An Empirical Study Of China’s Three-Factor Model Based On Chinese Investor Sentiment Index

Behavioral finance believes that due to differences in risk preferences,personality preferences,knowledge and experience,investors will have different expectations for the risks and returns of investment portfolios.Such error-prone expectations are called investor sentiment.In recent years,many scholars have studied the impact of investor sentiment on stock returns,and most of them have found that investor sentiment has an impact on stock returns,and it has a more significant impact on small-cap companies.In previous studies,the investor sentiment indicators selected by scholars are mainly objective indicators such as the number of IPOs and turnover rate,and some have added subjective indicators such as the consumer confidence index on the basis of objective indicators,but these indicators are not comprehensive.Reflect changes in investor sentiment.With the rapid development of Internet finance,stockholders and netizens are highly overlapped.More and more scholars are beginning to use deep learning to measure emotions based on text big data on the Internet.They believe that the emotions constructed with this method are relatively more objective than objective indicators accurate.The National Development Institute of Peking University and Percentage Co.,Ltd.built the Chinese Investor Sentiment Index(CISI)based on hundreds of millions of financial texts on the Internet reflecting investor sentiment and using deep learning methods.Jianan Liu et al.(2019)are based on the Fama-French three-factor model(FF-3),and aiming at the specific conditions of the Chinese market,they use the method to eliminate the 30%of the smallest stocks in the Chinese market to reduce the impact of shell value pollution.The book-to-market value ratio(BM)factor was replaced with an earnings-to-market price ratio(EP)factor,and finally a Chinese version of the three-factor model(CH-3)that fits China’s conditions was constructed,and it was found that CH-3 is more applicable to China’ s stock market than FF-3 has better applicability in the Chinese stock market.This article adds the Chinese investor sentiment index CISI based on text big data to CH-3 as the sentiment factor and the information condition respectively,in an attempt to improve the explanatory ability of the model.And observe whether the impact of investor sentiment based on text big data measurement on stock excess return is consistent with the conclusions of previous studies that scholars have used objective indicators to measure investor sentiment.This article first uses the data of the A-share market to construct CH-3 and perform regression testing on it.Secondly,on the basis of CH-3,a four-factor model was constructed with CISI as the mood factor,and 25 scale-EP independent variable combinations were used for regression testing.Then this paper selects the largest 5 combinations,10 combinations,15 combinations,20 combinations,and all 25 combinations to conduct GRS tests on the models.Finally,this article also adds CISI as an information condition to CH-3 for regression testing.By comparing the effects of these three models,conclusions can be drawn First,CH-3 is applicable to the Chinese A-share market,and the market premium factor,scale factor,and profit-to-market ratio factor are all significant.The overall model has a strong explanatory power.But the explanatory power for large-scale combination is weak.Second,adding CISI as an emotional factor to CH-3 improves the overall explanatory power of the model,and mainly improves the explanatory power of the model for large-scale combinations.After removing the small-scale combinations,it is found that the explanatory power of the four-factor model has improved more significantly than that of the CH-3,especially for the 20%portfolio with the highest market value in the stock pool,the explanatory power of the four-factor model has improved greatly.Third,taking CISI as the information condition model.From the regression results,the overall CISI×MKT factor(information condition factor)is not significant,but it has a significant impact on the excess return rate of large-scale portfolios.The phenomenon found in the model of the emotion factor is consistent.From the results of the GRS test,the explanatory power of the model with information conditions has not improved,but has declined,indicating that using CISI as an information condition is not as effective as an emotional factor.Fourth,in the past studies on selecting objective indicators to represent investor sentiment,most of them found that investor sentiment has a more significant impact on small market capitalization companies.However,this article selects CISI constructed based on textual big data to represent investor sentiment.The results show that investment The influence of investor sentiment on large market capitalization companies is more significant,which is the opposite of previous research.This may be caused by the different indicators selected to represent investor sentiment.Objective indicators reflect more trading behavior and cannot fully reflect sentiment.This can also explain to a certain extent,investor sentiment measured based on text big data is very different from investor sentiment measured by objective indicators.