Empirical Analysis Of The Impact Of Stock Events On Abnormal Volatility Of Stock Prices

With the increasing popularity of the Internet and the growing stock market in China,people are getting more and more information about stocks from the Internet.At present,the stock market has more than 3,600 shares and continues to increase with the normalization of new stock issuance.It is more and more impossible to classify stock announcement information completely by manpower.Whether different information has the same impact on stocks has long been a problem for investors.Is it not easy to find out whether different stocks respond to the same information in different environments or different kinds of stocks? More and more shareholders and stock analysts hope that there is a way to help them automatically separate useful information from the stock market,and find that news announcements affect the law of stock prices.In this paper,the above problems are studied and attempted from two aspects.Firstly,this paper tries to classify the major events of individual stocks by text categorization.Secondly,through the event study method combined with hypothesis test,this paper analyses how news events affect the abnormal volatility of stock prices.In this paper,a total of 87 stocks were collected from January 1,2014 to March 26,2019,with 20728 text data.After self-defined word segmentation and self-defined stop word library to stop words,a total of 16602-dimensional word bag is obtained.In this paper,we modify the TF-IDF algorithm according to the characteristics of text,change IDF as a reverse index into a positive index,and use the modified algorithm to calculate the weight of feature words to get the feature vector.In this paper,K-means clustering is used to cluster texts,which are clustered into 13 categories,and each category of keywords is mapped according to its weight.From the cloud chart,we can see that nine of the 13 categories are well interpretable,they are "stock issuance category","stock dividend category","stock pledge and de-pledge category","performance forecast category","stock suspension and re-license category","increasing and reducing shareholdings category","company financial report category","stock price abnormal fluctuation reminder category" and "asset restructuring category".Through the empirical analysis of event study method,this paper finds that "dividend-sharing event","performance forecasting event","asset reorganization event" and "financial reporting event" have no significant impact on the abnormal fluctuations of stock prices.And the reason is that the message itself has two sides.In view of the "increase or decrease" event,this paper finds that the market reacts more intensely to the event of increase or decrease than to the event of increase or decrease.Through variance analysis,this paper observes whether the abnormal volatility caused by five kinds of information is significant in the three cases of "bull market","bear market" and "shock market",and draws the conclusion that the abnormal reaction of stock price to information is more intense in the bear market.The difference is used to analyze whether the abnormal volatility caused by five kinds of information is significant in three different types of stocks,namely "small market value","medium market value" and "large market value".It is concluded that the response of small and medium market value stocks to information is more intense than that of large market value stocks in most cases.In view of the above conclusions,this paper gives reasons for speculation or proof,and lists real examples in the market to support.