Time Variation in Extrapolation and Anomalies

Wei He,Yuehan Wang,Jianfeng Yu
DOI: https://doi.org/10.2139/ssrn.3564119
2020-01-01
SSRN Electronic Journal
Abstract:We find that the degree of extrapolative weighting in investors' belief (DOX) proposed by Cassella and Gulen (2018) has strong predictive power for a broad set of overreaction-related anomalies in the stock market. The average return spread of these anomalies is about 0.81% per month following high DOX periods, and -0.22% per month following low DOX periods. In sharp contrast, DOX has opposite, but weaker, predictive power for under-reaction-related anomalies. In addition, the predictive power of DOX is robust after controlling for a broad set of economic forces including investor sentiment and the consumption surplus ratio. Moreover, most of the DOX effect on long-short anomaly returns derives from the short legs of these overreaction-related anomalies, suggesting that time variation in DOX leads to more time variation in overpricing than in under-pricing, probably because of short-sale impediments.
What problem does this paper attempt to address?