Digesting anomalies in emerging European markets: A comparison of factor
pricing models

Adam Zaremba, Anna Czapkiewcz

PII: S1566-0141(16)30174-1
DOI: doi: 10.1016/j.ememar.2016.12.002
Reference: EMEMAR 490

To appear in: Emerging Markets Review

Received date: 31 May 2016

Revised date: 11 December 2016
Accepted date: 16 December 2016

Please cite this article as: Zaremba, Adam, Czapkiewcz, Anna, Digesting anomalies in
emerging European markets: A comparison of factor pricing models, Emerging Markets
Review (2016), doi: 10.1016/j.ememar.2016.12.002

This is a PDF le of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its nal form. Please note that during the production process
errors may be discovered which could aect the content, and all legal disclaimers that
apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Digesting Anomalies in Emerging European Markets: A Comparison of Factor

Pricing Models

PT
RI
Adam Zaremba
(corresponding author)

SC
Poznan University of Economics and Business
adam.zaremba@ue.poznan.pl

NU
Anna Czapkiewcz
MA
AGH University of Science and Technology
gzrembie@cyf-kr.edu.pl
ED
PT
CE
AC
 ACCEPTED MANUSCRIPT

Digesting Anomalies in Emerging European Markets: A Comparison of

Factor Pricing Models

T
Abstract

P
This study compares the performance of four popular factor pricing modelsthe capital

RI
asset-pricing model (Sharpe, 1964), the three-factor model of Fama and French (1993), the

SC
four-factor model of Carhart (1997), and the five-factor model of Fama and French (2015a)
testing their explanatory power over a broad range of cross-sectional return patterns in

NU
emerging European markets. We identify, classify, and replicate 100 anomalies documented in
the financial literature. Only 20 (32) of the capitalization-weighted (equal-weighted) anomaly
portfolios are significantly profitable. We show that the five-factor model best explains the
MA
returns of anomaly portfolios and verify its superiority over the other models.

Keywords: asset pricing, factor models, anomalies, emerging European markets,

ED

emerging markets, cross section of returns, size, value, momentum, profitability, asset growth
PT

JEL Codes: G11, G12, G14

CE
AC

1. Introduction
A key characteristic of any effective factor pricing model is its power to explain
patterns in the cross section of stock returns. Presenting their empirical three-factor model in
1996, Fama and French demonstrated it could adequately summarize all state-of-the-art cross-
sectional patterns known to science at that time. This notion has subsequently been challenged
by several researchers, such as Green et al. (2016), Jacobs (2015), Harvey et al. (2015), and
Hou et al. (2014), who have discovered dozens of anomalies unexplainable by the Fama-
French three-factor framework. This shortcoming has spurred the creation of a new generation
of models incorporating an array of new factors related to momentum, investment, and
profitability (Carhart, 1997; Fama & French, 2015a).
 ACCEPTED MANUSCRIPT

This study aims to conduct a comprehensive out-of-sample test of each of the most
popular factor pricing models 1 to verify their power to explain cross-sectional patterns
(anomalies) in emerging European stock markets. We evaluate four popular factor pricing
models: (a) the capital asset-pricing model (CAPM) (Sharpe, 1964); (b) the Fama-French

T
(1993) three-factor model (FF3) capturing the size and value effects; (c) the four-factor model

P
(C4), following the original idea by Carhart (1997) and extended to take into account the

RI
momentum phenomenon; and (d) the new Fama-French (2015a) five-factor model (FF5) in
which the momentum effect is replaced with profitability and investment factors.

SC
To evaluate the performance of these factor pricing models, we replicate 100 cross-
sectional anomalies in a cross section of returns already identified in developed countries.

NU
Using sorting procedures, we form equal-weighted and capitalization-weighted long-short
anomaly portfolios and categorize the anomalies into 16 distinct groups, including
MA
momentum, value versus growth, quality investing, and seasonal effects. We then apply each
multifactor pricing model to the anomalies to examine their power to explain abnormal
returns. We supplement these examinations with further validation of model usefulness for
ED

asset pricing in emerging European markets.

We chose emerging European markets for a few reasons. First, we intend to make our
PT

research largely out-of-sample. Most of the anomalies in this study (including pricing factors,
even those underlying the FF5 model) have never been examined in this region. Most stock
CE

market anomalies were first discovered in the U.S. market and have yet to be measured in
other markets. Interestingly, Dimson and Marsh (1999) and McLean and Pontiff (2016)
uncover that various anomalies frequently escape out-of-sample studies (i.e., are not subjected
AC

to out-of-sample testing). Importantly, among studies evidencing the failure of even the most
prominent anomalies in emerging markets, Cakici, Fabozzi, and Tan (2013) find no evidence
of momentum in Eastern Europe. Our sample could also reveal differences in the cross-
sectional patterns of returns.
Second, according to widely-accepted belief, anomalies tend to intensify in less
efficient markets and, as such, should be relatively robust in emerging economies which are
markedly less liquid (Lesmond et al., 2004) and are characterized by distinctly higher trading
costs (Investment Technology Group, 2015). The higher liquidity constraints and transactions
costs could translate to abnormally elevated returns on stock market anomalies.

1
We refer to the models as factor pricing models, following the convention used by Cochrane (2005, p. 78).
The same models are also described as factor asset-pricing models (e.g., by Fama and French [2015a]).
 ACCEPTED MANUSCRIPT

Third, whereas anomalies tend to appear particularly strong on the short side
(Stambaugh et al., 2012), the short sale in emerging markets is rarely available. Thus, short
sale constraints in emerging European markets could contribute to the magnitude of abnormal
returns.

T
Last, but not least, the stock markets in emerging Europe have been growing rapidly,

P
both in terms of market capitalization and the absolute number of stocks. This reflects the

RI
increasing importance of international investors who still seek portfolio diversification in
emerging markets despite the ongoing integration between the emerging and developed

SC
markets in the post-liberalization period (Bekaert & Harvey, 2002).
The principal findings of this study can be summarized as follows: only 32 (20) of the

NU
100 equal-weighted (capitalization-weighted) anomaly portfolios display mean returns that
are both positive and significantly different from zero, mostly within the value investing,
MA
profitability, and issuance categories. Interestingly, few return patterns specific for mature
markets are observed within our sample, including the failure of the most notable momentum
anomaly to deliver significant and robust abnormal returns.
ED

Our research also shows the FF5 model clearly outperforms the earlier models in
terms of its explanatory power over patterns in the cross section of returns. The CAPM fails to
PT

account for the returns on numerous capitalization-weighted anomaly portfolios, and both the
FF3 and C4 models turn out to be efficient almost exclusively in value-versus-growth
CE

strategies, and small-firm and low-price effects. Against this background, the FF5 model not
only delivers a better explanation of value versus growth patterns, but also correctly explains
many of the profitability anomalies, making it the model best-suited to the data from
AC

emerging European markets. We validate the usefulness of the FF5 model for asset pricing by
designing various sets of portfolios from double sorts and confirm its ability to explain the
cross section of returns. We also show that all component factors of the model bear significant
risk premia, except for the conservative minus aggressive (CMA) factor representing
investment patterns.
Our study contributes in several ways. First, we conduct the first comprehensive
comparison of performance of factor pricing modelsincluding the recent FF5not only in
Eastern Europe, but in any emerging markets. Earlier studies were focused solely on the
performance of these models in developed markets (e.g., Fama & French, 2015a; Chiah et al.,
2016). Although Fama and French (2015b) and Cakici (2015) report the results of research in
the markets of 23 different countries, none of these samples include any emerging markets.
Furthermore, the existing investigations of asset-pricing models in emerging Europe did not
 ACCEPTED MANUSCRIPT

consider the five-factor asset-pricing models at all (e.g., Borys, 2011; Foye et al., 2013;
Waszczuk, 2013; Zaremba, 2015; Czapkiewicz & Wjtowicz, 2015).
Second, we review and test the broadest possible sample of equity anomalies in the
emerging European markets. We conduct these wide-ranging examinations of stock market

T
anomalies in the spirit of Green et al. (2016), Jacobs (2015), and Harvey et al. (2015),

P
although none of these studies that focus on comparable arrays of anomalies include

RI
emerging markets in their samples. The articles available to an international audience focus
on not more than a few anomalies, particularly those that have included emerging European

SC
markets in the research (e.g., Cakici, Fabozzi, & Tan, 2013; Waszczuk, 2013). To our
knowledge, when general emerging markets are considered, the broadest study ever

NU
conducted is by Li et al. (2016), who investigates 16 well-known predictive signals in
multiple countries.
MA
Third, by examining these 100 anomalies in the context of emerging European
markets, our research also contributes to the fast-growing strain of academic studies tracking
commonalities in cross-sectional return patterns in global equity markets (e.g., Asness,
ED

Moskowitz, and Pedersen (2013) for value and momentum, Frazzini and Pedersen (2014) for
beta, Blackburn and Cakici (2016) for long-term reversal, or Jacobs (2016) for a set of 11
PT

popular anomalies). We document which of the return patterns have their parallels in
emerging Europe.
CE

Finally, we also provide fresh new evidence supporting the hypothesis that numerous
anomalies are likely to be a result of data-mining. This is in line with the findings of Dimson
and Marsh (1999) and McLean and Pontiff (2016), showing that numerous return-predictive
AC

signals perform poorly in out-of-sample tests due to either investor learning or data-mining
biases. Neither of these studies focused on emerging Europe or emerging markets in general.
Summing up, by researching the above-discussed issues, our study aims to provide new
insights into asset pricing in emerging markets, as well as important results for both the
academic community and market practitioners.
Section 2 lays out both the data and methods we used. Section 3 presents the findings,
while section 4 presents our research conclusions.

2. Data and Methods

To compare the performance of the factor pricing models, we first form long-short
zero-investment portfolios based on 100 anomalies derived from stock market returns and
 ACCEPTED MANUSCRIPT

chronicled in the academic literature. We subsequently attempt to explain the returns with the
models factors and validate the model that best explains the cross-sectional patterns with
additional asset-pricing tests.

T
Data Source and Sample Preparation

P
Our sample includes the five most important stock markets in Eastern Europe: the Czech

RI
Republic, Hungary, Poland, Russia, and Turkey. The selection stems from the composition of
the MSCI Emerging Europe index, and is identical to the Eastern European samples used in

SC
other studies (e.g., Cakici et al., 2013).
We source both international stock returns and accounting data from Bloomberg,

NU
including both listed and delisted companies to eliminate any survivorship bias. 2 Our
computations are based on a monthly time series, and the returns are adjusted for corporate
MA
actions and cash distributions. The sample period of returns extends from December 1997 to
November 2015. We also used earlier data when necessary for calculating certain anomalies
(e.g., long-term reversal).
ED

A company is included in the sample when both its return in month t and its total
capitalization at the end of month t-1 can be determined. To ensure the quality of our sample
PT

and align with market practices, we adopt several static and dynamic filters. As the sample
contains only common stocks, we exclude closed-end funds, ETFs, GDRs, and similar
CE

investment vehicles, allowing only the securities for which the Czech Republic, Hungary,
Poland, Russia, or Turkey were the primary markets. Considering the practical problems with
so-called penny stocks, we eliminate any company from the sample in month t when at the end
AC

of month t-1 either its nominal share price drops below 0.30 USD or the total stock market
capitalization sinks below 8 million USD. 3 Finally, following Rouwenhorst (1999), we
manually screen the data for suspicious returns. We use all the companies available in
Bloomberg, and our final sample includes 1913 companies. The basic composition of the
sample, with the number of stocks varying every month, is presented in Figure A1, Appendix 1.
All data are converted into USD. Whenever a strategy relies upon accounting data, we
use lagged values from month t-5 to eliminate the look-ahead bias. Finally, to be consistent
with our use of USD, we use the 1-month U.S. T-bill rate as a proxy for the risk-free rate.

2
The list of companies is compiled using quarterly downloads and the equity screening function (EQS). We
select listed, delisted, liquidated, withdrawn, and acquired companies.
3
A review of methods used to eliminate the penny stocks is provided in Waszczuk (2014).
 ACCEPTED MANUSCRIPT

The Factor Pricing Models

The asset-pricing formula can be expressed as follows:

  =  +   + +   , (1)
with the corresponding seemingly unrelated regressions (SUR) model:

T
 =  +   + . +  + , (2)

P
where  is a vector of portfolios excess returns at time t,  ,,  are risk factors observed

RI
at time t,  .  are vectors of risk factor sensitivities or loadings,  is a vector of
intercepts, and is a vector of disturbances. The  ,,  denote the risk premium

SC
parameters associated with the corresponding risk factors.
First, we consider the classical CAPM model, represented by the following SUR

NU
model:
 =  +    + , (3)
MA
where   is the excess return on the market portfolio observed at time t.
Tested second is the FF3 model, where portfolios excess returns depend additionally
on small minus big () and high minus low () factors that reflect the size and value
ED

effects in equity markets, i.e.,

 =  +    +   +   + . (4)
PT

The third model tested is the C4 model, which incorporates the winners minus losers
(WML) factor representing the momentum effect, where
CE

 =  +    +   +   +   + . (5)
The last model tested is the FF5 model, where the regression equations in the SUR
AC

model represent the relationship between excess returns and five risk factors:
 =  +    +   +   +   +   +  . (6)
The model represented by formula (6) replaces  in (5) with two additional
factorsrobust minus weak (  ) and conservative minus aggressive (  )which
correspond with the profitability and investment patterns of the examined companies.

Asset-Pricing Factors
Our models rely on six distinct factors derived from cross-sectional data observed at
time t:  , , , , , and . The market risk factor,  , is calculated
as an excess return over the risk-free rate of a capitalization-weighted portfolio formed from all
securities in the sample. To compute the five remaining classical cross-sectional factors, the
companies are sorted by their B/M ratio, size (total stock market capitalization), momentum
 ACCEPTED MANUSCRIPT

(lagged cumulative return in months t-12 to t-2), operating profitability (trailing 12 months
operating income minus trailing 12 months interest expense, divided by total equity at t-1), and
investment (percentage change of total assets in the 12 months ending t-1) at time t-1. Big and
small companies are categorized by capitalization above or below the median at time t-1,

T
respectively. The B/M ratio breakpoints are the 30th and 70th percentiles of the B/M ratio for all

P
the companies measured at time t-1. The intersection of the independent 23 sorts on size and

RI
B/M ratio produced six portfoliosSG, SN, SV, BG, BN, and BVwhere B and S indicate big
or small, and G, N, and V stand for growth, neutral, and value (bottom 30%, middle 40%, and

SC
top 30% of B/M), respectively. Next, we compute the monthly capitalization-weighted returns
for all six portfolios. The t-month return on the size factor, , is calculated as the mean

NU
return on the three small company portfolios from the 23 size-B/M sorts minus the mean
return of the three big company portfolios. The return on the value factor,  , is the
MA
difference between the mean return of the value portfolios (BV, SV) and the mean return of the
growth portfolios (BG, SG). The three remaining factors  ,  , and  are
calculated in the same way as  with the B/M ratio of the stocks sorted by momentum,
ED

operating profitability, and investment, respectively.

The fundamental statistics of the asset-pricing factors used in the study are displayed in
PT

Table 1. Interestingly, only two of these factors, HML and RMW, deliver positive mean
monthly returns that differ significantly from zero. The returns on MKT and SMB are
CE

positive, but not statistically significant. Similarly, the CMA factor fails to deliver a
significant mean, which corresponds with the findings of Jacobs (2015) that the asset growth
anomaly was not significantly profitable in Poland, Russia, and Turkey. The WML portfolio
AC

displays a mean monthly return virtually equal to zero. Although this observation starkly
contradicts the evidence from developed markets, it is consistent with other similar studies of
the region (e.g., Cakici et al., 2013).
[Insert Table 1 here]

Anomaly Portfolios
We examine anomaly portfolios based on 100 distinct stock market cross-sectional
patterns. The selection of the anomalies was motivated by previous research studies on cross-
sectional return patterns, and specifically included the selections made by Green et al. (2016),
Hou et al. (2014), and Jacobs (2015), which we supplement with a few additional screens. The
anomaly is first computed using accounting and market data from standard databases, such as
Bloomberg. As we require a monthly frequency of the anomaly, it is derived from monthly
 ACCEPTED MANUSCRIPT

data. We attain the anomaly returns through the long-short portfolios based on cross-sectional
rankings of securities and determine the anomaly using the data available in emerging
European markets.
A detailed description of the anomalies together with the related portfolio formation

T
procedures is presented in Table A1, Appendix 1. The 100 anomalies are classified into 16

P
categories based on the underlying economic concept. These are presented in summary in

RI
Table 2.
[Insert Table 2 here]

SC
Based on the 100 anomalies, we design portfolios applying uniform procedures across
all strategies. All stocks within the sample are sorted on the anomaly-relevant metrics as at the

NU
end of each month. All the metrics are detailed in Table A1, Appendix 1. Next, unless stated
otherwise in Table A1, all the securities from the top and bottom quintiles of the rankings are
MA
used to form equal-weighted and capitalization-weighted portfolios. Last, we form zero-
investment portfolios, which are essentially standard long-short portfolios.
We assume a long (short) position in the portfolio if it is expected to provide higher
ED

(lower) returns based on the available empirical evidence. Therefore, we expect all zero-
investment portfolios to display positive returns.
PT

Portfolios from Double Sorts

Having identified the model which best explains the anomaly returns, we validate its
CE

performance using portfolios from double sorts following Fama and French (2012) and Cakici
et al. (2013). The 44 portfolios from double sorts combine the variables underlying the
factors in the FF5 model (i.e., size, B/M ratio, investment, and operating profitability, defined
AC

as in the previous subsection). The methods for calculating the variables remains identical to
that for the asset-pricing factors. All the stocks within the sample are first ranked based on the
selected variable and subsequently grouped according to the 25th, 50th, and 75th percentile
breakpoints. Independent 44 sorts on two selected variables produces 16 double-sorted and
capitalization-weighted portfolios.
In our baseline approach we include three sets of portfolios from sorts on (a) size and
B/M ratio; (b) size and operating profitability; and (c) size and investment. As a robustness
check, we also validate the results by examining all remaining combinations (i.e., value and
profitability; value and investment; and investment and profitability).
Estimation Methods and Testing
Our investment universe initially comprises the 100 portfolios formed from the 100
anomalies. As we aim to evaluate the factor pricing models ability to explain the cross-
 ACCEPTED MANUSCRIPT

sectional patterns in returns, we first filter out from the sample the return-predictive signals
with insignificant means. In the tests, we consider only those anomalies which show mean
monthly returns that are both positive and significantly different from zero at the 10% level.
Subsequently, we employ a two-step approach. First, we determine which model best explains

T
the returns on the anomalies. Second, based on separate sets of portfolios from double sorts

P
we confirm the explanatory power of the best model.

RI
In the first step, we construct the SUR model like (2), where  is assumed to be
a vector of these selected anomalies. We assess whether the betas in (2) correctly capture the

SC
cross-sectional variation in the returns,  , of the ith portfolio at time t (i.e., if the intercepts
alpha simultaneously equaled zero in all the portfolios). We test the null hypothesis  :  = 0

NU
against  with two tests: the GRS test, proposed by Gibbons, Ross, and Shanken (1989), and
the generalized moment method (GMM)-based test. Although the GRS test is very popular,
MA
the major advantage of the GMM approach is its robustness to both conditional
heteroscedasticity and serial correlation in the returns on the explanatory factors and tested
portfolios. The procedure for testing  in the GMM framework is derived from MacKinlay
ED

and Richardson (1991) and Cochrane (2005) and is detailed in Appendix 2.

In the second step, to validate the models usefulness for asset pricing, we also
PT

construct the SUR model following (2) and introducing  , the vector of excess returns on the
all-stock portfolios. A model is deemed appropriate for asset pricing when its betas in (1)
CE

correctly explain the cross-section averages of portfolios excess returns. Thus, in model
validation we focus on testing the null hypothesis that the intercepts in (2) are simultaneously
null and the risk premia in (1) are positive, with  remaining insignificant.
AC

We test whether intercepts are jointly null with the use of two tests: the GRS test and
the GMM-based procedure. The risk premia parameters in (1) are also estimated in the GMM
framework that is robust to both conditional heteroscedasticity and serial correlation because
this approach requires no assumption of the normality of returns. To implement the GMM
method, we use a two-step GMM procedure following Cochrane (2005) and Shanken and
Zhou (2007). The detailed description of this procedure is provided in Appendix 2.
3. Results
In this section, we first present the results of our examination of the anomalies in
emerging European markets, and describe the model that best explains the anomalies we
observe. Based on the sets of double-sorted portfolios, we confirm that models usefulness for
asset pricing in emerging European markets.
Cross-Sectional Patterns in Emerging European Market Returns
 ACCEPTED MANUSCRIPT

Table 3 summarizes the fundamental statistics of the 100 anomaly portfolios.

Profitable anomalies are concentrated in several key categories.
[Insert Table 3 here]
Within group 1 (value versus growth), all the strategies produce positive average

T
payoffs, with the means in 8 out of 10 cases significantly different from zero in both the

P
equal-weighting and capitalization-weighting approaches: EP (1), BM (2), CFP (3), SP (4),

RI
EBEV (5), EBP (6), BMCap (9), and BMGPA (10). All the mean monthly returns exceed 1%,
and in three cases (EP [1], EBP [7], and BMCap [9]) even reach 2%. These results are

SC
consistent with earlier studies of the value effect (proxied by the book-to-market ratio) in the
Eastern European region (cf., Cakici et al., 2013; Zaremba, 2015; Hanauer & Linhart, 2015).

NU
Within group 2 (dividends), only the equal-weighted zero-investment portfolio formed
on dividends (DY [11]) delivers a mean monthly return significantly different from zero. The
MA
return was calculated as 1.03% per month.
The following portfolios from group 3 (profitability) display positive means significantly
different from zero under both the equal-weighted and capitalization-weighted approach:
ED

ROA (14), ROE (15), ROIC (16), GPA (19), and SGIG (20). In group 4 (distress risk), only
two strategies show significant, positive means in the monthly returns: the change in the
PT

current ratio (CRCh [29)]) and the change in quick ratio (QRCh [31]). However, once the
portfolios have been capitalization-weighted, the anomalies linked to the change in leverage
CE

(i.e., LevCh (26) and LTDCha [32]) also displayed positive and significant means.
The results from group 5 (investment patterns) contradict the cross-sectional patterns
observed in developed markets with only one out of seven strategiesInvest [34]being
AC

profitable. The results from group 6 (issuance anomalies) closely follow their counterparts
from the developed markets, yet with the means being statistically significant only in the
equal-weighted approach. Three strategiesIPO [40], CEI [41], and age [43]generate
means significantly exceeding zero and ranging from 0.53% (IPO) to 0.77% (CEI) per month.
Group 7 (accrual anomalies) provides no evidence of profitability, as none of the
portfolios (OA [44], TA [45], POA [46], or PTA [47]) shows mean returns significantly
higher than zero. In fact, only one zero-investment portfolio displays a significant mean (the
equal-weighted portfolio based on sorts by inventory growthIng [51]).
 ACCEPTED MANUSCRIPT

The next three groups of anomalies (group 8: liquidity; group 9: low-volatility; group
10: extreme risk) show no resemblance to the cross-sectional patterns in the developed
markets, recording no significantly profitable anomalies.4
In group 11 (long-term reversal), all three strategies (LtRev36 [64], LtRev60 [65],

T
LtRevIvol [66]) are characterized by sizable and positive payoffs. Yet, as the returns on all the

P
portfolios are highly volatile, only one portfolio produces positive returns with a mean

RI
significantly higher than zero: the equal-weighted portfolio based on the idiosyncratic
volatility-enhanced long-term reversal, which yields an average of 1.79% per month with a

SC
standard deviation of 10.57%.
The results in group 12 (momentum) disappoint and, even though the momentum

NU
effect is both extremely powerful and pervasive (Asness, Moskowitz, & Pedersen, 2013), the
evidence for its presence in emerging European markets seems rather weak. None of the
MA
capitalization-weighted dollar-neutral portfolios delivers returns with means significantly
higher than zero, and the equal-weighted portfolios are equally dissatisfying. The classical
momentum strategies, e.g., 6-month momentum (67) or 12-month momentum (68), produce
ED

modest returns with insignificant means, whereas the four enhanced-momentum strategies
MomAge (70), MomTR (74), Mom52H (75), and MomCons (78)displayed significant
PT

payoffs with mean monthly returns ranging from 0.87% (Mom52H) to 1.68% (MomTR). This
poor profitability of the momentum effect is consistent with the earlier studies of, among
CE

others, Cakici et al. (2013), who investigated data for the years 19902011 and from similar
regions, only to find no evidence of the momentum anomaly.
Profitability in group 13 (technical analysis) visibly outperforms most of its
AC

counterparts. The portfolio constructed according to the 52-week high strategy52HA (88)
and four equal-weighted portfolios based on moving averagesMA200A (82), MA250A
(83), MA200B (84), and MA250B (85)see returns with significant means. Nevertheless, as
with the momentum effect, none of the capitalization-weighted portfolios produces any
significant payoffs.
We identify no significant means of returns in group 14 (seasonal effect). Neither the
seasonality momentum effect (90) nor the January effect (92) is confirmed in the emerging
European data.

4
The lack of significance of the liquidity-related patterns might be a result of strong restrictions imposed on
penny stocks in this study. In fact, an earlier study of Central-Eastern European markets by Zaremba (2014)
documented that portfolios from sorts on turnover ratio or bid-ask spread indeed display abnormal returns.
 ACCEPTED MANUSCRIPT

The patterns found in group 15 (analyst coverage) diverge from the cross-sectional
effects captured in the developed markets, with only one strategy producing returns with a
mean significantly different from zero: the zero-investment portfolio based on the change in
the forecasted EPS (96).

T
Group 16 (market frictions) gives rise to only two capitalization-weighted portfolios

P
with means significantly more than zero: one portfolio constructed according to the small-

RI
firm strategy (Cap [98]) and one constructed on the low-price effect strategy (LP [99]), with
the former being consistent with the results of Cakici et al. (2013).

SC
In conclusion, attempting to replicate the return-predictive signals from developed
markets delivered disappointing results. Even when we apply a not too restrictive minimum

NU
significance level of 10%, only 32 equal-weighted and 20 capitalization-weighted anomaly
portfolios display positive means of returns that differ significantly from zero. This may result
MA
from the relatively short time frame and the small number of stocks available in the cross
section of the emerging markets, as many of the replicated strategies delivered positive
returns, but with means that failed to reach statistical significance. On the other hand, a
ED

portion of the originally discovered anomalies may only stem from the extensive data
snooping, as the performance of many anomalies has frequently are disappointing in out-of-
PT

sample studies (cf., Dimson & Marsh, 1999; McLean & Pontiff, 2016).
For further tests, we include only the anomalies which produce positive means
CE

significantly different from zero at the 10% significance level, therefore continuing with 32
equal-weighted and 20 capitalization-weighted anomaly portfolios.5 The anomalies represent
multiple categories, although they are not completely independent. The selection includes
AC

both close variations of the same anomalies and enhanced versions of other anomalies,
however, the average pair-wise correlation of the returns on the equal-weighted
(capitalization-weighted) anomaly portfolios amounts to merely 0.10 (0.14), indicating a
diverse set of return patterns.6
Comparison of the Factor Pricing Models
The results of the examinations of the factor pricing models are summarized in Table
4. In addition to the aggregate GMM and GRS tests, we also summarize the intercepts from
separate multiple regressions applied to individual anomalies. In Panel A, that shows the

6
We report the pair-wise correlations in Tables A2 and A3 in Appendix 1.
 ACCEPTED MANUSCRIPT

results of the 20 capitalization-weighted portfolios, the first column on the left displays the
outcome of the CAPM model, which clearly shows no cross-sectional patterns of anomalies.
The average of the estimated intercepts equaled 1.36%, with 18 out of 20 anomalies showing
significant intercepts at the 10% significance level.7

T
[Insert Table 4 here]

P
The application of the FF3 model factors visibly improves the models ability to

RI
explain abnormal returns. The average of the estimated intercepts decreases to 1.03% (i.e., by
about 24% compared to the CAPM). At the 10% significance level, the model explains 8 out

SC
of 20 anomalies, leaving 12 anomalies with significant intercepts. Table A4 in Appendix 1
contains the alphas of various factor pricing models applied to explain capitalization-weighted

NU
anomaly portfolios. The main advantage of the FF3 model over the CAPM model is its
superior ability to explain both value-related anomalies and the effects related to market
MA
frictions. The model captures all significant abnormal returns on more than half of the value
vs. growth anomalies (EP, BM, SP, EBEV, and BMGPA), as well as on the small-cap (Cap
[98]) and low-price (LP [99)]) patterns. It is unable, however, to cover most anomalies related
ED

to profitability (group 3), distress risk (group 4), and investment (group 5).
Including the WML factor in the C4 model fails to significantly improve its
PT

explanatory power. The mean intercept of 0.98% is only marginally lower than the mean
intercept calculated for the three factors.
CE

The last column in Panel A, Table 3 presents the alpha estimates for the five factors
whose usefulness in explaining the anomaly portfolios clearly outperforms all the previous
models. The average alpha declines to 0.81 (i.e., by 17% compared with C4), leaving only
AC

eight anomalies unexplained at the 10% significance level.

For testing purposes, we adopt the GRS statistic and the GMM-based approach. For
the CAPM model, the p value obtained from the two tests fails to allow us to reject our null
hypothesis, whereas in the FF3 and C4 models both tests result in a rejection of the null
hypothesis at the 10% significance level. Eventually, only FF5 confirms our basic hypothesis,
sufficiently explains the cross-sectional patterns in all anomaly returns, and is superior to
other classical models (i.e., the CAPM, FF3, and C4).
Set against the capitalization-weighted portfolios, the performance of the 32 equal-
weighted portfolios (Panel B, Table 3) paints a very different picture. Undoubtedly, the equal-
weighted portfolios challenge the pricing models more than do the capitalization-weighted

7
We always report Newsy-West (1987) adjusted t-statistics.
 ACCEPTED MANUSCRIPT

portfolios, given the variation in the results. However, all the models factors fail to explain
the cross-sectional patterns. As in previous studies, the CAPM models factors explain the
smallest number of anomalies, detecting 31 alphas significantly different from zero with the
significance level at 10%. The other models explain merely four to seven anomalies, leaving

T
2427 alphas significantly different from zero at the 10% significance level. The average

P
intercept ranges from a low of 0.93% in the all-factor model to 1.17% in the FF3 model. The

RI
relative advantage of these other models in comparison to the CAPM model stems from their
superior ability to explain abnormal returns on the portfolios formed on dividends (DY [11]),

SC
inventory growth (Ing [51]), idiosyncratic volatility-enhanced long-term reversal (LtRevIVol
[66]), and partly the portfolios formed on anomalies related to group 5 (investment) and group

NU
6 (issuance). More details on the performance of individual equal-weighted anomaly
portfolios under various models are displayed in Table A5, Appendix 1. Interestingly, even
MA
models including the WML factor produce highly significant positive intercepts on all
anomalies stemming from momentum and technical analysis. Overall, the models all perform
poorly within the equal-weighted portfolios where both the GRS and the GMM tests
ED

hypotheses s are strongly rejected even with significance at 1%. In other words, no model
successfully captures the cross-sectional patterns in equal-weighted zero-investment
PT

portfolios.
There are two reasons why the returns on equal-weighted anomaly portfolios cannot
CE

be easily explained with factor pricing models. First, compared to the value-weighted
portfolios, the equal-weighted portfolios allocate significantly more capital to the smallest
stocks, a usual niche of inefficiencies and stock market anomalies. Thus, although the equal-
AC

weighted portfolio results frequently outperform their value-weighted counterparts, this

results from their disproportionately large positions in small stocks, which are typically
illiquid and expensive to trade or rebalance. For equal-weighted strategies, trading costs are,
on average, two to three times higher than for capitalization-weighted portfolios, which
usually turn out to be less profitable once adjusted for transaction costs (Novy-Marx &
Velikov, 2015). This might be particularly important in the emerging European markets
where trading costs are usually higher than in developed markets (cf., Zaremba & Konieczka,
2015).
Second, the abnormal performance of equal-weighted portfolios can result from the
diversification return (i.e., an incremental return earned by a rebalanced portfolio of assets;
Willenbrock, 2011), which as an additional source of profit is particularly strong in frequently
rebalanced portfolios composed of many volatile assets (Erb & Harvey, 2006). Thus, it can
 ACCEPTED MANUSCRIPT

particularly boost the performance of the equal-weighted arbitrage portfolios, reducing the
explanatory power of the standard factor pricing models with capitalization-weighted factors.
Evaluating the explanatory power of the factor pricing models over various return-
predictive signals in emerging European markets, we find the FF5 model best covers the

T
cross-sectional patterns in returns, outperforms other classical models and renders the

P
momentum factor redundant. Nonetheless, none of the models is able to deal with the

RI
abnormal performance of equal-weighted zero-investment portfolios.
Validation of the Five-Factor Pricing Model

SC
As the FF5 model demonstrates the highest explanatory power over many anomalies,
we progress to confirm the models usefulness for asset pricing in emerging European

NU
markets by examining the explanatory power over returns on the sets of 16 capitalization-
weighted portfolios from double sorts. 8 We adopt the characteristics underlying the
MA
construction of the factor portfolios in the FF5 model as sorting variables in the double-sorted
portfolios. In other words, we consider three sets of portfolios constructed from all stocks in
the emerging European markets based on the following pairs of variables: (a) size and B/M
ED

ratio, (b) size and asset growth, and (c) size and operating profitability.9
The performance of the portfolios from double sorts on combinations of size, B/M
PT

ratio, operating profitability, and asset growth is presented in Table 5. Starting with the
portfolios from sorts on size and B/M ratio (Panel A), the cross-sectional pattern in returns
CE

related to the B/M ratio stood out across all the size classes, while the variability remains the
lowest among large stocks. We detect no clear pattern related to size and the high B/M stocks
seem slightly riskier in terms of standard deviation.
AC

[Insert Table 5 here]

The profitability effect (Panel B) emerges across all the class sizes, thriving more
among the small stocks than the large ones. Interestingly, the least profitable companies
display the highest volatility.
Finally, the return patterns related to asset growth clearly underperform. The eyeball
test provides no clear regularities in the cross section of returns, regardless of the company
size.10

8
We concentrate solely on the capitalization-weighted portfolios as we have shown the factor pricing models
based on the capitalization-weighted factors are unsuitable to explain returns on equal-weighted portfolios.
9
For simplicity, we also refer to the stock market capitalization as size, B/M ratio as value, asset growth as
investment, and operating profitability as profitability.
10
The returns on the alternative double sorts are reported in Table A6 in Appendix 1.
 ACCEPTED MANUSCRIPT

We adopt the SUR model (2) for each of the three portfolio sets to explain the excess
returns. We verify the conjecture that MKT, SMB, HML, RMW, and CMA factors would
generate efficient portfolios and, thus, the vector of intercepts would equal zero. We apply
both the GRS and GMM tests to the sets of 16 double-sorted portfolios formed from all the

T
securities in our investment universe. The results confirm our initial expectations.

P
As presented in Table 6 which displays the results of the asset-pricing tests performed

RI
on the portfolios from double sorts, none of the portfolio sorts on the GRS or GMM tests
hypotheses are disqualified.

SC
[Insert Table 6 here]
Table 6 also includes the beta estimates of model (2) for portfolios sorted on: size and

NU
value, size and asset growth, and size and profitability.11 Panel A displays the portfolios from
sorts on B/M ratio and stock market capitalization. In line with our expectations, the small
MA
portfolios have the highest values of
parameters and the big portfolios have the lowest
values. Furthermore, the estimates of
make the cross-sectional patterns even more
evident than do the estimates of
. Within each size quartile,
 estimates increase
ED

monotonically from high B/M to low B/M stock portfolios. Combining this observation with
the findings in Table 4 leads us to assess the high B/M portfolios as riskier and more
PT

profitable than low B/M portfolios. Finally, also in line with our initial expectations, we
observe no regularity in the loadings of either the RMW factor or CMA factors for the size-
CE

value sorts.
Panel B of Table 6 displays the betas for the portfolios from sorts on stock market
capitalization and asset growth. Again, the value of
parameters is the highest for small
AC

portfolios and the lowest for big ones. Although these parameters do not decrease
monotonically, they indicate that small firms are riskier and outperform big companies.
As the estimates of
increase monotonically from aggressive to conservative stock
portfolios within each size quartile, conservative stock portfolios appear to be riskier and to
generate higher profits than aggressive ones. For portfolios sorted on size and asset growth,
estimates of loadings of HML factors or estimates of loadings of RMW factors display no
visible regularities.
The results for portfolios sorted on size and profitability are presented in Panel C of
Table 6. The estimates of
show a clear cross-sectional pattern:
 increases

11
As all are significantly positive at the 1% level and all the values approximate 1 (indicating that the
market portfolio fails to explain cross-sectional returns variability), we exclude the results from Table 6.
 ACCEPTED MANUSCRIPT

monotonically from robust to weak stock portfolios within each quartile. To be exact, in the
first three profitability quartiles all
 remain negative whereas in the last quartile almost
all the estimates of profitability loadings emerge as positive. Accordingly, the portfolios of
profitable stocks display the highest payoffs (Table 4), whereas the findings in portfolios

T
sorted on size and profitability resemble the previous analysis.12

P
For additional validation, we also calculate the risk premia rewarding investors for

RI
taking one unit of risk associated with a corresponding risk factor. To verify the risk premia
related to MKT, SMB, HML, RMW, and CMA factors, we consider model (1) and apply the

SC
GMM method.
In Table 7, we present the results for the cross-sectional regression based on model

NU
(1), the parameters of which are estimated using the GMM method described in the previous
section. The table summarizes the results of six sets of portfolios: the three basic sets (i.e.,
MA
size and value; size and asset growth; and size and profitability) and the portfolios from sorts
on value and asset growth; value and profitability; and asset growth and profitability.
[Insert Table 7 here]
ED

The parameter is insignificant for all sets of portfolios, which indicates that the
variation of average returns on portfolios depends solely on the variation of systematic risks
PT

related to the underlying factors. The parameters are significant for all sets of portfolios.
The monthly risk premium associated with size varies significantly from zero, ranging from
CE

0.3% for size-value portfolios to 1.0% for size-asset growth portfolios. The parameter
 also diverges significantly from zero, which confirms that the value effect exists in
emerging European markets. The risk premium associated with value approximates 1.5% per
AC

month. The loadings on RMW also represent an important cross-sectional determinant of

average returns for the set of portfolios sorted for example, on profitability. The risk premium
approximates 0.8% in size-profitability portfolios and 1.0% in asset growth-profitability
portfolios 13 whereas the risk premium related to the asset growth effect is altogether
insignificant.
When describing the goodness of fit for these models, we follow the cross-sectional
 measure by Jagannathan and Wang (1998), who measure unconditional deviation from the

12
To confirm the validity of the results, we analyze portfolios sorted on: (a) value and asset growth; (b) value
and profitability; and (c) asset growing and profitability. The outcomes display no major qualitative
inconsistencies (see Table A6, Appendix 1).
13
We indicate only statistically significant values.
 ACCEPTED MANUSCRIPT

model and show the fraction of cross-sectional variation in average returns explained by the
model. For example, for portfolios sorted on size and value,  reached 73.2%; for portfolios
sorted on size and asset growth, it scores 75.9%; and in portfolios sorted on size and
profitability, 69%. Such high values of the  statistics support the view that the FF5 model

T
fits well to the emerging European market.14

P
4. Concluding Remarks

RI
In this paper, we aimed to compare the explanatory power of the four popular factor
pricing modelsthe CAPM, the three-factor model by Fama and French (1993), the four-

SC
factor model by Carhart (1997), and the five-factor model by Fama and French (2015a)over
various cross-sectional patterns in the emerging European stock returns. We therefore

NU
identified, classified, and replicated 100 anomalies described in the financial literature. In our
study, only 20 (32) of the capitalization-weighted (equal-weighted) anomaly portfolios turn
MA
out to be significantly profitable in emerging European markets. Importantly, the five-factor
model explains the returns of all the capitalization-weighted and many of the equal-weighted
cross-sectional patterns, clearly outperforming the other models. Additional tests also confirm
ED

its usefulness in pricing assets in emerging European markets.

Further research could be pursued in several directions. First, a deeper investigation
PT

into the sources of the weak momentum performance in the emerging European markets could
yield interesting insights. Second, the standard factor pricing models largely disregard the
CE

effects of transaction costs, which may prove important in emerging markets. The research
could also be extended to include other asset-pricing models that capture alternative pricing
return patterns related to liquidity (Pastor & Stambaugh, 2003; Bekaert, Harvey, & Lundblad,
AC

2007; Wagner & Winter, 2013) or the low-risk effects (Frazzini & Pedersen, 2014). Although
the anomalies related to liquidity and low-risk effects are insignificant in our sample,
including these variables in the asset-pricing models might modify the findings and yield
some further interesting insights, particularly into the emerging markets.

References
Bali, T.G., Engle, R.F., & Murray, S. (2016). Empirical asset pricing: The cross section of
stock returns. Wiley, Hoboken.
Bekaert, G., & Harvey, R., (2002). Research in emerging markets finance: Looking to the
future. Emerging Markets Review, 3, 429-448.

14
We also present the cross-sectional regressions in Figure A3, Appendix 1.
 ACCEPTED MANUSCRIPT

Bekaert, G., Harvey, C.R., & Lundblad, C. (2007). Liquidity and expected returns: Lessons
from emerging markets. Review of Financial Studies, 20(6),
http://dx.doi.org/10.1093/rfs/hhm030
Blackburn, D.W., & Cakici, N. (2016). Overreaction and the cross-section of returns:

T
International evidence. Available at SSRN: http://ssrn.com/abstract=2800188.

P
Borys, M. (2011). Testing multi-factor asset pricing models in the Visegrad Countries. Czech

RI
Journal of Economics and Finance, 61, 118139.
Cakici, N. (2015). The five-factor Fama-French model: International evidence. Available at

SC
SSRN: http://dx.doi.org/10.2139/ssrn.2601662.
Cakici, N., Fabozzi, F.J., & Tan, S. (2013). Size, value, and momentum in emerging market

NU
stock returns. Emerging Markets Review, 16, 46-65.
Carhart, M.M. (1997). On persistence in mutual fund performance. Journal of Finance, 52,
MA
57-82.
Chan, L., Karceski, J., & Lakonishok, J. (1998). The risk and return from factors. Journal of
Financial and Quantitative Analysis, 33, 159188.
ED

Chiah, M., Chai, D., Zhong, A., & Li, S. (2016). A better model? An empirical investigation
of the Fama-French five-factor model in Australia. International Review of Finance,
PT

forthcoming.
Cochrane, J.H. (2005). Asset pricing. Revised edition. Princeton University Press, Princeton.
CE

Czapkiewicz, A. & Wojtowicz, T. (2014). The four-factor asset pricing model on the Polish
stock market. Economic Research-Ekonomska Istrazivanja, 27(1), 771-783.
Dimson, E. & Marsh, P. (1999). Murphy's law and market anomalies. Journal of Portfolio
AC

Management, 25(2), 53-69.

Erb, C.B. & Harvey, C.R. (2006). Strategic and tactical value of commodity futures.
Financial Analyst Journal, 62(2), 69-97.
Fama, E.F. & French, K.R. (1993). Common risk factors in the returns on stocks and bonds.
Journal of Financial Economics, 33, 3-56.
Fama, E.F. & French, K.R. (1996). Multifactor explanations of asset pricing anomalies.
Journal of Finance, 51(1), 55-84.
Fama, E.F. & French, K.R. (2012). Size, value, and momentum in international stock returns
Journal of Financial Economics, 105(3), 45-472.
Fama, E.F. & French, K.R. (2015a). A five-factor asset pricing model. Journal of Financial
Economics, 116(1), 1-22.
 ACCEPTED MANUSCRIPT

Fama, E.F. & French, K.R. (2015b). International tests of a five-factor asset pricing model.
Available at SSRN: http://dx.doi.org/10.2139/ssrn.2622782.
Foye, J., Mramor, D., & Pahor, M. (2013). A respecified Fama-French three-factor model for
the new European Union member states. Journal of International Financial

T
Management & Accounting, 24, 325.

P
Frazzini, A., & Pedersen, L.H. (2014). Betting against beta. Journal of Financial Economics,

RI
111(1), 1-25.
Gibbons, M., Ross, S., & Shanken, J. (1989). A test of the efficiency of a given portfolio.

SC
Econometrica, 57, 11211152.
Green, J., Hand, J.R.M., & Zhang, F. (2016). The characteristics that provide independent

NU
information about average U.S. monthly stock returns. Available at SSRN:
http://dx.doi.org/10.2139/ssrn.2262374.
MA
Hanauer, M.X. & Linhart, M. (2015). Size, value, and momentum in emerging market stock
returns: Integrated or segmented pricing? Asia-Pacific Journal of Financial Studies, 44,
175-214.
ED

Harvey, C.R., Liu, Y., & Zhu., H. (2015). ... and the cross section of expected returns. Review
of Financial Studies, forthcoming.
PT

Hou, K., Xue, C., & Zhang, L. (2014). Digesting anomalies: An investment approach. Review
of Financial Studies, forthcoming.
CE

Investment Technology Group (2015). Global cost review, Q3/2014. Available at:
http://www.itg.com/2015/02/05/itg-peer-analysis-global-cost-review-q3-2014/.
Jacobs, H. (2015). What explains the dynamics of 100 anomalies? Journal of Banking &
AC

Finance, 57, 65-86.

Jacobs, H. (2016). Market maturity and mispricing. Journal of Financial Economics, 122(2),
270-287.
Jagannathan, R. & Wang, Z. (1998). An asymptotic theory for estimating beta-pricing models
using cross-sectional regression. Journal of Finance, 57, 2337-2367.
Lesmond, D. A., Schill, M. J., & Zhou, C. (2004). The illusory nature of momentum profits.
Journal of Financial Economics, 71(2), 349380.
Li, P.M., Zhang, Q., Cai, C.X., & Keasey, K. (2016). Understanding asset pricing anomalies
across the globe: The role of newswatchers. Available at SSRN:
http://dx.doi.org/10.2139/ssrn.2839799.
MacKinlay, A.C., & Richardson, M.P. (1991). Using generalized method of moments to test
mean-variance efficiency. Journal of Finance. 46(2), 511-527.
 ACCEPTED MANUSCRIPT

McLean, D.R., & Pontiff, J. (2016). Does academic research destroy stock return
predictability? Journal of Finance, forthcoming.
Newey, W.K., West, K.D. (1987). A simple positive-definite heteroskedasticity and
autocorrelation consistent covariance matrix. Econometrica, 55, 703-708.

T
Novy-Marx, R., & Velikov, M. (2015). A taxonomy of anomalies and their trading costs.

P
Review of Financial Studies, 29(1), 104-147.

RI
Pastor, L., & Stambaugh, R.F. (2003). Liquidity risk and expected returns. Journal of
Political Economy, 111(3), 642-685.

SC
Rouwenhorst, K. G. (1999). Local return factors and turnover in emerging stock markets. The
Journal of Finance, 54(4), 14391464.

NU
Shanken, J., & Zhou G. (2007). Estimating and testing beta pricing models: alternative
methods and their performance in simulations. Journal of Financial Economics, 84, 40
MA
86.
Sharpe, W.F., (1964). Capital asset prices: A theory of market equilibrium under conditions of
risk. Journal of Finance, 19, 425-442.
ED

Stambaugh, R.F., Yu, J., & Yuan, Y. (2012). The short of it: Investor sentiment and
anomalies. Journal of Financial Economics, 104(2), 288-302.
PT

Wagner, N., & Winter, E. (2013). A new family of equity style indices and mutual fund
performance: Do liquidity and idiosyncratic risk matter? Journal of Empirical Finance,
CE

21, 69-85.
Waszczuk, A. (2013). A risk-based explanation of return patternsEvidence from the Polish
stock market. Emerging Markets Review, 15, 186-210.
AC

Waszczuk, A. (2014). Assembling international equity datasets review of studies on the

cross-section of returns. Procedia Economics and Finance, 15, 1603-1612.
Willenbrock, S. (2011). Diversification return, portfolio rebalancing, and the commodity
return puzzle. Financial Analyst Journal, 67 (4), 42-49.
Zaremba, A. & Konieczka, P. (2015). Are value, size and momentum premiums in cee
emerging markets only illusionary? Finance a vr-Czech Journal of Economics and
Finance, 65(1), 84-104.
Zaremba, A. (2014). Quality investing in CEE emerging markets. Business, Management and
Education, 12(2), 159-180.
Zaremba, A. (2015). Value, size, momentum, and unique role of microcaps in CEE market
stock returns. Eastern European Economics, 53(3), 221-241.
 ACCEPTED MANUSCRIPT

Tables

Table 1
Monthly Returns on the Asset-Pricing Factors
This table reports the summary statistics of monthly returns of the asset-pricing factors investigated in this study:

T
MKT (market risk), SMB (small minus big), HML (high minus low), WML (winners minus losers), RMW

P
(robust minus weak), and CMA (conservative minus aggressive). The means, standard deviations, returns in

RI
extreme months, and intercepts are expressed in percentage terms. Asterisks *, **, and *** indicate values
significantly different from zero at the 10%, 5%, and 1% levels, respectively. The values in brackets are t-

SC
statistics, and with significance at the 10% level signified by bolded text. Panel A displays basic return
characteristics, whereas Panel B shows Pearson pair-wise correlation coefficients.

NU
MKT SMB HML WML RMW CMA
Panel A: Basic statistics
Mean 0.85 0.34 1.23*** 0.15 0.73** -0.09
(1.31) (0.97) (2.78) (0.27) (1.99) (-0.24)
MA
Standard deviation 9.49 5.15 6.53 8.14 5.39 5.45
Skewness -0.09 -0.39 0.84 -0.27 1.32 -0.56
Kurtosis 3.74 2.04 4.22 3.47 8.52 5.15
Worst month -39.11 -24.10 -20.82 -30.48 -16.36 -28.98
ED

Best month 46.75 13.85 28.77 36.72 36.22 18.66

Panel B: Pair-wise correlation coefficients.
SMB -0.33***
PT

(-5.39)
HML 0.37*** -0.38***
(6.35) (-6.39)
WML -0.25*** 0.01 -0.35***
CE

(-3.92) (0.09) (-5.89)

RMW -0.21*** -0.11 -0.14** 0.03
(-3.19) (-1.56) (-2.07) (0.45)
AC

CMA -0.06 0.06 0.21*** -0.04 -0.25***

(-0.82) (0.94) (3.25) (-0.55) (-3.99)
 ACCEPTED MANUSCRIPT

Table 2

Anomalies Examined in the Study

This table provides detailed information on the anomalies examined in this study. No. is the running number
used to identify the anomalies in the text, and Abbr. is the symbol of an anomaly utilized in the study. The

T
detailed description of the anomalies along with the reference literature is provided in Table A1 in Appendix 1.

P
No. Abbr. Name No. Abbr. Name

RI
Group 1: Value vs growth
1 EP Earnings-to-price 6 SEV Sales-to-EV

SC
2 BM Book-to-market 7 EBP EBITDA-to-price
3 CFP Cash flow-to-price 8 SG Sales growth
4 SP Sales-to-price 9 BMCap Size-enhanced book-to-market ratio
Gross profitability-enhanced book-to-market
5 EBEV EBITDA-to-EV 10 BMGPA

NU
ratio
Group 2: Dividends
11 DY Dividend yield 13 DCh Change in absolute dividends
12 DYCh Change in dividend yield
MA
Group 3: Profitability
14 ROA Return on assets 20 SGIG Sales growth-to-inventory growth
15 ROE Return on equity 21 GMCh Change in gross margin
16 ROIC Return on invested capital 22 GMGSG Gross margin growth minus sales growth
ED

17 GM Gross margin 23 EarVol Earnings volatility

18 AT Asset turnover 24 CfVol Cash flow volatility
19 GPA Gross profitability
Group 4: Credit risk
PT

25 DM Leverage 29 CRCh Change in current ratio

26 LevCh Change in leverage 30 QR Quick ratio
27 CH Cash holdings 31 QRCh Change in quick ratio
CE

28 CR Current ratio 32 LTDCh Change in long-term debt

Group 5: Investment and intangibles.
33 AG Asset growth 37 CIA Capital investments
34 Invest Investment 38 I2Ch 2-year change in investments.
AC

35 IG Investment growth 39 OL Operating leverage

36 HR Hiring rate
Group 6: Issuance
40 IPO Initial public offerings 42 NSI Net stock issuance
41 CEI Composite equity issuance 43 Age Age
Group 7: Accruals
44 OA Operating accruals 49 NOAg Net operating assets growth
45 TA Total accruals 50 NOAc Net operating assets change
46 POA Percent operating accruals 51 InG Inventory growth
47 PTA Percent total accruals 52 InC Inventory change
Idiosyncratic volatility-enhanced
48 AcIvol
accruals
Group 8: Liquidity
53 Turn Turnover 55 TRV Turnover ratio variability
54 TR Turnover ratio 56 TurnV Turnover variability
Group 9: Low-volatility
57 Beta Beta 59 Ivol Idiosyncratic volatility
58 SD Volatilty
Group 10: Extreme and downside risk
 ACCEPTED MANUSCRIPT

60 DownVol Downside volatility 62 Skew Skewness

61 VaR Value at risk 63 Kurt Kurtosis
Group 11: Long-term reversal
Idiosyncratic volatility-enhanced long-term
64 LtRev36 Long-term reversal (36 months) 66 LtRevIvol
reversal
65 LtRev60 Long-term reversal
Group 12: Momentum

T
67 StMom Short-term momentum 75 Mom52H 52-week high-enhanced momentum

P
68 LtMom Long-term momentum 76 MomNeg Analyst coverage-enhanced momentum
69 IntMom Intermediate momentum 77 MomR2 R2-enhanced momentum

RI
70 MomAge Age-enhanced momentum 78 MomCons Return consistency-enhanced momentum
Idiosyncratic volatility-enhanced
71 MomIvol 79 RALtMom Risk-adjusted momentum
momentum

SC
72 MomSmall Size-enhanced momentum 80 TSMom Time series momentum
Book-to-market ratio-enhanced
73 MomBM 81 MomAcc Momentum acceleration
momentum

NU
74 MomTR Liquidity-enhanced momentum
Group 13: Technical analysis
82 MA200A 200-day moving average (absolute) 86 MA200Q 200-day moving average (ratio)
83 MA250A 250-day moving average (absolute) 87 MA250Q 250-day moving average (ratio)
MA
84 MA200B 200-day moving average (band) 88 52HA 52-week high (absolute)
85 MA250B 250-day moving average (band) 89 52HQ 52-week high (ratio)
Group 14: Seasonalities
90 SeasMom Seasonality momentum 91 OtherJan The other January effect
ED

Group 15: Analyst coverage

92 Neg Analyst coverage 95 FEPS Next-year forecasted EPS growth
93 CovCh Change in analyst coverage 96 FEPSCh Change in EPS forecast
94 LEPSF Long-term forecasted EPS growth
PT

Group 16: Market frictions

97 StRev Short-term reversal 99 LP Price
98 Cap Total market capitalization 100 Spread Bid-ask spread
CE
AC
 ACCEPTED MANUSCRIPT

Table 3

Monthly Returns on the Anomaly Portfolios

This table reports the mean monthly returns on the equal-weighted and capitalization-weighted zero-investment
portfolios based on stock market anomalies. No. is the running number used to identify the anomalies in the text,

T
and Abbr. is the symbol of an anomaly used in the study. R is the mean monthly return, Vol is the standard

P
deviation of the monthly returns, Skew is skewness, Kurt is kurtosis, and N is the number of observations. The

RI
mean significance is expressed in percentage terms. Asterisks *, **, and *** indicate values that are positive and
significantly different from zero at the 10%, 5%, and 1% levels, respectively. The numbers in brackets are t-

SC
statistics, and the values simultaneously consistent with the economic expectations and significantly different
from zero at the 10% significance level are signified by bolded text. The symbols of the strategies are explained
in Table A1 in Appendix 1.

No. Abbr.
Panel A: Equal-weighted portfolios
R t-stat Vol Skew Kurt NU Panel B: Capitalization-weighted portfolios
R t-stat Vol Skew Kurt
N
MA
Group 1: Value vs growth
1EP 2.00*** 3.97 7.40 0.91 3.02 2.05*** 3.15 9.54 0.99 3.42 216
2BM 1.70*** 3.62 6.87 1.04 4.92 1.17** 2.11 8.15 0.39 3.07 216
3CFP 1.27*** 2.86 6.48 0.35 2.82 1.64*** 2.86 8.40 1.10 2.82 216
ED

4SP 1.18*** 2.63 6.60 0.77 3.38 1.41** 2.43 8.53 0.80 3.59 216
5EBEV 1.55*** 3.75 6.05 -0.65 4.07 1.13* 1.82 9.06 -0.16 1.95 216
6SEV 0.56 1.24 6.66 -0.33 6.01 0.54 1.04 7.63 -0.40 3.10 216
7EBP 2.00*** 4.46 6.57 0.38 2.50 1.72*** 2.83 8.92 0.83 1.69 216
PT

8SG 0.58 0.79 10.70 -0.37 17.36 0.54 0.61 12.69 -0.06 8.75 211
9BMCap 2.50*** 3.75 9.77 0.57 2.56 2.34*** 3.30 10.39 0.47 3.51 216
10BMGPA 1.78*** 4.03 6.49 0.32 2.41 1.70*** 3.09 8.07 0.27 3.19 216
CE

Group 2: Dividends
11DY 1.03*** 2.90 5.22 0.82 2.74 0.53 1.04 7.47 0.02 2.43 216
12DYCh 0.66 1.60 6.03 0.08 6.23 0.86 1.29 9.81 -0.09 6.96 216
13DCh 0.20 0.66 4.37 0.44 3.60 0.44 0.68 9.42 0.91 5.26 216
AC

Group 3: Profitability
14ROA 1.14*** 3.12 5.37 0.57 5.82 1.19* 1.95 8.98 0.81 5.55 216
15ROE 1.37*** 4.12 4.86 0.76 4.14 1.32** 2.49 7.75 0.40 2.10 216
16ROIC 0.68** 2.09 4.74 -0.33 1.50 1.28* 1.70 10.98 2.85 22.65 216
17GM 0.72 1.10 9.58 0.56 5.08 0.46 0.68 9.95 0.82 3.41 216
18AT 0.38 1.20 4.59 0.41 7.20 0.16 0.32 7.33 -1.03 10.64 216
19GPA 0.59* 1.95 4.46 -0.64 2.27 0.99* 1.81 7.99 0.25 4.37 216
20SGIG 0.89*** 3.30 3.97 0.21 5.90 1.56*** 2.87 7.98 0.52 1.09 216
21GMCh 0.72 1.11 9.53 -0.15 6.03 0.82 1.01 11.89 0.31 7.24 216
22GMGSG -0.52 -1.02 7.56 -0.71 6.87 -1.36 -1.95 10.20 -0.83 4.11 216
23EarVol -0.27 -0.55 7.11 0.32 4.27 -0.21 -0.28 10.93 0.21 2.56 216
24CFVol -0.13 -0.29 6.38 -0.23 3.95 -1.89 -3.00 9.22 -1.88 10.83 216
Group 4: Distress risk
25DM -0.33 -0.79 6.01 -0.67 3.40 -0.09 -0.14 9.28 -1.04 3.25 216
26LevCh 0.47 1.57 4.41 0.33 2.83 0.82* 1.83 6.56 0.37 1.09 216
27CH 0.13 0.47 4.03 -0.06 1.28 -0.13 -0.26 7.15 -1.39 9.66 216
28CR -0.05 -0.17 4.07 -0.52 2.38 0.53 1.11 7.04 1.19 7.34 216
29CRCh 0.92** 2.50 5.37 0.67 13.01 1.29** 2.48 7.63 0.91 7.24 216
30QR 0.00 -0.01 4.19 -0.11 1.39 -0.03 -0.07 6.42 0.27 3.56 216
 ACCEPTED MANUSCRIPT

31QRCh 0.80** 2.49 4.73 0.49 9.05 2.12*** 3.91 7.95 1.89 10.47 216
32LTDCh 0.32 1.36 3.41 0.27 2.03 1.07** 2.29 6.88 0.06 3.58 216
Group 5: Investment
33AG 0.13 0.35 5.48 -0.10 3.09 -0.01 -0.02 8.18 -0.66 5.19 216
34Invest 0.97** 2.24 6.35 0.95 5.42 1.76*** 2.68 9.61 0.95 3.71 216
35IG -0.06 -0.17 5.38 -0.61 6.90 0.01 0.02 10.32 1.39 10.88 216
36HR -0.43 -1.24 5.06 -0.73 7.33 -1.31 -2.24 8.59 -1.65 7.31 216

T
37CIA 0.03 0.09 4.48 -0.23 3.22 0.86 1.31 9.61 2.01 15.67 216

P
38I2Ch 0.20 0.48 5.96 0.31 5.61 -1.43 -1.99 10.51 -1.13 10.14 216
39OL -0.23 -0.60 5.74 -0.21 3.39 -0.57 -1.13 7.45 -0.84 4.89 216

RI
Group 6: Issuance
40IPO 0.53* 1.89 4.14 0.78 2.91 0.40 0.86 6.90 0.98 5.11 216

SC
41CEI 0.77** 2.30 4.59 0.55 3.37 0.50 0.99 6.93 -0.29 0.11 190
42NSI 0.90 1.51 8.70 0.67 4.48 0.53 0.72 10.77 0.74 5.52 216
43Age 0.69** 1.99 5.11 0.19 0.25 1.17 1.65 10.41 1.19 4.18 216
Group 7: Accruals

NU
44OA 0.09 0.19 6.94 0.03 5.60 -0.57 -0.90 9.25 -0.20 2.57 216
45TA -0.65 -1.53 6.23 0.34 5.18 -0.51 -0.78 9.62 -0.41 3.03 216
46POA 0.01 0.03 5.62 -0.38 4.62 -0.73 -1.18 9.06 -1.59 10.91 216
47PTA -0.05 -0.16 4.64 1.22 6.45 0.55 1.13 7.16 0.99 3.88 216
MA
48AcIvol 0.51 0.41 17.97 -0.49 12.00 0.84 0.64 19.28 -0.59 9.48 216
49NOAg 0.27 0.71 5.56 0.10 2.87 0.20 0.35 8.15 0.21 2.41 216
50NOAc 0.39 1.41 4.11 0.29 3.37 0.05 0.13 6.05 -0.77 3.68 216
51InG 0.57** 1.97 4.22 0.16 1.84 0.05 0.10 6.73 0.52 0.82 216
ED

52InC 0.39 1.41 4.11 0.29 3.37 0.05 0.13 6.05 -0.77 3.68 216
Group 8: Liquidity
53Turn 0.69 1.26 7.50 0.15 2.52 0.64 1.00 8.69 0.88 2.38 188
54TR 0.11 0.17 8.59 0.11 1.50 -0.23 -0.36 8.72 0.11 1.13 187
PT

55TRV 0.63 1.13 7.51 -0.17 0.20 0.06 0.10 8.60 -0.06 1.32 181
56TurnV 0.69 1.41 6.56 0.06 0.41 0.57 0.94 8.12 1.09 4.19 181
Group 9: Low-volatility
CE

57Beta -0.41 -0.90 6.72 -0.30 4.06 -0.40 -0.59 9.98 -0.10 2.24 216
58SD -0.15 -0.21 10.88 -0.11 3.88 -0.39 -0.53 10.82 -0.95 4.37 216
59Ivol -0.33 -0.50 9.63 -0.40 3.48 -0.57 -0.78 10.69 -1.96 8.40 216
Group 10: Extreme risk
AC

60DownVol 0.44 0.60 10.70 0.08 2.78 0.87 1.15 11.15 0.49 2.22 216
61VaR 0.15 0.23 9.66 0.01 4.00 0.25 0.29 12.46 1.67 6.29 216
62Skew 0.39 1.05 5.41 0.06 2.66 -0.70 -0.96 10.64 -2.10 7.74 216
63Kurt -0.35 -1.42 3.65 -0.42 1.90 -0.62 -1.08 8.37 -1.90 18.90 216
Group 11: Long-term reversal
64LtRev36 0.52 0.96 7.87 1.23 4.21 1.66 1.54 15.79 6.24 61.37 216
65LtRev60 0.78 1.43 7.98 -1.11 8.95 1.82 1.62 16.54 5.79 57.89 216
66LtRevIVol 1.79** 2.48 10.57 1.06 11.88 2.08 1.60 19.06 3.73 45.88 216
Group 12: Momentum
67StMom 0.25 0.42 8.74 -1.20 4.92 -1.06 -1.23 12.65 -1.42 4.19 216
68LtMom 0.52 0.83 9.07 -0.24 5.34 -0.55 -0.62 12.87 -0.80 3.64 216
69IntMom 0.64 1.15 8.12 -0.15 5.27 -0.08 -0.10 11.97 -1.11 6.16 216
70MomAge 1.26** 1.98 9.36 0.20 5.68 0.88 0.91 14.23 0.07 2.84 216
71MomIvol 0.29 0.49 8.60 -0.29 2.12 -0.19 -0.21 13.58 -1.04 6.36 216
72MomSmall 0.49 0.69 10.39 -0.63 5.35 0.69 0.96 10.51 -0.63 4.85 216
73MomBM 1.09 1.52 10.53 0.02 5.31 -0.67 -0.76 12.93 -0.06 3.04 216
74MomTR 1.68*** 3.21 7.12 -0.41 2.84 0.68 0.90 10.29 -0.51 3.59 187
75Mom52H 0.87** 1.99 6.39 0.08 1.78 0.19 0.28 9.90 -1.02 5.88 216
76MomNeg 0.70 1.11 9.29 -0.13 4.13 -0.32 -0.34 13.59 -1.36 8.52 216
 ACCEPTED MANUSCRIPT

77MomR2 0.07 0.12 9.08 -0.28 3.97 -2.00 -2.00 14.65 -1.38 7.01 216
78MomCons 1.34* 1.82 10.79 -0.87 4.12 0.25 0.23 15.35 -1.12 5.56 216
79RALtMom 0.81 1.53 7.76 -0.67 3.79 0.27 0.34 11.48 -0.73 3.84 216
80TSMom 0.43 1.15 5.45 -0.69 3.97 -0.58 -0.84 10.03 -1.34 8.79 216
81MomAcc -0.06 -0.11 7.31 -0.70 4.31 -0.23 -0.33 10.23 -0.48 3.29 216
Group 13: Technical analysis
82MA200A 0.65* 1.78 5.34 -0.31 3.97 0.04 0.09 7.55 -0.90 4.89 216

T
83MA250A 0.71** 1.98 5.27 -0.20 3.64 0.17 0.33 7.70 -0.68 5.04 216

P
84MA200B 1.71** 2.44 10.29 -0.14 2.67 0.37 0.35 15.73 0.11 3.95 216
85MA250B 1.54** 2.32 9.70 -0.25 2.89 -0.07 -0.07 14.90 0.07 4.48 216

RI
86MA200Q 0.45 0.71 9.34 -0.78 4.54 -0.68 -0.80 12.45 -0.99 3.10 216
87MA250Q 0.53 0.83 9.41 -0.49 4.67 -0.69 -0.80 12.74 -0.68 3.48 216

SC
8852HA 0.93** 2.09 6.52 -0.69 3.95 0.04 0.07 8.35 -0.11 2.88 216
8952HQ 0.80 1.22 9.64 -0.58 4.64 -0.54 -0.64 12.25 -0.96 2.74 216
Group 14: Seasonal effects
90SeasMom -0.21 -0.53 5.85 -0.16 3.28 -0.12 -0.17 10.04 -1.15 8.48 216

NU
91OtherJan -0.04 -0.09 5.89 -1.17 4.72 -0.37 -0.55 9.81 -0.74 4.99 216
Group 15: Analyst coverage
92Neg -0.15 -0.56 3.89 0.08 1.77 0.02 0.03 8.68 2.32 10.74 216
93CovCh -0.27 -0.69 5.79 -0.33 17.24 -0.17 -0.32 7.72 1.11 12.41 216
MA
94LEPSF -0.98 -2.00 5.40 -0.17 0.19 -0.50 -0.86 6.44 0.33 0.03 123
95FEPS 1.46 1.44 11.13 1.15 3.15 1.17 0.96 13.46 0.58 2.45 122
96FEPSCh 2.53** 2.17 10.21 0.58 4.03 1.96 1.52 11.35 0.09 3.66 78
Group 16: Market frictions
ED

97StRev -0.64 -1.20 7.84 0.50 4.94 0.84 1.12 11.08 0.19 2.52 216
98Cap 0.37 1.26 4.30 0.13 0.44 0.74* 1.67 6.49 1.83 14.83 216
99LP 0.37 0.45 12.26 0.91 5.92 1.90* 1.87 14.95 2.55 11.98 216
100Spread 0.88 0.83 15.25 5.12 59.36 0.47 0.41 16.62 3.78 42.57 208
PT
CE
AC
 ACCEPTED MANUSCRIPT

Table 4
Summary of the Factor Pricing Models Ability to Equity Anomalies
This table reports the explanatory power over the anomalies of various factor models examined in this study:
CAPM: the Capital Asset-Pricing Model (Sharpe, 1964); FF3: the three-factor model (Fama & French, 1993);
C4: the four-factor model (Carhart, 1997); and FF5: the five-factor model (Fama & French, 2015a). The

T
description of the models is presented in the Methods section. The models are examined within the anomaly

P
portfolios that produced positive means of monthly returns significantly different from zero at the 10%

RI
significance level (i.e., there are 20 capitalization-weighted anomaly portfolios displayed in Table A2 of
Appendix 1 [Panel A] and 32 equal-weighted anomaly portfolios displayed in Table A3 of Appendix 1 [Panel

SC
B]). The average absolute intercepts and their standard deviations are expressed as a percentage. Intercepts sign.
diff. from zero is the number of intercepts significantly different from zero at 10%, 5%, and 1% significance
levels. T-statistics are corrected according to Newey and West (1987). GRS is the F-statistic from the GRS test

NU
of Gibbons et al. (1989), and GMM is the statistic based on the GMM method estimated according to the
procedure described in the Methods. Numbers in brackets are corresponding p-values expressed as a percentage,
MA
and significance at the 10% level is signified by bolded text. Asterisks *, **, and *** indicate values
significantly different from zero at the 10%, 5%, and 1% levels, respectively.

CAPM FF3 C4 FF5

ED

Panel A: Capitalization-weighted portfolios

Average intercept 1.36 1.03 0.98 0.81
Standard deviation of intercepts 0.35 0.58 0.54 0.50
Average t-statistic 2.25 1.78 1.72 1.49
PT

Standard deviation of t-statistics 0.50 0.99 0.96 0.92

Intercepts sign. diff. from 0 at 10% 18 12 13 8
Intercepts sign. diff. from 0 at 5% 14 9 8 7
CE

Intercepts sign. diff. from 0 at 1% 6 4 2 2

GRS 2.37*** 1.83** 1.71** 1.56*
(0.001) (0.02) (0.034) (0.066)
AC

GMM 43.79*** 29.55* 29.61* 27.56

(0.002) (0.078) (0.076) (0.12)
Panel B: Equal-weighted portfolios
Average intercept 1.23 1.17 0.98 1.16
Standard deviation of intercepts 0.53 0.71 0.59 0.70
Average t-statistic 2.61 2.63 2.38 2.65
Standard deviation of t-statistics 0.62 1.06 1.01 0.92
Intercepts sign. diff. from 0 at 10% 31 25 24 27
Intercepts sign. diff. from 0 at 5% 28 24 24 24
Intercepts sign. diff. from 0 at 1% 11 16 15 17
GRS 3.16*** 3.27*** 3.48*** 3.10***
(0.000) (0.000) (0.000) (0.000)
GMM 110.61*** 126.27*** 134.77*** 127.02***
(0.000) (0.000) (0.000) (0.000)

29
 ACCEPTED MANUSCRIPT

Table 5

Averages and Standard Deviations of the Portfolios from Double Sorts on Total Stock Market
Capitalization, Book-to-Market Ratio, Past Return, Operating Profitability, and Asset
Growth.

P T
This table reports the means and standard deviations of excess returns on the sets of 16 capitalization-weighted
portfolios formed from double sorts on pairs of following variables: book-to-market ratio, market value (total

RI
stock market capitalization), operating profitability, and asset growth. The variables are described in detail in the

SC
Methods section. The means and standard deviations are expressed as a percentage.

Average Standard deviation

Low 2 3 High Low 2 3 High

NU
Panel A: Book-to-market ratio
Low -0.32 0.15 1.14 1.84 9.75 9.68 9.99 11.56
Market

2 0.77 0.85 1.14 1.48 10.19 10.04 11.63 11.69

value

MA
3 -0.25 0.58 1.65 1.95 9.60 9.72 10.52 12.62
High 0.18 0.29 0.71 1.46 9.67 11.83 12.72 12.58
Panel B: Operating profitability
Low 0.27 0.80 1.46 1.58 11.36 9.69 10.32 10.67
Market

2 0.65 0.46 1.26 1.47 12.18 10.70 10.05 10.01

ED
value

3 0.43 0.78 0.56 0.96 11.58 9.79 8.72 10.66

High 0.67 0.29 -0.05 1.32 12.44 10.31 10.85 11.48
Panel C: Asset growth
PT

Low 1.19 0.94 1.58 0.84 10.31 10.61 9.87 11.56

Market

2 1.20 1.65 1.09 1.08 10.84 11.19 10.09 11.07

value

3 0.24 0.76 1.11 0.96 10.06 10.22 9.82 11.64

CE

High 0.87 0.22 0.83 0.92 11.06 10.39 10.75 11.49

AC

30
 ACCEPTED MANUSCRIPT

Table 6

Estimation of the Regression Coefficients of the Five-Factor Model

This table reports estimates of regression coefficients of equation (6) as follows:

= + + + + + + ,

T
where is a vector of excess returns of portfolios constructed as independent double sorts on combinations of

P
capitalization, book-to-market ratio, asset growth, and operating profitability. GRS is the F-statistic from the

RI
GRS test of Gibbons et al. (1989), and GMM is the statistic based on the GMM method estimated according to a
procedure described in the Methods section. Asterisks *, **, and *** indicate values significantly different from

SC
zero at the 10%, 5%, and 1% levels, respectively. The 10% significance level is signified by bolded text.
Small 2 3 Big Small 2 3 Big
Market capitalization

NU

Panel A: Book-to-market ratio

Low 0.80*** 0.96*** 0.41*** -0.30 -0.20 -0.34 -0.19 -0.38

2 0.75*** 0.89*** 0.28*** -0.29 -0.16 -0.01 0.18 0.04
MA
3 0.81*** 0.73*** 0.39*** -0.26 0.20** 0.35*** 0.34*** 0.37***
High 0.59*** 0.73*** 0.52*** -0.08 0.46*** 0.66*** 0.71*** 0.68***

Low -0.21 -0.30 -0.26 -0.23 0.07 -0.05 0.11 0.08
ED

2 -0.30 -0.11 -0.29 -0.40 0.04 -0.03 0.18* 0.05

3 -0.23 -0.33 -0.37 -0.20 0.14** -0.01 0.11 0.03
High -0.21 -0.33 -0.29 -0.19 0.01 0.13 0.16 -0.06
PT

GRS=1.12 p-value=0.29 GMM=21.05 p-value=0.27

Low 0.93*** 0.99*** 0.41*** 0.06 0.22** 0.21** 0.13 0.26***
CE

2 0.81*** 1.03*** 0.58*** 0.07 0.35*** 0.31*** 0.27** -0.10

Panel B: Asset growth

3 0.90*** 0.74*** 0.56*** 0.03 0.25** 0.30*** 0.21*** 0.32***

High 0.98*** 0.84*** 0.60*** 0.30*** 0.45*** 0.15** 0.22*** 0.12
AC

Low -0.14 -0.25 -0.36 0.04 0.43*** 0.51*** 0.45*** 0.76***
2 -0.31 -0.16 -0.30 -0.13 -0.10 0.24*** 0.14* 0.38***
3 -0.11 -0.21 -0.20 -0.17 -0.15 0.04 -0.02 0.07
High -0.17 -0.29 -0.33 0.09 -0.28 -0.33 -0.20 -0.53
GRS=1.18 p-value=0.28 GMM=26.42 p-value=0.10

Panel C: Operating profitability

Low 0.92*** 0.93*** 0.60*** 0.13 0.08 -0.01 0.15 0.21

2 0.79*** 0.72*** 0.45*** 0.15 0.13 0.17 0.05 -0.20
3 0.99*** 0.83*** 0.42*** -0.13 -0.14 0.18*** -0.03 0.10
High 0.91*** 0.67*** 0.58*** 0.39*** 0.03 0.00 -0.07 0.30***

Low -0.52 -0.64 -0.59 -0.96 0.19 0.23* 0.22** -0.10
2 -0.27 -0.32 -0.37 -0.58 -0.14 0.21 0.16 0.29**
3 -0.35 -0.25 -0.31 -0.10 -0.06 0.01 0.06 0.15
High -0.11 -0.03 -0.28 0.70*** 0.31* 0.30*** 0.29** -0.21
GRS=1.34 p-value=0.18 GMM=23.71 p-value=0.17

31
 ACCEPTED MANUSCRIPT

Table 7
Estimation of the Risk Premia in the Five-Factor Model
This table reports the risk premia in the five-factor model expressed by equation (6) as follows:
( ) = + + + + + ,
where is a vector of excess returns of portfolios constructed as independent double sorts on combinations of

T
capitalization, book-to-market ratio, asset growth, and operating profitability. The risk premia are expressed as a

P
percentage. Asterisks *, **, and *** indicate values significantly different from zero at the 10%, 5%, and 1%

RI
levels, respectively. The numbers in brackets are test statistics, and significance at the 10% significance level is
signified by bolded text.

SC

Market capitalization & BM ratio 2.6 -2.2 0.3** 1.5*** -0.6 -3.3** 72.3
(1.54) (-1.18) (2.14) (11.65) (-0.48) (-2.06)

NU
Market capitalization & asset growth -0.1 0.6 1.0*** 0.5 0.6 0.0 75.9
(-0.08) (0.47) (5.33) (0.84) (0.84) (0.50)
Market capitalization & profitability 1.1 0.0 0.8*** -1.1 0.8*** -1.1* 69.0
MA
(0.85) (0.00) (3.14) (-1.40) (7.80) (-1.77)
BM ratio & asset growth 0.6 -0.4 2.0 1.4** 0.1 -0.3 55.8
(0.39) (-0.24) (1.48) (2.29) (0.15) (-0.72)
BM ratio & profitability -0.8 1.9 0.3 1.1* 0.5 -0.4 76.2
ED

(-0.59) (1.15) (0.26) (1.73) (1.19) (-0.59)

Asset growth & profitability 1.3 -2.6 0.3 0.1 1.0*** -0.2 47.4
(1.42) (-1.63) (0.30) (0.13) (3.98) (-0.68)
PT
CE
AC

32
 ACCEPTED MANUSCRIPT

Highlights
We compare the explanatory power of four popular factor pricing models in emerging
European markets.
We identify, classify, and replicate 100 anomalies from the finance literature.
Only a fraction of the anomaly portfolios are profitable.
The five-factor model vividly outperforms all the other models.

T
The five-factor model satisfactorily explains the returns on anomalies.

P
RI
SC
NU
MA
ED
PT
CE
AC

33