The Research On The High-frequency Data Volatility Of The Chinese Stock Market

In recent years, with the rapid development of financial market, trading in the marketin the short period is more and more frequent and the trading volume is also growing, theresearches using of low-frequency data in the past are difficult to meet the needs of thedevelopment of financial market. So, gradually, people began to shift to the high-frequencydata fields where have more sophisticated requirements for the time scale. Typically,high-frequency data is divided into two categories: one is time interval intraday data with ahigher frequency, which is known as traditional high-frequency data; the other istransaction by transaction and recorded by second, which is known as the UHF (Ultra HighFrequency data). With the higher collection and containing a large number of marketinformation, the high-frequency data is an important factor in the microscopic structure ofthe market research. Therefore, the research of financial market high-frequency data canhelp to understand the microscopic structure of the financial market，and also can guidemarket investors to invest，to be the monitoring tools for market oversight bodies. In thispaper study, we begin with the background of high-frequency data, taking HS300index asthe sample, and then do the research of high-frequency data of the Chinese stock marketvolatility mainly in three parts.In the first part, we first describe the characteristics of high frequency time series data,and then collect the intraday data with the interval of1minute,5minutes,10minutes,15minutes,30minutes,60minutes to analyze the statistical characteristics of thehigh-frequency data of market microstructure and study the intraday pattern of the yield.The result show that the distribution of micro-market return is non-normality, and thehigher collection frequency, the more significant non-normality; and also the result verifythe return of high-frequency time series has a clear intraday pattern.Subsequently, we study the volatility characteristics of high-frequency data in thesecond and third parts, with considering the yield volatility in the spatial direction andduration volatility in the time direction. In the second part, we use the ARCH model to study volatility characteristics of high-frequency data with equal time interval in the spatialdirection. After model fitting and comparison, we conclude that: EGARCH (1,1) modelcan fit the volatility of stock index returns in China’s stock market better, and at the currentsituation, the influence on return of the good news is greater than the bad news.In the third part, we introduce the ACD model and research of the duration of UHFdata with unequal time interval in the time direction. The ACD model takes transactionduration as a marked point process, which can effectively solve the problem of modeling ofthe UHF data. Finally, after model fitting and comparison, we conclude that: EACD(1,1)can well study the transaction duration of autoregressive situation, and in the securitiesmarket, the transaction duration has strong clustering phenomena and the impact of thepresent transaction duration on the future duration may be decreasing as exponential form.In this paper, we study the Chinese stock market volatility from the perspective of thehigh-frequency data, mainly has the following innovation:(1) Comparison of characteristics of multiple frequency interval frequency time series,and then study the market microstructure changes with the increase in the frequency ofdata collection.(2) Based on the spatial direction and time direction, we analyze the characteristics ofthe Chinese stock market volatility from a three-dimensional perspective.(3) Using the ACD model to analyze the transaction duration of UHF data, afterComparison and research of the model, we obtain the mode that fitting of Chinese tradingduration.