Investor Sentiment Based on Naive Bayes Method and Its Impact on Stock Idiosyncratic Risk
With the vigorous development of web2.0 and the rapid development of text mining technology,dif-ferent subjects create and share content on social media.In the process of information dissemination and shar-ing,the information structure of the stock market has been changed,which further strengthens the sentiment tendency of investors and ultimately affects their decisions.Investors driven by sentiment will induce noise trad-ing and cause changes in stock prices,which is manifested as changes in stock idiosyncratic risk.However,the influence of investor sentiment on idiosyncratic risk is rarely mentioned.Therefore,by crawling the real-time post content of the individual stock bar of Oriental Fortune.com,the daily investor sentiment of individual stock is constructed using the Naive Bayesian method.A panel regression model is used to investigate the influence of investor sentiment on idiosyncratic risk.The research is supplemented on the impact of investor sentiment and the influencing factors of idiosyncratic risk,which is helpful for a deeper understanding of the influence mechanism of investor sentiment on stock risk from the perspective of a high frequency,and has certain guid-ance and reference value for investors,listed companies and regulators.The results show that(1)Investor sentiment in the lagging period and the current period all have a significant positive impact on idiosyncratic risk(IR),indicating that idiosyncratic risk increases when investor sentiment tends to be optimistic.(2)The effect of investor sentiment on idiosyncratic risk varies with the change of the difficulty of stock valuation.Stocks with a lower book-to-market and analyst tracking number are more difficult to value,and investor sentiment has a greater impact on idiosyncratic risk.(3)Considering the effect of arbitrage restriction,with the increase in the degree of short-selling restrictions,the effect of investor sentiment on idiosyncratic risk is significantly enhanced.After a series of robust tests,the above conclusions are still robust.
NETL
NSTL
万方数据
