Time-Varying Skewness and Momentum Crashes

Abstract Returns on conventional momentum portfolios exhibit time-varying skewness that deepens during momentum crashes. We exploit this feature and propose a crash indicator—based on the interaction between conditional volatility and skewness—that provides a measure of downside risk directly from the return distribution. This indicator significantly predicts left-tail realizations of momentum returns at daily frequency, capturing information about crash risk beyond volatility alone. Building on this predictability, a skewness-based dynamic allocation improves daily downside risk management and earns significant alphas over existing momentum-timing approaches. We also show that momentum skewness cannot be fully reconciled with asymmetric market exposure. (JEL G10, G11, G12, C22)