Forecasting VaR using realized EGARCH model with skewness and kurtosis

Abstract This paper proposes an extension of the realized EGARCH (RealEGARCH) model, namely the RealEGARCH model with Skewness and Kurtosis (RealEGARCH-SK model) for forecasting VaR. The model is able to account for time-varying non-Gaussianities (time-varying skewness and kurtosis). The empirical analysis using the Chinese stock indices, the Shanghai Stock Exchange Composite (SSEC) and the Shenzhen Stock Exchange Component (SZSEC), demonstrates that the RealEGARCH-SK model produces more accurate extreme VaR forecasts than the standard realized GARCH (RealGARCH) and RealEGARCH models with the normal distribution and the RealEGARCH model with the skewed t -distribution (RealEGARCH-ST model).