CSU's Predictive Power for Stock Expected Returns

Abstract

This paper aims to investigate the predictive power of cross-sectional uncertainty (CSU) on expected stock returns. Building upon the research of Deshui Yu and Difang Huang, this study extends the sample period to December 2022 and empirically finds that CSU has a strong predictive ability for expected stock returns, outperforming the 13 macroeconomic variables and 4 measures of economic uncertainty proposed by Goyal and Welch (GW). The results show that CSU's predictive power strengthens as the forecast horizon lengthens and passes tests both in-sample and out-of-sample. Moreover, the combined forecast of CSU with macroeconomic variables and economic uncertainty variables can produce better out-of-sample forecast results, especially over longer periods. Economic tests indicate that using CSU to predict stock returns can bring higher economic utility to risk-averse investors. Therefore, this paper concludes that CSU is an effective forecasting tool that can provide investors with additional information for predicting market excess returns.