Description
Most studies on cost-based decision-making examine the profit impact of cost reports that rely on different methods
to allocate costs. In practice, firms’ cost reports often employ the same cost allocation method with subtle variations in
the way that the cost data are presented
The interplay between cost accounting knowledge
and presentation formats in cost-based decision-making
Eddy Cardinaels
*
Tilburg University, Department of Accountancy, P.O. Box 90153, 5000 LE Tilburg, The Netherlands
Abstract
Most studies on cost-based decision-making examine the pro?t impact of cost reports that rely on di?erent methods
to allocate costs. In practice, ?rms’ cost reports often employ the same cost allocation method with subtle variations in
the way that the cost data are presented. This paper examines experimentally the pro?t impact of a cost report’s pre-
sentation format in relation to a decision maker’s level of cost accounting knowledge. Using a customer pro?tability
report prepared using activity-based costing and presented in either a tabular or a graphical format, participants ana-
lyze a complex pricing and resource allocation task that a?ects ?rm pro?tability. The results suggest a strong relation
between presentation format and cost accounting knowledge. Speci?cally, decision makers with a low level of cost
accounting knowledge attain higher pro?ts when they use a graphical format in comparison to a tabular format. More
surprisingly, graphs (versus tables) have an adverse e?ect on pro?ts for users with a high level of cost knowledge. This
result has broad implications: in order to facilitate the decisions of a variety of users of accounting data (e.g. managers,
external investors, etc.), ?rms may need to adapt the presentation format of their accounting data to the level of
accounting sophistication of the users.
Ó 2007 Elsevier Ltd. All rights reserved.
Introduction
The performance e?ects of di?erent types of
cost reports (variable versus absorption costing,
volume-based versus activity-based costing) in
relation to a number of contextual variables is
the key focus of several previous studies on cost-
based decision-making (e.g. Briers, Chow, Hwang,
& Luckett, 1999; Drake, Haka, & Ravenscroft,
1999; Gupta & King, 1997; Waller, Shapiro, &
Sevcik, 1999). This paper presents the results of
an experiment conducted to study how di?erent
representations of identical underlying cost data
a?ect cost-based decision-making and ?rm pro?t-
ability. Speci?cally, I ?nd unique evidence suggest-
ing that the pro?t impact of tabular versus
graphical representations of activity-based costing
0361-3682/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2007.06.003
*
Tel.: +31 13 4668231; fax: +31 13 4668001.
E-mail address: [email protected]
Available online at www.sciencedirect.com
Accounting, Organizations and Society 33 (2008) 582–602
www.elsevier.com/locate/aos
(ABC) data is dependent upon the accounting
sophistication of the user, i.e. his or her level of
cost accounting knowledge.
Studying the joint e?ects of presentation format
and a decision maker’s level of accounting knowl-
edge is important for several reasons. First, di?er-
ent managers clearly have di?erent levels of
accounting knowledge (Dearman & Shields,
2001; Stone, Hunton, & Wier, 2000). Firms’ infor-
mation systems provide managers with reports
that range from traditional tabular formats to
graphical displays (So & Smith, 2002; Sullivan,
1988). Many managers use elaborate cost reports
for their daily decisions. Others encourage the
use of ‘‘easy-to-understand’’ graphs (Mooney,
Rogers, & Wright, 2000; Remus, 1987; Yates,
1985) in the belief that a graphical representation
(versus a tabular format) makes cost data accessi-
ble to all members of the ?rm, irrespective of their
level of accounting knowledge. In spite of evidence
of di?erential managerial knowledge, however, the
extant literature does not indicate how managerial
decision-making and, in turn, ?rm pro?ts are
a?ected (Sprinkle, 2003) when information is pre-
sented in di?erent report formats to decision mak-
ers with di?erent levels of accounting knowledge
(Haynes & Kachelmeier, 1998; Libby, 1981).
Second, recent studies in accounting only
address the separate e?ects of expertise and report
format (Sprinkle, 2003), though such studies ?nd
that both a?ect cost-based decision-making. For
instance, Dearman and Shields (2001) show that
the level of a manager’s cost accounting knowl-
edge is linked with the ability to correct for vol-
ume-based cost bias, and Bucheit (2003) shows
that investment decisions change when cost
reports explicitly contain the cost of capacity
(compared to reports that do not). There is also
some evidence that suggests the interaction of the
two variables. For instance, Vera-Mun˜ oz, Kinney,
and Bonner (2001) show that the impact of alter-
native task representations depends upon the deci-
sion maker’s expertise. However, they employ only
tabular reports based on either historical earnings
or historical cash ?ows. Thus, the impact of tabu-
lar versus graphical representation of identical
data in relation to a decision maker’s knowledge
is an open question.
Finally, although there is a strong belief that
decision makers should bene?t from graphical rep-
resentation (Harvey & Bolger, 1996), research that
compares the relative impact of graphical versus
tabular formats remains inconclusive (Vessey,
1991). In an attempt to resolve the controversy, a
few studies suggest looking at individual di?er-
ences among the users of information (Amer,
1991; Chandra & Krovi, 1999; Ganzach, 1993).
The current study sheds light on this debate by
testing whether accounting knowledge as a mana-
gerial characteristic helps to explain when certain
report formats are associated with stronger perfor-
mance than others in a cost-based decision task.
To investigate these joint considerations, I con-
duct an experiment with presentation format as
the between subjects factor. I measure a partici-
pant’s level of accounting knowledge, in addition
to some common control variables, using research
instruments suggested in prior studies (Bonner &
Lewis, 1990; Cloyd, 1997; Dearman & Shields,
2005). I create a complex task in which the partic-
ipant’s realized pro?t depends upon both price and
resource allocation decisions for a heterogeneous
set of customers. All participants receive ABC-dri-
ven customer pro?tability data, presented in either
a tabular or a graphical format (multicolored bar
charts and trend charts). I measure pro?t perfor-
mance objectively as the di?erence between a par-
ticipant’s realized pro?t and the maximum pro?t
that could be achieved in performing the task.
After controlling for di?erences in ability and
work experience, I ?nd evidence of an e?ect reversal
across knowledge levels: decision makers with a low
level of cost accounting knowledge perform better
with a graphical ABC format, and decision makers
with a high level of cost knowledge obtain superior
pro?ts with a tabular ABC format. Further evi-
dence indicates that graphical formats tend to
reduce task complexity for a low-knowledge deci-
sion maker, whereas tables support the information
search of a more knowledgeable user. This result
provides important theoretical and practical
insights, suggesting that (1) cost accounting knowl-
edge is a crucial managerial characteristic that
should be taken into account when a ?rm presents
cost reports to a decision maker, and (2) managerial
cost accounting knowledge and data representation
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 583
jointly determine the direction on ?rmpro?tability.
This has broad implications for the many informa-
tion ?ows within the ?rm. To convey accounting
data to both internal and external parties (e.g. man-
agers, investors), an exclusive focus on one format
is probably not advisable. Instead, ?rms can facili-
tate decision-making by adapting the presentation
format to the level of the accounting sophistication
of their intended audience.
Prior research
Although accounting systems vary greatly with
respect to howthey package the same kind of infor-
mation, systematic research on this issue remains
scarce (Haynes & Kachelmeier, 1998; Luft, 1994).
Most prior work examines the pro?t impact of cost
reports that are based on di?erent assumptions,
such as activity- versus volume-based costing or
absorption versus variable costing (e.g. Drake
et al., 1999; Gupta & King, 1997; Waller et al.,
1999). This study is the ?rst to disentangle the dif-
ferential impact of tabular versus graphical repre-
sentations of identical ABC cost data. The unique
contribution of this study is that it links di?erential
information representation to the decision maker’s
level of accounting sophistication. Nevertheless,
most evidence on either the e?ects of presentation
format or accounting knowledge is found in sepa-
rate research streams. In what follows I review this
literature and I provide arguments for why we need
to consider the joint e?ects of knowledge and pre-
sentation formats.
With regard to information representation,
some studies focus on salience e?ects. Bucheit
(2003) ?nds that capacity investment decisions
are suboptimal if cost reports explicitly contain
capacity costs (versus reports that exclude these
costs). Maines and McDaniel (2000) show that
non-professional investors are in?uenced by com-
prehensive income information when it appears
in a separate statement, as opposed to when it is
masked in stockholders’ equity. In contrast to sal-
ience e?ects, evidence on the performance e?ects of
various presentation formats (Vessey, 1991) and
the belief that graphs are superior to tables (Har-
vey & Bolger, 1996) remains inconclusive.
Even research that focuses on task characteris-
tics (both task type and task complexity) to address
this controversy (Jarvenpaa, 1989; Remus, 1987;
Vessey, 1991) o?ers mixed results (Hwang, 1995).
Vessey (1991) argues that the representation of
the problem should match the type of task (cogni-
tive ?t) to enhance decision-making. Graphical for-
mats present spatial information (Vessey, 1991):
they tend to emphasize relationships in the data
and provide a holistic view by presenting the data
at a glance. Such formats are most appropriate
when decision makers have to compare alternatives
or tasks that require integration of data (Vessey,
1991). In contrast, tables present symbolic infor-
mation: by emphasizing discrete values, they pro-
vide an analytical view and facilitate the
extraction of speci?c values (item-by-item evalua-
tion) and the evaluation of changes in variables
(Mackay & Villarreal, 1987; Vessey, 1991). Tables
should be more appropriate when the task requires
the manager to extract and act on discrete data val-
ues (Vessey, 1991). Nonetheless, the evidence on
task type remains inconclusive. For instance, Amer
(1991) and Frownfelter-Lohrke (1998) show little
di?erence in decision accuracy between bar charts
and tables for tasks that are either spatial or sym-
bolic in nature. Also, the argument that graphical
formats are preferable for complex tasks (because
analytical processing is di?cult in complex tasks,
graphical formats that stimulate integrative pro-
cessing become more interesting, even for symbolic
tasks) receives little support (Hwang, 1995; Vessey,
1991). Studies in accounting have shown bene?cial
e?ects of graphs (e.g. Stock & Watson, 1984;
Wright, 1995) as well as favorable e?ects of tables
for tasks and questions that are su?ciently com-
plex (Davis, 1989; So & Smith, 2004).
One explanation for these disparate results is
that user characteristics are ignored (Chandra &
Krovi, 1999). Amer (1991, p. 30–31) argues that
subject variation could explain the failure to obtain
certain e?ects of presentation formats. This paper
is the ?rst to test whether speci?c knowledge (i.e.
cost accounting knowledge) is a critical individual
factor that can explain when graphics versus tables
increase or decrease performance (i.e. pro?ts from
cost-based decisions). Knowledge can in?uence
the internal memory representations of a decision
584 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
maker and could therefore be linked to the external
presentation format. Theory suggests that more
knowledgeable subjects possess internal schemata
that help them to better retrieve and search for rel-
evant data, along with internal rules for appropri-
ately weighing and acting on these data (Libby &
Luft, 1993; Rose & Wolfe, 2000; Spilker, 1995).
Knowledge has previously been studied in rela-
tion to task complexity (e.g. Bonner, 1994). Dear-
man and Shields (2001) show that managers with
more cost accounting knowledge perform better
in a judgment task that requires complex evalua-
tions of the level of cost distortions introduced in
a volume-based costing system. Other studies
argue that the e?ects of having more advanced
knowledge structures on the participants’ perfor-
mance in more complex tasks is dependent upon
other conditional factors such as accountability,
problem solving ability, and motivation (e.g.
Cloyd, 1997; Dearman & Shields, 2005; Tan &
Kao, 1999; Tan, Ng, & Mak, 2002). Yet, no study
has explored whether the e?ect of knowledge on
performance in a complex task depends on the
way we present information (i.e. presentation for-
mat) to the decision maker.
Only one recent study by Vera-Mun˜oz et al.
(2001) suggests that knowledge may be related to
the external representation of the problem. They
show that when cash-?ow data are presented in
an inappropriate format (i.e. in a statement of his-
torical earnings), experienced managers (with a
stronger knowledge base) are better able to deter-
mine the relevant cash-?ow items in advising cli-
ents than are less experienced managers.
Experience level has no e?ect, however, when the
information is presented appropriately (i.e. in a
cash ?ow statement). However, this ?nding is
based on tabular formats only and the underlying
data is based on di?erent assumptions (earnings
versus cash ?ow). Also, the performance conse-
quences (in terms of pro?ts) are not assessed.
Theory and hypothesis development
I study the joint e?ects of cost accounting
knowledge and presentation format in a complex
decision task with multiple cues and decision vari-
ables, which require a large amount of processing
(Bonner, 1994; Tan et al., 2002). Participants
simultaneously decide on price and resource allo-
cations for a heterogeneous set of customers. The
task requires participants to compare di?erent cus-
tomers in terms of cost and revenue potential, to
search for relevant information in an ABC-driven
report, and to explore the sensitivity of the results
to various decision rules. I argue that the direc-
tional e?ect on pro?ts of a tabular versus a graph-
ical ABC presentation format depends on the
user’s level of cost accounting knowledge: formats
that bene?t a low-knowledge user may hamper the
performance of a knowledgeable user.
Representational congruence and the cognitive
burden
The theory of representational congruence
(Arnold, Collier, Leech, & Sutton, 2004; Chandra &
Krovi, 1999) predicts a more favorable e?ect
on decision performance when the external presen-
tation format matches the user’s cognitive model or
internal representation (Chandra & Krovi, 1999).
A mismatch leads to a high cognitive load and a
less e?ective information retrieval, negatively
a?ecting the decision maker’s performance. Build-
ing on this work, the decision aid literature suggests
opportunities to reduce the cognitive load by
restoring the ?t between the user and the decision
aid (Rose & Wolfe, 2000; Rose, Douglas, & Rose,
2004). In particular, some authors suggest that in
comparison to tables, graphical formats can reduce
a decision makers’ cognitive burden. Stock and
Watson (1984) and Moriarity (1979) argue that
graphical data allow the decision maker to trigger
analogue graphical representations that are stored
in memory. These representations facilitate data
retrieval and information processing. Similarly,
Wright (1995) argues that graphical data help
reduce information overload by highlighting pat-
terns in the data, promoting the perception and
acquisition of information relationships in short-
term memory. Here, I presume that these bene?ts
of graphics apply mainly to decision makers with
a limited level of cost accounting knowledge.
Cost accounting knowledge is likely to be cru-
cial when decision makers receive a tabular ABC
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 585
report that contains very speci?c ABC-related
information. The advantage of greater cost
accounting knowledge is that it typically leads to
more advanced internal schemata (internal repre-
sentations) that are stored around various ABC
cost concepts (Dearman & Shields, 2001). These
internal schemata help decision makers search
for the relevant details within a tabular ABC cost
report (external representation). Decision makers
with less accounting knowledge, in contrast, lack
the su?cient internal ABC representations to
make a match and retrieve speci?c information
from a tabular ABC format. Based on the theory
of representational congruence (Chandra & Krovi,
1999), a graphical ABC format is likely to reduce
the cognitive burden of less knowledgeable deci-
sion makers by providing a ?t to analogue graph-
ical representations that are stored in memory
(Moriarity, 1979; Stock & Watson, 1984; Wright,
1995). Hence, this ?t should facilitate data retrie-
val and in turn their performance should improve.
Decision styles of users and the search for
information
The above theory says little about the perfor-
mance e?ects of graphical and tabular formats
for knowledgeable users. Such users have the
appropriate internal schemata to e?ciently use
the tabular ABC data for pro?t improvement. If
they use a graphical representation, on the other
hand, they may make parallel ?ts to analogue
graphical representations stored in memory in
the same way as their less knowledgeable counter-
parts. According to Umanath and Vessey (1994, p.
797), this may imply that the decisions of sophisti-
cated and unsophisticated users would not di?er
appreciably with a graphical representation. In
this section, I introduce the notion of the decision
style (Lucas, 1981; Sullivan, 1988) to argue that
graphical formats in comparison to tables could
actually produce adverse performance e?ects for
the knowledgeable decision maker. According to
Lucas (1981), analytical decision makers, or those
who speci?cally look for details, bene?t from
tables. Tables indeed o?er an analytical view of
the data that facilitates item-by-item evaluation
(Vessey, 1991). In contrast, heuristic decision mak-
ers, or those who look at the entire problem, ben-
e?t from graphs that emphasize an overview of the
same data (Lucas, 1981; Vessey, 1991).
There are reasons to believe that greater knowl-
edge in a particular domain leads to an analytical
focus. Factual knowledge can evolve into more
abstract (analytical) representations rather than
surface-level representations in memory (Ander-
son, 1990). Because an analytical approach takes
e?ort, a high degree of knowledge in a domain is
needed to engage in analytical processing (Alba &
Hutchinson, 1987, p. 418). Consistent with an ana-
lytical focus, knowledgeable users perform a more
focused information search prior to solving a prob-
lem(Cloyd, 1997; Wood &Lynch, 2002) and access
rules to distinguish between more and less relevant
information (Bonner, 1990; O’Donnell, Koch, &
Boone, 2005). Conversely, Chi, Feltovich, and Gla-
ser (1981), as cited in Vera-Mun˜ oz et al. (2001, p.
409), suggest that novices as opposed to experi-
enced physicists, tend to focus more on surface fea-
tures (heuristic style). Also, earlier work by
Benbasat and Schroeder (1977) suggests that users
with low-knowledge focus on an overview and
screen all available reports, whereas knowledgeable
decision makers search for speci?c details by
requesting a limited number of speci?c reports.
An unanticipated ?nding in Desanctis and Jar-
venpaa (1989) hints at the possibility that graphical
formats do not ?t the way certain users process
information: some users in their study started to
convert bar charts into numerical tables. In my
study, I predict lower performance with graphics,
particularly for high-knowledge decision makers,
given that knowledge in a domain leads to an
analytical focus. Because graphics emphasize an
overview of the data, changes in certain variables
may go unnoticed; and given that knowledgeable
users look for speci?c details (item-by-item pro-
cessing) we might expect them to perform better
with tables (Mackay & Villarreal, 1987). The fact
that graphics tend to lower decision times by
providing an overview (Hwang, 1995; Painton &
Gentry, 1985) may further lower the performance
of high-knowledge decision makers by impeding
detailed analytical processing.
On the basis of the above arguments, I hypoth-
esize the following:
586 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
HYPOTHESIS – The pro?t performance for
decision makers with high (low) cost accounting
knowledge is higher (lower) given a tabular ABC
presentation format than given a graphical ABC
presentation format.
Experimental method
The experimental task consists of a complex
management accounting problem. Participants
improve pro?ts by conducting eight trials of price
and resource allocation decisions for three custom-
ers of a case company. Presentation format is
manipulated as a between subjects factor: partici-
pants receive ABC-based customer pro?tability
reports presented in either a tabular or a graphical
format. Prior to administering the task, I measure
each participant’s level of knowledge, in addition
to common control variables such as ability and
work experience, using instruments similar to those
in Bonner and Lewis (1990), Dearman and Shields,
2005, and Cloyd (1997). I assume that these vari-
ables are randomly distributed across the between
subjects factor.
1
All experimental materials are
issued via computer in the following order. First,
participants provide demographic information,
which includes a measure of their work experience.
Second, participants complete two pre-tests to mea-
sure their cost accounting knowledge and general
ability. Third, the experimental taskis administered.
Finally, participants complete a debrie?ng ques-
tionnaire. The next subsections describe the partici-
pants, the experimental task and its procedures, and
the measurement of the relevant variables.
Participants
The participants were 55 students enrolled in a
four-year business program at a large west-Euro-
pean university. On average, participants were
22.0 years old (most students were about to start
or ?nish their ?nal year). All students had com-
pleted at least two accounting courses in which
ABC is covered. The advantage of using student
subjects and knowledge pre-tests is that I can better
capture the participant’s knowledge component
without the noise of expertise acquired via work
experience (Libby, 1995; Rose &Wolfe, 2000; Uma-
nath & Vessey, 1994). Nonetheless, further tests
control for self-reported work experience (see Sec-
tion ‘‘Control variables’’) as 93%of my participants
indicated some relevant part-time work experience,
with a mean of just over three years of self-reported
experience. Participants typically applied for part-
time jobs corresponding to their studies (e.g. in a
typical business setting) and one-quarter of them
interned with a Big-four accounting ?rm or with a
large controller department of a private ?rm.
To ensure that participants were motivated, they
received a participation fee of six euros along with
a lottery ticket that gave them a chance to win one
of four 50-euro bonus prizes (Bonner, Libby, &
Nelson, 1997). To stimulate ample cognitive e?ort
in both the pre-tests and the experimental task, par-
ticipants were informed that the chances of receiv-
ing this bonus increased with realized pro?t in the
task as well as with their knowledge and ability test
scores.
2
Two ?ve-point Likert-scaled items for
motivation (in the debrie?ng questionnaire)
showed that on average participants were highly
motivated (mean = 4.30, std. dev. = 0.58, a =
0.62), with no signi?cant di?erences across presen-
tation format and knowledge (all p’s > 0.28).
The experimental task and procedures
Participants play the role of a ?rm’s manager
and review descriptions of the case company and
1
In order to assess the validity of this assumption, t-test
results indicate no signi?cant di?erences (all p’s > 0.14) in the
mean ability and knowledge test scores as well as the mean level
of work experience across the two levels of presentation format.
2
It is not uncommon to also reward for knowledge and
ability scores (e.g. Vera-Mun˜ oz, 1998; Rose & Wolfe, 2000). In
my experiment, ticket numbers could appear 2, 6, 11, 15, 20, 21,
25, or 30 times (practical range: 2–25 tickets) in the lottery using
the following scheme. The person with the highest combined
score in the knowledge and ability test receives 10 tickets, the
next 50% all score ?ve tickets, the rest each receive one ticket.
The experimental task would be similar to that above. The
person with the highest mean pro?t realized over the eight
decision trials receives 20 tickets, the next 50% receive 10
tickets, the rest each receive one ticket. The bonus prices would
be drawn and paid out one week after the experiment.
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 587
its customers. The ?rm is an exclusive distributor
of a crucial product that is sold to three major cus-
tomers, denoted A, B, and C, respectively. Partic-
ipants are told that customers vary in revenue
potential and in customer-speci?c support costs
(delivery, logistics, and sales visit costs). Partici-
pants receive a detailed task description, together
with an initial ABC-driven customer pro?tability
report that presents a situation with ample room
for pro?t improvement.
3
Depending upon the
assigned condition, the cost report presents the
same underlying ABC data (calculations in the
Appendix) in either a tabular or a graphical format
(multicolored bar charts showing absolute data
and data on a per unit/percentage basis). Fig. 1
displays the examples. Participants are instructed
to carefully review this information.
4
Their objec-
tive is to improve pro?ts by altering price (p
i
)
and resource allocation (x
i
) for each customer
within a given range (p
i
2 [80, 140] and x
i
2 [30,
90]). The ABC report is updated after each of
the eight trials. As Fig. 1 shows, prior decisions
and outcomes are stored in either a trend chart
(graphical format) or a table.
As the Appendix shows, both prices and
resource allocations a?ect pro?ts and there is an
optimum level of each at which pro?ts are maxi-
mized. The task is relatively complex: (1) pro?ts
are a?ected by two simultaneous decisions, requir-
ing more mental capacity (Bonner, 1994; Tan
et al., 2002) than a task with one decision variable,
and (2) the task is unstructured (Vera-Mun˜ oz
et al., 2001). Participants know that customers
are heterogeneous in their required level of sales
support. Nevertheless, the cost reports only deli-
ver updated cost information in each trial. Consis-
tent with Christensen and Demski (1995) and
Datar and Gupta (1994), the ABC report does
not perfectly re?ect the true costs, nor are the
underlying cost parameters or functions revealed.
In order to improve pro?ts, participants must
make comparisons across the di?erent customers
and explore several decision rules in each trial to
retrieve, evaluate, and weigh the di?erent items
in their report.
Operational measures of the variables
Performance metrics
Table 1 contains an overview of the main vari-
ables used in the analyses. Consistent with Waller
et al. (1999), I measure each participant’s perfor-
mance as the deviation of his or her realized pro?t
from the optimal ?rm-wide pro?t (averaged over
eight trials). I also compute the mean deviation
of actual price and resource allocation decisions
for Customers A, B, and C against optimal price
and resource allocation decisions. The resulting
metrics in Table 1 are labeled as PROFIT and
DECISION, respectively. Smaller values signify
better performance. The hypothesis test is based
on the pro?t deviation (PROFIT); sensitivity
checks also explore the DECISION metric.
Presentation format
Twenty-eight participants receive their ABC
report and previous decision outcomes in a tabular
format. The remaining 27 participants receive the
same data in a graphical format (multicolored
bar charts of the ABC-data and trend charts for
the previous decision outcomes; e.g. Hwang,
1995; Jarvenpaa, 1989). Both formats are used in
ABC software applications (Mooney et al., 2000;
So & Smith, 2002).
It is important to note that the bar charts con-
tain most, but not all of the items in the table.
Because of computer screen space limitations, the
charts would become too crowded if all items of
3
The initial ABC report displays starting prices of 98, 97, and
104 euros, while sales visits are ?xed at 52, 60, and 54 euros for
Customers A, B and C, respectively. This provides room for
pro?t improvement: Firm-wide pro?t for the starting values is
only 1,614,542, whereas the maximum pro?t (see the Appendix)
equals 3,036,145 (achieved at P
a
= 115.5, P
b
= 124.9, P
c
= 97.6,
x
a
= 62.0, x
b
= 49.4, and x
c
= 83.1). Students are not given any
information about the optima.
4
It is important to note that the computer tracks the time
that participants use to review the problem description and the
initial ABC report, as supplementary analyses use this time as a
proxy for information search (see Section ‘‘Supplementary
analysis’’). The rationale is that the time expended on
information gathering prior to problem solving can provide
indications on how subjects with di?erent levels of knowledge
and presentation formats approach the problem (see Section
‘‘Theory and hypothesis development’’).
588 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
Fig. 1. Actual screenshots of the initial ABC-report issued prior to the task for the two levels of presentation format (between subjects
factor).
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 589
the table were shown, and the e?ectiveness of the
graph (Davis, 1989, p. 497; Tufte, 2001) would
decline. Nevertheless, I expect the two formats to
be essentially identical in information content. In
particular, most of the items are maintained in
the graphical format, including the most crucial
information (i.e. the customer-speci?c costs of
sales visits, internal logistics, and product delivery)
and the few omitted items convey little new infor-
mation. Both the tables as well as the bar charts
reveal that support costs are higher for Customer
B relative to purchasing costs and that the oppo-
site is true for Customer C. Inherent to the features
of presentation formats (Jarvenpaa, 1989; So &
Smith, 2004), the graphical format visualizes the
amount of costs per category by distances in the
space vector (without attaching a discrete value),
whereas tables display the speci?c ?gures.
5
Although bar charts do not convey the amount
of resources consumed per customer (e.g. volume
drivers for sales visits, deliveries, and stock pick-
ings), customer costs are calculated by multiplying
the amount of resources by a ?xed driver rate and
hence they have the same information content as
resources consumed (the two cues are perfectly
Table 1
Summary of variables’ operational de?nitions and descriptive statistics
Variable (De?nition) Mean
N = 55
Standard
deviation
Min. Max. Median
Dependent variables
a
PROFIT (pro?t deviation: mean distance of the total
realized pro?t to the optimal pro?t averaged over 8 trials)
b
761,491 292,732 307,758 1,447,754 773,486
DECISION (deviation of decisions: mean distance of the
actual decisions for Customer A, B, and C to the optimal
values averaged over 8 trials)
74.60 16.46 39.7 104.0 76.4
Independent variables
KNOWLEDGE
b
KNOWSCORE (score on the test for the level of cost
accounting content)
3.84 1.18 2 6 4
KNOW (mean split: high versus low level of accounting
knowledge)
– – 0 1 –
Low High
PRESENTATION (presentation format of ABC-data;
either graphical displays or tables)
c
– – 0 1 –
Table Graphic
Control variables
ABILITY (general ability test score)
d
7.00 1.96 1 11 7
WORKEXP (a participants’ self-reported relevant
part-time work experience in number of years)
3.06 2.37 0 9 2
a
PROFIT used the following formula: R
j
(P
*
À P
j
)/8 with P
*
optimal ?rm-wide pro?t and P
j
a participants’ realized ?rm-wide
pro?t in trial j = 1, . . . ,8. DECISION used the following formula: R
i
R
j
(jx
i*
À x
ij
j + jp
i*
À p
ij
j)/8 with pi*, xi* optimal solutions and x
ij
,
p
ij
a participants’ decision choices for customer i = A, B, C in trial j = 1,. . ., 8. Lower scores represent better performance.
b
The theoretical range of KNOWSCORE is from zero to six. For the mean split variable KNOW, 23 (32) subjects were classi?ed in
the low (high) accounting knowledge condition.
c
Twenty-eight (27) subjects received the ABC report in a tabular (graphical) format. The two reporting formats are displayed in
Fig. 1.
d
The theoretical range of ABILITY is from 0 to 11.
5
Bar charts give the di?erent customer cost categories about
the same color and use a di?erent color for the purchasing cost.
Consistent with the features described in the literature
(Umanath & Vessey, 1994) the bar charts provide an overview
and make certain trends (e.g. relative consumption of customer
costs across customers) more visible. These trends must be
determined by the participant in a tabular format. Nevertheless,
in a tabular format subjects have better access to the discrete
values in comparison to bar charts.
590 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
correlated). In the same way, sales volume is not
shown in the bar chart as volume is highly corre-
lated with revenues. For the total outcome feed-
back, I keep only the most relevant item in the
graphical format, namely total pro?ts. Other totals
in the tabular format can be automatically derived
and are less important, as the participants’ goal is
to increase pro?ts.
Accounting knowledge
Following prior research (Bonner & Lewis,
1990; Dearman & Shields, 2001; Dearman &
Table 2
Hypothesis test
Panel A: ANCOVA-models with LN(PROFIT) as dependent measure
a
Model 1: Controlling for ability and workexp Model 2: Controlling for ability, workexp and
W · K interaction
Source of variation Mean square F-stat p-value Mean square F-stat p-value
Factors
Presentation (P) 0.00146 0.01 0.923 0.01094 0.07 0.787
Know (K) 0.10341 0.67 0.417 0.51393 3.47 0.069
*
P · K 0.63676 4.12 0.048
**
0.75938 5.12 0.028
**
Covariates
Ability (A) 0.42559 2.75 0.103 0.35940 2.42 0.126
Workexp (W) 0.24561 1.59 0.213 0.37611 2.54 0.118
W · K – – – 0.45124 3.04 0.087
*
R
2
= 0.176; N = 55 R
2
= 0.225; N = 55
Panel B: Mean by mean comparison of the Presentation (P) by Know (K) interaction
b
400000
480000
560000
640000
720000
800000
880000
LOW HIGH
.047**
808915
(13.603)
648425
(13.382)
.108
.017**
.224
582432
(13.275)
Table
742561
(13.518)
Graphic
N=8
N=15
N=19
N=13
Model 1: Interaction P x K (F=4.12, p=.048**)
?
400000
480000
560000
640000
720000
800000
880000
LOW HIGH
.057*
804553
(13.598)
604644
(13.312)
.056*
.015**
.153
579436
(13.270)
Table
725387
(13.494)
Graphic
N=8
N=15
N=19
N=13
Model 2: Interaction P x K (F=5.12, p=.028**)
a
Variable de?nitions in Table 1.
**
,
*
Indicate signi?cance levels of 5 or 10%. The test uses the mean split variable for knowledge
(‘know’) to allow for a cell-by-cell comparison of the P · K interaction (see Panel B). Model 1 controls for the covariates work
experience and ability. Model 2 adds the WxK interaction as an additional covariate (experience matters more at lower-levels of cost
knowledge).
b
Detailed analyses of the P · K interaction of the models in Panel A. The horizontal axes represent the knowledge categories (low
versus high). The full lines represent the two presentation formats (tables versus graphics). The vertical axes display the mean pro?t
deviation (LN of pro?t between brackets) evaluated at covariates ‘workexp’ (mean = 3.06) and ‘ability’ (mean = 7.00). A lower pro?t
deviation represents better performance. N indicates the number of participants in each cell. Dotted arrows and associated ?gures
represent the direction and the p-value of a one-sided mean-by-mean comparison.
**
,
*
represents signi?cance at the 5 or 10% level.
a
Indicates that the e?ect of p = 0.108 becomes signi?cant at the 10% level (p = 0.097) when the last trial is excluded from the analysis
(end-trial e?ect).
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 591
Shields, 2005), I measure a participant’s knowl-
edge as the number of correct answers to six multi-
ple-choice questions that are intended to assess the
subject’s level of cost accounting knowledge.
6
The
questions are adapted from test banks of cost and
management accounting textbooks and are de?ni-
tional in nature. They assess conceptual knowl-
edge on ABC- and cost-related accounting topics
such as the e?ects of product complexity on prod-
uct costing, the e?ects associated with speci?c cost
allocation methods, the e?ects of value-added
costs and various types of cost drivers, the ABC
cost driver levels, and some statements on more
elaborate ABC-systems.
Table 1 shows a large variation in the partici-
pants’ knowledge, with scores (KNOWSCORE)
ranging between two and six. Studies with actual
managers tend to show similar variations in knowl-
edge (Dearman & Shields, 2001; Stone et al., 2000),
and reliability of the measure is comparable to other
studies on accounting knowledge.
7
To disentangle
the e?ects of presentation format by knowledge cat-
egory, the hypothesis tests will utilize a mean split
metric (KNOW) to classify participants into a
low- or a high-knowledge subgroup (see Bierstaker,
2003). Table 2 shows that about 23 participants, or
42%, fall into the low-knowledge subgroup. Their
average knowledge score of 2.65 is statistically
di?erent from the score of 4.69 for participants
in the high-knowledge subgroup (t = 12.09;
p < 0.01). Sensitivity checks will also repeat the tests
with the actual knowledge score (KNOWSCORE).
Control variables
Prior studies hypothesize positive relationships
between a decision maker’s ability and audit perfor-
mance (Bonner &Lewis, 1990; Libby &Luft, 1993).
Other studies suggest that work experience leads to
expertise (experience creates opportunities to gain
additional knowledge) that can a?ect performance
(Cloyd, 1997; Libby, 1995). Consistent with Vera-
Mun˜ oz (1998) and Cloyd (1997), I make no a priori
predictions about di?erences in ability and work
experience among the participants. However, I
introduce these variables to statistically control
for their e?ects on a participant’s task performance.
Work experience (WORKEXP) is measured by
participants’ self-reported relevant part-time work
experience in years. Table 1 shows that WORK-
EXP has a mean of 3.06 years, with a range of zero
to nine years of experience. Most of the participants
(92.7%) reported some work experience.
Similar to prior studies (e.g. Bonner & Lewis,
1990; Bonner, Davis, & Jackson, 1992; Cloyd,
1997; Dearman & Shields, 2005), general ability
(ABILITY) is measured as the number of correct
answers on an 11-item test derived from sections
of prior graduate record examinations (GREs),
with questions on problem solving, analytical abil-
ity, and data interpretation. The mean ABILITY
score reported in Table 1 is seven. Again, the reli-
ability of this test is satisfactory in comparison to
prior studies.
8
Experimental results
In the ?rst subsection below, I describe the
results for the hypothesis on the joint consider-
ations of knowledge and presentation format on
pro?t performance. The second subsection reports
6
One question is dropped because of interpretation problems
as indicated by a few participants.
7
Bonner and Walker (1994) argue that reliability measures
are not appropriate for accounting knowledge constructs.
Knowledge is typically an omnibus construct whereby each
item of a test measures a separate subconstruct of the overall
concept. Therefore, such tests often achieve a low level of
reliability. Depending on the items included, ex post reliability
of the current study’s test range from a = 0.15 to a = 0.47, with
reliability of the full six-item test equal to 0.21. These levels are
similar to those in studies with a comparable number of items.
Tan and Kao (1999) and Tan and Libby (1997) report a levels
ranging from 0.19 to 0.43 for auditing task knowledge or 0.39
for technical accounting knowledge. Dearman and Shields
(2001) obtain levels of 0.29 to 0.45 for similar constructs of cost
accounting content. Bonner and Lewis (1990) do not report a
levels for their knowledge tests. The mean score of 3.84 (64%
correct answers) is similar to that of prior studies (e.g. Cloyd,
1997).
8
The ex post reliability of the items in the current ability test
is a = 0.61. Dearman and Shields (2005) report a reliability of
a = 0.52, while both Bonner et al. (1992) and Cloyd (1997)
achieve a level of a = 0.63 for comparable ability tests. The
mean score of seven, or 63.6% correct answers, is comparable to
the ability test scores reported in prior studies (e.g. Cloyd,
1997).
592 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
some sensitivity checks. Finally, the third subsec-
tion further explores some motivations for the
joint e?ects of knowledge and presentation format.
Hypothesis test
I perform an ANCOVA analysis using PROFIT
as the dependent measure. Consistent with Waller
et al. (1999) and Gupta and King (1997), I employ
a log-linear relation speci?cation. The model
includes the presentation format factor and the
mean split variable KNOW as an additional factor.
Speci?cally, four unique cell means can be created –
lowKNOWor high KNOW, and tabular or graph-
ical format – from which cross-cell comparisons
can be made to better understand the interaction
between knowledge and presentation format. Work
experience and general ability serve as covariates.
The results are reported in Table 2, along with the
detailed tests of the interaction.
Panel A of Table 2 reports two models. Model 1
is an ANCOVA that controls for the main e?ects
of ability and work experience. Model 2 also
includes the interaction of work experience and
knowledge. This additional covariate is included
because the correlations suggest a strong e?ect of
work experience on the pro?t error for less knowl-
edgeable decision makers (r = À0.459, p < 0.03),
while for high-knowledge decision makers this
e?ect is insigni?cant (r = 0.0267, p = 0.88). Fol-
lowing Libby (1995), who suggests potential inter-
actions between indirect expertise (acquired via
work experience) and knowledge, Model 2 statisti-
cally controls for this e?ect.
9
The interaction between knowledge and presen-
tation format is signi?cant at the 5% level in both
models. The test provides strong evidence of an
e?ect reversal (the interaction is signi?cant; the
main factors are either not signi?cant or only mar-
ginally signi?cant). In line with my hypothesis, the
performance of decision makers with a high (low)
degree of accounting knowledge is higher (lower)
with tabular ABC presentation formats than with
graphical presentation formats. The plots of the
interaction in Panel B of Table 2 con?rm that
the pro?t deviation for high-knowledge decision
makers is signi?cantly larger for a graphical
ABC format than for a tabular ABC format
(p = 0.047 and p = 0.057 in Models 1 and 2,
respectively). Conversely, participants with a low
level of cost accounting knowledge perform better
when using graphics rather than tables. This e?ect
is signi?cant in Model 2 (p = 0.056), which con-
trols for the fact that a low-knowledge decision
maker bene?ts from work experience (W · K
interaction). The insigni?cant e?ect of presenta-
tion format in Model 1 (p = 0.108) also becomes
signi?cant (p = 0.097) when the last trial is
excluded from the analysis (end-trial e?ect). In line
with the theory on representational congruence
(see Section ‘‘Theory and hypothesis develop-
ment’’), improved cost accounting knowledge
especially matters with a tabular format (p =
0.017 and p = 0.015 in Models 1 and 2, respec-
tively). Participants with low cost accounting
knowledge presumably have more di?culty in
extracting data from a tabular ABC format due
to a lack of appropriate internal ABC schemata.
Conversely, accounting knowledge is not signi?-
cant under a graphical ABC format (p > 0.22 and
p > 0.15 in Models 1 and 2, respectively) because
a graphical format is bene?cial for the low-knowl-
edge decision maker but detrimental for users with
more knowledge.
Note that the ?ndings do not suggest a di?er-
ence in information content across formats.
Although bar charts do not capture all the items
of a table, low-knowledge users with graphs in
my experiment can realize the same pro?t as their
knowledgeable counterparts that use a table (i.e.
the two cells with the best performance). P-values
of this comparison are not signi?cant (p > 0.28 and
9
Libby (1995) and Tan and Kao (1999) also suggest that
interactions of knowledge and ability may a?ect performance
(especially when the task is very complex and accountability is
high, see Tan & Kao, 1999). Tan and Libby (1997) further study
the e?ects of ability and knowledge at di?erent levels of
experience. Accordingly, I perform an ANCOVA analysis
adding all the possible interactions of ability, knowledge, and
work experience; the results are not qualitatively altered in that
the P · K interaction still remains signi?cant (while the extra
covariates are not signi?cant). Although theory does not
provide guidance about the interactions of ability, work
experience, and presentation format, I also include these as
covariates in a further test. The P · K interaction remains
signi?cant (the covariates are again not signi?cant).
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 593
p > 0.40 in Models 1 and 2, respectively; compari-
son not shown in Panel B of Table 2). This sug-
gests that the graphical format maintains the
relevant items that are crucial for pro?t improve-
ment. Furthermore, I would probably not ?nd a
negative pro?t e?ect for tabular formats with
low-knowledge decision makers if the extra items
of a table revealed new information.
Sensitivity checks
Results remain similar when I use the log of
DECISION as a dependent variable. The interac-
tion of presentation format and knowledge remain
signi?cant (p = 0.07 in Model 1; p = 0.04 in Model
2). Simple presentation format e?ects at high or
low levels of knowledge corroborate the e?ect
reversal: low-knowledge participants perform bet-
ter with graphs than with tables (smallest p =
0.05, cf. p = 0.06 for the pro?t deviation) whereas
the opposite is true for high-knowledge partici-
pants (smallest p = 0.08, cf. p = 0.05 for the pro?t
deviation).
10
I also test whether the e?ect of the
interaction between presentation format and
knowledge (P · K) on pro?t persisted over trials.
A repeated measure test (on the log of pro?t in
each trial) does not reveal any signi?cant three-
way interaction among trial, presentation format,
and knowledge (F = 1.26, p > 0.27 in Model 1;
F = 1.13, p > 0.34 in Model 2). Conditional means
per trial con?rm that tables are always superior to
graphical formats for the high-knowledge decision
maker, where the opposite holds true for the low-
knowledge decision maker.
Additionally, I explore the interactive e?ect of
presentation format and knowledge on perfor-
mance (i.e. log of pro?t) by using the actual score
on the knowledge test (KNOWSCORE) as a con-
tinuous variable into two regressions that use the
same control variables as in the previous section
(Model 1 includes work experience and ability;
Model 2 additionally includes the interaction of
work experience and knowledge). I follow the pro-
cedure of Aiken and West (1991) whereby presen-
tation format is dummy-coded (tabular format =
0; graphical format = 1), and knowledge is mean-
centered in both main e?ect and the multiplicative
terms to avoid multicollinearity problems (see
Cloyd, 1997; Tan & Kao, 1999). Consistent with
my hypothesis, the parameter estimate for the
interaction term of presentation format and
knowledge is signi?cant in Models 1 (b = 0.191;
p = 0.05) and 2 (b = 0.186; p = 0.06).
I determine the nature of this signi?cant interac-
tionby calculating simple e?ects of presentationfor-
mat at one standard deviation above and below the
mean of knowledge (Aiken &West, 1991; Pedhazur,
1982).
11
For participants with high-knowledge
(mean plus one standard deviation), presentation
format has a positive sign signifying higher pro?t
deviations (or lower performance) under a graphi-
cal format than under a tabular format. The e?ect
is signi?cant in Model 1 (b = 0.278, p = 0.05) and
Model 2 (b = 0.229, p = 0.08). For participants with
low-knowledge (mean less one standard deviation),
the e?ect of presentation format is negative in both
models, consistent with the prediction that graphi-
cal formats lead to better performance (lower pro?t
deviation), although the e?ect is only signi?cant in
Model 2 (b = À0.216, p = 0.08; cf. Model 1:
b = À0.177, p = 0.13).
Panel A of Table 3 plots regression lines for
Models 1 and 2 (predicted values of the log of
pro?t were transformed to pro?t deviations).
Panel B of Table 3 displays the simple e?ects and
the signi?cance levels (p-values) of presentation
format at each level of knowledge, which is equiv-
alent to establishing regions of signi?cance (see
Aiken & West (1991, p. 132), for a more detailed
description). The results in Panel B of Table 3
10
The statistical tests increase in power if I consider that none
of the suggested variables has an e?ect on the decision errors for
Customer A (negative adj. R-square for all models, p’s
individual beta’s all >0.24). The e?ect reversal becomes highly
signi?cant when the analysis of ‘decision’ is based on the
resulting sum of decision errors on Customers B and C (the
least and the most pro?table customers, respectively).
11
Consistent with Pedhazur (1982, p. 440), the decomposition
of the interaction in simple e?ects of presentation format at
high or low-knowledge or at each speci?c knowledge score (as
reported in Table 3 below) is based on one-tailed statistics to
minimize the type II error. Based on the theory, I also predict
speci?c directions for the e?ect of presentation format at high
or low levels of accounting knowledge.
594 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
show that graphs as opposed to tables are bene?-
cial for people with a low-knowledge score (score
of 2) and disadvantageous for people with scores
higher than or equal to 5. People with scores near
the points of intersection (scores of 3 or 4; see
Panel A of Table 3; 3.576 in Model 1; 3.816 in
Model 2) are not a?ected by presentation format.
Supplementary analysis
The supplementary analyses in Table 4 explore
some reasons for the ?nding that graphical for-
mats are bene?cial to the low-knowledge decision
maker, whereas tables result in better performance
for a knowledgeable decision maker.
I ?rst study the e?ects of presentation format
and knowledge (KNOW) on self-perceived com-
plexity, as indicated by participants on a ?ve-point
Likert scale in the debrie?ng questionnaire. The
?ndings in Panel A of Table 4 show that users with
a low level of cost accounting knowledge perceive
the task to be more complex under tabular formats
than under graphical formats (p = 0.02). The pre-
sentation format does not a?ect the cognitive
Table 3
Sensitivity checks with KNOWSCORE (actual score) as covariate in Models 1 and 2
Panel A: Plots of the regression lines in Models 1 and 2 and p-value of P · K interaction
a
200000
300000
400000
500000
600000
700000
800000
900000
1000000
2 3 4 5 6
Knowledge
P
r
o
f
i
t
d
e
v
i
a
t
i
o
n
Point of intersection:
3,576
Graph
Tab
mean
Model 1: Interaction P x K (p=.05**)
200000
300000
400000
500000
600000
700000
800000
900000
1000000
2 3 4 5 6
Knowledge
P
r
o
f
i
t
d
e
v
i
a
t
i
o
n
Point of intersection:
3,816
Model 2: Interaction P x K (p=.06*)
mean Graph
Tab
Panel B: E?ect of presentation format at each level of the knowledge score
b
Model 1: Estimate, sign. level, e?ect size Model 2: Estimate, sign. level, e?ect size
Knowledge score b (format) p-value Di?erence in pro?t b (format) p-value Di?erence in pro?t
Dev. (graph-tab) Dev. (graph-tab)
2 À0.30184 0.08
*
À228545.2 À0.33777 0.05
*
À252933.0
3 À0.11039 0.21 À79108.9 À0.15172 0.13 À107341.3
4 0.08107 0.24 55201.4 0.03433 0.38 22988.3
5 0.27252 0.05
**
176698.1 0.22038 0.09
*
140315.3
6 0.46397 0.03
**
287388.1 0.40643 0.05
**
246599.2
a
The regressions use the LN(pro?t) as the dependent and Models 1 and 2 use similar control variables as reported in Table 2.
Consistent with Aiken and West (1991), knowledge was mean centered (Mean: 3.8364). Predicted values were calculated for each
knowledge score by subtracting the mean from the score. Predicted values of LN(pro?t) are transformed to actual deviations.
b
Following Aiken and West (1991, p. 132–133), I performed ?ve regressions of the following structure:
ln(pro?t) = b1 · PF + b2 · K + b3 · KP + (covariates of either Model 1 or Model 2), whereby k was equal to knowledge-2, knowl-
edge-3, . . ., knowledge-6 to evaluate the factor presentation format at each level of the knowledge score. Panel B displays the parameter
estimate of presentation format, the signi?cance level (one-tailed, see Pedhazur, 1982) and predicted di?erences in pro?t deviation
between graphs and tables (e?ect size). A negative (positive) signs denote that graphical formats lead to lower (higher) pro?t deviations
in comparison to tables.
**
,
*
indicates signi?cance at the 5 or 10% level.
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 595
burden for the high-knowledge subgroup (p >
0.45). An ANOVA-test of the measure of complex-
ity (not reported in Panel A of Table 4) suggests
that the e?ect of presentation format depends on
knowledge (P · K interaction: F = 6.49 and p =
0.02; main e?ects, p > 0.11). Consistent with the
theory on representational congruence (see Section
‘‘Theory and hypothesis development’’), I presume
that participants with low knowledge lack the
appropriate schemata to make a ?t with a tabular
format and hence ?nd the task to be more complex
than their counterparts who receive a graphical
format.
A second test analyzes the amount of time that
participants spend reviewing the task description
and the initial cost report (‘‘information search’’)
prior to making their decisions (Panel B of Table
4). Similar to Cloyd (1997), I control for a partic-
ipant’s reading ability (reading speed). Results are
analyzed via an ANCOVA in a similar way as
Table 4
Supplementary analyses of the level of perceived complexity and information search
Panel A: Self-perceived complexity (?ve-point Likert scale) and the e?ect of format (tables versus graphs) per knowledge category
a
Mean Median SD
Low-Knowledge subgroup (N = 23)
Table (N = 15) 3.20 3.00 0.86 Test of di?erence between Table-Graph
Graph (N = 8) 2.25 2.00 0.71 t-value: 2.67; p < .02
**
(v
2
: 5.598, p < .02
**
)
High-Knowledge subgroup (N = 32)
Table (N = 13) 2.31 2.00 0.95 Test of di?erence between Table-Graph
Graph (N = 19) 2.53 3.00 0.70 t-value: À.075; p > .45 (v
2
: 0.437, p > .51)
Panel B: ANCOVA analysis on information search (SEARCH) prior to the decision trials
b
Source of variation Mean square F-stat p-value
Factors
523.2
749.3
661.3
595.8
400
450
500
550
600
650
700
750
800
LOW HIGH
.062*
.147
.164
.001***
Graphic
Table
Presentation (P) 728 0.03 0.863
Know (K) 260313 10.69 0.002
***
P · K 80141 3.29 0.076
*
Covariates
Reading speed 1010714 41.52 0.001
***
R
2
= .555; N = 55
a
Analysis of the item ‘‘I considered the task as extremely complex’’. Due to the word ‘‘extremely’’, the means are not that high. The
tests employ a t-test assuming equal variances (a non-parametric Kruskal-Wallis test is reported between brackets). Note that for the
low-knowledge subgroup the number of users considering the task as complex (scores P3) dropped from 80% for tabular ABC-
reports to 37.5% for graphical ABC-reports. For the high-knowledge subgroup, this number increased (from 46% to 53%). A further
ANOVA-test also reveals a signi?cant P · K interaction: F = 6.49; p < 0.02 (main e?ects: p > 0.11).
b
The dependent variable of this ANCOVA is the time (recorded in seconds) that participants spent on the problem description that
contained the initial ABC report either in a graphical or tabular format (see Fig. 1). Subjects were instructed to thoroughly review this
information. Similar to Cloyd (1997), the test controls for reading speed, which is measured by the participants’ total time in the full
experiment. The ?gure on the right gives a detailed analysis of the P · K interaction of the model (i.e. similar test as reported in Panel B
of Table 2). A higher score represents more search time prior to the problem (scores evaluated at covariate reading speed = 2263.8);
***
,
**
,
*
indicates signi?cance at the 1, 5 or 10% level.
596 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
reported in Table 2. Panel B of Table 4 shows that
knowledgeable decision makers spent more time
on information search (KNOW; p = 0.002) and
there is also weak evidence that the e?ect of pre-
sentation format on information search depends
on knowledge (P · K interaction is marginally sig-
ni?cant; p = 0.076). As shown in the ?gure in
Panel B of Table 4, the di?erence between tables
and graphical formats is marginally signi?cant
(p = 0.062) for high-knowledge participants; tables
tend to increase their information search. The
e?ect of format on search time is not signi?cant
for low-knowledge decision makers (p = 0.147).
Consistent with theory, this ?nding could imply
that due to an analytical focus of high-knowledge
decision makers, tables that provide an analytical
view of the data better assist their information
search. Graphics display the same data at a glance
and are less appropriate for knowledgeable deci-
sion makers who approach the problem in an ana-
lytical way (and therefore search time is reduced).
The ?gure in Panel B of Table 4 also reveals that
low- and high-knowledge participants have the
same search time under graphics (p = 0.164);
knowledge level only makes a di?erence with
tables (p = 0.001).
Although other explanations are possible
besides the ones explored here, these metrics indi-
cate that low-knowledge decision makers seem to
bene?t from graphics because of a reduction in
complexity, whereas sophisticated users likely ben-
e?t from tables because tables enhance their infor-
mation search.
12
Conclusion
In many ?rms, managers with various levels of
cost accounting knowledge make cost-based deci-
sions on the basis of accounting information sys-
tems that provide ‘‘easy-to-understand’’ graphical
rather than tabular cost reports. Yet, the perfor-
mance e?ects of di?erent types of presentation for-
mats in relation to varied levels of cost accounting
knowledge have received scant attention in the lit-
erature. This study experimentally investigates
these joint e?ects in the context of a relatively com-
plex task that requires decision makers to make
customer-speci?c price and resource allocation
decisions that a?ect pro?tability.
This study advances our understanding of man-
agerial decision-making in several ways. First, I
?nd that di?erent formats in?uence pro?tability
di?erently, with the direction of impact contingent
upon the user’s level of accounting sophistication.
After controlling for di?erences in ability and
experience, I show that cost report users with a
low level of cost accounting knowledge achieve
higher pro?ts when ABC data are displayed in a
graphical format rather than in a tabular format.
More surprisingly, the opposite is true for sophis-
ticated users; their pro?ts are higher given a tabu-
lar format.
Second, this ?nding has important practical
implications. There is no unique way to present
cost data to managers. To extract maximum
potential for improved decision-making, account-
ing information systems may need to adjust a cost
report’s format to the user’s level of accounting
sophistication. In most ?rms, many non-accoun-
tants make use of accounting information systems
(Birnberg, 2000; Shields, 1995; Mauldin & Ruc-
hala, 1999). These non-accountants are better
served by graphical presentations of accounting
information. In contrast, accountants are likely
to receive appropriate access to the data by means
of traditional tables. In a broader context, with
respect to the debate on how non-professional ver-
sus professional investors acquire ?rm-speci?c
information (Maines & McDaniel, 2000), ?rms
can capitalize on my ?ndings and adjust the for-
mat of their disclosures to the level of accounting
knowledge of their investors.
12
I further explore these assertions by testing how these
metrics relate to pro?t. For each knowledge category, I regress
the measures for ‘search’ and ‘complexity’ together with the
e?ects of ability and work experience (control variables of Table
2) on the pro?t deviation. While overall pro?t is not a?ected,
analysis of the ?rst-trial pro?t deviation produces interesting
?ndings (it is logical to assume that only ?rst-trial pro?ts are
a?ected by information search; subsequent decisions are
anchored on and adjusted from prior decisions). For the
high-knowledge subgroup, ‘information search’ tends to reduce
the ?rst-trial pro?t deviation (t = À1.75, p = 0.092); the e?ect
of ‘complexity’ is not signi?cant (p > 0.19). For less knowl-
edgeable decision makers, more ‘complexity’ tends to increase
the ?rst-trial pro?t deviation (t = 1.76, p = 0.096), while
‘information search’ is not signi?cant (p > 0.68).
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 597
Third, from a theoretical viewpoint, this study
shows that knowledge is an important attribute that
explains when particular presentation formats are
likely to provide pro?t improvement. While most
research onpresentation formats has produced con-
?icting evidence on several aspects of the decision
task (Vessey, 1991), my ?ndings indicate that a
focus on a user’s characteristics may explain part
of these mixed results. I show that researchers can
expect adverse performance e?ects when the exter-
nal presentation format is not well aligned with a
decision maker’s mental model or decision style.
To focus on the joint e?ects of knowledge and
presentation format, I hold the level of task com-
plexity constant. Future research could vary task
complexity (e.g. Tan & Kao, 1999; Tan et al.,
2002) and explore how the joint e?ects of knowl-
edge and presentation format extend to di?erent
levels of task complexity. It might also be interest-
ing to allow decision makers to choose their for-
mat rather than simply assigning them graphical
or tabular formats, in order to study the types of
decision makers who are more likely to access
graphical rather than tabular representations.
Other dimensions of managerial expertise (junior-
versus senior-level managers) or job functions
(top- versus lower-level managers) could lead to
preferences for a speci?c type of format.
Finally, a potential limitation of the study is
that graphical formats did not contain all the items
of the table. Nevertheless, this probably did not
a?ect my ?ndings because the extra items con-
tained little extra information content for the
objective of maximizing pro?ts. Yet, some deci-
sions, like the evaluation of division managers,
involve a more subjective judgment and further
studies could explore whether extra (redundant)
cues do matter here. Evaluations may di?er when
managers have access to only a few or many met-
rics that are all equally informative about perfor-
mance. Again, one could explore whether
expertise of the evaluator matters herein. Also,
the practice of presenting these metrics to evalua-
tors strongly varies across ?rms. Sometimes bal-
anced scorecards use red, yellow, and green
ratings for their performance metrics (Malina &
Selto, 2001). In any event, further research that
links such variations in practice to a decision
maker’s expertise (Sprinkle, 2003) may provide
guidance for how ?rms can optimize the many
?ows of accounting information to internal and
external users of that information.
Acknowledgements
I want to thank Jan Bouwens, Lynn Hannan,
Laurence van Lent, Paula van Veen-Dirks, Luk
Warlop, and two anonymous reviewers for their
insightful comments. This study also bene?ts from
comments of participants at the Management
Accounting Section’s midyear meeting in Phoenix
(2005), the 4th conference on new directions in
management accounting of the European Institute
for Advanced Management Studies in Brussels
(2004), and the accounting seminar at Tilburg Uni-
versity (2004).
Appendix
This appendix describes the underlying func-
tions, parameters, and cost allocation assumptions
in my experiment. Customer demand (Equations 1
and 2 in Panel A of Table A1) is based on a com-
mon sales response function (Mantrala, Prabhak-
ant, & Zoltners, 1992): the price (p
i
) in?uences
potential demand, whereas the amount allocated
to sales visits (x
i
) determines how much demand
is realized. The gross contribution for a customer
is equal to revenues minus a ?xed portion of reve-
nues: d
i
represents the products’ purchase cost
(Equation 3). Given my focus on cost-based deci-
sion-making, I introduce a complex customer cost
function (Equation 4) with the following elements:
a ?xed cost (FC
i
) per customer (e.g. storage space);
a nonlinear distribution function (u
i
qe
i
+ v
i
qe
1:5
i
Þ
with costs increasing more than proportionally
with volume (Klincewicz, 1990); and a cost func-
tion for sales visits (x
i
w
i
), with sales visits (x
i
)
acquired at a speci?c rate (w
i
). Consistent with
practice, the set of customers is heterogeneous
(Kaplan & Narayanan, 2001). Compared to Cus-
tomers A and C, Customer B is a high-cost-to-
serve customer, consuming more ?xed costs
(FC), more resources in distribution (u
i
and v
i
),
598 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
and more in sales visits (w
i
). Customer pro?tability
(Equation 5) equals gross margin minus customer-
speci?c costs. Via substitution, one can rewrite
Equation 5 entirely in terms of the participant’s
decision choices (p
i
and x
i
). There is a unique opti-
mum for each customer (see the note to Table A1).
Given the focus on cost-based decisions, optima
are determined in large part by the cost variations
across customers. Firm-wide pro?t is equal to the
sum of customer pro?ts (Equation 6).
Both the graphical format as well as the tabu-
lar format present ABC-driven customer data in
Table A1
Underlying functions of the experimental task and ABC cost allocation assumptions
a
Panel A: Underlying functions of the experimental task (by substituting all are f(xi, pi))
Panel B: ABC-customer cost pools (CCP) and cost driver assumptions (i = Customer A, B, C)
a
Optimal solutions can be found by solving ?rst order conditions for pi and xi based on the parameter data for customers i = A,B, C
(cost parameters in bold). Optima are P
a
= 115.5; X
a
= 62.0; P
b
= 124.9; X
b
= 49.4; P
c
= 97.6; X
c
= 83.1; ?rm-wide pro?t: 3,036,145.
The ABC-reports (in Fig. 1) use these functions of Panel A. Save for customer costs, I introduced some small errors following common
assumptions of Datar and Gupta (1994) and Christensen and Demski (1995) that ABC seldom re?ects true costs.
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 599
addition to total ?rm-wide pro?t (actual outcome).
To generate the reports, I use the functions of
Panel A, except for the total customer costs, which
I allocate along the assumptions of Panel B of
Table A1. In the ?rst stage, I assign customer costs
to cost pools, and in the second stage, I assign
them to customers on the basis of cost drivers. I
introduce small errors by specifying the driver
for sales visits slightly di?erently from the actual
driver and by introducing linear cost drivers for
a nonlinear distribution function. Consistent with
Datar and Gupta (1994) and Christensen and
Demski (1995), the reports never re?ect ‘‘true’’
costs. Because Customer B requires more of each
cost driver, customer cost and pro?tability ?gures
remain quite accurate.
References
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing
and interpreting interactions. Newbury Park: Sage.
Alba, J. W., & Hutchinson, J. W. (1987). Dimensions of
consumer expertise. Journal of Consumer Research, 13(4),
411–454.
Amer, T. (1991). An experimental investigation of multi-cue
?nancial information display and decision-making. Journal
of Information Systems, 5(2), 18–34.
Anderson, J. R. (1990). Cognitive Psychology and its implica-
tions (third ed.). New York: W.H. Freeman & Co Ltd.
Arnold, V., Collier, P. A., Leech, S. A., & Sutton, S. G. (2004).
Impact of intelligent decision aids on expert and novice
decision-makers’ judgments. Accounting and Finance, 44, 1–26.
Benbasat, I., & Schroeder, R. G. (1977). An experimental
investigation of some MIS Design Variables. MIS quarterly,
1(1), 37–49.
Bierstaker, J. L. (2003). Auditor recall and evaluation of
internal control information: does task-speci?c knowledge
mitigate part-list interference?. Managerial Auditing Journal
18(1/2), 90–99.
Birnberg, J. G. (2000). The role of behavioral research in
management accounting education in the 21st century.
Issues in Accounting Education, 15(4), 713–728.
Bonner, S. E. (1990). Experience e?ects in auditing: the role of
task speci?c knowledge. Accounting Review, 65(1), 72–92.
Bonner, S. E. (1994). A model of the e?ects of audit task
complexity. Accounting, Organizations and Society, 19(3),
213–234.
Bonner, S. E., & Lewis, B. L. (1990). Determinants of auditor
expertise. Journal of Accounting Research, 28, 1–19.
Bonner, S. E., & Walker, P. L. (1994). The e?ects of
instructions and experience on the acquisition of audit
knowledge. Accounting Review, 69(1), 157–178.
Bonner, S. E., Davis, J. S., & Jackson, B. R. (1992). Expertise in
corporate tax planning: The issue identi?cation stage.
Journal of Accounting Research, 30, 1–28.
Bonner, S. E., Libby, R., & Nelson, W. (1997). Audit
category knowledge as a precondition to learning from
experience. Accounting, Organizations and Society, 22(5),
387–410.
Briers, M., Chow, C., Hwang, N., & Luckett, P. (1999). The
e?ects of alternative types of feedback on product-related
decision performance: A research note. Journal of Manage-
ment Accounting Research, 11, 75–92.
Bucheit, S. (2003). Reporting the cost of capacity. Accounting,
Organizations and Society, 28, 549–565.
Chandra, A., & Krovi, R. (1999). Representational congruence
and information retrieval: Towards an extended model of
cognitive ?t. Decision Support Systems, 25, 271–288.
Chi, M., Feltovich, P., & Glaser, R. (1981). Categorization and
representation of physics problems by experts and novices.
Cognitive Science, 5(2), 121–152.
Christensen, J., & Demski, S. J. (1995). The classical founda-
tions of modern costing. Management Accounting Research,
6(2), 13–32.
Cloyd, B. C. (1997). Performance in tax research tasks: The
joint e?ects of knowledge and accountability. Accounting
Review, 72(1), 111–131.
Datar, S. M., & Gupta, M. (1994). Aggregation, speci?cation
and measurement error errors in product costing. Account-
ing Review, 69(3), 567–591.
Davis, L. R. (1989). Report format and the decision makers
task: An experimental investigation. Accounting, Organiza-
tions and Society, 14(5/6), 495–508.
Dearman, D. T., & Shields, M. D. (2001). Cost knowledge and
cost-based judgment performance. Journal of Management
Accounting Research, 13, 1–18.
Dearman, D. T., & Shields, M. D. (2005). Reducing accounting
?xation: Determinants of cognitive adaptation to variation
in accounting method. Contemporary Accounting Research,
22(2), 351–384.
Desanctis, G., & Jarvenpaa, S. L. (1989). Graphical presenta-
tion of accounting data for ?nancial forecasting: An
experimental investigation. Accounting, Organizations and
Society, 14(5/6), 509–525.
Drake, A. R., Haka, S. F., & Ravenscroft, S. P. (1999). Cost
system and incentive structure e?ects on innovation,
e?ciency and pro?tability in teams. Accounting Review,
74(2), 323–345.
Frownfelter-Lohrke, C. (1998). The e?ects of di?ering infor-
mation presentations of general purpose ?nancial state-
ments on users’ decisions. Journal of Information Systems,
12(2), 99–107.
Ganzach, Y. (1993). Predictor representation and prediction
strategies. Organizational Behavior and Human Decision
Processes, 56, 190–212.
Gupta, M., & King, R. R. (1997). An experimental investiga-
tion of the e?ect of cost information and feedback on
product cost decisions. Contemporary Accounting Research,
14, 99–127.
600 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
Harvey, N., & Bolger, F. (1996). Graphs versus tables: E?ects
of data presentation format on judgmental forecasting.
International Journal of Forecasting, 12, 119–137.
Haynes, C. M., & Kachelmeier, S. J. (1998). The e?ects of
accounting contexts on accounting decisions: A synthesis
of cognitive and economic perspectives in accounting
experimentation. Journal of Accounting Literature, 17,
97–136.
Hwang, M. I. (1995). The e?ectiveness of graphic and tabular
presentation under time pressure and task complexity.
Information Resource Management Journal, 8(3), 25–31.
Jarvenpaa, S. L. (1989). The e?ects of task demands and
graphical format on information processing strategies.
Management Science, 35(3), 285–303.
Kaplan, R. S., & Narayanan, V. G. (2001). Measuring and
managing customer pro?tability. Journal of Cost Manage-
ment(September/October), 5–15.
Klincewicz, J. G. (1990). Solving a freight transport problem
using facility location techniques. Operations Research,
38(2), 99–109.
Libby, R. (1981). Accounting and Human Information Process-
ing. Englewood Cli?s, NJ: Prentice-Hall.
Libby, R. (1995). The role of knowledge and memory in audit
judgment. In R. Ashton & A. Ashton (Eds.), Judgment and
Decision-Making Research in Accounting and Auditing.
Cambridge, UK: Cambridge University Press.
Libby, R., & Luft, J. (1993). Determinants of judgment
performance in accounting settings: Ability, knowledge,
motivation, and environment. Accounting, Organizations
and Society, 18(5), 425–450.
Lucas, H. C. (1981). An experimental investigation of the use of
computer-based graphics in decision-making. Management
Science, 27(7), 757–768.
Luft, J. (1994). Bonus and penalty incentives: Contract choice
by employees. Journal of Accounting and Economics, 18,
181–206.
Mackay, D. B., & Villarreal, A. (1987). Performance di?erences
in the use of graphic and tabular displays of multivariate
data. Decision Sciences, 18, 535–546.
Maines, L. A., & McDaniel, L. S. (2000). E?ects of compre-
hensive-income characteristics on non-professional inves-
tors’ judgments: The role of ?nancial-statement
presentation format. Accounting Review, 75(2), 179–207.
Malina, M. A., & Selto, F. H. (2001). Communicating and
controlling strategy: An empirical study of the e?ectiveness
of the balanced scorecard. Journal of Management Account-
ing Research, 13, 47–90.
Mantrala, M. K., Prabhakant, S., & Zoltners, A. (1992). Impact
of resource allocation rules on marketing investment-level
decisions and pro?tability. Journal of Marketing Research,
29(May), 162–175.
Mauldin, E. G., & Ruchala, L. V. (1999). Towards a meta-
theory of accounting information systems. Accounting,
Organizations and Society, 24, 317–331.
Mooney, J. L., Rogers, C., & Wright, H. (2000). A guide to
selecting cost management software. Journal of Corporate
Accounting & Finance, 11(3), 19–27.
Moriarity, S. (1979). Communicating ?nancial information
through multidimensional graphics. Journal of Accounting
Research, 17(1), 205–224.
O’Donnell, E., Koch, B., & Boone, J. (2005). The in?uence of
domain knowledge and task complexity on tax profession-
als’ compliance recommendations. Accounting, Organiza-
tions and Society, 30(2), 145–165.
Painton, S., & Gentry, J. W. (1985). Another look at the impact
of information presentation format. Journal of Consumer
Research, 12(2), 240–244.
Pedhazur, E. J. (1982). Multiple regression in behavioral research
(2nd ed.). New York, NY: Holt, Rinehart, and Winston.
Remus, W. (1987). A study of graphical and tabular displays
and their interaction with environmental complexity. Man-
agement Science, 33(9), 1200–1204.
Rose, J. M., & Wolfe, C. J. (2000). The e?ects of system design
alternatives on the acquisition of tax knowledge from a
computerized tax decision aid. Accounting, Organizations
and Society, 25, 285–306.
Rose, J. M., Douglas, R. F., & Rose, A. M. (2004). A?ective
responses to ?nancial data and multimedia: The e?ects of
information load and cognitive load. International Journal
of Accounting Information Systems, 5, 5–24.
Shields, M. D. (1995). An empirical analysis of ?rms’ imple-
mentation experiences with activity-based costing. Journal
of Management Accounting Research, 7, 148–166.
So, S., & Smith, M. (2002). Colour graphics and task
complexity in multivariate decision-making. Accounting,
Auditing and Accountability Journal, 15(4), 565–592.
So, S., & Smith, M. (2004). Multivariate decision accuracy and
the presentation of accounting information. Accounting
Forum, 283–305.
Spilker, B. C. (1995). The e?ects of time pressure and
knowledge on key word selection behavior in tax research.
Accounting Review, 70, 514–528.
Sprinkle, G. B. (2003). Perspectives on experimental research in
managerial accounting. Accounting, Organizations and Soci-
ety, 28(2-3), 287–318.
Stock, D., & Watson, C. J. (1984). Human judgment accuracy,
multidimensional graphics, and human versus models.
Journal of Accounting Research, 22, 192–206.
Stone, D. N., Hunton, J. E., & Wier, B. (2000). Succeeding in
managerial accounting. Part 1: Knowledge, ability and
rank. Accounting, Organizations and Society, 25, 697–
715.
Sullivan, J. J. (1988). Financial presentation format and
managerial decision-making: Tables versus graphics. Man-
agement, 2(2), 194–216.
Tan, H. T., & Libby, R. (1997). Tacit managerial versus
technical knowledge as determinants of audit expertise in
the ?eld. Journal of Accounting Research, 35(1), 97–113.
Tan, H. T., & Kao, A. (1999). Accountability e?ects on
auditors’ performance: the in?uence of knowledge, problem
solving ability, and task complexity. Journal of Accounting
Research, 37(1), 209–223.
Tan, H. T., Ng, B. P., & Mak, W. Y. (2002). The e?ects of task
complexity on auditors’ performance: The impact of
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 601
accountability and knowledge. Auditing: A Journal of
Practice and Theory, 21(2), 81–95.
Tufte, E. (2001). The visual display of quantitative information
(second ed.). Cheshire, CT: Graphics Press.
Umanath, N. S., & Vessey, I. (1994). Multiattribute data
presentation and human judgment: A cognitive ?t perspec-
tive. Decision Sciences, 25, 795–823.
Vera-Mun˜ oz, S. C. (1998). The e?ects of accounting knowledge
and context on the omission of opportunity costs in
resource allocation decisions. Accounting Review, 73(1),
47–72.
Vera-Mun˜ oz, S. C., Kinney, W. R., & Bonner, S. E. (2001). The
e?ects of domain experience and task presentation format
on accountants’ information relevance assurance. Account-
ing Review, 76(3), 405–429.
Vessey, I. (1991). Cognitive ?t: A theory-based analysis of the
graphs versus tables literature. Decision Sciences, 22(2),
219–240.
Waller, W., Shapiro, B., & Sevcik, G. (1999). Do cost-based
pricing biases persist in laboratory markets? Accounting,
Organizations and Society, 24(8), 717–739.
Wood, S. L., & Lynch, J. G. (2002). Prior knowledge and
complacency in new product learning. Journal of Consumer
Research, 29(3), 416–426.
Wright, W. F. (1995). Superior loan collectibility judgments
given graphical displays. Auditing: A Journal of Theory and
Practice, 144–154.
Yates, J. (1985). Graphics as a managerial tool. A case study of
DuPont’s use of graphs in the early twentieth century.
Journal of Business Communications, 22, 5–33.
602 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
doc_998589789.pdf
Most studies on cost-based decision-making examine the profit impact of cost reports that rely on different methods
to allocate costs. In practice, firms’ cost reports often employ the same cost allocation method with subtle variations in
the way that the cost data are presented
The interplay between cost accounting knowledge
and presentation formats in cost-based decision-making
Eddy Cardinaels
*
Tilburg University, Department of Accountancy, P.O. Box 90153, 5000 LE Tilburg, The Netherlands
Abstract
Most studies on cost-based decision-making examine the pro?t impact of cost reports that rely on di?erent methods
to allocate costs. In practice, ?rms’ cost reports often employ the same cost allocation method with subtle variations in
the way that the cost data are presented. This paper examines experimentally the pro?t impact of a cost report’s pre-
sentation format in relation to a decision maker’s level of cost accounting knowledge. Using a customer pro?tability
report prepared using activity-based costing and presented in either a tabular or a graphical format, participants ana-
lyze a complex pricing and resource allocation task that a?ects ?rm pro?tability. The results suggest a strong relation
between presentation format and cost accounting knowledge. Speci?cally, decision makers with a low level of cost
accounting knowledge attain higher pro?ts when they use a graphical format in comparison to a tabular format. More
surprisingly, graphs (versus tables) have an adverse e?ect on pro?ts for users with a high level of cost knowledge. This
result has broad implications: in order to facilitate the decisions of a variety of users of accounting data (e.g. managers,
external investors, etc.), ?rms may need to adapt the presentation format of their accounting data to the level of
accounting sophistication of the users.
Ó 2007 Elsevier Ltd. All rights reserved.
Introduction
The performance e?ects of di?erent types of
cost reports (variable versus absorption costing,
volume-based versus activity-based costing) in
relation to a number of contextual variables is
the key focus of several previous studies on cost-
based decision-making (e.g. Briers, Chow, Hwang,
& Luckett, 1999; Drake, Haka, & Ravenscroft,
1999; Gupta & King, 1997; Waller, Shapiro, &
Sevcik, 1999). This paper presents the results of
an experiment conducted to study how di?erent
representations of identical underlying cost data
a?ect cost-based decision-making and ?rm pro?t-
ability. Speci?cally, I ?nd unique evidence suggest-
ing that the pro?t impact of tabular versus
graphical representations of activity-based costing
0361-3682/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2007.06.003
*
Tel.: +31 13 4668231; fax: +31 13 4668001.
E-mail address: [email protected]
Available online at www.sciencedirect.com
Accounting, Organizations and Society 33 (2008) 582–602
www.elsevier.com/locate/aos
(ABC) data is dependent upon the accounting
sophistication of the user, i.e. his or her level of
cost accounting knowledge.
Studying the joint e?ects of presentation format
and a decision maker’s level of accounting knowl-
edge is important for several reasons. First, di?er-
ent managers clearly have di?erent levels of
accounting knowledge (Dearman & Shields,
2001; Stone, Hunton, & Wier, 2000). Firms’ infor-
mation systems provide managers with reports
that range from traditional tabular formats to
graphical displays (So & Smith, 2002; Sullivan,
1988). Many managers use elaborate cost reports
for their daily decisions. Others encourage the
use of ‘‘easy-to-understand’’ graphs (Mooney,
Rogers, & Wright, 2000; Remus, 1987; Yates,
1985) in the belief that a graphical representation
(versus a tabular format) makes cost data accessi-
ble to all members of the ?rm, irrespective of their
level of accounting knowledge. In spite of evidence
of di?erential managerial knowledge, however, the
extant literature does not indicate how managerial
decision-making and, in turn, ?rm pro?ts are
a?ected (Sprinkle, 2003) when information is pre-
sented in di?erent report formats to decision mak-
ers with di?erent levels of accounting knowledge
(Haynes & Kachelmeier, 1998; Libby, 1981).
Second, recent studies in accounting only
address the separate e?ects of expertise and report
format (Sprinkle, 2003), though such studies ?nd
that both a?ect cost-based decision-making. For
instance, Dearman and Shields (2001) show that
the level of a manager’s cost accounting knowl-
edge is linked with the ability to correct for vol-
ume-based cost bias, and Bucheit (2003) shows
that investment decisions change when cost
reports explicitly contain the cost of capacity
(compared to reports that do not). There is also
some evidence that suggests the interaction of the
two variables. For instance, Vera-Mun˜ oz, Kinney,
and Bonner (2001) show that the impact of alter-
native task representations depends upon the deci-
sion maker’s expertise. However, they employ only
tabular reports based on either historical earnings
or historical cash ?ows. Thus, the impact of tabu-
lar versus graphical representation of identical
data in relation to a decision maker’s knowledge
is an open question.
Finally, although there is a strong belief that
decision makers should bene?t from graphical rep-
resentation (Harvey & Bolger, 1996), research that
compares the relative impact of graphical versus
tabular formats remains inconclusive (Vessey,
1991). In an attempt to resolve the controversy, a
few studies suggest looking at individual di?er-
ences among the users of information (Amer,
1991; Chandra & Krovi, 1999; Ganzach, 1993).
The current study sheds light on this debate by
testing whether accounting knowledge as a mana-
gerial characteristic helps to explain when certain
report formats are associated with stronger perfor-
mance than others in a cost-based decision task.
To investigate these joint considerations, I con-
duct an experiment with presentation format as
the between subjects factor. I measure a partici-
pant’s level of accounting knowledge, in addition
to some common control variables, using research
instruments suggested in prior studies (Bonner &
Lewis, 1990; Cloyd, 1997; Dearman & Shields,
2005). I create a complex task in which the partic-
ipant’s realized pro?t depends upon both price and
resource allocation decisions for a heterogeneous
set of customers. All participants receive ABC-dri-
ven customer pro?tability data, presented in either
a tabular or a graphical format (multicolored bar
charts and trend charts). I measure pro?t perfor-
mance objectively as the di?erence between a par-
ticipant’s realized pro?t and the maximum pro?t
that could be achieved in performing the task.
After controlling for di?erences in ability and
work experience, I ?nd evidence of an e?ect reversal
across knowledge levels: decision makers with a low
level of cost accounting knowledge perform better
with a graphical ABC format, and decision makers
with a high level of cost knowledge obtain superior
pro?ts with a tabular ABC format. Further evi-
dence indicates that graphical formats tend to
reduce task complexity for a low-knowledge deci-
sion maker, whereas tables support the information
search of a more knowledgeable user. This result
provides important theoretical and practical
insights, suggesting that (1) cost accounting knowl-
edge is a crucial managerial characteristic that
should be taken into account when a ?rm presents
cost reports to a decision maker, and (2) managerial
cost accounting knowledge and data representation
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 583
jointly determine the direction on ?rmpro?tability.
This has broad implications for the many informa-
tion ?ows within the ?rm. To convey accounting
data to both internal and external parties (e.g. man-
agers, investors), an exclusive focus on one format
is probably not advisable. Instead, ?rms can facili-
tate decision-making by adapting the presentation
format to the level of the accounting sophistication
of their intended audience.
Prior research
Although accounting systems vary greatly with
respect to howthey package the same kind of infor-
mation, systematic research on this issue remains
scarce (Haynes & Kachelmeier, 1998; Luft, 1994).
Most prior work examines the pro?t impact of cost
reports that are based on di?erent assumptions,
such as activity- versus volume-based costing or
absorption versus variable costing (e.g. Drake
et al., 1999; Gupta & King, 1997; Waller et al.,
1999). This study is the ?rst to disentangle the dif-
ferential impact of tabular versus graphical repre-
sentations of identical ABC cost data. The unique
contribution of this study is that it links di?erential
information representation to the decision maker’s
level of accounting sophistication. Nevertheless,
most evidence on either the e?ects of presentation
format or accounting knowledge is found in sepa-
rate research streams. In what follows I review this
literature and I provide arguments for why we need
to consider the joint e?ects of knowledge and pre-
sentation formats.
With regard to information representation,
some studies focus on salience e?ects. Bucheit
(2003) ?nds that capacity investment decisions
are suboptimal if cost reports explicitly contain
capacity costs (versus reports that exclude these
costs). Maines and McDaniel (2000) show that
non-professional investors are in?uenced by com-
prehensive income information when it appears
in a separate statement, as opposed to when it is
masked in stockholders’ equity. In contrast to sal-
ience e?ects, evidence on the performance e?ects of
various presentation formats (Vessey, 1991) and
the belief that graphs are superior to tables (Har-
vey & Bolger, 1996) remains inconclusive.
Even research that focuses on task characteris-
tics (both task type and task complexity) to address
this controversy (Jarvenpaa, 1989; Remus, 1987;
Vessey, 1991) o?ers mixed results (Hwang, 1995).
Vessey (1991) argues that the representation of
the problem should match the type of task (cogni-
tive ?t) to enhance decision-making. Graphical for-
mats present spatial information (Vessey, 1991):
they tend to emphasize relationships in the data
and provide a holistic view by presenting the data
at a glance. Such formats are most appropriate
when decision makers have to compare alternatives
or tasks that require integration of data (Vessey,
1991). In contrast, tables present symbolic infor-
mation: by emphasizing discrete values, they pro-
vide an analytical view and facilitate the
extraction of speci?c values (item-by-item evalua-
tion) and the evaluation of changes in variables
(Mackay & Villarreal, 1987; Vessey, 1991). Tables
should be more appropriate when the task requires
the manager to extract and act on discrete data val-
ues (Vessey, 1991). Nonetheless, the evidence on
task type remains inconclusive. For instance, Amer
(1991) and Frownfelter-Lohrke (1998) show little
di?erence in decision accuracy between bar charts
and tables for tasks that are either spatial or sym-
bolic in nature. Also, the argument that graphical
formats are preferable for complex tasks (because
analytical processing is di?cult in complex tasks,
graphical formats that stimulate integrative pro-
cessing become more interesting, even for symbolic
tasks) receives little support (Hwang, 1995; Vessey,
1991). Studies in accounting have shown bene?cial
e?ects of graphs (e.g. Stock & Watson, 1984;
Wright, 1995) as well as favorable e?ects of tables
for tasks and questions that are su?ciently com-
plex (Davis, 1989; So & Smith, 2004).
One explanation for these disparate results is
that user characteristics are ignored (Chandra &
Krovi, 1999). Amer (1991, p. 30–31) argues that
subject variation could explain the failure to obtain
certain e?ects of presentation formats. This paper
is the ?rst to test whether speci?c knowledge (i.e.
cost accounting knowledge) is a critical individual
factor that can explain when graphics versus tables
increase or decrease performance (i.e. pro?ts from
cost-based decisions). Knowledge can in?uence
the internal memory representations of a decision
584 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
maker and could therefore be linked to the external
presentation format. Theory suggests that more
knowledgeable subjects possess internal schemata
that help them to better retrieve and search for rel-
evant data, along with internal rules for appropri-
ately weighing and acting on these data (Libby &
Luft, 1993; Rose & Wolfe, 2000; Spilker, 1995).
Knowledge has previously been studied in rela-
tion to task complexity (e.g. Bonner, 1994). Dear-
man and Shields (2001) show that managers with
more cost accounting knowledge perform better
in a judgment task that requires complex evalua-
tions of the level of cost distortions introduced in
a volume-based costing system. Other studies
argue that the e?ects of having more advanced
knowledge structures on the participants’ perfor-
mance in more complex tasks is dependent upon
other conditional factors such as accountability,
problem solving ability, and motivation (e.g.
Cloyd, 1997; Dearman & Shields, 2005; Tan &
Kao, 1999; Tan, Ng, & Mak, 2002). Yet, no study
has explored whether the e?ect of knowledge on
performance in a complex task depends on the
way we present information (i.e. presentation for-
mat) to the decision maker.
Only one recent study by Vera-Mun˜oz et al.
(2001) suggests that knowledge may be related to
the external representation of the problem. They
show that when cash-?ow data are presented in
an inappropriate format (i.e. in a statement of his-
torical earnings), experienced managers (with a
stronger knowledge base) are better able to deter-
mine the relevant cash-?ow items in advising cli-
ents than are less experienced managers.
Experience level has no e?ect, however, when the
information is presented appropriately (i.e. in a
cash ?ow statement). However, this ?nding is
based on tabular formats only and the underlying
data is based on di?erent assumptions (earnings
versus cash ?ow). Also, the performance conse-
quences (in terms of pro?ts) are not assessed.
Theory and hypothesis development
I study the joint e?ects of cost accounting
knowledge and presentation format in a complex
decision task with multiple cues and decision vari-
ables, which require a large amount of processing
(Bonner, 1994; Tan et al., 2002). Participants
simultaneously decide on price and resource allo-
cations for a heterogeneous set of customers. The
task requires participants to compare di?erent cus-
tomers in terms of cost and revenue potential, to
search for relevant information in an ABC-driven
report, and to explore the sensitivity of the results
to various decision rules. I argue that the direc-
tional e?ect on pro?ts of a tabular versus a graph-
ical ABC presentation format depends on the
user’s level of cost accounting knowledge: formats
that bene?t a low-knowledge user may hamper the
performance of a knowledgeable user.
Representational congruence and the cognitive
burden
The theory of representational congruence
(Arnold, Collier, Leech, & Sutton, 2004; Chandra &
Krovi, 1999) predicts a more favorable e?ect
on decision performance when the external presen-
tation format matches the user’s cognitive model or
internal representation (Chandra & Krovi, 1999).
A mismatch leads to a high cognitive load and a
less e?ective information retrieval, negatively
a?ecting the decision maker’s performance. Build-
ing on this work, the decision aid literature suggests
opportunities to reduce the cognitive load by
restoring the ?t between the user and the decision
aid (Rose & Wolfe, 2000; Rose, Douglas, & Rose,
2004). In particular, some authors suggest that in
comparison to tables, graphical formats can reduce
a decision makers’ cognitive burden. Stock and
Watson (1984) and Moriarity (1979) argue that
graphical data allow the decision maker to trigger
analogue graphical representations that are stored
in memory. These representations facilitate data
retrieval and information processing. Similarly,
Wright (1995) argues that graphical data help
reduce information overload by highlighting pat-
terns in the data, promoting the perception and
acquisition of information relationships in short-
term memory. Here, I presume that these bene?ts
of graphics apply mainly to decision makers with
a limited level of cost accounting knowledge.
Cost accounting knowledge is likely to be cru-
cial when decision makers receive a tabular ABC
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 585
report that contains very speci?c ABC-related
information. The advantage of greater cost
accounting knowledge is that it typically leads to
more advanced internal schemata (internal repre-
sentations) that are stored around various ABC
cost concepts (Dearman & Shields, 2001). These
internal schemata help decision makers search
for the relevant details within a tabular ABC cost
report (external representation). Decision makers
with less accounting knowledge, in contrast, lack
the su?cient internal ABC representations to
make a match and retrieve speci?c information
from a tabular ABC format. Based on the theory
of representational congruence (Chandra & Krovi,
1999), a graphical ABC format is likely to reduce
the cognitive burden of less knowledgeable deci-
sion makers by providing a ?t to analogue graph-
ical representations that are stored in memory
(Moriarity, 1979; Stock & Watson, 1984; Wright,
1995). Hence, this ?t should facilitate data retrie-
val and in turn their performance should improve.
Decision styles of users and the search for
information
The above theory says little about the perfor-
mance e?ects of graphical and tabular formats
for knowledgeable users. Such users have the
appropriate internal schemata to e?ciently use
the tabular ABC data for pro?t improvement. If
they use a graphical representation, on the other
hand, they may make parallel ?ts to analogue
graphical representations stored in memory in
the same way as their less knowledgeable counter-
parts. According to Umanath and Vessey (1994, p.
797), this may imply that the decisions of sophisti-
cated and unsophisticated users would not di?er
appreciably with a graphical representation. In
this section, I introduce the notion of the decision
style (Lucas, 1981; Sullivan, 1988) to argue that
graphical formats in comparison to tables could
actually produce adverse performance e?ects for
the knowledgeable decision maker. According to
Lucas (1981), analytical decision makers, or those
who speci?cally look for details, bene?t from
tables. Tables indeed o?er an analytical view of
the data that facilitates item-by-item evaluation
(Vessey, 1991). In contrast, heuristic decision mak-
ers, or those who look at the entire problem, ben-
e?t from graphs that emphasize an overview of the
same data (Lucas, 1981; Vessey, 1991).
There are reasons to believe that greater knowl-
edge in a particular domain leads to an analytical
focus. Factual knowledge can evolve into more
abstract (analytical) representations rather than
surface-level representations in memory (Ander-
son, 1990). Because an analytical approach takes
e?ort, a high degree of knowledge in a domain is
needed to engage in analytical processing (Alba &
Hutchinson, 1987, p. 418). Consistent with an ana-
lytical focus, knowledgeable users perform a more
focused information search prior to solving a prob-
lem(Cloyd, 1997; Wood &Lynch, 2002) and access
rules to distinguish between more and less relevant
information (Bonner, 1990; O’Donnell, Koch, &
Boone, 2005). Conversely, Chi, Feltovich, and Gla-
ser (1981), as cited in Vera-Mun˜ oz et al. (2001, p.
409), suggest that novices as opposed to experi-
enced physicists, tend to focus more on surface fea-
tures (heuristic style). Also, earlier work by
Benbasat and Schroeder (1977) suggests that users
with low-knowledge focus on an overview and
screen all available reports, whereas knowledgeable
decision makers search for speci?c details by
requesting a limited number of speci?c reports.
An unanticipated ?nding in Desanctis and Jar-
venpaa (1989) hints at the possibility that graphical
formats do not ?t the way certain users process
information: some users in their study started to
convert bar charts into numerical tables. In my
study, I predict lower performance with graphics,
particularly for high-knowledge decision makers,
given that knowledge in a domain leads to an
analytical focus. Because graphics emphasize an
overview of the data, changes in certain variables
may go unnoticed; and given that knowledgeable
users look for speci?c details (item-by-item pro-
cessing) we might expect them to perform better
with tables (Mackay & Villarreal, 1987). The fact
that graphics tend to lower decision times by
providing an overview (Hwang, 1995; Painton &
Gentry, 1985) may further lower the performance
of high-knowledge decision makers by impeding
detailed analytical processing.
On the basis of the above arguments, I hypoth-
esize the following:
586 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
HYPOTHESIS – The pro?t performance for
decision makers with high (low) cost accounting
knowledge is higher (lower) given a tabular ABC
presentation format than given a graphical ABC
presentation format.
Experimental method
The experimental task consists of a complex
management accounting problem. Participants
improve pro?ts by conducting eight trials of price
and resource allocation decisions for three custom-
ers of a case company. Presentation format is
manipulated as a between subjects factor: partici-
pants receive ABC-based customer pro?tability
reports presented in either a tabular or a graphical
format. Prior to administering the task, I measure
each participant’s level of knowledge, in addition
to common control variables such as ability and
work experience, using instruments similar to those
in Bonner and Lewis (1990), Dearman and Shields,
2005, and Cloyd (1997). I assume that these vari-
ables are randomly distributed across the between
subjects factor.
1
All experimental materials are
issued via computer in the following order. First,
participants provide demographic information,
which includes a measure of their work experience.
Second, participants complete two pre-tests to mea-
sure their cost accounting knowledge and general
ability. Third, the experimental taskis administered.
Finally, participants complete a debrie?ng ques-
tionnaire. The next subsections describe the partici-
pants, the experimental task and its procedures, and
the measurement of the relevant variables.
Participants
The participants were 55 students enrolled in a
four-year business program at a large west-Euro-
pean university. On average, participants were
22.0 years old (most students were about to start
or ?nish their ?nal year). All students had com-
pleted at least two accounting courses in which
ABC is covered. The advantage of using student
subjects and knowledge pre-tests is that I can better
capture the participant’s knowledge component
without the noise of expertise acquired via work
experience (Libby, 1995; Rose &Wolfe, 2000; Uma-
nath & Vessey, 1994). Nonetheless, further tests
control for self-reported work experience (see Sec-
tion ‘‘Control variables’’) as 93%of my participants
indicated some relevant part-time work experience,
with a mean of just over three years of self-reported
experience. Participants typically applied for part-
time jobs corresponding to their studies (e.g. in a
typical business setting) and one-quarter of them
interned with a Big-four accounting ?rm or with a
large controller department of a private ?rm.
To ensure that participants were motivated, they
received a participation fee of six euros along with
a lottery ticket that gave them a chance to win one
of four 50-euro bonus prizes (Bonner, Libby, &
Nelson, 1997). To stimulate ample cognitive e?ort
in both the pre-tests and the experimental task, par-
ticipants were informed that the chances of receiv-
ing this bonus increased with realized pro?t in the
task as well as with their knowledge and ability test
scores.
2
Two ?ve-point Likert-scaled items for
motivation (in the debrie?ng questionnaire)
showed that on average participants were highly
motivated (mean = 4.30, std. dev. = 0.58, a =
0.62), with no signi?cant di?erences across presen-
tation format and knowledge (all p’s > 0.28).
The experimental task and procedures
Participants play the role of a ?rm’s manager
and review descriptions of the case company and
1
In order to assess the validity of this assumption, t-test
results indicate no signi?cant di?erences (all p’s > 0.14) in the
mean ability and knowledge test scores as well as the mean level
of work experience across the two levels of presentation format.
2
It is not uncommon to also reward for knowledge and
ability scores (e.g. Vera-Mun˜ oz, 1998; Rose & Wolfe, 2000). In
my experiment, ticket numbers could appear 2, 6, 11, 15, 20, 21,
25, or 30 times (practical range: 2–25 tickets) in the lottery using
the following scheme. The person with the highest combined
score in the knowledge and ability test receives 10 tickets, the
next 50% all score ?ve tickets, the rest each receive one ticket.
The experimental task would be similar to that above. The
person with the highest mean pro?t realized over the eight
decision trials receives 20 tickets, the next 50% receive 10
tickets, the rest each receive one ticket. The bonus prices would
be drawn and paid out one week after the experiment.
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 587
its customers. The ?rm is an exclusive distributor
of a crucial product that is sold to three major cus-
tomers, denoted A, B, and C, respectively. Partic-
ipants are told that customers vary in revenue
potential and in customer-speci?c support costs
(delivery, logistics, and sales visit costs). Partici-
pants receive a detailed task description, together
with an initial ABC-driven customer pro?tability
report that presents a situation with ample room
for pro?t improvement.
3
Depending upon the
assigned condition, the cost report presents the
same underlying ABC data (calculations in the
Appendix) in either a tabular or a graphical format
(multicolored bar charts showing absolute data
and data on a per unit/percentage basis). Fig. 1
displays the examples. Participants are instructed
to carefully review this information.
4
Their objec-
tive is to improve pro?ts by altering price (p
i
)
and resource allocation (x
i
) for each customer
within a given range (p
i
2 [80, 140] and x
i
2 [30,
90]). The ABC report is updated after each of
the eight trials. As Fig. 1 shows, prior decisions
and outcomes are stored in either a trend chart
(graphical format) or a table.
As the Appendix shows, both prices and
resource allocations a?ect pro?ts and there is an
optimum level of each at which pro?ts are maxi-
mized. The task is relatively complex: (1) pro?ts
are a?ected by two simultaneous decisions, requir-
ing more mental capacity (Bonner, 1994; Tan
et al., 2002) than a task with one decision variable,
and (2) the task is unstructured (Vera-Mun˜ oz
et al., 2001). Participants know that customers
are heterogeneous in their required level of sales
support. Nevertheless, the cost reports only deli-
ver updated cost information in each trial. Consis-
tent with Christensen and Demski (1995) and
Datar and Gupta (1994), the ABC report does
not perfectly re?ect the true costs, nor are the
underlying cost parameters or functions revealed.
In order to improve pro?ts, participants must
make comparisons across the di?erent customers
and explore several decision rules in each trial to
retrieve, evaluate, and weigh the di?erent items
in their report.
Operational measures of the variables
Performance metrics
Table 1 contains an overview of the main vari-
ables used in the analyses. Consistent with Waller
et al. (1999), I measure each participant’s perfor-
mance as the deviation of his or her realized pro?t
from the optimal ?rm-wide pro?t (averaged over
eight trials). I also compute the mean deviation
of actual price and resource allocation decisions
for Customers A, B, and C against optimal price
and resource allocation decisions. The resulting
metrics in Table 1 are labeled as PROFIT and
DECISION, respectively. Smaller values signify
better performance. The hypothesis test is based
on the pro?t deviation (PROFIT); sensitivity
checks also explore the DECISION metric.
Presentation format
Twenty-eight participants receive their ABC
report and previous decision outcomes in a tabular
format. The remaining 27 participants receive the
same data in a graphical format (multicolored
bar charts of the ABC-data and trend charts for
the previous decision outcomes; e.g. Hwang,
1995; Jarvenpaa, 1989). Both formats are used in
ABC software applications (Mooney et al., 2000;
So & Smith, 2002).
It is important to note that the bar charts con-
tain most, but not all of the items in the table.
Because of computer screen space limitations, the
charts would become too crowded if all items of
3
The initial ABC report displays starting prices of 98, 97, and
104 euros, while sales visits are ?xed at 52, 60, and 54 euros for
Customers A, B and C, respectively. This provides room for
pro?t improvement: Firm-wide pro?t for the starting values is
only 1,614,542, whereas the maximum pro?t (see the Appendix)
equals 3,036,145 (achieved at P
a
= 115.5, P
b
= 124.9, P
c
= 97.6,
x
a
= 62.0, x
b
= 49.4, and x
c
= 83.1). Students are not given any
information about the optima.
4
It is important to note that the computer tracks the time
that participants use to review the problem description and the
initial ABC report, as supplementary analyses use this time as a
proxy for information search (see Section ‘‘Supplementary
analysis’’). The rationale is that the time expended on
information gathering prior to problem solving can provide
indications on how subjects with di?erent levels of knowledge
and presentation formats approach the problem (see Section
‘‘Theory and hypothesis development’’).
588 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
Fig. 1. Actual screenshots of the initial ABC-report issued prior to the task for the two levels of presentation format (between subjects
factor).
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 589
the table were shown, and the e?ectiveness of the
graph (Davis, 1989, p. 497; Tufte, 2001) would
decline. Nevertheless, I expect the two formats to
be essentially identical in information content. In
particular, most of the items are maintained in
the graphical format, including the most crucial
information (i.e. the customer-speci?c costs of
sales visits, internal logistics, and product delivery)
and the few omitted items convey little new infor-
mation. Both the tables as well as the bar charts
reveal that support costs are higher for Customer
B relative to purchasing costs and that the oppo-
site is true for Customer C. Inherent to the features
of presentation formats (Jarvenpaa, 1989; So &
Smith, 2004), the graphical format visualizes the
amount of costs per category by distances in the
space vector (without attaching a discrete value),
whereas tables display the speci?c ?gures.
5
Although bar charts do not convey the amount
of resources consumed per customer (e.g. volume
drivers for sales visits, deliveries, and stock pick-
ings), customer costs are calculated by multiplying
the amount of resources by a ?xed driver rate and
hence they have the same information content as
resources consumed (the two cues are perfectly
Table 1
Summary of variables’ operational de?nitions and descriptive statistics
Variable (De?nition) Mean
N = 55
Standard
deviation
Min. Max. Median
Dependent variables
a
PROFIT (pro?t deviation: mean distance of the total
realized pro?t to the optimal pro?t averaged over 8 trials)
b
761,491 292,732 307,758 1,447,754 773,486
DECISION (deviation of decisions: mean distance of the
actual decisions for Customer A, B, and C to the optimal
values averaged over 8 trials)
74.60 16.46 39.7 104.0 76.4
Independent variables
KNOWLEDGE
b
KNOWSCORE (score on the test for the level of cost
accounting content)
3.84 1.18 2 6 4
KNOW (mean split: high versus low level of accounting
knowledge)
– – 0 1 –
Low High
PRESENTATION (presentation format of ABC-data;
either graphical displays or tables)
c
– – 0 1 –
Table Graphic
Control variables
ABILITY (general ability test score)
d
7.00 1.96 1 11 7
WORKEXP (a participants’ self-reported relevant
part-time work experience in number of years)
3.06 2.37 0 9 2
a
PROFIT used the following formula: R
j
(P
*
À P
j
)/8 with P
*
optimal ?rm-wide pro?t and P
j
a participants’ realized ?rm-wide
pro?t in trial j = 1, . . . ,8. DECISION used the following formula: R
i
R
j
(jx
i*
À x
ij
j + jp
i*
À p
ij
j)/8 with pi*, xi* optimal solutions and x
ij
,
p
ij
a participants’ decision choices for customer i = A, B, C in trial j = 1,. . ., 8. Lower scores represent better performance.
b
The theoretical range of KNOWSCORE is from zero to six. For the mean split variable KNOW, 23 (32) subjects were classi?ed in
the low (high) accounting knowledge condition.
c
Twenty-eight (27) subjects received the ABC report in a tabular (graphical) format. The two reporting formats are displayed in
Fig. 1.
d
The theoretical range of ABILITY is from 0 to 11.
5
Bar charts give the di?erent customer cost categories about
the same color and use a di?erent color for the purchasing cost.
Consistent with the features described in the literature
(Umanath & Vessey, 1994) the bar charts provide an overview
and make certain trends (e.g. relative consumption of customer
costs across customers) more visible. These trends must be
determined by the participant in a tabular format. Nevertheless,
in a tabular format subjects have better access to the discrete
values in comparison to bar charts.
590 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
correlated). In the same way, sales volume is not
shown in the bar chart as volume is highly corre-
lated with revenues. For the total outcome feed-
back, I keep only the most relevant item in the
graphical format, namely total pro?ts. Other totals
in the tabular format can be automatically derived
and are less important, as the participants’ goal is
to increase pro?ts.
Accounting knowledge
Following prior research (Bonner & Lewis,
1990; Dearman & Shields, 2001; Dearman &
Table 2
Hypothesis test
Panel A: ANCOVA-models with LN(PROFIT) as dependent measure
a
Model 1: Controlling for ability and workexp Model 2: Controlling for ability, workexp and
W · K interaction
Source of variation Mean square F-stat p-value Mean square F-stat p-value
Factors
Presentation (P) 0.00146 0.01 0.923 0.01094 0.07 0.787
Know (K) 0.10341 0.67 0.417 0.51393 3.47 0.069
*
P · K 0.63676 4.12 0.048
**
0.75938 5.12 0.028
**
Covariates
Ability (A) 0.42559 2.75 0.103 0.35940 2.42 0.126
Workexp (W) 0.24561 1.59 0.213 0.37611 2.54 0.118
W · K – – – 0.45124 3.04 0.087
*
R
2
= 0.176; N = 55 R
2
= 0.225; N = 55
Panel B: Mean by mean comparison of the Presentation (P) by Know (K) interaction
b
400000
480000
560000
640000
720000
800000
880000
LOW HIGH
.047**
808915
(13.603)
648425
(13.382)
.108
.017**
.224
582432
(13.275)
Table
742561
(13.518)
Graphic
N=8
N=15
N=19
N=13
Model 1: Interaction P x K (F=4.12, p=.048**)
?
400000
480000
560000
640000
720000
800000
880000
LOW HIGH
.057*
804553
(13.598)
604644
(13.312)
.056*
.015**
.153
579436
(13.270)
Table
725387
(13.494)
Graphic
N=8
N=15
N=19
N=13
Model 2: Interaction P x K (F=5.12, p=.028**)
a
Variable de?nitions in Table 1.
**
,
*
Indicate signi?cance levels of 5 or 10%. The test uses the mean split variable for knowledge
(‘know’) to allow for a cell-by-cell comparison of the P · K interaction (see Panel B). Model 1 controls for the covariates work
experience and ability. Model 2 adds the WxK interaction as an additional covariate (experience matters more at lower-levels of cost
knowledge).
b
Detailed analyses of the P · K interaction of the models in Panel A. The horizontal axes represent the knowledge categories (low
versus high). The full lines represent the two presentation formats (tables versus graphics). The vertical axes display the mean pro?t
deviation (LN of pro?t between brackets) evaluated at covariates ‘workexp’ (mean = 3.06) and ‘ability’ (mean = 7.00). A lower pro?t
deviation represents better performance. N indicates the number of participants in each cell. Dotted arrows and associated ?gures
represent the direction and the p-value of a one-sided mean-by-mean comparison.
**
,
*
represents signi?cance at the 5 or 10% level.
a
Indicates that the e?ect of p = 0.108 becomes signi?cant at the 10% level (p = 0.097) when the last trial is excluded from the analysis
(end-trial e?ect).
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 591
Shields, 2005), I measure a participant’s knowl-
edge as the number of correct answers to six multi-
ple-choice questions that are intended to assess the
subject’s level of cost accounting knowledge.
6
The
questions are adapted from test banks of cost and
management accounting textbooks and are de?ni-
tional in nature. They assess conceptual knowl-
edge on ABC- and cost-related accounting topics
such as the e?ects of product complexity on prod-
uct costing, the e?ects associated with speci?c cost
allocation methods, the e?ects of value-added
costs and various types of cost drivers, the ABC
cost driver levels, and some statements on more
elaborate ABC-systems.
Table 1 shows a large variation in the partici-
pants’ knowledge, with scores (KNOWSCORE)
ranging between two and six. Studies with actual
managers tend to show similar variations in knowl-
edge (Dearman & Shields, 2001; Stone et al., 2000),
and reliability of the measure is comparable to other
studies on accounting knowledge.
7
To disentangle
the e?ects of presentation format by knowledge cat-
egory, the hypothesis tests will utilize a mean split
metric (KNOW) to classify participants into a
low- or a high-knowledge subgroup (see Bierstaker,
2003). Table 2 shows that about 23 participants, or
42%, fall into the low-knowledge subgroup. Their
average knowledge score of 2.65 is statistically
di?erent from the score of 4.69 for participants
in the high-knowledge subgroup (t = 12.09;
p < 0.01). Sensitivity checks will also repeat the tests
with the actual knowledge score (KNOWSCORE).
Control variables
Prior studies hypothesize positive relationships
between a decision maker’s ability and audit perfor-
mance (Bonner &Lewis, 1990; Libby &Luft, 1993).
Other studies suggest that work experience leads to
expertise (experience creates opportunities to gain
additional knowledge) that can a?ect performance
(Cloyd, 1997; Libby, 1995). Consistent with Vera-
Mun˜ oz (1998) and Cloyd (1997), I make no a priori
predictions about di?erences in ability and work
experience among the participants. However, I
introduce these variables to statistically control
for their e?ects on a participant’s task performance.
Work experience (WORKEXP) is measured by
participants’ self-reported relevant part-time work
experience in years. Table 1 shows that WORK-
EXP has a mean of 3.06 years, with a range of zero
to nine years of experience. Most of the participants
(92.7%) reported some work experience.
Similar to prior studies (e.g. Bonner & Lewis,
1990; Bonner, Davis, & Jackson, 1992; Cloyd,
1997; Dearman & Shields, 2005), general ability
(ABILITY) is measured as the number of correct
answers on an 11-item test derived from sections
of prior graduate record examinations (GREs),
with questions on problem solving, analytical abil-
ity, and data interpretation. The mean ABILITY
score reported in Table 1 is seven. Again, the reli-
ability of this test is satisfactory in comparison to
prior studies.
8
Experimental results
In the ?rst subsection below, I describe the
results for the hypothesis on the joint consider-
ations of knowledge and presentation format on
pro?t performance. The second subsection reports
6
One question is dropped because of interpretation problems
as indicated by a few participants.
7
Bonner and Walker (1994) argue that reliability measures
are not appropriate for accounting knowledge constructs.
Knowledge is typically an omnibus construct whereby each
item of a test measures a separate subconstruct of the overall
concept. Therefore, such tests often achieve a low level of
reliability. Depending on the items included, ex post reliability
of the current study’s test range from a = 0.15 to a = 0.47, with
reliability of the full six-item test equal to 0.21. These levels are
similar to those in studies with a comparable number of items.
Tan and Kao (1999) and Tan and Libby (1997) report a levels
ranging from 0.19 to 0.43 for auditing task knowledge or 0.39
for technical accounting knowledge. Dearman and Shields
(2001) obtain levels of 0.29 to 0.45 for similar constructs of cost
accounting content. Bonner and Lewis (1990) do not report a
levels for their knowledge tests. The mean score of 3.84 (64%
correct answers) is similar to that of prior studies (e.g. Cloyd,
1997).
8
The ex post reliability of the items in the current ability test
is a = 0.61. Dearman and Shields (2005) report a reliability of
a = 0.52, while both Bonner et al. (1992) and Cloyd (1997)
achieve a level of a = 0.63 for comparable ability tests. The
mean score of seven, or 63.6% correct answers, is comparable to
the ability test scores reported in prior studies (e.g. Cloyd,
1997).
592 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
some sensitivity checks. Finally, the third subsec-
tion further explores some motivations for the
joint e?ects of knowledge and presentation format.
Hypothesis test
I perform an ANCOVA analysis using PROFIT
as the dependent measure. Consistent with Waller
et al. (1999) and Gupta and King (1997), I employ
a log-linear relation speci?cation. The model
includes the presentation format factor and the
mean split variable KNOW as an additional factor.
Speci?cally, four unique cell means can be created –
lowKNOWor high KNOW, and tabular or graph-
ical format – from which cross-cell comparisons
can be made to better understand the interaction
between knowledge and presentation format. Work
experience and general ability serve as covariates.
The results are reported in Table 2, along with the
detailed tests of the interaction.
Panel A of Table 2 reports two models. Model 1
is an ANCOVA that controls for the main e?ects
of ability and work experience. Model 2 also
includes the interaction of work experience and
knowledge. This additional covariate is included
because the correlations suggest a strong e?ect of
work experience on the pro?t error for less knowl-
edgeable decision makers (r = À0.459, p < 0.03),
while for high-knowledge decision makers this
e?ect is insigni?cant (r = 0.0267, p = 0.88). Fol-
lowing Libby (1995), who suggests potential inter-
actions between indirect expertise (acquired via
work experience) and knowledge, Model 2 statisti-
cally controls for this e?ect.
9
The interaction between knowledge and presen-
tation format is signi?cant at the 5% level in both
models. The test provides strong evidence of an
e?ect reversal (the interaction is signi?cant; the
main factors are either not signi?cant or only mar-
ginally signi?cant). In line with my hypothesis, the
performance of decision makers with a high (low)
degree of accounting knowledge is higher (lower)
with tabular ABC presentation formats than with
graphical presentation formats. The plots of the
interaction in Panel B of Table 2 con?rm that
the pro?t deviation for high-knowledge decision
makers is signi?cantly larger for a graphical
ABC format than for a tabular ABC format
(p = 0.047 and p = 0.057 in Models 1 and 2,
respectively). Conversely, participants with a low
level of cost accounting knowledge perform better
when using graphics rather than tables. This e?ect
is signi?cant in Model 2 (p = 0.056), which con-
trols for the fact that a low-knowledge decision
maker bene?ts from work experience (W · K
interaction). The insigni?cant e?ect of presenta-
tion format in Model 1 (p = 0.108) also becomes
signi?cant (p = 0.097) when the last trial is
excluded from the analysis (end-trial e?ect). In line
with the theory on representational congruence
(see Section ‘‘Theory and hypothesis develop-
ment’’), improved cost accounting knowledge
especially matters with a tabular format (p =
0.017 and p = 0.015 in Models 1 and 2, respec-
tively). Participants with low cost accounting
knowledge presumably have more di?culty in
extracting data from a tabular ABC format due
to a lack of appropriate internal ABC schemata.
Conversely, accounting knowledge is not signi?-
cant under a graphical ABC format (p > 0.22 and
p > 0.15 in Models 1 and 2, respectively) because
a graphical format is bene?cial for the low-knowl-
edge decision maker but detrimental for users with
more knowledge.
Note that the ?ndings do not suggest a di?er-
ence in information content across formats.
Although bar charts do not capture all the items
of a table, low-knowledge users with graphs in
my experiment can realize the same pro?t as their
knowledgeable counterparts that use a table (i.e.
the two cells with the best performance). P-values
of this comparison are not signi?cant (p > 0.28 and
9
Libby (1995) and Tan and Kao (1999) also suggest that
interactions of knowledge and ability may a?ect performance
(especially when the task is very complex and accountability is
high, see Tan & Kao, 1999). Tan and Libby (1997) further study
the e?ects of ability and knowledge at di?erent levels of
experience. Accordingly, I perform an ANCOVA analysis
adding all the possible interactions of ability, knowledge, and
work experience; the results are not qualitatively altered in that
the P · K interaction still remains signi?cant (while the extra
covariates are not signi?cant). Although theory does not
provide guidance about the interactions of ability, work
experience, and presentation format, I also include these as
covariates in a further test. The P · K interaction remains
signi?cant (the covariates are again not signi?cant).
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 593
p > 0.40 in Models 1 and 2, respectively; compari-
son not shown in Panel B of Table 2). This sug-
gests that the graphical format maintains the
relevant items that are crucial for pro?t improve-
ment. Furthermore, I would probably not ?nd a
negative pro?t e?ect for tabular formats with
low-knowledge decision makers if the extra items
of a table revealed new information.
Sensitivity checks
Results remain similar when I use the log of
DECISION as a dependent variable. The interac-
tion of presentation format and knowledge remain
signi?cant (p = 0.07 in Model 1; p = 0.04 in Model
2). Simple presentation format e?ects at high or
low levels of knowledge corroborate the e?ect
reversal: low-knowledge participants perform bet-
ter with graphs than with tables (smallest p =
0.05, cf. p = 0.06 for the pro?t deviation) whereas
the opposite is true for high-knowledge partici-
pants (smallest p = 0.08, cf. p = 0.05 for the pro?t
deviation).
10
I also test whether the e?ect of the
interaction between presentation format and
knowledge (P · K) on pro?t persisted over trials.
A repeated measure test (on the log of pro?t in
each trial) does not reveal any signi?cant three-
way interaction among trial, presentation format,
and knowledge (F = 1.26, p > 0.27 in Model 1;
F = 1.13, p > 0.34 in Model 2). Conditional means
per trial con?rm that tables are always superior to
graphical formats for the high-knowledge decision
maker, where the opposite holds true for the low-
knowledge decision maker.
Additionally, I explore the interactive e?ect of
presentation format and knowledge on perfor-
mance (i.e. log of pro?t) by using the actual score
on the knowledge test (KNOWSCORE) as a con-
tinuous variable into two regressions that use the
same control variables as in the previous section
(Model 1 includes work experience and ability;
Model 2 additionally includes the interaction of
work experience and knowledge). I follow the pro-
cedure of Aiken and West (1991) whereby presen-
tation format is dummy-coded (tabular format =
0; graphical format = 1), and knowledge is mean-
centered in both main e?ect and the multiplicative
terms to avoid multicollinearity problems (see
Cloyd, 1997; Tan & Kao, 1999). Consistent with
my hypothesis, the parameter estimate for the
interaction term of presentation format and
knowledge is signi?cant in Models 1 (b = 0.191;
p = 0.05) and 2 (b = 0.186; p = 0.06).
I determine the nature of this signi?cant interac-
tionby calculating simple e?ects of presentationfor-
mat at one standard deviation above and below the
mean of knowledge (Aiken &West, 1991; Pedhazur,
1982).
11
For participants with high-knowledge
(mean plus one standard deviation), presentation
format has a positive sign signifying higher pro?t
deviations (or lower performance) under a graphi-
cal format than under a tabular format. The e?ect
is signi?cant in Model 1 (b = 0.278, p = 0.05) and
Model 2 (b = 0.229, p = 0.08). For participants with
low-knowledge (mean less one standard deviation),
the e?ect of presentation format is negative in both
models, consistent with the prediction that graphi-
cal formats lead to better performance (lower pro?t
deviation), although the e?ect is only signi?cant in
Model 2 (b = À0.216, p = 0.08; cf. Model 1:
b = À0.177, p = 0.13).
Panel A of Table 3 plots regression lines for
Models 1 and 2 (predicted values of the log of
pro?t were transformed to pro?t deviations).
Panel B of Table 3 displays the simple e?ects and
the signi?cance levels (p-values) of presentation
format at each level of knowledge, which is equiv-
alent to establishing regions of signi?cance (see
Aiken & West (1991, p. 132), for a more detailed
description). The results in Panel B of Table 3
10
The statistical tests increase in power if I consider that none
of the suggested variables has an e?ect on the decision errors for
Customer A (negative adj. R-square for all models, p’s
individual beta’s all >0.24). The e?ect reversal becomes highly
signi?cant when the analysis of ‘decision’ is based on the
resulting sum of decision errors on Customers B and C (the
least and the most pro?table customers, respectively).
11
Consistent with Pedhazur (1982, p. 440), the decomposition
of the interaction in simple e?ects of presentation format at
high or low-knowledge or at each speci?c knowledge score (as
reported in Table 3 below) is based on one-tailed statistics to
minimize the type II error. Based on the theory, I also predict
speci?c directions for the e?ect of presentation format at high
or low levels of accounting knowledge.
594 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
show that graphs as opposed to tables are bene?-
cial for people with a low-knowledge score (score
of 2) and disadvantageous for people with scores
higher than or equal to 5. People with scores near
the points of intersection (scores of 3 or 4; see
Panel A of Table 3; 3.576 in Model 1; 3.816 in
Model 2) are not a?ected by presentation format.
Supplementary analysis
The supplementary analyses in Table 4 explore
some reasons for the ?nding that graphical for-
mats are bene?cial to the low-knowledge decision
maker, whereas tables result in better performance
for a knowledgeable decision maker.
I ?rst study the e?ects of presentation format
and knowledge (KNOW) on self-perceived com-
plexity, as indicated by participants on a ?ve-point
Likert scale in the debrie?ng questionnaire. The
?ndings in Panel A of Table 4 show that users with
a low level of cost accounting knowledge perceive
the task to be more complex under tabular formats
than under graphical formats (p = 0.02). The pre-
sentation format does not a?ect the cognitive
Table 3
Sensitivity checks with KNOWSCORE (actual score) as covariate in Models 1 and 2
Panel A: Plots of the regression lines in Models 1 and 2 and p-value of P · K interaction
a
200000
300000
400000
500000
600000
700000
800000
900000
1000000
2 3 4 5 6
Knowledge
P
r
o
f
i
t
d
e
v
i
a
t
i
o
n
Point of intersection:
3,576
Graph
Tab
mean
Model 1: Interaction P x K (p=.05**)
200000
300000
400000
500000
600000
700000
800000
900000
1000000
2 3 4 5 6
Knowledge
P
r
o
f
i
t
d
e
v
i
a
t
i
o
n
Point of intersection:
3,816
Model 2: Interaction P x K (p=.06*)
mean Graph
Tab
Panel B: E?ect of presentation format at each level of the knowledge score
b
Model 1: Estimate, sign. level, e?ect size Model 2: Estimate, sign. level, e?ect size
Knowledge score b (format) p-value Di?erence in pro?t b (format) p-value Di?erence in pro?t
Dev. (graph-tab) Dev. (graph-tab)
2 À0.30184 0.08
*
À228545.2 À0.33777 0.05
*
À252933.0
3 À0.11039 0.21 À79108.9 À0.15172 0.13 À107341.3
4 0.08107 0.24 55201.4 0.03433 0.38 22988.3
5 0.27252 0.05
**
176698.1 0.22038 0.09
*
140315.3
6 0.46397 0.03
**
287388.1 0.40643 0.05
**
246599.2
a
The regressions use the LN(pro?t) as the dependent and Models 1 and 2 use similar control variables as reported in Table 2.
Consistent with Aiken and West (1991), knowledge was mean centered (Mean: 3.8364). Predicted values were calculated for each
knowledge score by subtracting the mean from the score. Predicted values of LN(pro?t) are transformed to actual deviations.
b
Following Aiken and West (1991, p. 132–133), I performed ?ve regressions of the following structure:
ln(pro?t) = b1 · PF + b2 · K + b3 · KP + (covariates of either Model 1 or Model 2), whereby k was equal to knowledge-2, knowl-
edge-3, . . ., knowledge-6 to evaluate the factor presentation format at each level of the knowledge score. Panel B displays the parameter
estimate of presentation format, the signi?cance level (one-tailed, see Pedhazur, 1982) and predicted di?erences in pro?t deviation
between graphs and tables (e?ect size). A negative (positive) signs denote that graphical formats lead to lower (higher) pro?t deviations
in comparison to tables.
**
,
*
indicates signi?cance at the 5 or 10% level.
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 595
burden for the high-knowledge subgroup (p >
0.45). An ANOVA-test of the measure of complex-
ity (not reported in Panel A of Table 4) suggests
that the e?ect of presentation format depends on
knowledge (P · K interaction: F = 6.49 and p =
0.02; main e?ects, p > 0.11). Consistent with the
theory on representational congruence (see Section
‘‘Theory and hypothesis development’’), I presume
that participants with low knowledge lack the
appropriate schemata to make a ?t with a tabular
format and hence ?nd the task to be more complex
than their counterparts who receive a graphical
format.
A second test analyzes the amount of time that
participants spend reviewing the task description
and the initial cost report (‘‘information search’’)
prior to making their decisions (Panel B of Table
4). Similar to Cloyd (1997), I control for a partic-
ipant’s reading ability (reading speed). Results are
analyzed via an ANCOVA in a similar way as
Table 4
Supplementary analyses of the level of perceived complexity and information search
Panel A: Self-perceived complexity (?ve-point Likert scale) and the e?ect of format (tables versus graphs) per knowledge category
a
Mean Median SD
Low-Knowledge subgroup (N = 23)
Table (N = 15) 3.20 3.00 0.86 Test of di?erence between Table-Graph
Graph (N = 8) 2.25 2.00 0.71 t-value: 2.67; p < .02
**
(v
2
: 5.598, p < .02
**
)
High-Knowledge subgroup (N = 32)
Table (N = 13) 2.31 2.00 0.95 Test of di?erence between Table-Graph
Graph (N = 19) 2.53 3.00 0.70 t-value: À.075; p > .45 (v
2
: 0.437, p > .51)
Panel B: ANCOVA analysis on information search (SEARCH) prior to the decision trials
b
Source of variation Mean square F-stat p-value
Factors
523.2
749.3
661.3
595.8
400
450
500
550
600
650
700
750
800
LOW HIGH
.062*
.147
.164
.001***
Graphic
Table
Presentation (P) 728 0.03 0.863
Know (K) 260313 10.69 0.002
***
P · K 80141 3.29 0.076
*
Covariates
Reading speed 1010714 41.52 0.001
***
R
2
= .555; N = 55
a
Analysis of the item ‘‘I considered the task as extremely complex’’. Due to the word ‘‘extremely’’, the means are not that high. The
tests employ a t-test assuming equal variances (a non-parametric Kruskal-Wallis test is reported between brackets). Note that for the
low-knowledge subgroup the number of users considering the task as complex (scores P3) dropped from 80% for tabular ABC-
reports to 37.5% for graphical ABC-reports. For the high-knowledge subgroup, this number increased (from 46% to 53%). A further
ANOVA-test also reveals a signi?cant P · K interaction: F = 6.49; p < 0.02 (main e?ects: p > 0.11).
b
The dependent variable of this ANCOVA is the time (recorded in seconds) that participants spent on the problem description that
contained the initial ABC report either in a graphical or tabular format (see Fig. 1). Subjects were instructed to thoroughly review this
information. Similar to Cloyd (1997), the test controls for reading speed, which is measured by the participants’ total time in the full
experiment. The ?gure on the right gives a detailed analysis of the P · K interaction of the model (i.e. similar test as reported in Panel B
of Table 2). A higher score represents more search time prior to the problem (scores evaluated at covariate reading speed = 2263.8);
***
,
**
,
*
indicates signi?cance at the 1, 5 or 10% level.
596 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
reported in Table 2. Panel B of Table 4 shows that
knowledgeable decision makers spent more time
on information search (KNOW; p = 0.002) and
there is also weak evidence that the e?ect of pre-
sentation format on information search depends
on knowledge (P · K interaction is marginally sig-
ni?cant; p = 0.076). As shown in the ?gure in
Panel B of Table 4, the di?erence between tables
and graphical formats is marginally signi?cant
(p = 0.062) for high-knowledge participants; tables
tend to increase their information search. The
e?ect of format on search time is not signi?cant
for low-knowledge decision makers (p = 0.147).
Consistent with theory, this ?nding could imply
that due to an analytical focus of high-knowledge
decision makers, tables that provide an analytical
view of the data better assist their information
search. Graphics display the same data at a glance
and are less appropriate for knowledgeable deci-
sion makers who approach the problem in an ana-
lytical way (and therefore search time is reduced).
The ?gure in Panel B of Table 4 also reveals that
low- and high-knowledge participants have the
same search time under graphics (p = 0.164);
knowledge level only makes a di?erence with
tables (p = 0.001).
Although other explanations are possible
besides the ones explored here, these metrics indi-
cate that low-knowledge decision makers seem to
bene?t from graphics because of a reduction in
complexity, whereas sophisticated users likely ben-
e?t from tables because tables enhance their infor-
mation search.
12
Conclusion
In many ?rms, managers with various levels of
cost accounting knowledge make cost-based deci-
sions on the basis of accounting information sys-
tems that provide ‘‘easy-to-understand’’ graphical
rather than tabular cost reports. Yet, the perfor-
mance e?ects of di?erent types of presentation for-
mats in relation to varied levels of cost accounting
knowledge have received scant attention in the lit-
erature. This study experimentally investigates
these joint e?ects in the context of a relatively com-
plex task that requires decision makers to make
customer-speci?c price and resource allocation
decisions that a?ect pro?tability.
This study advances our understanding of man-
agerial decision-making in several ways. First, I
?nd that di?erent formats in?uence pro?tability
di?erently, with the direction of impact contingent
upon the user’s level of accounting sophistication.
After controlling for di?erences in ability and
experience, I show that cost report users with a
low level of cost accounting knowledge achieve
higher pro?ts when ABC data are displayed in a
graphical format rather than in a tabular format.
More surprisingly, the opposite is true for sophis-
ticated users; their pro?ts are higher given a tabu-
lar format.
Second, this ?nding has important practical
implications. There is no unique way to present
cost data to managers. To extract maximum
potential for improved decision-making, account-
ing information systems may need to adjust a cost
report’s format to the user’s level of accounting
sophistication. In most ?rms, many non-accoun-
tants make use of accounting information systems
(Birnberg, 2000; Shields, 1995; Mauldin & Ruc-
hala, 1999). These non-accountants are better
served by graphical presentations of accounting
information. In contrast, accountants are likely
to receive appropriate access to the data by means
of traditional tables. In a broader context, with
respect to the debate on how non-professional ver-
sus professional investors acquire ?rm-speci?c
information (Maines & McDaniel, 2000), ?rms
can capitalize on my ?ndings and adjust the for-
mat of their disclosures to the level of accounting
knowledge of their investors.
12
I further explore these assertions by testing how these
metrics relate to pro?t. For each knowledge category, I regress
the measures for ‘search’ and ‘complexity’ together with the
e?ects of ability and work experience (control variables of Table
2) on the pro?t deviation. While overall pro?t is not a?ected,
analysis of the ?rst-trial pro?t deviation produces interesting
?ndings (it is logical to assume that only ?rst-trial pro?ts are
a?ected by information search; subsequent decisions are
anchored on and adjusted from prior decisions). For the
high-knowledge subgroup, ‘information search’ tends to reduce
the ?rst-trial pro?t deviation (t = À1.75, p = 0.092); the e?ect
of ‘complexity’ is not signi?cant (p > 0.19). For less knowl-
edgeable decision makers, more ‘complexity’ tends to increase
the ?rst-trial pro?t deviation (t = 1.76, p = 0.096), while
‘information search’ is not signi?cant (p > 0.68).
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 597
Third, from a theoretical viewpoint, this study
shows that knowledge is an important attribute that
explains when particular presentation formats are
likely to provide pro?t improvement. While most
research onpresentation formats has produced con-
?icting evidence on several aspects of the decision
task (Vessey, 1991), my ?ndings indicate that a
focus on a user’s characteristics may explain part
of these mixed results. I show that researchers can
expect adverse performance e?ects when the exter-
nal presentation format is not well aligned with a
decision maker’s mental model or decision style.
To focus on the joint e?ects of knowledge and
presentation format, I hold the level of task com-
plexity constant. Future research could vary task
complexity (e.g. Tan & Kao, 1999; Tan et al.,
2002) and explore how the joint e?ects of knowl-
edge and presentation format extend to di?erent
levels of task complexity. It might also be interest-
ing to allow decision makers to choose their for-
mat rather than simply assigning them graphical
or tabular formats, in order to study the types of
decision makers who are more likely to access
graphical rather than tabular representations.
Other dimensions of managerial expertise (junior-
versus senior-level managers) or job functions
(top- versus lower-level managers) could lead to
preferences for a speci?c type of format.
Finally, a potential limitation of the study is
that graphical formats did not contain all the items
of the table. Nevertheless, this probably did not
a?ect my ?ndings because the extra items con-
tained little extra information content for the
objective of maximizing pro?ts. Yet, some deci-
sions, like the evaluation of division managers,
involve a more subjective judgment and further
studies could explore whether extra (redundant)
cues do matter here. Evaluations may di?er when
managers have access to only a few or many met-
rics that are all equally informative about perfor-
mance. Again, one could explore whether
expertise of the evaluator matters herein. Also,
the practice of presenting these metrics to evalua-
tors strongly varies across ?rms. Sometimes bal-
anced scorecards use red, yellow, and green
ratings for their performance metrics (Malina &
Selto, 2001). In any event, further research that
links such variations in practice to a decision
maker’s expertise (Sprinkle, 2003) may provide
guidance for how ?rms can optimize the many
?ows of accounting information to internal and
external users of that information.
Acknowledgements
I want to thank Jan Bouwens, Lynn Hannan,
Laurence van Lent, Paula van Veen-Dirks, Luk
Warlop, and two anonymous reviewers for their
insightful comments. This study also bene?ts from
comments of participants at the Management
Accounting Section’s midyear meeting in Phoenix
(2005), the 4th conference on new directions in
management accounting of the European Institute
for Advanced Management Studies in Brussels
(2004), and the accounting seminar at Tilburg Uni-
versity (2004).
Appendix
This appendix describes the underlying func-
tions, parameters, and cost allocation assumptions
in my experiment. Customer demand (Equations 1
and 2 in Panel A of Table A1) is based on a com-
mon sales response function (Mantrala, Prabhak-
ant, & Zoltners, 1992): the price (p
i
) in?uences
potential demand, whereas the amount allocated
to sales visits (x
i
) determines how much demand
is realized. The gross contribution for a customer
is equal to revenues minus a ?xed portion of reve-
nues: d
i
represents the products’ purchase cost
(Equation 3). Given my focus on cost-based deci-
sion-making, I introduce a complex customer cost
function (Equation 4) with the following elements:
a ?xed cost (FC
i
) per customer (e.g. storage space);
a nonlinear distribution function (u
i
qe
i
+ v
i
qe
1:5
i
Þ
with costs increasing more than proportionally
with volume (Klincewicz, 1990); and a cost func-
tion for sales visits (x
i
w
i
), with sales visits (x
i
)
acquired at a speci?c rate (w
i
). Consistent with
practice, the set of customers is heterogeneous
(Kaplan & Narayanan, 2001). Compared to Cus-
tomers A and C, Customer B is a high-cost-to-
serve customer, consuming more ?xed costs
(FC), more resources in distribution (u
i
and v
i
),
598 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
and more in sales visits (w
i
). Customer pro?tability
(Equation 5) equals gross margin minus customer-
speci?c costs. Via substitution, one can rewrite
Equation 5 entirely in terms of the participant’s
decision choices (p
i
and x
i
). There is a unique opti-
mum for each customer (see the note to Table A1).
Given the focus on cost-based decisions, optima
are determined in large part by the cost variations
across customers. Firm-wide pro?t is equal to the
sum of customer pro?ts (Equation 6).
Both the graphical format as well as the tabu-
lar format present ABC-driven customer data in
Table A1
Underlying functions of the experimental task and ABC cost allocation assumptions
a
Panel A: Underlying functions of the experimental task (by substituting all are f(xi, pi))
Panel B: ABC-customer cost pools (CCP) and cost driver assumptions (i = Customer A, B, C)
a
Optimal solutions can be found by solving ?rst order conditions for pi and xi based on the parameter data for customers i = A,B, C
(cost parameters in bold). Optima are P
a
= 115.5; X
a
= 62.0; P
b
= 124.9; X
b
= 49.4; P
c
= 97.6; X
c
= 83.1; ?rm-wide pro?t: 3,036,145.
The ABC-reports (in Fig. 1) use these functions of Panel A. Save for customer costs, I introduced some small errors following common
assumptions of Datar and Gupta (1994) and Christensen and Demski (1995) that ABC seldom re?ects true costs.
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 599
addition to total ?rm-wide pro?t (actual outcome).
To generate the reports, I use the functions of
Panel A, except for the total customer costs, which
I allocate along the assumptions of Panel B of
Table A1. In the ?rst stage, I assign customer costs
to cost pools, and in the second stage, I assign
them to customers on the basis of cost drivers. I
introduce small errors by specifying the driver
for sales visits slightly di?erently from the actual
driver and by introducing linear cost drivers for
a nonlinear distribution function. Consistent with
Datar and Gupta (1994) and Christensen and
Demski (1995), the reports never re?ect ‘‘true’’
costs. Because Customer B requires more of each
cost driver, customer cost and pro?tability ?gures
remain quite accurate.
References
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing
and interpreting interactions. Newbury Park: Sage.
Alba, J. W., & Hutchinson, J. W. (1987). Dimensions of
consumer expertise. Journal of Consumer Research, 13(4),
411–454.
Amer, T. (1991). An experimental investigation of multi-cue
?nancial information display and decision-making. Journal
of Information Systems, 5(2), 18–34.
Anderson, J. R. (1990). Cognitive Psychology and its implica-
tions (third ed.). New York: W.H. Freeman & Co Ltd.
Arnold, V., Collier, P. A., Leech, S. A., & Sutton, S. G. (2004).
Impact of intelligent decision aids on expert and novice
decision-makers’ judgments. Accounting and Finance, 44, 1–26.
Benbasat, I., & Schroeder, R. G. (1977). An experimental
investigation of some MIS Design Variables. MIS quarterly,
1(1), 37–49.
Bierstaker, J. L. (2003). Auditor recall and evaluation of
internal control information: does task-speci?c knowledge
mitigate part-list interference?. Managerial Auditing Journal
18(1/2), 90–99.
Birnberg, J. G. (2000). The role of behavioral research in
management accounting education in the 21st century.
Issues in Accounting Education, 15(4), 713–728.
Bonner, S. E. (1990). Experience e?ects in auditing: the role of
task speci?c knowledge. Accounting Review, 65(1), 72–92.
Bonner, S. E. (1994). A model of the e?ects of audit task
complexity. Accounting, Organizations and Society, 19(3),
213–234.
Bonner, S. E., & Lewis, B. L. (1990). Determinants of auditor
expertise. Journal of Accounting Research, 28, 1–19.
Bonner, S. E., & Walker, P. L. (1994). The e?ects of
instructions and experience on the acquisition of audit
knowledge. Accounting Review, 69(1), 157–178.
Bonner, S. E., Davis, J. S., & Jackson, B. R. (1992). Expertise in
corporate tax planning: The issue identi?cation stage.
Journal of Accounting Research, 30, 1–28.
Bonner, S. E., Libby, R., & Nelson, W. (1997). Audit
category knowledge as a precondition to learning from
experience. Accounting, Organizations and Society, 22(5),
387–410.
Briers, M., Chow, C., Hwang, N., & Luckett, P. (1999). The
e?ects of alternative types of feedback on product-related
decision performance: A research note. Journal of Manage-
ment Accounting Research, 11, 75–92.
Bucheit, S. (2003). Reporting the cost of capacity. Accounting,
Organizations and Society, 28, 549–565.
Chandra, A., & Krovi, R. (1999). Representational congruence
and information retrieval: Towards an extended model of
cognitive ?t. Decision Support Systems, 25, 271–288.
Chi, M., Feltovich, P., & Glaser, R. (1981). Categorization and
representation of physics problems by experts and novices.
Cognitive Science, 5(2), 121–152.
Christensen, J., & Demski, S. J. (1995). The classical founda-
tions of modern costing. Management Accounting Research,
6(2), 13–32.
Cloyd, B. C. (1997). Performance in tax research tasks: The
joint e?ects of knowledge and accountability. Accounting
Review, 72(1), 111–131.
Datar, S. M., & Gupta, M. (1994). Aggregation, speci?cation
and measurement error errors in product costing. Account-
ing Review, 69(3), 567–591.
Davis, L. R. (1989). Report format and the decision makers
task: An experimental investigation. Accounting, Organiza-
tions and Society, 14(5/6), 495–508.
Dearman, D. T., & Shields, M. D. (2001). Cost knowledge and
cost-based judgment performance. Journal of Management
Accounting Research, 13, 1–18.
Dearman, D. T., & Shields, M. D. (2005). Reducing accounting
?xation: Determinants of cognitive adaptation to variation
in accounting method. Contemporary Accounting Research,
22(2), 351–384.
Desanctis, G., & Jarvenpaa, S. L. (1989). Graphical presenta-
tion of accounting data for ?nancial forecasting: An
experimental investigation. Accounting, Organizations and
Society, 14(5/6), 509–525.
Drake, A. R., Haka, S. F., & Ravenscroft, S. P. (1999). Cost
system and incentive structure e?ects on innovation,
e?ciency and pro?tability in teams. Accounting Review,
74(2), 323–345.
Frownfelter-Lohrke, C. (1998). The e?ects of di?ering infor-
mation presentations of general purpose ?nancial state-
ments on users’ decisions. Journal of Information Systems,
12(2), 99–107.
Ganzach, Y. (1993). Predictor representation and prediction
strategies. Organizational Behavior and Human Decision
Processes, 56, 190–212.
Gupta, M., & King, R. R. (1997). An experimental investiga-
tion of the e?ect of cost information and feedback on
product cost decisions. Contemporary Accounting Research,
14, 99–127.
600 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
Harvey, N., & Bolger, F. (1996). Graphs versus tables: E?ects
of data presentation format on judgmental forecasting.
International Journal of Forecasting, 12, 119–137.
Haynes, C. M., & Kachelmeier, S. J. (1998). The e?ects of
accounting contexts on accounting decisions: A synthesis
of cognitive and economic perspectives in accounting
experimentation. Journal of Accounting Literature, 17,
97–136.
Hwang, M. I. (1995). The e?ectiveness of graphic and tabular
presentation under time pressure and task complexity.
Information Resource Management Journal, 8(3), 25–31.
Jarvenpaa, S. L. (1989). The e?ects of task demands and
graphical format on information processing strategies.
Management Science, 35(3), 285–303.
Kaplan, R. S., & Narayanan, V. G. (2001). Measuring and
managing customer pro?tability. Journal of Cost Manage-
ment(September/October), 5–15.
Klincewicz, J. G. (1990). Solving a freight transport problem
using facility location techniques. Operations Research,
38(2), 99–109.
Libby, R. (1981). Accounting and Human Information Process-
ing. Englewood Cli?s, NJ: Prentice-Hall.
Libby, R. (1995). The role of knowledge and memory in audit
judgment. In R. Ashton & A. Ashton (Eds.), Judgment and
Decision-Making Research in Accounting and Auditing.
Cambridge, UK: Cambridge University Press.
Libby, R., & Luft, J. (1993). Determinants of judgment
performance in accounting settings: Ability, knowledge,
motivation, and environment. Accounting, Organizations
and Society, 18(5), 425–450.
Lucas, H. C. (1981). An experimental investigation of the use of
computer-based graphics in decision-making. Management
Science, 27(7), 757–768.
Luft, J. (1994). Bonus and penalty incentives: Contract choice
by employees. Journal of Accounting and Economics, 18,
181–206.
Mackay, D. B., & Villarreal, A. (1987). Performance di?erences
in the use of graphic and tabular displays of multivariate
data. Decision Sciences, 18, 535–546.
Maines, L. A., & McDaniel, L. S. (2000). E?ects of compre-
hensive-income characteristics on non-professional inves-
tors’ judgments: The role of ?nancial-statement
presentation format. Accounting Review, 75(2), 179–207.
Malina, M. A., & Selto, F. H. (2001). Communicating and
controlling strategy: An empirical study of the e?ectiveness
of the balanced scorecard. Journal of Management Account-
ing Research, 13, 47–90.
Mantrala, M. K., Prabhakant, S., & Zoltners, A. (1992). Impact
of resource allocation rules on marketing investment-level
decisions and pro?tability. Journal of Marketing Research,
29(May), 162–175.
Mauldin, E. G., & Ruchala, L. V. (1999). Towards a meta-
theory of accounting information systems. Accounting,
Organizations and Society, 24, 317–331.
Mooney, J. L., Rogers, C., & Wright, H. (2000). A guide to
selecting cost management software. Journal of Corporate
Accounting & Finance, 11(3), 19–27.
Moriarity, S. (1979). Communicating ?nancial information
through multidimensional graphics. Journal of Accounting
Research, 17(1), 205–224.
O’Donnell, E., Koch, B., & Boone, J. (2005). The in?uence of
domain knowledge and task complexity on tax profession-
als’ compliance recommendations. Accounting, Organiza-
tions and Society, 30(2), 145–165.
Painton, S., & Gentry, J. W. (1985). Another look at the impact
of information presentation format. Journal of Consumer
Research, 12(2), 240–244.
Pedhazur, E. J. (1982). Multiple regression in behavioral research
(2nd ed.). New York, NY: Holt, Rinehart, and Winston.
Remus, W. (1987). A study of graphical and tabular displays
and their interaction with environmental complexity. Man-
agement Science, 33(9), 1200–1204.
Rose, J. M., & Wolfe, C. J. (2000). The e?ects of system design
alternatives on the acquisition of tax knowledge from a
computerized tax decision aid. Accounting, Organizations
and Society, 25, 285–306.
Rose, J. M., Douglas, R. F., & Rose, A. M. (2004). A?ective
responses to ?nancial data and multimedia: The e?ects of
information load and cognitive load. International Journal
of Accounting Information Systems, 5, 5–24.
Shields, M. D. (1995). An empirical analysis of ?rms’ imple-
mentation experiences with activity-based costing. Journal
of Management Accounting Research, 7, 148–166.
So, S., & Smith, M. (2002). Colour graphics and task
complexity in multivariate decision-making. Accounting,
Auditing and Accountability Journal, 15(4), 565–592.
So, S., & Smith, M. (2004). Multivariate decision accuracy and
the presentation of accounting information. Accounting
Forum, 283–305.
Spilker, B. C. (1995). The e?ects of time pressure and
knowledge on key word selection behavior in tax research.
Accounting Review, 70, 514–528.
Sprinkle, G. B. (2003). Perspectives on experimental research in
managerial accounting. Accounting, Organizations and Soci-
ety, 28(2-3), 287–318.
Stock, D., & Watson, C. J. (1984). Human judgment accuracy,
multidimensional graphics, and human versus models.
Journal of Accounting Research, 22, 192–206.
Stone, D. N., Hunton, J. E., & Wier, B. (2000). Succeeding in
managerial accounting. Part 1: Knowledge, ability and
rank. Accounting, Organizations and Society, 25, 697–
715.
Sullivan, J. J. (1988). Financial presentation format and
managerial decision-making: Tables versus graphics. Man-
agement, 2(2), 194–216.
Tan, H. T., & Libby, R. (1997). Tacit managerial versus
technical knowledge as determinants of audit expertise in
the ?eld. Journal of Accounting Research, 35(1), 97–113.
Tan, H. T., & Kao, A. (1999). Accountability e?ects on
auditors’ performance: the in?uence of knowledge, problem
solving ability, and task complexity. Journal of Accounting
Research, 37(1), 209–223.
Tan, H. T., Ng, B. P., & Mak, W. Y. (2002). The e?ects of task
complexity on auditors’ performance: The impact of
E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602 601
accountability and knowledge. Auditing: A Journal of
Practice and Theory, 21(2), 81–95.
Tufte, E. (2001). The visual display of quantitative information
(second ed.). Cheshire, CT: Graphics Press.
Umanath, N. S., & Vessey, I. (1994). Multiattribute data
presentation and human judgment: A cognitive ?t perspec-
tive. Decision Sciences, 25, 795–823.
Vera-Mun˜ oz, S. C. (1998). The e?ects of accounting knowledge
and context on the omission of opportunity costs in
resource allocation decisions. Accounting Review, 73(1),
47–72.
Vera-Mun˜ oz, S. C., Kinney, W. R., & Bonner, S. E. (2001). The
e?ects of domain experience and task presentation format
on accountants’ information relevance assurance. Account-
ing Review, 76(3), 405–429.
Vessey, I. (1991). Cognitive ?t: A theory-based analysis of the
graphs versus tables literature. Decision Sciences, 22(2),
219–240.
Waller, W., Shapiro, B., & Sevcik, G. (1999). Do cost-based
pricing biases persist in laboratory markets? Accounting,
Organizations and Society, 24(8), 717–739.
Wood, S. L., & Lynch, J. G. (2002). Prior knowledge and
complacency in new product learning. Journal of Consumer
Research, 29(3), 416–426.
Wright, W. F. (1995). Superior loan collectibility judgments
given graphical displays. Auditing: A Journal of Theory and
Practice, 144–154.
Yates, J. (1985). Graphics as a managerial tool. A case study of
DuPont’s use of graphs in the early twentieth century.
Journal of Business Communications, 22, 5–33.
602 E. Cardinaels / Accounting, Organizations and Society 33 (2008) 582–602
doc_998589789.pdf