Tweet Sentiments and Crowd-Sourced Earnings
Estimates as Valuable Sources of Information around
Earnings Releases

Abstract

In this article, the authors examine the confluence of two important financial social media databases—Estimize and iSentium. Both databases capture crowdsourced information that appears increasingly more important for financial market research. In particular, the authors investigate the event of the earnings announcement. First, they confirm that crowdsourced/Estimize’s consensus earnings are slightly more accurate than Wall Street’s consensus earnings, a result that has been robust over the past two years (2013–2014). Second, the authors document that the objectivity of the crowd is one reason why it is more accurate. Wall Street’s consensus is biased due to the lowballing phenomenon that is pervasive in the industry. Wall Street’s consensus earnings are 65%–68% lower than the actual reported earnings, whereas the crowd’s consensus is 52%–54% lower. Third, the authors find economically and statistically significant evidence that tweet sentiment contains distinct information that is not contained in the traditional preannouncement variables such as forecasts error, earnings surprise, bias, coverage, track record, and earnings volatility. Fourth, they show that tweet sentiment prior to the earnings announcement date can predict postannouncement risk-adjusted excess returns over the short term (i.e., a few days). This predictive relationship holds even in the presence of the earnings surprise variable. Fascinatingly enough, the market quickly incorporates this information, and the statistical significance of this relationship wanes after only a few days. The authors estimate that gross of costs, the alpha from tweet sentiments post–earnings announcement may be as high as approximately 10%–20% per year.

TOPICS: Fundamental equity analysis, big data/machine learning, information providers/credit ratings, performance measurement