Research On Modeling And Forecasting Of Chinese Stock Market Volatility Based On Overnight Information

The modeling and forecasting of stock market volatility has always been an important part of financial economics research.It is important for portfolio selection,pricing of financial assets and their derivatives,and risk management of financial institutions.Since the 1980s,domestic and foreign scholars have proposed the GARCH class and SV class models based on low frequency data to estimate and forecast the stock market volatility,which portrays the agglomeration and time-varying characteristics of stock market volatility.In the 21st century,the modeling and forecasting of stock market volatility based on high frequency data has become a new research trend.On the basis of Realized Volatility,ARFIMA and HAR models are able to describe the long-term memory and heterogeneous characteristics of stock market volatility.The dissemination and diffusion of information is the internal cause of the stock market volatility.As the trading period of the stock market is short,the stock market has accumulated a lot of information during the non-trading hours between two trading days,which is the so-called overnight information.Due to policy transmission,economic trends,international linkage and other factors,the generation of overnight information involves many aspects.Therefore,it is of great significance to study the influence of overnight information on Chinese stock market volatility forecasting.It is of great significance to study modeling and forecasting stock market volatility with overnight information as the new entry point.In academic terms,the definition and classification of overnight information is extended,and the impact of overnight information on stock market volatility are used for modeling and forecasting,which enriches the theoretical space of financial volatility modeling.In theory terms,it provides theoretical basis for policy makers,information disclosers and stock market managers.This study is designed to enable decision-making authorities to clearly understand the impact of these changes on stock market volatility,with the proper control objectives and information disclosure purpose.In this way,they can creat a sound institutional system to effectively reduce the impact on the stock market and finance system and to maintain financial market stability.In practice terms,to correctly understand the impact of overnight information on the volatility is of great guiding significance for the stock market investors to make a correct judgment.Stock market volatility is not a random behavior,but influenced by the overnight information and other factors.On the one hand,the correct understanding of investors can reduce speculation in the market.On the other hand,it is conducive to make reasonable investment decisions with their full use of overnight information.First,we review research results of the domestic and foreign scholars in this field,and establish the research direction,the theoretical basis,and the empirical support for the establishment of the model.On this basis,we then expound the influence of overnight information on stock market volatility from both theoretical and empirical aspects.In the theoretical aspect,the connotation and classification of overnight information are defined,and the relevant theory of information and volatility,the microcosmic basis and the mechanism of overnight information influencing stock market volatility are demonstrated.In the empirical aspect,the overnight information,stock market volatility and its overnight performance and jump behavior are measured.Then we use Granger causality test and mediation effect analysis to test the influence of overnight information on stock market volatility.Last but not least,three kinds of stock market volatility modeling methods based on overnight information are proposed respectively,and they are compared with the traditional volatility modeling methods.Among them,the multi-factor and variable-coefficient models and the HAR-CJI models are constructed by means of the mediating effects of the overnight information on the stock market volatility,the overnight performance and the jumping behavior,to improve the performance of the existing classical stock market volatility model.The composite models are based on the BP neural network model,combining the estimation results of the classical volatility model with the overnight information.Through the empirical test of the three models,it is found that the overnight information can improve the fitting effect and predictive performance of the volatility models.Among them,the former two models have greater theoretical explanatory ability,while the latter have better predictive effect.The results of this study are reflected in three aspects.First of all,as for the impact of overnight information on stock market volatility,macroeconomic policy indicators,international market information and information disclosure of listed companies have different effects on stock market volatility.In particular,the changes in macroeconomic policy indicators such as the benchmark interest rate,the deposit reserve ratio and the purchasing manager index,the changes in international market information such as the international oil price,the London gold price and the Nasdaq index,the improvement in information disclosure of listed companies and the discontinuity between trading days have an increasing effect on intraday volatility.At the same time,overnight information can affect the stock market by overnight performance and its jump behavior,which plays an important role in the stock market volatility forecast.On the one hand,overnight performance is the mediation variable of all the overnight information variables mentioned that affect the volatility of the stock market,while the jump behavior shows a certain degree of mediating effect in some overnight information variables' impact on the volatility.Secondly,from the perspective of stock market volatility model construction based on overnight information,this paper proposes multi-factor and variable-coefficient models,HAR-CJI models and BP-neural-network based composite models.In terms of fitting effect and forecasting ability,they perform better than the classical volatility models and neural network nonparametric models.Finally,as for the prediction ability of the new models,considering overnight information improves the forecasting accuracy of the models in terms of direction change and the magnitude of the volatility,and it also improves the stability of nonparametric models.Among them,the improvement of the forecast direction is mainly manifested in the accuracy of the stock market volatility's positive changes.The classical linear models and the neural network models based on the overnight information are different in explaining the theoretical significance,the prediction method and the predicting performance of the stock market volatility.