A0855
Title: Value-at-Risk forecasts under misspecified conditional models
Authors: Wilson Chen - The University of Sydney (Australia) [presenting]
Bradley Rava - University of Sydney (Australia)
Nam Ho-Nguyen - The University of Sydney (Australia)
Abstract: GARCH-type models are commonly used to capture the dynamics of conditional volatility in financial time series. Vast empirical evidence suggests that the conditional distributions of financial returns tend to be heavy-tailed and asymmetric. To avoid the challenging task of finding the correct parametric family of innovation distributions, Gaussian quasi-maximum likelihood estimators are frequently used to obtain consistent parameter estimates. However, such misspecified conditional models will lead to unsatisfactory Value-at-Risk (VaR) forecasts. A method is provided for obtaining adequate VaR forecasts under models with misspecified innovation distributions.
