The Research On The Stock Market Volatility In China Based On Mixed Frequency Data

There are two phenomena in exsiting researches on stock market volatility. One is that the stock market data used in research are same frequency. They can be low frequency data, such as monthly or quarterly data. They also can be high frequency data, such as daily or intraday 5 minute data. The other is how to keep the same data freuency between the stock market data and macroeconomic data in the research on the relationship between the stock market and macroeconomic. Mostly use low frequency stock market data, such as monthly, quarterly or annual data. These approaches do not make full use of available financial data information, unavoidably importantly affecting the study. To make full use of the data information, it is necessary to use different frequency data in a model of the study. That is some variables using daily data, some using weekly or monthly data, this is the basic motive that why the study is based on the mixed frequency data. This paper mainly studies two aspects of the stock market volatility based on mixed frequency data. One is about the stock market volatility forecast, the other is the impact of macroeconomic on the stock market volatility. That is, the paper focuses on the prediction of stock market volatility and the economic root of the stock market volatility.Firstly, this paper uses the autoregressive distributed lag mixed data sampling regression model (ADL-MIDAS) and GARCH model to predict weekly volatility of the stock market separately. ADL-MIDAS model uses the mixed data (daily data and weekly data) of the Shanghai composite index and the Shenzhen component index, while GARCH model just uses weekly data. We make regression of the two models and compare the in-sample predicting properties of two models. Next, we compare the out-sample prediction accuracy of the two models in the period of the rest time through the non-nested models comparison test which was proposed by West in 2006. The result shows that there is no definite evidence to prove that ADL-MDDAL model is better than GARCH without using nowcasting. However, ADL-MEDAS model outperforms GARCH model on both the in-sample properties and out-sample prediction after the adoption of nowcasting, based on the data of Shanghai composite index and the Shenzhen component index. And we also make a series of robustness analysis to show that the conclusion is bound.Then, this paper uses the GARCH-MIDAS to seperate the compositions of the stock market volatility respectively. On this basis, we directly incorporate the macroeconomic variables in the above model to study the relationship between stock market volatility and macroeconomic. In order to ensure the robustness and convincingness of the conclusion, we establish a multiple regression model on the long-run volatility of the stock market and some macroeconomic variables to explore the relationship between the stock market volatility and macroeconomic. The empirical results show that the GARCH-MIDAS model based on the RV can resolve the short-term and long-term component of Chinese stock market well and that there is a significant relationship between macro economy and stock market volatility. When considering the most macroeconomic factors, the current value of the macro-economic factors affect the stock market volatility significantly.