Premium
Investor Overreaction to Earnings Surprises and Post‐Earnings‐Announcement Reversals
Author(s) -
Bathke Allen W.,
Mason Terry W.,
Morton Richard M.
Publication year - 2019
Publication title -
contemporary accounting research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.769
H-Index - 99
eISSN - 1911-3846
pISSN - 0823-9150
DOI - 10.1111/1911-3846.12491
Subject(s) - earnings , post earnings announcement drift , uncorrelated , economics , earnings response coefficient , negative correlation , quarter (canadian coin) , monetary economics , autocorrelation , positive correlation , financial economics , econometrics , finance , mathematics , medicine , statistics , archaeology , history
Prior literature suggests that the market underreacts to the positive correlation in a typical firm's seasonal earnings changes, which leads to a post‐earnings‐announcement drift (PEAD) in prices. We examine the market reaction for a distinct set of firms whose seasonal earnings changes are uncorrelated and show that the market incorrectly assumes that the earnings changes of these firms are positively correlated. We also document that positive (negative) seasonal earnings changes in the current quarter are associated with negative (positive) abnormal returns in the next quarter. Thus, we observe a reversal of abnormal returns, consistent with a systematic overreaction to earnings, rather than the previously documented PEAD. Additional analysis indicates that financial analysts similarly overestimate the autocorrelation of these firms, although to a lesser extent. We also find that the magnitude of overestimation and the subsequent price reversal are inversely related to the richness of the information environment. Our results challenge the notion that investors recognize but consistently underestimate earnings correlation and provide a new perspective on the inability of prices to fully reflect the implications of current earnings for future earnings. That is, we show that investors predictably overestimate correlation when it is lacking, but underestimate it when it is present.