Font Size: a A A

The Impact Of Investor Attention And Sentiment On Fund Returns

Posted on:2024-05-16Degree:MasterType:Thesis
Country:ChinaCandidate:C Y ShiFull Text:PDF
GTID:2568307073976559Subject:Applied statistics
Abstract/Summary:PDF Full Text Request
Behavioral finance suggests that asset prices in the securities market are not only determined by the intrinsic value of the security,but are also influenced to a large extent by investor behavior and psychology.In the Internet era,a large amount of market information and investors’ behavioral information is generated every day,and this information brings a fresh approach to the measurement of investor concerns and sentiment.In this thesis,we obtain the content of investor postings under the China Merchants CSI White Wine Index Fund in the fund bar of the daily fund website through web crawler technology,and obtain a total of about 580,000 posting data from January 1,2018 to June30,2022 after data pre-processing,and then construct an investor concern index and an investor sentiment index based on the number of monthly postings and the sentiment expressed by the monthly posting content,respectively.At the same time,this thesis also constructs a comprehensive sentiment index based on six commonly used sentiment proxies through the principal component analysis method.The investor sentiment index is constructed by first manually labeling the dataset,then training classification models through algorithms such as Random Forest,Na?ve Bayes,Support Vector Machine and Text CNN,and using the model with the best classification effect to predict the sentiment tendency of investors’ monthly posting content,and finally constructing an investor sentiment index based on the number of investors’ monthly posting as "positive" and "negative".The Text CNN model with Word2 vec as the word embedding among the classification models has the highest prediction accuracy,reaching 80.70%.In this thesis,we construct VAR models for two indexes of investor attention and investor sentiment together with liquor fund returns and investigate the leading lag relationship between the three variables.The results of the study find that fund returns are Granger causes of investor sentiment and investor sentiment is Granger cause of investor attention,which means that investor sentiment leads investor attention and fund returns lead investor sentiment.Both investor attention and sentiment show a short-term positive impulse response to fund returns,decaying to zero at about a six-period lag.The reason for this is that when fund returns change,investor sentiment tends to change at the same time,while investor attention takes a while to respond due to the transmission of information.Finally,by comparing the investor sentiment index with the composite sentiment index,it is found that the investor sentiment index constructed through textual big data has a more pronounced effect on liquor returns,verifying the effectiveness of the textually constructed investor sentiment index in measuring investor sentiment in individual fund areas.
Keywords/Search Tags:Investor attention, Investor sentiment, Fund returns, Text classification, Vector autoregressive model
PDF Full Text Request
Related items