Investor Sentiment Predicts Stock Return

Abstract:To more accurately quantify the true feelings of investors for individual stock, we try to explore big data mining method about the social networking platform-micro-blog. Thus building a individual stocks’ micro-blog sentiment index and texting its ability to predict stock return.In order to ensure that each stock is mentioned more frequently on the micro-blog each day, we pick out the top48stocks in CSI100constituents. The posted interval is January1,2013to June30,2013. Using micro-blog capture tool, we gain537789micro-blogs referring to sample stocks. Through sentiment propensity analysis tool, we analyze every micro-blogs’ emotion and ultimately construct individual stocks’ micro-blog sentiment indicator. By correlation analysis between this indicator and consumer confidence index, we find they are significantly correlated. Besides, this emotion indicator shows convergence trend to the Shanghai Composite Index.Through the basic statistical analysis, each sample stock has59micro-blogs on daily average bases. Grouping sample stocks into ten groups according to the extent of micro-blogs’ emotion, we find that the minimum emotion set of stocks show low stock return, whereas high emotion set of stocks show high stock return, indicating that micro-blog sentiment indicator and stock return are positive related.Further analying the impact of micro-blog sentiment index on individual stock return, we perform regression analysis. Using mixed data regression model, the multiple regression results show micro-blog mood and individual stock return are significantly positive correlated at the1%level, which verifies the assumption that micro-blog emotion positive predict the stock return. In addition, this predictive ability still exists in a relatively long period(in three days). In perspective of time series data, the granger causality test results of micro-blog sentiment sequence and stock return sequence exhibit that micro-blog mood is the granger cause of stock return when lag stage are3and5. All of this means that individual stocks’micro-blog sentiment index has a certain ability to predict stock return.