A Re‐examination of Analyst Under‐reaction

• Date: 01 October 2013
• DOI: http://doi.org/10.1111/ajfs.12029
• Published date: 01 October 2013
• Author: Su Jeong Lee,Bok Baik

A Re-examination of Analyst

Under-reaction*

Bok Baik**

College of Business, Seoul National University

Su Jeong Lee

College of Business, Seoul National University

Received 6 December 2010; Accepted 24 May 2013

Abstract

Prior studies suggest that analysts under-react to past information (Klein, 1990; Abarbanell

and Bernard, 1992; Easterwood and Nutt, 1999). We examine whether evidence interpreted

as analyst under-reaction to negative news can be attributed to analyst self-selection. Analyst

self-selection arises when analysts’ information production efforts are focused toward stocks

that analysts expect to perform well. Using a large sample of firms for the period of 1983–

2004, we find evidence suggesting that self-selection is responsible, at least in part, for analyst

under-reaction to past negative information about the firm’s performance.

Keywords Analyst under-reaction; Self-selection; Cognitive bias; Earnings forecasts; Financial

analysts

JEL Classification: G14,M41

1. Introduction

This paper re-examines the stylized facts that are interpreted as evidence for analyst

under-reaction to unfavorable news and suggests one underlying reason for seem-

ingly irrational analyst reaction to negative news. We propose that part of evidence

interpreted as apparent analyst under-reaction to past negative news is a manifesta-

tion of self-selection (i.e. analysts’ tendency to withhold negative reports in order to

curry favor with management). More specifically, we argue that a positive relation

*Acknowledgments: We appreciate helpful comments by two anonymous reviewers and An-

wer Ahmed, Bruce Billings, Hemang Desai, Bruce Johnson, Terry Marsh, Rick Morton, Kathy

Petroni, Mort Pincus, Tom Rothenberg, and Brett Trueman. Particularly, we are grateful to

Reuven Lehavy for his encouragement and insightful comments. Baik acknowledges financial

support from the Institute of Management Research, Seoul National University. Any errors

are our own.

**Corresponding author: Bok Baik, College of Business, Seoul National University, Seoul,

Korea. Tel.: +82-2-880-2524, Fax: +82-2-878-3154, email: bbaik@snu.ac.kr.

Asia-Pacific Journal of Financial Studies (2013) 42, 724–754 doi:10.1111/ajfs.12029

724 ©2013 Korean Securities Association

between forecast errors and the negative information in prior stock returns, prior

earnings changes and prior forecast surprises is, in part, attributable to analysts’

self-selection bias.

Prior research indicates that investors do not fully reflect the implication of cur-

rent earnings for future earnings (Bernard and Thomas, 1989; Brown and Han,

2000). Particularly, several papers imply that analysts may be the source of investor

failure to appreciate fully the implications of current earnings for futur e earnings.

For example, prior studies report that forecast errors are positively associated with

past negative news, such as past negative earnings changes (Abarbanell and Bernard,

1992; Easterwood and Nutt, 1999; Ahmed et al., 2002), past negative returns (Klein,

1990; Lys and Sohn, 1990), and past negative forecast errors (Mendenhall, 1991).

Prior research interprets this association as support for the hypothesis that analysts

under-react to prior bad news. Given that forecast accuracy affects analysts’ com-

pensation and reputation (Gu and Wu, 2003; Hong and Kubik, 2003), it is puzzling

that analysts do not undo the systematic relation between forecast errors and past

information in forecasting future earnings.

Several studies, however, imply that this forecast bias could potentially result

from analysts’ economic incentives. McNichols and O’Brien (1997) and Hayes

(1998) document that analysts promote the trading of stocks, and some analysts

have incentives to choose not to say anything rather than release a low forecast for

those firms with unfavorable information. This self-selection likely leads to an over-

optimistic ex post observed distribution of forecasts, even though analysts issue ex

ante unbiased forecasts. Interestingly, analyst under-reaction seems to overlap with

self-selection. Prior research finds that forecast errors are positively associated with

past negative news. Self-selection is more severe for firms with negative information

(Hayes and Levine, 2000). While numerous papers provide evidence consistent with

analyst under-reaction (e.g. Easterwood and Nutt, 1999; Ahmed et al., 2002), few

academic studies investigate the underlying reasons for this apparent analyst irratio-

nality from an analyst incentive perspective. In this study, we assert that analyst

self-selection partially affects the association between forecast errors and past news,

resulting in an asymmetry in the association between good and bad news. We are

unaware of any paper that investigates whether the positive association between

forecast errors and prior information is a manifestation of analyst self-selection.

Using a large sample of 25 690 firm-years over the period 1983–2004, we find

evidence suggesting that prior evidence for analyst under-reaction to negative infor-

mation is partially attributable to analyst self-selection. We document that the asso-

ciation between forecast errors and negative prior returns becomes significantly

weakened once we control for self-selection, using our selection measures, TRUNC1

and TRUNC2. Specifically, we find that the positive relation between forecast errors

and negative prior returns decreases when the first selection measure (TRUNC1) is

controlled. Also, we find that the association is significantly mitigated using the sec-

ond measure (TRUNC2). We also draw similar conclusions from the analyses on

the positive relation between forecast errors and negative prior earnings changes

A Re-examination of Analyst Under-reaction

©2013 Korean Securities Association 725