Do Features That Associate Managers With a Message
Magnify Investors’ Reactions to Narrative Disclosures?*
H. SCOTT ASAY, University of Iowa
ROBERT LIBBY, Cornell University†
KRISTINA M. RENNEKAMP, Cornell University
*The authors would like to acknowledge helpful comments from Tim Bauer, Rob Bloomfield,
Scott Emett, Brooke Elliott, Stephanie Grant, Jessen Hobson, Frank Hodge, Ken Merkley, Mark
Nelson, Mark Peecher, and Hun-Tong Tan and participants at the Cornell University, University
of Illinois, and University of North Carolina Accounting Workshops and 2013 BYU Accounting
Research Symposium.
†Corresponding author: 383 Sage Hall, Cornell University, Ithaca, New York, 14853. Phone:
607-255-3348. Fax: 607-254-4590. Email: [email protected].
Do Features that Associate Managers with a Message
Magnify Investors’ Reactions to Narrative Disclosures?
Abstract
We test whether investors react more strongly to narrative disclosures when the CEO’s presence is more
salient or prominent in the disclosure. In our first experiment, holding all other information constant, we
manipulate whether a disclosure uses more personal pronouns (e.g., “I” and “our” rather than “the
company” and “its”). Again holding all other information constant, our second experiment manipulates
whether a disclosure does or does not contain a photo of the CEO. Both manipulations lead to stronger
reactions from investors in between-subjects tests. That is, when news is positive (negative), including
either more personal pronouns or the CEO’s photo leads to more positive (negative) reactions from
investors. This provides convergent evidence for our hypothesis. We also find that, within-subjects, both
manipulations are perceived as indicating greater association with the message, but participants do not
expect an effect on investment evaluations. In our third experiment, we provide additional evidence that
personal pronoun usage affects investor reactions by increasing the perceived credibility of the disclosure.
Keywords: language, voluntary disclosure, investor judgment, Plain English
This research did not receive any specific grant from funding agencies in the public, commercial, or not-
for-profit sectors.
1
1. Introduction
Managers play a key role in communicating firm performance, and the disclosures they provide
can vary in the extent to which the manager’s presence is salient or prominent in the disclosure (hereafter,
managers’ “association with the message”). For example, some types of disclosures (e.g., conference
calls, letters to shareholders, MD&A, etc.) might associate managers more with the message than other
types of disclosures (e.g., press releases, 10-Ks, etc.), and firms are increasingly using disclosure
mediums that could more closely associate managers with the information in their disclosures (e.g., online
video disclosures as in Elliott et al. 2012, or social media disclosures as in Cade 2016). Even within these
different types of disclosures there is variation in the extent to which disclosure choices associate
managers with the message they are communicating. In this paper, we examine how two such disclosure
choices – a manager’s use of personal pronouns and the inclusion of the manager’s photo in the disclosure
– affect investors’ reactions to the information in the disclosure.
Drawing on prior work in communications and psycholinguistics (Cohn, Mehl, and Pennebaker
2004; Hyland 2005a, b; Pennebaker 2011), we predict that these disclosure choices signal to investors that
managers believe the information in the disclosure, thereby magnifying investors’ reactions to that
information. Understanding these effects is important for several reasons. First, there is a great deal of
variation in practice in the use of personal pronouns and photos. For example, recent work in the
accounting literature documents variation in managers’ use of personal pronouns within firm disclosures
(Gow, Kaplan, Larcker, and Zakolyukina 2015; Brochet, Miller, Naranjo, and Yu 2016). Likewise, the
Letters to Shareholders in the 2015 annual reports of Procter and Gamble, Walmart, and Target all include
photos of their respective CEOs, whereas the Letters to Shareholders in the 2015 annual reports for Ford,
Apple, and Kroger do not. Second, the literature has suggested a variety of reasons for this variation,
including deception, self-serving attribution bias, and manager traits such as overconfidence, narcissism,
and cultural background. Each of these explanations suggests that personal pronouns and photos may be
signals of negative managerial traits. The theory that we rely on instead focuses on investors’ reactions to
these disclosure choices, and suggests that they may magnify investors’ positive or negative reactions to
2
news. Third, prior archival work indicates that personal pronoun and photo usage may be related to firm
performance (Chatterjee and Hambrick 2007; Chen and Loftus 2017), which suggests the possibility of
strategic use. Users would benefit from knowing about the possible effects of these strategies on their
investment evaluations, especially given that psychology research suggests that they may be unaware of
these effects (Pennebaker 2011). Finally, the effects of personal pronouns are of particular interest given
that the SEC describes them as one of the components of Plain English disclosures (SEC 1998). The SEC
requires firms to use Plain English in some disclosures and encourages firms to use Plain English in all
disclosures.1 Our study examines whether doing so may have an effect on users that the SEC has not
considered.
We test our predictions using a series of experiments. In Experiment 1, we use a 2 x 2 between-
subjects design. Within an abbreviated letter to shareholders, we manipulate (1) whether the disclosure
contains good or bad news and (2) whether the CEO refers to the company in first or third person in the
final sentence of the disclosure when discussing future performance expectations (e.g., “I” and “our” vs.
“the company” and “its”). We isolate our manipulation to the final sentence in order to provide a strong
test of our theory without changing other characteristics of the disclosure. Consistent with our predictions,
we find that participants’ reactions to the disclosure are stronger when the disclosure contains more first-
person personal pronouns. That is, participants’ responses are more positive when good news is conveyed
using first-person personal pronouns in the final sentence of the disclosure and more negative when bad
news is conveyed using first-person personal pronouns in the final sentence of the disclosure. Within-
participant analyses indicate that participants do believe that personal pronouns increase managers’
association with the message, but do not anticipate an effect on valuation judgments. Findings of
valuation differences using between-subjects treatments but not using within-subjects treatments suggests
1 See, for example, SEC Rule 421(d), Rules 13a-20 and 15d-20, and SEC Releases No. 33-8998 and 34-38164. As
stated in Release No. 34-38164 “[The SEC’s] ultimate goal is to have all disclosure documents written in plain
English.”
3
that participants may be unaware of the effects of the personal pronouns in our between-subjects test
(Kahneman and Tversky 1996; Libby, Bloomfield, and Nelson 2002).
In Experiment 2, we provide convergent evidence by using a different disclosure choice to
manipulate the extent to which managers associate themselves with the message within the same
abbreviated letter to shareholders. We use a 2 x 2 between-subjects design and manipulate (1) whether the
disclosure contains good or bad news and (2) whether the CEO’s photo is present in the disclosure.
Again, consistent with our predictions, we find that participants’ reactions to the disclosure are stronger
when the CEO’s photo is included. That is, participants’ responses are more positive when a photo of the
CEO is included in a good-news disclosure and more negative when a photo of the CEO is included in a
bad-news disclosure. As in Experiment 1, within-participants analyses indicate that participants believe
that the photo increases the managers’ association with the message but do not anticipate an effect on
valuation judgments, suggesting that participants may be unaware of the effects of the photo in our
between-subjects test.
While we find that investors’ valuation judgments are affected by our manipulations in both
Experiments 1 and 2, we do not find evidence that these reactions are explained by measures of the extent
to which participants are willing to rely on the disclosure or the extent to which they think that
management believes the information in the disclosure, both of which may be thought of as proxies for
disclosure credibility. Similarly, we do not find evidence that these reactions are explained by measures
of management’s trustworthiness or competence, two dimensions of management credibility (Mercer
2004). We conduct a third experiment to provide additional evidence regarding the process underlying
our results. In Experiment 3, we use a 2 x 2 between-subjects design similar to that of Experiment 1.
Specifically, we manipulate (1) whether the disclosure contains good or bad news and (2) whether the
CEO refers to the company in the first or third person. In order to increase our ability to shed light on the
underlying process, we use a shorter disclosure, manipulate the usage of personal pronouns throughout
the disclosure (rather than in only the final sentence), and increase our sample size. Further, we also
collect additional measures to capture alternative processes that might explain our results (psychological
4
distance, perceived management control, and involvement). Results from Experiment 3 indicate that
personal pronouns increase the extent to which participants felt like they could rely on the information in
the disclosure, the extent to which participants felt like management believed the information in the
disclosure, and perceptions of management trustworthiness. These measures, in turn, led to more positive
reactions to good news and more negative reactions to bad news. In contrast, the alternative process
measures do not explain our results. These findings provide additional support for the idea that disclosure
choices that associate managers with the message increase the credibility of the disclosure and magnify
investors’ reactions to the information contained therein.
Our findings complement related accounting research that examines disclosure attributes that
magnify investor reactions through perceptions of either disclosure credibility or management credibility
(Jennings 1987; Mercer 2004). Our findings suggest that managers can enhance the credibility of their
disclosures by increasing their personal salience or prominence in the disclosure. These findings have
potentially important implications for preparers and users of disclosures, as managers might seek to
influence investors’ reactions to firm communications by selectively adopting disclosure choices that
more closely associate themselves with, or distance themselves from, the information in disclosures.
Consistent with this idea, recent archival evidence suggests that the use of personal pronouns and CEO
photos are related to firm performance (e.g., Chatterjee and Hambrick 2007; Chen and Loftus 2017).
Managers might also opt for disclosure mediums that influence their association with the information in
their disclosures (e.g., online video vs. textual disclosures as in Elliott et al. 2012) and make other choices
within a disclosure medium to associate themselves or distance themselves from the message (e.g.,
personal vs. corporate Twitter accounts).
Our findings also complement archival studies in accounting, finance, and management and have
important implications for researchers who use CEO photos and personal pronouns as proxies for
conceptual variables. For example, personal pronouns and CEO photos have both been used as proxies for
narcissism (Chatterjee and Hambrick 2007) and overconfidence (Pennebaker 2011; Schrand and Zechman
2012). In addition, personal pronouns have been used as a proxy for deception (Larcker and Zakolyukina
5
2012) and self-serving attribution bias (Li 2011; Chen and Loftus 2017). While much of the prior work
tends to view these disclosure choices as indicators of negative managerial traits, our results suggest that
users of disclosures do not view them as uniformly negative. Instead, these disclosure choices lead
participants to respond more strongly to the news in a disclosure, which results in either a more positive
or negative reaction.
Finally, our results should also be of interest to regulators who have emphasized the importance
of disclosure language (e.g., SEC 1998). As part of the Plain English initiative, the SEC encourages firms
to more frequently use personal pronouns to increase disclosure clarity (SEC 1998). However, little is
known about how personal pronoun usage might influence investors’ interpretation of, or reaction to,
disclosures. Our findings suggest one potentially unanticipated effect of encouraging an increase in
personal pronoun usage – personal pronouns may cause users to react more strongly to information in a
disclosure.
The remainder of this paper proceeds as follows. Section 2 provides background information and
develops our hypotheses. Sections 3 and 4 describe methods and results for Experiments 1 and 2,
respectively. Section 5 describes our supplemental experiment that tests the robustness of our personal
pronoun results. Section 6 concludes.
2. Background and Development of Hypotheses
The impact of narrative disclosures
Qualitative, non-numeric information makes up an important part of accounting disclosures.
Narratives in disclosures allow firms to provide additional detail and explanations for firm performance
and to convey additional information about firm strategy and goals for the future. The literature
examining the characteristics of narrative disclosures has grown recently, in part due to the electronic
availability of textual data and tools that allow these data to be analyzed efficiently (for a review, see Li
2010a). We classify research on the qualitative characteristics of disclosures as primarily focusing on
disclosure content or disclosure style. Content refers to what information is explicitly stated (e.g., “Sales
6
in consumer markets increased”). Style refers to how that information is communicated and has little
impact on the literal meaning of communications.
Disclosure content
Many of the studies of narrative disclosure characteristics examine what these disclosures
explicitly state by developing measures based on nouns, adjectives, regular verbs, and adverbs. Since
linguists commonly refer to these types of words as “content” words, we refer to these studies as
examining disclosure content choices. Content words label objects or actions, typically have a culturally
shared meaning, and are “absolutely necessary to convey an idea to someone else” (Pennebaker 2011,
21). If content choices are removed from a communication, the communication becomes meaningless.
Existing studies of disclosure content choices focus on attributes including disclosure tone (e.g., Kothari,
Li, and Short 2009; Feldman, Govindaraj, Livnat, and Segal 2010; Li 2010b; Davis, Piger, and Sedor
2012; Davis and Tama-Sweet 2012; Demers and Vega 2014; Huang, Teoh, and Zhang 2014), language
vividness (Hales, Kuang, and Venkataraman 2011), and disclosure quantity (e.g., Merkley 2014; see
Merkl-Davies and Brennan 2007 and Li 2010a for recent reviews).
Disclosure style
Recent studies have also begun to examine how disclosures are communicated, which we refer to
as disclosure style choices. Relative to disclosure content choices, disclosure style choices have little
effect on the literal meaning of the disclosure.2 Style choices that have received attention in recent studies
include disclosure attributes such as disclosure readability (Li 2008; You and Zhang 2009; Miller 2010;
Rennekamp 2012; Tan et al. 2014; Tan et al. 2015), disclosure medium (Elliott et al. 2012), and emphasis
(Bowen, Davis, and Matsumoto 2005; Elliott 2006). This literature suggests that style choices can have
important effects on user decisions. For example, empirical evidence suggests that bad news disclosures
tend to be longer and less readable than good news disclosures (Li 2008). Further, longer, less readable
2 That is, the meaning of communications can still largely be understood if style choices are removed from a
communication. Our distinction between content choices and style choices is adapted from Pennebaker’s (2011)
classification of content words and style words.
7
disclosures are associated with lower trading and weaker reactions (You and Zhang 2009; Miller 2010;
Rennekamp 2012; Tan et al. 2014; Tan et al. 2015) and increase investors’ propensity to seek out and rely
on outside sources of information about the firm (Asay et al. 2017).
Recent work has also examined the usage of personal pronouns and CEO photos. Concerning
personal pronoun usage, research suggests that their use might reflect managerial narcissism (Chatterjee
and Hambrick 2007), deception (Larcker and Zakolyukina 2012), or managers’ susceptibility to self-
serving attribution bias (Li 2011; Chen and Loftus 2017). The prevalence of CEO photos has been
interpreted as indicating managerial overconfidence (Schrand and Zechman 2012) or managerial
narcissism (Chatterjee and Hambrick 2007; Olsen, Dworkis, and Young 2013). While these studies have
treated personal pronouns and CEO photos as proxies for managerial traits, we are interested in
examining how users react to these style choices in an experimental setting that controls for underlying
firm characteristics.
Disclosure credibility and association with the message
Prior work argues that users’ reactions to a disclosure depend, in part, on the disclosure’s
credibility (Jennings 1987; Mercer 2004), which Mercer (2004) defines as “investors’ perceptions of the
believability of a particular disclosure” (p. 186). This literature suggests that the market responds more
strongly to more credible disclosures (Hutton and Stocken 2009; Ng, Tuna, and Verdi 2013) and that
disclosure credibility is influenced by what is communicated as well as by how it is communicated. For
example, disclosures are more credible if they provide more precise information (Hassell et al. 1988;
Hirst, Koonce, and Miller 1999; Hirst, Koonce, and Venkataraman 2007) or additional supporting
information (Hutton, Miller, and Skinner 2003; Baginski, Hassell, and Kimbrough 2004). Holding
constant what is explicitly stated in the disclosure, disclosure credibility is also influenced by the
disclosure’s time horizon (Pownall et al. 1993), medium (Elliott et al. 2012), and the credibility of the
individual or firm providing the disclosure (Hirst, Koonce, and Miller 1999; Hodge, Hopkins, and Pratt
2006).
8
We suggest that the extent to which managers choose to associate themselves with a disclosure
can play a similar role. When managers are highly associated with a disclosure, they assume greater
responsibility for its content. For example, statements made by managers can be directly attributed to
them and presumably reflect their views or beliefs. In contrast, when statements are made by “the
company,” responsibility is more diffuse. Because disclosures that turn out to be inaccurate (e.g.,
restatements, missed forecasts, etc.) elicit negative market reactions and can damage managers’
reputations (Desai, Hogan, and Wilkins 2006; Hutton and Stocken 2009), managers may be hesitant to
make their presence salient or prominent in the disclosure unless they are highly confident in the accuracy
of its content. This suggests that managers are able to signal the extent to which they believe a
disclosure’s content by choosing to associate themselves with (or distance themselves from) that
disclosure. As a result, disclosures that feature the manager more prominently might elicit stronger
reactions from investors by increasing their willingness to rely on the information therein.
This view is supported by prior communications research suggesting that self-mention (i.e.,
greater use of personal pronouns) is a rhetorical device that associates authors with their arguments
(Hyland 2001). Moreover, this literature suggests that self-mention emphasizes that the author believes
what they are saying (Hyland 2005a, b). Consistent with this idea, prior psycholinguistics research argues
that personal pronouns are reliable markers of a speaker’s belief in their message (Pennebaker 2011).
Prior psycholinguistics literature also finds that individuals with a desire to distance themselves from an
event use fewer first-person pronouns (Cohn, Mehl, and Pennebaker 2004).
Because the inclusion of a photo associates a manager more closely with a disclosure message,
the presence or absence of a manager’s photo could also reflect the manager’s degree of belief in the
information. Further, photos help individuals envision communicators actually speaking (Jameson 2000),
so a CEO’s photo may serve as an important cue to investors that the words represent the CEO’s
interpretation of the information and therefore their belief in the validity of the message. The
psycholinguistics, managerial narcissism, and managerial overconfidence literatures are largely silent on
how investors would respond to variation in managers’ pronoun or photo usage.
9
In accounting, our predictions are consistent with prior literature showing that style choices can
magnify investor reactions to disclosures. Sedor (2002), Rennekamp (2012), and Tan et al. (2014, 2015)
focus on the effects of style changes in the message. Elliott et al. (2012) show that video disclosures
increase participants’ reactions to the disclosure, even beyond the effects of a photo. They predict and
find that video increases participants’ reactions to the disclosure by increasing the salience of the CEO’s
perceived trustworthiness. Although Elliott et al. (2012) do not focus on how their video manipulations
associate managers with the disclosure, video disclosures may further associate a manager with a message
above and beyond the inclusion of a photo in a textual disclosure.3 Thus, our predictions may also offer an
interesting interpretation of results in prior literature.
In summary, we suggest that style choices that more closely associate managers with the
information in their disclosures (e.g., greater use of personal pronouns or the inclusion of a CEO’s photo)
will magnify investors’ reactions to firm disclosures.
HYPOTHESIS 1. Greater use of personal pronouns in a narrative disclosure will
magnify investor reactions to the news in the disclosure.
HYPOTHESIS 2. The inclusion of a CEO’s photo in a narrative disclosure will magnify
investor reactions to the news in the disclosure.
Disclosure design goals
Our experiments present participants with a short letter to shareholders and ask them to provide a
valuation judgment based on background information and the letter. Our first major goal in the design of
the letter is to vary the sign of the news without varying other content or style elements. We manipulate
the sign of news by substituting antonyms for the important nouns, adjectives, and adverbs used in the
3Our manipulation and contribution differs from Elliott et al. (2012) in several important respects. First, Elliott et al.
(2012) expose all participants to a photo of the CEO in their study, to ensure that all participants are familiar with
the CEO’s appearance. Second, the use of audio and video allows participants to make inferences about verbal and
non-verbal cues, which could include information about a firm’s future performance (Mayew and Venkatachalam
2012), and convey more than our construct of interest, association with the message. Finally, their theory relies more
on users’ “rich sensory experience provided by video”, and conscious interpretations of the trustworthiness of
managers. In contrast, our manipulation lacks the richness of video and our predictions do not rely on sensory
experience.
10
letter to indicate the sign of the news (above-below; higher-lower; strength-weakness; positively-
negatively; increased-decreased; grow-slow; satisfied-dissatisfied; improvement-deterioration). Our
second goal is to vary personal pronoun use (Experiments 1 and 3) or photo inclusion (Experiment 2)
without affecting disclosure content (nouns, adjectives, regular verbs, and adverbs) or other style
attributes, such as readability (SEC 1998). To accomplish this goal, we change only the last sentence of
the letter in Experiment 1 and include/exclude a single photo with no changes in the wording of the letter
in Experiment 2. In Experiment 3, we manipulate personal pronoun usage throughout a shorter disclosure,
increase our sample size, and collect other process measures in order to increase our ability to provide
insight about the process underlying our results. The stimuli are kept short and require only 3 to 4 minutes
to read and evaluate.
3. Experiment 1
Method
Design
In Experiment 1, we use a 2 x 2 between-subjects design, and manipulate (1) whether an
abbreviated letter to shareholders for a fictitious soft drink company conveys good or bad news about the
firm and (2) the rate of personal pronoun usage in the letter. Specifically, in the final sentence of the letter
to shareholders containing good (bad) news,4 the CEO indicates one of the following (italics added here
to highlight the pronoun manipulations; see Appendix 1 for illustrations of our Good News conditions):
“The Company is satisfied (dissatisfied) with its accomplishments in 2012, and expects further
improvement (deterioration) in operations in the near future.”
“I am satisfied (dissatisfied) with our accomplishments in 2012, and I expect further improvement
(deterioration) in operations in the near future.”
4 In the bad news condition, each of the words in parentheses in the examples is substituted for the immediately
preceding word.
11
The CEO’s photo is not included in any of the disclosures in Experiment 1, and the language used
throughout the disclosure is identical throughout the disclosure until the final sentence.5
Participants
Participants in Experiment 1 are 138 individuals recruited from Amazon’s Mechanical Turk
platform. Since our task requires that participants be able to read and understand English and pay careful
attention to the task, we recruit participants who are native English speakers and have an approval rate of
at least 95 percent.6 On average, participants are 32 years old and have 11 years of full-time work
experience. Fifty-five percent are male. Participants take an average of 3.2 minutes to complete the task
and receive $0.25 of compensation.
Procedures
Participants are randomly assigned to one of the four experimental conditions. Participants begin
the study by reading background information about a fictitious company (Cooper Soda Co.). They then
provide an initial judgment about the appropriate common stock valuation for the firm. Next, they view
the abbreviated letter to shareholders that contains our manipulations. After reviewing the letter,
participants again indicate their beliefs about the appropriate common stock valuation of the firm. They
then answer several debriefing and demographic questions. Once they move on to the next step, they are
prohibited from changing their answers to earlier parts of the task.
Dependent measure
Following prior literature, we elicit valuation judgments by asking participants to indicate on a
101-point scale what they believe to be an appropriate common stock valuation for the firm, ranging from
0 (“low”) to 100 (“high”) (Koonce and Lipe 2010). Participants make this judgment before and after
5 The abbreviated letter to shareholders is adapted from Rennekamp (2012).
6 We initially recruited 150 participants for Experiment 1. To prevent Workers from having multiple accounts, AMT
requires U.S. based workers to provide their social security number or individual tax identification number. In
addition, AMT prevents Workers from participating multiple times on the same task, and Qualtrics (which we use to
distribute our survey) uses cookies to prevent participants from participating multiple times in the same study. 10
participants were excluded for being repeat participants from pilot testing, and two participants were excluded for
being non-Native English speakers. Results from testing our predicted interaction are still significant if we include
all 150 participants.
12
viewing the letter to shareholders that contains our manipulations, and the difference between the two
measures serves as our primary dependent measure. This change measure captures the direction and
magnitude of participants’ reactions to the disclosure. Eliciting our dependent measure in this way is
designed to reduce noise in participants’ final judgments by starting them from a similar baseline of
beliefs, which should help us better detect the effects of our manipulations.
Follow-up questions
Following the valuation judgment, participants respond to several additional between-subjects
questions intended to measure whether pronoun use increases the extent to which participants believe (1)
that they could rely on the disclosure (“I felt like I could rely on the information in the Letter to
Shareholders”), (2) that the manager believes the disclosure’s message (“I felt like the management of
Cooper Soda really believed the information in the Letter to Shareholders”), and (3) that the manager is
competent (“How competent or incompetent do you believe the management of Cooper Soda to be?”) and
(4) that the manager is trustworthy (“How trustworthy or untrustworthy do you believe the management
of Cooper Soda to be?”). The first two measures test whether our manipulations affect participants’
perceptions about credibility of the disclosure itself (investors’ willingness to rely on the statements and
the strength of managers’ belief in the information), whereas the third and fourth measures test whether
our manipulations affect participants’ perceptions of the credibility of the management delivering the
information (competence and trustworthiness) (e.g., Mercer 2005; Koonce and Lipe 2010).
Results
Between-subjects tests
Our expectation is that investors will react more positively (negatively) to good (bad) news when
managers associate themselves with the disclosure by using personal pronouns at a higher rate in the
disclosure. Descriptive statistics for our primary dependent measure are presented in panel A of Table 1.7
<INSERT TABLE 1 HERE>
7 There are no significant differences in initial valuation judgments (i.e., before participants view the disclosure
containing our manipulations) across conditions (p=0.861, two-tailed).
13
Consistent with an effective manipulation of news valence, we find that participants adjust their
valuations positively after viewing the good news disclosure (p < 0.001, one-tailed, untabulated) and
negatively after viewing the bad news disclosure (p < 0.001, one-tailed). In addition, we find that this
effect is magnified when the CEO uses personal pronouns at a higher rate in the disclosure (p = 0.034,
one-tailed) (see panel B of Table 1).8 Further, we also find that the pronouns lead to both more negative
reactions for the bad news disclosure (p = 0.097, one-tailed), and more positive reactions for the good
news disclosure (p = 0.099, one-tailed), as indicated in panel C of Table 1.
Potential mediators
Participants perceive that management believes the bad news more than the good news (p =
0.003, two-tailed), that they can rely more on the bad news than good news (p < 0.001, two-tailed), and
that management is more competent in the good news condition than the bad (p < 0.001, two-tailed).
News does not affect the perceived trustworthiness of management (p = 0.360, two-tailed). In sum, the
sign of the news affects our measure of the strength of the manager’s belief in the message, participants’
willingness to rely on the information in the disclosure, and one of the two measures of manager
credibility. However, pronoun usage does not have a significant main effect on any of these possible
mediators: how strongly they agree that management believes the message being conveyed, how strongly
they believe that they can rely on the information, or perceptions of management’s competence and
trustworthiness (p = 0.656, p = 0.441, p = 0.528, and p = 0.310, two-tailed, respectively). Personal
pronoun usage also does not interact with news valence to affect any of these additional measures (all p >
0.409, two-tailed).9 Thus, while the overall pattern of results is consistent with H1, these measures do not
shed much light on the theoretical process underlying these results.
8 Note that all reported one-tailed p-values associated with tests using the F-distribution are one-tailed equivalents,
given our directional predictions. 9 After viewing the disclosures that included our manipulations, we also asked participants about their perceptions of
the firm’s future earnings, as well as the desirability of the firm overall as an investment. While correlated with our
primary valuation judgments (ρ >0.85 across both measures), these judgments are noisier in that we did not elicit a
pre-manipulation judgment to serve as a control. For both the measure of perceptions of future earnings and
14
Within-subjects tests
We next use a within-subjects study to test whether our pronoun manipulation operationalizes our
construct of interest and to examine whether participants consciously expect the manipulation to affect
valuation judgments. Participants for the within-subjects tests are also recruited from AMT (with the
requirement that they did not complete our primary experiments) and are randomly assigned to either the
good news or bad news condition. To facilitate comparison, we present participants with the disclosure
containing more personal pronouns side-by-side with the disclosure containing fewer personal pronouns.
The order of the two letters is varied and has no effect on the results. Participants rate their agreement
with the statement, “The letter that includes more personal pronouns (e.g., “I” and “our”) more closely
associates the CEO with the message contained in the letter” (on a 6-point scale ranging from 1 = strongly
disagree to 6 = strongly agree). Values of 4, 5, and 6 on the scale represent responses in agreement with
the statement. 98 percent of participants agree that the disclosure containing more pronouns more closely
associates the CEO with the message in the disclosure (p < 0.001, one-tailed). We interpret these findings
as supporting the validity of our operationalization of association with the message. In addition,
participants agree, on average, that the two disclosures would lead to the same valuation judgments by
investors (p < 0.001, one-tailed). 72 percent of the participants believe that investors would make the
same valuation in both cases. This suggests that the majority are not aware of the effects of personal
pronouns on investment evaluations, even when variation in pronoun usage is made obvious to them by
the side-by-side presentation.
Summary
In summary, we find that personal pronoun use magnifies participants’ reactions to the valence of
the message in a disclosure. Personal pronouns even enhance the effect of bad news, which is judged by
our participants to be more credible than good news. Participants’ beliefs about disclosure credibility and
management credibility do not mediate the effects on valuation judgments, and participants in our within-
desirability of the firm as an investment, results are directionally consistent with the valuation measure, although
insignificant.
15
subjects analysis do not expect personal pronoun use to affect investment evaluations. Together, this
evidence suggests that the between-subjects effect may be subconscious, consistent with prior research
suggesting that individuals often lack awareness of others’ or even their own use of personal pronouns
(Pennebaker 2011).
4. Experiment 2
Method
Design and procedures
In Experiment 2, we use a 2 x 2 between-subjects design and manipulate (1) whether an
abbreviated letter to shareholders conveys good or bad news about the firm and (2) whether the letter
contains a photo of the CEO (see Appendix 2 for illustrations of our Good News conditions). All other
aspects of the task and procedure are identical to Experiment 1.
The photo we use portrays a smiling CEO. To ensure that this is consistent with CEO photos in
practice, we examine CEO photos contained in the letters to shareholders for firms in the S&P 500. We
rank firms by their change in earnings for the most recent year (scaled by market cap) and categorize the
100 firms in the top (bottom) quintile as a “high performer” (“low performer”). On average, high
performers have a positive change in earnings while low performers have a negative change in earnings.
From this sample of 200 firms, we then collect the first CEO photo (if any) contained in their letter to
shareholders.
After collecting photos, two independent raters who are unaware of our hypotheses look at each
photo from our sample of S&P 500 firms individually, as well as the CEO photo we use in our study, and
indicate how happy each CEO appears in the photo on a 9-point scale (1 = “Very Unhappy”, 5 =
“Neutral”, and 9 = “Very Happy”; Intra-class correlation = 0.81). Performance does not affect the
likelihood of issuing a letter to shareholders (p = 0.480, two-tailed), but low performers are more likely to
include a photo in their letter to shareholders (p = 0.051, two-tailed). On average, CEOs are rated above
the midpoint (mean = 6.39; median = 6.5; p < 0.001, two-tailed), suggesting they generally look happy in
their photos. This result holds for low performers (mean = 6.43; median = 6.5; p < 0.001, two-tailed) and
16
high performers (mean = 6.34; median = 6.5; p < 0.001, two-tailed), and this difference is not significant
(p = 0.660, two- tailed). Finally, both raters rate our CEO as a “7”, which is about one-half of one
standard deviation above the mean rating for both high and low performers in our S&P 500 sample.10
Overall, these results suggest our photo is well within the normal range for CEO photos that appear in
letters to shareholders following both good and bad performance.
Participants
Participants in Experiment 2 are 104 Workers recruited from Amazon’s Mechanical Turk
platform.11
We again restrict our sample to Workers within the United States who are native English
speakers and have an approval rate of at least 95 percent. On average, participants are 32 years old and
have 12 years of full-time work experience. 53 percent are female. Participants take an average of 3.5
minutes to complete the task and receive $0.25 of compensation.
Results
Between-subjects tests
Our expectation is that investors will react more positively to good news and more negatively to
bad news when the manager associates himself with the disclosure by including his photo in the
disclosure. Descriptive statistics for our primary dependent measure are presented in panel A of Table 2.12
<INSERT TABLE 2 HERE>
Consistent with an effective manipulation of news valence, we find that participants adjust their
valuations positively after viewing the good news disclosure (p < 0.001, one-tailed) and negatively after
viewing the bad news disclosure (p < 0.001, one-tailed). In addition, we find that this effect is magnified
10
The photo that we use in our study is presented randomly within the sample of real photos from S&P 500 firms.
The two raters are unaware that the photo used in our study is not of a real CEO. 11
We initially recruited 120 participants for our task, although 121 participated. 16 participants have the same IP
address or AMT ID as participants from pilot testing, and these participants are excluded from analyses. We further
exclude one participant that identifies as a non-native English speaker because our task requires that participants be
able to read and interpret a disclosure. Results from testing for our predicted interaction are still significant if all
participants are included. 12
There are no significant differences in initial valuation judgments (i.e., before participants view the disclosure
containing our manipulations) across conditions (p=0.444, two-tailed).
17
when the CEO’s photo is included in the disclosure (p = 0.005, one-tailed) (see panel B of Table 2).
Further, as indicated in panel C of Table 2, we find that the photo leads to more positive reactions for the
good news disclosure (p = 0.098, one-tailed), but more negative reactions for the bad news disclosure (p =
0.009, one-tailed).
Mediators
News valence significantly affects only the two management credibility measures (p < 0.002,
two-tailed). Consistent with Experiment 1, the photo does not affect participants’ perceptions about how
strongly management believes the message being conveyed (p=0.395, two-tailed). Similarly, we find no
evidence that the photo increases their rating of the degree to which they could rely on the information (p
= 0.921, two-tailed), or perceptions of management’s competence or trustworthiness (p = 0.504 and p =
0.566, two-tailed, respectively), the two common components of perceived credibility of management in
prior research. The photo also does not interact with news valence to affect any of these additional
measures (all p > 0.365, two-tailed).13
Within-Subjects Tests
We next use a within-subjects study to test whether our photo manipulation operationalizes our
construct of interest and to examine whether participants consciously expect the manipulation to affect
valuation judgments. Participants for the within-subjects tests are also recruited from AMT (with the
requirement that they did not complete our primary experiments) and are randomly assigned to either the
good news or bad news condition. As in our personal pronoun study, we present participants with the
disclosure containing a photo side-by-side with the disclosure without the photo. Participants rate their
agreement with the statement, “The letter that includes the photo of the CEO more closely associates the
CEO with the message contained in the letter” (on a 6-point scale ranging from 1 = strongly disagree to 6
= strongly agree). Values of 4, 5, and 6 on the scale represent responses in agreement with the statement.
13
Consistent with the results for the main dependent variable, we find that the photo magnifies participants’
perceptions of future earnings (p = 0.046, one-tailed). The photo does not affect the perceived desirability of the
company as an investment opportunity (p = 0.872, two-tailed).
18
89 percent of the participants agree that the disclosure containing the CEO’s photo more closely
associates the CEO with the message in the disclosure (p < 0.001, one-tailed). We interpret these findings
as supporting the validity of our operationalization of association with the message. In addition,
participants agree, on average, that the two disclosures would lead to the same valuation judgments by
investors (p < 0.001, one-tailed). 69 percent of the participants believe that investors would make the
same valuation in both cases. This suggests that the majority are not aware of the effects of CEO photos
on investment evaluations, even when their usage is made obvious to them.
Summary
These results complement our results from Experiment 1. Specifically, we operationalize our
construct of interest (association with the message) using a very different style choice. Consistent with
our predictions, we document a similar pattern of results, as each style choice magnifies participants’
reactions to good and bad news disclosures. As in Experiment 1, responses to our follow-up questions do
not provide additional insight with respect to the process underlying these results. The within-subjects
tests and the mediation analysis using follow-up questions may reflect participants’ lack of insight into
the underlying process (Griffith et al. 2015).
5. Experiment 3
Our primary results from Experiments 1 and 2 are consistent with our predictions. However,
participants’ responses to follow-up questions did not provide additional insight into the underlying
process in either experiment. There are at least three possible explanations for this inconsistency. First,
it’s possible that participants have relatively little self-insight about the cognitive processes underlying
their judgments (Nisbett and Wilson 1977; Pennebaker 2011; Griffith et al. 2015). Second, the subtle
nature of our manipulations may limit our ability to observe the effects on our measures of the intervening
process. And third, it’s possible that our primary results are driven by an alternative process. For example,
managers’ association with the message might operate by changing participants’ psychological distance
from the firm or by changing perceived management control and involvement. We shed some additional
light on this issue in Experiment 3.
19
Method
Design and procedures
In Experiment 3, we provide an additional test of our theory using a 2 x 2 experiment. As in
Experiment 1, we manipulate (1) whether an abbreviated letter to shareholders conveys good or bad news
about the firm and (2) the rate of personal pronoun usage in the letter. In order to increase our ability to
provide insight about the process underlying our results, we use a shorter disclosure and manipulate
personal pronoun usage throughout that disclosure (rather than in only the final sentence) (see Appendix
3). We also increase our sample size and collect process measures to capture participants’ psychological
distance from the firm, perceptions of management control, and perceptions of management involvement.
Experimental procedures are otherwise identical to Experiments 1 and 2.
Follow-up questions
Following the valuation judgment, participants respond to several additional between-subjects
questions. As in Experiments 1 and 2, participants indicate the extent to which (1) they could rely on the
disclosure, (2) they believe that the manager believes the disclosure’s message (3) they believe that the
manager is competent, and (4) that the manager is trustworthy. In addition, participants rate the extent to
which they “believe the CEO has control over the performance of Cooper Soda” and the extent to which
they “believe the CEO is involved in the operations of Cooper Soda.” Participants are also asked to
estimate (in miles) the distance from their current location to Cooper Soda’s headquarters in Portland
Oregon. The difference between their estimate and the actual distance calculated from their IP address
serves as a measure of psychological distance (Elliott, Rennekamp, and White 2015). The order of these
follow-up questions is randomized between subjects.
Participants
Participants in Experiment 3 are 302 Workers recruited from Amazon’s Mechanical Turk
platform. We again restrict our sample to Workers within the United States who are native English
20
speakers and have an approval rate of at least 95 percent.14
On average, participants are 37.2 years old.
Fifty-six percent are female. Participants take an average of 4.92 minutes to complete the task and receive
$0.50 of compensation.
Results
Between-subjects tests
Our expectation is that investors will react more positively to good news and more negatively to
bad news when the manager associates himself with the disclosure. Descriptive statistics for our primary
dependent measure are presented in panel A of Table 3.15
<INSERT TABLE 3 HERE>
Consistent with an effective manipulation of news valence, we find that participants adjust their
valuations positively after viewing the good news disclosure (p < 0.001, one-tailed) and negatively after
viewing the bad news disclosure (p < 0.001, one-tailed). In addition, we find that this effect is magnified
when personal pronouns are used at a higher rate in the disclosure (p = 0.069, one-tailed) (see panel B of
Table 3). As indicated in panel C of Table 3, we find that reactions to the good news disclosure are
directionally more positive when the disclosure contains more personal pronouns, though this difference
is not significant (p = 0.256, one-tailed). In contrast, reactions to the bad news disclosure are more
negative when the disclosure contains more personal pronouns (p = 0.075, one-tailed).
Mediators
Disclosure Credibility: Consistent with the idea that greater use of personal pronouns increases
disclosure credibility, we find that greater use of personal pronouns increases the extent to which
participants felt they could rely on the disclosure (p = 0.036, two-tailed) and the extent to which
participants felt like the manager believed the information in the disclosure (p = 0.072, two-tailed). When
news is good, participants’ valuation judgments are positively associated with each of these measures
14
We exclude one Worker who is not a native English speaker. Inferences are unchanged if the Worker is included. 15
There are no significant differences in initial valuation judgments (i.e., before participants view the disclosure
containing our manipulations) across conditions (p=0.671, two-tailed).
21
(both p < 0.001, two-tailed). In contrast, when news is bad, participants’ valuation judgments are
negatively associated with each of these measures (both p < 0.001, two-tailed). A principal component
analysis for these two measures indicates that they load on a single factor (eigenvalue = 1.784, variance
explained = 89.2%), and the news contained in the disclosure moderates the effect of this factor on
participants’ valuation judgments (F(5, 295) = 37.18; p < 0.001, two-tailed). A path analysis confirms that
greater pronoun usage increases valuation judgments through this factor when news is good (z = 1.77; p =
0.077, two-tailed) and decreases valuation judgments through this factor when news is bad (z = -1.86; p =
0.062, two-tailed) (see Figure 1).16
These findings suggest that greater use of personal pronouns affects
participants’ valuation judgments by increasing disclosure credibility.
[INSERT FIGURE 1]
Management Credibility: Greater use of personal pronouns also increases the extent to which
participants view management as trustworthy (p = 0.030, two-tailed). Pronoun use does not affect
participants’ perceptions of management’s competence (p = 0.453, two-tailed). When news is good,
participants’ valuation judgments are positively associated with participants’ perceptions of management
trustworthiness (p = 0.011, two-tailed). However, when news is bad, participants’ valuation judgements
are not associated with their perceptions of management trustworthiness (p = 0.982, two-tailed).
Management Control and Involvement: Personal pronoun use does not affect the extent to
which participants believe the CEO has control over firm performance (p = 0.568, two-tailed) or the
extent to which participants believe the CEO is involved in the operations of the firm (p = 0.978, two-
tailed).
Psychological Distance: Greater use of personal pronouns also marginally increases the
perceived distance between participants’ current location and the firm’s headquarters (p = 0.087, two-
tailed), suggesting that greater personal pronoun usage might lead to greater psychological distance.
16
Following Hayes (2013), we estimate these conditional indirect effects using a moderated mediation model based
on 10,000 bootstrap samples.
22
However, this measure is not associated with participants’ valuation judgments when news is good (p =
0.217, two-tailed) or when news is bad (p = 0.746).
Summary
Overall, these results provide additional support for our predictions. Across three experiments, we
generally find that investors react more strongly to disclosures that feature the manager more
prominently.17
In addition, the results of the process measures in Experiment 3 are supportive of the idea
that managers’ association with the message increases the perceived credibility of the disclosure, which in
turn leads to stronger reactions to the information in the disclosure.
6. Conclusion
We predict that disclosure choices that more closely associate managers with the message in their
disclosures should magnify investors’ reactions to disclosures. Consistent with this theory, we find that
participants in within-subjects tests interpret both first-person pronoun use and inclusion of a CEO photo
as indicating greater association with the message. Between-subjects, we find that investors’ reactions to
disclosures are magnified when those disclosures use first-person personal pronouns at a higher rate or
contain a CEO’s photo, leading investors to react more positively when news is good, but more
negatively when news is bad. Within-subjects tests indicate that participants do not believe that their
investment decisions would be affected by either the use of pronouns or photos in the disclosures. The
combination of effects using between-subjects treatments but not using what are normally more powerful
within-subjects treatments, as has been demonstrated here, suggests that the between-subjects differences
are unintentional (Kahneman and Tversky 1996; Libby, Bloomfield, and Nelson 2002; Libby,
17
This conclusion is further supported by a supplemental experiment in which we again manipulate whether a
disclosure contains good or bad news and the use of personal pronouns. The disclosure consists of a short forward-
looking claim about the likely outcome of a company lawsuit. Participants are 50 accounting students from a top
undergraduate program, the dependent variable is a direct likelihood measure of belief as opposed to a valuation
judgment, and the first-person pronoun used is “we” (rather than “I”). We again find that increased use of personal
pronouns magnifies participants’ reactions to the disclosure (p < 0.001, one-tailed). Specifically, we find that
pronoun usage decreases participants’ likelihood assessments when news is bad (p = 0.010, one-tailed), but
increases likelihood assessments when news is good (p = 0.009, one-tailed). These results suggest that the results
from Experiments 1 and 3 are not specific to the AMT participants, the valuation dependent variable, or the use of
“I” versus “we.”
23
Rennekamp, and Seybert 2015). The process measures collected in Experiment 3 provide additional
support for the idea that managers’ association with the message increases disclosure credibility and
generally do not provide support for other potential mediators (e.g., management credibility, management
control and involvement, or psychological distance).
Our paper makes several contributions. First, these findings are related to prior research on the
inclusion of credibility-enhancing devices in disclosures. For example, prior research has examined
disclosures that include more forward-looking verifiable statements (Hutton, Miller, and Skinner 2003),
greater disaggregation of forecast components (Hirst, Koonce, and Venkataraman 2007; Lansford, Lev,
and Tucker 2013; Merkley, Bamber, and Christensen 2013), or less self-serving attributions (Baginski,
Hassell, and Hillison 2000; Baginski, Hassell, and Kimbrough 2004). While the judgment effects we
predict are similar to that of these credibility-enhancing devices, our paper differs from prior studies in
that most of the credibility-enhancing devices examined in these studies involve content, rather than style
choices. Further, our participants do not appear to consciously react to our pronoun and photo
manipulations.
Second, prior work in accounting, finance, and management has used CEO photos and personal
pronouns as proxies for unobservable management characteristics like narcissism and overconfidence,
and these characteristics have been interpreted predominantly as negative. Our findings suggest that
investors do not react to personal pronouns and photos as negative indicators of narcissism and
overconfidence, as those would likely predict a negative main effect on investment decisions, not the
interaction we find. Similarly, while photos with certain characteristics may elicit a positive or negative
response, these characteristics should also result in a main effect, and do not explain our interaction
pattern of results. Our pattern of results also cannot be explained by investors reacting to perceptions that
managers’ attributions in the disclosure were self-serving, given that we hold attributions constant across
both good and bad news conditions. Our theory and evidence presents a more nuanced view of the effect
of these style choices – style choices that more closely associate managers with their message magnify
investors’ reactions to disclosures.
24
Third, the SEC explicitly recommends the use of personal pronouns in firm disclosures (SEC
1998), suggesting a positive view of personal pronouns. Our results suggest one potential unintended
consequence of this recommendation. Because this style choice increases managers’ association with the
message, increasing personal pronoun usage might also magnify non-professional investors’ reactions to
disclosures. If managers’ use of personal pronouns or photos does not reflect the strength of their belief in
the message conveyed by the disclosure, this increased reaction may be unwarranted.
Finally, our paper also contributes to the broader literature on how style choices in corporate
disclosures affect investors’ judgments and decisions. Prior literature has either measured (see Li 2010a)
or manipulated (Sedor 2002; Elliott et al. 2012; Rennekamp 2012; Tan et al. 2014, 2015; Elliott,
Rennekamp, and White 2015; Asay et al. 2017) features like readability, concreteness, and/or disclosure
medium. We contribute to this literature by demonstrating that style choices that associate management
with the message have similar effects on non-professional investors’ judgments. A fruitful direction for
future research is to examine other style differences that may affect association with the message. For
example, managers could more closely associate themselves with a written disclosure message by
mentioning their names more frequently in the disclosure, including more direct quotes, or more clearly
indicating themselves as the speaker or narrator of the message. Managers’ association with the message
may also be influenced by disclosure medium and disclosure venue. For example, verbal disclosures (e.g.,
online videos or conference calls) may increase association with the message to a greater extent than
written disclosures, and disclosures provided via social media may increase association with the message
by making the disclosure seem more like a direct communication from management (e.g., a CEO vs.
corporate Twitter account). Thus, future work might extend our results by providing additional evidence
on features that associate management with the information in their disclosures. We leave it to future
research to investigate how investors react to other features that more closely associate managers with a
disclosure message, and what might be driving these responses.
Our disclosures were designed to be short, and our participants’ evaluations can best be described
as initial impressions. Given that initial impressions can be formed very quickly, and can have strong and
25
lasting impressions on subsequent evaluation of information (Russo, Meloy, and Wilks 2000; Wilks
2002), our results suggest that very subtle changes in disclosure features could have relatively strong
effects on investors’ reactions to information in the disclosure. However, the exact nature of their effects
on interpretation of more complex disclosures is a topic for future research.
26
Appendix 1 – Good News Conditions for Experiment 1*
Fewer Personal Pronouns: More Personal Pronouns:
* The bolded words in the final sentence of the disclosures were not bolded in the actual experiment. Emphasis has been added here to highlight the subtle
differences between the two disclosures.
27
Appendix 2 – Good News Conditions for Experiment 2
Photo Absent: Photo Present:
28
Appendix 3 – Good News Conditions for Experiment 3*
Fewer Personal Pronouns: More Personal Pronouns:
29
REFERENCES
Asay, S., W. B. Elliott, and K. Rennekamp. 2017. Disclosure readability and the sensitivity of investors’
valuation judgments to outside information. The Accounting Review 92 (4): 1-25.
Baginski, S. P., J. M. Hassell, and W. A. Hillison. 2000. Voluntary causal disclosures: tendencies and
capital market reaction. Review of Quantitative Finance and Accounting 15 (4): 371-389.
Baginski, S. P., J. M. Hassell, and M. Kimbrough. 2004. Why do managers explain their earnings
forecasts? Journal of Accounting Research 42 (1): 1-29.
Bowen, R. M., A. K. Davis, and D. A. Matsumoto. 2005. Emphasis on pro forma versus GAAP earnings
in quarterly press releases: Determinants, SEC intervention, and market reactions. The
Accounting Review 80 (4): 1011-1038.
Brochet, F., Miller, G., Naranjo, P., and G. Yu. 2016. Managers’ cultural background and disclosure
attributes. Working paper.
Cade, N. 2016. Corporate social media: How two-way disclosure channels influence investors. Working
paper.
Chatterjee, A. and D. C. Hambrick. 2007. It’s all about me: Narcissistic chief executive officers and their
effects on company strategy and performance. Administrative Science Quarterly 52 (3): 351-386.
Chen, Z. and S. Loftus 2017. Managers’ Self-inclusive language in conference calls: Multi-method
evidence. Working paper. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2950702
Cohn, M. A., M. R. Mehl, and J. W. Pennebaker. 2004. Linguistic markers of psychological change
surrounding September 11, 2001. Psychological Science 15 (10): 687-693.
Davis, A. K., J. M. Piger, and L. M. Sedor. 2012. Beyond the numbers: Measuring the information
content of earnings press release language. Contemporary Accounting Research 29 (3): 845-868.
Davis, A. K. and I. Tama-Sweet. 2012. Managers’ use of language across alternative disclosure outlets:
Earnings press releases versus MD&A. Contemporary Accounting Research 29 (3): 804-837.
Demers, E. and C. Vega. 2014. The impact of credibility on the Pricing of managerial textual content.
Working paper, The University of Virginia and the Board of Governors of the Federal Reserve
System. Available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1153450.
Desai, H., C. E. Hogan, and M. S. Wilkins. 2006. The reputational penalty for aggressive accounting:
Earnings restatements and management turnover. The Accounting Review 81 (1): 83-112.
Elliott, W. B. 2006. Are investors influenced by pro forma emphasis and reconciliation in earnings
announcements? The Accounting Review 81 (1): 113-133.
Elliott, W. B., F. D. Hodge, and L. M. Sedor. 2012. Using online video to announce a restatement:
Influences on investor decisions and the mediating role of trust. The Accounting Review 87 (2):
513-535.
Elliott, W. B., K. Rennekamp, and B. White. 2015 Does concrete language in disclosures increase
willingness to invest? Review of Accounting Studies 20 (2): 839-865.
30
Feldman, R., S. Govindaraj, J. Livnat, and B. Segal. 2010. Management’s tone change, post earnings
announcement drift and accruals. Review of Accounting Studies 15 (4): 915-953.
Gow, I.D., Kaplan, S.N., Larcker, D.F., and A. Zakolyukina. 2015. CEO personality and firm policies.
Working paper.
Griffith, E., Kadous, K., and D. Young. 2015. How insights from the “new” JDM research can improve
auditor judgment: Fundamental research questions and methodological advice. Auditing: A
Journal of Practice & Theory 35 (2): 1-22.
Hales, J., X. Kuang, and S. Venkataraman. 2011. Who believes the hype? An experimental examination
of how language affects investor judgments. Journal of Accounting Research 49 (1): 223-255.
Hassell, J. M., R. H. Jennings, and D. Lasser. 1988. Management earnings forecasts: Their usefulness as a
source of firm-specific information to security analysts. Journal of Financial Research 11
(Winter): 303-319.
Hayes, A.F. 2013. Introduction to mediation, moderation, and condition process analysis: A regression-
based approach. Guilford Press.
Hirst, D. E., L. Koonce, and J. Miller. 1999. The joint effect of management’s prior forecast accuracy and
the form of its financial forecasts on investor judgment. Journal of Accounting Research 37 (1):
101-124.
Hirst, D. E., L. Koonce, and S. Venkataraman. 2007. How disaggregation enhances the credibility of
management earnings forecasts. Journal of Accounting Research 45 (4): 811-837.
Hodge, F., P. Hopkins, and J. Pratt. 2006. Management reporting incentives and classification credibility:
The effects of reporting discretion and reputation. Accounting, Organizations and Society 31 (7):
623-634.
Horton, J. J., D. G. Rand, and R. J. Zeckhauser. 2011. The online laboratory: Conducting experiments in a
real labor market. Experimental Economics 14 (1): 399-425.
Huang X., S. H. Teoh, and Y. Zhang. 2014. Tone management. The Accounting Review 89 (3): 1083-
1113.
Hutton, A., G. Miller, and D. Skinner. 2003. The role of supplementary statements with management
earnings forecasts. Journal of Accounting Research 41 (5): 867-890.
Hutton, A., and P.C. Stocken. 2009. Prior forecasting accuracy and investor reaction to management
earnings forecasts. Working paper.
Hyland, K. 2001. Humble servants of the discipline? Self-mention in research articles. English for
Specific Purposes 20 (3): 207-226.
Hyland, K. 2005a. Metadiscourse: Exploring Interaction in Writing. London: Continuum.
Hyland, K. 2005b. Stance and engagement: a model of interaction in academic discourse. Discourse
Studies 7 (2): 173-192.
Jameson, D. 2000. Telling the investment story: a narrative analysis of shareholder reports. Journal of
Business Communication 37 (1): 7-38.
31
Jennings, R. 1987. Unsystematic security price movements, management earnings forecasts, and revisions
in consensus analyst earnings forecasts. Journal of Accounting Research 25 (1): 90-110.
Kahneman, D., & Tversky, A. 1996. On the reality of cognitive illusions. Psychological Review 103 (3):
582–591.
Koonce, L. and M. Lipe. 2010. Earnings trend and performance relative to benchmarks: How consistency
influences their joint use. Journal of Accounting Research 48 (4): 859-884.
Kothari, S. P., X. Li, and J. E. Short. 2009. The effect of disclosures by management, analysts, and
business press on cost of capital, return volatility, and analyst forecasts: A study using content
analysis. The Accounting Review 84 (5): 1639-1670.
Lansford, B., B. Lev, and J. W. Tucker. 2013. Causes and consequences of disaggregating earnings
guidance. Journal of Business Finance and Accounting 40 (1-2): 26-54.
Larcker, D. F. and A. A. Zakolyukina. 2012. Detecting deceptive discussions in conference calls. Journal
of Accounting Research 50 (2): 495-540.
Li, F. 2008. Annual report readability, current earnings, and earnings persistence. Journal of Accounting
and Economics 45 (2-3): 221-247.
Li, F. 2010a. Textual analysis of corporate disclosures: A survey of the literature. Journal of Accounting
Literature 29 (1): 143-165.
Li, F. 2010b. The information content of forward-looking statements in corporate filings – A naïve
Bayesian machine learning approach. Journal of Accounting Research 48 (5): 1049-1102.
Li, F. 2011. Managers’ self-serving attribution bias and corporate financial policies. Working paper.
Libby, R., R. Bloomfield, and M. W. Nelson. 2002. Experimental research in financial accounting.
Accounting, Organizations and Society 27 (8): 775–810
Libby, R., K. Rennekamp, and N. Seybert. 2015. Regulation and the interdependent roles of managers,
auditors and directors on earnings management and accounting choice. Accounting,
Organizations and Society 47 (November): 25-42.
Mayew, W. J. and M. Venkatachalam. 2012. The power of voice: Managerial affective states and future
firm performance. The Journal of Finance 67 (1): 1-43.
Mercer, M. 2004. How do investors assess the credibility of management disclosures? Accounting
Horizons 18 (3): 185-196.
Mercer, M. 2005. The fleeting effects of disclosure forthcomingness on management’s reporting
credibility. The Accounting Review 80 (2): 723-744.
Merkl-Davies, D. M. and N. Brennan. 2007. Discretionary disclosure strategies in corporate narratives:
incremental information or impression management? Journal of Accounting Literature 26 (1):
116-194.
Merkley, K. J., L. S. Bamber, and T. E. Christensen. 2013. Detailed management earnings forecasts: do
analysts listen? Review of Accounting Studies 18 (2): 479-521.
32
Merkley, K. J. 2014. Narrative disclosure and earnings performance: Evidence from R&D disclosures.
The Accounting Review 89 (2): 725-757.
Miller, B. 2010. The effects of reporting complexity on small and large investor trading. The Accounting
Review 85 (6): 2107-2143.
Ng, J., I. Tuna, and R. Verdi. 2013. Management forecast credibility and underreaction to news. Review
of Accounting Studies 18 (4): 956-986.
Nisbett, R. E. and T. D. Wilson. 1977. Telling more than we can know: Verbal reports on mental
processes. Psychological review 84 (3): 231-259.
Olsen, K., K. K. Dworkis and S. M. Young. 2013. CEO narcissism and accounting: A picture of profits.
Journal of Management Accounting Research 26 (2014): 243-267.
Pennebaker, J. W. 2011. The Secret Life of Pronouns: What Our Words Say about Us. New York, NY:
Bloomsbury Press.
Pownall, G., C. Wasley, and G. Waymire. 1993. The stock price effects of alternative types of
management earnings forecasts. The Accounting Review 68 (4): 896-912.
Rennekamp, K. 2012. Processing fluency and investors’ reactions to disclosure readability. Journal of
Accounting Research 50 (5): 1319-1354.
Russo, J. E., M. G. Meloy, and T. J. Wilks. 2000. Predecisional distortion of information by auditors and
salespersons. Management Science 46 (1): 13-27.
Schrand, C. and S. L. C. Zechman. 2012. Executive overconfidence and the slippery slope to financial
misreporting. Journal of Accounting and Economics 53 (1-2): 311-329.
Securities and Exchange Commission (SEC). 1998. A Plain English Handbook: How to Create Clear
SEC Disclosure. Washington, DC: SEC Office of Investor Education and Assistance. Retrieved
from the SEC’s website: http://www.sec.gov/pdf/handbook.pdf.
Sedor, L. M. 2002. An explanation for unintentional optimism in analysts’ earnings forecasts. The
Accounting Review 77 (4): 731-753.
Tan, H-T., E. Y. Wang, and B. Zhou. 2014. When the use of positive language backfires: The joint effect
of tone, readability and investor sophistication on earnings judgments. Journal of Accounting
Research 52 (1): 273-302.
Tan, H-T., E. Y. Wang, and B. Zhou. 2015. How does readability influence investors' judgments?
Consistency of benchmark performance matters. The Accounting Review 90 (1): 371-393.
Wilks, T. J. 2002. Predecisional distortion of evidence as a consequence of real-time audit review. The
Accounting Review 77 (1): 51-71.
You, H. and X. Zhang. 2009. Financial reporting complexity and investor underreaction to 10-K
information. Review of Accounting Studies 14 (4): 559-586.
33
FIGURE 1
Indirect Effect of Personal Pronoun Usage (through Disclosure Credibility) on Change in Valuations
(Experiment 3)
This figure displays the observed coefficients and (standard errors) from a moderated mediation model (Hayes 2013)
for Experiment 3. Pronoun Usage is coded as -1 when the disclosure contained fewer personal pronouns and 1 when
the disclosure contained more personal pronouns. News is coded as -1 when the disclosure contained bad news and
1 when the disclosure contained good news. Disclosure credibility is a factor estimated from a principal component
analysis on participants’ ratings of (1) the extent to which they felt they could rely on the disclosure and (2) the
extent to which they felt like the manager believed the information in the disclosure. These two measures load on a
single factor (eigenvalue = 1.784, variance explained = 89.2%).
Pronoun
Usage
Disclosure
Credibility
(Factor)
Change in
Valuation
News
Disclosure
Credibility
x
News
0.159
(0.077)
p = 0.038
-0.512
(0.549)
p = 0.351
-0.605
(0.742)
p = 0.415
17.379
(0.749)
p < 0.001
3.590
(0.549)
p < .001
34
TABLE 1
Descriptive Statistics and Tests of H1 (Experiment 1)
Panel A: Descriptive Statistics – Change in Valuation
Fewer Personal
Pronouns
More Personal
Pronouns Row Total
n
Mean
[std dev]
n
Mean
[std dev]
n
Mean
[std dev]
Good News 35 16.14
[9.14] 34
19.82
[13.37] 69
17.96
[11.21]
Bad News 34 -22.82
[13.46] 35
-26.54
[11.39] 69
-24.71
[12.50]
Column Total 69 -3.06
[22.69] 69
-3.70
[26.28]
Panel B: ANOVA Model of Change in Valuation
Source of Variation SS df MS F-stat p-value
News 62791.317 1 62791.317 449.8798 <0.001a
Personal Pronouns 0.013 1 0.013 0.0001 0.992b
News x Personal Pronouns 472.206 1 472.206 3.3832 0.034a
Error 18702.854 134 139.574
Panel C: Follow-up Tests of Simple Effects
Source of Variation df F-stat p-value
Effect of personal pronouns given good news. 1 1.674 0.099a
Effect of personal pronouns given bad news. 1 1.709
0.097a
This table presents descriptive statistics (Panel A) and tests of H1 (Panels B and C) for Experiment 1, where we use
a 2x2 between-subjects design and manipulate (1) whether the disclosure conveys good or bad news and (2) whether
the CEO uses personal pronouns in the last sentence of the disclosure. Before receiving the disclosure containing our
manipulations, participants provide judgments on a 101-point scale about the appropriate valuation for the firm (0 =
“Low” to 100 = “High”). Participants again provide the judgment after receiving the disclosure containing our
manipulations. This table summarizes the change between the initial and final judgment. a One-tailed equivalents, given our directional predictions.
b Two-tailed.
35
TABLE 2
Descriptive Statistics and Tests of H2 (Experiment 2)
Panel A: Descriptive Statistics – Change in Valuation
CEO Photo
Absent
CEO Photo
Present Row Total
n
Mean
[std dev]
n
Mean
[std dev]
n
Mean
[std dev]
Good News 27 15.81
[9.95] 28
19.54
[10.55] 55
17.71
[10.34]
Bad News 24 -17.88
[14.15] 25
-25.24
[6.89] 49
-21.63
[11.55]
Column Total 51 -0.04
[20.78] 53
-1.58
[24.27]
Panel B: ANOVA Model of Change in Valuation
Source of Variation SS df MS F-stat p-value
News 39871.256 1 39871.256 352.2126 <0.001a
CEO Photo 85.997 1 85.997 0.7597 0.386b
News x CEO Photo 795.873 1 795.873 7.0305 0.005a
Error 11324.558 101 112.100
Panel C: Follow-up Tests of Simple Effects
Source of Variation df F-stat p-value
Effect of CEO photo given good news. 1 1.681 0.098a
Effect of CEO photo given bad news. 1 5.867
0.009a
This table presents descriptive statistics (Panel A) and tests of H2 (Panels B and C) for Experiment 2, where we use
a 2x2 between-subjects design and manipulate (1) whether the disclosure conveys good or bad news and (2) whether
a photo of the CEO is included with the disclosure. Before receiving the disclosure containing our manipulations,
participants provide judgments on a 101-point scale about the appropriate valuation for the firm (0 = “Low” to 100 =
“High”). Participants again provide the judgment after receiving the disclosure containing our manipulations. a One-tailed equivalents, given our directional predictions.
b Two-tailed.
36
TABLE 3
Descriptive Statistics and Tests of H1 (Experiment 3)
Panel A: Descriptive Statistics – Change in Valuation
Fewer Personal
Pronouns
More Personal
Pronouns Row Total
n
Mean
[std dev]
n
Mean
[std dev]
n
Mean
[std dev]
Good News 76 10.30
[13.22] 75
11.80
[10.69] 151
11.05
[12.01]
Bad News 76 -22.16
[17.13] 74
-25.46
[14.12] 151
-23.79
[15.75]
Column Total 152 -5.93
[22.31] 149
-6.70
[22.47]
Panel B: ANOVA Model of Change in Valuation
Source of Variation SS df MS F-stat p-value
News 91434.074 1 91434.074 467.331 <0.001a
Personal Pronouns 61.230 1 61.230 0.313 0.576b
News x Personal Pronouns 433.194 1 433.194 2.214 0.069a
Error
Panel C: Follow-up Tests of Simple Effects
Source of Variation df F-stat p-value
Effect of personal pronouns given good news. 1 0.433 0.256a
Effect of personal pronouns given bad news. 1 2.089 0.075a
This table presents descriptive statistics (Panel A) and tests of H1 (Panels B and C) for Experiment 3, where we use
a 2x2 between-subjects design and manipulate (1) whether the disclosure conveys good or bad news and (2) the rate
of personal pronoun usage in the disclosure. Before receiving the disclosure containing our manipulations,
participants provide judgments on a 101-point scale about the appropriate valuation for the firm (0 = “Low” to 100 =
“High”). Participants again provide the judgment after receiving the disclosure containing our manipulations. This
table summarizes the change between the initial and final judgment. a One-tailed equivalents, given our directional predictions.
b Two-tailed.