Skewness from High-Frequency Data Predicts the Cross-Section of Stock Returns

Abstract

Theoretical and empirical studies document a negative relation between stock returns and individual skewness. In these studies, individual skewness has been defined with predictive models, industry groups and even with options' skewness. However, measures of skewness computed only from stock returns, such as historical skewness, do not confirm this negative relation. We propose a model-free measure of individual skewness directly obtained from high-frequency intraday prices, which we call realized skewness. We test whether realized skewness predicts future stock returns by sorting stocks every week according to realized skewness, forming five portfolios and analyzing subsequent weekly returns. We find a negative relation between realized skewness and stock returns in the cross section. A trading strategy that buys stocks in the lowest realized skewness quintile and sells stocks in the highest realized skewness quintile generates an average raw return of 38 basis points per week with a t-statistic of 9.15. This result is robust to different market periods, portfolio weightings, firm characteristics proxies and is not explained by the Fama-French-Carhart factors.