Description
We conduct an experiment with MBA students where we manipulate whether participants
are exposed to an analyst’s name in Stage 1, and whether they are given a cue in Stage 2
about the particular analyst’s prior performance as an All-star analyst. We find that in
the absence of a favorable performance cue about the analyst, mere exposure to the analyst’s
name enhances perceived analyst credibility, which in turn influences the investors’
earnings estimates. This suggests a benefit to analysts in terms of building credibility
merely through media exposure that cannot be explained by performance. In fact, a diagnostic
cue such as the analyst’s high prior performance no longer matters to investors once
they have prior exposure to the analyst’s name.
Judgment effects of familiarity with an analyst’s name
Wei Chen
a
, Hun-Tong Tan
b,?
a
Australian Business School, University of New South Wales, Sydney 2052, Australia
b
Nanyang Business School, Nanyang Technological University, Nanyang Avenue, Singapore 639798, Singapore
a b s t r a c t
We conduct an experiment with MBA students where we manipulate whether participants
are exposed to an analyst’s name in Stage 1, and whether they are given a cue in Stage 2
about the particular analyst’s prior performance as an All-star analyst. We ?nd that in
the absence of a favorable performance cue about the analyst, mere exposure to the ana-
lyst’s name enhances perceived analyst credibility, which in turn in?uences the investors’
earnings estimates. This suggests a bene?t to analysts in terms of building credibility
merely through media exposure that cannot be explained by performance. In fact, a diag-
nostic cue such as the analyst’s high prior performance no longer matters to investors once
they have prior exposure to the analyst’s name. However, this enhancement of an analyst’s
credibility through investors’ prior exposure to his/her name is reversed when the analyst’s
forecast turns out to be inaccurate.
Ó 2013 Elsevier Ltd. All rights reserved.
Introduction
In this study, we examine whether investors’ perception
on the analyst’s credibility and their earnings estimates
made in reaction to the analyst’s forecast are in?uenced
by the joint effect of prior exposure to the analyst’s name
and their awareness of the analyst’s prior performance.
Financial analysts and their reports often receive coverage
in the media. Such media coverage can increase the sal-
ience and, thus, familiarity to investors of those analysts.
In addition, media coverage of an analyst sometimes
includes the prior performance of the analyst (such as
All-Star status) but sometimes not (see Appendix A for
examples of such media coverage).
1
Thus, an analyst’s fore-
cast received by investors can vary in terms of two analyst
attributes: whether the analyst’s name has received prior
exposure by the investors, and whether the analyst’s prior
performance is made known to the investor.
0361-3682/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.aos.2013.02.001
?
Corresponding author. Tel.: +65 6790 4819; fax: +65 6793 7956.
E-mail address: [email protected] (H.-T. Tan).
1
For instance, some data sources/media (e.g. Starmine, a division of
Thomson Reuters providing equity research service; Yahoo Finance!;
Institutional Investor; Wall Street Transcript) always include the analysts’
prior forecast accuracy ratings with their names, but other media sources
do not always do so (e.g. PR Newswire reports on earnings conference calls
generally do not report other information about the analyst other than the
brokerage ?rm name). As evidence of this latter point, we randomly select
20 highly-exposed analysts whose media mention is higher than the mean
mention documented in Bonner, Hugon, and Walther (2007), half of whom
hold All-Star or All-American status and the other half have no such award.
We collect the media mentions, other than those from Institutional
Investor, Wall Street Transcript, and earnings conference calls (484 in total
without duplicates) for these 20 analysts in the year 2010 from the
FACTIVA database and code the information about the analyst accompa-
nying his/her name. There is only one instance where there is mention of
the award status (or related performance-related information) of the
analyst.
Accounting, Organizations and Society 38 (2013) 214–227
Contents lists available at SciVerse ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
The joint effect of these two analyst attributes on inves-
tors’ judgments has not been investigated, but is important
for a few reasons. Familiarity through mere exposure to an
analyst’s name per se (i.e., without any performance mea-
sures) is an irrelevant cue, and psychology research sug-
gests that this can positively affect investors’ judgments
(Zajonc, 1968). Following psychology literature (Zajonc,
1968), we use the term ‘‘mere exposure effect’’ to refer to
the enhancement of investors’ favorability (in terms of reli-
ance on the analyst’s forecast) towards an analyst through
mere exposure of the individual to the analyst’s name. To
the extent this mere exposure effect occurs in an invest-
ment setting, it has implications on how analysts can en-
hance their credibility with and impact on investors,
even without any change in their forecasting performance.
On the other hand, an analyst’s performance status (e.g.,
All-star status) is a relevant cue that is positively associ-
ated with forecast accuracy (Fang & Yasuda, 2009; Stickel,
1992), and should be relied upon. However, these analyst
attributes (prior exposure and performance status) are
sometimes present in isolation and sometimes at the same
time, and theory suggests that they have substitutive ef-
fects (Dhami & Harries, 2001; Gigerenzer & Gaissmaier,
2011; Gigerenzer & Goldstein, 1996). For instance, while
investors’ in?uence by an irrelevant cue such as their
familiarity with the analyst’s name may be an area of con-
cern, the presence of an analyst’s performance status (a
relevant cue juxtaposed with the analyst’s name) may
actually alleviate this concern as it can dominate this mere
exposure effect. However, the reverse is also possibly true.
That is, in the presence of prior exposure to the analyst’s
name, performance cue may no longer matter, which is
an area of concern.
Recent archival research provides some evidence of a
mere exposure effect, and indicates that the celebrity sta-
tus of analysts (measured by the quantity of media cover-
age analysts receive in the major media sources) positively
affects investors’ reactions to forecast revisions (Bonner
et al., 2007). However, they also ?nd that media coverage
is positively associated with ex ante forecast performance
such as All-Star status,
2
and thus, it is possible that media
coverage is proxying for analyst performance. As Table 1
shows, All-American/All-Star analysts are about three/two
times more likely to be in the high versus low media cover-
age group (22.60% vs. 6.94%; Pearson chi-square = 978.48,
p < 0.01/18.42% vs. 9.41%, Pearson chi-square = 341.77,
p < 0.01).
3
Furthermore, because the analyst’s name is some-
times accompanied by his/her prior performance in such
media coverage but sometimes not, it is unclear whether
the media coverage effect is due to investors’ exposure to
the analyst’s name, his/her prior performance, or an interac-
tion effect between the two. The authors acknowledge that
‘‘an alternative explanation of our results is that, consistent
with prior work in the area, market participants react more
strongly to forecast revisions issued by analysts with supe-
rior performance’’ (Bonner et al., 2007, p. 483).
We also examine whether the credibility enhancement
established by prior exposure persists when the analyst’s
actual performance disappoints investors. We suggest that
while there is a bene?t to the analyst from increased media
exposure, this bene?t is short-lived should the analyst
make a forecast error. This ?nding informs us about a pos-
sible negative effect of exposure on the analyst’s credibility
over multi-period interactions between analysts and inves-
tors. This ?nding is also important to other market partici-
pants such as managers, who may also consider exposure as
one way to enhance their credibility with investors.
We conduct an experiment to investigate these issues.
The key advantage of the experimental approach is that
it allows us to manipulate participants’ awareness of the
prior performance of the analyst. Although it is possible
to use archival data to measure both media coverage and
analysts’ prior performance (e.g., as in Bonner et al.,
2007), investors exposed to a high media coverage analyst
may or may not be aware of the analyst’s prior perfor-
mance. Therefore, it is dif?cult to assess whether the stron-
ger reactions to the celebrity analysts’ earnings forecast are
caused by their prior performance, investors’ exposure to
the celebrity analysts’ names, or both. In addition, using
the experimental approach enables us to hold constant
other information such as the strength of the analyst’s
arguments (Hirst, Koonce, & Simko, 1995). Other factors
determining the appearance of analyst’s name in the media
Table 1
A 2 Â 2 frequency table of media coverage and analyst’s perceived
performance.
a
Panel A:
Performance
[Institutional Investors All-American
Award]
No Yes
Media Coverage Low 9407 (93.06%) 701 (6.94%)
High 7760 (77.40%) 2266 (22.60%)
Panel B:
Performance
[Wall Street Journal All-Star Award]
No Yes
Media Coverage Low 9157 (90.59%) 951 (9.41%)
High 8179 (81.58%) 1847 (18.42%)
Number in parentheses refers to the percentage of observations within
each media coverage group.
a
This frequency table is based on the sample of Bonner et al. (2007)
(20,134 analyst-?rm-quarter observations) and shows the number of
observations (the percentage of observations within each media coverage
group) in each cell. Media coverage of the analyst is classi?ed as low/high
using median split. Performance of the analyst is measured by the award
status (Institutional Investors All-American award, Panel A; Wall Street
Journal All-Star award, Panel B).We also conduct the chi-square test of
independence and ?nd that media coverage is associated with both All-
American and All-Star award status (Pearson chi-square = 978.48/341.77,
p < 0.01 for both award status measures).
2
A recent working paper by Rees, Sharp, and Twedt (2011) using a
subset of the media that Bonner et al. (2007) cover (top 5 business press;
86% of Rees et al.’s (2011) sample comes from Financial Times and The Wall
Street Journal) also ?nds similar results for the positive correlation between
All-Star status and media coverage. They ?nd a positive association
(regression coef?cient of 1.277, odds ratio = 3.586) between media cover-
age and All-Star status in the logistic regression, suggesting that the odds of
being covered by the top business press increase by 258.6 percent with All-
Star status.
3
We appreciate the help of Beverly Walther in providing the analysis
included in Table 1.
W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227 215
press can also be controlled. For example, the media may
selectively report those analysts who are more active,
available, in?uential, or who provide eye-catching news.
Finally, the experimental method offers an opportunity to
test the cognitive mechanism by which investors react to
an earnings forecast provided by a familiar analyst.
We manipulate prior exposure of the analyst’s name
(zero, high) and favorable performance cue (absent, pres-
ent) in our between-subjects experiment. Participants are
MBA students, and we manipulate the exposure of the ana-
lyst’s name in stage one of the experiment by ?rst showing
all participants twenty slides containing nine names, but
where the target analyst’s name is mentioned six times in
one condition and none in the other condition. Participants
are merely shown the names with no other information
associatedwiththe names at this stage. We manipulate par-
ticipants’ awareness of the analyst’s award status by either
providing or not providing this favorable performance cue
along with the analyst’s quarterly earnings forecast in the
second stage of the experiment, where participants are told
that the analyst is a Wall Street Journal All-Star analyst or
they are not given any information about the analyst’s prior
performance. Finally, in the third stage of the experiment,
we provide participants with the actual earnings, which
indicate that the analyst has made a forecast error.
Results indicate that in the absence of a favorable per-
formance cue about the analyst, investors’ perceived ana-
lyst credibility is higher in the presence of prior exposure
to the analyst’s name than in its absence, and investors rely
more on the highly-exposed analyst’s forecast to make
their own earnings estimates. When the favorable perfor-
mance cue about the analyst is present, exposure to the
analyst’s name has no incremental effect on investors’ per-
ceived analyst credibility and earnings estimates. Further-
more, with prior exposure to the analyst’s name, investors’
perceived credibility assessments and earnings estimates
are no different with or without awareness of the prior
high performance of the analyst. This suggests that inves-
tors implicitly equate a familiar analyst name with a level
of high prior performance, or in the minimum, the analyst’s
high prior performance no longer matters given prior
exposure to his/her name. Perceived analyst credibility
mediates investors’ earnings estimates and willingness to
rely on the analyst’s future report. Finally, after the actual
earnings are announced and the analyst’s forecast turns
out to be inaccurate, investors lower their credibility rat-
ings on the highly-exposed (versus the non-exposed) ana-
lyst to a greater extent when the investor has not received
a prior favorable performance cue about the analyst.
This study contributes to the literature by partitioning
the effects of prior exposure to an analyst’s name from
investors’ awareness of the performance cue about the ana-
lyst. Our examinationof the interactionbetweenprior expo-
sure and performance cue (rather than a main effect of prior
exposure) allows us to contribute to the literature in three
important ways. First, our documentation of a prior expo-
sure effect in the absence of a favorable performance cue
provides evidence of a direct causal relation between prior
exposure to an analyst’s name and investors’ positive cred-
ibility judgments, which is hard to test using archival data.
Second, we ?nd that given prior exposure to the analyst
name, investors’ credibility judgments are no different
whether they are giventhe analyst’s high prior performance
or not. An inference then is that a diagnostic cue such as the
analyst’s highprior performance no longer matters to inves-
tors once they have prior exposure to the analyst’s name.
This is potentially worrisome given that, among high media
coverage analysts, the majority (82%/77%, Table 1) do not
have All-Star/All-American award status. Under these cir-
cumstances, it wouldappear bene?cial for investors to place
more reliance on this smaller group of high prior perfor-
mance analysts. However, our results show that they do
not do so. Third, by including the favorable performance
cue as a moderator, we ?nd that there is no prior exposure
effect in the presence of a performance cue (i.e., investors
are aware of the analyst’s All-Star status). This suggests a
boundary condition to the media exposure effect.
In addition, we extend the literature by showing that
the mechanism by which prior exposure to an analyst’s
name in?uences investors’ earnings judgments is via their
assessments of the analyst’s credibility. Our results show
that analysts can enhance their perceived credibility sim-
ply by increasing their media exposure frequency, even if
it pertains to their names only.
However, while an analyst’s credibility can be enhanced
when investors have prior exposure to his/her name, this
prior exposure effect is short-lived when the analyst’s fore-
cast turns out to be inaccurate. We show that the analyst’s
credibility takes a bigger hit when the investor has prior
exposure to his/her name. This effect occurs because when
the analyst’s forecast is inaccurate, investors realize that
prior enhanced credibility is misplaced.
In the next section of the paper, we review the related
literature and develop the hypotheses. We then describe
the research design and experiment procedure, analyze
the experimental results, and conclude the paper.
Hypothesis development
Effects of exposure and performance cue on investors’
judgments
Zajonc (1968) documents that mere exposure of the
individual to a stimulus enhances his/her favorability to-
ward it. Mere exposure refers to a condition in which prior
exposure makes the given stimulus more accessible to the
individual’s memory and in?uences his/her perception. Za-
jonc (1968) manipulates the exposure frequency of non-
sense Chinese characters and asks participants to guess
their meanings on a good-bad scale. He ?nds a signi?cantly
positive exposure-favorability relationship. Following Za-
jonc (1968), other psychology studies ?nd that increased
exposure results in favorable evaluations of a variety of
properties such as liking, goodness, attractiveness, and
pleasantness measures (Bornstein, 1989). This suggests
that credibility assessments will also be likewise positively
in?uenced by prior exposure.
The perceptual ?uency model offers an explanation for
this mere exposure effect (Bornstein &D’Agostino, 1994; Ja-
coby, Toth, Lindasy, & Debner, 1992; Mandler, Nakamura,
Shebo, & Zandt, 1987). This model is based on the concept
of ‘‘perceptual ?uency,’’ which refers to the ease with which
216 W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227
people perceive, encode andprocess information. According
to this model, prior exposure to a stimulus results in the for-
mationof a perceptual representationof the stimulus. When
people are asked to evaluate the previously-exposed stimu-
lus, the perceptual representation of the stimulus is acti-
vated and facilitates the encoding and processing of the
stimulus and enhances the perceptual ?uency for that stim-
ulus. Perceptual ?uency serves as the basis for the feeling of
familiarity in the sense that stimulus perceived more
quickly tends to be identi?ed as familiar (Johnston, Dark, &
Jacoby, 1985). People often attribute the cause of the per-
ceptual ?uencytothe stimulus’ status onthe dimensionthat
is salient at that moment. Such attribution of perceptual ?u-
ency to the exposed stimulus produces higher ratings onthe
stimulus. For example, Mandler et al. (1987) ?nd that when
participants are asked to evaluate the stimuli in the particu-
lar dimensions (such as preference, brightness or darkness),
exposed irregular shapes are perceived as preferred, bright-
er or darker than the unexposed shapes.
Psychology research suggests that people can automati-
cally encode frequency information (Zacks, Hasher, & Sanft,
1982) and that this process uses a minimal amount of atten-
tional effort (Hasher & Zacks, 1979). Prior studies in the
auditing and accounting literature have investigated the ef-
fect of frequency of exposure to relevant cues. For instance,
Butt (1988) ?nds that for both students and auditors, fre-
quency judgments based on direct experience (by repeated
individual presentations of ?nancial statement errors) are
more accurate than those based on indirect experience (by
receiving summary data). Joe (2003) ?nds that auditors
whoreceiveboththe?nancial statements andapress release
of a client’s debt default (which is the repeated information
that has beendisclosedinthe ?nancial statements notes ear-
lier) perceive a client’s bankruptcy probability to be higher
than in a situation when there is no press coverage. In an
experiment, Hugon(2004) ?nds that redundancyof ?nancial
performance informationis positively associated with infor-
mationcredibility. All these studies investigate the effects of
repeating decision-relevant information, but not decision-
irrelevant information (such as an analyst’s name).
Most related to this study is the archival study by Bon-
ner et al. (2007). They de?ne a celebrity as a famous or
well-publicized person, being well-known in addition to
his performance-related qualities (Bonner et al., 2007, p.
482). They use the quantity of media coverage analysts re-
ceive as their empirical proxy for celebrity and measure to-
tal media coverage as the number of appearances of an
analyst’s name associated with his brokerage house em-
ployer in all media sources. Bonner et al. (2007) ?nd that
investors react more strongly to the celebrity analysts’ re-
vised earnings forecasts. They also ?nd a positive associa-
tion between media coverage and an analyst’s ex ante
performance, and obtain the same ?ndings after including
the analysts’ performance as control variables in their
regression models.
4
However, an analyst’s name is some-
times accompanied by his/her prior performance status in
these media coverage, and Bonner et al. (2007) do not exam-
ine whether investors are aware of the analyst’s prior
performance, nor how this knowledge interacts with prior
exposure to the analyst’s name.
Based on the mere exposure effect, we expect that the
mere prior exposure of the name enhances investors’ per-
ceptual ?uency for the name. When the name appears in
the press release as the source of the earnings forecast,
investors will view the person (the analyst) more favorably
in terms of credibility. Higher perceived analyst credibility
increases the believability of the earnings forecast issued
by the analyst (Hirst, Koonce, & Venkataraman, 2007;
Williams, 1996). In turn, investors are more likely to rely
on the highly-exposed analyst’s forecast, such that inves-
tors’ own earnings estimates are more in?uenced by the
analyst’s forecast.
Prior research also suggests that investors’ reactions to
analysts’ (management) earnings forecasts depend on the
analysts’ (management) prior forecast accuracy or factors
associated with current forecast accuracy (Bonner, Wal-
ther, & Young, 2003; Clement & Tse, 2003; Hirst, Koonce,
& Miller, 1999). Therefore, we expect that investors’ reac-
tions to the analyst’s earnings forecast will be stronger
when the analyst is known to be an All-Star analyst.
5
Psychology research suggests that these two factors—
prior exposure to the analyst’s name (absent, present)
and favorable performance cue (absent, present)—have
substitutive effects. In particular, the one-reason decision
making model provides a detailed discussion about how
people make judgments with multiple cues (Dhami & Har-
ries, 2001; Gigerenzer & Gaissmaier, 2011; Gigerenzer &
Goldstein, 1996). This model suggests that people make
decisions or judgments based on a single cue, instead of
integrating all available information. In other words, multi-
ple cues substitute for each other, and are not compensa-
tory. Previous research on persuasion also ?nds a
substitutive effect between source credibility and other
source variables (Joseph, 1977; Maddux & Rogers, 1980;
Pornpitakpan, 2004). For example, Joseph (1977) shows
that physical attractiveness has little effect on participants’
preference when the source is known as an expert, while
the effect of attractiveness is signi?cant for a non-expert.
Maddux and Rogers (1980) document that in terms of per-
suasiveness of an argument, the superiority of the expert
to the non-expert is greater when the source is unattrac-
tive than when the source is attractive.
As we mentioned earlier, we expect that either prior
exposure to the analyst’s name or awareness of the ana-
lyst’s prior good performance leads investors to perceive
the analyst to have higher credibility, which also leads
them to rely more on the analyst’s earnings forecast when
making their own earnings estimates. The one-reason deci-
sion making model and related studies discussed above
suggest that these two factors are substitutes, such that a
4
They do not examine whether there is an interaction effect between
media coverage and performance. Statistics textbooks (e.g., Kutner,
Nachtsheim, Neter, & Li, 2005) argue that once there is a signi?cant
interaction effect between two variables, any main effect of one variable
has to be interpreted conditional on the other effect.
5
We focus on a favorable performance cue (i.e., the analyst’s award
status) in this study. The reason is that the analysts and the media are
generally reluctant to disclose the performance for bad performers (for
example, Yahoo! Finance only displays the performance ratings for those
top performers; i.e. those analysts whose ratings are higher than average).
W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227 217
pronounced mere exposure effect (from prior exposure to
the analyst’s name) exists in the absence of the perfor-
mance cue about the analyst; this effect is likely to be sub-
dued when a performance cue about the analyst
accompanies reports containing the analyst’s views. Our
hypothesis is formally stated below.
Hypothesis 1. Prior exposure to an analyst’s name has a
more positive effect on investors’ analyst credibility
assessments and their reliance on the analyst’s forecast
when a favorable analyst performance cue is absent than
when it is present.
The one-reason decision model also suggests that the
positive effect of the performance cue is likely to be sub-
dued when investors have prior exposure to the analyst’s
name. This indicates a potential adverse effect on investors
arising from prior exposure to the analyst’s name in that
relevant cues such as favorable prior performance of the
analyst are under- (not) weighted with prior exposure to
the analyst’s name. Our hypothesis is stated below:
Hypothesis 2. A favorable analyst performance cue has a
more positive effect on investors’ analyst credibility
assessments and their reliance on the analyst’s forecast
when the investor has not had prior exposure to the
analyst name than when the investor has had prior
exposure.
Effects of exposure and performance cue on investors’
judgments after actual earnings announcement
Hypotheses 1 and 2 relate to positive effects accruing to
the analyst from investors’ prior exposure to his/her name
or a favorable prior performance status. Our next research
question relates to whether the positive effects accruing to
the analyst due to this prior name exposure or a favorable
prior performance persist. We do so in a setting where the
analyst’s actual current performance is known and turns
out to be inaccurate. InHypothesis 1, we predict that the po-
sitive effect of prior exposure is stronger in the absence
(than in the presence) of a favorable performance cue about
the analyst. Whenthe analyst’s forecast turns out to be inac-
curate, the presence of forecast error reminds investors that
their greater con?dence and reliance placed earlier on the
highly-exposed analyst are not warranted; the prior en-
hanced credibility assessments will be correspondingly re-
duced. This downward adjustment is particularly greater
where the favorable performance cue is absent since the
prior exposure effect is initially stronger here. Similarly, in
Hypothesis 2, we predict that the positive effect of a favor-
able performance cue is stronger in the absence of prior
exposure to the analyst’s name than in its presence. Again,
the presence of a forecast error leads to an undoing of the
initial enhanced con?dence and reliance placed on the ana-
lyst witha prior favorable performance cue. Again, the effect
is magni?edwhen there is no prior exposure to the analyst’s
namesince the prior performance cue effect is stronger here.
Our hypotheses are formally stated as follows.
Hypothesis 3. Prior exposure to an analyst’s name pro-
duces a more negative reaction to a forecast error when the
investor has not received a favorable analyst performance
cue than when the investor has received a favorable
analyst performance cue.
Hypothesis 4. Prior exposure to a favorable analyst perfor-
mance cue produces a more negative reaction to a forecast
error when the investor has not had prior exposure to the
analyst’s name than when the investor has had prior expo-
sure to the analyst’s name.
Fig. 1 provides a graphical summary for Hypotheses 1
and 2 (panel A) and Hypotheses 3 and 4 (panel B).
Method
Participants
We conduct an experiment with 89 M.B.A. students
from a major Singapore university as proxy for non-profes-
sional investors (Elliott, Hodge, Kennedy, & Pronk, 2007).
6
Panel A: Graphical summary of Hypotheses 1 and 2
Panel B: Graphical summary of Hypotheses 3 and 4
Zero High
Perceived
Analyst
Credibility
Exposure
Performance Cue Present
Performance Cue Absent
Zero High
Change in
Perceived
Analyst
Credibility
Exposure
0
Performance Cue Present
Performance Cue Absent
Fig. 1. Effects of exposure and performance cue on investors’ judgments.
Notes: Panel A plots predicted results associated with Hypotheses 1 and 2,
which predict that perceived analyst credibility is a joint function of
exposure and performance cue. Panel B plots predicted results associated
with Hypotheses 3 and 4, which predict that after the actual earnings
announcement, the change in perceived analyst credibility is a joint
function of exposure and performance cue.
6
We dropped one participant whose responses indicated that he
misread the case (i.e., initial earnings estimate relative to consensus
forecast is both directionally opposite to all other participants’ estimates,
and extreme at 3.2 the standard deviation). When we include this
participant in the analysis, the contrast test for H1&H2 is signi?cant for
the credibility measure (F = 12.50, p < 0.01) but not for the change in
earnings estimates measure (F = 1.40, p = 0.12). The contrast test for H3 and
H4 is signi?cant for the change in perceived credibility measure (F = 3.20,
p = 0.04).
218 W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227
The participants have a mean (median) working experience
of 7.34 (5.58) years. On average, the participants have taken
2.66 (3.66) accounting (?nance) courses. Seventy-three
(Fifty-six) percent of the participants have investment expe-
rience in stock (fund) market. Each student is paid twenty
Singapore dollars for participating in the experiment.
Design
We employ a two-way (exposure; performance cue) be-
tween-subjects design to test the hypotheses. The ?rst
independent variable is the exposure frequency of the tar-
get analyst’s name (zero, high). The second independent
variable is the analyst’s favorable performance cue (absent,
present). In the performance cue present condition, the
target analyst name is accompanied by a description in
the press release that the analyst is a ‘‘Wall Street Journal
All-Star analyst.’’ In contrast, in the performance cue ab-
sent condition, only the target analyst name is provided
in the press release, so participants are not aware of the
All-star status of the analyst.
Procedure
Participants are randomly assigned to the treatment
conditions. At the beginning of the experiment, partici-
pants are told that they will complete two studies; a visual
study followed by an earnings forecast study. The reason
we choose to label the ?rst part a visual study (where
the exposure frequency of analyst’s name is manipulated)
is to avoid any possible demand effects on participants.
In the visual study, participants are told to focus on the
screen and watch the slide show. Twenty slides containing
nine different English names (?rst name plus last name)
are presented one by one in the projector screen at two
seconds interval. These nine names include both the target
(alternative) analyst’s name in the high (zero) exposure
condition and eight other ?ller names. The target analyst’s
name is the name whose exposure frequency is manipu-
lated, and is shown as the analyst who makes the earnings
forecast in the press release in the main experiment for all
conditions. In the high exposure condition, the target name
appears a total of six times out of twenty slides in the slide
show, whereas in the zero exposure condition, an alterna-
tive name replaces the target name in the corresponding
positions in the slide show and the target name never ap-
pears in the visual study.
In selecting the target, alternative and ?ller names, we
follow several rules. First, English names are used because
the largest brokerage ?rms are U.S. companies, and the use
of English names helps to increase the external validity of
the results. Second, we choose the middle-ranked frequent
last and ?rst names from the U.S. Census Bureau 1990 to
remove possible semantic implication of the name. Finally,
we choose ?rst names that are usually used as male names,
since most analysts are male (Green, Jegadeesh, & Tang,
2009). In the experiment, we use ‘‘Devon Fraley’’ as the tar-
get analyst’s name. In the high exposure condition, ‘‘Devon
Fraley’’ appears six times in the slide show while in the
zero exposure condition, ‘‘Kirby Sikora’’ replaces ‘‘Devon
Fraley’’ in the corresponding positions; the other names
are all the same for the two conditions. In order to avoid
any possible recency or primacy effect, the ?rst and the last
name shown are all ?ller names.
After the slide show, participants are asked two ques-
tions unrelated to current experiment to clear participants’
short memory. Then, participants are told to put aside the
materials for the visual study and continue to do the
earnings forecast study. After reading the background
information about a listed manufacturing company ‘‘Theta
Inc.,’’ participants are shown the 5-year ?nancial summary
for the period 2004–2008. This is followed by the quarterly
earnings for the past two years and for the ?rst two quar-
ters of ?nancial-year 2009, and consensus EPS forecast for
the third quarter of 2009 ($0.18) as well as the consensus
12-month EPS forecast for 2009 ($0.80). Participants are
asked to provide their earnings forecasts for the third quar-
ter and full year of 2009, the con?dence in their earnings
estimates, and evaluations of Theta’s earnings growth
potential and stock appreciation potential in the next
12 months.
Participants then proceed to open Envelope A, which
contains the press release issued on August 31, 2009 about
the analyst’s earnings forecast for Theta’s third quarter
earnings. In the performance cue absent condition, partic-
ipants read the press release stating that ‘‘(a)nalyst, Devon
Fraley, estimates Theta’s earnings per share for the third-
quarter ending September 30, 2009 will be $0.20.’’ In the
performance cue present condition, participants read the
same statement containing the analyst’s forecast, along
with the information that the analyst is a Wall Street Jour-
nal All-Star analyst.
We set the analyst’s forecast to be good news compared
to the consensus analysts forecast ($0.18) and the same
quarter realization in the prior year because this is consis-
tent with prior research that investors expect analysts to
issue good news forecast (Hirst et al., 1995), and good
news earnings forecast may be less credible (Hutton, Mill-
er, & Skinner, 2003). Therefore, in this scenario, the role of
an analyst’s credibility is critical to enhance the believabil-
ity of his/her earnings forecast (Hirst et al., 2007; Mercer,
2004, 2005; Williams, 1996).
After reading the press release, participants are asked to
give their updated earnings forecasts for the third quarter
and full-year 2009, their con?dence in the forecasts, and
future earnings growth and stock price appreciation poten-
tials in the next 12 months. In addition, participants are
asked to indicate their willingness to rely on the analyst’s
report in the future, the analyst’s competence, trustworthi-
ness, reputation, their expectation that the analyst’s fore-
cast to be accurate, the believability of the forecast, the
likelihood that the analyst was intentionally misguiding
the market. All these questions are assessed by 11-point
scales (0 = Extremely Low and 10 = Extremely High), with
the exception of the con?dence question, where the scale
is from 0.0 to 1.0.
After the participants make their assessments of the
analyst based on the press release, they are asked to put
all materials to Envelope A and open Envelope B, which
contains the actual earnings announcement stating that
W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227 219
‘‘(o)n October 14, 2009, Theta announces that the earnings
per share for the third-quarter ending September 30, 2009
are $0.16.’’ Participants are then asked to evaluate Theta’s
full-year EPS for the year 2009, earnings growth potential
and stock price appreciation potential, their willingness
to rely on the analyst’s report in the future, the analyst’s
absolute and relative competence, trustworthiness, reputa-
tion and the likelihood that the analyst was intentionally
misguiding the market. Following this, participants
complete other post-experimental questions including
the manipulation checks and demographic questions about
their working and investment experience.
Results
Manipulation checks
To check whether the high exposure frequency
increases the familiarity of the target analyst’s name, in
the post-experimental questionnaire, participants are
asked to rate their familiarity with the analyst’s name on
an 11-point scale ranging from 0 (not familiar at all) to
10 (completely familiar). Participants in the high exposure
condition rate the analyst name mentioned in the press
release to be signi?cantly more familiar (mean = 6. 84)
than those in the zero exposure condition (mean = 2.67;
F = 47.29, p < 0.01).
7
Seventy-two percent of all participants
correctly answer the question on whether the analyst men-
tioned in the press release is a Wall Street Journal All-Star
analyst. We include all participants in our analyses and
results are similar if we exclude those who fail the manipu-
lation checks.
In our post-experimental questionnaire, we also ask
participants to indicate whether the press release men-
tioned the speci?c analyst’s name and whether the ana-
lyst’s name was (was not) shown in the visual study.
Seventy-three percent of the participants in the high expo-
sure condition correctly remember that the press release
mentioned the speci?c analyst’s name, while only 36% of
the participants in the zero exposure condition correctly
remember this (F = 13.94, p < 0.01). All (59%) of the partic-
ipants in the high (zero) exposure condition correctly rec-
ognize that the analyst’s name was (was not) shown in the
visual study (F = 28.59, p < 0.01). This suggests that the
exposure frequency in?uences participants’ attention to
the analyst name in the press release and recognition of
the analyst.
Test of Hypotheses 1 and 2
Perceived analyst credibility and earnings estimates
Prior literature (Hovland, Janis, & Kelley, 1953; Mercer,
2004, 2005) documents that credibilityhas twodimensions:
competence and trustworthiness. We take the average of
participants’ ratings on the analyst’s competence and trust-
worthiness as a single composite credibility measure (Cron-
bach’s alpha = 0.81). Table 2 summarizes the results of
participants’ perceived analyst credibility after they receive
the press release about the analyst’s earnings forecast.
8
The
ANOVA results indicate signi?cant main effects of exposure
(p = 0.04) and performance cue (p < 0.01), along with a signif-
icant interaction effect (p = 0.07).
As an overall test of the ordinal interaction implied in
Hypotheses 1 and 2 (Buckless & Ravenscroft, 1990; Rosnow
& Rosenthal, 1995), we conduct a contrast test with the
weights À3 in the ‘‘zero exposure + performance cue absent’’
condition, +1 in the ‘‘high exposure + performance cue ab-
sent’’, ‘‘zero exposure + performance cue present’’, and ‘‘high
exposure + performance cue present’’ conditions. The con-
trast test is signi?cant (F = 12.65, p < 0.01, Panel C, Table 2).
9
Results are consistent with Hypothesis 1. Speci?cally,
when participants have no information about the perfor-
mance cue, they perceive the highly-exposed analyst to
be more credible (mean = 5.69) relative to the non-exposed
analyst (mean = 4.98, F = 5.11, p = 0.01). Further, when the
analyst is known to be an All-Star analyst, exposure of the
analyst’s name has no signi?cant effect on perceived ana-
lyst credibility (mean for the zero/high exposure condi-
tion = 5.95/6.00, F = 0.02, p = 0.44; panel C Table 2).
Results are also consistent with Hypothesis 2. We ?nd
that with no prior exposure to the analyst’s name, partici-
pants attend to the high prior performance of the analyst in
that credibility assessments are higher with the presence
rather than absence of a performance cue (F = 9.82,
p < 0.01). In addition, with prior exposure to the analyst’s
name, participants appear to pay short shrift to prior per-
formance information. Their credibility assessments do
not differ whether the performance cue is present or ab-
sent (F = 1.00, p = 0.16). This suggests that prior exposure
to the analyst’s name has an impact on participants’ judg-
ments such that a diagnostic cue such as prior performance
information has no incremental effect. In fact, we also ?nd
no difference in participants’ credibility between the ‘‘high
exposure + performance cue absent’’ condition and the
‘‘zero exposure + performance cue present’’ condition
(p = 0.20). This suggests that mere exposure to the ana-
lyst’s name offers the same positive reputational and abil-
ity-to-in?uence impact as good prior performance.
We focus onparticipants’ quarterlyearnings estimates to
examine their reliance on the analyst’s forecast for the third
quarter. We expect that higher perceived analyst credibility
makes the analyst’s forecast more believable, which in turn
7
The p-value ?gures are all one-tailed for directional predictions, unless
otherwise indicated.
8
We ask participants to indicate their perceptions on the analyst’s
competence/trustworthiness/reputation in absolute (an 11-point scale) and
relative terms (compared to the peers). Factor analysis suggests that these
questions load to two factors and we label them as perceived absolute/
relative analyst credibility. Following prior research (e.g., Barton & Mercer,
2005; Cianci & Kaplan, 2010; Mercer 2005), we use the perceived absolute
competence and trustworthiness to construct the perceived analyst
credibility measure in the main analysis. We obtain similar results when
we use perceived relative analyst credibility or perceived absolute/relative
analyst reputation measures. Results for all tests of hypotheses are similar
after controlling for the participants’ working experience, and number of
accounting and ?nance courses they have taken.
9
We also test alternative contrast weights À3/+1/0/+2 in the ‘‘zero
exposure + performance cue absent’’/‘‘high exposure + performance cue
absent’’/‘‘zero exposure + performance cue present’’/‘‘high exposure + per-
formance cue present’’ condition. The contrast test is signi?cant for the
perceived analyst credibility (F = 11.14, p < 0.01) and change in quarterly
earnings estimates (F = 1.73, p = 0.10).
220 W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227
leads tohigher investors’ reliance onthe analyst’s forecast.
10
In our experiment, the target analyst’ earnings forecast is
optimistic compared to the consensus forecast. Therefore,
higher reliance on the target analyst’s forecast implies higher
third-quarter earnings estimates. Participants are asked to
provide their earnings estimates twice, before and after the
press release about the analyst forecast. Therefore, we use
the change in the participants’ third-quarter earnings esti-
mates as the dependent variable. ANOVA results (Table 3, pa-
nel B) indicates a signi?cant interaction effect (p = 0.05). As
with the credibility measure, the contrast test with the same
weights (À3/+1/+1/+1) is also signi?cant (F = 2.67, p = 0.05).
We ?nd a signi?cant exposure effect on the change in partic-
ipants’ third quarter earnings estimates in the absence of the
performance cue (p = 0.04), and an insigni?cant exposure ef-
fect in its presence (p = 0.29). A favorable performance cue
has a signi?cant effect on the change in participants’ earnings
estimates in the absence of prior exposure to the analyst’s
name (p = 0.08), but not in its presence (p = 0.17). No signi?-
cant difference in earnings estimates is found between the
‘‘high exposure + performance cue absent’’ condition and
the ‘‘zero exposure + performance cue present’’ condition
(p = 0.33). In sum, Hypotheses 1 and 2 are supported for both
credibility and earnings estimates measures.
11
Process model tests
We rely on structural equation analysis to verify that
the exposure effect on investors’ judgments and decisions
is due to the underlying mechanisms we propose. We posit
that exposure and performance cue jointly in?uence
Table 2
Perceived analyst credibility after the analyst’s forecast.
Panel A: Mean perceived analyst credibility (standard deviation) [rank]
a
Performance cue
c
Exposure
b
Absent Present Row Mean
Zero 4.98 (1.19) [6.09] 5.95 (1.12) [7.91] 5.47 (1.25) [7.00]
N = 22 N = 22 N = 44
[Condition 1] [Condition 3]
High 5.69 (0.94) [7.38] 6.00 (0.86) [8.00] 5.86 (0.90) [7.71]
N = 21 N = 24 N = 45
[Condition 2] [Condition 4]
Column mean 5.33 (1.12) [6.72] 5.98 (0.98) [7.96]
N = 43 N = 46
Panel B: ANOVA results
Source df Mean square F p-Value
Intercept 1 2840.08 2654.83
We conduct an experiment with MBA students where we manipulate whether participants
are exposed to an analyst’s name in Stage 1, and whether they are given a cue in Stage 2
about the particular analyst’s prior performance as an All-star analyst. We find that in
the absence of a favorable performance cue about the analyst, mere exposure to the analyst’s
name enhances perceived analyst credibility, which in turn influences the investors’
earnings estimates. This suggests a benefit to analysts in terms of building credibility
merely through media exposure that cannot be explained by performance. In fact, a diagnostic
cue such as the analyst’s high prior performance no longer matters to investors once
they have prior exposure to the analyst’s name.
Judgment effects of familiarity with an analyst’s name
Wei Chen
a
, Hun-Tong Tan
b,?
a
Australian Business School, University of New South Wales, Sydney 2052, Australia
b
Nanyang Business School, Nanyang Technological University, Nanyang Avenue, Singapore 639798, Singapore
a b s t r a c t
We conduct an experiment with MBA students where we manipulate whether participants
are exposed to an analyst’s name in Stage 1, and whether they are given a cue in Stage 2
about the particular analyst’s prior performance as an All-star analyst. We ?nd that in
the absence of a favorable performance cue about the analyst, mere exposure to the ana-
lyst’s name enhances perceived analyst credibility, which in turn in?uences the investors’
earnings estimates. This suggests a bene?t to analysts in terms of building credibility
merely through media exposure that cannot be explained by performance. In fact, a diag-
nostic cue such as the analyst’s high prior performance no longer matters to investors once
they have prior exposure to the analyst’s name. However, this enhancement of an analyst’s
credibility through investors’ prior exposure to his/her name is reversed when the analyst’s
forecast turns out to be inaccurate.
Ó 2013 Elsevier Ltd. All rights reserved.
Introduction
In this study, we examine whether investors’ perception
on the analyst’s credibility and their earnings estimates
made in reaction to the analyst’s forecast are in?uenced
by the joint effect of prior exposure to the analyst’s name
and their awareness of the analyst’s prior performance.
Financial analysts and their reports often receive coverage
in the media. Such media coverage can increase the sal-
ience and, thus, familiarity to investors of those analysts.
In addition, media coverage of an analyst sometimes
includes the prior performance of the analyst (such as
All-Star status) but sometimes not (see Appendix A for
examples of such media coverage).
1
Thus, an analyst’s fore-
cast received by investors can vary in terms of two analyst
attributes: whether the analyst’s name has received prior
exposure by the investors, and whether the analyst’s prior
performance is made known to the investor.
0361-3682/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.aos.2013.02.001
?
Corresponding author. Tel.: +65 6790 4819; fax: +65 6793 7956.
E-mail address: [email protected] (H.-T. Tan).
1
For instance, some data sources/media (e.g. Starmine, a division of
Thomson Reuters providing equity research service; Yahoo Finance!;
Institutional Investor; Wall Street Transcript) always include the analysts’
prior forecast accuracy ratings with their names, but other media sources
do not always do so (e.g. PR Newswire reports on earnings conference calls
generally do not report other information about the analyst other than the
brokerage ?rm name). As evidence of this latter point, we randomly select
20 highly-exposed analysts whose media mention is higher than the mean
mention documented in Bonner, Hugon, and Walther (2007), half of whom
hold All-Star or All-American status and the other half have no such award.
We collect the media mentions, other than those from Institutional
Investor, Wall Street Transcript, and earnings conference calls (484 in total
without duplicates) for these 20 analysts in the year 2010 from the
FACTIVA database and code the information about the analyst accompa-
nying his/her name. There is only one instance where there is mention of
the award status (or related performance-related information) of the
analyst.
Accounting, Organizations and Society 38 (2013) 214–227
Contents lists available at SciVerse ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
The joint effect of these two analyst attributes on inves-
tors’ judgments has not been investigated, but is important
for a few reasons. Familiarity through mere exposure to an
analyst’s name per se (i.e., without any performance mea-
sures) is an irrelevant cue, and psychology research sug-
gests that this can positively affect investors’ judgments
(Zajonc, 1968). Following psychology literature (Zajonc,
1968), we use the term ‘‘mere exposure effect’’ to refer to
the enhancement of investors’ favorability (in terms of reli-
ance on the analyst’s forecast) towards an analyst through
mere exposure of the individual to the analyst’s name. To
the extent this mere exposure effect occurs in an invest-
ment setting, it has implications on how analysts can en-
hance their credibility with and impact on investors,
even without any change in their forecasting performance.
On the other hand, an analyst’s performance status (e.g.,
All-star status) is a relevant cue that is positively associ-
ated with forecast accuracy (Fang & Yasuda, 2009; Stickel,
1992), and should be relied upon. However, these analyst
attributes (prior exposure and performance status) are
sometimes present in isolation and sometimes at the same
time, and theory suggests that they have substitutive ef-
fects (Dhami & Harries, 2001; Gigerenzer & Gaissmaier,
2011; Gigerenzer & Goldstein, 1996). For instance, while
investors’ in?uence by an irrelevant cue such as their
familiarity with the analyst’s name may be an area of con-
cern, the presence of an analyst’s performance status (a
relevant cue juxtaposed with the analyst’s name) may
actually alleviate this concern as it can dominate this mere
exposure effect. However, the reverse is also possibly true.
That is, in the presence of prior exposure to the analyst’s
name, performance cue may no longer matter, which is
an area of concern.
Recent archival research provides some evidence of a
mere exposure effect, and indicates that the celebrity sta-
tus of analysts (measured by the quantity of media cover-
age analysts receive in the major media sources) positively
affects investors’ reactions to forecast revisions (Bonner
et al., 2007). However, they also ?nd that media coverage
is positively associated with ex ante forecast performance
such as All-Star status,
2
and thus, it is possible that media
coverage is proxying for analyst performance. As Table 1
shows, All-American/All-Star analysts are about three/two
times more likely to be in the high versus low media cover-
age group (22.60% vs. 6.94%; Pearson chi-square = 978.48,
p < 0.01/18.42% vs. 9.41%, Pearson chi-square = 341.77,
p < 0.01).
3
Furthermore, because the analyst’s name is some-
times accompanied by his/her prior performance in such
media coverage but sometimes not, it is unclear whether
the media coverage effect is due to investors’ exposure to
the analyst’s name, his/her prior performance, or an interac-
tion effect between the two. The authors acknowledge that
‘‘an alternative explanation of our results is that, consistent
with prior work in the area, market participants react more
strongly to forecast revisions issued by analysts with supe-
rior performance’’ (Bonner et al., 2007, p. 483).
We also examine whether the credibility enhancement
established by prior exposure persists when the analyst’s
actual performance disappoints investors. We suggest that
while there is a bene?t to the analyst from increased media
exposure, this bene?t is short-lived should the analyst
make a forecast error. This ?nding informs us about a pos-
sible negative effect of exposure on the analyst’s credibility
over multi-period interactions between analysts and inves-
tors. This ?nding is also important to other market partici-
pants such as managers, who may also consider exposure as
one way to enhance their credibility with investors.
We conduct an experiment to investigate these issues.
The key advantage of the experimental approach is that
it allows us to manipulate participants’ awareness of the
prior performance of the analyst. Although it is possible
to use archival data to measure both media coverage and
analysts’ prior performance (e.g., as in Bonner et al.,
2007), investors exposed to a high media coverage analyst
may or may not be aware of the analyst’s prior perfor-
mance. Therefore, it is dif?cult to assess whether the stron-
ger reactions to the celebrity analysts’ earnings forecast are
caused by their prior performance, investors’ exposure to
the celebrity analysts’ names, or both. In addition, using
the experimental approach enables us to hold constant
other information such as the strength of the analyst’s
arguments (Hirst, Koonce, & Simko, 1995). Other factors
determining the appearance of analyst’s name in the media
Table 1
A 2 Â 2 frequency table of media coverage and analyst’s perceived
performance.
a
Panel A:
Performance
[Institutional Investors All-American
Award]
No Yes
Media Coverage Low 9407 (93.06%) 701 (6.94%)
High 7760 (77.40%) 2266 (22.60%)
Panel B:
Performance
[Wall Street Journal All-Star Award]
No Yes
Media Coverage Low 9157 (90.59%) 951 (9.41%)
High 8179 (81.58%) 1847 (18.42%)
Number in parentheses refers to the percentage of observations within
each media coverage group.
a
This frequency table is based on the sample of Bonner et al. (2007)
(20,134 analyst-?rm-quarter observations) and shows the number of
observations (the percentage of observations within each media coverage
group) in each cell. Media coverage of the analyst is classi?ed as low/high
using median split. Performance of the analyst is measured by the award
status (Institutional Investors All-American award, Panel A; Wall Street
Journal All-Star award, Panel B).We also conduct the chi-square test of
independence and ?nd that media coverage is associated with both All-
American and All-Star award status (Pearson chi-square = 978.48/341.77,
p < 0.01 for both award status measures).
2
A recent working paper by Rees, Sharp, and Twedt (2011) using a
subset of the media that Bonner et al. (2007) cover (top 5 business press;
86% of Rees et al.’s (2011) sample comes from Financial Times and The Wall
Street Journal) also ?nds similar results for the positive correlation between
All-Star status and media coverage. They ?nd a positive association
(regression coef?cient of 1.277, odds ratio = 3.586) between media cover-
age and All-Star status in the logistic regression, suggesting that the odds of
being covered by the top business press increase by 258.6 percent with All-
Star status.
3
We appreciate the help of Beverly Walther in providing the analysis
included in Table 1.
W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227 215
press can also be controlled. For example, the media may
selectively report those analysts who are more active,
available, in?uential, or who provide eye-catching news.
Finally, the experimental method offers an opportunity to
test the cognitive mechanism by which investors react to
an earnings forecast provided by a familiar analyst.
We manipulate prior exposure of the analyst’s name
(zero, high) and favorable performance cue (absent, pres-
ent) in our between-subjects experiment. Participants are
MBA students, and we manipulate the exposure of the ana-
lyst’s name in stage one of the experiment by ?rst showing
all participants twenty slides containing nine names, but
where the target analyst’s name is mentioned six times in
one condition and none in the other condition. Participants
are merely shown the names with no other information
associatedwiththe names at this stage. We manipulate par-
ticipants’ awareness of the analyst’s award status by either
providing or not providing this favorable performance cue
along with the analyst’s quarterly earnings forecast in the
second stage of the experiment, where participants are told
that the analyst is a Wall Street Journal All-Star analyst or
they are not given any information about the analyst’s prior
performance. Finally, in the third stage of the experiment,
we provide participants with the actual earnings, which
indicate that the analyst has made a forecast error.
Results indicate that in the absence of a favorable per-
formance cue about the analyst, investors’ perceived ana-
lyst credibility is higher in the presence of prior exposure
to the analyst’s name than in its absence, and investors rely
more on the highly-exposed analyst’s forecast to make
their own earnings estimates. When the favorable perfor-
mance cue about the analyst is present, exposure to the
analyst’s name has no incremental effect on investors’ per-
ceived analyst credibility and earnings estimates. Further-
more, with prior exposure to the analyst’s name, investors’
perceived credibility assessments and earnings estimates
are no different with or without awareness of the prior
high performance of the analyst. This suggests that inves-
tors implicitly equate a familiar analyst name with a level
of high prior performance, or in the minimum, the analyst’s
high prior performance no longer matters given prior
exposure to his/her name. Perceived analyst credibility
mediates investors’ earnings estimates and willingness to
rely on the analyst’s future report. Finally, after the actual
earnings are announced and the analyst’s forecast turns
out to be inaccurate, investors lower their credibility rat-
ings on the highly-exposed (versus the non-exposed) ana-
lyst to a greater extent when the investor has not received
a prior favorable performance cue about the analyst.
This study contributes to the literature by partitioning
the effects of prior exposure to an analyst’s name from
investors’ awareness of the performance cue about the ana-
lyst. Our examinationof the interactionbetweenprior expo-
sure and performance cue (rather than a main effect of prior
exposure) allows us to contribute to the literature in three
important ways. First, our documentation of a prior expo-
sure effect in the absence of a favorable performance cue
provides evidence of a direct causal relation between prior
exposure to an analyst’s name and investors’ positive cred-
ibility judgments, which is hard to test using archival data.
Second, we ?nd that given prior exposure to the analyst
name, investors’ credibility judgments are no different
whether they are giventhe analyst’s high prior performance
or not. An inference then is that a diagnostic cue such as the
analyst’s highprior performance no longer matters to inves-
tors once they have prior exposure to the analyst’s name.
This is potentially worrisome given that, among high media
coverage analysts, the majority (82%/77%, Table 1) do not
have All-Star/All-American award status. Under these cir-
cumstances, it wouldappear bene?cial for investors to place
more reliance on this smaller group of high prior perfor-
mance analysts. However, our results show that they do
not do so. Third, by including the favorable performance
cue as a moderator, we ?nd that there is no prior exposure
effect in the presence of a performance cue (i.e., investors
are aware of the analyst’s All-Star status). This suggests a
boundary condition to the media exposure effect.
In addition, we extend the literature by showing that
the mechanism by which prior exposure to an analyst’s
name in?uences investors’ earnings judgments is via their
assessments of the analyst’s credibility. Our results show
that analysts can enhance their perceived credibility sim-
ply by increasing their media exposure frequency, even if
it pertains to their names only.
However, while an analyst’s credibility can be enhanced
when investors have prior exposure to his/her name, this
prior exposure effect is short-lived when the analyst’s fore-
cast turns out to be inaccurate. We show that the analyst’s
credibility takes a bigger hit when the investor has prior
exposure to his/her name. This effect occurs because when
the analyst’s forecast is inaccurate, investors realize that
prior enhanced credibility is misplaced.
In the next section of the paper, we review the related
literature and develop the hypotheses. We then describe
the research design and experiment procedure, analyze
the experimental results, and conclude the paper.
Hypothesis development
Effects of exposure and performance cue on investors’
judgments
Zajonc (1968) documents that mere exposure of the
individual to a stimulus enhances his/her favorability to-
ward it. Mere exposure refers to a condition in which prior
exposure makes the given stimulus more accessible to the
individual’s memory and in?uences his/her perception. Za-
jonc (1968) manipulates the exposure frequency of non-
sense Chinese characters and asks participants to guess
their meanings on a good-bad scale. He ?nds a signi?cantly
positive exposure-favorability relationship. Following Za-
jonc (1968), other psychology studies ?nd that increased
exposure results in favorable evaluations of a variety of
properties such as liking, goodness, attractiveness, and
pleasantness measures (Bornstein, 1989). This suggests
that credibility assessments will also be likewise positively
in?uenced by prior exposure.
The perceptual ?uency model offers an explanation for
this mere exposure effect (Bornstein &D’Agostino, 1994; Ja-
coby, Toth, Lindasy, & Debner, 1992; Mandler, Nakamura,
Shebo, & Zandt, 1987). This model is based on the concept
of ‘‘perceptual ?uency,’’ which refers to the ease with which
216 W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227
people perceive, encode andprocess information. According
to this model, prior exposure to a stimulus results in the for-
mationof a perceptual representationof the stimulus. When
people are asked to evaluate the previously-exposed stimu-
lus, the perceptual representation of the stimulus is acti-
vated and facilitates the encoding and processing of the
stimulus and enhances the perceptual ?uency for that stim-
ulus. Perceptual ?uency serves as the basis for the feeling of
familiarity in the sense that stimulus perceived more
quickly tends to be identi?ed as familiar (Johnston, Dark, &
Jacoby, 1985). People often attribute the cause of the per-
ceptual ?uencytothe stimulus’ status onthe dimensionthat
is salient at that moment. Such attribution of perceptual ?u-
ency to the exposed stimulus produces higher ratings onthe
stimulus. For example, Mandler et al. (1987) ?nd that when
participants are asked to evaluate the stimuli in the particu-
lar dimensions (such as preference, brightness or darkness),
exposed irregular shapes are perceived as preferred, bright-
er or darker than the unexposed shapes.
Psychology research suggests that people can automati-
cally encode frequency information (Zacks, Hasher, & Sanft,
1982) and that this process uses a minimal amount of atten-
tional effort (Hasher & Zacks, 1979). Prior studies in the
auditing and accounting literature have investigated the ef-
fect of frequency of exposure to relevant cues. For instance,
Butt (1988) ?nds that for both students and auditors, fre-
quency judgments based on direct experience (by repeated
individual presentations of ?nancial statement errors) are
more accurate than those based on indirect experience (by
receiving summary data). Joe (2003) ?nds that auditors
whoreceiveboththe?nancial statements andapress release
of a client’s debt default (which is the repeated information
that has beendisclosedinthe ?nancial statements notes ear-
lier) perceive a client’s bankruptcy probability to be higher
than in a situation when there is no press coverage. In an
experiment, Hugon(2004) ?nds that redundancyof ?nancial
performance informationis positively associated with infor-
mationcredibility. All these studies investigate the effects of
repeating decision-relevant information, but not decision-
irrelevant information (such as an analyst’s name).
Most related to this study is the archival study by Bon-
ner et al. (2007). They de?ne a celebrity as a famous or
well-publicized person, being well-known in addition to
his performance-related qualities (Bonner et al., 2007, p.
482). They use the quantity of media coverage analysts re-
ceive as their empirical proxy for celebrity and measure to-
tal media coverage as the number of appearances of an
analyst’s name associated with his brokerage house em-
ployer in all media sources. Bonner et al. (2007) ?nd that
investors react more strongly to the celebrity analysts’ re-
vised earnings forecasts. They also ?nd a positive associa-
tion between media coverage and an analyst’s ex ante
performance, and obtain the same ?ndings after including
the analysts’ performance as control variables in their
regression models.
4
However, an analyst’s name is some-
times accompanied by his/her prior performance status in
these media coverage, and Bonner et al. (2007) do not exam-
ine whether investors are aware of the analyst’s prior
performance, nor how this knowledge interacts with prior
exposure to the analyst’s name.
Based on the mere exposure effect, we expect that the
mere prior exposure of the name enhances investors’ per-
ceptual ?uency for the name. When the name appears in
the press release as the source of the earnings forecast,
investors will view the person (the analyst) more favorably
in terms of credibility. Higher perceived analyst credibility
increases the believability of the earnings forecast issued
by the analyst (Hirst, Koonce, & Venkataraman, 2007;
Williams, 1996). In turn, investors are more likely to rely
on the highly-exposed analyst’s forecast, such that inves-
tors’ own earnings estimates are more in?uenced by the
analyst’s forecast.
Prior research also suggests that investors’ reactions to
analysts’ (management) earnings forecasts depend on the
analysts’ (management) prior forecast accuracy or factors
associated with current forecast accuracy (Bonner, Wal-
ther, & Young, 2003; Clement & Tse, 2003; Hirst, Koonce,
& Miller, 1999). Therefore, we expect that investors’ reac-
tions to the analyst’s earnings forecast will be stronger
when the analyst is known to be an All-Star analyst.
5
Psychology research suggests that these two factors—
prior exposure to the analyst’s name (absent, present)
and favorable performance cue (absent, present)—have
substitutive effects. In particular, the one-reason decision
making model provides a detailed discussion about how
people make judgments with multiple cues (Dhami & Har-
ries, 2001; Gigerenzer & Gaissmaier, 2011; Gigerenzer &
Goldstein, 1996). This model suggests that people make
decisions or judgments based on a single cue, instead of
integrating all available information. In other words, multi-
ple cues substitute for each other, and are not compensa-
tory. Previous research on persuasion also ?nds a
substitutive effect between source credibility and other
source variables (Joseph, 1977; Maddux & Rogers, 1980;
Pornpitakpan, 2004). For example, Joseph (1977) shows
that physical attractiveness has little effect on participants’
preference when the source is known as an expert, while
the effect of attractiveness is signi?cant for a non-expert.
Maddux and Rogers (1980) document that in terms of per-
suasiveness of an argument, the superiority of the expert
to the non-expert is greater when the source is unattrac-
tive than when the source is attractive.
As we mentioned earlier, we expect that either prior
exposure to the analyst’s name or awareness of the ana-
lyst’s prior good performance leads investors to perceive
the analyst to have higher credibility, which also leads
them to rely more on the analyst’s earnings forecast when
making their own earnings estimates. The one-reason deci-
sion making model and related studies discussed above
suggest that these two factors are substitutes, such that a
4
They do not examine whether there is an interaction effect between
media coverage and performance. Statistics textbooks (e.g., Kutner,
Nachtsheim, Neter, & Li, 2005) argue that once there is a signi?cant
interaction effect between two variables, any main effect of one variable
has to be interpreted conditional on the other effect.
5
We focus on a favorable performance cue (i.e., the analyst’s award
status) in this study. The reason is that the analysts and the media are
generally reluctant to disclose the performance for bad performers (for
example, Yahoo! Finance only displays the performance ratings for those
top performers; i.e. those analysts whose ratings are higher than average).
W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227 217
pronounced mere exposure effect (from prior exposure to
the analyst’s name) exists in the absence of the perfor-
mance cue about the analyst; this effect is likely to be sub-
dued when a performance cue about the analyst
accompanies reports containing the analyst’s views. Our
hypothesis is formally stated below.
Hypothesis 1. Prior exposure to an analyst’s name has a
more positive effect on investors’ analyst credibility
assessments and their reliance on the analyst’s forecast
when a favorable analyst performance cue is absent than
when it is present.
The one-reason decision model also suggests that the
positive effect of the performance cue is likely to be sub-
dued when investors have prior exposure to the analyst’s
name. This indicates a potential adverse effect on investors
arising from prior exposure to the analyst’s name in that
relevant cues such as favorable prior performance of the
analyst are under- (not) weighted with prior exposure to
the analyst’s name. Our hypothesis is stated below:
Hypothesis 2. A favorable analyst performance cue has a
more positive effect on investors’ analyst credibility
assessments and their reliance on the analyst’s forecast
when the investor has not had prior exposure to the
analyst name than when the investor has had prior
exposure.
Effects of exposure and performance cue on investors’
judgments after actual earnings announcement
Hypotheses 1 and 2 relate to positive effects accruing to
the analyst from investors’ prior exposure to his/her name
or a favorable prior performance status. Our next research
question relates to whether the positive effects accruing to
the analyst due to this prior name exposure or a favorable
prior performance persist. We do so in a setting where the
analyst’s actual current performance is known and turns
out to be inaccurate. InHypothesis 1, we predict that the po-
sitive effect of prior exposure is stronger in the absence
(than in the presence) of a favorable performance cue about
the analyst. Whenthe analyst’s forecast turns out to be inac-
curate, the presence of forecast error reminds investors that
their greater con?dence and reliance placed earlier on the
highly-exposed analyst are not warranted; the prior en-
hanced credibility assessments will be correspondingly re-
duced. This downward adjustment is particularly greater
where the favorable performance cue is absent since the
prior exposure effect is initially stronger here. Similarly, in
Hypothesis 2, we predict that the positive effect of a favor-
able performance cue is stronger in the absence of prior
exposure to the analyst’s name than in its presence. Again,
the presence of a forecast error leads to an undoing of the
initial enhanced con?dence and reliance placed on the ana-
lyst witha prior favorable performance cue. Again, the effect
is magni?edwhen there is no prior exposure to the analyst’s
namesince the prior performance cue effect is stronger here.
Our hypotheses are formally stated as follows.
Hypothesis 3. Prior exposure to an analyst’s name pro-
duces a more negative reaction to a forecast error when the
investor has not received a favorable analyst performance
cue than when the investor has received a favorable
analyst performance cue.
Hypothesis 4. Prior exposure to a favorable analyst perfor-
mance cue produces a more negative reaction to a forecast
error when the investor has not had prior exposure to the
analyst’s name than when the investor has had prior expo-
sure to the analyst’s name.
Fig. 1 provides a graphical summary for Hypotheses 1
and 2 (panel A) and Hypotheses 3 and 4 (panel B).
Method
Participants
We conduct an experiment with 89 M.B.A. students
from a major Singapore university as proxy for non-profes-
sional investors (Elliott, Hodge, Kennedy, & Pronk, 2007).
6
Panel A: Graphical summary of Hypotheses 1 and 2
Panel B: Graphical summary of Hypotheses 3 and 4
Zero High
Perceived
Analyst
Credibility
Exposure
Performance Cue Present
Performance Cue Absent
Zero High
Change in
Perceived
Analyst
Credibility
Exposure
0
Performance Cue Present
Performance Cue Absent
Fig. 1. Effects of exposure and performance cue on investors’ judgments.
Notes: Panel A plots predicted results associated with Hypotheses 1 and 2,
which predict that perceived analyst credibility is a joint function of
exposure and performance cue. Panel B plots predicted results associated
with Hypotheses 3 and 4, which predict that after the actual earnings
announcement, the change in perceived analyst credibility is a joint
function of exposure and performance cue.
6
We dropped one participant whose responses indicated that he
misread the case (i.e., initial earnings estimate relative to consensus
forecast is both directionally opposite to all other participants’ estimates,
and extreme at 3.2 the standard deviation). When we include this
participant in the analysis, the contrast test for H1&H2 is signi?cant for
the credibility measure (F = 12.50, p < 0.01) but not for the change in
earnings estimates measure (F = 1.40, p = 0.12). The contrast test for H3 and
H4 is signi?cant for the change in perceived credibility measure (F = 3.20,
p = 0.04).
218 W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227
The participants have a mean (median) working experience
of 7.34 (5.58) years. On average, the participants have taken
2.66 (3.66) accounting (?nance) courses. Seventy-three
(Fifty-six) percent of the participants have investment expe-
rience in stock (fund) market. Each student is paid twenty
Singapore dollars for participating in the experiment.
Design
We employ a two-way (exposure; performance cue) be-
tween-subjects design to test the hypotheses. The ?rst
independent variable is the exposure frequency of the tar-
get analyst’s name (zero, high). The second independent
variable is the analyst’s favorable performance cue (absent,
present). In the performance cue present condition, the
target analyst name is accompanied by a description in
the press release that the analyst is a ‘‘Wall Street Journal
All-Star analyst.’’ In contrast, in the performance cue ab-
sent condition, only the target analyst name is provided
in the press release, so participants are not aware of the
All-star status of the analyst.
Procedure
Participants are randomly assigned to the treatment
conditions. At the beginning of the experiment, partici-
pants are told that they will complete two studies; a visual
study followed by an earnings forecast study. The reason
we choose to label the ?rst part a visual study (where
the exposure frequency of analyst’s name is manipulated)
is to avoid any possible demand effects on participants.
In the visual study, participants are told to focus on the
screen and watch the slide show. Twenty slides containing
nine different English names (?rst name plus last name)
are presented one by one in the projector screen at two
seconds interval. These nine names include both the target
(alternative) analyst’s name in the high (zero) exposure
condition and eight other ?ller names. The target analyst’s
name is the name whose exposure frequency is manipu-
lated, and is shown as the analyst who makes the earnings
forecast in the press release in the main experiment for all
conditions. In the high exposure condition, the target name
appears a total of six times out of twenty slides in the slide
show, whereas in the zero exposure condition, an alterna-
tive name replaces the target name in the corresponding
positions in the slide show and the target name never ap-
pears in the visual study.
In selecting the target, alternative and ?ller names, we
follow several rules. First, English names are used because
the largest brokerage ?rms are U.S. companies, and the use
of English names helps to increase the external validity of
the results. Second, we choose the middle-ranked frequent
last and ?rst names from the U.S. Census Bureau 1990 to
remove possible semantic implication of the name. Finally,
we choose ?rst names that are usually used as male names,
since most analysts are male (Green, Jegadeesh, & Tang,
2009). In the experiment, we use ‘‘Devon Fraley’’ as the tar-
get analyst’s name. In the high exposure condition, ‘‘Devon
Fraley’’ appears six times in the slide show while in the
zero exposure condition, ‘‘Kirby Sikora’’ replaces ‘‘Devon
Fraley’’ in the corresponding positions; the other names
are all the same for the two conditions. In order to avoid
any possible recency or primacy effect, the ?rst and the last
name shown are all ?ller names.
After the slide show, participants are asked two ques-
tions unrelated to current experiment to clear participants’
short memory. Then, participants are told to put aside the
materials for the visual study and continue to do the
earnings forecast study. After reading the background
information about a listed manufacturing company ‘‘Theta
Inc.,’’ participants are shown the 5-year ?nancial summary
for the period 2004–2008. This is followed by the quarterly
earnings for the past two years and for the ?rst two quar-
ters of ?nancial-year 2009, and consensus EPS forecast for
the third quarter of 2009 ($0.18) as well as the consensus
12-month EPS forecast for 2009 ($0.80). Participants are
asked to provide their earnings forecasts for the third quar-
ter and full year of 2009, the con?dence in their earnings
estimates, and evaluations of Theta’s earnings growth
potential and stock appreciation potential in the next
12 months.
Participants then proceed to open Envelope A, which
contains the press release issued on August 31, 2009 about
the analyst’s earnings forecast for Theta’s third quarter
earnings. In the performance cue absent condition, partic-
ipants read the press release stating that ‘‘(a)nalyst, Devon
Fraley, estimates Theta’s earnings per share for the third-
quarter ending September 30, 2009 will be $0.20.’’ In the
performance cue present condition, participants read the
same statement containing the analyst’s forecast, along
with the information that the analyst is a Wall Street Jour-
nal All-Star analyst.
We set the analyst’s forecast to be good news compared
to the consensus analysts forecast ($0.18) and the same
quarter realization in the prior year because this is consis-
tent with prior research that investors expect analysts to
issue good news forecast (Hirst et al., 1995), and good
news earnings forecast may be less credible (Hutton, Mill-
er, & Skinner, 2003). Therefore, in this scenario, the role of
an analyst’s credibility is critical to enhance the believabil-
ity of his/her earnings forecast (Hirst et al., 2007; Mercer,
2004, 2005; Williams, 1996).
After reading the press release, participants are asked to
give their updated earnings forecasts for the third quarter
and full-year 2009, their con?dence in the forecasts, and
future earnings growth and stock price appreciation poten-
tials in the next 12 months. In addition, participants are
asked to indicate their willingness to rely on the analyst’s
report in the future, the analyst’s competence, trustworthi-
ness, reputation, their expectation that the analyst’s fore-
cast to be accurate, the believability of the forecast, the
likelihood that the analyst was intentionally misguiding
the market. All these questions are assessed by 11-point
scales (0 = Extremely Low and 10 = Extremely High), with
the exception of the con?dence question, where the scale
is from 0.0 to 1.0.
After the participants make their assessments of the
analyst based on the press release, they are asked to put
all materials to Envelope A and open Envelope B, which
contains the actual earnings announcement stating that
W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227 219
‘‘(o)n October 14, 2009, Theta announces that the earnings
per share for the third-quarter ending September 30, 2009
are $0.16.’’ Participants are then asked to evaluate Theta’s
full-year EPS for the year 2009, earnings growth potential
and stock price appreciation potential, their willingness
to rely on the analyst’s report in the future, the analyst’s
absolute and relative competence, trustworthiness, reputa-
tion and the likelihood that the analyst was intentionally
misguiding the market. Following this, participants
complete other post-experimental questions including
the manipulation checks and demographic questions about
their working and investment experience.
Results
Manipulation checks
To check whether the high exposure frequency
increases the familiarity of the target analyst’s name, in
the post-experimental questionnaire, participants are
asked to rate their familiarity with the analyst’s name on
an 11-point scale ranging from 0 (not familiar at all) to
10 (completely familiar). Participants in the high exposure
condition rate the analyst name mentioned in the press
release to be signi?cantly more familiar (mean = 6. 84)
than those in the zero exposure condition (mean = 2.67;
F = 47.29, p < 0.01).
7
Seventy-two percent of all participants
correctly answer the question on whether the analyst men-
tioned in the press release is a Wall Street Journal All-Star
analyst. We include all participants in our analyses and
results are similar if we exclude those who fail the manipu-
lation checks.
In our post-experimental questionnaire, we also ask
participants to indicate whether the press release men-
tioned the speci?c analyst’s name and whether the ana-
lyst’s name was (was not) shown in the visual study.
Seventy-three percent of the participants in the high expo-
sure condition correctly remember that the press release
mentioned the speci?c analyst’s name, while only 36% of
the participants in the zero exposure condition correctly
remember this (F = 13.94, p < 0.01). All (59%) of the partic-
ipants in the high (zero) exposure condition correctly rec-
ognize that the analyst’s name was (was not) shown in the
visual study (F = 28.59, p < 0.01). This suggests that the
exposure frequency in?uences participants’ attention to
the analyst name in the press release and recognition of
the analyst.
Test of Hypotheses 1 and 2
Perceived analyst credibility and earnings estimates
Prior literature (Hovland, Janis, & Kelley, 1953; Mercer,
2004, 2005) documents that credibilityhas twodimensions:
competence and trustworthiness. We take the average of
participants’ ratings on the analyst’s competence and trust-
worthiness as a single composite credibility measure (Cron-
bach’s alpha = 0.81). Table 2 summarizes the results of
participants’ perceived analyst credibility after they receive
the press release about the analyst’s earnings forecast.
8
The
ANOVA results indicate signi?cant main effects of exposure
(p = 0.04) and performance cue (p < 0.01), along with a signif-
icant interaction effect (p = 0.07).
As an overall test of the ordinal interaction implied in
Hypotheses 1 and 2 (Buckless & Ravenscroft, 1990; Rosnow
& Rosenthal, 1995), we conduct a contrast test with the
weights À3 in the ‘‘zero exposure + performance cue absent’’
condition, +1 in the ‘‘high exposure + performance cue ab-
sent’’, ‘‘zero exposure + performance cue present’’, and ‘‘high
exposure + performance cue present’’ conditions. The con-
trast test is signi?cant (F = 12.65, p < 0.01, Panel C, Table 2).
9
Results are consistent with Hypothesis 1. Speci?cally,
when participants have no information about the perfor-
mance cue, they perceive the highly-exposed analyst to
be more credible (mean = 5.69) relative to the non-exposed
analyst (mean = 4.98, F = 5.11, p = 0.01). Further, when the
analyst is known to be an All-Star analyst, exposure of the
analyst’s name has no signi?cant effect on perceived ana-
lyst credibility (mean for the zero/high exposure condi-
tion = 5.95/6.00, F = 0.02, p = 0.44; panel C Table 2).
Results are also consistent with Hypothesis 2. We ?nd
that with no prior exposure to the analyst’s name, partici-
pants attend to the high prior performance of the analyst in
that credibility assessments are higher with the presence
rather than absence of a performance cue (F = 9.82,
p < 0.01). In addition, with prior exposure to the analyst’s
name, participants appear to pay short shrift to prior per-
formance information. Their credibility assessments do
not differ whether the performance cue is present or ab-
sent (F = 1.00, p = 0.16). This suggests that prior exposure
to the analyst’s name has an impact on participants’ judg-
ments such that a diagnostic cue such as prior performance
information has no incremental effect. In fact, we also ?nd
no difference in participants’ credibility between the ‘‘high
exposure + performance cue absent’’ condition and the
‘‘zero exposure + performance cue present’’ condition
(p = 0.20). This suggests that mere exposure to the ana-
lyst’s name offers the same positive reputational and abil-
ity-to-in?uence impact as good prior performance.
We focus onparticipants’ quarterlyearnings estimates to
examine their reliance on the analyst’s forecast for the third
quarter. We expect that higher perceived analyst credibility
makes the analyst’s forecast more believable, which in turn
7
The p-value ?gures are all one-tailed for directional predictions, unless
otherwise indicated.
8
We ask participants to indicate their perceptions on the analyst’s
competence/trustworthiness/reputation in absolute (an 11-point scale) and
relative terms (compared to the peers). Factor analysis suggests that these
questions load to two factors and we label them as perceived absolute/
relative analyst credibility. Following prior research (e.g., Barton & Mercer,
2005; Cianci & Kaplan, 2010; Mercer 2005), we use the perceived absolute
competence and trustworthiness to construct the perceived analyst
credibility measure in the main analysis. We obtain similar results when
we use perceived relative analyst credibility or perceived absolute/relative
analyst reputation measures. Results for all tests of hypotheses are similar
after controlling for the participants’ working experience, and number of
accounting and ?nance courses they have taken.
9
We also test alternative contrast weights À3/+1/0/+2 in the ‘‘zero
exposure + performance cue absent’’/‘‘high exposure + performance cue
absent’’/‘‘zero exposure + performance cue present’’/‘‘high exposure + per-
formance cue present’’ condition. The contrast test is signi?cant for the
perceived analyst credibility (F = 11.14, p < 0.01) and change in quarterly
earnings estimates (F = 1.73, p = 0.10).
220 W. Chen, H.-T. Tan/ Accounting, Organizations and Society 38 (2013) 214–227
leads tohigher investors’ reliance onthe analyst’s forecast.
10
In our experiment, the target analyst’ earnings forecast is
optimistic compared to the consensus forecast. Therefore,
higher reliance on the target analyst’s forecast implies higher
third-quarter earnings estimates. Participants are asked to
provide their earnings estimates twice, before and after the
press release about the analyst forecast. Therefore, we use
the change in the participants’ third-quarter earnings esti-
mates as the dependent variable. ANOVA results (Table 3, pa-
nel B) indicates a signi?cant interaction effect (p = 0.05). As
with the credibility measure, the contrast test with the same
weights (À3/+1/+1/+1) is also signi?cant (F = 2.67, p = 0.05).
We ?nd a signi?cant exposure effect on the change in partic-
ipants’ third quarter earnings estimates in the absence of the
performance cue (p = 0.04), and an insigni?cant exposure ef-
fect in its presence (p = 0.29). A favorable performance cue
has a signi?cant effect on the change in participants’ earnings
estimates in the absence of prior exposure to the analyst’s
name (p = 0.08), but not in its presence (p = 0.17). No signi?-
cant difference in earnings estimates is found between the
‘‘high exposure + performance cue absent’’ condition and
the ‘‘zero exposure + performance cue present’’ condition
(p = 0.33). In sum, Hypotheses 1 and 2 are supported for both
credibility and earnings estimates measures.
11
Process model tests
We rely on structural equation analysis to verify that
the exposure effect on investors’ judgments and decisions
is due to the underlying mechanisms we propose. We posit
that exposure and performance cue jointly in?uence
Table 2
Perceived analyst credibility after the analyst’s forecast.
Panel A: Mean perceived analyst credibility (standard deviation) [rank]
a
Performance cue
c
Exposure
b
Absent Present Row Mean
Zero 4.98 (1.19) [6.09] 5.95 (1.12) [7.91] 5.47 (1.25) [7.00]
N = 22 N = 22 N = 44
[Condition 1] [Condition 3]
High 5.69 (0.94) [7.38] 6.00 (0.86) [8.00] 5.86 (0.90) [7.71]
N = 21 N = 24 N = 45
[Condition 2] [Condition 4]
Column mean 5.33 (1.12) [6.72] 5.98 (0.98) [7.96]
N = 43 N = 46
Panel B: ANOVA results
Source df Mean square F p-Value
Intercept 1 2840.08 2654.83