Analysts’ and Investors’ Reactions to Consistent Earnings Signals

• Author: Peter M. Johnson,Marcus L. Caylor,Theodore E. Christensen,Thomas J. Lopez
• Published date: 01 November 2015
• Date: 01 November 2015
• DOI: http://doi.org/10.1111/jbfa.12163

Journal of Business Finance & Accounting

Journal of Business Finance & Accounting, 42(9) & (10), 1041–1074, November/December 2015, 0306-686X

doi: 10.1111/jbfa.12163

Analysts’ and Investors’ Reactions

to Consistent Earnings Signals

MARCUS L. CAYLOR,THEODORE E. CHRISTENSEN,PETER M. JOHNSON

AND THOMAS J. LOPEZ∗

Abstract: We investigate (1) whether the trajectory of the current-quarter earnings ex-

pectation path (defined by the signs of the forecast revision and the earnings surprise)

provides information about future firm performance, and (2) the extent to which analysts

and investors react to that information. Our results indicate that analysts underreact more to

earnings information revealed by consistent-signal earnings expectation paths than to earnings

information communicated by inconsistent-signal expectation paths. We also find that the

current earnings expectation path provides incremental explanatory power for future abnormal

returns, even after controlling for the sign and magnitude of the earnings surprise. Overall,

our evidence is consistent with underreaction stemming from analysts’ and investors’ bias in

processing the information in consistent-signal earnings expectation paths.

Keywords: analysts, earnings expectation path, forecasts, trading strategies, representativeness

bias, market efficiency

1. INTRODUCTION

Prior research suggests that analysts and investors do not fully incorporate past earn-

ings information into their expectations about future performance and market prices

(e.g., Lys and Sohn, 1990; Mendenhall, 1991; Abarbanell and Bernard, 1992; and

Ali et al., 1992). The extant literature indicates that a primary factor contributing to

market underreaction is the level of information uncertainty inherent in the earnings

signal (e.g., Miller, 2005; and Francis et al., 2007). In particular, this literature reveals

that unexpected earnings signals characterized as having less (more) information

∗The first author is from the School of Accountancy, Coles College of Business, Kennesaw State University,

Kennesaw, Georgia, USA. The second author is from the J.M. Tull School of Accounting, Terry College

of Business, University of Georgia, Athens, Georgia, USA. The third and fourth authors are from the

Culverhouse School of Accountancy, Culverhouse College of Commerce and Business Administration,

University of Alabama, Tuscaloosa, Alabama, USA. The authors would like to thank John Barrick, Larry

Brown, Michael Clement, Steve Glover,Scott Jackson, Grant McQueen, Rick Mendenhall, Mike Pinegar, Phil

Shane, Kay Stice, and workshop participants at BYU and the University of Alabama workshop for helpful

comments and suggestions. They are also grateful to Thomson Financial for providing earnings forecast

data, available through the Institutional Brokers Estimate System. These data have been provided as part of

their broad academic program to encourage earnings expectations research.

Address for correspondence: Peter M. Johnson, Culverhouse School of Accountancy, Culverhouse College

of Commerce and Business Administration, University of Alabama, Tuscaloosa, AL 35487, USA

e-mail: pjohnson@cba.ua.edu

C

2015 John Wiley & Sons Ltd 1041

1042 CAYLOR, CHRISTENSEN, JOHNSON AND LOPEZ

uncertainty have more robust contemporaneous (PEAD) returns. In addition, prior

research also finds compelling evidence that not all earnings surprises of the same sign

and of similar magnitude have the same ability to predict future performance (e.g.,

Bartov et al., 2002). In particular, Bartov et al. (2002) report that when the forecast

revision during the quarter and the earnings surprise at the end of the quarter provide

consistent good (bad) news signals (i.e., the forecast revision and earnings surprise

have the same sign), future earnings are higher (lower) than when the forecast revision

and the earnings surprise provide conflicting news (i.e., they have opposite signs).1

Our research has two primary objectives. First, we investigate whether analysts’ earn-

ings forecasts accurately reflect the predictable information about future performance

conveyed by the current earnings expectation path.2Second, we investigate whether

the current earnings expectation path is predictive of future stock returns and if

this predictive ability is incremental to the explanatory power of the post-earnings-

announcement drift (PEAD), price momentum, and revision return trading strategies.

Our investigation is motivated by three streams of research suggesting that earnings

expectation paths may differentially affect analysts’ forecasts of future earnings

and future stock returns. First, the earnings expectation path literature provides

evidence that consistent positive (negative) earnings signals lead to higher (lower):

(1) contemporaneous returns, (2) future earnings expectations, and (3) future

earnings realizations (Bartov et al., 2002; Miller, 2005, 2006; and Caylor et al., 2007).3

Second, prior experimental and empirical evidence indicates that behavioral biases in

capital markets create biased forecasts and market prices (e.g., Bernard and Thomas,

1990; Bernard, 1992; Elliott et al., 1995; and Barberis et al., 1998). Chan et al.

(2004) suggest that while consistency in the sequence of past performance might

induce representativeness (i.e., overreaction), a lack of consistency (or a consistent

sequence over a relatively short horizon) might lead to investor conservatism (i.e.,

underreaction).4Taken together, this literature predicts that earnings expectation

paths with consistent positive (negative) earnings signals will lead to higher (lower)

future earnings forecast errors and stock returns.

Finally, our research is motivated by the literature that examines the relation

between information uncertainty and market returns. Francis et al. (2007) report that

unexpected earnings signals characterized as having greater information uncertainty

(IU) have more muted initial market reactions and more robust post-earnings-

announcement returns. In other words, unexpected earnings signals characterized

as having low IU have more complete initial market reactions and less robust post-

earnings-announcement returns. Bartov et al. (2002) and Caylor et al. (2007) provide

1 In other words, an earnings surprise of $1.00 for one firm may not have the same predictive ability

for future performance as a $1.00 earnings surprise for another firm. Bartov et al.’s (2002) evidence

suggests that a primary contributing factor in the relation between the current earnings surprise and future

performance is the sign of the forecast revision prior to the earnings surprise.

2 We define the revision in analysts’ forecasts during the quarter as the last quarterly earnings per share

(EPS) forecast prior to the current-quarter earnings announcement minus the first quarterly EPS forecast

issued during the quarter following the prior quarter’s earnings announcement. We define the earnings

surprise as actual quarterly earnings per share minus the last quarterly analyst EPS forecast prior to the

current-quarter earnings announcement. Consistent with Bartov et al. (2002), we identify an expectation

path based on a combination of the sign of the forecast revision and the sign of the earnings surprise.

3 We define consistent earnings signals as situations in which the forecast revision earnings news and the

earnings surprise have the same sign.

4 This interpretation is consistent with Barberis et al.’s (1998) conceptual model where investor underreac-

tion is attributable to a conservative bias in processing information that runs counter to recent trends.

C

2015 John Wiley & Sons Ltd

REACTIONS TO CONSISTENT EARNINGS SIGNALS 1043

evidence suggesting that investors’ response to consistent signal expectation paths is

more complete than their response to inconsistent paths. Miller (2005) reports that

analysts make larger future earnings forecast revisions when preannouncements result

in consistent earnings news paths than when preannouncements result in inconsistent

paths. Prior research reports that firms with consistently increasing earnings have

significantly larger earnings response coefficients and contemporaneous market

returns (Barth et al., 1999; and Myers et al., 2007). Similarly, Lopez and Rees (2002)

report evidence that earnings response coefficients are significantly larger for firms

that consistently beat analysts’ earnings forecasts. The empirical results in each of these

studies suggest a more complete contemporaneous market response for intertemporal

earnings signals that are consistent in sign (i.e., lower information uncertainty).

We empirically investigate these competing theories to see which better explains

how investors and analysts use the information signaled in the earnings expectation

path. Consistent with Bartov et al. (2002), we find that consistent-signal “good news”

(“bad news”) expectation paths are associated with higher (lower) future profitability

than inconsistent-signal good (bad) news expectation paths.5In addition, our evidence

suggests that analysts appear to underreact more to the information in consistent-

signal expectation paths than inconsistent-signal expectation paths. Consistent with

this notion, we find that the probability of a positive (negative) forecast error in the

subsequent year is significantly greater for consistent-signal “good news” (“bad news”)

expectation paths than inconsistent-signal paths.

We also explore analysts’ efficiency in incorporating information in the current

earnings surprise following Abarbanell and Bernard (1992) and Abarbanell and

Bushee (1997). Consistent with Abarbanell and Bernard’s (1992) results, we find

significant autocorrelation in I/B/E/S earnings surprises. However, when we compare

consistent-signal path observations with inconsistent-signal path observations, we find

substantially higher autocorrelation in earnings surprises for consistent-signal versus

inconsistent-signal paths. This evidence suggests that analysts underreact to the

earnings information in the current earnings expectation path and the underreaction

is more pronounced for consistent-signal expectation paths. We also estimate annual

and quarterly versions of Abarbanell and Bushee’s (1997) Table 5 analyses. Overall, the

results from these analyses suggest that the information about future earnings changes

contained in consistent signal expectation paths is not subsumed by the information in

fundamental signals. More importantly, these results suggest that even though analysts

use the information in consistent signals in revising their forecasts, they do not use the

information in the signals efficiently.

Given these results, we investigate whether the current-quarter earnings expectation

path is predictive of future stock returns. To do so, we perform regression analyses

following Livnat and Mendenhall (2006) and find that a current-quarter, consistent-

signal earnings expectation path hedge portfolio trading strategy earns significant

abnormal returns that are incremental to trading strategies based on (1) the post earn-

ings announcement drift, (2) forecast revisions, and (3) price momentum. Specifically,

our regression results indicate that a hedge portfolio that is long in the highest decile

5 We define a good (bad) news earnings expectation path as a path that has a positive (negative) earnings

surprise (SURP). A consistent-signal good (bad) news expectation path results when the positive (negative)

SURP is preceded by a positive (negative) forecast revision (REV) during the quarter.On the other hand, an

inconsistent-signal good (bad) news expectation path results when a positive (negative) SURP is preceded

by a negative (positive) REV. Weexplain earnings expectation paths in more detail later.

C

2015 John Wiley & Sons Ltd