The Calculation And Empirical Analysis Of A Risk Management New Method-VaR Of Financial Market

In this paper, we studied the calculated methods of value at risk (VaR) in financial market. At the same time, we verified them using Chinese security market data. We introduced some calculated methods of VaR with static and unconditional distribution, and analyzed the time-variation of volatility using ARCH model and its general forms detailsly. Because return of Chinese security market is non-normal distribution, so we use GARCH-t model which can describe the time-variation of volatility and the high-peaked and heavy-tailed characteristics of return to calculate VaR value of market index. From empirical results we know that this model is efficient. The full text consists of seven parts.Chapter 1 reviewed history and current situation of financial risk management, we reminded VaR especially. At the same time, we gave sample which be used in this paper and introduced some statistic software. We proposed problems studied in this thesis lastly.Chapter 2 introduced concept, calculated principle and methods of VaR. We compared their advantages and disadvantages and pointed out their suitable regions.Chapter 3 studied the accuracy verified of VaR models. We introduced three tested methods and obtained the partition of sample using Basel Committee' s method.Chapter 4 tested normal distribution of return time series of Chinese stock market. Resulted show that the distribution of return time series is non-normal, it has high-peaked and heavy-tailed characteristics. At the same time, we explain the reason of it. We tested the time-variation of means and volatility, results show that the time-variation of means is not obvious, but the time-variation of volatility is very clear. We also verified stochastic feature of return time series, from results we see that Chinese stock market is weak form efficient.Chapter 5 studied the volatility of return time series using correlated analysis. From results we know that correlation of return time series is not obvious, but correlation of the square time series of return, i.e., variance time series, is clear. So we use GARCH model to estimate conditional variance, and calculated parameters in model by the way.Chapter 6 calculated VaR values of composite index in Shanghai Stock Exchange and component index in Shenzhen Stock Exchange at difference confidence level using difference methods, and we compared these results with real profit-loss. From results we know that VaR values of two index based on GARCH(1, l)-t model is accuracy that others.Chapter 7 summarized results in this paper and the problems to research are given.