Do US Stock Markets Typically Overreact to Corporate News Stories?

Abstract:     
It is widely believed that once news is made public the information is fully reflected in prices within at most a day or two (the efficient market hypothesis). We test this idea using the set of 245,429 Wall Street Journal corporate news stories from 1973 to 2001. Using computational linguistics methods we classify the stories according to topic, and for each topic with a sufficient number of identified events, we run an event study. Our results differ from popular impressions in several ways. 1. On average there is a reversal or overreaction, so that pre-event and post-event abnormal returns have the opposite sign. 2. Statistically significant return momentum is observed for many days after publication. 3. As a result, the inference to be drawn from an event study is often very sensitive to the assumed event window. 4. The average news story has a bigger and more prolonged impact during a recession than during an expansion.

Keywords: stock returns, news, event study, overreaction, publication bias

JEL Classifications: G12, G14

Working Paper Series