Value at Risk Forecasting with GAS-Dynamics Factor Copulas

Flavio Ziegelmann (UFRGS)

Abstract: We forecast Value at Risk (VaR) for large dimensional portfolios via copula modelling. For that we explore factor copulas, with both static and dynamic fittings. In the dynamic case the dependence parameters are driven by a GAS (Generalized Autoregressive Score) model. Our empirical results for assets negotiated at Brazilian BOVESPA stock market suggest that, compared to the other copula models, the GAS dynamic factor copula approach has a superior performance in terms of AIC and a non-inferior performance with respect to VaR forecasting.