02% 6.54% 2.68% 11.17% 2.56% 11.86% 1.40%Q4 00 N 12.91% 47.52% 12.13% 24.74% 27.31% 35.09% 35.95% 12.51% 28.28%Q1 01 N 6.27% 18.49% 9.16% 14.88% 20.30% 10.56% 18.68% 16.07% 18.72%Q2 01 + 7.35% 22.31% 5.27% 9.53% 8.96% −2.40% −0.26% 8.23% 6.50%Q3 01 N 6.21% 17.50% 5.71% 16.31% 14.99% 13.16% 9.82% 13.83% 9.39%Q4 01 + 2.93% −6.83% 5.29% −18.90% −22.13% −18.33% −13.59% −13.28% −7.57%Q1 02 N 6.32% 10.69% 9.76% 9.53% 5.67% 4.74% 3.60% 10.03% 9.04%Q2 02 N 9.53% 32.12% 13.81% 25.04% 18.76% 20.03% 17.53% 22.49% 18.99%Q3 02 N 6.66% −8.93% 6.18% 26.31% 29.98% 18.30% 26.14% 17.09% 20.57%Q4 02 + −1.45% −7.21% −6.51% −9.27% −13.45% −10.42% −14.57% −13.54% −14.92%Q1 03 N −0.65% 8.06% −2.02% 9.98% 9.51% 7.04% 4.67% 5.76% 5.18%Q2 03 + 0.98% 3.16% 1.09% −12.87% −13.07% −8.67% −12.45% −7.31% −10.79%Q3 03 N −4.86% 2.89% 2.36% −2.17% −0.19% 7.15% 9.70% 8.72% 13.43%Q4 03 + −3.46% −22.91% −3.63% −0.27% 9.85% 1.11% 4.96% 1.34% 0.31%Q1 04 + 6.52% 6.34% −0.14% 2.32% −1.26% 0.78% −2.40% 8.01% 2.06%Q2 04 + −0.65% 3.40% 2.55% 3.64% −1.75% 0.61% −2.58% 8.29% 1.84%Q3 04 N 5.56% 21.48% 6.67% 10.01% 6.66% 13.13% 8.20% 9.87% 7.56%Q4 04 + 3.32% −14.47% 1.63% 6.85% 5.28% 6.07% 3.90% 4.72% 1.91%Q1 05 + −2.82% 19.78% 0.91% 2.88% 4.35% 3.06% 6.13% 5.88% 1.41%Q2 05 + 4.60% 13.84% 5.81% 6.96% 2.26% 5.55% 4.46% 11.07% 6.53%Q3 05 + −2.99% 1.44% −1.03% 2.79% 0.27% 4.89% 4.02% 6.20% −0.45%Q4 05 + −1.35% −19.27% 0.94% −5.46% −4.76% 1.97% 3.20% 7.48% 4.80%Q1 06 + −4.39% 14.44% −0.18% 5.24% 5.41% 9.91% 5.16% 12.34% 5.01%Q2 06 N 4.36% 12.98% 3.39% 4.31% 2.65% −6.46% −0.92% −2.14% 0.89%Q3 06 + −4.73% 14.61% −0.88% 0.68% 0.54% 3.63% 3.59% −0.50% −3.05%Q4 06 + 1.25% 6.43% 0.69% 7.26% 0.47% 8.77% 5.08% 10.27% 5.94%Q1 07 + −2.18% −24.32% 0.40% 7.63% 1.75% 4.86% −4.50% 8.25% 2.26%Q2 07 + 1.89% 14.13% 8.24% 6.04% 2.09% 5.61% 0.61% 6.33% 2.03%Q3 07 N 0.09% 2.21% 3.15% 1.34% −0.31% 2.45% 1.83% 4.78% 0.43%Q4 07 + −0.55% −5.82% −2.83% 2.94% −1.87% −5.82% −3.01% −7.10% −8.27%Q1 08 N 0.67% 11.04% 0.95% 7.17% −0.76% −2.01% −2.51% 2.98% 0.82%Q2 08 N 1.47% 1.25% 7.72% −5.49% 8.22% −7.93% 7.82% 2.68% 12.75%Q3 08 N −3.07% −16.21% −4.37% −12.87% 1.99% −7.27% −12.16% −20.98% −14.62%Q4 08 N 8.47% −4.80% −4.22% 4.37% 13.97% −12.61% −15.35% 3.08% 9.23%Q1 09 N 4.46% 10.97% 1.96% 10.00% 8.89% 3.91% 2.74% 8.09% 7.84%Q2 09 + 0.22% 7.39% 12.03% −9.97% −21.10% 1.01% −2.40% 5.46% −3.90%

Mean 1.61% 3.76% 2.22% 3.60% 3.45% 2.54% 2.65% 3.88% 3.24%t-statistic 2.2553 2.0388 3.4445 2.7318 2.6653 1.7398 1.7271 3.3196 2.8646p-value 0.0278 0.0460 0.0011 0.0083 0.0099 0.0871 0.0894 0.0015 0.0058st. dev. 5.53% 14.28% 4.99% 10.20% 10.04% 11.30% 11.91% 9.06% 8.76%Number 60 60 60 60 60 60 60 60 60st. dev. (ann.) 11.07% 28.56% 9.99% 20.39% 20.07% 22.60% 23.81% 18.13% 17.52%SR 0.58 0.53 0.89 0.71 0.69 0.45 0.45 0.86 0.74

Quarterly percentage portfolio return differences (value minus glamour) based on portfolios formed from EuroStoxx constituents using different one- and two-dimensionalvalue classification schemes. Portfolios reformation occurs yearly at the beginning of July during the period from 1994/95 to 2008/09. In addition, years in which the marketindexes declined are marked by “N”, years in which it rose are indicated by a “+”. The t-statistic for the test of the hypothesis that the differences in returns between valueand glamour are equal to zero, the corresponding p-value and the quarterly standard deviation of returns are indicated in the bottom lines. The return differences’ standarddeviation (annualized) and the Sharp Ratio (SR) are indicated for convenience as further useful information.

G. Elze / The Quarterly Review of Economics and Finance 50 (2010) 527–537 535

Table 4Average monthly returns in different market environments: worst 25 down months, next 50 down months, positive 80 up months, best 25 up months.

Panel A: Dividend yield

1 2 3 4 5 6 7 8 9 10 Eq.-w. 10−1 t-statistic p-Value

W25 −0.0954 −0.0891 −0.0939 −0.0851 −0.0852 −0.0848 −0.0793 −0.0842 −0.0844 −0.0720 −0.0852 0.0234 3.0424 0.0056N53 −0.0247 −0.0202 −0.0198 −0.0150 −0.0200 −0.0188 −0.0119 −0.0108 −0.0159 −0.0127 −0.0168 0.0120 2.0129 0.0506P80 0.0209 0.0278 0.0251 0.0254 0.0247 0.0256 0.0261 0.0271 0.0240 0.0279 0.0255 0.0070 2.3613 0.0205B25 0.0874 0.0803 0.0840 0.0790 0.0813 0.0788 0.0682 0.0766 0.0716 0.0721 0.0778 −0.0153 −1.5309 0.1389

Panel B: DY (t0) & RoE (t0)

3/3 3/2 2/3 3/1 2/2 1/3 2/1 1/2 1/1 Eq.-w. Index 1/1−3/3 t-statistic p-Value

W25 −0.0946 −0.0900 −0.0937 −0.0921 −0.0822 −0.0784 −0.0773 −0.0811 −0.0784 −0.0855 0.0000 0.0162 2.6067 0.0155N50 −0.0268 −0.0209 −0.0202 −0.0203 −0.0141 −0.0164 −0.0164 −0.0132 −0.0105 −0.0173 0.0000 0.0163 3.3483 0.0017P80 0.0194 0.0269 0.0253 0.0287 0.0262 0.0237 0.0255 0.0288 0.0261 0.0257 0.0000 0.0067 2.4413 0.0167B25 0.0972 0.0791 0.0909 0.0831 0.0695 0.0815 0.0779 0.0691 0.0687 0.0786 0.0000 −0.0285 −3.6663 0.0012

Panel C: DY (t0) & Levy27

3/3 3/2 2/3 3/1 2/2 1/3 2/1 1/2 1/1 Eq.-w. 1/1−3/3 t-statistic p-value

W25 −0.1071 −0.0872 −0.1021 −0.0837 −0.0804 −0.0959 −0.0743 −0.0727 −0.0692 −0.0853 0.0379 2.5519 0.0175N50 −0.0309 −0.0184 −0.0182 −0.0155 −0.0177 −0.0180 −0.0109 −0.0117 −0.0038 −0.0164 0.0270 3.3124 0.0019P80 0.0209 0.0224 0.0225 0.0293 0.0254 0.0265 0.0285 0.0256 0.0290 0.0257 0.0080 2.1662 0.0331B25 0.1026 0.0782 0.0887 0.0775 0.0729 0.0817 0.0723 0.0677 0.0769 0.0781 −0.0257 −1.7922 0.0857

Panel D: DY (t0) & MO3m

3/3 3/2 2/3 3/1 2/2 1/3 2/1 1/2 1/1 Eq.-w. 1/1−3/3 t-statistic p-Value

W25 −0.1112 −0.0895 −0.1005 −0.0773 −0.0833 −0.0972 −0.0734 −0.0758 −0.0581 −0.0853 0.0531 4.0470 0.0005N50 −0.0300 −0.0193 −0.0182 −0.0157 −0.0142 −0.0155 −0.0151 −0.0113 −0.0072 −0.0161 0.0229 3.5922 0.0008P80 0.0221 0.0261 0.0257 0.0268 0.0256 0.0276 0.0274 0.0257 0.0289 0.0261 0.0068 1.7503 0.0837B25 0.1042 0.0785 0.0874 0.0778 0.0730 0.0843 0.0746 0.0657 0.0719 0.0779 −0.0323 −1.9776 0.0596

Panels A–D present the average percentage decile portfolio returns, the average total performance, the return differences (value minus glamour) on a monthly basisrepresentatively for one-dimensional value strategies based on DY; for two-dimensional strategies based on DY and RoE, DY and Levy27, DY and MO3m. The performanceo 25 wor otheri atisticp

aketarao

Wf5iuiSwmcggo

p2

a

ostsIsibme

mtroysc

f our portfolios are divided into four states of general market environments; theemaining 50 negative months other than the 25 worst, the 80 positive up monthsn returns between value and glamour for each state is also reported along with t-st-value.

nd RoE, Levy27 and MO3m during four states of the global mar-et; the 25 worst stock return months36 in the sample based on thequally weighted index, the remaining 50 negative months otherhan the 25 worst, the 80 positive months other than the 25 best,nd the 25 best months in the sample. The average difference ineturns between the value and glamour portfolio for each state islso reported along with t-statistics for the test that the differencef returns is equal to zero and its corresponding p-value.37

The results in this table are ambiguous (due to low t-statistics).hile for most classification schemes, the value portfolio outper-

ormed glamour in the market’s worst 25 months and in the next0 negative months, results in all cases are only statistically signif-

cant for the category “Next negative 50 months”.38 For example,sing DY, the value portfolio lost an average of 7.2% of its value

n the worst 25 months, whereas glamour lost 9.5% of its value.imilarly, the value portfolio outperformed glamour in the nextorst 50 months in which the index declined. It lost 1.5% in theseonths while glamour experienced a 2.5% decline. Similar results

an be observed for other one- and two-dimensional value strate-ies as well (refer to the appendix section). The value strategy didenerally better when the market fell. In the 80 positive monthsther than the best 25, one- and two-dimensional value strategies

36 We widely follow Lakonishok et al. (1994) in the structure of this analysis.37 Slight differences in returns and risk measures are due to the fact that negativearameter characteristics are not in included in this analysis. Refer also to Section.38 Illustrations are presented for dividend yield (DY). For a complete list includingll variable examined refer to the appendix section. Similar results can be observed.

(f

tas

v

d

rst stock return months in the sample based on the equally weighted index, thethan the 25 best, and the 25 best up months in the sample. The average differences for the test that the differences of returns are equal to zero and its corresponding

utperformed glamour slightly. However, results lack statisticalignificance in some cases. In the very best 25 months, one- andwo-dimensional value strategies underperformed glamour39 sub-tantially. Return results however also lack statistical significance.f anything, the superior performance of the value strategy iskewed toward negative market return months rather than pos-tive market return months. Overall, the value strategy performedetter than the glamour in all states other then extreme upwardovements. Table 4 thus indicates that the value strategy does not

xpose investors to greater downside risk.Finally, for completeness, Table 5, Panels A–D present some

ore traditional risk measures for portfolios using our classifica-ion schemes. These risk measures are calculated using quarterlyeturn measurement intervals over the sample period. For eachf our strategies, we have 60 quarterly return observations in theear following the first formation and hence we can compute thetandard deviation of quantile returns. We also have calculatedorresponding returns on the total return of the EuroStoxx indexequally weighted) and can calculate a beta for each quantile port-olio.

First, the beta of the equally weighted index is lower then forhe market-capitalized index. Secondly betas for value portfoliosre similar or much smaller than glamour portfolios.40 Only forome single value strategies are betas of the value portfolios slightly

39 These results are consisted with earlier observations. For a complete list of allariables examined refer to the appendix section.40 Refer also to the appendix section for variable combinations other than withividend yield.

536 G. Elze / The Quarterly Review of Economics and Finance 50 (2010) 527–537

Table 5Traditional risk measures: beta, standard deviation.

Panel A: Dividend yield

1 2 3 4 5 6 7 8 9 10 Eq.-w.

Beta ann. 1.0836 1.0481 1.0962 0.9987 1.0077 1.0012 0.8911 0.9784 0.9907 0.9265 0.9079St. dev. (ann) 0.2098 0.1994 0.2050 0.1860 0.1870 0.1862 0.1671 0.1835 0.1852 0.1801 0.1796

Panel B: DY (t0) & RoE (t0)

3/3 3/2 2/3 3/1 2/2 1/3 2/1 1/2 1/1 Eq.-w.

Beta ann. 1.1358 1.0360 1.1008 1.0642 0.9211 0.9837 0.9364 0.9544 0.9199 0.9122St. dev. (ann) 0.2187 0.1947 0.2059 0.2068 0.1723 0.1926 0.1819 0.1793 0.1776 0.1742

Panel C: DY (t0) & Levy27

3/3 3/2 2/3 3/1 2/2 1/3 2/1 1/2 1/1 Eq.-w.

Beta ann. 1.2331 1.0090 1.1042 0.9885 0.9497 1.1074 0.9047 0.8805 0.9046 0.9093St. dev. (ann) 0.2477 0.1899 0.2119 0.1927 0.1773 0.2105 0.1735 0.1668 0.2030 0.1914

Panel D: DY (t0) & MO3m

3/3 3/2 2/3 3/1 2/2 1/3 2/1 1/2 1/1 Eq.-w.

Beta ann. 1.2773 1.0326 1.1206 0.9496 0.9590 1.1341 0.8966 0.8920 0.7953 0.9088St. dev. (ann) 0.2550 0.1930 0.2141 0.1872 0.1774 0.2181 0.1745 0.1691 0.1743 0.1920

Panel A shows the beta with respect to the European market index EuroStoxx, the decile return standard deviations and the return value premium’s standard deviations( arterld t to tht o-dimq July d

hsssdvdoi

dqdgtrrhdtvdbr

5

tD

v

ww

ttwaskmvo

poectbai

s3(ssts

annualized = ann.) for various one-dimensional value strategies (DY) based on quuring the period from 1994/95 to 2008/09. Panels B–D show the beta with respeche return value premium’s standard deviations (annualized = ann.) for various twuarterly performance data. Portfolio reformation occurs yearly at the beginning of

igher than for the glamour portfolios. In addition, it seems thattrategy combinations based on dividend yield (DY) exhibit themallest betas in the value quantile portfolios. If anything, theuperior performance of the strategies occurs disproportionallyuring “bad” performances of the stock market (refer to the pre-ious pages: performance in different market environments). Theifference in betas of less then 0.1 can explain only a small portionf return difference41 and surely not the roughly 5–17% differencen returns that we find.

Additionally, Table 5, Panels A–D present the annualized stan-ard deviations of one- and two-dimensional strategy portfoliouantile returns based on DY. It shows that value quantile portfolioso not generally have higher standard deviations of returns thenlamour portfolios.42 For example, using the DY, RoE classification,he value quantile portfolio has an average standard deviation ofeturns of 17.8% compared to 21.9% for the glamour quantile. Someemarks about these results are necessary. First, because of its muchigher mean return, the value strategy’s higher standard deviationoes not necessarily translate into greater downside risk. Second,he small differences in standard deviations of returns betweenalue and glamour portfolios are quite small in comparison to theifference in average return (around 5–17% per year). A risk modelased on differences in standard deviation cannot explain the supe-ior returns on these strategies.

. Conclusion

Value investing is an investment paradigm that derives fromhe ideas on investment and speculation that Ben Graham & Davidodd began teaching at Columbia Business School in 1928. Since

41 This conclusion is not valid for value strategies based on P/B where betas for thealue quantile portfolios seem to be substantially higher than for glamour portfolios.42 It seems however that glamour and value strategies generally exhibit some-hat higher standard deviations than other portfolio quantiles or the total equallyeighted index.

sirffii

y performance data. Portfolio reformation occurs yearly at the beginning of Julye European market index EuroStoxx, the quantile return standard deviations andensional value strategies (DY and RoE, DY and Levy27, DY and MO3m) based onuring the period from 1994/95 to 2008/09.

hen numerous empirical academic studies have consistently foundhat value stocks outperform glamour stocks and the market as ahole. This article extended prevailing research on existing value

nomalies. It evaluated simple value strategies for the Europeantock market (compared to many other studies that test mar-et data on a country-by-country basis) as well as sophisticatedulti-dimensional value strategies that also include capital return

ariables (Consistent Earner Strategy) and momentum factors (Rec-gnized Value Strategy).

In Section 3 (a) of this analysis it was shown that a variety of sim-le classification schemes sorting value and glamour stocks basedn dividend yield (DY), price-to-book ratio (P/B) and price-to-arnings ratio (P/E) produced superior returns for value portfoliosompared to glamour. As market proxy for the European markethe EuroStoxx index was selected. Return differentials (premiums)etween value and glamour varied between 5.40% and 12.66% pernnum on average depending on the selection criteria chosen dur-ng the period from July 15, 1994/95 to June 30, 2008/09.

Motivated by these results we subsequently examined portfoliotrategies based on two-dimensional selection criteria in Section.1 (b). First simple value measures (as evaluated in Section 3a)) were combined. It was shown that two-dimensional valuetrategies (Multiple Value) based on a combination of simple valuetrategies do not further improve investment performance and sta-istical significance (in fact, investment returns were smaller andtatistically not significant).

Subsequently more sophisticated two-dimensional valuetrategies were evaluated. The Consistent Earner Strategy includ-ng capital return variables (e.g. RoE) resulted in investmenteturns similar to simple value strategies but much better than

or single capital return variables. Return differences (premiums)all in a range between 6.43% and 14.78%. Statistical significancemproved substantially.43 The Consistent Earner Strategy mimicsnvestment styles of well-know investors like Warren Buffett or

43 Refer to Table 3, page 14 to compare statistical significances.

onomi

JbpagG

adpIasaah

owscoqs

R

B

–

B

BC

D

F

F

F

G

GGKK

L

L

L

Levy, Robert A. (1967). Relative strength as a criterion for investment selection.Journal of Finance, V22, 595–610.

Oyefeso, O. (2004). Literature survey of measurement of risk: The value premium.Journal of Asset Management, 5(4), p. 277–288, 12p.

G. Elze / The Quarterly Review of Ec

oel Greenblatt who further developed the value investing concepty focusing on “finding an outstanding company at a sensiblerice” or buying “cheap and good companies with competitivedvantages indicated by a high return on capital” rather thaneneric companies at a bargain price as originally promoted byraham and Dodd.

With regard to strategies combining momentum and value vari-bles (Recognized Value Strategy), both investment performanceifferences (premiums) and statistical significance improved com-ared to simple value and/or simple momentum variables.

nvestment returns fell in a range between 11.08% and 17.05% pernnum on average.44 The Recognized Value Strategy is based on thetock momentum life cycle hypothesis45 stating that stocks movelternately through periods of relative “glamour” and “neglect”ttempting to reconcile intermediate horizon momentum and longorizon-reversals of behavioral finance theories.

Finally in Section 4 the question of whether strategies basedn investment criteria previously chosen are fundamentally riskieras evaluated. Evidence could be provided that, in general, value

trategies based both on one- and two-dimensional simple valueriteria as well as “sophisticated” strategies including capital returnr momentum variables have outperformed glamour strategiesuite consistently without support for the hypothesis that valuetrategies are fundamentally riskier.

eferences

all, R. (1978). Anomalies in relationships between securities’ yields and yield-surrogates. Journal of Financial Economics, 6, 103–126.

44 Momentum criteria here included representatively Levy27 and relative strength3 month (MO3m).

45 Lee & Swaminathan (2000). Price momentum & trading value. Journal of Finance.

R

cs and Finance 50 (2010) 527–537 537

asu, S. (1977). Investment performance of common stocks in relation to their priceearnings ratios: A test of the efficient market hypothesis. Journal of Finance, 32,663–682.

asu, S. (1983). Earnings’ yield and the size effect. Journal of Finance.han, L., & Lakonishok, J. (2004). Value and growth investing: Review and update.

Financial Analysts Journal, 60(1), 71–86. January/February.e Bondt, W, & Thaler, R. (1985). Does the stock market overreact? Journal of Finance,

40, 793–805.ama, & French. (1992). The cross-section of expected stock returns. Journal of

Finance, 46, 427–466.ama, & French. (1998). Value versus Growth the International Evidence. Journal of

Finance.ama, E., & MacBeth, J. (1973). Risk, return and equilibrium: Empirical tests. Journal

of Political Economy, 81, 607–636.ilovich, T., Griffin, D., & Kahneman, D. (Eds.). (2002). Heuristics and biases: The

psychology of intuitive judgment. Cambridge University Press.raham, B., & Dodd, D. (2005). Security analysis. McGraw-Hill.reenblatt, J. (2006). The little book that beats the market. Wiley & Sons.ahneman, & Tversky. (1979). Prospect Theory. Econometrica, 17, 263–291.ahneman, D., & Riepe, M. (1998). Aspects of investor psychology. The Journal of

Portfolio Management, 24, 52–65.akonishok, Hamao, & Chan. (1991). Fundamentals and stock returns in Japan. Jour-

nal of Finance.akonishok, Shleifer, & Vishny. (1994). Contrarian investment, extrapolation, and

risk. Journal of Finance.ee, C., & Swaminathan, B. (2000). Price momentum & trading value. Journal of

Finance.

osenberg, B., Reid, K., & Lanstein, R. (1984). Persuasive evidence of market ineffi-ciency. Journal of Portfolio Management, 11, 9–17.