A1708
Topic: Contributed on Risk and volatility modelling
Title: Modelling conditional densities with asymmetric tails
Authors: Steve Thiele - Queensland University of Technology (Australia) [presenting]
Abstract: A new time varying density model is proposed based on the DCS/GAS approach that allows for time varying volatility and different upper/lower tail thickness. The asymptotic properties of the model are presented together with simulations showing that such distributional features can be estimated with practical sample sizes. The importance of tail asymmetry for a range of equity index returns is first demonstrated in-sample relative to existing heavy tailed/skewed models. Application to risk management is then considered, with out-of-sample results for volatility, Value-at-Risk and expected shortfall forecasting shows strong performance relative to leading models from the literature.