A multivariate GARCH model for the non-normal behaviour of financial assets

This thesis extends the dynamic conditional correlation (DCC) model proposed in Engle (2002) to the case of conditional returns supposed to follow an asymmetric multivariate Laplace (AML) distribution as presented in Kotz, Kozubowsky and Podgorski (2003). We prove that maximum likelihood estimator provides optimal estimates of the relevant parameters estimated. We show the applicability of our approach in a comprehensive set of risk management implementations where we compute Value-at-Risk and Expected-Shorfall measures for portfolios composed by a large number of assets