The Research Of CSI300 Stock Index Futures Risk Measuring

The stock index futures traded in the China financial futures exchange on the April 16, 2010,marking the end of the era of "unilateral market".As China’s first financial futures products in the true sense, it has a huge influence of development of our financial market in China.Since listing, the trade has expanded year by year, the size is much higher than those of the Shanghai stock exchange stock turnover.However, at the same time of rapid development, there are some prominent problems. More constraint to the qualification such as investment, margin system and trading system are in conflict with the spot market, lack of phase two cities now joint manipulation of identification and punishment system, hedge participation mechanism is lower.These problems to a certain extent, increased the risk of stock index futures market.And high lever of stock index futures, sensitivity to price changes, and the difficulty of the trading skills, etc., make it from the spot market risk is greatly increased. Therefore carries on the accurate risk measure is particularly important, must adopt scientific and reasonable risk management methods to prevent and control of the risk.In this paper, on the basis of literature study, the selection range of the most widely used now are VaR technology as the basis of the CSI300 stock index futures risk measurement, with the CSI300 index futures data and 30 minutes of high-frequency data as the research object, by introducing GARCH class models, SGED distribution to achieve accurate measurement of risk, through the comparative analysis of different samples in the prediction effect, combined with posterior inspection and sample prediction effect, the selection of the optimal risk measurement model. The results show that:(1) the CSI 300 stock index futures returns distribution has the characteristics of peak, thick tail, biased and fluctuation has gathered and leverage.(2) in terms of volatility estimates, the model under the normal distribution assumption has the worst goodness fit. the model under the assumption of t, GED, SGED distribution can significantly improve the goodness fit, including APARCH model fitting effect is superior to the GARCH model, EGARCH model is insufficient performance to some extent.(3) in samples prediction and out of sample prediction has a big difference,according to the comprehensive performance of t test and Kupiec test of VaR, no matter day data or 30 minutes, EGARCH-M-GED has achieved the best out of sample prediction.