Based On POT Model Study Of Dynamic Value-at-risk In China Stock Market

Many studies suggest that financial asset yields sequence data with spikes, heavy tail characteristic, asymmetry and volatility clustering, traditional financial market return distribution assumption of normal distribution with seriously underestimated risks. In order to improve the accuracy of risk measures, a large number of scholars using extreme value theory EVT (Extreme Value Theory), to carry out risk assessment, modeling a trailer of this method only, do not make any assumptions the overall distribution of data, highlighting the tail risk characterization of conducive to grasp conditions of extreme risk. This paper extreme value theory in the POT (Peaks Over Threold) volatility model in combination with other models on the original dynamic risk measures yields sequence. Currently financial markets most common of risk management method still for VaR (Value at Risk), due to the method not meet consistency, and is not consider beyond VaR Hou of market risk,, makes just application the method to for risk metric exists larger of defects, to the good VaR in risk metric aspects of defects, this will ES (Expected Shortfall) method as risk metric of effective tool to intends fill VaR method of defects. Observe the investment behaviors of investors can be found, average investors are diversifying their investment funds investing in different assets, portfolio investments, few investors will all all their money into a single asset. And related structures between different assets for portfolio investment in research has become extremely important, Copula functions among assets described as powerful tool for structure, makes its investment portfolio value-at-risk analysis has been widely applied. Based on above analysis, this of research work will from following several aspects expand:1) TARCH model used to filter raw yield data to know the standard yield data, using commonly used generalized error distribution, t distribution, normal distribution, t distribution and generalized Pareto distribution based on POT model GPD (Generalized Pareto Distribution) after fitting the filter criteria yields a sequence, and with their dynamic VaR measure. VaR further deduced on the basis of measuring formulas of different distribution of ES, its further dynamic ES measure. And separately under different distribution:an empirical comparison of the CSI series dynamic VaR and ES of back-test results, proven POT model in one-dimensional asset value-at-risk measure of superiority.2) using JC-Copula function to connect standard yields the marginal distribution of the sequence, respectively, assuming their marginal distributions such as the normal distribution, t distribution, received marginal distributions subject to different forms of dynamic measuring VaR and ES. Again:an empirical comparison of the CSI series portfolio in different distribution of dynamic VaR, ES back-test results, proven POT model in multi asset value-at-risk measure of superiority.CSI and formed by a sequence of one-dimensional asset portfolio of dynamic VaR, ES back-test and inspection showed that ES method is more robust than the VaR method, and the TARCH-GPD and JC-Copula-TARCH-GPD models of dynamic ES back-test results achieved better than other models. The research is sponsored by National Natural Science Foundation of China:Research on Jump Behavior of Financial Assets Based on Realized Measurement Methods (NO.71171056).