Informational Content of CEO Tweets and Stock Market Predictability

This paper shows that CEO tweets contain informational content on the U.S. stock markets and provide investors with value-relevant information on predicting the stock price movement. We create a large, unique sample of CEO users on Twitter, extract hashtags and sentiments that can be used as features for prediction from large, unstructured tweet text, and construct hashtag and sentiment time series data. To prove the stock market predictability of CEO tweets using machine learning, we predict three numeric stock market indicators as a regression problem and the direction of stock prices as a classification problem. Findings confirm that the select list of hashtags and sentiments have predictive power on the stock return, trading volume, volatility, and stock price direction. We also find that the predictive power of CEO sentiments still stands after controlling for well-known macroeconomic and financial variables