HAR-RV Based On Classified Information Model Of Chinese Stock Market Volatility

As volatility is unobserved,the research on volatility has become more and more important,occupying a significant position in financial market research.In the course of theoretical research,the measurement of financial risk,the pricing of financial assets,and the pricing of financial derivatives all have a relationship with volatility.In terms of investment practice,the changes of volatility in financial markets are also the most concerned issues for financial investors.It is crucial to measure and predict volatility.In recent years,with the rapid development of computer technology,the acquisition and storage of high-frequency data have become more and more convenient.The use of high-frequency data to study the realized volatility has gradually attracted the attention of scholars and has become an important perspective of volatility research.Based on the heterogeneous market hypothesis and the mixture distribution hypothesis,this paper took the classified information into the HAR model to analyze the impact of classified information on China stock market realized volatility and explore a better model with better fitting effect and robust prediction ability.This paper used the method of empirical analysis,and selected the high-frequency data of the Shanghai Composite Index from January 4,2012 to November 30,2018 as the research object.First,we calculated the realized volatility and analyzed the characteristics of the realized volatility.Secondly,the HAR-RV model was constructed based on the heterogeneous market hypothesis theory,the residual sequence of the model had been tested and we found that heteroscedasticity existed,so we constructed the HAR-RV-GARCH model.Then,based on the mixture distribution hypothesis theory and its extension theory,we used volume as a proxy for information flow.According to the impact strength,the volume of serial correlation removed was decomposed into four kinds of information flows entering the market--"strong positive information flow" and "weak positive information flow","Strong negative information flow" and "weak negative information flow",which were proxy of the classified information flow.We took these variables into the HAR-RV-GARCH model to investigate the impact of the classified information on the realized volatility.Finally,this paper employed the rolling window method for out-of-sample forecasting,and used loss function and Mincer-Zernowitz regression to evaluate the predicting power of HAR-RV,HAR-RV-GARCH and HAR-RV-GARCH-V model soundly and reliability.Through the different loss function,the method of out-of-sample forecast was used to study the prediction ability of HAR-RV model,HAR-RV-GARCH model and HAR-RV-GARCH-V model with classification information.The empirical conclusions show that: firstly,the volatility of China's stock market has the characteristics of leptokurtosis and fatter tails,long memory,and skewed to right.Second,China's stock market has characteristics of investor heterogeneity.Among them,short-term investors' trading behavior has the greatest impact on market volatility,followed by medium-term investors,while long-term investors' trading behaviors' impact on market volatility is minimal.Thirdly,by taking classification information into the realized volatility model,we can obtain that the influence of ?strong positive information flow? and ?strong negative information flow? on volatility is significant.There is a significant negative correlation between ?strong positive information flow? and the realized volatility.There is a significant positive correlation between the ?strong negative information flow? and the volatility,and the positive and negative information flows have asymmetry on the impact of volatility.Fourthly,based on the results of loss function and Mincer-Zarnowitz regression method,we can obtain that the HAR-RV-GARCH-V model with classified information is optimal in prediction ability,which also shows that taking the classified information into model is effective for predicting the volatility.