(Ab)Normal Returns in an Emerging Stock Market: International Investor Perspective

ABSTRACT

This article studies the comparative attractiveness of public equity investments in the Polish (emerging) and in the U.S. (advanced) stock markets in the years 2000–2013. Through an original implementation strategy based on several one- and multifactor asset pricing models (APMs), we find that the potential for “beating the market” in the form of abnormal profits is higher in the Polish stock market, specifically related to size and profitability anomalies. The Fama–French five-factor model fares best in an international setting and yields additional monthly abnormal returns of 0.19 pp. An international investor should apply local, rather than global, risk factors to properly assess relative abnormal investment opportunities between markets.

KEY WORDS:

• abnormal returns
• advanced market
• asset pricing model
• emerging market
• international portfolio diversification

JEL CLASSIFICATION CODES:

• G12
• G14
• G15

Acknowledgments

We gratefully acknowledge the guidance and many helpful comments from William Fuchs, Nicolae Garlenau, Emir Hrnjić, Robert Jalali, Dmitry Livdan, Terrance Odean, Marcin Owczarczuk, Panos Patatoukas, David Sraer, visiting scholar seminar participants at the Haas School of Business, UC Berkeley, seminar participants at the Craig School of Business, CS Fresno, and conference participants at the World Finance & Banking Symposium, Bangkok. We thank the editor and three anonymous referees who helped us improve the article.

Notes

1. An informational efficiency in asset pricing itself is untestable (Fama 1970). Noteworthy, the joint hypothesis problem has not been a problem in validating conclusions in empirical asset pricing studies (Binsbergen and Opp, 2017).

2. We narrow our study to the single class of multifactor models that are known for their empirical ability to describe stock returns. We do not look for perfect specification of APT or inter-temporal pricing models. Accordingly, we do not propose any factors that would be priced in Poland but not in the U.S.

3. Interpreting Polish risk premia, we acknowledge that not only is the time series short but also the market is relatively new. Risk premia are not stable in time (Ferson and Harvey, 1991; Welch and Goyal, 2008). Rather, they tend to decrease, especially after the introduction of market indexes that allow for easier portfolio diversification (less transaction costs and lower risk, among others). International investors tend to invest in index funds in emerging markets. Thus, we expect that a growing engagement of foreign investors in the WSE will reduce risk premia over time.

4. Phrase borrowed from the following crowd-sourced content service for financial markets: seekingalpha.com.

5. We test the statistical significance of the models’ intercept α using the GRS test introduced by Gibbons, Ross, and Shanken (1989). The null hypothesis of the joint insignificance of the alphas, which allows us to examine the models’ explanatory power, is

$\mathrm{\forall }i\alpha i=0$

The GRS test is illustrated by the following formula:

$\left(\frac{T}{N}\right)\left(\frac{T-N-L}{T-L-1}\right)\left[\frac{{\hat{\alpha }}^{\prime }{\hat{\mathrm{\Sigma }}}^{-1}\hat{\alpha }}{1+{\bar{\mu }}^{\prime }{\hat{\mathrm{\Omega }}}^{-1}\bar{\mu }}\right]\sim F\left(N,T-N-L\right)$

where$T$ is the number of periods,$N$ is the number of assets (portfolios),$L$ is the number of factors,$\hat{\alpha }$ is an estimated intercept vector,$\hat{\mathrm{\Omega }}$ is an estimate of the factor covariance matrix,$\mu$ is a vector of the sample factor mean returns, and$\hat{\mathrm{\Sigma }}$ is an estimate of the covariance of residuals. The test uses the F distribution with$N$ and $T-N-L$ degrees of freedom. Apart from the F statistic, we provide the corresponding p-values. The entire set of models’ summary statistics allows us to gauge the robustness of their performance. The $\left(\frac{T-N-L}{T-L-1}\right)$ factor represents the adjustment that is preferred using the small sample.

6 Following Lewellen et al. (2010), we report that the Sharpe ratio is calculated as

$SR\left(\alpha \right)={\left({\hat{\alpha }}^{\prime }{\hat{\mathrm{\Sigma }}}^{-1}\hat{\alpha }\right)}^{1/2}.$