[Excerpt] In recent years, financial researchers have gradually accepted the notion that stock returns are partially predictable (Cochrane, 1999). Most often, the extent of return predictability is assessed from a statistical perspective, with the t-statistics and R2’s of predictive regressions guiding conclusions. Statistical ‘evidence’ of predictability, however, does not necessarily imply economic significance.
In this paper, we assess the significance of predictor variables within an asset allocation framework. Recent research shows that the optimal allocation to risky stocks is horizon dependent if stock returns are predictable. The extent of horizon effects, therefore, is a convenient metric of return predictability and our results are presented as plots of the optimal allocation to the risky asset as investment horizon increases. If a variable is useful for predicting stock returns, knowledge of that variable’s value will cause a utility-maximizing investor to alter her optimal allocation. Thus, the importance of predictor variables is judged from an economic perspective, not a statistical one.
Boudry, W. I., & Gray, P. (2003). Assessing the economic significance of return predictability: A research note [Electronic version]. Retrieved [insert date], from Cornell University, School of Hospitality Administration site: http://scholarship.sha.cornell.edu/articles/270