Out-of-Sample Industry Return Predictability: Evidence from a Large Number of Predictors

Abstract

We uncover extensive evidence of out-of-sample return predictability for industry portfolios based on a principal component approach that incorporates information from a large number of predictors. Moreover, we find substantial differences in the degree of return predictability across industries. To understand these differences, we propose a decomposition of out-of-sample industry return predictability into beta and alpha shares, where the former corresponds to a conditional beta pricing model. A conditional version of the popular Fama-French three-factor model accounts for nearly all out-of-sample industry return predictability, with exposures to time-varying market and size risk premiums especially important for explaining differences in return predictability across industries. We also show that out-ofsample return predictability is economically important from an asset allocation perspective and can be exploited to improve portfolio performance for industry-rotation investment strategies. JEL classifications: C22, C53, G11, G12, G17