The Markowitz portfolio allocation model is not extremely used by professionals. This is due to the nature of the information available. Portfolios designed with the use of historical data suffer from two shortcomings. The risk that generation processes may change and that of their estimation are not accounted for. This paper studies the estimation risk. Simulations are run in order to measure losses of efficiency generated when historical data are used and to determine the adjustments capable of reducing these losses.
AFTALION, F. and CHEMAMA, G. (2000). Nouvelles mesures statistiques du risque d'estimation.