The Realized Volatility And Its Empirical Study On Value-at-Risk

The risk of the financial market is the focus of global financing institution and supervisory authorities. Correspondingly, the accurate measure of volatility is central to the measure of the Value-at-Risk. About the measurements of volatility, the first method is the initial classic model that the volatility is estimated from the financial analysis model (such as Black-Scholes). The second method is the ARCH model and the SV model. The third method is realized volatility which is based upon high frequency data. From these, we can see the measure of volatility is developed quickly. In this paper, we not only study the Value-at-Risk using realized volatility about China's stock markets, but also study the VaR based on the GARCH model. Finally we compare and study the two results.In this paper, we firstly introduce the study background, significance and domestic and international research dynamic states. Secondly, we recommend the theories and methods about Value-at-Risk, the theory and arithmetic about realized volatility, the choice of the optimum data interval, the types of ARFIMA model and the methods about estimating paraments. Thirdly, on the basis of the foregoing statements, we compute the optimum data interval, analyse the distribution characteristics of returns and realized volatility, study the long memory effect of volatility, establish ARFIMA model on these basis. At last, we establish the conditional realized volatility, calculate the VaR using realized volatility and GARCH model. Then we compute the LR statistic and the failure rate to compare the results of two different methods.The results show that the performance forecasting of Value-at-Risk using realized volatility is good under the Student's T distribution and Generalize error distribution. Compared with the Value-at-Risk using GARCH model, the performace of realized volatility is better.