ABSTRACT
Scenario Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409 19.5 Computational Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
19.5.1 The Data Set and Methodology . . . . . . . . . . . . . . . . . . . . . . . . 410 19.5.2 In-Sample Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410 19.5.3 Out-of-Sample Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
19.6 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416
Appendix 19.A: The General Case of a Positive Semi-Definite
Covariance Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418 Appendix 19.B: The Composition of Efficient Portfolios . . . . . . . . . . . . . . . . . 420 Appendix 19.C: The In-Sample Parameters for the Return Distributions of
Efficient Portfolios. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422
M EAN-RISK MODELS ARE STILL THE MOST WIDELY USED APPROACH in the practice ofportfolio selection. With mean-risk models, return distributions are characterized and compared using two statistics: the expected value and the value of a risk measure.
Thus, mean-risk models have a ready interpretation of results and in most cases are
convenient from a computational point of view. Sceptics on the other hand may question
these advantages since the practice of describing a distribution by just two parameters
involves great loss of information.