Decoding Anomalies through Alpha Dynamics

Abstract

This paper studies how alphas of the characteristic-sorted stock portfolios evolve over the months after the sorting date, which I refer to as "alpha dynamics". I develop new tests to examine the alpha dynamics predicted by economic theories. The results provide new insights relevant to assessing whether anomalies (1) are attributable to collective "data snooping" or are real, (2) can be attributed at least in part to mispricing, and (3) imply potential profits after considering trading costs. I study 205 published anomalies and find that t-tests of whether average alphas equal zero fail to detect many real anomalies, a problem that becomes more severe with higher t-statistic cutoffs. Further, the observed alpha dynamic pattern conforms to existing behavioral models rather than rational models for about sixty percent of characteristics, including net share issuance, idiosyncratic volatility, and momentum. Moreover, I show that after-cost profitability is significantly underestimated when alpha dynamics are not allowed for.

Keywords: Anomalies, Alpha Dynamics, Existence, Profitability, Mispricing

JEL Classification: G02,G10,G11,G14