GARCH/TARCH Modeling And Empirical Analyzing Base On Different Distribution Of Value At Risk

Under the influence of factors such as economic globalization and finance integration, competing and relaxing control and finance innovation and technological progress, global financial environment and financial market have changed greatly. Meanwhile, the fluctuation of the financial market and system risk are aggravated greatly.However,in order to grasp the core technology of financial risk management.Only depending on measuring risk stake correctly, risk management could be achieved. As one of the risk management technology ,the VaR method is produced naturally. The stock market of our country have developed more than 10 years, much successful experience has already made, but there are a lot of unripe and nonstandard aspects. So the stock market of our countryoften fluctuates radically, and the market fluctuation is far higher than the western ripe stock market of developed country.And it is also influenced by country policy in a great degree.so it is called"policy market". Because the stock market of our country is at a specific developing stage at present, it is imperative to strengthen risk management. So,VaR model has a great meaning to the risk management of stock market of our country.In this paper, A deeper research has been done on VaR system including its background of analyzing, calculation principle, advantage and disadvantage of VaR model. Then, it profoundly studies the typical three methods for calculating VaR at present---the Variance and Covariance Approach,the Historrical Simulation Method ,the Monte Carlo Simulation. Ande gives overall compare and analysis about the three methods. Among them,As to calculate VaR, the traditional calculation methods is established in the premise of Normal distribution.But in China, In the empirical study we also found the phenomenon of "fat tail" distribution, conditional variance and asymmetric variance of stock index return.So we unite the GARCH-model and TARCH-model and VaR-model together and research the heteroscedasticity conditions in stock market at the following hypothsis—the Normal distribution ,the T distribution and the GED distribution.and then use the Kupeic to test those model.Finally,We get a meaningful conclusion:the GED- distribution is the best to simulate the Chinese stock market yield distribution,And so improve the accuracy of VaR calculation.