Selección de portafolio : estimación no paramétrica y robusta

Abstract

The classic Markowitz model is very sensitive to estimation erros of its parameters. It generates high uctuations in its weights after every rebalance, therefore high transaction costs. In this article we evaluate non parametric and robust techniques in order to improve the model and correct these errors. The method for incorporating these techniques consists in replacing the estimation of the variance and covariance matrix with alternative tools. For the nonparametric estimation we replace the Pearson correlation coeffcient for the Kendall coeffcient and the Spearman coeffcient. In the robust section we replace it with the comedian matrix.