The Research On Range Risk Of Index Futures Market Based On Bayesian Conditional Autoregressive Expectile Model

Stock index futures serve as financial derivatives,has played a positive role in completing market function,hedging,risk aversion and so on.Different from other financial derivatives,however,the trading mechanism of stock index futures possesses the characteristics of high leverage,price sensitivity,etc.At the same time,luck of effective market regulation makes the market contain great risk.Especially when the financial markets come down in a whole,futures market declines even more.Against this backdrop,quantizing the futures market risk scientifically and accurately is the core problem of market operation.This paper chooses the Shanghai and Shenzhen 300 Index Futures as the research object,and select the range price return data that reflect the market information representatively,use the more efficient empirical model—Bayesian conditional autoregressive Expectile model,to analyze dynamic characteristics of fluctuations of CSI300 IF market.And make comparison with the predicting performances of other models,aiming to explore the optimal model that conforms to Chinese stock futures market.According to above research line,the paper firstly defines the implication of range risk,the choose of risk indicators and conventional method of market risk measurement.Meanwhile,we recommend the risk modeling process of Expectile and propose the Asymmetric Inverse Gaussian(AIG)distribution that introduced the scale parameter as the linkage function in Bayesian Expectile model.Secondly,we develop the Bayesian Conditional Autoregressive Expectile model to analyze and predict the CSI300 IF market risk.Using the AIG distribution to design the Gibbs Sampling algorithm for MCMC estimation of model parameters.Lastly,selecting the CSI300 IF contract price data spanning from 2010.4.16 to 2018.3.21 as the sample,we investigate the advantages in parameters estimation and model selection of Bayesian Conditional Autoregressive Expectile model over traditional Conditional Autoregressive Expectile model,and employ rigorous VaR and ES backtesting to constrastively analysis predicting performance between the empirical model in our paper and commonly-used model in practice field.All the R code and empirical data has been uploeaded at: https://github.com/ChuangWan/MasterThesisEmpirical results show that:(1)the CSI300 IF range risk yield rate presents remarkable negative-skewed,peak and fat tailed characteristics.The market risk appears obvious autocorrelation feature and is influenced by positive and negative messages asymmetric.(2)The estimation effect based on MCMC is superior to that based on traditional ALS method,and can control estimation risk and selection risk effectively,making the Bayesian model surpass the traditional model.(3)Comparing with general risk model,the Bayesian conditional autoregressive Expectile model has higher prediction accuracy.Especially in extreme risk prediction,the Asymmetric Slope(AS)model form with leverage show up superior predicting ability.