Portfolio optimization: Empirical test on MV, MAD and CVAR models

Abstract

Consulta en la Biblioteca ETSI Industriales (Riunet)

[EN] One of the major problems faced by investors is how much and in which quantity should they allocate their assets; which is the equilibrium point between risk and return? This issue has been thoroughly studied since 1952, when Harry Markowitz presented his Portfolio Theory, and new models have been developed to try to improve it. This paper focuses in three models that share many assumptions when constructing them, being the assumption that asset returns follow a normal distribution, the outstanding one. These models are the Mean-Variance, Mean-Absolute Variance and Conditional Value-At-Risk. The aim is to test them and analyse their outcomes before and after the European crisis: what would have happened if an optimal portfolio was chosen with data before the crisis, and after the crisis. To do this, real-data from the Dow Jones Industrial Average Stock for different periods will be introduced in the models. In this paper, the description of the models and explanation of the optimization problems will be showed helping the reader to understand more in depth how do the stated portfolio optimizations work. Moreover, possible improvements of the models will be introduced. Due to the simplicity of the models compared to the behaviour of markets in real life, issues such as multi-period and introduction of different constraints should help to forecast better future return vs risk outcomes using the same historical data.