Computing Conditional VaR using Time-varying CopulasComputing Conditional VaR using Time-varying Copulas

It is now widespread the use of Value-at-Risk (VaR) as a canonical measure at risk. Most accurate VaR measures make use of some volatility model such as GARCH-type models. However, the pattern of volatility dynamic of a portfolio follows from the (univariate) behavior of the risk assets, as well as from the type and strength of the associations among them. Moreover, the dependence structure among the components may change conditionally t past observations. Some papers have attempted to model this characteristic by assuming a multivariate GARCH model, or by considering the conditional correlation coefficient, or by incorporating some possibility for switches in regimes. In this paper we address this problem using time-varying copulas. Our modeling strategy allows for the margins to follow some FIGARCH type model while the copula dependence structure changes over time.