Modeling And Forecasting Realized Covariance Based On Overnight Information

With the development and improvement of transportation and computer science and technology industry,the financial market in China has gone through a rapid growth in the recent 20 years.The global financial markets has also shown a trend of integration between different markets.And it has not only inspired more investors poured into the capital market,but also built a bridge of information communication among different capital markets.So that more and more information from foreign markets could be cumulated during the non-trading time of China stock market,which is called the overnight information in our paper.It also seems that the study of realized covariance between different assets is still in the early stage of development.try to go deeply in this area is vary meaningful and necessary.In this paper,based on the perspective of empirical analysis,using the HAR models,we systematically studied the different time span of the overnight market information,as well as the overnight assets information on assets,and try to make an effort on the prediction of covariance matrix among different assets.We also presents the statistical test,which is called SPA and put forward by other scholars,and the economical test,based on then optimal portfolio(minimum variance)approach,to test whether the addition of overnight information into HAR models is a good decision.And based on the characteristics of HAR model,this paper analyzes the contribution of overnight information on daily covariance,the weekly average covariance and he monthly average covariance.We found that the investors reflect sensitively both on the trading day overnight information and the holiday overnight information,the impact time period is not the same,the overnight information between trading day has a quick effect on the volatility of next trading day,but the holiday overnight information release slowly and affect the whole month on average volatility in the future.By the comparison of SPA testing and related parameters of portfolio,we found that by adding market overnight information,or assets overnight information into the volatility and covariance timing model,the power of prediction becomes stronger.We also find that investors react to the overnight information of their holding assets more sensitively,and the reflection to the market overnight information of investors depends on the market of overnight information is good news or bad news.