Robust Portfolio Selection.

In this paper, we discuss one of the reasons leading practitionners to the rejection of the Markowitz model and propose a new stastistical method to avoid this problem. To be more precise, we discuss the problem of statistical robustness of the Markowitz optimizer and show that the latter is not robust, meaning that a few extreme assets prices or returns can lead to irrelevant 'optimal' portfolios. We then propose a robust Markowitz optimizer and show that it is far more stable than the classical version.