Tail Risk Forecasting Using Bayesian Realized Egarch Models

This paper develops a Bayesian framework for the realized exponential generalized autoregressive conditional heteroskedasticity (realized EGARCH) model and adopts a standardized Student-t and a standardized skewed Student-t distributions for the return equation. The Bayesian estimators show more favorable results than maximum likelihood estimators in a simulation study. We apply the proposed models by including three realized volatility measures to forecast one-step-ahead tail risks over a period of 1000 days of seven stock market indices. The proposed model yields more accurate tail risk forecasts than competing models. The sub-sampled realized range is found to be a crucial volatility measure