Research Of China’s Stock Market Risk Measurement

In the recent years, Global financial market booming, at the same time volatility is rising. How to effectively manage and accurately predict financial risk become an important problem for government and financial institutions to be solved. While important index for measuring and researching financial market risk is one is volatility, accurating measurement and predicting not only helps to understand the operation of the financial market characteristic, but also is helpful for investors to accurately predict the investment risk,so as to make right investment decision. Since based on highfrequency data to modeling and forecasting of volatility, having some advantage of using information fully and not need to estimate the parameters, realized volatility model gets a promotion. In the study, heterogeneous market hypothesis is one of more famous. On the basis of this hypothesis, HAR model was put forward. Second, Value at Risk, for financial institutions and financial regulators, in addition to measure and forecast of volatility, and should prepare risk response ahead of time. This involve another import index in risk management, Value at Risk(VaR).Through empirical analysis on the characteristics of the realized volatility of Shanghai stock index and related inspection, the conclusion is that it doesn’t obey the normal distribution, has extreme kurtosis and starboard,long memory characteristic, and HAR-RV model is applicable in China’s stock market. Based on existing in China’s market mainly confined to HAR-RV model that a constant coefficient,we put forward the dynamic coefficient of nonparametric HAR-RV model(TVC-HAR-RV model),with local linear method and strong local weighted method to estimate the nonparametric dynamic coefficient and its confidence interval, and model feasibility test and Conditional Predictive Ability testing. In addition to this, we explored the form of coefficient of TVC-HAR-RV model, and construct the TVC(1)-HAR-RV model. For China’s stock market empirical analysis, making the following conclusions:(1) the coefficient of HAR-RV model is dynamic change, not constant;(2)testing HAR-RV model, TVC-HAR-RV model and TVC(1)-HAR-RV model’s prediction ability in outsample for 1、5、21 day, the conclusion is for realized measurement and prediction ability test,on the whole, TVC-HAR-RV model is superior to HAR-RV model and TVC-HAR-RV model;(3) strong local weighted method relative to the local linear method,it improves model’s prediction ability.At last, in order to accurately measure VaR, this paper uses the parameter & semiparametric method to measure the standardizing returns’ distribution of GED, SGT and GPD and the accuracy of risk measure in the GARCH, EGARCH and PGARCH models based on FSH, respectively. By this method, twelve kinds of risk measure VaR models have been build. With these models, the dynamic VaR value of Shanghai Stock market index on an out-of-sample evaluation. In different condidence level can be gotten. After that, model’s accuracy can be tested in a composite way with LR, LRuc and DQ. In the end, the model’s priority can be measured by taking advantage of the lost function.The result is that EGARCH will be more accurate in the confidence level of 99% and 99.5%,while PGARCH will be accurate in the confidence level of 95% and97.5%.Besides,when the confidence level of 99% and 99.5% semiparametric will be more accurate than parameter model.