On the Short-Term Predictability of Stock Returns: A Quantile Boosting Approach

Abstract

We study the predictability of stock returns using an iterative model-building approach known as quantile boosting. Examining alternative return quantiles that represent normal, bull and bear markets via recursive quantile regressions, we trace the predictive value of extensively studied predictors including the recently suggested short interest and sentiment variables. We find that short-term returns are predictable to some extent for extreme lower quantiles of the conditional distribution of returns. Interestingly, however, short-interest and sentiment variables do not add significant predictive power, challenging the recent findings on the predictive ability of short sellers for future cash flows and associated market returns.