Research Of HAR-RV GAS Model Based On Investor Sentiment

Since 1982,the volatility model has been proposed,but for a long time there has been no universally applicable observation-driven modeling framework.The GAS(Generalized Autoregressive Score)model proposed by Creal,Koopman and Lucas(2013)fills this gas and received a lot of attention.However,the development of the GAS model mostly lies in the application.There are few innovations about the model.At the same time,since Hurst(1951)proposed the feature of long memory,the long memory of financial market has been widely recognized.This finding is very important for financial risk management.Because once the long memory of the financial market is neglected,it will have a serious impact on the control the grasp of risks of market trends and asset pricing.At the same time,with the rise of behavioral finance,investors are considered to be bounded rational,and their limited rational behavior will cause the market to be ineffective and the asset price to deviate from the intrinsic value.The stock price will be affected by investor sentiment.Based on the GAS model,this paper combines the conditional variance with long memory and investor sentiment,and proposes the HAR-RV-SENT GAS model.In the numerical simulation,the three distributions of t,normal and gamma are compared,and the superiority of the HAR-RV-SENT GAS model is verified.From the empirical results,uses the 5-minute and daily yield data of the Shanghai and Shenzhen 300 constituents to examine the fitting effect and predictive ability of GAS,HAR-RV GAS and HAR-RV-SENT GAS models on market volatility.By comparing the theory and actual autocorrelation functions,it is found that the GAS model with long memory and investor sentiment can better capture the long-term correlation in the volatility,the prediction ability and fitting effect of the model population are better than the GAS model.Compared with the GAS model and the HAR-RV GAS model,the HAR-RV-SENT GAS model with long memory and investor sentiment is also optimal in predicting the conditional variance,also can more acutely capture the spikes and thick tail features in the market.The main content of this paper includes five chapters:In chapter ?,begins with a brief introduction to the research background and significance of this paper.Secondly,a systematic introduction to the volatility model,long memory and investor sentiment.Finally,outline the main content and innovations of this article.In chapter ?,HAR-RV and GAS models which are used in this paper are systematically introduced,and the SPA theory is briefly explained.In chapter ?,firstly,introduce the construction of HAR-RV-SENT GAS model.Secondly,the maximum likelihood method adopted in parameter estimation is introduced,and the theoretical proof of its large sample property is given.Finally,the superiority of the HAR-RV-SENT GAS model is verified by numerical simulation.In chapter ?,empirical analysis.In the context of China's Shanghai and Shenzhen 300 constituent stocks,we explore the long memory and investor sentiment in the stock market.The theory,sample autocorrelation function,SPA test and other tools are used to compare the performance of the volatility model in fitting and prediction.In chapter ?,the main conclusions of this paper are summarized,the future research and development trend are prospected.