To what extent is resampling useful in portfolio management?

Petitjean, Mikael [FUCAM] Delcourt, François

Implementation of the Markowitz mean-variance portfolio construction technique requires knowledge of the expected returns of all assets as well as the covariances. Since these parameters are unknown and are subject to sampling error, portfolio weights may be unstable and unreliable. To address this issue; Michaud (1998) establishes the resampled efficiencyTM technique which has gained ground in portfolio management. In this paper, we take a new look on the resampled method and compare it with the Markowitz mean-variance portfolio construction technique by assessing the performance of three representative portfolios, i.e. the global minimum variance portfolio, the intermediate return portfolio, and the maximum return portfolio. We show that resampling leads to more stable and more diversified portfolios. However, the out-of-sample analysis shows that resampling does not systematically increase (decrease) the risk-adjusted performance (turnover) of the portfolios.