Description
Thisstudyinvestigatestheuseofauditevidencedocumentationinstructionsthatpromotethecollec-tionandprocessingofevidencewithhigh-levelconstruals(broad,abstractinterpretationsoftheevi-dence).Abstractioncanhelpapersonpiecetogetherindividualpiecesofinformationorevidenceandbetterenableapersontoseethebigpictureofwhatthecllectiveinformationportends.Theresultsofanexperimentsuggestthatauditorsthinkandactwithmoreprofessionalskepticismwhenusingthedocumentationinstructionsthatprmotehigh-levelconstrualsascomparedwithauditorsusingdocu-mentationinstructionspromotinglow-levelconstruals(specific,detailedinterpretationsoftheevidence,akintocurrentauditpractice)andwithauditorsnotgivendocumentationinstructions.Further,thehigh-levelconstrualsfosterbetterprocessingofthecollectedevidence.
Accounting, Organizations and Society 46 (2015) 44–55
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
Construal instructions and professional skepticism in evaluating
complex estimates
Jason Tyler Rasso
College of Charleston, United States
a r t i c l e i n f o
Article history:
Received 1 June 2014
Revised 16 March 2015
Accepted 17 March 2015
Available online 11 April 2015
a b s t r a c t
This study investigates the use of audit evidence documentation instructions that promote the collec-
tion and processing of evidence with high-level construals (broad, abstract interpretations of the evi-
dence). Abstraction can help a person piece together individual pieces of information or evidence and
better enable a person to see the big picture of what the collective information portends. The results
of an experiment suggest that auditors think and act with more professional skepticism when using the
documentation instructions that promote high-level construals as compared with auditors using docu-
mentation instructions promoting low-level construals (speci?c, detailed interpretations of the evidence,
akin to current audit practice) and with auditors not given documentation instructions. Further, the high-
level construals foster better processing of the collected evidence. The study also provides preliminary
evidence that task complexity could interfere with professional skepticism.
© 2015 Elsevier Ltd. All rights reserved.
Introduction
Regulators have recently criticized auditors for failing to dis-
play a proper amount of professional skepticism in their audits of
complex estimates (Bratten, Gaynor, McDaniel, Montague, & Sierra,
2013; IFIAR, 2012; PCAOB, 2012). One possible explanation for
the lack of professional skepticism is that audits of complex es-
timates require the processing of numerous pieces of audit evi-
dence collected over an extended period of time. This study ex-
plores whether and how construals (interpretations) of evidence
affect an auditor’s judgments and decisions relating to the collec-
tion and processing of audit evidence, particularly when all of the
evidence is not yet available.
Speci?cally, this paper examines whether documentation in-
structions that promote broad, abstract interpretations (high-level
construals) of the audit evidence can be helpful in promoting ef-
fective processing of information throughout the evidence collec-
tion process. Studying the process of evidence collection is im-
portant because a central component of professional skepticism
is the suspension of judgment until su?cient and competent ev-
idence has been obtained (AU 230.07-.08; Hurtt, 2010). Auditing
complex estimates, like many other audit tasks, involves a sequen-
tial judgment process through which auditors receive information
and then need to make important decisions based on that informa-
E-mail address: [email protected]
tion such as whether to continue collecting new evidence (Gibbins,
1984; Knechel & Messier, 1990). Failure to properly process evi-
dence, particularly early in the evidence collection process, could
lead to poor judgments such as a decision to end the evidence col-
lection process prematurely (which would be judged ex post as a
lack of professional skepticism). As support, experienced auditors
recently reported that they have trouble processing and assimi-
lating collected audit evidence, particularly negative evidence that
ought to increase skepticism and suggest to an auditor that evi-
dence collection should continue (Gri?th, Hammersley, & Kadous,
in press).
Construal-level theory (CLT) suggests that the manner in which
individuals construe (interpret) information and evidence impacts
their judgments and decisions. The theory describes two levels
of construals: low-level construals and high-level construals. Low-
level construals are speci?c and detailed while high-level constru-
als are broad and abstract (Trope & Liberman, 2003). Fundamental
to this study, one of the primary differences between construal lev-
els is that high-level construals make it easier for an individual to
process and assimilate numerous pieces of information, including
negative information, while low-level construals make such a task
more di?cult (Trope & Liberman, 2010, 2003).
Based on evidence from audit practice, I argue that auditors are
likely processing audit evidence with low-level construals. As de-
scribed previously, experienced auditors report having trouble pro-
cessing and assimilating audit evidence. These auditors describehttp://dx.doi.org/10.1016/j.aos.2015.03.003
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55 45
having trouble “seeing the big picture” surrounding the estimate
as well as having di?culty detecting patterns when given all of
the available information (Gri?th et al., in press). If an auditor
does not detect a problem with an estimate (because of a failure
to properly process negative evidence, for example) then the col-
lection of audit evidence phase of the audit could be terminated
too quickly, an act consistent with lower professional skepticism.
I posit that documentation instructions that promote high-level
construals have the potential to overcome these di?culties because
this level of construal facilitates better processing and assimila-
tion of all information, both positive and negative, which should
allow an auditor to properly determine whether the amount of ev-
idence collected is su?cient to render a judgment about the ap-
propriateness of the estimate (Trope & Liberman, 2010, 2003). To
test this prediction, I employ a 1 × 3 between-participants experi-
ment using auditors experienced in auditing complex estimates as
participants. One group received documentation instructions which
prime high-level construal thinking while another group received
documentation instructions that prime low-level construal think-
ing. A third group of participants did not receive documentation
instructions and served as a control condition.
In a setting calling for increased professional skepticism,
the participants performed a simulated fair value audit task
in which they collected and reviewed audit evidence that col-
lectively suggests the client’s estimate is on the high end of
a reasonable range (suggesting that the estimate is aggres-
sive). I capture multiple measures of professional skepticism be-
cause of the lack of consensus in the construct’s de?nition
(Hurtt, 2010; Hurtt, Brown-Libured, Earley, & Krishnamoorthy,
2013; Nelson, 2009). I predict that auditors using the documen-
tation instructions that promote high-level construals will dis-
play the highest level of professional skepticism and will ulti-
mately produce the highest judgments of risk involved with the
estimate.
Consistent with the sequential process of auditing (Gibbins,
1984; Knechel & Messier, 1990), the participants collect evidence
about the fair value estimate and make decisions concerning
whether they collected su?cient evidence to render judgments
about the estimate or need to continue gathering more evidence.
Participants given the high-level construal instructions document
the collected evidence while considering the evidence broadly and
why the estimate could be fairly presented or materially misstated.
Participants using the low-level construal instructions document
evidence while considering the evidence speci?cally and how the
estimate could be fairly presented or materially misstated. Partici-
pants in the control condition do not receive instructions and only
document evidence at the end of the collection process.
I ?nd that auditors given documentation instructions
priming high-level construals display higher levels of both
skeptical judgments and skeptical actions. Auditors given
these instructions spent signi?cantly more time on the
task, particularly in the collection and evaluation of au-
dit evidence phase of the task. Auditors using the high-
level construal instructions also collected signi?cantly more
evidence than did auditors using the low-level construal
instructions.
Furthermore, auditors using the high-level construal instruc-
tions rated the fair value estimate as signi?cantly more risky than
did their counterpart auditors in the other two conditions. Note
that the increases in skepticism do not result merely from the col-
lection of greater amounts of evidence. There is no difference in
skeptical judgments between users of low-level instructions and
the control group; however, the control group collected statisti-
cally the same amount of evidence as users of the high-level in-
structions. Interestingly, the use of the high-level construal instruc-
tions appears to have moderately increased perceptions of task
di?culty.
This study contributes to existing literature streams and an-
swers several calls for research. I extend the literature on auditing
and construal-level theory by showing that considering audit evi-
dence through high-level construals leads to enhanced professional
skepticism. This ?nding adds to the growing stream of literature
which suggests that interventions can be used to improve how au-
ditors cognitively process audit information (Brewster, 2011; Plum-
lee, Rixom, & Rosman, 2015).
Concurrent research documents alternative mechanisms that
help promote effective processing of information, including the de-
liberative mindset (Gri?th, Hammersley, Kadous, & Young, 2015)
and convergent and divergent thinking (Plumlee et al., 2015).
Backof, Thayer, and Carpenter (2014) also investigate the bene?ts
of construal-level theory when auditing complex estimates. These
studies, however, only show more effective processing of infor-
mation when auditors are given a relatively complete set of ev-
idence.
1
An important aspect of professional skepticism and of
audits of complex estimates in general is the collection of evi-
dence (AU 230.07-.08; Hurtt, 2010). My study shows that auditors
struggle with processing incomplete sets of evidence. This prob-
lem manifests as a lack of professional skepticism when the au-
ditor prematurely concludes the evidence collection phase of the
audit and potentially makes poor judgments based on the incom-
plete evidence. An important ?nding of my study is that audi-
tors using the high-level construal instructions can better pro-
cess incomplete sets of evidence and properly recognize that
they need more evidence in order to make a conclusion. Fur-
ther, when provided with the same amount of evidence, audi-
tors using the high-level construal instructions are more likely
to recognize the high risk level of a complex estimate when
the majority of the evidence suggests the estimate could be
overstated.
This study also contributes to practice by presenting an eas-
ily implementable technique that can be used to promote effective
processing of audit evidence. The documentation instructions that
prime high-level construals are simple to use, inexpensive, and can
easily be tailored for a ?rm’s speci?c needs or language. The in-
structions can be used to help overcome at least one of the de-
?ciencies noted by regulators in their inspection reports, namely
the lack of perceived professional skepticism (PCAOB, 2012, 2008).
Consequently, regulators should also ?nd these results to be infor-
mative.
The remainder of this paper is organized as follows. The next
section describes the problem, reviews the literature relevant to
this study’s setting, and develops the hypotheses. Next, I discuss
the method used to examine the hypotheses. The fourth section
presents the results, and the ?nal section offers a discussion of the
conclusions drawn from the study.
Literature review and hypothesis development
Audits of complex estimates and professional skepticism
One of the many di?culties involved in audits of complex es-
timates is that the audit takes place over the course of time. Au-
1
The Backof, Thayer, et al. (2014) study also differs from the current study
because it focuses on the reasonableness of management’s assumptions. The
focus on one particular aspect of the complex estimate is likely the rea-
son why their results differ from those in this study. Low-level constru-
als would aid an auditor focused on one aspect of the audit while high-
level construals would aid an auditor in assimilating evidence, both complete
sets and incomplete sets, and better see the “big picture” surrounding the
estimate. I discuss these differences in more detail in the ?nal section of
this paper.
46 J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55
dit evidence is collected piecemeal over that time rather than all
at once. Auditors interviewed by Gri?th et al. (in press) confess to
having di?culty processing and assimilating audit evidence, partic-
ularly negative evidence, and this problem can only be worse when
the auditors do not have the complete set of evidence available to
them.
Due to the sequential process of auditing, judgments are made
at several points during an audit as evidence is collected, including
the decision whether to continue or terminate the evidence collec-
tion process (Gibbins, 1984; Knechel & Messier, 1990). One of the
consequences of the problem noted above is that a failure to ad-
equately respond to negative evidence can result in a premature
termination of the evidence collection process. An auditor could
fail to see that an estimate is aggressive (as stated by some of the
auditors in Gri?th et al. (in press)) and decide that the evidence
collected so far is su?cient to conclude that the estimate is rea-
sonable. The suspension of judgment until an auditor obtains an
appropriate level of evidence is a key aspect of professional skepti-
cism (Hurtt, 2010; Nelson, 2009; PCAOB, 2014; AU 230.09).
Regulators, having the bene?t of a fuller set of evidence when
inspecting the audit months or years later, would likely conclude
that the auditor did not display su?cient professional skepticism
because the auditor failed to obtain an appropriate level of evi-
dence. As evidence, U.S. and international audit inspectors cite a
lack of professional skepticism as a primary reason for audit de?-
ciencies (IFIAR, 2012; PCAOB, 2012, 2008). In particular, a summary
report details that “In some instances, ?rms did not su?ciently
test or challenge management’s forecasts, views, or representations
that constituted critical support for amounts recorded in the ?nan-
cial statements” (PCAOB, 2008, p. 20). I posit that documentation
instructions employing the precepts of construal-level theory can
promote skepticism and enhance processing of evidence when us-
ing evidence sets at various levels of completion.
Construal-level theory
Construal-level theory (CLT) makes predictions of how
individuals construe, or interpret, information. Researchers
have used this theory to describe how individuals make
predictions and evaluate situations, how gambling prefer-
ences are affected, and how construals can affect interper-
sonal negotiation (Trope & Liberman, 2010, 2003). The level
of construal through which a person processes informa-
tion shapes how the information is encoded in a person’s
mind.
There are two levels of construals identi?ed by CLT: high-level
construals and low-level construals. Viewing information through
high-level construals makes information more abstract. This ab-
straction allows for easier assimilation of information because the
representations become less ambiguous, more coherent, and more
schematic (Trope & Liberman, 2010). High-level construals focus on
why something has been done (Fujita, Trope, Liberman, & Levin-
Sagi, 2006; Trope & Liberman, 2010, 2003). Low-level construals
are more detailed, transactional interpretations of information that
focus on how something has been done (Fujita et al., 2006; Trope
& Liberman, 2010, 2003).
To illustrate the differences in levels, consider the follow-
ing example of a jigsaw puzzle. A person viewing the puz-
zle pieces with high-level construals loses some of the spe-
ci?c details of the pieces (colors, shapes, end vs. middle piece,
etc.), but can better see how the pieces of the puzzle ?t to-
gether. In contrast, a person viewing the puzzle pieces with
low-level construals can likely tell you speci?c details about
many of the pieces, but the detailed level of focus restricts the
person from being able to see how the individual pieces ?t
together.
2
Interviews with experienced auditors suggest that auditors
could be using low-level construals when processing audit evi-
dence. The interviewees report having problems assimilating audit
evidence and seeing the c?big pictured? or the overall pattern sug-
gested by the evidence (Gri?th et al., in press). Due to the piece-
meal nature of the evidence collection process during the audit of
a complex estimate, auditors are likely processing and recording
each evidence item speci?cally as the evidence is received. This act
would be consistent with processing the evidence with low-level
construals, particularly if the auditors do not take the time to con-
sider how previously received evidence relates to the evidence just
received.
Audit plans and speci?c ?rm guidance/policy on how to process
and document audit evidence are generally not made available to
the public for reference. One recent exception is KPMG’s profes-
sional judgment framework. In this framework, KPMG recognizes
the critical role of gathering and evaluating information and es-
pouses the importance of seeing the big picture of information;
however, the framework does not reference the problem associ-
ated with collecting evidence over time nor does it offer sugges-
tions for how to effectively process previously obtained evidence
(KPMG, 2011).
Due to the lack of proper references concerning this issue, I
interviewed seven auditors concerning the practice of collecting
and evaluating audit evidence over time. These auditors averaged
8.14 years of audit experience and each had experience with au-
diting complex estimates. The auditors represented a broad cross-
section of ?rm type, ranging from Big 4 auditors (three), auditors
from national ?rms (two), and auditors from local or regional ?rms
(two).
Only one of the seven auditors mentioned that his or her ?rm
provided guidance on how to assimilate previously received evi-
dence with evidence currently received. The remaining six audi-
tors shared similar experiences of collecting and recording evi-
dence items individually as they were received, a manner consis-
tent with processing the evidence with low-level construals. None
of the seven auditors interviewed reported using any sort of rou-
tine that would resemble processing evidence, either currently or
previously received, with high-level construals.
Processing evidence with high-level construals can increase ef-
fective processing of information. In an auditing context, the ab-
straction facilitated by high-level construals makes the collected
audit evidence easier to assimilate and easier for an auditor to rec-
ognize patterns in the information (Trope & Liberman, 2010).
3
The
abstraction and subsequent assimilation of the collected evidence
will allow auditors to recognize the need for increased professional
skepticism if the evidence contains items suggesting something
unusual with the estimate (negative or discon?rming evidence).
For example, auditors could realize that the estimate is aggressive
and respond by collecting more evidence, concluding that the es-
timate is risky and making appropriate adjustments to the audit
plan. Thus, processing evidence with high-level construals should
help overcome the problems described above resulting from the
2
One can also imagine a person with his face low to the puzzle. This person can
see the details of the speci?c pieces; however, it’s not until he raises his face to a
higher level that he can see how the pieces ?t together.
3
As a colloquial example of how this process works, imagine ?ve very large let-
ters in front of you that spell a word. Each letter individually takes up your entire
?eld of view. By creating distance from the letters, you are better able to see the
entire word rather than just the individual letters.
J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55 47
piecemeal evidence collection process. Conversely, low-level con-
struals allow individual elements of evidence to dominate others
to the extent that these elements block an individual’s ability to
see the totality of what the evidence suggests (Trope & Liberman,
2010).
The hypotheses described below test the proposed advantages
of processing audit evidence with high-level construals.
4
H1. Auditors using documentation instructions that prime high-
level construals will display higher professional skepticism than
will auditors using documentation instructions that prime low-
level construals.
H2. Auditors using documentation instructions that prime high-
level construals will display higher professional skepticism than
will auditors not given documentation instructions.
Method
Participants
Fifty-eight auditors experienced in auditing complex estimates
and representing six accounting ?rms participated in the study.
Table 1 displays information about these participants. The audi-
tor participants averaged 5.4 years of audit work experience and
ranged in rank from staff auditor to partner. Ninety percent of the
participants had experience auditing fair value estimates.
5
Design
I use a 1 × 3 between-participants experimental design to
test my hypotheses. The independent variable in this experiment
is the type of documentation instructions given to participants,
which comprises three levels: high-level construal documentation
instructions (hereafter referred to as high-level instructions for
brevity), low-level construal documentation instructions (hereafter
referred to as low-level instructions), or no documentation instruc-
tions. The computerized experimental instrument randomly as-
signed participants to one of the three experimental conditions.
4
Although construal-level theory maps on to the speci?c features of the audit
setting described in this section, there are other related ideas and theories that
generally support the hypotheses. For example, Trope and Liberman (2010) also dis-
cuss the idea of psychological distance, the distance that an individual is removed
from something either spatially, temporally, or mentally. High-level construals can
increase psychological distance which, in an audit setting, can increase the mental
(psychological) distance between an auditor and the evidence. The increased dis-
tance could lead to less biased processing which would result in higher profes-
sional skepticism. Further, mindset theory appears to support the hypotheses. For
example, processing evidence with a deliberative mindset can be bene?cial because
individuals in this mindset tend to think more critically about the information pre-
sented to them (Gri?th et al., 2015) and tend to remain impartial and objective
longer (Gollwitzer, 1990).
5
I collected responses from the auditors in person whenever possible. However,
28 participants (48.3% of the participants) completed the study on their own time
due to ?rm requests or logistical problems. I found signi?cant differences in gen-
eral work experience and the type of ?rm that employed each participant, but the
inclusion of these two variables or a testing location variable does not affect the re-
sults reported later in this paper. The differences in ?rm type can be explained by
the fact that all the participants from local, regional, and national accounting ?rms
completed the study in a controlled setting while only 35% of the participants from
international ?rms completed the study in a controlled setting. The differences in
work experience is likely due to outliers in the variables resulting from two part-
ners with work experience of 25 and 30 years, respectively completing the study in
a controlled setting. No other potential covariates, including scores from the Hurtt
(2010) trait professional skepticism scale, are signi?cant.
Task
I use a case adapted from Kohlbeck, Cohen, and Holder-Webb
(2009) which involves an audit of a client’s intangible asset ac-
count (reacquired franchise rights). Each participant began the ex-
periment by reviewing background and ?nancial information about
the client, American Pizza Company (APC). The information in-
cluded a description of the company and its franchising activities.
A table presented information about the client’s ?nancial state-
ments, including the value of its reacquired franchise rights ac-
count (an intangible asset recorded at fair value). Participants also
read information about the client’s accounting policies related to
reacquired franchise rights.
The next screen contained information relevant to the experi-
mental task. Each participant read that their ?rm has audited APC
for over ten years and that the client is a signi?cant source of rev-
enue for the ?rm. The participants considered speci?c information
about the re-acquired franchise rights account and learned that the
fair value of the account was greater than the book value of the
account. The current screen also displayed management’s assump-
tions used in generating the fair value estimate.
After reviewing the client and task information, the partici-
pants began to collect evidence relating to the task. I created thirty
pieces of evidence that related to the client and/or the client’s
reacquired franchise rights accounts. In order to create a setting
that called for increased professional skepticism, I generated 12 ev-
idence items which collectively suggested that the client’s estimate
was aggressive (at the high end of a reasonable range of the esti-
mate). I also generated six evidence items which collectively sug-
gested that the client’s estimate was fairly stated and 12 neutral
evidence items which were largely irrelevant to the task.
6
In order to reinforce the fact that time spent collecting audit
evidence is not costless, the computerized instrument displayed a
time budget to all participants. The use of a time budget in this
task is consistent with audit practice. On the ?rst screen of evi-
dence, each participant learned that 70% of the current audit bud-
get has been expended to generate the initial round of evidence.
After learning the percent of audit hours used, each participant
read the following statement:
Your audit manager will look favorably upon you if you com-
plete the fair value audit using as few audit hours as possible;
however, you should continue searching for evidence until you
have satis?ed yourself that you have obtained su?cient appro-
priate evidential matter to provide you with a reasonable basis
for forming an opinion.
Each successive round of evidence collection incremented the
percentage used by 10%. The initial collection given to each par-
ticipant contained the same 10 evidence items (four negative, four
neutral, and two positive, all randomly displayed). Each participant
had the ability to continue collecting evidence up to four times
with each subsequent collection revealing ?ve new evidence items
in the same ratio of two negative, two neutral, and one positive
items each.
7
Participants using all available rounds of collection
6
I gave the list of evidence to several current and former auditors who agreed
that each item of evidence was appropriately labeled as either positive, negative, or
neutral evidence. I also gave an unlabeled, alphabetized list to three audit mangers
who rated the strength of each evidence item on a scale of ?5 (Extremely nega-
tive) to +5 (Extremely positive), with 0 labeled as neutral. The average strength of
the negative evidence was ?2.53 (s.d. = 1.73). The average strength of the positive
evidence was 2.17 (s.d. = 1.82) while the average strength of the neutral evidence
was ?0.08 (s.d. = 1.44). These ratings by the audit managers provide support for
the claimed directionality of the evidence as being either positive, negative, or neu-
tral.
7
Across all experimental conditions, each round of collection provided the same
48 J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55
Table 1
Participant demographic statistics.
High-level
instructions condition
mean (s.d. or percent
of sample) (n = 19)
Low-level
instructions condition
mean (s.d. or percent
of sample) (n = 19)
Control condition
mean (s.d. or percent
of sample) (n = 20)
Total mean (s.d. or
percent of sample) (n
= 58)
One-way ANOVA
p-value
Gender 0.529
Male 10 (52.6%) 11 (57.9%) 8 (40.0%) 29 (50.0%)
Female 9 (47.4%) 8 (42.1%) 12 (60.0%) 29 (50.0%)
Age 28.05 (4.034) 29.89 (4.267) 30.70 (5.823) 29.57 (4.842) 0.221
General work experience (years) 7.11 (5.322) 8.42 (6.086) 9.35 (6.930) 8.31 (6.125) 0.526
Audit work experience (years) 4.63 (3.095) 5.26 (4.107) 6.25 (5.077) 5.40 (4.171) 0.481
Experience auditing fair value estimates 16 (84%) 17 (89%) 19 (95%) 52 (90%) 0.556
Hurtt professional skepticism score 131.74 (16.251) 124.05 (16.092) 133.15 (17.942) 129.71 (16.990) 0.204
Auditor rank 0.496
Staff 3 (15.8%) 2 (10.5%) 3 (15.0%) 8 (13.8%)
Senior 13 (68.4%) 12 (63.2%) 10 (50.0%) 35 (60.3%)
Manager 3 (15.8%) 4 (21.1%) 6 (30.0%) 13 (22.4%)
Partner 0 (0.0%) 1 (5.2%) 1 (5.0%) 2 (3.5%)
Firm type 0.923
Local 1 (5.3%) 1 (5.3%) 2 (10.0%) 4 (6.9%)
Regional 3 (15.8%) 3 (15.8%) 2 (10.0%) 8 (13.8%)
National 0 (0.0%) 1 (5.3%) 2 (10.0%) 3 (5.2%)
International 15 (78.9%) 14 (73.7%) 14 (70.0%) 43 (74.1%)
went over budget by 10% (used 110% of the audit hours allocated
to the task by the audit manager). The computerized instrument
randomized the order of the evidence items in all conditions.
After each round of collection, participants given documenta-
tion instructions responded to the questions designed to invoke
a distinct construal-level. Participants given the high-level instruc-
tions responded to the following two questions when considering
the audit evidence:
c?Thinking broadly about all of the evidence collectively, list
reasons why management’s estimate could be fairly presented.
Thinking broadly about all of the evidence collectively, list
reasons why management’s estimate could be materially mis-
stated.d?
I drew the wording for this condition from prior literature
which shows that thinking broadly about a situation and why
the situation occurred invokes high-level construals of informa-
tion (Backof, Bamber, & Carpenter, 2014; Fujita et al., 2006; Rim,
Uleman, & Trope, 2009). Having the participants think about both
sides of the situation is consistent with the KPMG judgment frame-
work’s suggestion to have auditors consider both the pros and cons
of a situation (KPMG, 2011) and with current accounting research
that also investigates the effect of using a judgment framework
based on construal-level theory, but in a much different setting
(Backof, Bamber, et al., 2014).
8
The participants viewed these in-
structions both before and after considering the audit evidence
produced by each of their collections.
The participants given the low-level instructions saw the fol-
lowing wording when considering audit evidence:
c?Thinking speci?cally about each evidence item, list reasons
how management’s estimate could be fairly presented.
Thinking speci?cally about each evidence item, list reasons
?ve evidence items presented randomly. For example, on the second collection
round all conditions viewed evidence items 11 through 15 presented randomly. On
the third collection round all conditions viewed evidence items 16 through 20 pre-
sented randomly, and so on. The evidence sequence can be found in Appendix A.
8
Considering reasons why the estimate could be fairly presented biases against
?nding results in this study.
how management’s estimate could be materially misstated.d?
I drew the wording for this condition from prior literature
which shows that thinking about a situation with speci?c details
of the c?howd? of a situation will invoke low-level construals of
information (Backof, Bamber, et al., 2014; Fujita et al., 2006; Rim
et al., 2009). The participants in the control condition did not re-
ceive documentation instructions but were still required to doc-
ument the reasoning for their judgments and actions. The control
group documented evidence at the end of the task concurrent with
providing their judgments.
Dependent variables
I captured multiple measures of professional skepticism due to
the debate surrounding a proper de?nition of professional skep-
ticism (Glover & Prawitt, 2013; Hurtt et al., 2013; Nelson, 2009).
Consistent with the model developed by Nelson (2009), I use sep-
arate measures for skeptical judgment and skeptical action. Af-
ter determining that they collected su?cient evidence, the partic-
ipants reported their perceived risk that the estimate is materially
misstated and their perceived likelihood that they would recom-
mend an adjustment to the client’s reacquired franchise rights ac-
count. I consider these two variables to be measures of skeptical
judgment consistent with Hurtt et al. (2013).
9
The computerized instrument allowed me to record the amount
of time each auditor spent collecting and reviewing evidence as
well as the total time spent on the task. The instrument also al-
lowed me to view the number of times each auditor collected ad-
ditional audit evidence. I consider each of these variables to be a
measure of skeptical action consistent with the de?nition provided
by Hurtt et al. (2013). Table 2, Panel A displays univariate statistics
for each of the measures described above while Panel B displays
the correlations between the dependent variables. The participants
9
Two of the measures of skeptical action (time spent collecting and reviewing
audit evidence and the number of searches for evidence) come before the mea-
sures of skeptical judgment. Normally judgments precede actions and I do not ar-
gue differently here. Auditing is a process that involves sequential decision making
(Gibbins, 1984; Knechel & Messier, 1990). I do not explicitly capture the judgments
that precede the actions described, but presumably the initial judgments would be
whether or not the accumulated evidence is su?cient to render a conclusion or
whether more evidence should be collected.
J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55 49
Table 2
Dependent variable information.
High-level instructions
condition mean (s.d.)
Low-level instructions
condition mean (s.d.)
Control condition mean
(s.d.) Total mean (s.d.)
(n = 19) (n = 19) (n = 20) (n = 58)
Panel A: Descriptive statistics
Risk of material misstatement 7.53 (2.245) 4.00 (2.380) 4.25 (2.074) 5.24 (2.723)
Likelihood of recommending an adjustment 6.74 (2.997) 3.16 (2.522) 4.00 (2.103) 4.62 (2.943)
Time spent collecting and reviewing audit evidence (in minutes) 16.65 (6.371) 11.15 (11.778) 5.14 (2.973) 10.88 (9.067)
Total time spent on the study (in minutes) 26.68 (6.407) 19.09 (13.312) 17.59 (6.470) 21.06 (9.946)
Number of additional collections of evidence 2.63 (1.535) 1.37 (1.571) 2.50 (1.100) 2.17 (1.500)
Risk of material
misstatement
Likelihood of
recommending an
adjustment
Time spent collecting
and reviewing audit
evidence
Total time spent on the
study
Number of additional
collections of evidence
Panel B: Correlations between dependent variables
Risk of material misstatement ? 0.844 0.509 0.454 0.453
0.000 0.000 0.000 0.000
Likelihood of recommending an
adjustment
0.815 ? 0.304 0.286 0.393
0.000 0.020 0.029 0.002
Time spent collecting and reviewing
audit evidence
0.567 0.439 ? 0.926 0.549
0.000 0.001 0.000 0.000
Total time spent on the study 0.477 0.381 0.888 ? 0.610
0.000 0.003 0.000 0.000
Number of additional collections of
evidence
0.442 0.433 0.554 0.612 ?
0.001 0.001 0.000 0.000
Pearson product-moment correlation coe?cients are displayed above the ?’s while the Spearman’s rank correleation coe?cients are displayed below the ?’s. The p-values
for the correlations are listed beneath the coe?cients.
concluded the task by completing a post-experimental question-
naire that included a manipulation check, the Hurtt (2010) profes-
sional skepticism scale, and demographic questions.
10
Results
Manipulation check
I had two coders blind to the hypotheses code each partici-
pant’s documentation. I gave the coders training in the differences
between high-level construals and low-level construals and then
asked them to code each group of responses as either a +1 if the
responses appeared to be high-level construals or a ?1 if the re-
sponses appeared to be low-level construals.
11
Participants using
the high-level instructions averaged a response of 0.465 indicating
that, on average, their responses were consistent with processing
evidence with high-level construals.
In contrast, participants using the low-level instructions aver-
aged a response of ?0.407 while participants not receiving doc-
umentation instructions averaged a response of ?0.158. Both of
these response levels indicate that, on average, users of the low-
level instructions and participants not given documentation in-
structions processed evidence with low-level construals. There is a
signi?cant difference in construal levels between the three groups
overall (p-value < 0.001).
I note that there is not a signi?cant difference in construal
level between users of the low-level instructions and participants
not given documentation instructions (planned contrast p-value =
0.283). This ?nding provides validation of the claim that using the
low-level instructions is akin to current audit practice regarding
10
I tested all responses in my post-experimental questionnaire, including re-
sponses to the Hurtt (2010) professional skepticism scale. The inclusion of these
possible control variables does not affect the inferences drawn from the results de-
scribed in the next section.
11
The coders worked out any differences in coding in a single round of reconcili-
ation resulting in 100% agreement in the ?nal coding set.
the collection and processing of audit evidence. Presumably au-
ditors not receiving documentation instructions will fall back on
collection and processing methods that they use in practice. The
lack of a signi?cant difference in construal level shows that the
two groups of auditors are both processing evidence with low-level
construals on average.
12
Test of hypotheses – Skeptical judgment dependent variables
The ?rst skeptical judgment dependent variable is the partic-
ipants’ assessments of the risk that the client’s estimate is ma-
terially misstated. The participants entered their responses on a
scale of 1 to 10 where 1 signi?ed c?Very Unlikelyd? and 10 sig-
ni?ed c?Very Likely.d? The second skeptical judgment dependent
variable is the participants’ assessments of the likelihood that they
would recommend an adjustment to the client’s estimate. The par-
ticipants entered their responses on a scale of 1 to 10 where 1
signi?ed c?Very Unlikelyd? and 10 signi?ed c?Very Likely.d? Given
the high correlation (see Table 2, Panel B) as well as a high Cron-
bach’s alpha (0.914) between these two variables, for parsimony, I
combine the variables and describe the results based on tests of
the combined skeptical judgment variable.
13
Hypothesis 1 predicts that auditors using the high-level instruc-
tions will display higher professional skepticism than will auditors
12
There is generally a rather broad distinction between responses coded as high-
level construals and those coded as low-level construals. The majority of responses
coded as low-level construals simply repeated individual items of evidence without
making any links between the evidence. For example, one responses coded as a
low-level construal was: c?the estimate was prepared by an accounting o?cial. The
use of the model conforms with GAAP.d? The majority of responses coded as high-
level construal demonstrated an assimilation of evidence items (past and current).
For example, one auditor wrote: c?I think the evidence collectively supports that
the estimate is extremely high. The client’s overall choices seem to indicate that
they want the estimate high, and this type of pressure would be a good reason
why the estimate could be materially misstated.d?
13
Inferences drawn from the results described hereafter are the same as those
that would be drawn from an analysis of the individual measures.
50 J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55
Table 3
Tests of hypotheses, combined skeptical judgment dependent variable.
Source df Mean square F-statistic p-Value
Panel A: ANOVA results
Construal-level instructions 2 69.867 13.627 0.000
???
Error 55 5.127
Total 57
Contrast Difference Std. error t-statistic df p-Value (one-tailed)
Panel B: Planned contrasts
High-level > low-level 3.58 0.735 4.836 55 0.000
???
High-level > control 3.01 0.725 4.145 55 0.001
???
Low-level < control 0.55 0.725 0.753 55 0.455
???
Signi?cant at the 0.001 level.
using the low-level instructions. Hypothesis 2 predicts that audi-
tors using high-level instructions will display higher professional
skepticism than will auditors not given documentation instruc-
tions. I use one-way Analysis of Variance (ANOVA) and planned
contrasts to test the hypotheses.
14
Panel A of Table 3 displays the results of the ANOVA analy-
sis for the combined skeptical judgment dependent variable. This
test reveals that there are signi?cant differences in the dependent
variable between conditions (p-value < 0.001) but does not reveal
where these differences can be found. I therefore use planned con-
trasts to test Hypotheses 1 and 2. Panel B of Table 3 shows the re-
sults of the planned contrasts using the combined dependent vari-
able.
The ?rst contrast tests whether the skeptical judgment dis-
played by participants using the high-level instructions is higher
than that displayed by participants using the low-level instruc-
tions. The signi?cant p-value ( low-level 5.49 3.072 1.787 27.703 0.043
?
Contrast Difference Std. error t-statistic df p-Value (one-tailed)
Panel D: Planned contrast of total time spent on the experiment
High-level > low-level 7.59 3.389 2.238 25.914 0.017
?
?
Signi?cant at the 0.05 level.
??
Signi?cant at the 0.01 level.
???
Signi?cant at the 0.001 level.
Table 5
Tests of hypotheses, number of evidence collections dependent variable.
Source df Mean square F-statistic p-Value
Panel A: ANOVA on number of collections of audit evidence
Construal-level instructions 2 9.217 4.615 0.014
?
Error 55 1.997
Total 57
Contrast Difference Std. error t-statistic df p-Value (one-tailed)
Panel B: Planned contrasts of number of collections of audit evidence
High-level > low-level 1.26 0.459 2.755 55 0.004
??
High-level > control 0.13 0.453 0.291 55 0.386
Low-level < control 1.13 0.453 2.499 55 0.008
??
???
Signi?cant at the 0.001 levels, respectively.
?
Signi?cant at the 0.05 levels, respectively.
??
Signi?cant at the 0.01 levels, respectively.
instructions appear to be better at processing incomplete sets of
evidence than auditors using low-level instructions, leading them
to more frequently realize the need for greater evidence before ar-
riving at a conclusion.
The second contrast shows that auditors using high-level in-
structions conducted only 0.13 more collections, on average, than
auditors not given documentation instructions. This result is
not signi?cant (p-value = 0.386) and provides no support for
Hypothesis 2. This lack of a ?nding is curious considering that au-
ditors using high-level instructions displayed higher professional
skepticism than auditors not given documentation instructions
with the measures of skeptical judgment discussed earlier.
One possible reason why auditors in these two conditions do
not differ in their number of collections could be a difference in
their perceptions of the task di?culty. The participants not given
documentation instructions could have continued collecting evi-
dence simply because it was less cognitively taxing to do so since
they did not have the requirement to document evidence each
time. Also, control participants did not have to explicitly consider
reasons why the estimate could be fairly presented or materially
misstated after each collection. I asked the participants in the post-
experimental questionnaire to rate their perceived di?culty of the
task. Participants using high-level instructions rated the task as
signi?cantly more di?cult than did participants in the other two
conditions (p-value = 0.049). However, there is not a signi?cant
correlation between the number of collections dependent variable
and the task di?culty covariate (Pearson correlation -0.027, p-value
= 0.841).
In order to more de?nitively test my conjecture, I test for this
correlation again using only participants given high-level instruc-
tions and participants not given documentation instructions. This
test reveals a negative correlation between the number of collec-
tions and perceived task di?culty just short of conventional sig-
ni?cance (Pearson correlation 0.310, p-value = 0.055). The change
from almost no relation to a moderately signi?cant relation pro-
vides fuel for future research into the possible relation between
task di?culty and professional skepticism.
Despite conducting statistically the same number of rounds of
evidence collection, recall that participants using high-level in-
structions properly rated the estimate’s risk as high given that the
collected evidence pointed to the estimate residing on the high
end of a reasonable range. Judgments of the perceived risk of the
estimate did not differ between users of low-level instructions and
participants not given documentation instructions. These results
52 J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55
collectively suggest that, when given the same level of informa-
tion, using high-level instructions allows a user to more effectively
process information.
Comparison of results with concurrent research
Overall, I ?nd that documentation instructions based on high-
level construals can increase professional skepticism in a setting
calling for higher skepticism. There are two concurrent research
studies that largely speak to the same issues as the ones noted
in my study: Backof, Thayer, et al. (2014) and Gri?th et al. (2015).
Backof, Thayer, et al. (2014) also use construal-level theory in their
research design while Gri?th et al. (2015) use mindset theory,
a theory very related to construal-level theory.
16
The results of
Gri?th et al. (2015) appear to complement the results of
my study in that the deliberative mindset (a mindset that
promotes broad thinking about a situation similar to high-
level construal thinking) appears to increase professional skep-
ticism of a client’s fair value estimate in a situation call-
ing for increased skepticism.
17
Similarly, the Gri?th et al.
study reveals that the implemental mindset (a mindset that
promotes speci?c thinking about a situation similar to low-
level construal thinking) appears to work against professional
skepticism.
The results of the Backof, Thayer, et al. (2014) study, how-
ever, appear on their face to con?ict with the results of my study.
Their study reveals that auditors considering evidence with low-
level construals display higher professional skepticism, particularly
when presented with evidence in graphical form. A possible rea-
son for the differences in results between our three studies can be
found in an examination of the experimental tasks.
All three of our studies charge the participants with evaluating
evidence regarding a fair value estimate put forth by management.
On closer examination there appears to be a distinct difference
in the scope of the evaluation performed by the auditor partici-
pants in the experimental tasks. The auditor participants in Backof,
Thayer, et al. (2014) focus speci?cally on two particular assump-
tions used by management in their discounted cash ?ow (DCF)
model: the location growth rate and the revenue growth rate.
The auditor participants in Gri?th et al. (2015) receive evi-
dence relating to four major assumptions of the DCF model: rev-
enue projections, expense projections, capital expenditures projec-
tions, and the discount rate. The evidence set used in this experi-
mental task is thus broader than the evidence set used in Backof,
Thayer, et al. (2014). This difference requires greater assimilation of
the evidence, particularly since the authors seeded important evi-
dence about certain assumptions in the evidence collected about
other assumptions. Importantly, the participants in both Gri?th
et al. (2015) and Backof, Thayer, et al. (2014) receive one set of
evidence to evaluate and do not have the ability to collect further
evidence.
The auditor participants in my study could access a much
broader set of evidence. While I include evidence about the DCF
16
I note that the description of the deliberative mindset written by Gri?th et al.
(2015) contains many similarities to descriptions of high-level construals noted in
psychology research (such as Trope & Liberman, 2010, 2003). For example, Gri?th
et al. (2015) describe that the deliberative mindset facilitates broad consideration
of evidence that is less focused on task speci?c details. Gri?th et al.’s (2015) de-
scription of the implemental mindset also appears to be similar to descriptions of
low-level construals. For example, the authors note that the implemental mindset
focuses on the c?howd? of a situation and prompts a narrow focus on the speci?cs
of a situation rather than the overall big picture. Both of these aspects are identical
to descriptions of low-level construals described in construal-level theory research
(Trope & Liberman, 2010, 2003).
17
Gri?th et al. (2015) seed incidental information into their evidence that sug-
gests that the client’s estimate is materially overstated.
model used by management, I also include other information re-
lated to management’s construction of the fair value estimate (see
Appendix A for a listing of the evidence). Additionally, my partic-
ipants only receive a subset of the available evidence and have to
decide whether to continue collecting evidence before rendering a
judgment on the estimate.
I consider the differences in audit evidence and judgment fo-
cus described above to parallel the differences between low-level
and high-level construals. The Backof, Thayer, et al. (2014) task has
a speci?c, detailed focus: the reasonableness of two individual as-
sumptions of the DCF model. This is a speci?c, detailed task calling
for the speci?c, detailed manner of thinking promoted by low-level
construals. Thus, instructions priming low-level construals would
work better in this setting.
The Gri?th et al. (2015) task has a broader, but still some-
what limited focus: the four major assumptions of the DCF model.
The participants are given evidence concerning each of the as-
sumptions, and critical information about one assumption has been
seeded in the evidence of another assumption. This broader task
and the need to assimilate information from different sources
of evidence suggests that high-level construals (or the deliber-
ative mindset) are better suited for the task setting. However,
the Gri?th et al. (2015) setting speci?cally focuses on the DCF
model rather than the totality of audit evidence that would be
considered by an auditor in this situation, including manage-
ment’s motivation and the level of expertise of the estimate’s pre-
parer, for example. The fact that this setting appears to strad-
dle the line between speci?c and broad thinking could explain
why Gri?th et al. (2015) ?nd certain inconsistencies in their
results.
18
My setting is the broadest of the three experimental tasks set-
ting and requires participants to consider evidence relating to sev-
eral different aspects of the estimate, including evidence relating
to the DCF model assumptions, management’s motivation, and the
experience level of the estimate’s preparer, among others. The very
broad nature of the evidence suggests that high-level construals of
the evidence would work best in terms of mental processing, and
that is what I document in this study.
Further, my participants do not receive a complete set of ev-
idence to consider, a situation which simulates the nature of
audit evidence collection in practice. Auditors receive an initial
set of evidence and then must determine whether the accumu-
lated evidence is su?cient to render a judgment or whether
more evidence must be collected (Gibbins, 1984; Knechel &
Messier, 1990). There is a greater need for assimilation and
abstraction of evidence in my study to support this determi-
nation since the auditors need to assess the big picture of
what the evidence reveals collectively rather than individually.
High-level construals support this initial determination and also
facilitate the assimilation of evidence collected in subsequent
rounds.
Conclusion
In this study, I investigate if audit documentation instructions
can help auditors construe (interpret) audit evidence in a man-
ner that enhances processing of the evidence and promotes pro-
fessional skepticism. My results indicate that auditors using docu-
mentation instructions that promote high-level construals demon-
18
For example, Gri?th et al. (2015) ?nd that there is no difference in target issues
with the estimate identi?ed by their mindset (non-control) participants. Also, the
authors ?nd that, based on a conventional statistics level, there is no difference
between mindset participants in decisions to contact the audit manager.
J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55 53
strate a higher level of professional skepticism. Further, relative to
auditors receiving documentation instructions that promote low-
level construals and auditors not receiving instructions, auditors
using high-level instructions are able to more effectively process
audit evidence, particularly incomplete sets of evidence.
My results also suggest that auditors in practice collect and pro-
cess information with low-level construals. Speci?cally, I ?nd no
signi?cant differences in the perceived risk of the estimate pro-
vided by users of low-level instructions and auditors not given
documentation instructions. Further, the auditors in these two
groups documented evidence in a manner consistent with low-
level construals of the audit evidence. According to several auditors
interviewed for this study, this problem could be due to a lack of
training in processing incomplete sets of evidence. My ?ndings can
thus be extremely bene?cial to current accounting practice. The
high-level instructions are easy to use, provide important bene?ts,
and will be inexpensive for ?rms to incorporate.
The ?ndings in this study also add to the current research
stream relating to construal-level theory and mindsets. Backof,
Thayer, et al. (2014) provide evidence that low-level constru-
als bene?t auditors when they must focus on speci?c details
of an audit of a complex estimate (two particular assumptions
used by management in their discounted cash ?ow analysis, for
example). Gri?th et al. (2015) demonstrate that a deliberative
mindset, akin to thinking with high-level construals, helps audi-
tors think more critically about their audits of complex estimate
when the auditors must broaden their focus to the entire dis-
counted cash ?ow model rather than just a few of the assump-
tions. The results of my study complement these ?ndings by show-
ing that high-level construals aid auditors in processing evidence
and maintaining professional skepticism in the broadest setting of
a complex estimate audit in which the auditors receive incom-
plete sets of evidence relating to matters beyond just manage-
ment’s assumptions. Considering the results of all three studies as
well as key aspects of construal-level theory, I infer that broader
tasks bene?t from high-level construals while speci?c tasks ben-
e?t from low-level construals. Future research can test this
conjecture.
Every experimental study is subject to limitations, but these
limitations often represent excellent opportunities for future re-
search. One limitation is that the study does not directly test the
conjecture in the previous section that task complexity is neg-
atively related with professional skepticism. More di?cult tasks
could actually induce higher skeptical action due to the increase
in focus required by the higher di?culty. On the other hand, as
suggested by the results in this study, higher complexity could
cause an auditor to prematurely conclude a task, such as searching
for audit evidence, simply because the task is cognitively draining.
Such an end to the task could be later interpreted that the auditor
failed to display an appropriate amount of professional skepticism.
Another question that could not be answered by this study
is whether the high-level instructions increases skeptical judg-
ments and actions when such an increase is not warranted. For
example, future research could test for increases in skeptical judg-
ments and actions in a situation where the client’s estimate is
fairly presented. The high-level instructions could be deemed in-
e?cient if there are increases in skepticism and the increases
are to the level of those displayed in the current study, although
this ine?ciency would still be consistent with Nelson’s (2009)
c?presumptive doubtd? notion of professional skepticism.
The documentation instructions provided to the auditors in
this experiment asked them to consider reasons both for and
against the reasonableness of the estimate. Considering both the
pros and the cons is consistent with current audit practice and
with recommendations of the KPMG judgment framework (KPMG,
2011). However, considering both sides could potentially balance
the competing effects of the positive and negative information. Fu-
ture research can examine different wording of the instructions.
I included a control condition in which participants did not re-
ceive documentation instructions and therefore were not explicitly
asked to consider how or why the fair value estimate was either
fairly presented or materially misstated. The control condition also
differed from the experimental conditions in that their documenta-
tion occurred concurrently with their judgments rather than hav-
ing them document after each round of evidence collection. These
issues limit inferences that can be drawn from comparisons of the
control group with the experimental groups.
Finally, the experimental instrument used to test the hypothe-
ses is based on aspects of construal-level theory. Construal-level
theory espouses that high-level construals can be primed by hav-
ing a participant think broadly and about c?whyd? something hap-
pens while low-level construals can be primed by having a par-
ticipant think speci?cally and about c?howd? something happens.
Neither this study nor any construal-level theory studies, to my
knowledge, explore whether the how vs. why aspect of the prime
or the broad vs. speci?c aspect of the prime is more powerful.
19
Future research can help determine if one of these aspects domi-
nates the other or whether combining the aspects provide additive
or multiplicative power to the manipulation.
Acknowledgements
I thank my dissertation committee: Uday Murthy (chair), Lisa
Gaynor, Mark Mellon, and Joe Vandello. This manuscript bene?ted
from excellent comments provided by audience members at the
2014 Accounting, Organizations & Society Conference on Account-
ing Estimates, the editors of the conference (Lisa Koonce, Mark
Peecher, and Hun Tong Tan) and the anonymous reviewers. Special
thanks to Deloitte LLP for its sponsorship of the conference and
its process. I also thank Stephen Fuller, Erin Hamilton, Bob Libby,
Rina Limor, Ivo Tafkov, and participants from workshops at Ap-
palachian State University, College of Charleston, Georgia State Uni-
versity, Florida Atlantic University, Kent State University, the Uni-
versity of South Carolina, and the University of South Florida. Fi-
nally, I thank all of my auditor participants and their respective
audit ?rms for allowing me a portion of their valuable time.
Appendix A. – Evidence provided to participants
Evidence Screen 1 (2 positive, 4 neutral, 4 negative):
1. You test management’s mathematical calculations and formulas.
You do not discover any errors.
2. The use of fair value for the reacquired franchise rights account
is appropriate and conforms with GAAP.
3. APC produces pizzas of higher quality than most other pizza
franchises.
4. APC’s controller produced the fair value estimate.
5. There is a lack of objective data that can be used when calcu-
lating the estimate for the reacquired franchise rights account.
6. You interview some of American Pizza Company’s lower level
accountants, but they have no knowledge of how the reacquired
franchise rights account is prepared.
19
That said, there are psychology studies that only use the how vs. why aspect
in their manipulations with success (Fujita et al. (2006) and Freitas, Gollwitzer, and
Trope (2004), for example).
54 J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55
7. You spend time preparing a benchmark model to compare the
output with management’s estimate. Management’s estimate
appears at the high end of your reasonable range, but any dif-
ferences could be attributed to fundamental differences in the
models used.
8. The demand for high quality pizza is decreasing due to the cur-
rent economic environment. Although APC produces pizzas of
generally higher quality than most pizza chains, the company’s
pizza product would be considered on the low end of the high
quality pizza scale.
9. Management’s estimate appears to be moderately to highly
sensitive to changes in the basic assumptions used to calculate
the estimate.
10. You compile a report which suggests that Pennsylvania could
be saturated with pizza restaurants.
Evidence Screen 2 (1 positive, 2 neutral, 2 negative):
1. The model used by management is the same model used by
other companies in the pizza franchise industry and is consis-
tent with the model used by APC’s main competitors.
2. You determine that there is only a moderate risk associated
with the expected future cash ?ows related to the reacquired
franchise rights account.
3. You review the minutes of APC’s last two board meetings but
do not ?nd any mention of management’s intentions relative to
the reacquired franchise rights account.
4. You ask a specialist in your ?rm to prepare a benchmark esti-
mate. His estimate is lower than management’s fair value es-
timate. However, the specialist states that the difference could
be due to circumstances unique to the company.
5. Restaurant franchises similar to American Pizza Company have
been growing at a rate of 3.5 restaurants per year.
Evidence Screen 3 (1 positive, 2 neutral, 2 negative):
1. The model used by management is consistent with what com-
panies in other industries use when estimating reacquired fran-
chise rights. The model appears to be the standard model used
by over 95% of companies when estimating reacquired franchise
rights.
2. American Pizza Company focuses on a dine-in experience as
opposed to the traditional pizza delivery business.
3. The increase in fuel costs has caused an increase in the expense
ratio of other pizza companies, but APC focuses on creating
dine-in restaurants. High fuel costs do not signi?cantly affect
APC’s expense ratio.
4. A detailed examination of American Pizza Company’s history
shows that the company meets its target growth rate approx-
imately 80% of the time, but it growth rate during the other
20% of the time is only two restaurants per year.
5. The person who prepared APC’s fair value estimate does not
have prior experience in preparing fair value estimates.
Evidence Screen 4 (1 positive, 2 neutral, 2 negative):
1. As of the date of your review, APC appears to have the eco-
nomic ability to follow through with its planned restaurant
growth.
2. The CFO of APC tells you about the company’s plans relative to
reacquired franchise rights. The plan appears to be consistent
with the assumptions used in management’s estimation model.
3. American Pizza Company’s marketing expenses have not in-
creased signi?cantly over the past three years.
4. The management of APC did not consider using a third party to
prepare the estimate.
5. An interview with the person who prepared the fair value esti-
mate indicated that the CEO of APC expressed a desire for the
estimate to be as high as possible. The preparer stated that no
one exerted undue in?uence on him, though, and the preparer
does not have any incentives to overstate the estimate.
Evidence Screen 5 (1 positive, 2 neutral, 2 negative):
1. Upon reviewing budgets set by APC’s management, you dis-
cover evidence that APC plans to follow through with its
planned restaurant growth.
2. APC’s pizza receives rave reviews from its customers.
3. Hungry Howie’s is closing down most of its restaurants in
Pennsylvania, but Hungry Howie’s is not in the same pizza mar-
ket segment as American Pizza Company. The closure of the
restaurants should not signi?cantly impact APC’s business in
Pennsylvania.
4. Management has used a different model in the past when
preparing the estimate for its reacquired franchise rights ac-
count. The estimate for this account would be lower when us-
ing the old model. Management justi?ed the change by saying
that the model used this year is the model used by the majority
of its competitors.
5. The discount rate used by management could be a little low
given the current economic conditions. The use of a lower rate
could lead to management’s estimate being slightly higher than
would otherwise be expected.
References
Backof, A. G., Bamber, E. M., & Carpenter, T. D. (2014). Do auditor judgment frame-
works help in constraining aggressive reporting? Evidence under more precise and
less precise accounting standards. Working paper: University of Georgia and Uni-
versity of Virginia.
Backof, A. G., Thayer, J., & Carpenter, T. D. (2014b). Auditing complex estimates:
Management-provided evidence and auditors’ consideration of inconsistent evi-
dence. Working paper, University of Virginia & University of Georgia.
Bratten, B., Gaynor, L. M., McDaniel, L., Montague, N. R., & Sierra, G. E. (2013). The
audit of fair values and other estimates: The effects of underlying environmen-
tal, task, and auditor-speci?c factors. Auditing: A Journal of Practice & Theory,
32(Supplement 1), 7–44.
Brewster, B. E. (2011). How a systems perspective improves knowledge acquisition
and performance in analytical procedures. The Accounting Review, 86(3), 915–
943.
Freitas, A. L., Gollwitzer, P., & Trope, Y. (2004). The in?uence of abstract and concrete
mindsets on anticipating and guiding others’ self-regulatory efforts. Journal of
Experimental Social Psychology, 40, 739–752.
Fujita, K., Trope, Y., Liberman, N., & Levin-Sagi, M. (2006). Construal levels and self-
control. Journal of Personality and Social Psychology, 90(3), 351–367.
Gibbins, M. (1984). Propositions about the psychology of professional judgment in
public accounting. Journal of Accounting Research, 22(1), 103–125.
Glover, S. M., & Prawitt, D. F. (2013). Enhancing auditor professional skepticism. .
Gollwitzer, P. M. (1990). Action phases and mind-sets. In E. T. Higgins, & R. M. Sor-
rentino (Eds.), Handbook of motivation and cognition: Foundations of social behav-
ior (pp. 53–92). New York: Guilford Press.
Gri?th, E. E., Hammersley, J. S., & Kadous, K. (in press). Audits of complex es-
timates as veri?cation of management numbers: How institutional pressures
shape practice. Contemporary Accounting Research.
Gri?th, E. E., Hammersley, J. S., Kadous, K., & Young, D. (2015). Auditor mindsets
and audits of complex estimates. Journal of Accounting Research, 53(1), 49–77.
Hurtt, R. K. (2010). Development of a scale to measure professional skepticism. Au-
diting: A Journal of Practice & Theory, 29(1), 149–171.
Hurtt, R. K., Brown-Libured, H., Earley, C. E., & Krishnamoorthy, G. (2013). Research
on auditor professional skepticism: Literature synthesis and opportunities for
future research. Auditing: A Journal of Practice & Theory, 32(Supplement 1), 45–
97.
International Forum of Independent Audit Regulators (IFIAR), (2012). Summary
report of inspection ?ndings. Available at: .
Knechel, W. R., & Messier, W. F., Jr. (1990). Sequential auditor decision making: In-
formation search and evidence evaluation. Contemporary Accounting Research,
6(2), 386–406.
J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55 55
Kohlbeck, M. J., Cohen, J. R., & Holder-Webb, L. L. (2009). Auditing intangible assets
and evaluating fair market value: The case of reacquired franchise rights. Issues
in Accounting Education, 24(1), 45–61.
KPMG (2011). Elevating professional judgment in auditing and accounting: The
KPMG professional judgment framework. .
Nelson, M. W. (2009). A model and literature review of professional skepticism in
auditing. Auditing: A Journal of Practice & Theory, 28(2), 1–34.
Public Company Accounting Oversight Board (PCAOB) (2008). Report on the
PCAOB’s 2004, 2005, 2006, and 2007 inspections of domestic annually in-
spected ?rms. PCAOB Release No. 2007-008. .
PCAOB (2012). Maintaining and applying professional skepticism in audits. PCAOB Staff
Audit Practice Alert No. 10. December 4, 2012. Washington, D.C.: PCAOB.
PCAOB (2014). Due professional care in the performance of work. AU Section 230.
.
Plumlee, D., Rixom, B. A., & Rosman, A. J. (2015). Training auditors to solve ill-
structured tasks using metacognitive skills. The Accounting Review, 90(1), 351–
369.
Rim, S., Uleman, J. S., & Trope, Y. (2009). Spontaneous trait inference and construal
level theory: Psychological distance increases nonconscious trait thinking. Jour-
nal of Experimental Social Psychology, 45, 1088–1097.
Trope, Y., & Liberman, N. (2003). Temporal construal. Psychological Review, 110(3),
403–421.
Trope, Y., & Liberman, N. (2010). Construal-level theory of psychological distance.
Psychological Review, 117(2), 440–463.
doc_352638728.pdf
Thisstudyinvestigatestheuseofauditevidencedocumentationinstructionsthatpromotethecollec-tionandprocessingofevidencewithhigh-levelconstruals(broad,abstractinterpretationsoftheevi-dence).Abstractioncanhelpapersonpiecetogetherindividualpiecesofinformationorevidenceandbetterenableapersontoseethebigpictureofwhatthecllectiveinformationportends.Theresultsofanexperimentsuggestthatauditorsthinkandactwithmoreprofessionalskepticismwhenusingthedocumentationinstructionsthatprmotehigh-levelconstrualsascomparedwithauditorsusingdocu-mentationinstructionspromotinglow-levelconstruals(specific,detailedinterpretationsoftheevidence,akintocurrentauditpractice)andwithauditorsnotgivendocumentationinstructions.Further,thehigh-levelconstrualsfosterbetterprocessingofthecollectedevidence.
Accounting, Organizations and Society 46 (2015) 44–55
Contents lists available at ScienceDirect
Accounting, Organizations and Society
journal homepage: www.elsevier.com/locate/aos
Construal instructions and professional skepticism in evaluating
complex estimates
Jason Tyler Rasso
College of Charleston, United States
a r t i c l e i n f o
Article history:
Received 1 June 2014
Revised 16 March 2015
Accepted 17 March 2015
Available online 11 April 2015
a b s t r a c t
This study investigates the use of audit evidence documentation instructions that promote the collec-
tion and processing of evidence with high-level construals (broad, abstract interpretations of the evi-
dence). Abstraction can help a person piece together individual pieces of information or evidence and
better enable a person to see the big picture of what the collective information portends. The results
of an experiment suggest that auditors think and act with more professional skepticism when using the
documentation instructions that promote high-level construals as compared with auditors using docu-
mentation instructions promoting low-level construals (speci?c, detailed interpretations of the evidence,
akin to current audit practice) and with auditors not given documentation instructions. Further, the high-
level construals foster better processing of the collected evidence. The study also provides preliminary
evidence that task complexity could interfere with professional skepticism.
© 2015 Elsevier Ltd. All rights reserved.
Introduction
Regulators have recently criticized auditors for failing to dis-
play a proper amount of professional skepticism in their audits of
complex estimates (Bratten, Gaynor, McDaniel, Montague, & Sierra,
2013; IFIAR, 2012; PCAOB, 2012). One possible explanation for
the lack of professional skepticism is that audits of complex es-
timates require the processing of numerous pieces of audit evi-
dence collected over an extended period of time. This study ex-
plores whether and how construals (interpretations) of evidence
affect an auditor’s judgments and decisions relating to the collec-
tion and processing of audit evidence, particularly when all of the
evidence is not yet available.
Speci?cally, this paper examines whether documentation in-
structions that promote broad, abstract interpretations (high-level
construals) of the audit evidence can be helpful in promoting ef-
fective processing of information throughout the evidence collec-
tion process. Studying the process of evidence collection is im-
portant because a central component of professional skepticism
is the suspension of judgment until su?cient and competent ev-
idence has been obtained (AU 230.07-.08; Hurtt, 2010). Auditing
complex estimates, like many other audit tasks, involves a sequen-
tial judgment process through which auditors receive information
and then need to make important decisions based on that informa-
E-mail address: [email protected]
tion such as whether to continue collecting new evidence (Gibbins,
1984; Knechel & Messier, 1990). Failure to properly process evi-
dence, particularly early in the evidence collection process, could
lead to poor judgments such as a decision to end the evidence col-
lection process prematurely (which would be judged ex post as a
lack of professional skepticism). As support, experienced auditors
recently reported that they have trouble processing and assimi-
lating collected audit evidence, particularly negative evidence that
ought to increase skepticism and suggest to an auditor that evi-
dence collection should continue (Gri?th, Hammersley, & Kadous,
in press).
Construal-level theory (CLT) suggests that the manner in which
individuals construe (interpret) information and evidence impacts
their judgments and decisions. The theory describes two levels
of construals: low-level construals and high-level construals. Low-
level construals are speci?c and detailed while high-level constru-
als are broad and abstract (Trope & Liberman, 2003). Fundamental
to this study, one of the primary differences between construal lev-
els is that high-level construals make it easier for an individual to
process and assimilate numerous pieces of information, including
negative information, while low-level construals make such a task
more di?cult (Trope & Liberman, 2010, 2003).
Based on evidence from audit practice, I argue that auditors are
likely processing audit evidence with low-level construals. As de-
scribed previously, experienced auditors report having trouble pro-
cessing and assimilating audit evidence. These auditors describehttp://dx.doi.org/10.1016/j.aos.2015.03.003
0361-3682/© 2015 Elsevier Ltd. All rights reserved.
J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55 45
having trouble “seeing the big picture” surrounding the estimate
as well as having di?culty detecting patterns when given all of
the available information (Gri?th et al., in press). If an auditor
does not detect a problem with an estimate (because of a failure
to properly process negative evidence, for example) then the col-
lection of audit evidence phase of the audit could be terminated
too quickly, an act consistent with lower professional skepticism.
I posit that documentation instructions that promote high-level
construals have the potential to overcome these di?culties because
this level of construal facilitates better processing and assimila-
tion of all information, both positive and negative, which should
allow an auditor to properly determine whether the amount of ev-
idence collected is su?cient to render a judgment about the ap-
propriateness of the estimate (Trope & Liberman, 2010, 2003). To
test this prediction, I employ a 1 × 3 between-participants experi-
ment using auditors experienced in auditing complex estimates as
participants. One group received documentation instructions which
prime high-level construal thinking while another group received
documentation instructions that prime low-level construal think-
ing. A third group of participants did not receive documentation
instructions and served as a control condition.
In a setting calling for increased professional skepticism,
the participants performed a simulated fair value audit task
in which they collected and reviewed audit evidence that col-
lectively suggests the client’s estimate is on the high end of
a reasonable range (suggesting that the estimate is aggres-
sive). I capture multiple measures of professional skepticism be-
cause of the lack of consensus in the construct’s de?nition
(Hurtt, 2010; Hurtt, Brown-Libured, Earley, & Krishnamoorthy,
2013; Nelson, 2009). I predict that auditors using the documen-
tation instructions that promote high-level construals will dis-
play the highest level of professional skepticism and will ulti-
mately produce the highest judgments of risk involved with the
estimate.
Consistent with the sequential process of auditing (Gibbins,
1984; Knechel & Messier, 1990), the participants collect evidence
about the fair value estimate and make decisions concerning
whether they collected su?cient evidence to render judgments
about the estimate or need to continue gathering more evidence.
Participants given the high-level construal instructions document
the collected evidence while considering the evidence broadly and
why the estimate could be fairly presented or materially misstated.
Participants using the low-level construal instructions document
evidence while considering the evidence speci?cally and how the
estimate could be fairly presented or materially misstated. Partici-
pants in the control condition do not receive instructions and only
document evidence at the end of the collection process.
I ?nd that auditors given documentation instructions
priming high-level construals display higher levels of both
skeptical judgments and skeptical actions. Auditors given
these instructions spent signi?cantly more time on the
task, particularly in the collection and evaluation of au-
dit evidence phase of the task. Auditors using the high-
level construal instructions also collected signi?cantly more
evidence than did auditors using the low-level construal
instructions.
Furthermore, auditors using the high-level construal instruc-
tions rated the fair value estimate as signi?cantly more risky than
did their counterpart auditors in the other two conditions. Note
that the increases in skepticism do not result merely from the col-
lection of greater amounts of evidence. There is no difference in
skeptical judgments between users of low-level instructions and
the control group; however, the control group collected statisti-
cally the same amount of evidence as users of the high-level in-
structions. Interestingly, the use of the high-level construal instruc-
tions appears to have moderately increased perceptions of task
di?culty.
This study contributes to existing literature streams and an-
swers several calls for research. I extend the literature on auditing
and construal-level theory by showing that considering audit evi-
dence through high-level construals leads to enhanced professional
skepticism. This ?nding adds to the growing stream of literature
which suggests that interventions can be used to improve how au-
ditors cognitively process audit information (Brewster, 2011; Plum-
lee, Rixom, & Rosman, 2015).
Concurrent research documents alternative mechanisms that
help promote effective processing of information, including the de-
liberative mindset (Gri?th, Hammersley, Kadous, & Young, 2015)
and convergent and divergent thinking (Plumlee et al., 2015).
Backof, Thayer, and Carpenter (2014) also investigate the bene?ts
of construal-level theory when auditing complex estimates. These
studies, however, only show more effective processing of infor-
mation when auditors are given a relatively complete set of ev-
idence.
1
An important aspect of professional skepticism and of
audits of complex estimates in general is the collection of evi-
dence (AU 230.07-.08; Hurtt, 2010). My study shows that auditors
struggle with processing incomplete sets of evidence. This prob-
lem manifests as a lack of professional skepticism when the au-
ditor prematurely concludes the evidence collection phase of the
audit and potentially makes poor judgments based on the incom-
plete evidence. An important ?nding of my study is that audi-
tors using the high-level construal instructions can better pro-
cess incomplete sets of evidence and properly recognize that
they need more evidence in order to make a conclusion. Fur-
ther, when provided with the same amount of evidence, audi-
tors using the high-level construal instructions are more likely
to recognize the high risk level of a complex estimate when
the majority of the evidence suggests the estimate could be
overstated.
This study also contributes to practice by presenting an eas-
ily implementable technique that can be used to promote effective
processing of audit evidence. The documentation instructions that
prime high-level construals are simple to use, inexpensive, and can
easily be tailored for a ?rm’s speci?c needs or language. The in-
structions can be used to help overcome at least one of the de-
?ciencies noted by regulators in their inspection reports, namely
the lack of perceived professional skepticism (PCAOB, 2012, 2008).
Consequently, regulators should also ?nd these results to be infor-
mative.
The remainder of this paper is organized as follows. The next
section describes the problem, reviews the literature relevant to
this study’s setting, and develops the hypotheses. Next, I discuss
the method used to examine the hypotheses. The fourth section
presents the results, and the ?nal section offers a discussion of the
conclusions drawn from the study.
Literature review and hypothesis development
Audits of complex estimates and professional skepticism
One of the many di?culties involved in audits of complex es-
timates is that the audit takes place over the course of time. Au-
1
The Backof, Thayer, et al. (2014) study also differs from the current study
because it focuses on the reasonableness of management’s assumptions. The
focus on one particular aspect of the complex estimate is likely the rea-
son why their results differ from those in this study. Low-level constru-
als would aid an auditor focused on one aspect of the audit while high-
level construals would aid an auditor in assimilating evidence, both complete
sets and incomplete sets, and better see the “big picture” surrounding the
estimate. I discuss these differences in more detail in the ?nal section of
this paper.
46 J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55
dit evidence is collected piecemeal over that time rather than all
at once. Auditors interviewed by Gri?th et al. (in press) confess to
having di?culty processing and assimilating audit evidence, partic-
ularly negative evidence, and this problem can only be worse when
the auditors do not have the complete set of evidence available to
them.
Due to the sequential process of auditing, judgments are made
at several points during an audit as evidence is collected, including
the decision whether to continue or terminate the evidence collec-
tion process (Gibbins, 1984; Knechel & Messier, 1990). One of the
consequences of the problem noted above is that a failure to ad-
equately respond to negative evidence can result in a premature
termination of the evidence collection process. An auditor could
fail to see that an estimate is aggressive (as stated by some of the
auditors in Gri?th et al. (in press)) and decide that the evidence
collected so far is su?cient to conclude that the estimate is rea-
sonable. The suspension of judgment until an auditor obtains an
appropriate level of evidence is a key aspect of professional skepti-
cism (Hurtt, 2010; Nelson, 2009; PCAOB, 2014; AU 230.09).
Regulators, having the bene?t of a fuller set of evidence when
inspecting the audit months or years later, would likely conclude
that the auditor did not display su?cient professional skepticism
because the auditor failed to obtain an appropriate level of evi-
dence. As evidence, U.S. and international audit inspectors cite a
lack of professional skepticism as a primary reason for audit de?-
ciencies (IFIAR, 2012; PCAOB, 2012, 2008). In particular, a summary
report details that “In some instances, ?rms did not su?ciently
test or challenge management’s forecasts, views, or representations
that constituted critical support for amounts recorded in the ?nan-
cial statements” (PCAOB, 2008, p. 20). I posit that documentation
instructions employing the precepts of construal-level theory can
promote skepticism and enhance processing of evidence when us-
ing evidence sets at various levels of completion.
Construal-level theory
Construal-level theory (CLT) makes predictions of how
individuals construe, or interpret, information. Researchers
have used this theory to describe how individuals make
predictions and evaluate situations, how gambling prefer-
ences are affected, and how construals can affect interper-
sonal negotiation (Trope & Liberman, 2010, 2003). The level
of construal through which a person processes informa-
tion shapes how the information is encoded in a person’s
mind.
There are two levels of construals identi?ed by CLT: high-level
construals and low-level construals. Viewing information through
high-level construals makes information more abstract. This ab-
straction allows for easier assimilation of information because the
representations become less ambiguous, more coherent, and more
schematic (Trope & Liberman, 2010). High-level construals focus on
why something has been done (Fujita, Trope, Liberman, & Levin-
Sagi, 2006; Trope & Liberman, 2010, 2003). Low-level construals
are more detailed, transactional interpretations of information that
focus on how something has been done (Fujita et al., 2006; Trope
& Liberman, 2010, 2003).
To illustrate the differences in levels, consider the follow-
ing example of a jigsaw puzzle. A person viewing the puz-
zle pieces with high-level construals loses some of the spe-
ci?c details of the pieces (colors, shapes, end vs. middle piece,
etc.), but can better see how the pieces of the puzzle ?t to-
gether. In contrast, a person viewing the puzzle pieces with
low-level construals can likely tell you speci?c details about
many of the pieces, but the detailed level of focus restricts the
person from being able to see how the individual pieces ?t
together.
2
Interviews with experienced auditors suggest that auditors
could be using low-level construals when processing audit evi-
dence. The interviewees report having problems assimilating audit
evidence and seeing the c?big pictured? or the overall pattern sug-
gested by the evidence (Gri?th et al., in press). Due to the piece-
meal nature of the evidence collection process during the audit of
a complex estimate, auditors are likely processing and recording
each evidence item speci?cally as the evidence is received. This act
would be consistent with processing the evidence with low-level
construals, particularly if the auditors do not take the time to con-
sider how previously received evidence relates to the evidence just
received.
Audit plans and speci?c ?rm guidance/policy on how to process
and document audit evidence are generally not made available to
the public for reference. One recent exception is KPMG’s profes-
sional judgment framework. In this framework, KPMG recognizes
the critical role of gathering and evaluating information and es-
pouses the importance of seeing the big picture of information;
however, the framework does not reference the problem associ-
ated with collecting evidence over time nor does it offer sugges-
tions for how to effectively process previously obtained evidence
(KPMG, 2011).
Due to the lack of proper references concerning this issue, I
interviewed seven auditors concerning the practice of collecting
and evaluating audit evidence over time. These auditors averaged
8.14 years of audit experience and each had experience with au-
diting complex estimates. The auditors represented a broad cross-
section of ?rm type, ranging from Big 4 auditors (three), auditors
from national ?rms (two), and auditors from local or regional ?rms
(two).
Only one of the seven auditors mentioned that his or her ?rm
provided guidance on how to assimilate previously received evi-
dence with evidence currently received. The remaining six audi-
tors shared similar experiences of collecting and recording evi-
dence items individually as they were received, a manner consis-
tent with processing the evidence with low-level construals. None
of the seven auditors interviewed reported using any sort of rou-
tine that would resemble processing evidence, either currently or
previously received, with high-level construals.
Processing evidence with high-level construals can increase ef-
fective processing of information. In an auditing context, the ab-
straction facilitated by high-level construals makes the collected
audit evidence easier to assimilate and easier for an auditor to rec-
ognize patterns in the information (Trope & Liberman, 2010).
3
The
abstraction and subsequent assimilation of the collected evidence
will allow auditors to recognize the need for increased professional
skepticism if the evidence contains items suggesting something
unusual with the estimate (negative or discon?rming evidence).
For example, auditors could realize that the estimate is aggressive
and respond by collecting more evidence, concluding that the es-
timate is risky and making appropriate adjustments to the audit
plan. Thus, processing evidence with high-level construals should
help overcome the problems described above resulting from the
2
One can also imagine a person with his face low to the puzzle. This person can
see the details of the speci?c pieces; however, it’s not until he raises his face to a
higher level that he can see how the pieces ?t together.
3
As a colloquial example of how this process works, imagine ?ve very large let-
ters in front of you that spell a word. Each letter individually takes up your entire
?eld of view. By creating distance from the letters, you are better able to see the
entire word rather than just the individual letters.
J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55 47
piecemeal evidence collection process. Conversely, low-level con-
struals allow individual elements of evidence to dominate others
to the extent that these elements block an individual’s ability to
see the totality of what the evidence suggests (Trope & Liberman,
2010).
The hypotheses described below test the proposed advantages
of processing audit evidence with high-level construals.
4
H1. Auditors using documentation instructions that prime high-
level construals will display higher professional skepticism than
will auditors using documentation instructions that prime low-
level construals.
H2. Auditors using documentation instructions that prime high-
level construals will display higher professional skepticism than
will auditors not given documentation instructions.
Method
Participants
Fifty-eight auditors experienced in auditing complex estimates
and representing six accounting ?rms participated in the study.
Table 1 displays information about these participants. The audi-
tor participants averaged 5.4 years of audit work experience and
ranged in rank from staff auditor to partner. Ninety percent of the
participants had experience auditing fair value estimates.
5
Design
I use a 1 × 3 between-participants experimental design to
test my hypotheses. The independent variable in this experiment
is the type of documentation instructions given to participants,
which comprises three levels: high-level construal documentation
instructions (hereafter referred to as high-level instructions for
brevity), low-level construal documentation instructions (hereafter
referred to as low-level instructions), or no documentation instruc-
tions. The computerized experimental instrument randomly as-
signed participants to one of the three experimental conditions.
4
Although construal-level theory maps on to the speci?c features of the audit
setting described in this section, there are other related ideas and theories that
generally support the hypotheses. For example, Trope and Liberman (2010) also dis-
cuss the idea of psychological distance, the distance that an individual is removed
from something either spatially, temporally, or mentally. High-level construals can
increase psychological distance which, in an audit setting, can increase the mental
(psychological) distance between an auditor and the evidence. The increased dis-
tance could lead to less biased processing which would result in higher profes-
sional skepticism. Further, mindset theory appears to support the hypotheses. For
example, processing evidence with a deliberative mindset can be bene?cial because
individuals in this mindset tend to think more critically about the information pre-
sented to them (Gri?th et al., 2015) and tend to remain impartial and objective
longer (Gollwitzer, 1990).
5
I collected responses from the auditors in person whenever possible. However,
28 participants (48.3% of the participants) completed the study on their own time
due to ?rm requests or logistical problems. I found signi?cant differences in gen-
eral work experience and the type of ?rm that employed each participant, but the
inclusion of these two variables or a testing location variable does not affect the re-
sults reported later in this paper. The differences in ?rm type can be explained by
the fact that all the participants from local, regional, and national accounting ?rms
completed the study in a controlled setting while only 35% of the participants from
international ?rms completed the study in a controlled setting. The differences in
work experience is likely due to outliers in the variables resulting from two part-
ners with work experience of 25 and 30 years, respectively completing the study in
a controlled setting. No other potential covariates, including scores from the Hurtt
(2010) trait professional skepticism scale, are signi?cant.
Task
I use a case adapted from Kohlbeck, Cohen, and Holder-Webb
(2009) which involves an audit of a client’s intangible asset ac-
count (reacquired franchise rights). Each participant began the ex-
periment by reviewing background and ?nancial information about
the client, American Pizza Company (APC). The information in-
cluded a description of the company and its franchising activities.
A table presented information about the client’s ?nancial state-
ments, including the value of its reacquired franchise rights ac-
count (an intangible asset recorded at fair value). Participants also
read information about the client’s accounting policies related to
reacquired franchise rights.
The next screen contained information relevant to the experi-
mental task. Each participant read that their ?rm has audited APC
for over ten years and that the client is a signi?cant source of rev-
enue for the ?rm. The participants considered speci?c information
about the re-acquired franchise rights account and learned that the
fair value of the account was greater than the book value of the
account. The current screen also displayed management’s assump-
tions used in generating the fair value estimate.
After reviewing the client and task information, the partici-
pants began to collect evidence relating to the task. I created thirty
pieces of evidence that related to the client and/or the client’s
reacquired franchise rights accounts. In order to create a setting
that called for increased professional skepticism, I generated 12 ev-
idence items which collectively suggested that the client’s estimate
was aggressive (at the high end of a reasonable range of the esti-
mate). I also generated six evidence items which collectively sug-
gested that the client’s estimate was fairly stated and 12 neutral
evidence items which were largely irrelevant to the task.
6
In order to reinforce the fact that time spent collecting audit
evidence is not costless, the computerized instrument displayed a
time budget to all participants. The use of a time budget in this
task is consistent with audit practice. On the ?rst screen of evi-
dence, each participant learned that 70% of the current audit bud-
get has been expended to generate the initial round of evidence.
After learning the percent of audit hours used, each participant
read the following statement:
Your audit manager will look favorably upon you if you com-
plete the fair value audit using as few audit hours as possible;
however, you should continue searching for evidence until you
have satis?ed yourself that you have obtained su?cient appro-
priate evidential matter to provide you with a reasonable basis
for forming an opinion.
Each successive round of evidence collection incremented the
percentage used by 10%. The initial collection given to each par-
ticipant contained the same 10 evidence items (four negative, four
neutral, and two positive, all randomly displayed). Each participant
had the ability to continue collecting evidence up to four times
with each subsequent collection revealing ?ve new evidence items
in the same ratio of two negative, two neutral, and one positive
items each.
7
Participants using all available rounds of collection
6
I gave the list of evidence to several current and former auditors who agreed
that each item of evidence was appropriately labeled as either positive, negative, or
neutral evidence. I also gave an unlabeled, alphabetized list to three audit mangers
who rated the strength of each evidence item on a scale of ?5 (Extremely nega-
tive) to +5 (Extremely positive), with 0 labeled as neutral. The average strength of
the negative evidence was ?2.53 (s.d. = 1.73). The average strength of the positive
evidence was 2.17 (s.d. = 1.82) while the average strength of the neutral evidence
was ?0.08 (s.d. = 1.44). These ratings by the audit managers provide support for
the claimed directionality of the evidence as being either positive, negative, or neu-
tral.
7
Across all experimental conditions, each round of collection provided the same
48 J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55
Table 1
Participant demographic statistics.
High-level
instructions condition
mean (s.d. or percent
of sample) (n = 19)
Low-level
instructions condition
mean (s.d. or percent
of sample) (n = 19)
Control condition
mean (s.d. or percent
of sample) (n = 20)
Total mean (s.d. or
percent of sample) (n
= 58)
One-way ANOVA
p-value
Gender 0.529
Male 10 (52.6%) 11 (57.9%) 8 (40.0%) 29 (50.0%)
Female 9 (47.4%) 8 (42.1%) 12 (60.0%) 29 (50.0%)
Age 28.05 (4.034) 29.89 (4.267) 30.70 (5.823) 29.57 (4.842) 0.221
General work experience (years) 7.11 (5.322) 8.42 (6.086) 9.35 (6.930) 8.31 (6.125) 0.526
Audit work experience (years) 4.63 (3.095) 5.26 (4.107) 6.25 (5.077) 5.40 (4.171) 0.481
Experience auditing fair value estimates 16 (84%) 17 (89%) 19 (95%) 52 (90%) 0.556
Hurtt professional skepticism score 131.74 (16.251) 124.05 (16.092) 133.15 (17.942) 129.71 (16.990) 0.204
Auditor rank 0.496
Staff 3 (15.8%) 2 (10.5%) 3 (15.0%) 8 (13.8%)
Senior 13 (68.4%) 12 (63.2%) 10 (50.0%) 35 (60.3%)
Manager 3 (15.8%) 4 (21.1%) 6 (30.0%) 13 (22.4%)
Partner 0 (0.0%) 1 (5.2%) 1 (5.0%) 2 (3.5%)
Firm type 0.923
Local 1 (5.3%) 1 (5.3%) 2 (10.0%) 4 (6.9%)
Regional 3 (15.8%) 3 (15.8%) 2 (10.0%) 8 (13.8%)
National 0 (0.0%) 1 (5.3%) 2 (10.0%) 3 (5.2%)
International 15 (78.9%) 14 (73.7%) 14 (70.0%) 43 (74.1%)
went over budget by 10% (used 110% of the audit hours allocated
to the task by the audit manager). The computerized instrument
randomized the order of the evidence items in all conditions.
After each round of collection, participants given documenta-
tion instructions responded to the questions designed to invoke
a distinct construal-level. Participants given the high-level instruc-
tions responded to the following two questions when considering
the audit evidence:
c?Thinking broadly about all of the evidence collectively, list
reasons why management’s estimate could be fairly presented.
Thinking broadly about all of the evidence collectively, list
reasons why management’s estimate could be materially mis-
stated.d?
I drew the wording for this condition from prior literature
which shows that thinking broadly about a situation and why
the situation occurred invokes high-level construals of informa-
tion (Backof, Bamber, & Carpenter, 2014; Fujita et al., 2006; Rim,
Uleman, & Trope, 2009). Having the participants think about both
sides of the situation is consistent with the KPMG judgment frame-
work’s suggestion to have auditors consider both the pros and cons
of a situation (KPMG, 2011) and with current accounting research
that also investigates the effect of using a judgment framework
based on construal-level theory, but in a much different setting
(Backof, Bamber, et al., 2014).
8
The participants viewed these in-
structions both before and after considering the audit evidence
produced by each of their collections.
The participants given the low-level instructions saw the fol-
lowing wording when considering audit evidence:
c?Thinking speci?cally about each evidence item, list reasons
how management’s estimate could be fairly presented.
Thinking speci?cally about each evidence item, list reasons
?ve evidence items presented randomly. For example, on the second collection
round all conditions viewed evidence items 11 through 15 presented randomly. On
the third collection round all conditions viewed evidence items 16 through 20 pre-
sented randomly, and so on. The evidence sequence can be found in Appendix A.
8
Considering reasons why the estimate could be fairly presented biases against
?nding results in this study.
how management’s estimate could be materially misstated.d?
I drew the wording for this condition from prior literature
which shows that thinking about a situation with speci?c details
of the c?howd? of a situation will invoke low-level construals of
information (Backof, Bamber, et al., 2014; Fujita et al., 2006; Rim
et al., 2009). The participants in the control condition did not re-
ceive documentation instructions but were still required to doc-
ument the reasoning for their judgments and actions. The control
group documented evidence at the end of the task concurrent with
providing their judgments.
Dependent variables
I captured multiple measures of professional skepticism due to
the debate surrounding a proper de?nition of professional skep-
ticism (Glover & Prawitt, 2013; Hurtt et al., 2013; Nelson, 2009).
Consistent with the model developed by Nelson (2009), I use sep-
arate measures for skeptical judgment and skeptical action. Af-
ter determining that they collected su?cient evidence, the partic-
ipants reported their perceived risk that the estimate is materially
misstated and their perceived likelihood that they would recom-
mend an adjustment to the client’s reacquired franchise rights ac-
count. I consider these two variables to be measures of skeptical
judgment consistent with Hurtt et al. (2013).
9
The computerized instrument allowed me to record the amount
of time each auditor spent collecting and reviewing evidence as
well as the total time spent on the task. The instrument also al-
lowed me to view the number of times each auditor collected ad-
ditional audit evidence. I consider each of these variables to be a
measure of skeptical action consistent with the de?nition provided
by Hurtt et al. (2013). Table 2, Panel A displays univariate statistics
for each of the measures described above while Panel B displays
the correlations between the dependent variables. The participants
9
Two of the measures of skeptical action (time spent collecting and reviewing
audit evidence and the number of searches for evidence) come before the mea-
sures of skeptical judgment. Normally judgments precede actions and I do not ar-
gue differently here. Auditing is a process that involves sequential decision making
(Gibbins, 1984; Knechel & Messier, 1990). I do not explicitly capture the judgments
that precede the actions described, but presumably the initial judgments would be
whether or not the accumulated evidence is su?cient to render a conclusion or
whether more evidence should be collected.
J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55 49
Table 2
Dependent variable information.
High-level instructions
condition mean (s.d.)
Low-level instructions
condition mean (s.d.)
Control condition mean
(s.d.) Total mean (s.d.)
(n = 19) (n = 19) (n = 20) (n = 58)
Panel A: Descriptive statistics
Risk of material misstatement 7.53 (2.245) 4.00 (2.380) 4.25 (2.074) 5.24 (2.723)
Likelihood of recommending an adjustment 6.74 (2.997) 3.16 (2.522) 4.00 (2.103) 4.62 (2.943)
Time spent collecting and reviewing audit evidence (in minutes) 16.65 (6.371) 11.15 (11.778) 5.14 (2.973) 10.88 (9.067)
Total time spent on the study (in minutes) 26.68 (6.407) 19.09 (13.312) 17.59 (6.470) 21.06 (9.946)
Number of additional collections of evidence 2.63 (1.535) 1.37 (1.571) 2.50 (1.100) 2.17 (1.500)
Risk of material
misstatement
Likelihood of
recommending an
adjustment
Time spent collecting
and reviewing audit
evidence
Total time spent on the
study
Number of additional
collections of evidence
Panel B: Correlations between dependent variables
Risk of material misstatement ? 0.844 0.509 0.454 0.453
0.000 0.000 0.000 0.000
Likelihood of recommending an
adjustment
0.815 ? 0.304 0.286 0.393
0.000 0.020 0.029 0.002
Time spent collecting and reviewing
audit evidence
0.567 0.439 ? 0.926 0.549
0.000 0.001 0.000 0.000
Total time spent on the study 0.477 0.381 0.888 ? 0.610
0.000 0.003 0.000 0.000
Number of additional collections of
evidence
0.442 0.433 0.554 0.612 ?
0.001 0.001 0.000 0.000
Pearson product-moment correlation coe?cients are displayed above the ?’s while the Spearman’s rank correleation coe?cients are displayed below the ?’s. The p-values
for the correlations are listed beneath the coe?cients.
concluded the task by completing a post-experimental question-
naire that included a manipulation check, the Hurtt (2010) profes-
sional skepticism scale, and demographic questions.
10
Results
Manipulation check
I had two coders blind to the hypotheses code each partici-
pant’s documentation. I gave the coders training in the differences
between high-level construals and low-level construals and then
asked them to code each group of responses as either a +1 if the
responses appeared to be high-level construals or a ?1 if the re-
sponses appeared to be low-level construals.
11
Participants using
the high-level instructions averaged a response of 0.465 indicating
that, on average, their responses were consistent with processing
evidence with high-level construals.
In contrast, participants using the low-level instructions aver-
aged a response of ?0.407 while participants not receiving doc-
umentation instructions averaged a response of ?0.158. Both of
these response levels indicate that, on average, users of the low-
level instructions and participants not given documentation in-
structions processed evidence with low-level construals. There is a
signi?cant difference in construal levels between the three groups
overall (p-value < 0.001).
I note that there is not a signi?cant difference in construal
level between users of the low-level instructions and participants
not given documentation instructions (planned contrast p-value =
0.283). This ?nding provides validation of the claim that using the
low-level instructions is akin to current audit practice regarding
10
I tested all responses in my post-experimental questionnaire, including re-
sponses to the Hurtt (2010) professional skepticism scale. The inclusion of these
possible control variables does not affect the inferences drawn from the results de-
scribed in the next section.
11
The coders worked out any differences in coding in a single round of reconcili-
ation resulting in 100% agreement in the ?nal coding set.
the collection and processing of audit evidence. Presumably au-
ditors not receiving documentation instructions will fall back on
collection and processing methods that they use in practice. The
lack of a signi?cant difference in construal level shows that the
two groups of auditors are both processing evidence with low-level
construals on average.
12
Test of hypotheses – Skeptical judgment dependent variables
The ?rst skeptical judgment dependent variable is the partic-
ipants’ assessments of the risk that the client’s estimate is ma-
terially misstated. The participants entered their responses on a
scale of 1 to 10 where 1 signi?ed c?Very Unlikelyd? and 10 sig-
ni?ed c?Very Likely.d? The second skeptical judgment dependent
variable is the participants’ assessments of the likelihood that they
would recommend an adjustment to the client’s estimate. The par-
ticipants entered their responses on a scale of 1 to 10 where 1
signi?ed c?Very Unlikelyd? and 10 signi?ed c?Very Likely.d? Given
the high correlation (see Table 2, Panel B) as well as a high Cron-
bach’s alpha (0.914) between these two variables, for parsimony, I
combine the variables and describe the results based on tests of
the combined skeptical judgment variable.
13
Hypothesis 1 predicts that auditors using the high-level instruc-
tions will display higher professional skepticism than will auditors
12
There is generally a rather broad distinction between responses coded as high-
level construals and those coded as low-level construals. The majority of responses
coded as low-level construals simply repeated individual items of evidence without
making any links between the evidence. For example, one responses coded as a
low-level construal was: c?the estimate was prepared by an accounting o?cial. The
use of the model conforms with GAAP.d? The majority of responses coded as high-
level construal demonstrated an assimilation of evidence items (past and current).
For example, one auditor wrote: c?I think the evidence collectively supports that
the estimate is extremely high. The client’s overall choices seem to indicate that
they want the estimate high, and this type of pressure would be a good reason
why the estimate could be materially misstated.d?
13
Inferences drawn from the results described hereafter are the same as those
that would be drawn from an analysis of the individual measures.
50 J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55
Table 3
Tests of hypotheses, combined skeptical judgment dependent variable.
Source df Mean square F-statistic p-Value
Panel A: ANOVA results
Construal-level instructions 2 69.867 13.627 0.000
???
Error 55 5.127
Total 57
Contrast Difference Std. error t-statistic df p-Value (one-tailed)
Panel B: Planned contrasts
High-level > low-level 3.58 0.735 4.836 55 0.000
???
High-level > control 3.01 0.725 4.145 55 0.001
???
Low-level < control 0.55 0.725 0.753 55 0.455
???
Signi?cant at the 0.001 level.
using the low-level instructions. Hypothesis 2 predicts that audi-
tors using high-level instructions will display higher professional
skepticism than will auditors not given documentation instruc-
tions. I use one-way Analysis of Variance (ANOVA) and planned
contrasts to test the hypotheses.
14
Panel A of Table 3 displays the results of the ANOVA analy-
sis for the combined skeptical judgment dependent variable. This
test reveals that there are signi?cant differences in the dependent
variable between conditions (p-value < 0.001) but does not reveal
where these differences can be found. I therefore use planned con-
trasts to test Hypotheses 1 and 2. Panel B of Table 3 shows the re-
sults of the planned contrasts using the combined dependent vari-
able.
The ?rst contrast tests whether the skeptical judgment dis-
played by participants using the high-level instructions is higher
than that displayed by participants using the low-level instruc-
tions. The signi?cant p-value ( low-level 5.49 3.072 1.787 27.703 0.043
?
Contrast Difference Std. error t-statistic df p-Value (one-tailed)
Panel D: Planned contrast of total time spent on the experiment
High-level > low-level 7.59 3.389 2.238 25.914 0.017
?
?
Signi?cant at the 0.05 level.
??
Signi?cant at the 0.01 level.
???
Signi?cant at the 0.001 level.
Table 5
Tests of hypotheses, number of evidence collections dependent variable.
Source df Mean square F-statistic p-Value
Panel A: ANOVA on number of collections of audit evidence
Construal-level instructions 2 9.217 4.615 0.014
?
Error 55 1.997
Total 57
Contrast Difference Std. error t-statistic df p-Value (one-tailed)
Panel B: Planned contrasts of number of collections of audit evidence
High-level > low-level 1.26 0.459 2.755 55 0.004
??
High-level > control 0.13 0.453 0.291 55 0.386
Low-level < control 1.13 0.453 2.499 55 0.008
??
???
Signi?cant at the 0.001 levels, respectively.
?
Signi?cant at the 0.05 levels, respectively.
??
Signi?cant at the 0.01 levels, respectively.
instructions appear to be better at processing incomplete sets of
evidence than auditors using low-level instructions, leading them
to more frequently realize the need for greater evidence before ar-
riving at a conclusion.
The second contrast shows that auditors using high-level in-
structions conducted only 0.13 more collections, on average, than
auditors not given documentation instructions. This result is
not signi?cant (p-value = 0.386) and provides no support for
Hypothesis 2. This lack of a ?nding is curious considering that au-
ditors using high-level instructions displayed higher professional
skepticism than auditors not given documentation instructions
with the measures of skeptical judgment discussed earlier.
One possible reason why auditors in these two conditions do
not differ in their number of collections could be a difference in
their perceptions of the task di?culty. The participants not given
documentation instructions could have continued collecting evi-
dence simply because it was less cognitively taxing to do so since
they did not have the requirement to document evidence each
time. Also, control participants did not have to explicitly consider
reasons why the estimate could be fairly presented or materially
misstated after each collection. I asked the participants in the post-
experimental questionnaire to rate their perceived di?culty of the
task. Participants using high-level instructions rated the task as
signi?cantly more di?cult than did participants in the other two
conditions (p-value = 0.049). However, there is not a signi?cant
correlation between the number of collections dependent variable
and the task di?culty covariate (Pearson correlation -0.027, p-value
= 0.841).
In order to more de?nitively test my conjecture, I test for this
correlation again using only participants given high-level instruc-
tions and participants not given documentation instructions. This
test reveals a negative correlation between the number of collec-
tions and perceived task di?culty just short of conventional sig-
ni?cance (Pearson correlation 0.310, p-value = 0.055). The change
from almost no relation to a moderately signi?cant relation pro-
vides fuel for future research into the possible relation between
task di?culty and professional skepticism.
Despite conducting statistically the same number of rounds of
evidence collection, recall that participants using high-level in-
structions properly rated the estimate’s risk as high given that the
collected evidence pointed to the estimate residing on the high
end of a reasonable range. Judgments of the perceived risk of the
estimate did not differ between users of low-level instructions and
participants not given documentation instructions. These results
52 J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55
collectively suggest that, when given the same level of informa-
tion, using high-level instructions allows a user to more effectively
process information.
Comparison of results with concurrent research
Overall, I ?nd that documentation instructions based on high-
level construals can increase professional skepticism in a setting
calling for higher skepticism. There are two concurrent research
studies that largely speak to the same issues as the ones noted
in my study: Backof, Thayer, et al. (2014) and Gri?th et al. (2015).
Backof, Thayer, et al. (2014) also use construal-level theory in their
research design while Gri?th et al. (2015) use mindset theory,
a theory very related to construal-level theory.
16
The results of
Gri?th et al. (2015) appear to complement the results of
my study in that the deliberative mindset (a mindset that
promotes broad thinking about a situation similar to high-
level construal thinking) appears to increase professional skep-
ticism of a client’s fair value estimate in a situation call-
ing for increased skepticism.
17
Similarly, the Gri?th et al.
study reveals that the implemental mindset (a mindset that
promotes speci?c thinking about a situation similar to low-
level construal thinking) appears to work against professional
skepticism.
The results of the Backof, Thayer, et al. (2014) study, how-
ever, appear on their face to con?ict with the results of my study.
Their study reveals that auditors considering evidence with low-
level construals display higher professional skepticism, particularly
when presented with evidence in graphical form. A possible rea-
son for the differences in results between our three studies can be
found in an examination of the experimental tasks.
All three of our studies charge the participants with evaluating
evidence regarding a fair value estimate put forth by management.
On closer examination there appears to be a distinct difference
in the scope of the evaluation performed by the auditor partici-
pants in the experimental tasks. The auditor participants in Backof,
Thayer, et al. (2014) focus speci?cally on two particular assump-
tions used by management in their discounted cash ?ow (DCF)
model: the location growth rate and the revenue growth rate.
The auditor participants in Gri?th et al. (2015) receive evi-
dence relating to four major assumptions of the DCF model: rev-
enue projections, expense projections, capital expenditures projec-
tions, and the discount rate. The evidence set used in this experi-
mental task is thus broader than the evidence set used in Backof,
Thayer, et al. (2014). This difference requires greater assimilation of
the evidence, particularly since the authors seeded important evi-
dence about certain assumptions in the evidence collected about
other assumptions. Importantly, the participants in both Gri?th
et al. (2015) and Backof, Thayer, et al. (2014) receive one set of
evidence to evaluate and do not have the ability to collect further
evidence.
The auditor participants in my study could access a much
broader set of evidence. While I include evidence about the DCF
16
I note that the description of the deliberative mindset written by Gri?th et al.
(2015) contains many similarities to descriptions of high-level construals noted in
psychology research (such as Trope & Liberman, 2010, 2003). For example, Gri?th
et al. (2015) describe that the deliberative mindset facilitates broad consideration
of evidence that is less focused on task speci?c details. Gri?th et al.’s (2015) de-
scription of the implemental mindset also appears to be similar to descriptions of
low-level construals. For example, the authors note that the implemental mindset
focuses on the c?howd? of a situation and prompts a narrow focus on the speci?cs
of a situation rather than the overall big picture. Both of these aspects are identical
to descriptions of low-level construals described in construal-level theory research
(Trope & Liberman, 2010, 2003).
17
Gri?th et al. (2015) seed incidental information into their evidence that sug-
gests that the client’s estimate is materially overstated.
model used by management, I also include other information re-
lated to management’s construction of the fair value estimate (see
Appendix A for a listing of the evidence). Additionally, my partic-
ipants only receive a subset of the available evidence and have to
decide whether to continue collecting evidence before rendering a
judgment on the estimate.
I consider the differences in audit evidence and judgment fo-
cus described above to parallel the differences between low-level
and high-level construals. The Backof, Thayer, et al. (2014) task has
a speci?c, detailed focus: the reasonableness of two individual as-
sumptions of the DCF model. This is a speci?c, detailed task calling
for the speci?c, detailed manner of thinking promoted by low-level
construals. Thus, instructions priming low-level construals would
work better in this setting.
The Gri?th et al. (2015) task has a broader, but still some-
what limited focus: the four major assumptions of the DCF model.
The participants are given evidence concerning each of the as-
sumptions, and critical information about one assumption has been
seeded in the evidence of another assumption. This broader task
and the need to assimilate information from different sources
of evidence suggests that high-level construals (or the deliber-
ative mindset) are better suited for the task setting. However,
the Gri?th et al. (2015) setting speci?cally focuses on the DCF
model rather than the totality of audit evidence that would be
considered by an auditor in this situation, including manage-
ment’s motivation and the level of expertise of the estimate’s pre-
parer, for example. The fact that this setting appears to strad-
dle the line between speci?c and broad thinking could explain
why Gri?th et al. (2015) ?nd certain inconsistencies in their
results.
18
My setting is the broadest of the three experimental tasks set-
ting and requires participants to consider evidence relating to sev-
eral different aspects of the estimate, including evidence relating
to the DCF model assumptions, management’s motivation, and the
experience level of the estimate’s preparer, among others. The very
broad nature of the evidence suggests that high-level construals of
the evidence would work best in terms of mental processing, and
that is what I document in this study.
Further, my participants do not receive a complete set of ev-
idence to consider, a situation which simulates the nature of
audit evidence collection in practice. Auditors receive an initial
set of evidence and then must determine whether the accumu-
lated evidence is su?cient to render a judgment or whether
more evidence must be collected (Gibbins, 1984; Knechel &
Messier, 1990). There is a greater need for assimilation and
abstraction of evidence in my study to support this determi-
nation since the auditors need to assess the big picture of
what the evidence reveals collectively rather than individually.
High-level construals support this initial determination and also
facilitate the assimilation of evidence collected in subsequent
rounds.
Conclusion
In this study, I investigate if audit documentation instructions
can help auditors construe (interpret) audit evidence in a man-
ner that enhances processing of the evidence and promotes pro-
fessional skepticism. My results indicate that auditors using docu-
mentation instructions that promote high-level construals demon-
18
For example, Gri?th et al. (2015) ?nd that there is no difference in target issues
with the estimate identi?ed by their mindset (non-control) participants. Also, the
authors ?nd that, based on a conventional statistics level, there is no difference
between mindset participants in decisions to contact the audit manager.
J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55 53
strate a higher level of professional skepticism. Further, relative to
auditors receiving documentation instructions that promote low-
level construals and auditors not receiving instructions, auditors
using high-level instructions are able to more effectively process
audit evidence, particularly incomplete sets of evidence.
My results also suggest that auditors in practice collect and pro-
cess information with low-level construals. Speci?cally, I ?nd no
signi?cant differences in the perceived risk of the estimate pro-
vided by users of low-level instructions and auditors not given
documentation instructions. Further, the auditors in these two
groups documented evidence in a manner consistent with low-
level construals of the audit evidence. According to several auditors
interviewed for this study, this problem could be due to a lack of
training in processing incomplete sets of evidence. My ?ndings can
thus be extremely bene?cial to current accounting practice. The
high-level instructions are easy to use, provide important bene?ts,
and will be inexpensive for ?rms to incorporate.
The ?ndings in this study also add to the current research
stream relating to construal-level theory and mindsets. Backof,
Thayer, et al. (2014) provide evidence that low-level constru-
als bene?t auditors when they must focus on speci?c details
of an audit of a complex estimate (two particular assumptions
used by management in their discounted cash ?ow analysis, for
example). Gri?th et al. (2015) demonstrate that a deliberative
mindset, akin to thinking with high-level construals, helps audi-
tors think more critically about their audits of complex estimate
when the auditors must broaden their focus to the entire dis-
counted cash ?ow model rather than just a few of the assump-
tions. The results of my study complement these ?ndings by show-
ing that high-level construals aid auditors in processing evidence
and maintaining professional skepticism in the broadest setting of
a complex estimate audit in which the auditors receive incom-
plete sets of evidence relating to matters beyond just manage-
ment’s assumptions. Considering the results of all three studies as
well as key aspects of construal-level theory, I infer that broader
tasks bene?t from high-level construals while speci?c tasks ben-
e?t from low-level construals. Future research can test this
conjecture.
Every experimental study is subject to limitations, but these
limitations often represent excellent opportunities for future re-
search. One limitation is that the study does not directly test the
conjecture in the previous section that task complexity is neg-
atively related with professional skepticism. More di?cult tasks
could actually induce higher skeptical action due to the increase
in focus required by the higher di?culty. On the other hand, as
suggested by the results in this study, higher complexity could
cause an auditor to prematurely conclude a task, such as searching
for audit evidence, simply because the task is cognitively draining.
Such an end to the task could be later interpreted that the auditor
failed to display an appropriate amount of professional skepticism.
Another question that could not be answered by this study
is whether the high-level instructions increases skeptical judg-
ments and actions when such an increase is not warranted. For
example, future research could test for increases in skeptical judg-
ments and actions in a situation where the client’s estimate is
fairly presented. The high-level instructions could be deemed in-
e?cient if there are increases in skepticism and the increases
are to the level of those displayed in the current study, although
this ine?ciency would still be consistent with Nelson’s (2009)
c?presumptive doubtd? notion of professional skepticism.
The documentation instructions provided to the auditors in
this experiment asked them to consider reasons both for and
against the reasonableness of the estimate. Considering both the
pros and the cons is consistent with current audit practice and
with recommendations of the KPMG judgment framework (KPMG,
2011). However, considering both sides could potentially balance
the competing effects of the positive and negative information. Fu-
ture research can examine different wording of the instructions.
I included a control condition in which participants did not re-
ceive documentation instructions and therefore were not explicitly
asked to consider how or why the fair value estimate was either
fairly presented or materially misstated. The control condition also
differed from the experimental conditions in that their documenta-
tion occurred concurrently with their judgments rather than hav-
ing them document after each round of evidence collection. These
issues limit inferences that can be drawn from comparisons of the
control group with the experimental groups.
Finally, the experimental instrument used to test the hypothe-
ses is based on aspects of construal-level theory. Construal-level
theory espouses that high-level construals can be primed by hav-
ing a participant think broadly and about c?whyd? something hap-
pens while low-level construals can be primed by having a par-
ticipant think speci?cally and about c?howd? something happens.
Neither this study nor any construal-level theory studies, to my
knowledge, explore whether the how vs. why aspect of the prime
or the broad vs. speci?c aspect of the prime is more powerful.
19
Future research can help determine if one of these aspects domi-
nates the other or whether combining the aspects provide additive
or multiplicative power to the manipulation.
Acknowledgements
I thank my dissertation committee: Uday Murthy (chair), Lisa
Gaynor, Mark Mellon, and Joe Vandello. This manuscript bene?ted
from excellent comments provided by audience members at the
2014 Accounting, Organizations & Society Conference on Account-
ing Estimates, the editors of the conference (Lisa Koonce, Mark
Peecher, and Hun Tong Tan) and the anonymous reviewers. Special
thanks to Deloitte LLP for its sponsorship of the conference and
its process. I also thank Stephen Fuller, Erin Hamilton, Bob Libby,
Rina Limor, Ivo Tafkov, and participants from workshops at Ap-
palachian State University, College of Charleston, Georgia State Uni-
versity, Florida Atlantic University, Kent State University, the Uni-
versity of South Carolina, and the University of South Florida. Fi-
nally, I thank all of my auditor participants and their respective
audit ?rms for allowing me a portion of their valuable time.
Appendix A. – Evidence provided to participants
Evidence Screen 1 (2 positive, 4 neutral, 4 negative):
1. You test management’s mathematical calculations and formulas.
You do not discover any errors.
2. The use of fair value for the reacquired franchise rights account
is appropriate and conforms with GAAP.
3. APC produces pizzas of higher quality than most other pizza
franchises.
4. APC’s controller produced the fair value estimate.
5. There is a lack of objective data that can be used when calcu-
lating the estimate for the reacquired franchise rights account.
6. You interview some of American Pizza Company’s lower level
accountants, but they have no knowledge of how the reacquired
franchise rights account is prepared.
19
That said, there are psychology studies that only use the how vs. why aspect
in their manipulations with success (Fujita et al. (2006) and Freitas, Gollwitzer, and
Trope (2004), for example).
54 J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55
7. You spend time preparing a benchmark model to compare the
output with management’s estimate. Management’s estimate
appears at the high end of your reasonable range, but any dif-
ferences could be attributed to fundamental differences in the
models used.
8. The demand for high quality pizza is decreasing due to the cur-
rent economic environment. Although APC produces pizzas of
generally higher quality than most pizza chains, the company’s
pizza product would be considered on the low end of the high
quality pizza scale.
9. Management’s estimate appears to be moderately to highly
sensitive to changes in the basic assumptions used to calculate
the estimate.
10. You compile a report which suggests that Pennsylvania could
be saturated with pizza restaurants.
Evidence Screen 2 (1 positive, 2 neutral, 2 negative):
1. The model used by management is the same model used by
other companies in the pizza franchise industry and is consis-
tent with the model used by APC’s main competitors.
2. You determine that there is only a moderate risk associated
with the expected future cash ?ows related to the reacquired
franchise rights account.
3. You review the minutes of APC’s last two board meetings but
do not ?nd any mention of management’s intentions relative to
the reacquired franchise rights account.
4. You ask a specialist in your ?rm to prepare a benchmark esti-
mate. His estimate is lower than management’s fair value es-
timate. However, the specialist states that the difference could
be due to circumstances unique to the company.
5. Restaurant franchises similar to American Pizza Company have
been growing at a rate of 3.5 restaurants per year.
Evidence Screen 3 (1 positive, 2 neutral, 2 negative):
1. The model used by management is consistent with what com-
panies in other industries use when estimating reacquired fran-
chise rights. The model appears to be the standard model used
by over 95% of companies when estimating reacquired franchise
rights.
2. American Pizza Company focuses on a dine-in experience as
opposed to the traditional pizza delivery business.
3. The increase in fuel costs has caused an increase in the expense
ratio of other pizza companies, but APC focuses on creating
dine-in restaurants. High fuel costs do not signi?cantly affect
APC’s expense ratio.
4. A detailed examination of American Pizza Company’s history
shows that the company meets its target growth rate approx-
imately 80% of the time, but it growth rate during the other
20% of the time is only two restaurants per year.
5. The person who prepared APC’s fair value estimate does not
have prior experience in preparing fair value estimates.
Evidence Screen 4 (1 positive, 2 neutral, 2 negative):
1. As of the date of your review, APC appears to have the eco-
nomic ability to follow through with its planned restaurant
growth.
2. The CFO of APC tells you about the company’s plans relative to
reacquired franchise rights. The plan appears to be consistent
with the assumptions used in management’s estimation model.
3. American Pizza Company’s marketing expenses have not in-
creased signi?cantly over the past three years.
4. The management of APC did not consider using a third party to
prepare the estimate.
5. An interview with the person who prepared the fair value esti-
mate indicated that the CEO of APC expressed a desire for the
estimate to be as high as possible. The preparer stated that no
one exerted undue in?uence on him, though, and the preparer
does not have any incentives to overstate the estimate.
Evidence Screen 5 (1 positive, 2 neutral, 2 negative):
1. Upon reviewing budgets set by APC’s management, you dis-
cover evidence that APC plans to follow through with its
planned restaurant growth.
2. APC’s pizza receives rave reviews from its customers.
3. Hungry Howie’s is closing down most of its restaurants in
Pennsylvania, but Hungry Howie’s is not in the same pizza mar-
ket segment as American Pizza Company. The closure of the
restaurants should not signi?cantly impact APC’s business in
Pennsylvania.
4. Management has used a different model in the past when
preparing the estimate for its reacquired franchise rights ac-
count. The estimate for this account would be lower when us-
ing the old model. Management justi?ed the change by saying
that the model used this year is the model used by the majority
of its competitors.
5. The discount rate used by management could be a little low
given the current economic conditions. The use of a lower rate
could lead to management’s estimate being slightly higher than
would otherwise be expected.
References
Backof, A. G., Bamber, E. M., & Carpenter, T. D. (2014). Do auditor judgment frame-
works help in constraining aggressive reporting? Evidence under more precise and
less precise accounting standards. Working paper: University of Georgia and Uni-
versity of Virginia.
Backof, A. G., Thayer, J., & Carpenter, T. D. (2014b). Auditing complex estimates:
Management-provided evidence and auditors’ consideration of inconsistent evi-
dence. Working paper, University of Virginia & University of Georgia.
Bratten, B., Gaynor, L. M., McDaniel, L., Montague, N. R., & Sierra, G. E. (2013). The
audit of fair values and other estimates: The effects of underlying environmen-
tal, task, and auditor-speci?c factors. Auditing: A Journal of Practice & Theory,
32(Supplement 1), 7–44.
Brewster, B. E. (2011). How a systems perspective improves knowledge acquisition
and performance in analytical procedures. The Accounting Review, 86(3), 915–
943.
Freitas, A. L., Gollwitzer, P., & Trope, Y. (2004). The in?uence of abstract and concrete
mindsets on anticipating and guiding others’ self-regulatory efforts. Journal of
Experimental Social Psychology, 40, 739–752.
Fujita, K., Trope, Y., Liberman, N., & Levin-Sagi, M. (2006). Construal levels and self-
control. Journal of Personality and Social Psychology, 90(3), 351–367.
Gibbins, M. (1984). Propositions about the psychology of professional judgment in
public accounting. Journal of Accounting Research, 22(1), 103–125.
Glover, S. M., & Prawitt, D. F. (2013). Enhancing auditor professional skepticism. .
Gollwitzer, P. M. (1990). Action phases and mind-sets. In E. T. Higgins, & R. M. Sor-
rentino (Eds.), Handbook of motivation and cognition: Foundations of social behav-
ior (pp. 53–92). New York: Guilford Press.
Gri?th, E. E., Hammersley, J. S., & Kadous, K. (in press). Audits of complex es-
timates as veri?cation of management numbers: How institutional pressures
shape practice. Contemporary Accounting Research.
Gri?th, E. E., Hammersley, J. S., Kadous, K., & Young, D. (2015). Auditor mindsets
and audits of complex estimates. Journal of Accounting Research, 53(1), 49–77.
Hurtt, R. K. (2010). Development of a scale to measure professional skepticism. Au-
diting: A Journal of Practice & Theory, 29(1), 149–171.
Hurtt, R. K., Brown-Libured, H., Earley, C. E., & Krishnamoorthy, G. (2013). Research
on auditor professional skepticism: Literature synthesis and opportunities for
future research. Auditing: A Journal of Practice & Theory, 32(Supplement 1), 45–
97.
International Forum of Independent Audit Regulators (IFIAR), (2012). Summary
report of inspection ?ndings. Available at: .
Knechel, W. R., & Messier, W. F., Jr. (1990). Sequential auditor decision making: In-
formation search and evidence evaluation. Contemporary Accounting Research,
6(2), 386–406.
J.T. Rasso / Accounting, Organizations and Society 46 (2015) 44–55 55
Kohlbeck, M. J., Cohen, J. R., & Holder-Webb, L. L. (2009). Auditing intangible assets
and evaluating fair market value: The case of reacquired franchise rights. Issues
in Accounting Education, 24(1), 45–61.
KPMG (2011). Elevating professional judgment in auditing and accounting: The
KPMG professional judgment framework. .
Nelson, M. W. (2009). A model and literature review of professional skepticism in
auditing. Auditing: A Journal of Practice & Theory, 28(2), 1–34.
Public Company Accounting Oversight Board (PCAOB) (2008). Report on the
PCAOB’s 2004, 2005, 2006, and 2007 inspections of domestic annually in-
spected ?rms. PCAOB Release No. 2007-008. .
PCAOB (2012). Maintaining and applying professional skepticism in audits. PCAOB Staff
Audit Practice Alert No. 10. December 4, 2012. Washington, D.C.: PCAOB.
PCAOB (2014). Due professional care in the performance of work. AU Section 230.
.
Plumlee, D., Rixom, B. A., & Rosman, A. J. (2015). Training auditors to solve ill-
structured tasks using metacognitive skills. The Accounting Review, 90(1), 351–
369.
Rim, S., Uleman, J. S., & Trope, Y. (2009). Spontaneous trait inference and construal
level theory: Psychological distance increases nonconscious trait thinking. Jour-
nal of Experimental Social Psychology, 45, 1088–1097.
Trope, Y., & Liberman, N. (2003). Temporal construal. Psychological Review, 110(3),
403–421.
Trope, Y., & Liberman, N. (2010). Construal-level theory of psychological distance.
Psychological Review, 117(2), 440–463.
doc_352638728.pdf