Has the last generation of academic research clarified which factors/characteristics/indicators predict which stocks will outperform and which stocks will not? How can academia do better? In his August 2009 paper entitled “The Cross-Section of Expected Stock Returns: What Have We Learnt from the Past Twenty-Five Years of Research?”, Avanidhar Subrahmanyam reviews recent research on cross-sectional predictors of stock returns at monthly or longer horizons and offers observations on how to improve this research. Citing a large number of relevant studies, he concludes that:
- Research addresses more than 50 variables (based on informal arguments, risk-return models, behavioral biases and trading frictions) as predictors of which stocks will outperform and underperform.
- Studies generally employ one of two types of analysis:
- Regression analysis with controls or risk adjustments based on the Fama-French market, size and value factors, momentum, liquidity, macroeconomic indicators and/or dataset-derived factors.
- Portfolio analysis (ranking and sorting).
- Issues across this body of research include:
- Which factors deserve status as standard controls/risk-adjusters?
- How should researchers account for errors/uncertainties in variable measurement?
- What rules should govern selection of the historical interval of measurement?
- What methods are most appropriate for analysis of time variation in predictive power?
- There is more work to be done on how factors/characteristics/indicators interrelate and whether past results survive different methodologies.
“It appears that the cross-section of expected stock returns is subject to myriad empirical influences. The research at this point presents a rather unsatisfying picture of a morass of variables, and an inability of us finance researchers to understand which effects are robust and which do not survive simple variations in methodology and use of alternative controls.”
In summary, investors should adopt a stance of considerable skepticism about stock-picking research.
