Multivariate volatility models in financial risk management and portfolio selection

Resumen

Multivariate volatility modeling is now established as one of the most influential and challenging areas in financial econometrics. Rather than modeling assets separately in a traditional univariate way, research in econometric modeling of volatility has been evolving towards the extension of the univariate framework through the development of multivariate specifications able to model and predict the temporal dependence in the second-order moments of many assets in a portfolio or in different markets taking into account their correlated behavior. Therefore, the use of multivariate volatility models in quantitative risk management has gained increased importance among academics and practitioners concerned with measuring and managing financial risks. In this thesis we study multivariate volatility models in problems involving quantitative market risk measurement and management. First, we consider the risk measurement problem of forecasting value-at-risk (VaR) using multivariate models vis-à-vis traditional univariate models in problems involving diversified portfolios with a large number of assets. Second, we present a novel active risk management approach based on current regulatory criteria to select optimal portfolio compositions. Finally, I discuss the implications, advantage and caveats of using multivariate volatility models, and propose research lines that can contribute to guide further research in this area.