The efficacy of optimization modeling as a retirement strategy in the presence of estimation error

Abstract

We examine the time series performance of mean variance efficient portfolios in the retirement setting. Using a rolling period optimization model we create portfolios with the same ex ante risk as several naive 1/n strategies to discern whether optimization can improve return performance. Data are simulated from TIAA-CREF retirement accounts during 1994 through 2004. We correct for estimation error using weight constraints and James-Stein adjustments. Overall results indicate optimization does outperform most naive investment strategies. The investor's terminal wealth improves 2-30%, depending on the underlying asset allocation and assumed time to retirement. Adjustments for estimation error do little to further enhance investment returns.