Study On CSI 300 Stock Index Futures' Overnight Risk Based On CAViaR Model

Since CSI 300 stock index futures can't trade in the night, so the investors who hold futures position will exposure to the market during the non-trading period. Moreover, the risk will enlarge because it belong to financial derivatives. As a result, scientific and accurate measurement of overnight risk of CSI 300 stock index futures has of great practical significance. This paper adopts CAViaR model to directly calculate the VaR of overnight return, and analyzes the dynamic characteristic of different quantiles both in left tail and right tail. Moreover, this paper tries to combine extreme value theory and CAViaR model to estimate the extreme overnight risk. Then both likelihood ratio test and dynamic quantile test are used to backtest the accuracy of these models.The empirical results shows that :(1) the overnight return have stylized facts of negative skewness, lack of long-term memory and leptokurtosis.(2) The CAViaR model has strong predictivity power to the common overnight risk. The SAV-CAViaR and AS-CAViaR model perform better than IGARCH model. After adding the extreme theory,the CAViaR model still can accurately depict the dynamic process of overnight risk in extreme quantiles, all three models used in this paper have no significance differences on accuracy for CSI300 stock index futures under given condition.(3)The GJR model based on leverage effect and FIAPARCH model based on long memory and leverage effect do not perform better than GARCH model. Meanwhile, skewed student distribution is more appropriate for conditional return distribution than normal distribution and generalized error distribution(GED).(4)The CAViaR model based on semi-parametric method shows absolutely excellent predictive ability compared with the GARCH-type models to the normal overnight risk and extreme overnight risk. Overall, CAViaR model is a relatively reasonable model in predicting dynamic risk of stock index futures.