Improvement On The Computational Method Of EVaR And Some Relevant Empirical Researches

This paper first proposes an estimation of the Conditional autoregressive expectile (CARE) model based on the empirical likelihood method, then provides the computational method of parametric model of EVaR based on the generalized autoregressive conditional heteroskedasticity(GARCH) model and stochastic volatility(SV) model. The paper adopts two fail-ratio tests of the out-of-sample predicted EVaR to evaluate the computational models under known distributions. Empirical studies with daily log returns of domestic and foreign stock market indexes over the period 2004-01-05 to 2009-12-30 are illustrated.The results show that using two simulated time series data under student t distribution, the out-of-sample predicted EVaR of GARCH-type is superior to the CARE model. During the period of worldwide financial crisis, the EVaR prediction of parametric model performs better than the semiparametric model.