Global asset allocation: from the efficient frontier to the reduction of the instability due to estimation error

Starting from the doubts which accompany the practical application of Markowitz’s model, in this paper the procedure of resampling is analysed. This methodology can be used to provide a partial solution to the problem of estimation error in the parameters used in asset allocation models. The logic at the basis of the mean-variance portfolio theory is still valid. The statistical procedure, examined here, is fed with the results from Markowitz’s model, using the efficient frontier as a fundamental input for the construction of resampled portfolios. Therefore, a strong opposition between the efficient frontier and the resampled one is incorrect. However, in the real world, as opposed to the Markowitz’s ideal world, errors can be made at the input estimate stage. The estimation of the error component, during the build up stage of portfolios, becomes a necessary step.