VaR/CVaR Analysis For Stock Market Based On EVT And GARCH Modles

The risk of financial market resulted in the resource competition among various financial institutions to gradually translate into the competition in inner management,operation innovation and corporation cultures, along with the economy worldwise and increasingly prick up in investment open. The situations make the risk management in financial institutions become the foundation of management and development in modern finance corporations. The management of financial market risk increasingly stands out, how to determine and control market risk becomes the issure anxious to solve in financial securities institution,investor and correlative supervisor organization. Under the background, various overseas financial institutions excessive pay attention to the determine and control about the risk. How to design the appropriate models and use the suitable measures to measure risk turn into a pop topic in finance research domanial. Value-at-Risk is a more popular method used to measure risk, which has the characteristic of concision and perspicuous, At the aspect of researching the market volatility, GARCH model and its many relatives are well used for non-linear financial time series.In the paper, we give a systems analysis to the distributing characters of financial time series in stock market. After that, we study the VaR and conditional VaR with the Value-at-Risk method, using the most typical GARCH model and its many relatives including EGARCH,PARCH models and so on to depict conditional variance in finance domain. Besides, we compute the VaR and CVaR about shanghai stock index, using its practical data from January 2, 1997 to May 23, 2008. The paper contains following studies and results: (1) we analyse the statistical characteristics of the log-return rate sequence about shanghai stock index, then we prove that the sequence have the characteristics: fat tails and sharp kurtosis of return distribution and dynamic volatility resulting in the phenomenon of volatility clustering and persistence; there is leverage effect in the market and remarkable asymmetry effect in volatility. We also know the negative shock give a more volatility than the positive one. (2) We get the methodology of VaR and conditional VaR based on the GARCH and EGARCH models with various distributions (include,normal,Student-t and GED) and hence modify the methodology of VaR and CVaR. The outcome tell us that the fat tail of return sequence can be better described with the GED and Student-t distributions than with the normal distribution. In addition, EGARCH model is able to capture the leverage effect in stock market, correspondence specking, it is asymmetric to the effect of negative shock and positive shock, which is more accurately to describe the stock volatility and we get a better outcome based EGARCH model than GARCH model. (3) While the application of Extreme Value Theory is relatively congruent in tracking the fat tails of asset return distribution, we choose the POT method, and the phenomenon of volatility clustering has been extensively studied using GARCH model and its many relatives. The paper combines the two methodologies to come up with a robust dynamic risk management method: GARCH,etc-EVT-VaR/CVaR. Further we analyse and compare the outcome of each model based on Kupiec LR failure rate test. We get the method of GARCH,etc-EVT can more accurately forecast the risk in stock market. (4) At last, we put forward some relevant polices on the basis of the demonstration study about the models we get in the front chapters.