An Application of a Multicriteria Approach to Portfolio Comparisons

Abstract

AbstractMost portfolio selection models found in the financial literature rely on an estimate of the probability distribution of returns to describe risk. Recent controversy over the appropriate description of risk, the psychological (and controversial) nature of probabilistic estimates and the potential bias introduced by estimation errors has raised doubts as to the adequacy of the frequency function as a unique measure of risk for portfolio selection. This paper applies an alternative approach, namely the Electre methods, to portfolio comparisons. The application of this analytical technique, based on `outranking' relationships, brings in a new perspective to this problem. It puts particular emphasis on a multicriteria analysis that integrates into a systematic framework the various dimensions of risk seen by portfolio managers. Problems with respect to the heterogeneity of measurement units, the conflicting nature of portfolio evaluations and the imprecise nature of the data used are taken care of in our approach, where the various criteria are used to derive the efficient set.Keywords: Electre methodsmulticriteria analysisoutranking relationshipsportfolio comparison