Research On The Influence Of Investor Sentiment On Stock Returns

The stock market is often regarded as a barometer reflecting the national economic situation,which is an important factor in analyzing and predicting the economy.The fluctuation of stock prices can sometimes reflect the capital operation and market views of industries or enterprises in a country or region to a certain extent.It is an important reference for analyzing and predicting the economy.Also,the proportion of individual investors in the Chinese stock market is relatively high,making it easier for investor sentiment to lead to certain irrational investment behaviors.With the development of mobile Internet,more and more investors collect stock information online.As one of the Internet social media,the stock bar is an important place for individual investors to express and exchange their opinions.Therefore,this paper selects the information in stock bar to construct investor sentiment indexes.This paper takes 5.85 million post information of the constituent stocks of the SSE 50 Index from January 1,2017 to December 31,2021 as a research sample,and uses a Bi-LSTM to classify emotions.Then,this paper construct weekly investor sentiment indicators,investor attention indicators,and investor opinion convergence indicators,and use fixed-effect regression,Granger causality test,heterogeneity test,robustness test to explore the relationship between stock yield and investor sentiment across different emotional states and industries in bear or bull market,before or after the COVID-19.The research shows:(1)There is a significant positive correlation between stock yield and investor sentiment,a significant negative correlation between investor opinion convergence index and stock yield,and investor attention has no significant effect on stock yield.(2)Relationship between stock yield and investor sentiment is reciprocal causation,and the causal relationship between investor opinion convergence and stock yield is unidirectional.(3)Investor sentiment in different market situation,before or after the epidemic,different emotional states,and different industries has different effects on stock yield respectively.Investor bearish sentiment has a greater influence.Investor sentiment has a greater influence in bear markets.Investor sentiment has a greater influence under the COVID-19 pandemic.Investor sentiment has a greater influence in the IT sector.(4)Compared with the strategy based on investor sentiment index,the strategy based on investor concern index has higher risk and higher return,and both performances are higher than the benchmark strategy.(5)After comparing different sentiment classification methods,it is found that among the classification methods such as LSTM,Bi-LSTM,SVM,and Naive Bayes,Bi-LSTM performs the best in the training samples of this paper,with an accuracy rate of 82.27%.