Effects Behind Words: Quantifying Qualitative Information in Corporate Announcements

Abstract

We develop a sentiment metric to analyze the tone and information amount in financial corporate announcements. We improve existing text processing methods by developing a different word selection approach that allows quantifying the sentiment of financial announcements in an intuitive, but effective manner. Applying our approach to corporate announcements, we find, first, that our metric provides superior explanatory power for abnormal returns during the day of the announcement. Second, analyst EPS forecast revisions during the month after the announcement show a significant structural relation to relevant announcement text components based on Tonality. Third, observed short-term stock price reactions immediately following the announcement are less related to EPS forecast revisions than relevant text components, allowing a new perspective on analyzing post-announcement-drift based on direct cause-and-effect relation.

Keywords: sentiment analysis, announcement study, asset pricing, analyst forecast revisions

JEL Classification: G11, G12, G14, G17