Vortrag
Risk Forecasting in (T)GARCH Models with Uncorrelated Dependent Innovations

Benjamin Beckers, Helmut Herwartz, Moritz Seidel

IAAE 2014 Annual Conference
London, Großbritannien, 26.06.2014 - 28.06.2014



Abstract:
Starting with their introduction by Engle (1982) and Bollerslev (1986) (G)ARCH type models have proven quite useful for the dynamic modelling and forecasting of risk attached to speculative asset returns. While the symmetric and conditionally Gaussian GARCH model has been generalized in a magnifold of directions, model innovations are uniformly presumed to stem from an underlying iid distribution. From an empirical perspective we notice that GARCH implied model innovations are likely at odds with the commonly held iid assumption. Drawing on this observation on the one hand, and noticing the heterogeneity of actual dependence patterns on the other hand, we follow two (complementary) strategies to evaluate the conditional distributions of consecutive GARCH innovations, a nonparametric approach and a semiparametric model class comprising standardized copula distributions. In the framework of an out-of-sample analysis for a cross section of 18 major stock markets, higher order dependence patterns are quantified and found to improve standard (threshold) GARCH implied conditional value-at-risk and expected shortfall forecasts that rely on the notion of iid innovations.