Statistical and economic significance of stock return predictability : a mean-variance analysis

In this paper, we use mean-variance analysis to investigate the statistical and economic significance of stock return predictability as documented in Fama and French (1992). We ask if investors who form mean-variance efficient portfolios conditioned on the predictive variables can earn higher expected returns and higher expected utility than those who optimize unconditionally. The result shows that, by and large, they cannot. We examine various reasons for the lack of economic gains from return predictability and conclude that, while the in-sample relation between returns and predetermined firm-specific variables is strong in terms of the t-statistics, it is not stable enough to produce better out-of-sample predictions