The Prediction And Impact Of Different Media Sentiment On Stock Returns

With the development of information processing technology and big data,various media platforms have become an important link connecting market participants’ emotional cognition and behavioral decision-making.Media public opinion causes changes in investor sentiment,which affects investment behavior and ultimately impacts the formation of asset prices.Nowadays,the diversification of media platforms and the rapid flow of information have created conditions for exploring the influence mechanism of media opinion on stock prices.Based on big data text processing technology and machine learning method,this thesis discusses the influence of media sentiments on stock return from the perspective of big data text emotion of different media sources,to provide an important theoretical basis for the influence of media sentiments on asset pricing.At the same time,this study provides very important practical significance for further improving asset allocation and risk management and maintaining the healthy development of the capital market.More importantly,in the context of the continuous emergence of new media and the sign of trouble with media platforms that may cause large fluctuations in stock prices,this thesis combines group psychology,information communication,behavioral finance,and other disciplines to explore the influence mechanism of asset prices.Firstly,based on the sentiment classification method of media big data,this thesis constructs three types of media sentiment indices for the Chinese stock market,including social media sentiment index,traditional newspaper sentiment index and Internet news sentiment index,and further discusses the following questions:(1)Do the newly constructed three types of media sentiment indices have different time series effects on stock market returns? Is there any difference in the predictive ability of the three media sentiments? Can the aggregate sentiment index extracted by dimensionality reduction technology improve the predictive accuracy of stock returns?(2)What are the differences in the influence of the three types of media sentiments on the crosssectional return of stocks? In portfolio analysis,do three types of media sentiments have different impacts on portfolio future returns? Are there differences in the performance of the three media sentiments in the Fama-Macbeth cross-sectional regression?(3)In order to better explain the differential influence of media public opinion on stock returns,this thesis attempts to explore the flow trend among the three types of media sentiment information,and give a reasonable explanation for the heterogeneous influence of media sentiments on stock returns.Firstly,among the newly constructed sentiment indices of social media,traditional newspapers and Internet news,this study tries to find out the media sentiment indices with the best performance in predicting stock market returns.The evidence shows that the social media sentiment and Internet news sentiment index have excellent forecasting ability for stock returns,which is far better than macroeconomic predictors.However,the prediction effect of newspaper sentiment index based on traditional newspapers is not ideal.There are also significant differences in the estimation results of different business cycles.In the bull market stage,the social media sentiment index has a good ability to predict stock returns,while in the bear market stage,the Internet news sentiment index can significantly improve the prediction accuracy of stock returns.Secondly,the aggregate sentiment indicators extracted by three dimensionality reduction techniques PCA,PLS and s PCA(scaled PCA)contain more incremental information about Chinese stock market returns relative to popular method forecasting models and macroeconomic variables.At the same time,from the perspective of asset allocation,it is found that the aggregate sentiment indicators generate significant and considerable economic value.Thirdly,the portfolio analysis and Fama-Macbeth regression model are used to explore the cross-section effects of social media sentiment,newspaper news sentiment and Internet news sentiment on stock returns.The empirical evidence suggests that social media sentiment has a significant effect on the cross-section returns of stocks.Small,growth and illiquid stocks are more likely to be driven by social media buzz.However,neither Internet news sentiment nor newspaper news sentiment has prominent price discovery ability for cross-section stock returns.By discussing the heterogeneous influence of three types of media sentiment on capital price,this thesis further expands the research on the media effect on cross-section stock returns.The conclusion also highlights the powerful price-discovery power of social media at the individual stock level.Finally,in order to explain the differential impact of target media sentiment on asset prices,this thesis further discusses the trend of information flow among the three media sentiments.Compared with Internet news and traditional newspaper sentiments,social media opinion plays a “dominant” role in the Chinese stock market,consistent with the market characteristic of the high proportion of retail investors.The extreme Granger causality test results show that in the short term,slight fluctuations of social media sentiment cause excessive positive and negative sentiment on Internet news,while in the long term,the overheated social media sentiment will lead to fluctuations of Internet news sentiment.At the same time,extremely negative Internet news opinions quickly spread to social media,confirming the proverb that “bad news travels fast”.In this study,heterogeneous media sentiment is incorporated into the influencing mechanism of stock prices,and the forecasting performance of different media sentiments on stock returns is investigated,which complements and expands the research on the return predictability of the Chinese stock market.The research on the predictability of the media sentiment indices and aggregate sentiment indicators on the Chinese stock market provides practical support for promoting the application of advanced technology in forecasting stock returns.Meanwhile,this thesis comprehensively discusses the influence mechanism of stock returns from time series and cross sections,and broadens the research boundary of the media effect theory of the stock market.From the perspective of information dissemination,this thesis explores the information flow trend and communication effect between different media platforms and provides a new idea for the effective governance of China’s stock market.