Modeling And Forecasting Volatility In Chinese Stock Market Based On Mixed-frequency Models

Financial market as a link between the real economy and the financial economy the important link,in optimizing the allocation of resources,to maintain stable economic growth,scattered market risk play an irreplaceable role,normal fluctuation in the stock market can promote economic growth,optimizing the allocation of resources,but the excessive volatility of investors can also cause psychological panic,triggering an economic crisis,the growing social unrest.Therefore,it is of great practical significance for both investors and policy makers to accurately predict stock market fluctuations.At present,most of the research focuses on the traditional same-frequency model.This has led many studies to focus on low frequency stock market data,so that the data can have the same frequency as the data of macro exogenous explanatory variables.When data processing is carried out for time series of different frequencies,the data of high frequency is usually transformed into data of the same frequency,which is likely to lose the information contained in high-frequency data.It is under such circumstances that this paper proposes to take the mixed frequency data as the research object.Based on the GARCH-MIDAS model of generalized autoregressive conditional heteroscedastic-to-mixing data sampling,this paper takes the asymmetric effect and uncertainty into consideration for empirical study.The asymmetric part is studied by using the Shanghai stock market data.The empirical results of the sample data show that both short-term and long-term leverage effects exist.In the short term,negative return can cause stronger volatility than positive return,while in the long term,positive return can cause stronger volatility than negative return.The calculation results of out-of-sample data show that the predictive ability of the GARCH-MIDAS model is improved after taking the asymmetry into account.In this paper,three loss functions are used to test them,and the results show that the improved model can improve the predictive volatility of the original GARCH-MIDAS model by up to 2.1%.The data of seven major global stock markets were used to study the uncertainty part.The empirical results of the data in the sample showed that the uncertainty could explain the long-term volatility in the whole sample,especially in China and Japan.The out-of-sample results show that the extended model with uncertainty has better predictive ability than the original GARCH-MIDAS model,which is more obvious during the financial crisis.This can help investors better understand the risks of the stock market and accurately grasp the volatility of the stock market.For policy makers,it can help develop a healthy stock market,guide rational investment and improve the disclosure mechanism.