Cross-sectional uncertainty and expected stock returns

We study the predictability of cross-sectional uncertainty (CSU) for stock returns. We find that CSU exhibits significant power for predicting monthly stock returns both in and out of sample with annual R-2 of 11.89% and 6.34%, respectively, greater than popular predictors. A bivariate combination forecast using CSU with one of the alternative predictors yields annual out-of-sample R-2 up to 18.08%. CSU generates significant economic gains for a mean-variance investor with a utility gain of over 400 basis points per annum. A vector autoregression decomposition shows that the source of predictability mainly comes from a cash flow channel.

Automated summary

This study finds that cross-sectional uncertainty (CSU) has significant power in predicting both in-sample and out-of-sample monthly stock returns with annual R-2 of 11.89% and 6.34% respectively, surpassing popular predictors. A bivariate combination forecast using CSU with one of the alternative predictors generates an annual out-of-sample R-2 up to 18.08%. CSU generates significant economic gains for a mean-variance investor, with a utility gain of over 400 basis points per annum. A vector autoregression decomposition shows that the predictability mainly comes from a cash flow channel.