Research On SSE 50ETF Volatility Based On Realized Jump GARCH Model

Since the launch of the Exchange Traded Fund(ETF),it has always received people's attention and attention.ETF products are playing a more and more important role in improving the structure of the securities market,diversifying market risks,and optimizing the allocation of market resources.The main member of the company is not only in the forefront of the market in terms of size and trading volume,but also the derivatives products based on the SSE 50 ETF fund are also rich.significance.With the development of computer technology,the theoretical research on volatility under highfrequency data has been the focus of attention of scholars,but there have also been a lot of confusions in the research process.In summary,there are four main problems: First,how to use the classic GARCH model to model high-frequency data? Second,among the existing measures of realized volatility,such as realized volatility(RV),realized kernel(RK),and Realized Bi-power Variation(BPV),which measurement method is most suitable for China's ETF What about market research? Third,under the assumption of different residual distributions,are the models based on thick-tailed distributions(tdistribution and GED distribution)more suitable for describing the objective situation of the market? Finally,is there a jump phenomenon in China's fund market,and if there is a jump,can the estimation of volatility be improved by establishing a jump model? This article studies the volatility of the SSE 50 ETF in view of the above issues,which can not only uncover the volatility of the SSE 50 ETF,but also solve the actual market problems.At the same time,it can answer some of the confusions in the above theoretical study of volatility,which has both practical and theoretical significance.This paper constructs a Realized GARCH model,and expands on three aspects based on this model.First,on the realized measure,the realized volatility(RV),realized kernel(RK),and Realized Bi-power Variation(BPV)are respectively applied to the Realized GARCH model.The research found that the realized kernel(RK)and the Realized Bi-power Variation(BPV)are superior to the realized volatility(RV)in the performance of the model.Specifically,the realization of double power variation(BPV)can significantly improve the fitting effect and prediction effect of Realized GARCH model,and the realized kernel(RK)can improve the prediction effect of Realized GARCH model.Secondly,this paper uses Realize the difference between the realized volatility(RV)and the Realized Bi-power Variation(BPV)to build a jump factor,extend the Realized GARCH model,and empirically test that the Realized Jump GARCH model with the jump factor can improve the fit Effect and prediction effect,but the promotion effect is limited;Finally,the paper analyzes the comparative effects of the model under the assumptions of normal distribution,t distribution and GED distribution,and finds that based on the thick tail distribution(t distribution and GED distribution)The modeling effect has been significantly improved.Specifically,the t-distribution-based model can significantly improve the model's ability to fit and predict,and the GED-based model can significantly improve the model's ability to fit.Finally,it was found that the comprehensive effect of Realized Jump GARCH based on t-distribution and Realized Bi-power Variation(BPV)was the best.In this paper,by studying the SSE 50 ETF volatility,we find that the SSE 50 ETF high-frequency data has obvious spikes fat tails,and long memory characteristics.The market is more stable and suitable for investors' long-term investment.