Performance of Characteristics-Based Portfolio Choice

Abstract

We compare the performance of the characteristics-based parametric portfolio approach introduced by Brandt, Santa Clara and Valkanov (RFS 2009) with standard optimal portfolio investments on the basis of S&P-500 stocks. We establish that the characteristics-based parametric portfolio approach can only be exactly justified as optimal investment under exceedingly strong assumptions. Moreover, we find that the empirical implementation of the parametric portfolio policy approach to US-data (for 1995-2013) runs into difficulties at moderate levels of relative risk aversion. Without refinements an interior maximum of the expected utility functional need not exist. Finally, and despite the strong implicit assumptions required by theory, the characteristics-based approach compares reasonably well with optimal portfolio investment for absolute risk aversion even in the absence of transactions costs, when we run horse-races in simulated and empirical data.