Persistent Predictors and the Cross-Section of Stock Returns

Abstract

We show that when returns are predictable, persistent predictors, known to bias time-series predictive regressions, also bias the estimation of the cross-sectional moments of asset return distribution, especially the variance-covariance matrix of returns. Hence, they also bias the estimation of the various stochastic discount factor bounds, Shanken's (1985) cross-sectional regression test statistic, and Hansen and Jagannathan's (1997) distance measure, except when conditioning information is used efficiently. However, the empirical distributions of these statistics are biased, even with efficient use of conditioning information. We define a test statistic that is immune to this bias and shows consistent results across the persistence levels in our simulations.

Keywords: asset pricing, persistent predictors, stochastic discount factor bounds, conditioning information

JEL Classification: G11, G12