Study On The Risk Prediction Of CSI 300 Index Based On GARCH Family Models

The launch of CSI 300 Index is helpful for the investors to make a comprehensive grasp of the stock market; it also contributes to provide basic conditions for the innovation and development of the Index product. This kind of Index has become a set of investment, speculation and hedging Index product with the successful launch of CSI 300 Index futures. Perhaps most investors hope to obtain a kind of high expected return without any risk on the condition of the smooth operation of CSI 300 Index, but in reality, the operation of the Index seem to be no laws, and high yield always accompany high risk. Therefore, how to measure and effectively control the risk of CSI 300 Index will become a very concerned problem for the investors.I introduced the basic theory of the GARCH family models and the VaR methods in the first place, and then applied these theories to study the risk characteristics of CSI 300 Index and calculate the VaR value. Overall, the results of this thesis can be summarized as one study and two options. One study refers to using a variety of GARCH family models to study the risk characteristics of CSI 300 Index. The first option is to choose the most appropriate model which can fits the return series of CSI 300 Index among the established GARCH family models. The other option is to choose the most proper approach to mesure the risk of CSI 300 Index, in VaR-GARCH method, historical simulation method and Monte Carlo simulation method.This thesis includes the following five parts:In the first part, I described the purpose and significance of this thesis, reviewed the research at home and abroad, discussed the main research methods, in addition to present the main contents, innovations and weaknesses of this thesis,In the second part, I made the data analysis of CSI 300 Index, and studied the probability distribution, stability, and conditional heteroskedasticity effect. Then I established six kinds of GARCH famiiy models, and studied on each of them in detail Finally, I chose the best fitting model on the basis of the evaluation of the prediction effect of the established models; it is ready for calculating the VaR value of CSI 300 Index.In the third part, I tried to calculate the VaR value of CSI 300 Index using the VaR-GARCH method, historical simulation method and Monte Carlo simulation method respectively and ultimately determined the most accurate method according to return test results.In the fourth part, I talked about some application of the results of this thesis. On the one hand, I elaborated the significance of the results in predicting the risk of China's stock market; On the other hand, I gave some suggestions in formulating investment strategy, setting investment principles, and controlling risk effectively for the investors in stock index futures.The fifth part is the conclusion; I summarized the main research result of this thesis in this part.The method I adopted is empirical research method. Through the empirical research, I found that the yield series of CSI 300 Index subject to Student's t distribution, and have a significant characteristic of clustering volatility, and the expected return is impacted by the expected risk on a certain positive effect. The results show that, when we are going to study the volatility of CSI 300 Index and calculate its VaR value, we can use GARCH family models and adopt VaR-GARCH method to calculate the VaR value. In addition, the results of this thesis have important reference value in predicting the risk of China's stock market, and formulating the strategy of index futures investment.I have made several innovations as following aspects, and made a contribution in the studying on CSI 300 Index. I corrected the error form of the VaR-GARCH method in some common references, namely we can't use the formula of normal analysis to calculate the VaR value when the yield series of CSI 300 Index have a conditional heteroskedasticity effect. In that case, we should use the conditional standard deviation to represent the loss of CSI 300 Index, and then determine an appropriate sub-site under a given confidence level, ultimately calculate the VaR value. In addition, I tried to apply the results of this thesis to predicting the risk of China's stock market, and formulating the strategy of index futures investment, in order to apply the theoretical research results to practice.Although this thesis has achieved fruitful results, there are still some deficiencies. The main disadvantage is that I failed to build an accurate model to simulate the yield series of CSI 300 Index, resulting in the model risk when I use the Monte Carlo simulation method, which seriously affected the validity of Monte Carlo simulation method.