Risk Measurement Of China's Stock Market Based On Quantile Regression Model

As society advances, the financial market are playing an increasingly important role in the economic development of countries.The financial market's stability and development is one of the factors of social stability and development.But under the influence of finance innovation and technological progress,economic globalization and finance integration,and so on. The fluctuation of the fiancial market and system risk are aggravated greatly, financial market and global financial environment have changed greatly.Under this environment risk management have become more important.,the foundation of risk management is to measure the risk.As a new tool of risk measuring and management,VaR technique is used widely since it was born,and at present has become the major technique to measure market risk abroad,but there are still large gaps between the application of VaR method in China and the international financial market risk management techniques, so how to learn from more mature international method of calculating VaR model have important application value and practical significance for financial market risk management in China.This paper used both the quantile regression method and GARCH method to make models, and then study the Shanghai Composite index's VaR. In GARCH method we used four different models (GARCH, TGARCH, EGARCH and PGARCH) to calculate the Shanghai Composite Index's VaR under the t distribution and the generalized error distribution., While the quantile regression method also established two different models to calculate the Shanghai Composite index's VaR. Then in this paper, the results were tested by Kupiec likelihood ratio test, which compared the merits of the quantile regression method and the GARCH mothod.From the model test overcome, at 95% confidence level,the LR statistics were less than the critical value, shows that the model can be accepted, while at 99% confidence level, the GARCH, TGARCH, EGARCH model of GED Distribution did not pass inspection, the model can not be accepted, indicating that GED distribution may not be applicable to the stock market in China.The result also showed that the LR statistics based on the quantile regression models were less than the LR statistics based on the GARCH models, indicating that the quantile regression model measured market risk more accurately.From the failure rate,we can find that:at 95% confidence level, quantile regression models have the lowest failure rate, GARCH (1,1) of GED Distribution have the highest failure rate, while at 99% confidence leve, the quantile regression models still have the lowest failure rates.In general the quantile regression is better than the GARCH method. Therefore the quantile regression models have a good nature at both the failure rate and LR statistics. This fully shows the quantile regression models have statistical superiority.So using quantile regression techniques to make model not only has a good nature, but also maintain the statistical advantages.Quantile regression techniques open a new path for risk measurement,and it will have broad prospects.