Explaining Equity Anomalies in Frontier Markets: A Horserace of Factor Pricing Models

ABSTRACT

We are the first to compare the explanatory power of the major empirical asset pricing models over equity anomalies in the frontier markets. We replicate over 160 stock market anomalies in 23 frontier countries for years 1996–2017 and evaluate their performance with the factor models. The Carhart’s four-factor model outperforms both the recent Fama and French five-factor model and the q-model by Hou, Xue, and Zhan. Its superiority is driven by the ability to explain the momentum-related anomalies. Inclusion of additional profitability and investment factors lead to no further major improvement in the performance. Nonetheless, none of the models is able to fully explain the abnormal returns on all of the anomaly portfolios.

KEY WORDS:

• empirical asset pricing
• equity anomalies
• factor models
• frontier stock markets
• the cross-section of returns

JEL:

• G11
• G12
• G15

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website

Acknowledgments

This paper is a part of the project no. 2016/23/B/HS4/00731 of the National Science Centre of Poland.

Notes

1. The SEC definition is available at https://www.sec.gov/reportspubs/investor-publications/investorpubsmicrocapstockhtm.html.

2. Whenever our calculations rely on the accounting data, we utilize lagged values from month t-5 to avoid the look-ahead bias.

3. We thank prof. Kenneth R. French for making this data available on his website: http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html .

4. For robustness, we also compute the factor returns using the original methodology of Fama and French (1993), who split the companies into large and small by their median capitalization. The examination yields qualitatively consistent results. However, we do not use this framework as our default approach. Frontier markets encompass a large quantity of tiny firms, so the companies with below-median capitalization account for less than 1% of total market capitalization in certain years. In consequence, not only the practical implementation of the trading strategies within these smallest firms may encounter significant obstacles, but also the economic importance of the results is very limited.

5. Importantly, when computing the international factor returns, we rely on breakpoints from the pooled international sample as in Cakici, Fabozzi, and Tan (2013), Blackburn and Cakici (2017), or Zaremba et al. (2018), rather than estimating local factors based on country-level breakpoints, and then weighting them according to, e.g., stock capitalization as in Frazzini and Pedersen Asness, Frazzini, and Pedersen (2017) or Frazzini and Pedersen (2014). The reason is that contrary to developed markets, in frontier and emerging equities a large part of the cross-sectional variation in returns comes from inter-market differences rather than intra-market differences (Kim 2012; Zaremba 2016), and we aim at capturing this information as well.

6. Analogous statistics for the quintile and tertile portfolios are reported in Table S2 in the Supplementary Material (available online).

7. Table S2 in the Supplementary Material (available online) uncovers the mean returns on the long-short anomaly portfolios based on quintiles and tertiles of stocks. The results display no qualitative differences.

8. Ahmed, Bu, and Tsvetanov (2018) survey a few further simple measures for comparison of factor pricing models based on absolute or squared intercepts, including their cross-sectional standard deviations, averages, both on “raw” basis or relative to the respective portfolio returns. We find that in our case these alternative measures lead to similar conclusions so, for brevity, we focus on average absolute alphas.

9. Notably, the recent study Kan, Robotti, and Shanken (2013) develops a method to formally test the statistical difference between R² coefficients.

10. Interestingly, C4 produces a large and negative alpha on the LtMom portfolio. The reason is that the UMD factor is formed on the equally weights momentum in small and large firms, while the LtMom, implemented with the value-weighting scheme, gravitates to small firms. The momentum effect is much stronger in smaller firms (see the performance of the MomCap [147] anomaly) –not represented largely in LtMom, but represented markedly in the UMD factor. Hence, LtMom regressed on UMD delivers negative payoffs.

11. Fama and French (2018) and Ahmed, Bu, and Tsvetanov (2018) report also the maximum Sharpe ratios of the intercepts. However, our sample considers a large array of different anomalies of which many are significant. Hence, for all of the models, our estimations of this measure are very elevated and similar in terms of their value. In consequence, we do not report these results as inconclusive.

12. Also, Barillas, Kan, Robotti, and Shanken (2017) establish a different way to test the significance of the difference between the Sharpe ratios in asset pricing tests.