The Application Of Extreme Value Theory To Stock Index Futures Margin Levels Setting

In stock index futures market, the setting of the margin plays a key role in the whole market. The level of the margin is too low, which will make the futures market excessive investment. So it is dangerous for the market; The level of the margin is too high, which will reduce the number of transactions. Reasonable margin level plays a major role in the stock index futures market, and it makes the stock index futures trades at high efficiency and safety. The traditional normal distribution cannot be well fitting of the tail of the data, thus it causes the setting of the margin unreasonable. So in the stock index futures market, the probability of the loss is higher than the estimation of normal distribution. Extreme Value theory can simulate the risk of tail of the stock index futures well. And then it can calculate the VaR and ES. Compared to VaR, the ES can cover the fluctuations of the price well. In this paper, we choose the ES as the margin level. The csi300stock index futures is the research object. At first, for the data, we do normality test. Through the parameters and graphics, it does not obey the normal distribution and it has the intense ARCH effect. The generalized pareto distribution of Extreme Value theory can fit the tail of the yield data well. The parameter estimation about the generalized pareto distribution can be realized through bootstrap sampling method that can overcame the small sample data. We can compare the margin level evaluated between the Extreme Value theory and Extreme Value theory based on the bootstrap sampling method. At last, we can find that the margin level based on the bootstrap sampling method can better cover the price fluctuation in the stock index futures.During the setting of the margin level process, we set different positions. For the different position, we compute the margin levels. So we can provide reference for stock index futures trading in our country. In this paper, the margin levels based on the extreme value theory and bootstrap sampling method are better.