Abstract:
The estimates of mean returns and covariances are essential to portfolio optimization. Confronted with limited resources, one might attempt to tackle just one of the two problems – accurately estimating either returns, either covariances – while using poorer forecast for the other. We attempt here to provide a realistic assessment of the loss one might incur as a result of using estimates with errors. Our results indicate that, confronted with the choice, one should definitely use his resources on estimating returns rather than covariances when assets belong to the equity class. However, when the skill in forecasting is not particularly high, the relative importance of covariances is much higher than previously believed.
