THE VALUE OF STOCK OPINIONS ON TWITTER: A CROSS-SECTIONAL STUDY OF ABNORMAL RETURNS DURING EARNINGS ANNOUNCEMENTS OF S&P500 FIRMS

Abstract

This paper examines the value of information contained in messages posted on Twitter. Focus is placed on the context of earnings announcements of S&P500 firms, where the relationships among tweet information, earnings surprise and abnormal returns were looked into. Apart from investigating the predictive power of pre-earnings announcement tweet information, the paper utilized a novel method in the field of financial studies and introduced a moderated mediation model to examine the role of tweet information during earnings announcement period. This model was analysed using Structural Equation Modelling and biascorrected bootstrapping. Two aspects of tweets were studied: twitter presence and tweet valence. Twitter presence was found to predict cumulative abnormal return (CAR) only when CAR was negative, lending support to negativity bias, where negative news tends to dominate. Tweet valence predicted CAR only when both pre- and during- earnings announcement period measures were applied. Mediation was found, where twitter presence and negative tweet valence mediated the relationship between earnings surprise and CAR when CAR was negative. The effects of earnings surprise on CAR was found to be indirect rather than direct. This pointed to strong associations between CAR and the mediators, twitter presence and negative tweet valence, that overshadowed the direct relationship between earnings surprise and CAR.