The Optimal Use of Return Predictability: An Empirical Analysis
Université Lille Nord de France - Skema Business School
University of Exeter Business School, University of Exeter
University of Warwick - Finance Group
AFA 2006 Boston Meetings Paper
In this paper we study the economic value and statistical significance of asset return predictability, based on a wide range of commonly used predictive variables. We assess the performance of actively managed portfolios strategies which optimally exploit such predictability, both in-sample as well as out-of-sample. Such strategies were first studied by Hansen and Richard (1987) and Ferson and Siegel 2001. Our criterion is to maximize various ex-post performance measures, including maximum Sharpe ratio, utility premia, and transaction costs. We develop a test statistic, based on the difference in maximum Sharpe ratio, that has both an intuitive economic interpretation as well as known statistical properties. We are thus able to assess the statistical significance of the economic gains from predictability. Our analysis allows us to compare and rank different predictor variables and also groups of predictor variables.
Overall we find that the optimal use of conditioning information does indeed significantly improve the risk-return tradeoff available to a mean-variance investor, relative to traditional buy-and-hold strategies. These findings are consistent across the different performance measures employed.
In addition we also compare the performance of the unconditionally efficient strategies with conditionally efficient strategies from an investment-based perspective. We find that the performance of the two strategies is quite different due to the differing response of the portfolio weights to changes in conditioning information.
Number of Pages in PDF File: 48
Keywords: return predictability, asset management
JEL Classification: C12, G11, G12
Date posted: March 22, 2005
© 2016 Social Science Electronic Publishing, Inc. All Rights Reserved.
This page was processed by apollobot1 in 0.266 seconds