Stock Index Volatility Forecast Method Improve

The stock market is an important part of China's capital market.Since the early 90s of last century,China has established two stock markets in Shanghai and Shenzhen,and the stock market has become the vane of economic development.With the continuous improvement of China's financial market mechanism,the reasonable control of the risk of stock price rise or fall has become the main focus of the community,especially for investors.Volatility is a measure of the uncertainty of the stock market returns,unable to understand the volatility means grasping and measuring the risk of the market will be so hard,not to mention carrying out effective asset allocation.The CSI 300 index which is an important basis to judge and study the fluctuation trend of stock prices and a barometer of China's economic development reflects the economic strength of our country to a certain extent.Therefore,the prediction of the CSI 300 index has important practical significance.Considering the experience of China's stock market,which started from a bull market to a mad cow market,however,finally evolved into the stock market crash,the volatility of the stock market shows that the prediction accuracy of stock market volatility needs to be further improved.Because the financial market volatility is difficult to observe directly and to forecast the volatility of stock returns,especially our country where such a financial market is still not mature,the stock price is affected by external factors,such as policy and so on.So establishing a forecasting model consistent with the actual situation of China's financial market to improve the prediction accuracy of volatility,not only conducive to boost investor confidence,but also play an important role in financial regulation to prevent stock market from changing radically in advance.The measurement of financial market volatility is the basis for studying the risk of financial market.Generally speaking,there are two ways commonly used to measure volatility,one is implied volatility,and the other is historical volatility.Among them,the implied volatility is less used for the fluctuation measure,because it substitutes the derivative transaction price into the model to retrodict volatility.While the historical volatility mainly forecasts the volatility from the historical trend of the market,which is a more commonly used method.The measurement method of historical volatility is mainly used for data prediction of small sample size,and the use of data is mainly based on low frequency data,such as month,week and day.But in the financial markets,the fluctuations in the stock market information is continuous,only using discrete data would cause the loss of information,while using the high frequency data can capture more financial information,and loss less information,furthermore forecast the fluctuation of financial market more accurately.Typically,transaction data calculated at hourly,minute,or per second is called high frequency data.High frequency data contain more information than low frequency data.Using high-frequency data to analyze the volatility of financial market can reflect the law of financial stock market volatility more clearly.Secondly,the improvement of the degree of information technology makes it easier to collect high-frequency financial data.With the help of high frequency data analysis and prediction,investors can make immediate decisions in time.At present,the basic model for the prediction of stock index volatility based on high frequency data is mainly the HAR-RV model and the HAR-RV-CJ model where the volatility RV has been further differentiated into jumping fluctuation JV and continuous wave CV on the basis of jumping,but the two models have not taken overnight information into consideration.Through the analysis,this paper finds that overnight information can affect the volatility of the stock market.Although in recent years,many scholars have paid more attention to the effect of the overnight information on financial market fluctuations.On this basis,this paper expands the contents of the original overnight information.Taking into account the source of overnight information,this paper selects three representative types of data-macro policy indicators(such as the deposit reserve ratio,purchasing manager index,etc.),overseas market trading information(such as the West Texas Intermediate base crude oil price,London gold fixed price,etc.)and the information disclosure of listed companies.And it is considered in the prediction of the volatility of the stock index.This paper take the factors of overnight information into consideration based the classical model HAR-RV created by the predecessors,and combined with the HAR-RV-CJ model where realized volatility is further divided into jump volatility and continuous volatility,we create HAR-RV-CJ-inf model.And we compares the prediction accuracy of HAR-RV model,HAR-RV-C J model and HAR-RV-CJ-inf model,using CSI 300 Index containg from September 1,2014 to 31 December 2016,580 trading days of the 5 minute high-frequency data as the research data in all.Based on the SPA test method,we use the four kinds of loss function to compare the prediction accuracy of the HAR-RV,HAR-RV-CJ,and HAR-RV-CJ-inf in the short,medium and long term volatility The p value of SPA test shows that the prediction accuracy of HAR-RV-CJ-inf model is better than that of HAR-RV model and HAR-RV-CJ model in short term,medium term and long term volatility.This conclusion will provide not only more accurate estimates for investors when making portfolio decisions,but also the reference significance for macroeconomic analysis.