Analyst Underreaction to Past Information About Earnings: reporting, processing or plain old misspecification bias?

Abstract

We revisit the debate concerning the interpretation given to prior year’s earnings changes in predicting future earnings as discussed by Abarbanell & Bernard (1992), Francis & Philbrick (1993) and Easterwood and Nutt (1999). We advance a new specification of this relationship which distinguishes between earnings reversion and momentum. On a large UK dataset, we find there is substantial underreaction, particularly in situations of earnings momentum, approximately six times as large as that identified by Abarbanell & Bernard. This suggests that analysts behaviour is still a candidate to explain post earnings announcement drift. We also show that our model performs well relative to a specification recently proposed by Easterwood and Nutt (1999).