Out-of-Sample Performance of Asset Allocation Strategies

Abstract:     
Using data for the S&P 500 Sector portfolios between 1989 and 2007, I find that sample-based mean-variance portfolios are very unstable and perform poorly out of sample in terms of Sharpe ratios, certainty equivalent returns and turnover. Minimum-variance and Bayes-Stein portfolios, which are supposed to be less susceptible to the estimation error plaguing mean-variance, also fall significantly short of a naive equally-weighted policy. Imposing shortsale or turnover constraints limits the fluctuation of portfolio weights and improves performance considerably. When the normality assumption about excess portfolio returns is relaxed, there is no evidence of the constrained portfolios performing worse than the exulted equally-weighted portfolio. I propose a sophisticated turnover constraint rule which recognizes the path dependency of the optimal portfolio policy and enhances the out-of-sample monthly Sharpe ratio by 13% relative to the best of the other strategies considered in this paper.

Number of Pages in PDF File: 40

Keywords: portfolio allocation, out-of-sample performance, turnover constraints

JEL Classification: G11

working papers series