The Economic Value of Estimated Portfolio Rules Under General Utility Specifications

Abstract

This paper addresses the economic value of estimated portfolio rules under general utility. Incorporating estimation risk magnifies errors associated with mean-variance approximations to the economic value of portfolio rules. In fact, for some preference specifications, including CRRA utility, the approximation error can be unboundedly large. The paper proposes two methods designed to curb the effects of estimation risk in a general utility setting. The first involves constraining portfolio weights to reside within the unit simplex. The second involves forming combinations of estimators and characterizes optimal combining weights using standard portfolio formulas. Both approaches dramatically improve portfolio performance in simulation and out-of-sample environments.

Keywords: Combined estimators, Shrinkage, Portfolio choice, Estimation risk

JEL Classification: G1, C13