Description
Internal auditors play an important role in influencing managers’ judgments. Yet, the
practitioner literature indicates that, because internal audit lacks the client services incen-
tives of external audit, internal auditors often adopt a ‘‘policeman approach’’ that can lead
to negative interpersonal relationships with managers. We investigate three variables fun-
damental to internal auditors’ ability to influence managers: (1) internal auditors’ interper-
sonal likability, (2) the information used to support their positions, and (3) whether they
present that information in a thematically organized argument. We find that managers
agree more with an internal auditor who is both likable and uses a thematically organized
argument. We find further that this joint effect occurs regardless of whether the internal
auditor’s information is relatively supportive or unsupportive of his position. Overall, our
theory and findings suggest that an internal auditor can achieve agreement from managers
on important corporate governance issues with this fairly straightforward presentation
tactic, even when the underlying information is relatively unsupportive and managers
otherwise tend not to agree with the internal auditor’s position
Internal auditors’ use of interpersonal likability, arguments, and
accounting information in a corporate governance setting
q
Kirsten Fanning
a
, M. David Piercey
b,?
a
College of Business, University of Illinois at Urbana-Champaign, Champaign, IL 61820, United States
b
Isenberg School of Management, University of Massachusetts, Amherst, MA 01003, United States
a b s t r a c t
Internal auditors play an important role in in?uencing managers’ judgments. Yet, the
practitioner literature indicates that, because internal audit lacks the client services incen-
tives of external audit, internal auditors often adopt a ‘‘policeman approach’’ that can lead
to negative interpersonal relationships with managers. We investigate three variables fun-
damental to internal auditors’ ability to in?uence managers: (1) internal auditors’ interper-
sonal likability, (2) the information used to support their positions, and (3) whether they
present that information in a thematically organized argument. We ?nd that managers
agree more with an internal auditor who is both likable and uses a thematically organized
argument. We ?nd further that this joint effect occurs regardless of whether the internal
auditor’s information is relatively supportive or unsupportive of his position. Overall, our
theory and ?ndings suggest that an internal auditor can achieve agreement from managers
on important corporate governance issues with this fairly straightforward presentation
tactic, even when the underlying information is relatively unsupportive and managers
otherwise tend not to agree with the internal auditor’s position. Our study contributes to
accounting, psychology, and writing and discourse theories with new evidence of the
effects of an argument structure (holding the underlying information constant) on users’
judgments, and how those effects depend on the likability of the source of information.
Our ?ndings have important implications for internal auditors, managers, external
auditors, and others interested in corporate governance.
Ó 2014 Elsevier Ltd. All rights reserved.
Introduction
In this study, we examine how three variables, each
fundamental to internal auditors’ interactions with manag-
ers, explain internal auditors’ in?uence on managers’ judg-
ments: (1) internal auditors’ interpersonal likability, (2)
the underlying information supporting their positions,
and (3) their use of thematically organized arguments to
present that information to managers. As Prawitt, Smith,
and Wood (2009, p. 1258) point out, internal auditors tend
to interact with managers frequently, and are ‘‘often the
party primarily responsible for the day-to-day monitoring
of management’s actions, including those related to
external ?nancial reporting’’ (see also Bariff, 2003; Cohen,http://dx.doi.org/10.1016/j.aos.2014.07.002
0361-3682/Ó 2014 Elsevier Ltd. All rights reserved.
q
We are grateful for the comments and assistance of Chris Agoglia, Jim
Bierstaker, Joe Brazel, Tom Clausen, Joshua Herbold, Jessen Hobson, Carol
Jessup, Eric Johnson, Kathryn Kadous, Tom Kida, Benjamin Luippold,
Mario Maletta, Mike Peters, Kimberly Moreno, Robert Resutek, Chad
Simon, Jim Smith, Dan Stone, Hun Tong Tan, David Wood, Arnie Wright,
May Zhang, and the editors and two anonymous reviewers. We also thank
workshop participants at the University of Massachusetts, Northeastern
University, and Villanova University, and conference participants at the
American Accounting Association (AAA) Annual Meeting, the AAA
Accounting, Behavior, & Organizations Research Conference, and Brigham
Young University Accounting Research Symposium for their helpful
comments. We thank Sudip Bhattacharjee, Kimberly Moreno, and Tracey
Riley for access to their experimental materials.
?
Corresponding author. Tel.: +1 217 300 1981/413 545 5585.
E-mail addresses: [email protected] (K. Fanning), piercey@isenberg.
umass.edu (M. David Piercey).
Accounting, Organizations and Society 39 (2014) 575–589
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
Hayes, Krishnamoorthy, Monroe, & Wright, 2013; Mercer,
2004).
The practitioner literature indicates that the tone of the
interactions between internal auditors and managers
varies widely in practice (Deloitte, 2010; Dittenhofer,
Ramamoorti, Ziegenfuss, & Evans, 2010; Pickett, 2010;
Ratliff & Brackner, 1998). Clearly, many internal auditors
have exceptional interpersonal interactions with manag-
ers. However, Pickett (2007, 2010) notes that internal audit
lacks the client services incentives of external audit, allow-
ing internal auditors to adopt a ‘‘policeman approach,’’
which places little emphasis on positive interpersonal
interactions with managers as clients, compared to exter-
nal audit. Deloitte (2010) similarly notes that the ‘‘police’’
approach to internal audit can harm the manager–internal
auditor relationship, and contend that ‘‘a dysfunctional
relationship [between managers and internal auditors] is
a contributing cause, and in some cases, a primary cause’’
of a variety of accounting problems, including ‘‘material
weaknesses, ?nancial restatement, regulatory compliance,
and the like’’ (p. 3). Despite the importance of the interper-
sonal relationship between internal auditors and manag-
ers, Archambeault, DeZoort, and Holt (2008), and Prawitt
et al. (2009) point out that it has received relatively little
research.
We develop theory and experimentally test how inter-
nal auditors can use arguments, personal likability, and
information to in?uence managers’ judgments. According
to theories of writing and discourse, an argument is a
?owing arrangement of information into thematically con-
nected groups in support of a particular conclusion (e.g.,
Conners, 1981; Smith, 2003). However, individuals often
simply provide information in support of a position as it
comes to mind, without organizing it into a structured
argument, leaving it to the user to decide how the pieces
?t together (Booher, 2001; Guffey, 2010; May & May,
2012). The internal audit practitioner literature indicates
a wide variance in the effectiveness of internal auditors’
use of arguments in practice (Chambers, 2009;
Dittenhofer et al., 2010). For example, Chambers (2009)
interviewed managers and executives who indicate that
internal auditors often do not provide information in a
manner that allows users to easily see how related pieces
of information connect together. In other words, internal
auditors provide information, but without always structur-
ing it into a coherent, thematically organized, ?owing
argument (cf. May & May, 2012).
Writing and discourse scholars distinguish arguments
(which organize information supporting a position into
thematically connected groups) from at least three other
forms of rhetoric: narration (connecting information into
temporal order), description, and exposition (e.g.,
Conners, 1981; Smith, 2003). As an example, an internal
auditor using an argument to recommend a write-down
of obsolete inventory (cf. KPMG, 2003) could structure
information into thematically connected groups by, for
example, ?rst introducing the state of the inventory and
its competition, then discussing information about how
slowly the inventory is selling, followed by details about
the technologically superior products, then discussing the
viability of any other sales prospects for the older
inventory, etc. Prior research has shown how other ways
of structuring information (holding the underlying
information constant) in?uences how users react to that
information (e.g., Lipe & Salterio, 2002; Ricchiute, 1992;
Sedor, 2002). For example, Sedor (2002) manipulated
whether analysts received optimistic earnings guidance
in temporal, causal narratives that linked past states of
the ?rm to current states to plans for the future, or the
same information in randomized order. Earnings forecasts
were more optimistic when a temporally causal narrative
format was used.
In our experiment, managers provide a controller with
their input into an inventory write-down judgment for
their divisions (e.g., Duncan, 2002), while interacting with
an internal auditor who prefers conservatively writing-
down the value of the inventory in the ?nancial statements
(e.g., KPMG, 2003; Mercer, 2004; Moeller, 2009; Prawitt
et al., 2009). We examine this setting within a 2 Â 2 Â 2
experimental framework, in which the internal auditor is
either interpersonally likable or dislikable, and presents
information that is either more supportive or less support-
ive of write-down, in either a coherent, thematically ?ow-
ing argument structure or not. Similar to Sedor’s (2002)
manipulation of temporal narratives, our manipulation of
arguments holds the underlying information about inven-
tory constant by comparing the argument condition to a
condition in which the internal auditor presents the same
statements about inventory in an unorganized order.
We combine theories from writing and discourse,
psychology, and accounting to build our predictions.
Because people ?nd the thematically structured ?ow of
an argument appealing, and because positive affective
states lead to heuristic processing, we predict that manag-
ers will heuristically agree more with an internal auditor
who is both likable and uses an argument structure,
beyond the effects of how supportive or unsupportive the
internal auditor’s information is of his position. Our ?nd-
ings are consistent with this hypothesis. First, our most
basic ?nding is that managers (unsurprisingly) agree more
with an auditor who uses more supportive information
than one who uses less supportive information. However,
beyond that, they also agree more with an internal auditor
who is both likable and uses a thematically organized argu-
ment structure, regardless of whether the information pre-
sented is relatively supportive or unsupportive of the
internal auditor’s position. In fact, our results demonstrate
that an internal auditor can achieve (on average) agree-
ment from managers simply because he is likable and uses
a ?owing argument structure, even when the underlying
information is relatively unsupportive and managers
otherwise (on average) do not to agree with the internal
auditor. Overall, our theory and ?ndings suggest that
internal auditors can achieve additional agreement from
managers on important corporate governance issues,
above and beyond how supportive or unsupportive their
information is, by using an argument structure and likabil-
ity jointly, as a fairly straightforward presentation tactic.
This is the ?rst study of which we are aware to demon-
strate how structuring information into thematically
organized arguments (holding constant the underlying
information presented) interacts with an information
576 K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589
provider’s likability to prompt additional agreement from
users that would not otherwise occur. As a result, we
contribute to both the research literature on interpersonal
likability (e.g., Kida, Moreno, & Smith, 2001; Moreno, Kida,
& Smith, 2002), as well as the literatures on persuasion and
other related theories (e.g., Kadous, Leiby, & Peecher, 2013;
Pennington & Hastie, 1986, 1988, 1993; Rich, Solomon, &
Trotman, 1997; Sedor, 2002; Suedfeld, Tetlock, &
Streufert, 1992). In addition, our study also has other
important implications for accounting theory, research
and practice. Accounting is not just about the mere aggre-
gation of data, but how accountants present that data to
other social beings in a social–organizational setting
(Bonner, 2008; Kachelmeier, 2010; Kinney, 2001). While
prior research has shown that internal auditors have at
least some impact on managers’ judgments (Prawitt
et al., 2009), our study can help develop a theory about
day-to-day behavioral factors that drive variance in inter-
nal auditors’ in?uence over managers’ judgments, as called
for by Prawitt et al. (2009) and Archambeault et al. (2008).
Internal auditing is a ‘‘relationship and communications
business’’ in which information is communicated within
an organizational context to in?uence managers
(Dittenhofer et al., 2010). As Kachelmeier (2010) puts it,
?rms do not make accounting decisions, people make
accounting decisions, and those decisions are shaped by
the behavioral interactions of individuals within the social
environment of their ?rms. Our study can help form an
empirical basis for the soft skills education and training
called for in the internal audit practitioner literature (e.g.,
Dittenhofer et al., 2010; IIA, 2013a, 2013b; Moeller, 2009;
Pickett, 2010). In addition, our ?ndings should also be
informative to external auditors, managers, researchers,
and others interested in in?uencing managers’ judgments
and corporate governance.
Theory
Interpersonal likability
As noted in the introduction, the practitioner litera-
ture indicates that the tone of the interactions between
internal auditors and managers varies widely in practice,
with some internal auditors occasionally adopting a
‘‘police’’ approach that can lead to dysfunctional
manager–internal audit relationships (e.g., Deloitte.,
2010; Pickett, 2007, 2010). Prior research suggests that
managers have affective reactions to others’ interper-
sonal likability as they interact with them within organi-
zational settings (Kida et al., 2001; Moreno et al., 2002).
Individuals are more likely to incorporate their affective
reactions into subsequent judgments when other task
factors prompt heuristic processing (e.g., Chaiken, 1980;
Finucane, Alhakami, Slovic, & Johnson, 2000; Schwarz,
Strack, Kommer, & Wagner, 1987; Shiv & Fedorikhin,
1999; Siemer & Reisenzein, 1998). This research suggests
that, for our setting, when managers do incorporate their
affective reactions into their subsequent judgments, they
would tend to respond positively (negatively) to a likable
(dislikable) internal auditor.
Argument structure and likability
An argument is a form of rhetoric which structures
underlying information into thematically related groups
of connected ideas that ?ow together to support a particu-
lar position (e.g., Conners, 1981; Smith, 2003). For exam-
ple, an individual trying to persuade others to support a
particular political candidate could organize supportive
information into thematically connected groups by, e.g.,
?rst discussing information about the high moral character
of the politician, then various advantages of the politician’s
social policy, followed by various advantages of the politi-
cian’s ?scal policy, then information about the politician’s
experience and history in past public of?ces, etc., in a
coherent argument structure. Alternatively, individuals
can convey the same information supporting a particular
position, but without structuring it into an argument
(May & May, 2012).
1
An argument structure provides an apparent, appealing
thematic ?owto the information. Kida (2006) suggests that
individuals naturally prefer verbally ?owing accounts to
raw presentations of data. We propose that an internal
auditor’s use of an argument structure (holding constant
the underlying information) can create additional agree-
ment from managers, without increased attention to how
supportive (or unsupportive) the presented information
actually is of the internal auditor’s position. Speci?cally,
this additional agreement is likely to be the result of a
heuristic whereby people simply respond favorably to the
apparent thematic ?ow that an argument structure pro-
vides. The strength of the underlying information would
still have a predictable directional effect on managers’
agreement with the internal auditor’s position. But the
1
Prior accounting research has examined constructs that share similar-
ities with arguments, yet also have important differences as well.
Compared to arguments (which arrange thematically related pieces of
information together in support of a position), the ‘‘scenarios’’ construct in
Sedor (2002) matches what writing and discourse scholars refer to as
‘‘narratives,’’ a different form of rhetoric from ‘‘arguments’’ that arranges
information temporally (e.g., Conners, 1981; Smith, 2003). In fact, Sedor
(2002) describes the ‘‘scenarios’’ construct as ‘‘narratives that concretely
describe the sequence of events in which proposed actions lead to future
outcomes’’ (p. 734), and the theory relies on narratives laying out the causal
links in the temporal sequence from past states to current states to future
outcomes in order to in?uence forecasts of the future. In addition, similar to
our setting of internal auditors providing input (with or without an
argument structure) to managers, Kadous et al. (2013) examine the effects
of an auditor providing input (with either high or low integrative
complexity) to other auditors. Low integrative complexity involves simple,
one-sided justi?cations of advice that do not go into much depth and are
not self-critical in the sense that they do not consider tradeoffs on both
sides of the issue. High integrative complexity involves deeper justi?ca-
tions that consider more information, including information on both sides
of the issue as well as more complex information (see Suedfeld et al., 1992).
As a result, integrative complexity is not the same construct as argument
structure, since integrative complexity cannot be reduced to a formatting
manipulation that holds the underlying information constant, as argument
structure can (also, e.g., Lipe & Salterio, 2002; Sedor, 2002). Additionally,
arguments and integrative complexity may create different effects. For
example, Kadous et al. (2013) predict that auditors become insensitive to
integrative complexity when receiving input from a coworker whom they
respect, know well and like, and have bene?tted from working with in the
past. In contrast, our theory will suggest that managers will become more
sensitive to argument structure when receiving input from a likable internal
auditor.
K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589 577
additional agreement prompted by such an ‘‘argument
heuristic’’ would be likely to occur over and above the
effects of the underlying information, and to be insensitive
to how strongly or weakly the underlying information
supports the internal auditor’s position.
However, whether such an ‘‘argument heuristic’’ would
in?uence managers’ judgments may depend on the
likability of the internal auditor providing the argument.
Psychology research indicates that individuals in positive
affective states are more likely to engage in heuristics than
are people in negative affective states, who tend to be
signi?cantly less heuristic (e.g., Bless et al., 1996;
Bodenhausen, Kramer, & Süsser, 1994; Hertel, Neuhof,
Theuer, & Kerr, 2000; Schwarz, 2002; Sinclair, 1988;
Sinclair & Mark, 1992). While these studies primarily use
mood and emotion to prompt the positive affective states
that lead to heuristic processes, Schwarz (2002) notes that
the likability of another person is likely to prompt similar
positive affective states. The interpersonal likability or dis-
likability of an internal auditor is likely to trigger positive or
negative affective reactions from managers. If so, then the
argument heuristic that we propose would be most likely
to occur among managers in positive affective states. That
is, managers would be likely to agree more with an internal
auditor’s position simply because (s)he is likable and uses
an argument structure for the information (s)he presents,
above and beyond the effects of how supportive or unsup-
portive that information is of the internal auditor’s position.
Our theory suggests that an internal auditor’s likability
is more likely to impact managers’ agreement with his or
her position when the internal auditor uses an argument
structure than when s(he) does not. Whether likability is
unlikely or just less likely to impact managers when the
internal auditor does not use a coherent argument struc-
ture is more dif?cult to predict ex ante. Chaiken (1980)
found that the likability of a communicator trying to
persuade participants only in?uenced their judgments
when other task factors were present that encouraged eas-
ier agreement. Thus, when the internal auditor uses an
argument structure that facilitates agreement, there is a
stronger ex ante basis for predicting that the internal audi-
tor’s likability will in?uence managers’ judgments. On the
other hand, when the internal auditor does not use a
coherent argument, this does not facilitate agreement,
and it may be more dif?cult for an internal auditor’s lik-
ability alone to in?uence managers without being offset
by self-correction processes that individuals can attempt
within social and persuasive settings (see, e.g., Fabrigar,
MacDonald, & Wegener, 2005; Wegener & Petty, 1995;
Wegener, Petty, Smoak, & Fabrigar, 2004). Still, likability
effects may still occur even under conditions that do not
maximize their likelihood (Sinclair & Mark, 1992; Slovic,
Finucane, Peters, & MacGregor, 2002). Thus, an internal
auditor’s likability is more likely to in?uence managers’
judgments when the internal auditor uses a coherent argu-
ment structure. When the internal auditor does not use an
argument structure, likability effects may or may not
occur, but if they do, they are unlikely to be as large as
when an argument structure is used.
By extension, our theoretical development also suggests
that use of an argument is more likely to achieve additional
agreement from managers when the internal auditor is
likable, than when (s)he is not. Whether a signi?cant argu-
ment effect is unlikely or just less likely to occur when peo-
ple are in negative affective states is more dif?cult to
predict ex ante. For example, Bodenhausen et al. (1994)
found across multiple experiments that stereotyping
heuristics only occur among people in positive affective
states, and Hertel et al. (2000) similarly found that social
norm heuristics in free-rider games only occur in positive
affective states. In contrast, Bless et al. (1996) found that
intrusion heuristics were less likely (but still occurred)
among people in negative affective states. Whether nega-
tive affective states simply reduce or impose a boundary
condition on argument effects is unclear. If the appealing
aspects of an argument structure facilitate agreement,
positive affective states may help individuals reach that
agreement more quickly and easily, without engaging in
self-correction processes (e.g., Fabrigar et al., 2005;
Wegener & Petty, 1995; Wegener et al., 2004). Overall,
the strongest ex ante case for making a ?rm directional
prediction for argument effects on agreement with the
internal auditor is when managers are in positive affective
states. Similar effects may or may not occur when the
internal auditor is dislikable, but those effects should be
stronger when the internal auditor is likable.
In summary, an internal auditor’s likability is most
likely to in?uence managers’ judgments when (s)he uses
an argument structure, and his or her use of an argument
structure is most likely to in?uence managers when the
internal auditor is likable. Whether likability and argument
structure will create additional agreement on their own
(i.e., without the other factor present) is less clear based
on prior theory and research. They may, or they may not.
Still, the best case for predicting additional agreement with
the internal auditor (beyond the effects of the internal
auditor’s underlying information) is when the internal
auditor is both likable and uses an argument structure. This
suggests the following hypothesis:
Hypothesis. Managers will agree relatively more with a
likable internal auditor using an argument structure,
and, in comparison, relatively less with a likable
internal auditor not using an argument structure, with
a dislikable internal auditor using an argument struc-
ture, or with a dislikable internal auditor not using an
argument structure.
Interpersonal likability, argument structure, and the
underlying information
Our hypothesis suggests that internal auditors can use
likability and arguments in combination to their advan-
tage, as a relatively straightforward presentation tactic.
While internal auditors can generally control how they
present information and how they present themselves to
managers, they may not always be able to directly control
how supportive the underlying information is of their
recommendation. As such, the information related to the
internal auditor’s recommendation may be relatively sup-
portive or relatively unsupportive of the internal auditor’s
preferred conclusion. Clearly, we would expect managers
578 K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589
to agree more with an internal auditor when the underly-
ing information is more supportive (vs. less supportive) of
the internal auditor’s recommendation.
Still, there are at least two advantages to examining the
effects of the internal auditor’s underlying information in
our setting. First, this enables us to test our hypothesis
under conditions of relatively supportive information and
relatively unsupportive information, and determine
whether our hypothesized joint effect appears to be robust
to internal auditors’ persuasion attempts given either type
of information. For example, the argument heuristic pre-
dicted by our theory suggests additional agreement from
managers without necessarily increased attention paid to
the underlying information. This suggests internal auditors
could potentially use likability and argument structure
jointly to their advantage even when the underlying infor-
mation is relatively unsupportive of the internal auditor’s
position. If so, then internal auditors could potentially
achieve agreement from managers by this fairly simple
joint presentation effect, even when the underlying infor-
mation is relatively unsupportive and managers otherwise
would not agree with the internal auditor. Manipulating
the underlying information allows us to test for this possi-
bility. Second, by manipulating the underlying informa-
tion, we can test for any interactions involving the
strength of the underlying information.
2
Method
Setting and participants
As Duncan (2002) notes, accounting controllers are spe-
cialists in accounting rules, but regularly consult with non-
accounting operational and divisional managers (who are
closest to the economics of the transactions) for their input
into subjective accounting estimates. For instance, Duncan
(2002, pp. 402–404) discusses examples of controllers
seeking input from operations managers, divisional sales
managers, and vice-presidents about the likelihoods of
inventory write-downs, write-offs of doubtful receivables,
and appropriate accruals of estimated operating liabilities.
Our experimental case places participants into the role of
a mid-level manager who provides input to a controller
about whether the value of inventory should be written-
down in the ?nancial statements as obsolete. Accordingly,
we recruited managers, executives, and other professionals
in management training programs to participate in the
study. Our 133 participants averaged 8.5 years of profes-
sional business experience and 4.4 years of managerial
experience. Consistent with their role in our experimental
task, approximately 71 percent of our participants (n = 95)
were experienced mid-level managers (averaging 9.6 years
of professional business experience and 5.3 years of mana-
gerial experience; e.g., productionmanagers, project manag-
ers, operations managers). Another 5 percent (n = 7) were
experienced upper-level managers (averaging 14.8 years of
professional business experience and 11.2 years of manage-
rial experience; e.g., C-suite executives, president), while 23
percent (n = 31) had no managerial experience (averaging
3.7 years of professional business experience; e.g., market-
ing, ?nancial, and operations personnel).
3
Participants ran-
ged from the beginning of their careers to 26 years of
professional business experience and 20 years of managerial
experience. We do not detect any signi?cant effects of years
of managerial experience or years of professional business
experience on any of our ?ndings, and therefore we included
all participants in our sample. Consistent with this,
Anderson, Jennings, Lowe, and Reckers (1997) indicate that
inventory write-down issues can be easily understood by
decision makers with a wide variety of experience.
In this task, participants form a judgment after receiv-
ing input from an internal auditor who believes that the
inventory in their division should be written down (cf.
Mercer, 2004; Moeller, 2009; Norman, Rose, & Suh,
2011). We adapted our materials from prior research
(e.g., Anderson et al., 1997; Bhattacherjee, Moreno, &
Riley, 2012), with input from a former internal auditor
and controller. Our completed instrument was also
reviewed separately by a CPA, CMA, and CISA with
experience as both an internal auditor and a controller,
who commented on the high realism of the experimental
setting and its relevance to practice.
Experimental task
Participants read case information as divisional manag-
ers of ManuTech, Inc., a hypothetical ?rm that designs,
manufactures, and sells electronic products for medical
and other industries. To encourage participants to pay
attention and think about the case, the instrument told
participants that they would document the reasons for
the judgments they would make in this task. The case
provided participants background information about
ManuTech, its medical products inventory, and the
research and development of its competitors toward a
new generation of ManuTech’s best-selling medical device.
The controller told participants in all conditions that she
wanted their input on the likelihood that the value of the
2
In related research, Sedor (2002) manipulated whether or not a ?rm
used narrative ‘‘scenarios’’ to provide its optimistic earnings guidance.
Speci?cally, that study compared narratives, which arranged the optimistic
earnings guidance temporally (along the causal path from past states to
present states to planned future outcomes), to an unorganized presentation
of the same optimistic information. Sedor (2002) suggested that the
temporally causal arrangement of the information would lead analysts to
process the underlying optimistic earnings guidance more easily, see the
causal linkages from past and present states to future outcomes better, and
ultimately form more optimistic future-oriented earnings forecasts. Con-
sistent with this, analysts’ earnings forecasts were higher when ?rms
structured the underlying optimistic information into narrative scenarios
leading to future outcomes. Compared to our setting, narratives arrange
information into temporal, causal order, while arguments do not (Conners,
1981; Smith, 2003). Therefore, arguments (with their simpler, thematically
organized, and less temporally causal presentation) may not lead managers
to process causal linkages within the underlying information more deeply
than other (similarly non-causal) ways of presenting the information
would. Still, by manipulating the internal auditor’s underlying information,
we can observe whether arguments similarly affect sensitivity to the
internal auditor’s information by testing for interactions involving argu-
ment structure and the underlying information.
3
We obtain statistically similar results and reach the same conclusions if
we include only the participants with managerial experience in our sample
(n = 102).
K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589 579
division’s inventory in that device may need to be written
down, and that they would receive related information
from an internal auditor (e.g., Bariff, 2003; KPMG, 2003;
Moeller, 2009). At the end of the task, participants judged
the likelihood that the value of the medical device
inventory would need to be written down (write-down
judgments, our dependent variable). We manipulated three
independent variables (likability, information, and argument
structure) in a 2 Â 2 Â 2 full-factorial between-subjects
design.
Likability
Likability manipulated whether the internal auditor
providing the information about inventory was either
interpersonally likable or dislikable (Appendix A).
While introducing the internal auditor as the source of
information about inventory, the controller described the
internal auditor in the likable (dislikable) condition as easy
to be around, down to earth, nice, and understanding (hard
to be around, arrogant, a jerk, and condescending). The
internal auditor then politely (rudely) interacted with
participants (adapted from Bhattacherjee et al., 2012),
expressing his opinion that the inventory should be writ-
ten down, and indicating that he would provide supporting
reasons for this conclusion in a separate email.
Information and argument structure
Participants then received the internal auditor’s email.
The information in this email was devoid of the likability
manipulation. Instead, it manipulated two variables: infor-
mation and argument structure (Appendix B).
Information manipulated whether the available informa-
tion was more or less supportive of write-down (adapted
from Anderson et al., 1997). Speci?cally, the competition
produced a new product that was technologically superior
but unproven in the marketplace, brought to market in a
rush with inadequate testing, while the company’s product
was technologically inferior but had an established reputa-
tion and alternative sales markets. The information less
(more) supportive of write-down manipulated whether
the competition’s product would be brought to market later
(or sooner) at a similar (or lower) price as the company’s
product, whether the company’s product was selling more
quickly (or more slowly), whether its alternative markets
had better (or worse) sales prospects, whether price
changes will (or will not) be necessary, etc.
4
Argument structure manipulated whether that informa-
tion was structured in a thematically ?owing, coherent
argument or not. In the argument condition, the internal
auditor’s email was organized so that thematically con-
nected pieces of information were joined together into
?owing paragraphs. Speci?cally, the internal auditor began
with an opening paragraph about the inventory and its
competition, and then discussed how well the inventory
is selling, then other technologically superior products on
the market, then the current viability of alternative sales
markets, etc., in a thematically structured argument. In
contrast, the no argument condition provided exactly the
same statements about inventory as in the argument con-
dition, but in an arbitrary order, following the design of
Sedor (2002). To control for any potential, but unantici-
pated, order effects within the no argument condition, par-
ticipants in this condition received one of four arbitrary
orderings of the same information about inventory, as in
Sedor (2002) and similar to Lipe and Salterio (2002). Our
participants’ responses within this condition are not sensi-
tive to which arbitrary ordering they received (all
p’s > 0.18).
5
Thus, our no argument condition presents the
same information, but without the thematically structured
?ow of the argument condition.
Case conclusion and response variables
The internal auditor’s email then closed politely
(rudely) in the likable (dislikable) conditions. We then
asked participants to provide on a scale from 0 to 100 their
beliefs about the likelihood that the value of the medical
device inventory would need to be written down
(write-down judgments; Appendix C).
Participants then completed a separate post-experimen-
tal questionnaire. This questionnaire asked participants to
rate on scales from 1 to 9 the extent to which they believed
the internal auditor to be a reliable (perceived reliability) and
objective (perceived objectivity) source of information, as
well as respond to manipulation check questions.
Results
Manipulation checks
As a likability manipulationcheck, the post-experimental
questionnaire asked participants to rate the likability of the
internal auditor’s attitude on a scale from 1 to 9. In the
dislikable condition, participants’ average response is
4
We manipulated information to achieve systematic variation in how
strongly the information supported write-down. The less supportive
information is not outright unsupportive in a way that would make the
internal auditor’s recommendation seem unreasonable or implausible. For
example, it is not as though the internal auditor was arguing write-down
for a cutting-edge product ?ying off the shelves; in both cases the inventory
faced competition from a new, technologically superior product. We
designed the less supportive information to be weaker than the more
supportive information but still at least somewhat suggestive of inventory
obsolescence so that the internal auditor’s preference for recognizing
obsolescence would remain plausible. The levels of actual inventory write-
down judgments in our results suggest that we achieved substantial
variation in information strength without switching to implausibility. Mean
judgments indicate at least some risk of write-down in every condition.
5
We adapt other design features that have been used in prior research
investigating the effects of an information structure on users’ judgments
(e.g., Lipe & Salterio, 2002; Sedor, 2002). Besides the four arbitrary
orderings in the no argument condition, the argument condition began
with the phrase ‘‘Let me start by saying that’’ immediately before the ?rst
statement about inventory, which was not present in the no argument
condition, in order to maintain a transitional ?ow into the introduction of
the argument that would be absent in the no argument condition (Graesser,
Millis, & Zwaan, 1997; Sedor, 2002). However, each individual statement
about inventory was present in both the argument and no argument
conditions (as in Sedor, 2002 Experiment 2), to hold the underlying
information about inventory constant.
580 K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589
signi?cantly below the midpoint 5 (2.1, t
118
= À10.72,
p
one-tail
< 0.001).
6
In the likable condition, participants’
average response is signi?cantly above the midpoint
(6.1, t
118
= 4.45, p
one-tail
< 0.001), and signi?cantly above the
dislikable group (6.1 vs. 2.1; t
118
= 10.86, p
one-tail
< 0.001).
7
As an information manipulation check, the post-experimental
questionnaire presented participants with the data fromboth
treatment levels of the information manipulation, and asked
them to identify which data set is more strongly supportive
of inventory write-down. Participants identi?ed the informa-
tion from the more (less) supportive condition as the more
(less) supportive information set 93 percent of the time,
signi?cantly better than chance (v
2
1
= 99.4, p < 0.001).
8
In addition, we conducted a follow-up study that
gathered additional manipulation checks and other post-
experimental measures as its dependent variables, using
a sample of 70 similar professionals from similar manage-
ment training programs. Participants in this follow-up
study averaged 11.0 years of professional business experi-
ence and 6.4 years of managerial experience, and had up to
30 and 29.9 years, respectively. Consistent with our main
experiment, most of these participants (84 percent,
n = 59) were middle-level managers, 7 percent (n = 5) were
upper-level executives, 9 percent (n = 6) had no managerial
experience, and experience does not affect our ?ndings in
this follow-up study. We manipulated argument structure
and likability exactly as in our main experiment, and
(because we do not ?nd any statistically signi?cant inter-
actions involving information in our main experiment),
we held information constant at the less supportive level.
For our argument manipulation checks, the theoretically
important aspects of arguments are (1) that they provide
a thematically ?owing structure to the information, and
(2) that people naturally respond favorably to this the-
matic structure. Accordingly, we asked participants in the
follow-up study to rate on 9-point Likert-type scales (1)
how structured they perceived the internal auditor’s
information about inventory to be, as well as (2) how per-
suasive they perceived the information to be. Participants
perceived the (otherwise identical) information to be more
structured (means (standard deviations) = 5.7 (2.2) vs. 3.6
(2.5), t
66
= 3.76, p
one-tail
< 0.001) and more persuasive
(means (standard deviations) = 5.6 (1.8) vs. 4.0 (2.2),
t
66
= 3.23, p
one-tail
= 0.001) in the argument conditions than
in the no argument conditions. In fact, in results not tabu-
lated in this paper, we ?nd that perceived structure fully
mediates the effect of arguments on perceived persuasive-
ness, consistent with our theory that it is the thematically
organized structure of information in an argument to which
individuals respond favorably.
9
Overall, participants’
responses to all of our manipulation checks across both
experiments vary as expected.
10
Hypothesis tests
Table 1 shows descriptive statistics for our dependent
variable, write-down judgments, by experimental condition,
and Table 2 (Panel A) shows an analysis of variance.
11
As
Table 2 shows, we ?nd a signi?cant main effect of
information. Speci?cally, across experimental conditions,
participants’ write-down judgments are, on average, 23.2
percentage points higher in the more supportive information
conditions than in the less supportive information conditions
(69.3 vs. 46.1, t
125
= 6.20, p
one-tail
< 0.001). Thus, overall,
managers reacted to the underlying information presented
by the internal auditor as expected.
The ANOVA table also shows a signi?cant argument
structure  likability interaction (F
1,125
= 11.17, p = 0.001;
Table 2). Fig. 1 shows the least-squares means of this
interaction from the ANOVA model.
12
This interaction from
6
We use one-tailed p-values for t-tests of manipulation checks, the
information main effect, and our hypothesis where we make directional
predictions, and denote them with a subscript, p
one-tail
. All other p-values
are two-tailed.
7
Given the variation in interpersonal likability that likely exists in
practice (Pickett, 2007), we wanted to achieve dislikable conditions that
were both strong (Kerlinger & Lee, 2000) and realistic. While the 2.1 mean
rating in the dislikable group was signi?cantly below the midpoint of the
scale, it was also signi?cantly above the lower endpoint of the scale
(F
1,118
= 16.8, p < 0.001). Cohen’s d (a measure of practical signi?cance) also
suggests that this rating is well above the lower endpoint of the scale
(d = 0.82).
8
Our objective with the information manipulation check is to provide
construct validation that the information we placed into the more
supportive information condition was in fact more supportive of inventory
obsolescence than the information we placed into the less supportive
information condition. Rather than the within-subjects manipulation check
we used, we could have tested this between-subjects by simply asking
participants how much the information from the internal auditor’s email
was supportive of inventory obsolescence. However, we did not use this
approach because, as Pedhazur and Schmelkin (1991, p. 263) suggest,
simply asking participants to rate how supportive the internal auditor’s
information was of his position might not accomplish our objective if
participants interpreted the question in light of how they responded to our
dependent variable measuring whether they agreed with the internal
auditor. Unlike our dependent variable write-down judgments, participants’
responses to our within-subjects manipulation check was not signi?cantly
in?uenced by our other manipulated factors (i.e., argument structure and
likability), whereas a between-subjects manipulation check might have
been. Reffett (2010) uses a within-subjects manipulation check for similar
reasons. The results of our manipulation check accomplishes its purpose of
providing empirical evidence that our more supportive information was
more suggestive of inventory obsolescence than our less supportive
information.
9
Additionally, while our main experiment asked participants to rate how
much they liked the individual auditor’s attitude, we asked participants in
this follow-up study to make the same rating of his attitude, plus ratings of
his interpersonal skills and of his personal likability. Findings for all three of
these variables are similar to those in our main experiment (all p
one-
tail
’s < 0.001).
10
Our manipulation checks only vary according to main effects of their
matching manipulated variables. For each, we do not ?nd any signi?cant
interactions or main effects involving the other manipulated variables. We
obtain statistically similar results when dropping observations that failed
our manipulation checks.
11
Tests of ANOVA model assumptions are well below levels that elevate
type I error rates (i.e., below levels that would suggest an increased risk of
?nding spurious effects; see Neter, Kutner, Nachtsheim, & Wasserman,
1996; Tabachnick & Fidell, 2001).
12
The least-squares means in Fig. 1 show participants’ responses
collapsed across information (Neter et al., 1996; Searle, Speed, & Milliken,
1980). Because information does not interact with argument structure or
likability in any two-way interactions or in a three-way interaction (Table 2,
Panel A; see also Table 3 Panel A), results within each level of the
information conditions are statistically similar to those in Fig. 1, with the
only signi?cant exception being that the means within the more supportive
information conditions are, on average, 23.2 percentage points higher than
those in the less supportive information conditions.
K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589 581
the ANOVA table indicates that an argument structure had a
greater impact on managers’ judgments when the internal
auditor was likable than when he was dislikable, consistent
with our theory.
Our hypothesis predicts an ordinal interaction such that
managers will agree more with the internal auditor when
he is both likable and uses an argument structure, and
relatively less under other conditions. To test this predic-
tion, we examine whether the observed means match the
pattern predicted by our hypothesis (Buckless &
Ravenscroft, 1990). First, as Fig. 1 shows, the means occur
in a pattern consistent with our expectations (see Keppel &
Wichens, 2004; e.g., Kadous et al., 2013). Second, we use
contrast weights of +3 for the likable/argument structure
condition, and À1 for the remaining conditions shown in
Fig. 1, and ?nd statistically signi?cant results (t
125
= 4.81,
p
one-tail
< 0.001; Table 2, Panel B). Moreover, the three con-
ditions assigned the weight of À1 do not differ signi?cantly
from each other (F
2,125
= 0.44, p = 0.61; Buckless &
Ravenscroft, 1990). These results support our hypothesis.
In addition, the means within the more supportive and less
supportive information conditions (see Table 1) also occur
in the expected pattern. Speci?cally, tests within each level
of information using contrast weights of +3, À1, À1, À1 pro-
vide statistically similar results (p
one-tail
’s 6 0.002; Table 2,
Panel B). Moreover, the three conditions within each level
of information assigned weights of À1 similarly do not
differ signi?cantly from one other, as well (p’s P0.34).
13
Supplementary analyses: Simple effects tests
Tests of the simple effects indicate the following results.
When the internal auditor is likable, the use of an
Table 2
Write-down judgments, analysis of variance, and hypothesis tests.
Panel A: Analysis of variance
Source Sum of squares df Mean square F p
Information 17466.6 1 17466.6 38.49
Internal auditors play an important role in influencing managers’ judgments. Yet, the
practitioner literature indicates that, because internal audit lacks the client services incen-
tives of external audit, internal auditors often adopt a ‘‘policeman approach’’ that can lead
to negative interpersonal relationships with managers. We investigate three variables fun-
damental to internal auditors’ ability to influence managers: (1) internal auditors’ interper-
sonal likability, (2) the information used to support their positions, and (3) whether they
present that information in a thematically organized argument. We find that managers
agree more with an internal auditor who is both likable and uses a thematically organized
argument. We find further that this joint effect occurs regardless of whether the internal
auditor’s information is relatively supportive or unsupportive of his position. Overall, our
theory and findings suggest that an internal auditor can achieve agreement from managers
on important corporate governance issues with this fairly straightforward presentation
tactic, even when the underlying information is relatively unsupportive and managers
otherwise tend not to agree with the internal auditor’s position
Internal auditors’ use of interpersonal likability, arguments, and
accounting information in a corporate governance setting
q
Kirsten Fanning
a
, M. David Piercey
b,?
a
College of Business, University of Illinois at Urbana-Champaign, Champaign, IL 61820, United States
b
Isenberg School of Management, University of Massachusetts, Amherst, MA 01003, United States
a b s t r a c t
Internal auditors play an important role in in?uencing managers’ judgments. Yet, the
practitioner literature indicates that, because internal audit lacks the client services incen-
tives of external audit, internal auditors often adopt a ‘‘policeman approach’’ that can lead
to negative interpersonal relationships with managers. We investigate three variables fun-
damental to internal auditors’ ability to in?uence managers: (1) internal auditors’ interper-
sonal likability, (2) the information used to support their positions, and (3) whether they
present that information in a thematically organized argument. We ?nd that managers
agree more with an internal auditor who is both likable and uses a thematically organized
argument. We ?nd further that this joint effect occurs regardless of whether the internal
auditor’s information is relatively supportive or unsupportive of his position. Overall, our
theory and ?ndings suggest that an internal auditor can achieve agreement from managers
on important corporate governance issues with this fairly straightforward presentation
tactic, even when the underlying information is relatively unsupportive and managers
otherwise tend not to agree with the internal auditor’s position. Our study contributes to
accounting, psychology, and writing and discourse theories with new evidence of the
effects of an argument structure (holding the underlying information constant) on users’
judgments, and how those effects depend on the likability of the source of information.
Our ?ndings have important implications for internal auditors, managers, external
auditors, and others interested in corporate governance.
Ó 2014 Elsevier Ltd. All rights reserved.
Introduction
In this study, we examine how three variables, each
fundamental to internal auditors’ interactions with manag-
ers, explain internal auditors’ in?uence on managers’ judg-
ments: (1) internal auditors’ interpersonal likability, (2)
the underlying information supporting their positions,
and (3) their use of thematically organized arguments to
present that information to managers. As Prawitt, Smith,
and Wood (2009, p. 1258) point out, internal auditors tend
to interact with managers frequently, and are ‘‘often the
party primarily responsible for the day-to-day monitoring
of management’s actions, including those related to
external ?nancial reporting’’ (see also Bariff, 2003; Cohen,http://dx.doi.org/10.1016/j.aos.2014.07.002
0361-3682/Ó 2014 Elsevier Ltd. All rights reserved.
q
We are grateful for the comments and assistance of Chris Agoglia, Jim
Bierstaker, Joe Brazel, Tom Clausen, Joshua Herbold, Jessen Hobson, Carol
Jessup, Eric Johnson, Kathryn Kadous, Tom Kida, Benjamin Luippold,
Mario Maletta, Mike Peters, Kimberly Moreno, Robert Resutek, Chad
Simon, Jim Smith, Dan Stone, Hun Tong Tan, David Wood, Arnie Wright,
May Zhang, and the editors and two anonymous reviewers. We also thank
workshop participants at the University of Massachusetts, Northeastern
University, and Villanova University, and conference participants at the
American Accounting Association (AAA) Annual Meeting, the AAA
Accounting, Behavior, & Organizations Research Conference, and Brigham
Young University Accounting Research Symposium for their helpful
comments. We thank Sudip Bhattacharjee, Kimberly Moreno, and Tracey
Riley for access to their experimental materials.
?
Corresponding author. Tel.: +1 217 300 1981/413 545 5585.
E-mail addresses: [email protected] (K. Fanning), piercey@isenberg.
umass.edu (M. David Piercey).
Accounting, Organizations and Society 39 (2014) 575–589
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
Hayes, Krishnamoorthy, Monroe, & Wright, 2013; Mercer,
2004).
The practitioner literature indicates that the tone of the
interactions between internal auditors and managers
varies widely in practice (Deloitte, 2010; Dittenhofer,
Ramamoorti, Ziegenfuss, & Evans, 2010; Pickett, 2010;
Ratliff & Brackner, 1998). Clearly, many internal auditors
have exceptional interpersonal interactions with manag-
ers. However, Pickett (2007, 2010) notes that internal audit
lacks the client services incentives of external audit, allow-
ing internal auditors to adopt a ‘‘policeman approach,’’
which places little emphasis on positive interpersonal
interactions with managers as clients, compared to exter-
nal audit. Deloitte (2010) similarly notes that the ‘‘police’’
approach to internal audit can harm the manager–internal
auditor relationship, and contend that ‘‘a dysfunctional
relationship [between managers and internal auditors] is
a contributing cause, and in some cases, a primary cause’’
of a variety of accounting problems, including ‘‘material
weaknesses, ?nancial restatement, regulatory compliance,
and the like’’ (p. 3). Despite the importance of the interper-
sonal relationship between internal auditors and manag-
ers, Archambeault, DeZoort, and Holt (2008), and Prawitt
et al. (2009) point out that it has received relatively little
research.
We develop theory and experimentally test how inter-
nal auditors can use arguments, personal likability, and
information to in?uence managers’ judgments. According
to theories of writing and discourse, an argument is a
?owing arrangement of information into thematically con-
nected groups in support of a particular conclusion (e.g.,
Conners, 1981; Smith, 2003). However, individuals often
simply provide information in support of a position as it
comes to mind, without organizing it into a structured
argument, leaving it to the user to decide how the pieces
?t together (Booher, 2001; Guffey, 2010; May & May,
2012). The internal audit practitioner literature indicates
a wide variance in the effectiveness of internal auditors’
use of arguments in practice (Chambers, 2009;
Dittenhofer et al., 2010). For example, Chambers (2009)
interviewed managers and executives who indicate that
internal auditors often do not provide information in a
manner that allows users to easily see how related pieces
of information connect together. In other words, internal
auditors provide information, but without always structur-
ing it into a coherent, thematically organized, ?owing
argument (cf. May & May, 2012).
Writing and discourse scholars distinguish arguments
(which organize information supporting a position into
thematically connected groups) from at least three other
forms of rhetoric: narration (connecting information into
temporal order), description, and exposition (e.g.,
Conners, 1981; Smith, 2003). As an example, an internal
auditor using an argument to recommend a write-down
of obsolete inventory (cf. KPMG, 2003) could structure
information into thematically connected groups by, for
example, ?rst introducing the state of the inventory and
its competition, then discussing information about how
slowly the inventory is selling, followed by details about
the technologically superior products, then discussing the
viability of any other sales prospects for the older
inventory, etc. Prior research has shown how other ways
of structuring information (holding the underlying
information constant) in?uences how users react to that
information (e.g., Lipe & Salterio, 2002; Ricchiute, 1992;
Sedor, 2002). For example, Sedor (2002) manipulated
whether analysts received optimistic earnings guidance
in temporal, causal narratives that linked past states of
the ?rm to current states to plans for the future, or the
same information in randomized order. Earnings forecasts
were more optimistic when a temporally causal narrative
format was used.
In our experiment, managers provide a controller with
their input into an inventory write-down judgment for
their divisions (e.g., Duncan, 2002), while interacting with
an internal auditor who prefers conservatively writing-
down the value of the inventory in the ?nancial statements
(e.g., KPMG, 2003; Mercer, 2004; Moeller, 2009; Prawitt
et al., 2009). We examine this setting within a 2 Â 2 Â 2
experimental framework, in which the internal auditor is
either interpersonally likable or dislikable, and presents
information that is either more supportive or less support-
ive of write-down, in either a coherent, thematically ?ow-
ing argument structure or not. Similar to Sedor’s (2002)
manipulation of temporal narratives, our manipulation of
arguments holds the underlying information about inven-
tory constant by comparing the argument condition to a
condition in which the internal auditor presents the same
statements about inventory in an unorganized order.
We combine theories from writing and discourse,
psychology, and accounting to build our predictions.
Because people ?nd the thematically structured ?ow of
an argument appealing, and because positive affective
states lead to heuristic processing, we predict that manag-
ers will heuristically agree more with an internal auditor
who is both likable and uses an argument structure,
beyond the effects of how supportive or unsupportive the
internal auditor’s information is of his position. Our ?nd-
ings are consistent with this hypothesis. First, our most
basic ?nding is that managers (unsurprisingly) agree more
with an auditor who uses more supportive information
than one who uses less supportive information. However,
beyond that, they also agree more with an internal auditor
who is both likable and uses a thematically organized argu-
ment structure, regardless of whether the information pre-
sented is relatively supportive or unsupportive of the
internal auditor’s position. In fact, our results demonstrate
that an internal auditor can achieve (on average) agree-
ment from managers simply because he is likable and uses
a ?owing argument structure, even when the underlying
information is relatively unsupportive and managers
otherwise (on average) do not to agree with the internal
auditor. Overall, our theory and ?ndings suggest that
internal auditors can achieve additional agreement from
managers on important corporate governance issues,
above and beyond how supportive or unsupportive their
information is, by using an argument structure and likabil-
ity jointly, as a fairly straightforward presentation tactic.
This is the ?rst study of which we are aware to demon-
strate how structuring information into thematically
organized arguments (holding constant the underlying
information presented) interacts with an information
576 K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589
provider’s likability to prompt additional agreement from
users that would not otherwise occur. As a result, we
contribute to both the research literature on interpersonal
likability (e.g., Kida, Moreno, & Smith, 2001; Moreno, Kida,
& Smith, 2002), as well as the literatures on persuasion and
other related theories (e.g., Kadous, Leiby, & Peecher, 2013;
Pennington & Hastie, 1986, 1988, 1993; Rich, Solomon, &
Trotman, 1997; Sedor, 2002; Suedfeld, Tetlock, &
Streufert, 1992). In addition, our study also has other
important implications for accounting theory, research
and practice. Accounting is not just about the mere aggre-
gation of data, but how accountants present that data to
other social beings in a social–organizational setting
(Bonner, 2008; Kachelmeier, 2010; Kinney, 2001). While
prior research has shown that internal auditors have at
least some impact on managers’ judgments (Prawitt
et al., 2009), our study can help develop a theory about
day-to-day behavioral factors that drive variance in inter-
nal auditors’ in?uence over managers’ judgments, as called
for by Prawitt et al. (2009) and Archambeault et al. (2008).
Internal auditing is a ‘‘relationship and communications
business’’ in which information is communicated within
an organizational context to in?uence managers
(Dittenhofer et al., 2010). As Kachelmeier (2010) puts it,
?rms do not make accounting decisions, people make
accounting decisions, and those decisions are shaped by
the behavioral interactions of individuals within the social
environment of their ?rms. Our study can help form an
empirical basis for the soft skills education and training
called for in the internal audit practitioner literature (e.g.,
Dittenhofer et al., 2010; IIA, 2013a, 2013b; Moeller, 2009;
Pickett, 2010). In addition, our ?ndings should also be
informative to external auditors, managers, researchers,
and others interested in in?uencing managers’ judgments
and corporate governance.
Theory
Interpersonal likability
As noted in the introduction, the practitioner litera-
ture indicates that the tone of the interactions between
internal auditors and managers varies widely in practice,
with some internal auditors occasionally adopting a
‘‘police’’ approach that can lead to dysfunctional
manager–internal audit relationships (e.g., Deloitte.,
2010; Pickett, 2007, 2010). Prior research suggests that
managers have affective reactions to others’ interper-
sonal likability as they interact with them within organi-
zational settings (Kida et al., 2001; Moreno et al., 2002).
Individuals are more likely to incorporate their affective
reactions into subsequent judgments when other task
factors prompt heuristic processing (e.g., Chaiken, 1980;
Finucane, Alhakami, Slovic, & Johnson, 2000; Schwarz,
Strack, Kommer, & Wagner, 1987; Shiv & Fedorikhin,
1999; Siemer & Reisenzein, 1998). This research suggests
that, for our setting, when managers do incorporate their
affective reactions into their subsequent judgments, they
would tend to respond positively (negatively) to a likable
(dislikable) internal auditor.
Argument structure and likability
An argument is a form of rhetoric which structures
underlying information into thematically related groups
of connected ideas that ?ow together to support a particu-
lar position (e.g., Conners, 1981; Smith, 2003). For exam-
ple, an individual trying to persuade others to support a
particular political candidate could organize supportive
information into thematically connected groups by, e.g.,
?rst discussing information about the high moral character
of the politician, then various advantages of the politician’s
social policy, followed by various advantages of the politi-
cian’s ?scal policy, then information about the politician’s
experience and history in past public of?ces, etc., in a
coherent argument structure. Alternatively, individuals
can convey the same information supporting a particular
position, but without structuring it into an argument
(May & May, 2012).
1
An argument structure provides an apparent, appealing
thematic ?owto the information. Kida (2006) suggests that
individuals naturally prefer verbally ?owing accounts to
raw presentations of data. We propose that an internal
auditor’s use of an argument structure (holding constant
the underlying information) can create additional agree-
ment from managers, without increased attention to how
supportive (or unsupportive) the presented information
actually is of the internal auditor’s position. Speci?cally,
this additional agreement is likely to be the result of a
heuristic whereby people simply respond favorably to the
apparent thematic ?ow that an argument structure pro-
vides. The strength of the underlying information would
still have a predictable directional effect on managers’
agreement with the internal auditor’s position. But the
1
Prior accounting research has examined constructs that share similar-
ities with arguments, yet also have important differences as well.
Compared to arguments (which arrange thematically related pieces of
information together in support of a position), the ‘‘scenarios’’ construct in
Sedor (2002) matches what writing and discourse scholars refer to as
‘‘narratives,’’ a different form of rhetoric from ‘‘arguments’’ that arranges
information temporally (e.g., Conners, 1981; Smith, 2003). In fact, Sedor
(2002) describes the ‘‘scenarios’’ construct as ‘‘narratives that concretely
describe the sequence of events in which proposed actions lead to future
outcomes’’ (p. 734), and the theory relies on narratives laying out the causal
links in the temporal sequence from past states to current states to future
outcomes in order to in?uence forecasts of the future. In addition, similar to
our setting of internal auditors providing input (with or without an
argument structure) to managers, Kadous et al. (2013) examine the effects
of an auditor providing input (with either high or low integrative
complexity) to other auditors. Low integrative complexity involves simple,
one-sided justi?cations of advice that do not go into much depth and are
not self-critical in the sense that they do not consider tradeoffs on both
sides of the issue. High integrative complexity involves deeper justi?ca-
tions that consider more information, including information on both sides
of the issue as well as more complex information (see Suedfeld et al., 1992).
As a result, integrative complexity is not the same construct as argument
structure, since integrative complexity cannot be reduced to a formatting
manipulation that holds the underlying information constant, as argument
structure can (also, e.g., Lipe & Salterio, 2002; Sedor, 2002). Additionally,
arguments and integrative complexity may create different effects. For
example, Kadous et al. (2013) predict that auditors become insensitive to
integrative complexity when receiving input from a coworker whom they
respect, know well and like, and have bene?tted from working with in the
past. In contrast, our theory will suggest that managers will become more
sensitive to argument structure when receiving input from a likable internal
auditor.
K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589 577
additional agreement prompted by such an ‘‘argument
heuristic’’ would be likely to occur over and above the
effects of the underlying information, and to be insensitive
to how strongly or weakly the underlying information
supports the internal auditor’s position.
However, whether such an ‘‘argument heuristic’’ would
in?uence managers’ judgments may depend on the
likability of the internal auditor providing the argument.
Psychology research indicates that individuals in positive
affective states are more likely to engage in heuristics than
are people in negative affective states, who tend to be
signi?cantly less heuristic (e.g., Bless et al., 1996;
Bodenhausen, Kramer, & Süsser, 1994; Hertel, Neuhof,
Theuer, & Kerr, 2000; Schwarz, 2002; Sinclair, 1988;
Sinclair & Mark, 1992). While these studies primarily use
mood and emotion to prompt the positive affective states
that lead to heuristic processes, Schwarz (2002) notes that
the likability of another person is likely to prompt similar
positive affective states. The interpersonal likability or dis-
likability of an internal auditor is likely to trigger positive or
negative affective reactions from managers. If so, then the
argument heuristic that we propose would be most likely
to occur among managers in positive affective states. That
is, managers would be likely to agree more with an internal
auditor’s position simply because (s)he is likable and uses
an argument structure for the information (s)he presents,
above and beyond the effects of how supportive or unsup-
portive that information is of the internal auditor’s position.
Our theory suggests that an internal auditor’s likability
is more likely to impact managers’ agreement with his or
her position when the internal auditor uses an argument
structure than when s(he) does not. Whether likability is
unlikely or just less likely to impact managers when the
internal auditor does not use a coherent argument struc-
ture is more dif?cult to predict ex ante. Chaiken (1980)
found that the likability of a communicator trying to
persuade participants only in?uenced their judgments
when other task factors were present that encouraged eas-
ier agreement. Thus, when the internal auditor uses an
argument structure that facilitates agreement, there is a
stronger ex ante basis for predicting that the internal audi-
tor’s likability will in?uence managers’ judgments. On the
other hand, when the internal auditor does not use a
coherent argument, this does not facilitate agreement,
and it may be more dif?cult for an internal auditor’s lik-
ability alone to in?uence managers without being offset
by self-correction processes that individuals can attempt
within social and persuasive settings (see, e.g., Fabrigar,
MacDonald, & Wegener, 2005; Wegener & Petty, 1995;
Wegener, Petty, Smoak, & Fabrigar, 2004). Still, likability
effects may still occur even under conditions that do not
maximize their likelihood (Sinclair & Mark, 1992; Slovic,
Finucane, Peters, & MacGregor, 2002). Thus, an internal
auditor’s likability is more likely to in?uence managers’
judgments when the internal auditor uses a coherent argu-
ment structure. When the internal auditor does not use an
argument structure, likability effects may or may not
occur, but if they do, they are unlikely to be as large as
when an argument structure is used.
By extension, our theoretical development also suggests
that use of an argument is more likely to achieve additional
agreement from managers when the internal auditor is
likable, than when (s)he is not. Whether a signi?cant argu-
ment effect is unlikely or just less likely to occur when peo-
ple are in negative affective states is more dif?cult to
predict ex ante. For example, Bodenhausen et al. (1994)
found across multiple experiments that stereotyping
heuristics only occur among people in positive affective
states, and Hertel et al. (2000) similarly found that social
norm heuristics in free-rider games only occur in positive
affective states. In contrast, Bless et al. (1996) found that
intrusion heuristics were less likely (but still occurred)
among people in negative affective states. Whether nega-
tive affective states simply reduce or impose a boundary
condition on argument effects is unclear. If the appealing
aspects of an argument structure facilitate agreement,
positive affective states may help individuals reach that
agreement more quickly and easily, without engaging in
self-correction processes (e.g., Fabrigar et al., 2005;
Wegener & Petty, 1995; Wegener et al., 2004). Overall,
the strongest ex ante case for making a ?rm directional
prediction for argument effects on agreement with the
internal auditor is when managers are in positive affective
states. Similar effects may or may not occur when the
internal auditor is dislikable, but those effects should be
stronger when the internal auditor is likable.
In summary, an internal auditor’s likability is most
likely to in?uence managers’ judgments when (s)he uses
an argument structure, and his or her use of an argument
structure is most likely to in?uence managers when the
internal auditor is likable. Whether likability and argument
structure will create additional agreement on their own
(i.e., without the other factor present) is less clear based
on prior theory and research. They may, or they may not.
Still, the best case for predicting additional agreement with
the internal auditor (beyond the effects of the internal
auditor’s underlying information) is when the internal
auditor is both likable and uses an argument structure. This
suggests the following hypothesis:
Hypothesis. Managers will agree relatively more with a
likable internal auditor using an argument structure,
and, in comparison, relatively less with a likable
internal auditor not using an argument structure, with
a dislikable internal auditor using an argument struc-
ture, or with a dislikable internal auditor not using an
argument structure.
Interpersonal likability, argument structure, and the
underlying information
Our hypothesis suggests that internal auditors can use
likability and arguments in combination to their advan-
tage, as a relatively straightforward presentation tactic.
While internal auditors can generally control how they
present information and how they present themselves to
managers, they may not always be able to directly control
how supportive the underlying information is of their
recommendation. As such, the information related to the
internal auditor’s recommendation may be relatively sup-
portive or relatively unsupportive of the internal auditor’s
preferred conclusion. Clearly, we would expect managers
578 K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589
to agree more with an internal auditor when the underly-
ing information is more supportive (vs. less supportive) of
the internal auditor’s recommendation.
Still, there are at least two advantages to examining the
effects of the internal auditor’s underlying information in
our setting. First, this enables us to test our hypothesis
under conditions of relatively supportive information and
relatively unsupportive information, and determine
whether our hypothesized joint effect appears to be robust
to internal auditors’ persuasion attempts given either type
of information. For example, the argument heuristic pre-
dicted by our theory suggests additional agreement from
managers without necessarily increased attention paid to
the underlying information. This suggests internal auditors
could potentially use likability and argument structure
jointly to their advantage even when the underlying infor-
mation is relatively unsupportive of the internal auditor’s
position. If so, then internal auditors could potentially
achieve agreement from managers by this fairly simple
joint presentation effect, even when the underlying infor-
mation is relatively unsupportive and managers otherwise
would not agree with the internal auditor. Manipulating
the underlying information allows us to test for this possi-
bility. Second, by manipulating the underlying informa-
tion, we can test for any interactions involving the
strength of the underlying information.
2
Method
Setting and participants
As Duncan (2002) notes, accounting controllers are spe-
cialists in accounting rules, but regularly consult with non-
accounting operational and divisional managers (who are
closest to the economics of the transactions) for their input
into subjective accounting estimates. For instance, Duncan
(2002, pp. 402–404) discusses examples of controllers
seeking input from operations managers, divisional sales
managers, and vice-presidents about the likelihoods of
inventory write-downs, write-offs of doubtful receivables,
and appropriate accruals of estimated operating liabilities.
Our experimental case places participants into the role of
a mid-level manager who provides input to a controller
about whether the value of inventory should be written-
down in the ?nancial statements as obsolete. Accordingly,
we recruited managers, executives, and other professionals
in management training programs to participate in the
study. Our 133 participants averaged 8.5 years of profes-
sional business experience and 4.4 years of managerial
experience. Consistent with their role in our experimental
task, approximately 71 percent of our participants (n = 95)
were experienced mid-level managers (averaging 9.6 years
of professional business experience and 5.3 years of mana-
gerial experience; e.g., productionmanagers, project manag-
ers, operations managers). Another 5 percent (n = 7) were
experienced upper-level managers (averaging 14.8 years of
professional business experience and 11.2 years of manage-
rial experience; e.g., C-suite executives, president), while 23
percent (n = 31) had no managerial experience (averaging
3.7 years of professional business experience; e.g., market-
ing, ?nancial, and operations personnel).
3
Participants ran-
ged from the beginning of their careers to 26 years of
professional business experience and 20 years of managerial
experience. We do not detect any signi?cant effects of years
of managerial experience or years of professional business
experience on any of our ?ndings, and therefore we included
all participants in our sample. Consistent with this,
Anderson, Jennings, Lowe, and Reckers (1997) indicate that
inventory write-down issues can be easily understood by
decision makers with a wide variety of experience.
In this task, participants form a judgment after receiv-
ing input from an internal auditor who believes that the
inventory in their division should be written down (cf.
Mercer, 2004; Moeller, 2009; Norman, Rose, & Suh,
2011). We adapted our materials from prior research
(e.g., Anderson et al., 1997; Bhattacherjee, Moreno, &
Riley, 2012), with input from a former internal auditor
and controller. Our completed instrument was also
reviewed separately by a CPA, CMA, and CISA with
experience as both an internal auditor and a controller,
who commented on the high realism of the experimental
setting and its relevance to practice.
Experimental task
Participants read case information as divisional manag-
ers of ManuTech, Inc., a hypothetical ?rm that designs,
manufactures, and sells electronic products for medical
and other industries. To encourage participants to pay
attention and think about the case, the instrument told
participants that they would document the reasons for
the judgments they would make in this task. The case
provided participants background information about
ManuTech, its medical products inventory, and the
research and development of its competitors toward a
new generation of ManuTech’s best-selling medical device.
The controller told participants in all conditions that she
wanted their input on the likelihood that the value of the
2
In related research, Sedor (2002) manipulated whether or not a ?rm
used narrative ‘‘scenarios’’ to provide its optimistic earnings guidance.
Speci?cally, that study compared narratives, which arranged the optimistic
earnings guidance temporally (along the causal path from past states to
present states to planned future outcomes), to an unorganized presentation
of the same optimistic information. Sedor (2002) suggested that the
temporally causal arrangement of the information would lead analysts to
process the underlying optimistic earnings guidance more easily, see the
causal linkages from past and present states to future outcomes better, and
ultimately form more optimistic future-oriented earnings forecasts. Con-
sistent with this, analysts’ earnings forecasts were higher when ?rms
structured the underlying optimistic information into narrative scenarios
leading to future outcomes. Compared to our setting, narratives arrange
information into temporal, causal order, while arguments do not (Conners,
1981; Smith, 2003). Therefore, arguments (with their simpler, thematically
organized, and less temporally causal presentation) may not lead managers
to process causal linkages within the underlying information more deeply
than other (similarly non-causal) ways of presenting the information
would. Still, by manipulating the internal auditor’s underlying information,
we can observe whether arguments similarly affect sensitivity to the
internal auditor’s information by testing for interactions involving argu-
ment structure and the underlying information.
3
We obtain statistically similar results and reach the same conclusions if
we include only the participants with managerial experience in our sample
(n = 102).
K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589 579
division’s inventory in that device may need to be written
down, and that they would receive related information
from an internal auditor (e.g., Bariff, 2003; KPMG, 2003;
Moeller, 2009). At the end of the task, participants judged
the likelihood that the value of the medical device
inventory would need to be written down (write-down
judgments, our dependent variable). We manipulated three
independent variables (likability, information, and argument
structure) in a 2 Â 2 Â 2 full-factorial between-subjects
design.
Likability
Likability manipulated whether the internal auditor
providing the information about inventory was either
interpersonally likable or dislikable (Appendix A).
While introducing the internal auditor as the source of
information about inventory, the controller described the
internal auditor in the likable (dislikable) condition as easy
to be around, down to earth, nice, and understanding (hard
to be around, arrogant, a jerk, and condescending). The
internal auditor then politely (rudely) interacted with
participants (adapted from Bhattacherjee et al., 2012),
expressing his opinion that the inventory should be writ-
ten down, and indicating that he would provide supporting
reasons for this conclusion in a separate email.
Information and argument structure
Participants then received the internal auditor’s email.
The information in this email was devoid of the likability
manipulation. Instead, it manipulated two variables: infor-
mation and argument structure (Appendix B).
Information manipulated whether the available informa-
tion was more or less supportive of write-down (adapted
from Anderson et al., 1997). Speci?cally, the competition
produced a new product that was technologically superior
but unproven in the marketplace, brought to market in a
rush with inadequate testing, while the company’s product
was technologically inferior but had an established reputa-
tion and alternative sales markets. The information less
(more) supportive of write-down manipulated whether
the competition’s product would be brought to market later
(or sooner) at a similar (or lower) price as the company’s
product, whether the company’s product was selling more
quickly (or more slowly), whether its alternative markets
had better (or worse) sales prospects, whether price
changes will (or will not) be necessary, etc.
4
Argument structure manipulated whether that informa-
tion was structured in a thematically ?owing, coherent
argument or not. In the argument condition, the internal
auditor’s email was organized so that thematically con-
nected pieces of information were joined together into
?owing paragraphs. Speci?cally, the internal auditor began
with an opening paragraph about the inventory and its
competition, and then discussed how well the inventory
is selling, then other technologically superior products on
the market, then the current viability of alternative sales
markets, etc., in a thematically structured argument. In
contrast, the no argument condition provided exactly the
same statements about inventory as in the argument con-
dition, but in an arbitrary order, following the design of
Sedor (2002). To control for any potential, but unantici-
pated, order effects within the no argument condition, par-
ticipants in this condition received one of four arbitrary
orderings of the same information about inventory, as in
Sedor (2002) and similar to Lipe and Salterio (2002). Our
participants’ responses within this condition are not sensi-
tive to which arbitrary ordering they received (all
p’s > 0.18).
5
Thus, our no argument condition presents the
same information, but without the thematically structured
?ow of the argument condition.
Case conclusion and response variables
The internal auditor’s email then closed politely
(rudely) in the likable (dislikable) conditions. We then
asked participants to provide on a scale from 0 to 100 their
beliefs about the likelihood that the value of the medical
device inventory would need to be written down
(write-down judgments; Appendix C).
Participants then completed a separate post-experimen-
tal questionnaire. This questionnaire asked participants to
rate on scales from 1 to 9 the extent to which they believed
the internal auditor to be a reliable (perceived reliability) and
objective (perceived objectivity) source of information, as
well as respond to manipulation check questions.
Results
Manipulation checks
As a likability manipulationcheck, the post-experimental
questionnaire asked participants to rate the likability of the
internal auditor’s attitude on a scale from 1 to 9. In the
dislikable condition, participants’ average response is
4
We manipulated information to achieve systematic variation in how
strongly the information supported write-down. The less supportive
information is not outright unsupportive in a way that would make the
internal auditor’s recommendation seem unreasonable or implausible. For
example, it is not as though the internal auditor was arguing write-down
for a cutting-edge product ?ying off the shelves; in both cases the inventory
faced competition from a new, technologically superior product. We
designed the less supportive information to be weaker than the more
supportive information but still at least somewhat suggestive of inventory
obsolescence so that the internal auditor’s preference for recognizing
obsolescence would remain plausible. The levels of actual inventory write-
down judgments in our results suggest that we achieved substantial
variation in information strength without switching to implausibility. Mean
judgments indicate at least some risk of write-down in every condition.
5
We adapt other design features that have been used in prior research
investigating the effects of an information structure on users’ judgments
(e.g., Lipe & Salterio, 2002; Sedor, 2002). Besides the four arbitrary
orderings in the no argument condition, the argument condition began
with the phrase ‘‘Let me start by saying that’’ immediately before the ?rst
statement about inventory, which was not present in the no argument
condition, in order to maintain a transitional ?ow into the introduction of
the argument that would be absent in the no argument condition (Graesser,
Millis, & Zwaan, 1997; Sedor, 2002). However, each individual statement
about inventory was present in both the argument and no argument
conditions (as in Sedor, 2002 Experiment 2), to hold the underlying
information about inventory constant.
580 K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589
signi?cantly below the midpoint 5 (2.1, t
118
= À10.72,
p
one-tail
< 0.001).
6
In the likable condition, participants’
average response is signi?cantly above the midpoint
(6.1, t
118
= 4.45, p
one-tail
< 0.001), and signi?cantly above the
dislikable group (6.1 vs. 2.1; t
118
= 10.86, p
one-tail
< 0.001).
7
As an information manipulation check, the post-experimental
questionnaire presented participants with the data fromboth
treatment levels of the information manipulation, and asked
them to identify which data set is more strongly supportive
of inventory write-down. Participants identi?ed the informa-
tion from the more (less) supportive condition as the more
(less) supportive information set 93 percent of the time,
signi?cantly better than chance (v
2
1
= 99.4, p < 0.001).
8
In addition, we conducted a follow-up study that
gathered additional manipulation checks and other post-
experimental measures as its dependent variables, using
a sample of 70 similar professionals from similar manage-
ment training programs. Participants in this follow-up
study averaged 11.0 years of professional business experi-
ence and 6.4 years of managerial experience, and had up to
30 and 29.9 years, respectively. Consistent with our main
experiment, most of these participants (84 percent,
n = 59) were middle-level managers, 7 percent (n = 5) were
upper-level executives, 9 percent (n = 6) had no managerial
experience, and experience does not affect our ?ndings in
this follow-up study. We manipulated argument structure
and likability exactly as in our main experiment, and
(because we do not ?nd any statistically signi?cant inter-
actions involving information in our main experiment),
we held information constant at the less supportive level.
For our argument manipulation checks, the theoretically
important aspects of arguments are (1) that they provide
a thematically ?owing structure to the information, and
(2) that people naturally respond favorably to this the-
matic structure. Accordingly, we asked participants in the
follow-up study to rate on 9-point Likert-type scales (1)
how structured they perceived the internal auditor’s
information about inventory to be, as well as (2) how per-
suasive they perceived the information to be. Participants
perceived the (otherwise identical) information to be more
structured (means (standard deviations) = 5.7 (2.2) vs. 3.6
(2.5), t
66
= 3.76, p
one-tail
< 0.001) and more persuasive
(means (standard deviations) = 5.6 (1.8) vs. 4.0 (2.2),
t
66
= 3.23, p
one-tail
= 0.001) in the argument conditions than
in the no argument conditions. In fact, in results not tabu-
lated in this paper, we ?nd that perceived structure fully
mediates the effect of arguments on perceived persuasive-
ness, consistent with our theory that it is the thematically
organized structure of information in an argument to which
individuals respond favorably.
9
Overall, participants’
responses to all of our manipulation checks across both
experiments vary as expected.
10
Hypothesis tests
Table 1 shows descriptive statistics for our dependent
variable, write-down judgments, by experimental condition,
and Table 2 (Panel A) shows an analysis of variance.
11
As
Table 2 shows, we ?nd a signi?cant main effect of
information. Speci?cally, across experimental conditions,
participants’ write-down judgments are, on average, 23.2
percentage points higher in the more supportive information
conditions than in the less supportive information conditions
(69.3 vs. 46.1, t
125
= 6.20, p
one-tail
< 0.001). Thus, overall,
managers reacted to the underlying information presented
by the internal auditor as expected.
The ANOVA table also shows a signi?cant argument
structure  likability interaction (F
1,125
= 11.17, p = 0.001;
Table 2). Fig. 1 shows the least-squares means of this
interaction from the ANOVA model.
12
This interaction from
6
We use one-tailed p-values for t-tests of manipulation checks, the
information main effect, and our hypothesis where we make directional
predictions, and denote them with a subscript, p
one-tail
. All other p-values
are two-tailed.
7
Given the variation in interpersonal likability that likely exists in
practice (Pickett, 2007), we wanted to achieve dislikable conditions that
were both strong (Kerlinger & Lee, 2000) and realistic. While the 2.1 mean
rating in the dislikable group was signi?cantly below the midpoint of the
scale, it was also signi?cantly above the lower endpoint of the scale
(F
1,118
= 16.8, p < 0.001). Cohen’s d (a measure of practical signi?cance) also
suggests that this rating is well above the lower endpoint of the scale
(d = 0.82).
8
Our objective with the information manipulation check is to provide
construct validation that the information we placed into the more
supportive information condition was in fact more supportive of inventory
obsolescence than the information we placed into the less supportive
information condition. Rather than the within-subjects manipulation check
we used, we could have tested this between-subjects by simply asking
participants how much the information from the internal auditor’s email
was supportive of inventory obsolescence. However, we did not use this
approach because, as Pedhazur and Schmelkin (1991, p. 263) suggest,
simply asking participants to rate how supportive the internal auditor’s
information was of his position might not accomplish our objective if
participants interpreted the question in light of how they responded to our
dependent variable measuring whether they agreed with the internal
auditor. Unlike our dependent variable write-down judgments, participants’
responses to our within-subjects manipulation check was not signi?cantly
in?uenced by our other manipulated factors (i.e., argument structure and
likability), whereas a between-subjects manipulation check might have
been. Reffett (2010) uses a within-subjects manipulation check for similar
reasons. The results of our manipulation check accomplishes its purpose of
providing empirical evidence that our more supportive information was
more suggestive of inventory obsolescence than our less supportive
information.
9
Additionally, while our main experiment asked participants to rate how
much they liked the individual auditor’s attitude, we asked participants in
this follow-up study to make the same rating of his attitude, plus ratings of
his interpersonal skills and of his personal likability. Findings for all three of
these variables are similar to those in our main experiment (all p
one-
tail
’s < 0.001).
10
Our manipulation checks only vary according to main effects of their
matching manipulated variables. For each, we do not ?nd any signi?cant
interactions or main effects involving the other manipulated variables. We
obtain statistically similar results when dropping observations that failed
our manipulation checks.
11
Tests of ANOVA model assumptions are well below levels that elevate
type I error rates (i.e., below levels that would suggest an increased risk of
?nding spurious effects; see Neter, Kutner, Nachtsheim, & Wasserman,
1996; Tabachnick & Fidell, 2001).
12
The least-squares means in Fig. 1 show participants’ responses
collapsed across information (Neter et al., 1996; Searle, Speed, & Milliken,
1980). Because information does not interact with argument structure or
likability in any two-way interactions or in a three-way interaction (Table 2,
Panel A; see also Table 3 Panel A), results within each level of the
information conditions are statistically similar to those in Fig. 1, with the
only signi?cant exception being that the means within the more supportive
information conditions are, on average, 23.2 percentage points higher than
those in the less supportive information conditions.
K. Fanning, M. David Piercey / Accounting, Organizations and Society 39 (2014) 575–589 581
the ANOVA table indicates that an argument structure had a
greater impact on managers’ judgments when the internal
auditor was likable than when he was dislikable, consistent
with our theory.
Our hypothesis predicts an ordinal interaction such that
managers will agree more with the internal auditor when
he is both likable and uses an argument structure, and
relatively less under other conditions. To test this predic-
tion, we examine whether the observed means match the
pattern predicted by our hypothesis (Buckless &
Ravenscroft, 1990). First, as Fig. 1 shows, the means occur
in a pattern consistent with our expectations (see Keppel &
Wichens, 2004; e.g., Kadous et al., 2013). Second, we use
contrast weights of +3 for the likable/argument structure
condition, and À1 for the remaining conditions shown in
Fig. 1, and ?nd statistically signi?cant results (t
125
= 4.81,
p
one-tail
< 0.001; Table 2, Panel B). Moreover, the three con-
ditions assigned the weight of À1 do not differ signi?cantly
from each other (F
2,125
= 0.44, p = 0.61; Buckless &
Ravenscroft, 1990). These results support our hypothesis.
In addition, the means within the more supportive and less
supportive information conditions (see Table 1) also occur
in the expected pattern. Speci?cally, tests within each level
of information using contrast weights of +3, À1, À1, À1 pro-
vide statistically similar results (p
one-tail
’s 6 0.002; Table 2,
Panel B). Moreover, the three conditions within each level
of information assigned weights of À1 similarly do not
differ signi?cantly from one other, as well (p’s P0.34).
13
Supplementary analyses: Simple effects tests
Tests of the simple effects indicate the following results.
When the internal auditor is likable, the use of an
Table 2
Write-down judgments, analysis of variance, and hypothesis tests.
Panel A: Analysis of variance
Source Sum of squares df Mean square F p
Information 17466.6 1 17466.6 38.49