PORTFOLIO OPTIMIZATION BASED ON CONDITIONAL CORRELATION MODELS

Abstract

Multivariate volatility models based on conditional correlations have been receiving increased attention due to the possibility of modeling the dynamics in the covariance matrices of financial asset returns. In this paper, the performance of this class of models is analyzed when applied to the problem of optimal portfolio selection of a portfolio of stocks. For that purpose, it is implemented four alternative parsimonious specifications for the class of conditional correlation Garch models. Using data from the Brazilian and North-American stock markets, it is obtained optimal minimum variance portfolios for each of the markets. The results indicate that the minimum variance portfolios obtained with conditional correlation models achieved superior results in terms of lower risk, higher risk-adjusted returns, higher cumulative gross returns and higher cumulative excess returns when compared to the main stock indices of each market.