.‘ . a . . . 7 gamma; 3.1%»?
W. , . . . . db... ..
.3Mwh. a" . . .

Orb
”maximum.
. :n. 5..

 

.3

-~a
p.“

. ;

 

 

IMWW‘W

This is to certify that the
dissertation entitled

The Effect of Industry Knowledge
0n Cost Driver Selection

presented by

Barbara Lamberton

has been accepted towards fulfillment
of the requirements for

Ph . D 1 degree in _Ac_c_an.r_;Lng

a, UL

Major )ruf nor

 

 

 

 

Date May 7. 1998

MSU is an Affirmative Action/Equal Opportunity Institution 0-12771

 

to remove thi
TO AVOID FINE-S retur

 

 

LIBRARY

Michigan State ,

University

 

 

PLACE IN RETURN BOX
5 checkout from your record.
n on or before date due.

 

DATE DUE

MTE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1/98 cICIRCJDIaDquGs-p.“

THE EFFECT OF INDUSTRY KNOWLEDGE ON COST DRIVER SELECTION

By

Barbara Lamberton

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Accounting

1 998

C0pyright by
BARBARA LAMBERTON
1998

ABSTRACT

THE EFFECT OF INDUSTRY KNOWLEDGE ON COST DRIVER SELECTION
BY

Barbara Lamberton

The purpose of this study is to examine determinants of
skilled cost driver selection through a controlled laboratory
experiment with objective performance criteria. Although auditing
research has investigated knowledge and ability effects on audit
performance, little behavioral research has been done on cost
driver selection.

Two types of knowledge were examined, industry specific
manufacturing knowledge and general management accounting
knowledge. The sample included student volunteers and was
comprised of participants with high and low levels of general
management accounting knowledge. To induce industry
knowledge, half of the participants were randomly assigned to a
training session related to the production process of a package
printing plant. The training session was derived from materials

used by the industry trade association.

The results suggest that superior management accounting
knowledge substitutes for low ability and lack of industry specific
manufacturing knowledge. In particular, superior management
accounting knowledge allowed participants to recognize highly
biased cost drivers without the benefit of specialized knowledge of
the manufacturing process. In contrast, for those with low levels
of management accounting knowledge, both industry specific
knowledge and ability had a significant effect on performance.

The study also suggests that both industry specific and
management accounting knowledge affect success at selecting the
driver with the lowest tracking cost out of several equally accurate
alternatives.

Demonstrating a substitution effect between knowledge and
ability provides a unique contribution to the accounting literature.
Previous accounting research has been unable to demonstrate that
one type of knowledge may be able to substitute for another type or
for weaknesses in ability. Examining the effects of different types
of knowledge on performance is a logical starting point for a
research agenda examining the relationship between technology

and individual differences in a management accounting setting.

This is dedicated to my loving husband and best friend, Donald.

ACKNOWLEDGMENTS

No dissertation is ever written alone, and this dissertation is no
exception. Without the guidance and support of a number of people, this
work could not have been accomplished.

My chairman, Fred Jacobs, had been a significant source of advice and
encouragement. Fred's keen research mind has been instrumental in ways
too numerous to list. Always with a view toward quality research, Fred has
been a true mentor.

The other members of my committee were very crucial to this work as
well. Sev Grabski provided invaluable insight on behavioral issues related to
systems. Joan Luft kept the focus on research design issues related to the
study of knowledge effects. Frank Boster brought to the committee his
incredible command of research design and statistical analysis issues. Bill
McCarthy provided the inspiration for a study of knowledge effects as it relates
to systems.

I also acknowledge the help of the industry expert who provided advice
as well as the means to acquire the industry training materials. I also
appreciate the time and effort of the midwest controller who reviewed the
experimental materials for realism.

I thank the student volunteers who so willingly gave of their time and
effort to serve as participants. I am also grateful to the Department of
Accounting at Michigan State, General Electric and the Patricia Roberts Harris
fellowship for important financial assistance.

Most importantly, I am grateful for the love, support and patience of a
good husband and family. I am fortunate to be part of a loving, caring family
that has instilled in me the importance of a spiritual perspective, perseverance
and faith.

vi

TABLE OF CONTENTS

List of Tables ............................................. x
List of Figures . . .. ......................................... xii

CHAPTERI INTRODUCTION
1.0 Overview ........................................
1.1 Research Question ...............................
1.2 Groundwork .....................................
1.3 Importance of the Research Question ................
1.4 The Institutional Setting: Package Printing ............
1.5 Summary ........................................

Scam—3.5.;

CHAPTER II LITERATURE REVIEW
2.0 Overview ........................................ 11
2.1 Cost Aggregation ................................. 11
2.2 Determinants of Skilled Performance ................. 15
2.2.1 Differences between exgerts versus novices. . . . 15

2.2.2 Tyges of knowledge and skilled gerformance . . . 21
2.2.3 The link between abilig and skilled gerformance 23

2.3 Hypotheses Development ........................... 27
2.3.1 The determinants of success at accuracy ...... 27
2.3.2 Tracking cost hygotheses ................... 33
2.4 Summary ........................................ 35
CHAPTER III METHODOLOGY
3.0 Overview ........................................ 37
3.1 Experimental Design ............................... 37
3.2 Participants ...................................... 39
3.2.1 Samgle statistics ........................... 39
3.2.1 Task Incentive ............................. 42

vii

3.3 Task ........................................... 42

 

3.3.1 Cost computation task ..................... 44

3.3.2 Basic ABC knowledge ...................... 50

3.3.3 Dependent variable one: Accuracy ............ 51

3.3.4. Dependent variable two: Tracking cost ........ 55

3.3.5 Pretests .................................. 56

3.3.6 Abilig variable ............................ 59

3.3.7 Manipulation checks ........................ 60

3.4 Methods of Analysis .............................. 61

3.4.1 Confirmatogy Factor Analysis ................ 61

3.4.2 ANOVA ................................... 61

3.4.3 Regression ............................... 61

3.5 Summary ........................................ 63

CHAPTER IV DATA ANALYSIS

4.0 Overview ......................................... 64

4.1 Confirmatory Factor Analysis ........................ 65

4.2 Manipulation Checks ............................... 68
4.2.1 Manipulation check: Familiarity

with printing ............................... 68

4.2.2 Manipulation check: Knowledge of colors ....... 71

4.3 An analysis of the properities of the key variables ...... 75

4.3.1 Ability .................................... 76

4.3.2 Accuracy .................................. 78

4.3.3 Tracking cost .............................. 82

4.4 H. and H2: Determinants of skill at accuracy ........... 85

4.4.1 Potential interactions effects: accuracy ........ 87

4.4.2 Logistic regression .......................... 91

4.4.3 OLS regression ............................. 92

viii

4.5 H3 and H4: Determinants of skill at noticing tracking cost. . 99

4.5.1 Potential interactions effects: tracking cost ..... 100
4.5.2 Logistic regression ......................... 107
4.5.3 OLS regression ............................. 109

4.5.4 OLS regression- without the borderline cases . . . 109
4.5.5 OLS regression --- corrected for measurement

error ...................................... 115

4.6 Summary ......................................... 119
CHAPTER V CONCLUSION

5.0 Overview ....................................... 122

5.1 Summary of results ............................... 122

5.2 Limitations ..................................... 123

5.3 Contributions and Future Extensions: Systems ....... 126

5.4 Contributions: Education .......................... 128

5.5 Future Extensions-«Economic implications ........... 128

APPENDIX A ............................................. 130

APPENDIX B ............................................. 133

REFERENCES ............................................. 146

3.1
3.2
3.3
3.4

3.5

3.6

3.7

4.1

4.2

4.3

4.4

4.5
4.6
4.7

4.8

4.9

LIST OF TABLES

Sample statistics ...................................... 41
Straightforward cost problem ............................ 46
Difficult problem ...................................... 48
Cost computation: Difficult problem ...................... 49
Magnitude of the batch error ............................. 50
Ratio of resource usage ................................. 55
Cost of Tracking ....................................... 56
Familiarity with printing ................................. 69
t test: Familiarity ....................................... 70
Colors ................................................ 72
t test: Colors .......................................... 73
t test: Colors by Genknow ............................... 74
t test of Ability: Full Sample ............................. 77
t test of Ability: Reduced Sample ......................... 79
Frequencies: Accuracy ................................. 81
Frequencies: Tracking cost .............................. 84

4.10

4.11

4.12

4.13

4.14

4.15

4.16
4.17
4.18
4.19

4.20

4.21

4.22

4.23

4.24

ANOVA of Accuracy: Full sample .......................... 89

ANOVAofAccuracy:Genknow=0........................ 90
Logistic Regression of Accdic ........................... 93
OLS Regression of Accuracy ............................. 94
Corrected OLS Regression of Accuracy ................... 96
ANOVA of Tracking cost ................................ 104
ANOVA of Trackdic ..................................... 105
Logistic Regression of Trackdic: N=116 ................... 108
Logistic Regression of Trackdic: N=1 14 ................... 1 10
OLS Regression of Tracking cost: N=116 .................. 111
OLS Regression of Trackdic: N=1 16 ....................... 112
OLS Regression of Tracking cost: N=114 ................... 113
OLS Regression of Trackdic: N=114 ....................... 114
Corrected OLS Regression of Tracking cost: N=116 .......... 1 16
Corrected OLS Regression of Tracking cost: N=114 .......... 117

xi

1.1

2.1 '

3.1

3.2

3.3

4.1

4.2

4.3

4.4

4.5

4.6

4.7

4.8

4.9

4.10

4.11

4.12

LIST OF FIGURES

Cost driver selection .................................. 4
Simple Heuristic ..................................... 20
Skilled Performer: Accuracy ............................ 54
Skilled Performer: Tracking cost ......................... 57
Unskilled Performer: Tracking cost ....................... 58
Ability Histogram ...................................... 76
Accuracy Histogram .................................... 78
Tracking cost Histogram ................................ 82
Genknow versus Industry: Accuracy ...................... 87
Industry versus Ability: Accuracy ......................... 91
Accuracy Hypotheses ................................... 97
Genknow versus Industry: Tracking cost .................. 101
Industry versus Ability: Tracking cost ..................... 102
Genknow by Ability: Tracking cost ........................ 103
Industry versus Ability: Tracking cost, Genknow = 1 ......... 106
Industry versus Ability: Tracking cost, Genknow = 0 ......... 107
Tracking cost Hypotheses .............................. 118

xii

CHAPTERI
INTRODUCTION

1.0 Overview

The research question is discussed in section 1, while
Section 2 lays the groundwork for the study. Contributions of the
study are discussed in Section 3 and the institutional setting,
package printing, is discussed in Section 4.
1.1 Research Question

The purpose of this study is to test the degree to which
variation in cost driver selection can be explained by individual
differences in knowledge and ability. This study shows that
individual differences affect the way decision makers use and
process the information needed to evaluate alternative systems
designs. Although auditing research has investigated knowledge
and ability effects on audit performance, the effects of these
variables on cost driver selection have not been examined.
1.2 Groundwork

This study starts with the assumption that the decision
maker (DM) makes rational choices when evaluating alternative
systems designs. The skilled DM is expected to have the
knowledge and/or training required to effectively evaluate systems

with alternative combinations of cost drivers using some type of

cost-benefit perspective.

2

As a minimum, the DM needs to be able to differentiate
between a system that provides accurate costs and another
system that results in high cost distortion. To eliminate highly
inaccurate drivers, the DM needs to have enough skill to identify
which cost drivers are clearly uncorrelated with a given activity.

In this study, competence at eliminating highly inaccurate drivers
is called the accuracy skill.

Although skill at recognizing the difference between
accurate and biased cost drivers is critical, it is only one aspect
of evaluating alternative systems designs. For example, it is
reasonable to assume that there may be several cost drivers that
provide the same benefit in terms of accuracy. If that is the
case, the skilled DM would be expected to recognize a situation in
which one set of cost drivers provides the same level of benefit as
another but at a lower tracking cost. By considering differences
in relative tracking costs, the DM reduces the chance of spending
more than necessary to obtain a given level of accuracy in the
cost system. In this study, competence at selecting the most
accurate, least costly system is called the tracking cost skill.

A controlled laboratory experiment was employed to test the
effects of knowledge and ability using an experimental stimuli with

objective criteria. To that end, participants were given a series

3

of problems and asked to recommend a design that provides
accurate information at the lowest tracking cost. The experiment
was designed to distinguish between skill at accuracy and skill at
tracking cost.

As shown in figure 1.1, this study seeks to demonstrate that
success at identifying accurate drivers and success at noticing
tracking costs represent sub-components of the cost driver
selection process. This distinction is important since it is
conceivable that the two sub-components correspond to separate
costs associated with making sub-optimal systems implementation
decisions. The first cost results from implementing a system with
highly distorted costs. To the extent that accurate costs are
imperative for decision making, relying on a system with distorted
costs may lead to poor decisions resulting in economic loss. The
second cost relates to the cost of maintaining a given system
design. The economically rational DM would be expected to
explicitly consider the relative cost of tracking various drivers to
avoid implementing a more expensive system than necessary. In
this study, skill at the accuracy part of the task proxies for skill at
quantifying the opportunity cost of a bad decision. Similarly, skill
at cost driver selection proxies for skill at quantifying the tracking

costs of a given system design. .

Figure 1.1
COST DRIVER SELECTION

 

Objective:
Determine the most accurate set of cost drivers subject to
minimizing tracking cost

   
 

 

ACCURACY :
Eliminate cost drivers that would provide highly distorted
costs.

TRACKING COST:
Select one set of cost drivers that is least costly to track.

   
 

 

5

The direct implication of using a cost-benefit perspective is
that the DM needs to have the knowledge and/or ability to quantify
both types of costs. Yet, it is not certain that the types of
knowledge required to quantify both types of cost are the same,
nor is it certain when and how the knowledge is acquired. In
fact, very little is known about the cognitive processes and
knowledge requirements associated with selecting cost drivers.

The primary message of this study is that knowledge and
ability have profound and different effects on the two sub-
components of the cost driver selection process. This study
suggests that various types of knowledge, such as industry
specific knowledge and general management accounting
knowledge, affect components of cost driver selection differently.
In addition, it is not clear how knowledge and ability relate to task
performance. According to Libby, "superior ability may allow
inferences to be made which may substitute for incomplete
knowledge."(1994, p. 13).

lntuitively, an individual with industry specific experience
would be expected to have substantive knowledge about the
production process in a particular institutional setting. It is
conceivable that such knowledge would make correlation among

competing cost drivers salient. In turn, this salience may reduce

6

the complexity of cost driver selection by reducing the number of
competing cost drivers that need to be evaluated.

Prior research has not provided much insight on effects of
different types of knowledge on performance of a management
accounting task. Nor has research indicated whether one type of
knowledge can substitute for another. A well-trained accountant
familiar with general management accounting concepts may be
able to perform at the same or better level than the industry
trained individual.

1.3 Why this question is important

Management accounting systems have been criticized as
being irrelevant and out of step with the information needs of an
advanced manufacturing environment. The implication is that the
information needed to develop cost savings' strategies,
investment justifications, and pricing decisions is simply not
available. Activity based accounting (ABC) has been proposed as
the solution to this problem. The focus of ABC is on collecting
and storing more detailed information, called a cost driver, about
manufacturing overhead costs. Proponents of ABC suggest that a
system with multiple cost drivers will enhance understanding of
costs and lead to better decisions.

Not all researchers agree that more is better than less when

7

it comes to cost drivers. The research of Datar and Gupta (1994),
Gupta (1993) and Banker and Potter (1993) suggest that caution
be used when deciding whether or not to increase the amount of
detailed information being tracked by a management accounting
system. Gupta (1993) demonstrates that increasing the number of
cost drivers does not always increase product cost accuracy.
Datar and Gupta (1994) show that careless selection of drivers
may lead to implementing a system that provides less accurate
costs. Banker and Potter (1993) identified specific situations in
which a firm would be better off economically with a single cost
driver system. The implication is that cost driver selection is a
critical decision in the design of management accounting systems.
Economic benefits would seem to be associated with careful
selection of cost drivers.

In terms of previous work on cost driver selection, analytical
research suggests that knowledge about the correlation among
cost drivers is crucial to efficient evaluation of alternative systems
designs (Dewan and Magee, 1992; Babad and Balachandran,
1993). Specifically, researchers have demonstrated that efficient
cost driver selection exploits correlations among cost drivers to
reduce the complexity of the task. In the current study it is shown

that some participants are better than others at identifying

8

meaningful resource consumption patterns about potential cost
drivers. In particular, it is shown that individual differences
affect success at recognizing patterns of high correlation among
competing cost drivers.

By increasing our understanding of the cost driver selection
process, this study has both practical and theoretical value. The
results are relevant to firms planning to implement changes in
their cost accounting system, such as ABC. Although researchers
have begun to investigate ABC empirically ( Foster and Gupta,
1990; Banker and Johnston, 1993) and analytically (Datar, et al.
1993; Hwang, et al. 1993; Gupta, 1993), little or no research has
been done examining the effect that individual differences have
on evaluation of alternative systems designs.

In the systems design area, the results should help in the
construction of more effective systems, decision aids, and
development teams. Demonstrating the effect of different types of
knowledge on performance is considered a logical starting point
for a research agenda examining the relationship between
technology, knowledge and ability in a manufacturing setting.

The results of this study should also be helpful in the design
of learning experiences that allow efficient acquisition of

knowledge for individuals of varying ability levels. This study is

9

expected to provide some evidence of the benefit of instructional
strategies that use real world manufacturing examples in the
classroom.

1.4 The Institutional Setting: Package Printing

The institutional setting used in the study reflects cost
behavior patterns of a package printer experiencing a change in
product mix. Prior to the mix change, the printer’s single cost
driver system was considered adequate for decision making
purposes. The current problem facing the printer is to determine
if and how the system needs to be upgraded.

The package printing industry was chosen due to the
inherent complexity of the manufacturing process and the
potential for diversity in product mix. Analytical research
demonstrates (Hwang, Evans, and Hedge, 1993) that the demand
for multiple cost drivers is a function of the heterogeneity of the
production process and the diversity of the product mix. Similarly,
Gupta (1993) found a positive effect between complexity and the
magnitude of cost distortions caused by using fewer cost drivers.
The package printing industry was also chosen due to the
availability of an industry expert and industry training materials.
The package printing industry is a major industry with whole-wide

sales of $120 billion.

10

Because of the inherent complexity of the production
process in package printing, the performance task focuses on
determining the best cost driver(s) to use in one major activity,
press setup. Press setup is a complex and costly activity for the
package printing industry. Just-in-time demands, changing mix
and other forces in the market have made press setup a
strategically critical activity in package printing.

1.5 Summary

This chapter introduced the research question in this study
and presented some contributions expected from this research.
There are four chapters that follow. Chapter II is a literature
review and Chapter III presents the methodology used in this
study. Chapter IV describes the data analysis. Limitations,

contributions and implications are summarized in Chapter V.

CHAPTER II
LITERATURE REVIEW

2.0 Overview

The theoretical background for this study utilizes literature
about cost aggregation and skilled performance. The issues
related to cost aggregation are addressed in the first section.
The determinants of skilled performance are addressed in the
second section. Development of the hypotheses is found in
section 3.
2.1 Cost Aggregation

A critical decision in designing a management accounting
system is determining the number and type of cost drivers to
include in the information system. Since measuring all potential
cost drivers and activities may be impractical, management
usually needs to limit the number and type of cost drivers to be
tracked by the information system. Some aggregation of
activities and cost drivers is typically part of design of the system
upgrade.

Historically, issues related to the cost aggregation problem
(CAP) have interested accounting researchers. In terms of
activity based costing, the CAP refers to the need to limit the

number of cost drivers being tracked by a company's system.

11

12

The CAP has received considerable amount of attention in the
analytical literature (Demski, 1980; Feltham, 1977; Demski and
Feltham, 1976; Demski and Feltham, 1972; Feltham and Demski,
1970). The information economics model with its emphasis on an
optimum solution and cost-benefit criterion has been viewed as
the theoretical standard for evaluating accounting choice
problems regarding aggregation.

Superficially, it would appear the information economics
model would provide a reasonable theoretical framework for
understanding cost driver selection. The task of selecting a cost
driver could be considered a sub-component of the cost
aggregation problem (CAP). Several authors (Dopuch, 1993;
Dewan and Magee, 1992), however, have suggested that the
information economics model may not be an appropriate and
practical reference point to guide research about the cost
aggregation decision. As a consequence research has shifted to
exploring heuristics used to solve the CAP (Dewan and Magee,
1992; Babad and Balachandran, 1993).

Dewan and Magee suggest that decision makers are likely to
rely on heuristics to reduce the amount of time required to solve
the CAP. Defining the objective function as minimizing the sum

of opportunity and tracking costs, Dewan and Magee used

13

simulations to evaluate heuristic approaches to solving the CAP.
Dewan and Magee’s objective function assumes that there are two
costs that need to be considered when solving CAP. The first
cost is the opportunity cost of a bad decision. The second cost
is the cost of tracking a given number of cost drivers. The
Dewan and Magee results indicate that heuristic performance is
significantly affected by the degree of correlation among the
various cost drivers.

In 1993, Babad and Balachandran took a slightly different
perspective from Dewan and Magee by explicitly incorporating
product cost accuracy in the cost driver selection process.

Unlike Dewan and Magee, the approach taken by Babad and
Balachandran involved several perfectly correlated cost drivers.
In addition, the objective function was less complex. The model
presented by Babad and Balachandran was based on maximizing
a given level of accuracy subject to minimizing tracking costs.

Like Dewan and Magee, Babad and Balachandran found that
the degree of correlation among the cost drivers was a major
factor affecting the process of cost driver selection. For
example, Babad and Balachandran demonstrated that perfect
correlation among cost drivers can be used to reduce the number

of different drivers that need to be evaluated. The authors

14
presented a proof demonstrating that perfectly correlated drivers
may be substituted for one another with no loss in product cost
accuracy.

From a behavioral perspective, the work of Babad and
Balachandran has some parallels to Dewan and Magee’s work on
the CAP. First, both papers present models with objective
criteria of success. For the behavioral researcher, the
availability of objective criteria for performance is potentially
valuable in an area where objective criteria are difficult to find
and support.

Second, both papers emphasize that characteristics of the
data, namely correlations among the cost drivers, have a
significant effect on cost driver selection. In that regard, both
papers presume that the decision maker can quickly recognize
strong versus weak correlations among the potential cost drivers.
In addition, in both papers, the decision maker needs to be able
to make complex computations and comparisons. The common
thread throughout these works is the lack of any explicit
discussion about the skill of decision makers. Both studies
assume that the decision maker is adept at identifying and
computing the net benefit of one system design versus another.

Examination of the effect of decision makers characteristics on

15

performance is left to future research.
2.2 Determinants of skilled performance

Our current understanding of skilled performance has
developed from over thirty years of research in cognitive
psychology and auditing behavioral research. Overall, the
research indicates that skilled performers, called experts, have
specific characteristics that differentiate them from less skilled
performers, called novices. The literature also provides support
for the concept that skilled performance is a function of different
types of knowledge and innate ability. Each of these issues will
be discussed in the following sections.

Section 2.2.1 summarizes the key research related to the
differences between experts and novices. Section 2.2.2 covers
behavioral research about the effects of various types of
knowledge on skilled performance. 2.2.3 presents the literature
about the link between ability and skilled performance.

2.2.1 Differencesbietween experts versus novices
Behavioral researchers in a variety of different domains have
investigated the differences between experts and novices. Prior
research suggests that skilled performers tend to view relevant
cues in a coherent, meaningful pattern (Newell and Simon,1972;

Chase and Simon, 19733). Researchers have also found that

16

experts are especially adept at classification and categorization
of various problem types (Hinsely, Hayes and Simon, 1978). The
consensus (Bedard and Biggs, 1991; Lesgold et al. 1988; Akin,
1980) is that experts tend to focus on salient characteristics of a
problem while novices tend to look at superficial properties.
Experts are thought to use knowledge about underlying principles
of their given domain to differentiate between significant and
superficial aspects of the problem. Novices, on the other hand,
are thought to rely primarily on superficial features which may be
irrelevant to the task at hand. The tendency for novices to rely
on potentially irrelevant factors suggests that their performance
deficiencies reflect deficiencies in knowledge.

To understand the differences in knowledge between
experts and novices, Chi et al. (1982) conducted eight studies
using the domain of physics. The stated objective of these
studies was to provide some empirical evidence about the
differences between experts and novices in a context which
requires command of a complex knowledge domain.

The particular area of physics chosen for the Chi et al.
studies was mechanics. The expert participants ranged from
physics professors to graduate students in physics. The novices

were students who had taken a mechanics course. A variety of

17

tasks were used in the studies including sorting problems, writing
assignments and protocol analysis. The results indicate that
experts categorize and represent problems in terms of specific
laws of physics, such as Newton's Second Law or the
Conservation of Energy Law. In contrast, the protocols of the
novices tend to be dominated by statements about the physical
aspects of the problem. For example, physics novices tend to
focus on the fact that the problem involves a spring or a pulley
rather than the law of physics involved.

Based on the results of a hierarchical sorting task and a
writing assignment, Chi et al. found evidence that classification
schemes used by expert physicists are more extensive, organized
and interrelated than those of novices. The experts used the
laws of physics as the primary classification category and
considered the surface features as subordinate categories. The
experts classification schemes took into consideration both
underlying principles of physics and superficial properties.
Novices, on the other hand, focused on superficial, physical
aspects of the problem. The implication of the Chi et al studies
is that the knowledge of experts and novices was different and the
difference in knowledge affected performance.

According to Bonner and Pennington (1991) the organized

18
and extensive knowledge of the expert translates to two distinct
advantages in performance. First, the expert’s knowledge
provides a reference point that aids in interpreting the facts of a
given problem. Knowledge assists the expert in matching the
pattern of the facts and features of the problem at hand to known
underlying principles of the given domain. Thus, this skill may be
a reflection of the tendency of experts to rely on knowledge to
represent a problem that requires combination of multiple cues.
Bonner and Pennington used the term “global interpretation of the
situation” to describe problem representation, skill at establishing
a framework for problem solving.

Second, the expert’s knowledge may include the actions and
procedures relevant to the problem at hand. According to Chi et
al. “experts’ schemata contain much more knowledge about the
explicit conditions of applicability of the major principles
underlying a problem” (1982,p.62). Two examples include the
chess masters command of defense and attack strategies (Chase
and Simon, 1973) and procedural knowledge demonstrated by
expert physicists (Chi et al., 1982).

Consistent with other domains, the auditing behavioral
research also suggests that novices tend to represent a problem

on a more superficial level than experts. Bedard and Biggs

19

(1991) imply that auditors that make errors may be focusing on
the surface features of the task rather than relying on their
knowledge of the underlying accounting principles.

Research from other domains would suggest that differences
in knowledge would have a profound effect on performance for
management accounting tasks, such as cost driver selection. In
presenting the topic of cost driver selection, many managerial
texts (Anderson and Sollenberger, 1994; Noreen and Garrison,
1996, Zimmerman, 1993) employ a heuristic that classifies
activities into mutually exclusive categories and uses these
categories to simplify cost driver selection. Using the heuristic,
activities such as assembly and fabrication are classified as
volume driven and a volume driver is selected. Similarly,
machine setup would be classified as batch-level and the use of
number of setups would be suggested. Once the classifications
of the activities and cost drivers have been learned, the heuristic
becomes simple to use. For example, the selection of a cost
driver for assembly is limited to a few volume-driver cost drivers.
Similarly, a batch-level cost driver, such as number of setups,
would be selected for a batch-level activity, such as machine
setup. In terms of the current study, Figure 2.1 shows how the

simplistic heuristic would be applied to the press setup activity in

20

Figure 2.1
Simple heuristic

Machine setup is a batch-level activity that requires a

batch-level cost driver.

Identify press setup as a type of machine setup.

Identify number of press setups as a type of batch-level

 

 
 

 
 

cost driver.

 

Select number of press setups as the cost driver.

21
package printing.

The concept of reviewing all possible combinations of cost
drivers is not typically discussed in managerial texts. Instead, a
number of problems in the text materials test the student’s
proficiency at classifying activities based on the heuristic just
discussed. For novices, the simplicity of the heuristic may hide
the underlying principle that, all else being the same, cost drivers
are selected for their correlation with a given activity. Evidence
from other domains would suggest that the novice decision
makers, when faced with cost driver selection, may ignore the
underlying principle of high correlation and focus on some
“surface feature” of the problem. Novice individuals may select
cost drivers based solely on the name of the activity entirely
ignoring resource consumption patterns. Skilled performers
would be expected to focus on the underlying principle that cost
drivers and their related activities need to be highly correlated.
In contrast, less knowledgeable decision makers may tend to
focus on surface features of the task, such as the name of the
activity.

2.2.2 Iypes of knowledge and skilled_performa_n_c_e_ In
1990, Bonner and Lewis (BL) proposed that skilled performance is

a function is different types of knowledge. To test the effect of

22

different types of knowledge on performance, the authors
constructed a series of knowledge tests and administered them to
auditors with varying amounts of experience. The types of
knowledge examined by BL include the following: (1) world
knowledge (2) general domain knowledge and (3) sub-specialty
knowledge. The definitions of general domain knowledge and
sub-specialty knowledge are most relevant to this study.

BL define general knowledge as the type of knowledge that
virtually everyone in a particular domain would have the
opportunity to acquire through instruction and/or experience.
Knowledge of internal controls, proficiency with certain audit
computations and an understanding of the basic accounting model
are examples of general knowledge in the audit domain.

Sub-specialty knowledge, as defined by BL, refers to the
knowledge that is acquired through experience with specific
industries and/or clients. Specific knowledge about interest rate
swaps and industry experience with manufacturing were two types
of sub-specialty knowledge tested by BL.

The BL study examined knowledge effects related to four
audit tasks that had been the subject of previous auditing
behavioral research. General accounting knowledge of internal

controls was positively related to performance of an internal

23

control task and knowledge of the analytical procedures was
positively related to ratio analysis. The specialized knowledge of
hedging transactions, a type of sub-specialty knowledge, was
positively related to performance of an audit financial instruments
task. The BL findings provide some preliminary evidence that,
for auditing, general and specialized knowledge are separate
constructs that have different effects on task performance.

2.2.3 The linkpetween a_bility aLd skilflperformyapnce
Research in psychology suggests that ability is another
determinant of skilled performance (Hunter, 1986; Lesgold, 1984;
Simon, 1979). Hunter (1986) summarized the results of 515
studies conducted by the US Employment Service and data from
nearly half a million military personnel. The results indicate that
while the predictive validity of ability is highest for complex jobs,
ability is nevertheless a valid predictor for virtually all jobs.

In studying the determinants of audit performance, BL argue
that certain types of tasks tend to require a certain level of ability.
Specifically, BL found that ability was positively correlated with
performance for analytical review and an earnings manipulation
task.

To increase understanding of the link between ability and

performance, Libby and Tan (1992) reexamined the BL data. As

24

part of this reexamination, Libby and Tan presented and tested a
classification scheme that categorized each of the BL tasks as
either structured or unstructured. The authors suggest that
unstructured tasks require problem solving ability while structured
tasks do not. According to Libby and Tan (1992), an
unstructured task is any task which requires, to some degree, the
need to “define the problem, generate alternative solutions,
search for information from disparate sources and make
computations.” Using this classification scheme, Libby and Tan
argued that the internal control and financial instruments tasks
are fairly structured. The internal control task required the
participant to: (1) list two financial statement errors that could
occur in spite of the internal control system and; (2) list two audit
procedures that would detect the errors. The financial
instruments task required the participant read about an interest
rate swap agreement, name the type of transaction involved and
the accounting required. In both tasks, the problem was well-
defined and there was no need to search for information from
different sources. Neither task required high levels of ability for
performance.

Libby and Tan(1992) classified the two other tasks studied

by BL, ratio analysis and an earning manipulation task, as being

25

unstructured. The ratio task requires the participant to identify
an accounting error that would account for unusual changes in
several financial ratios. The earnings manipulation task requires
the participant notice the relationship between a pattern of errors
and a management compensation agreement described in a
footnote. Since both tasks require computations, generation of
alternative solutions and search for information from disparate
sources, ability was predicted to affect performance. The results
indicated that ability was significant for the ratio analysis task and
marginally significant for the earning manipulation task.

BL (1990) and Libby and Tan (1992) were not the first
accounting researchers to link ability and performance. In 1979,
Benbasat and Dexter found an interaction between ability and
level of aggregation. In 1982, Otley and Dias studied the
combined effects of ability, aggregation level and information
content on performance. The authors predicted that the low
ability participants would have more difficulty in a management
accounting task than high ability participants. For a variety of
methodological reasons, the Otley and Dias experimental results
did not support a significant effect related to ability.

In the behavioral literature, the term, ability, is often used to

describe general intelligence. Since ability is a difficult construct

26

to measure, a number of different instruments have been
employed by researchers. For example, Hunter’s (1986) definition
of ability was based on the US. Employment General Aptitude
Test Battery (GATB). Other measures used in behavioral
research include selected GRE questions (Bonner and Lewis,
1990) and various timed tests.

In accounting, behavioral researchers have frequently relied
on the theory of field independence to define one kind of ability
that is believed to be relevant to certain accounting tasks (Awashi
& Pratt, 1990; Gul, 1984; Otley and Dias, 1982; Benbasat and
Dexter, 1979; Gul & Zaid, 1981; Lusk, 1973; Doktor, 1973).

Field independence theory considers an individual’s style of
perception as a type of ability. The theory focuses on the
individual’s skill at isolating simple figures from complex
diagrams. In field independence terminology, individuals who are
adept at noticing simple patterns in complex diagrams are said to
be field independent and thus high ability. In contrast,
individuals who have difficulty isolating simple patterns are called
field dependent or low ability. The theory predicts that field
independent individuals tend to perform relatively well at problem
solving and excel at analyzing and structuring certain tasks. In

this study, the construct of field independence was used to define

27

ability. Therefore, throughout the remainder of this paper, the
terms ability and field independence have been used
interchangeably.‘

The instrument most often used to measure field
independence is the embedded figures test, a visual perception
test (Wilkin et al. 1971 ). The test consists of series of exercises
that require the participant locate a simple geometric figure
embedded in more complex diagram. In this study, competence
at performing the embedded figures test provides a measure of
ability, as defined by field independence theory.

2.3 Hypotheses Development

Prior to hypothesis testing, confirmatory factor analysis was
performed confirming that the accuracy and tracking cost aspects
of the stimuli form two separate scales. The hypotheses related
to accuracy are discussed first followed by the hypotheses for
tracking cost.

2.3.1 The determinants of sgccess at accuracy Due to prior
ABC training, all participants are expected to be very familiar with
the simplistic cost driver selection heuristic described in section
2.2.1 and shown in Figure 2.1. In particular, all participants are

expected to easily recognize press setup as a type of machine

 

1 The Bonner 8. Lewis GRE questions were also administered but were not as successful as
the field independence instrument in explaining performance.

28

setup, a batch-level activity. According to the heuristic, as long
as resource usage is a flat amount per setup, a system based on
number of setups would provide accurate information. Use of the
heuristic is justified based on the presumption that the
components of setup cost strongly correlate with number of
setups. The underlying principle is that there would be little or
no benefit to tracking more information if costs are always the
same amount per setup.

It is the purpose of the accuracy part of the experimental
task to test skill at recognizing situations in which such a
simplistic approach to cost driver selection would result in an
inaccurate system. Since the case materials show resource
usage is not a flat amount per setup, the participant who selects
number of setups as the cost driver will be recommending a highly
inaccurate system. The key to success is recognizing that use of
a batch-level driver, such a number of setups, would result in a
system with highly distorted costs.

It should be noted that the case study materials provide all
the information needed to recognize that number of setups is not
the correct cost driver. The resource consumption patterns
include detail information about setup labor usage, ink waste,

number of colors, number of setups and number of orders. The

29

consumption patterns demonstrate that the number of colors
printed, setup labor and ink usage are perfectly correlated. The
participant needs to eliminate the system design that only tracks
number of setups selecting any one of the three perfectly
correlated drivers. Since colors, labor and ink usage are
perfectly correlated, any one of these three drivers would provide
accurate costs.

As a minimum, the participant needs to have a firm grasp on
basic management accounting concepts including knowledge of
generic cost behavior patterns and cost terminology. To perform
well, the participant also needs to be sufficiently familiar with data
analysis techniques to recognize the accounting significance of a
change in production complexity on indirect costs. Higher levels
of general management accounting knowledge are expected to
make the participant more sensitive to the importance of a strong
correlation between resource usage and cost drivers. Individuals
with high levels of management accounting knowledge are less
likely to view cost driver selection as narrowly defined by the
simplistic heuristic.

Although all participants are trained in basic ABC, by the
second management accounting class, participants are expected

to have relatively high levels of general management accounting

3O

knowledge. In this study, participants who are completing their
second management accounting class are classified as high
management accounting knowledge. Those participants
completing their first management accounting class are classified
as low management accounting knowledge.

High management accounting participants are expected to
have a better understanding of the factors that cause indirect
costs to change than the low management accounting
participants. In addition, it is reasonable to assume that
participants in the second management accounting class have had
more practice with data analysis than those in the first class.2
Better knowledge of data analysis techniques is expected to aid
participants in interpreting the resource consumption patterns.
Specifically, knowledge of data analysis techniques, such as
regression, is expected to make the underlying principle of
correlation among cost drivers especially salient for those in the
second management accounting class.

The high management accounting knowledge group is also
expected to approach the task with a broader definition of the

problem than their low management accounting knowledge

 

2 The second undergraduate management accounting class included specific lessons on
techniques such as linear programming and regression. Of the 78 participants classified as
high general knowledge, 75 demonstrated basic competence in regression through
completion of a class project unrelated to this study.

31
counterparts. Unlike the low management accounting knowledge
group, the high management accounting knowledge group is
expected to consider comparison of the resource patterns the
focal point of problem solving. For those with superior
management accounting knowledge, the superficial characteristics
of the task, such as names of the activities and drivers, are not
expected to play a prominent role in performing this task. For
virtually all of the high management accounting knowledge group,
eliminating highly distorted drivers is expected to be a well-
defined and straightforward task. As a consequence, neither
ability nor industry training are expected to affect performance for
those with superior levels of management accounting knowledge.
Little or no variance is expected for the high management
accounting knowledge group.

Due to the lack of variance for the high management
accounting knowledge group, the hypotheses for accuracy
examined determinants of performance for the low management
accounting knowledge group. For those with low levels of
management accounting knowledge, industry training is expected
to have a positive effect on performance by providing knowledge
about the activities and products involved in package printing.

Specifically, industry training is expected to increase awareness

32
that number of colors printed is a major factor driving the
complexity of the production process and that resource usage is
not a flat amount per batch. Knowledge of the link between
production complexity and the number of colors is expected to
provide a critical reference point in reviewing the facts
surrounding the case materials. Unlike the control group, the
industry trained group is expected to notice changes in resource

consumption patterns. Therefore:

H1: Industry training positively affects selection of an
accurate system for the low management accounting
knowledge group.

Ability is also expected to have a positive effect on
performance for the low management accounting knowledge
group. Higher ability is expected to aid the low management
accounting knowledge group in recognizing the need to broaden
the definition of the problem beyond that of the simplistic
heuristic. In contrast, lower ability participants are expected to
ignore resource consumption patterns, focus on the superficial

aspects of the task, and to continue use the simplistic heuristic.

33

Therefore:

H2: Ability positively affects selection of an accurate
system for the low management accounting knowledge
group.

2.3.2 Tracking cost hypotheses As shown in figure 1.1,

accuracy is a critical sub-component of the cost driver selection
process. Failing at the accuracy sub-component is a fatal error.
Hypothesis testing for the second component of the process,
tracking cost, focuses on a reduced sample composed only of
those participants who succeeded at the accuracy task.

To identify the lowest cost driver, the participants must
make comparisons beyond those required for the accuracy task.
Industry training is expected to aid both high and low management
accounting knowledge participants in isolating the least costly
driver, colors. Unlike the control group, the treatment group
comes to the problem aware of the relationship between colors
and the complexity of the production process. Essentially, prior
knowledge of the significance of colors to the setup activity is

expected to simplify the task for the treatment group. Therefore,

34

H3: Industry training is expected to positively affect
skill at identifying the least costly driver for both high
and low management accounting knowledge
participants.

The content of the industry training session focuses on the
complexity of the manufacturing process and the diversity of the
items produced. The slides and scripts contain no explicit
accounting information. To succeed, the participant needs to
recognize that the number of colors in the design is characteristic
of the product that correlates with two components of cost, labor
and ink. As a consequence, superior management accounting
knowledge is also expected to affect performance. Participants
with high levels of management accounting knowledge are
expected to notice a pattern that shows labor and ink use are
proportional to the colors printed. By helping the participant
recognize the accounting significance of the redundancy between
colors and resource use, more extensive management accounting
knowledge is expected to reduce the complexity of the task.

Such a finding would be consistent with previous research
suggesting that correlations among input cues can reduce task

complexity if the decision maker is aware of the redundancy.

35
(Bonner, 1994; Hammond, 1986, Naylor and Schenck, 1968).
Similarly, the positive effect of general knowledge would be
consistent with the BL (1990) finding that general knowledge of
analytical procedures was related to performance of a financial

instruments task. Therefore,

H4: Management accounting knowledge is expected to
positively affect skill at identifying the least costly
driver.

2.4 Summary

This chapter contained the literature review related to the
study. The research surrounding cost aggregation was
summarized in Section 1. Issues regarding the determinants of
skilled performance were discussed in Section 2.

This chapter also presented the hypotheses testing skill at
cost driver selection. The determinants of skill at accuracy were
presented in section 2.3.1 and hypotheses related to the tracking
cost portion of the process were presented in section 2.3.2.

The four hypotheses were subsequently examined through

36

an experiment described in the next chapter. Specific results are

presented in Chapter IV.

CHAPTER III
METHODOLOGY

3.0 Overview

The purpose of Chapter III is to discuss the hypotheses test
procedures. A controlled laboratory experiment is used to
investigate cost driver selection.

The first section presents an overview of the experiment and
the research design employed. The second section discusses
the participants. The third section is a detailed discussion of the
experimental stimuli. The fourth section discusses the methods
used in hypothesis testing.

3.1 Experimental Design

This study employed student participants to test the degree
to which variation in cost driver selection can be explained by
individual differences in knowledge and ability.

The experiment was conducted in two phases that took
place approximately one week apart. In the first session,
participants were given several tests including two ability
measures.

After the first session, half of the participants were randomly
assigned to an industry training session that provided an overview

of a manufacturing process similar to the one presented in the

37

38
performance task.

The training session was based on an audio-visual
presentation currently used by the industry trade association.
The industry audio-visual training session lasted approximately
twenty minutes and was conducted in the second session prior to
the performance task. The intent of the training session was to
provide an overview of the major activities involved in package
printing. (See Appendix A). The content of the scripts and the
slides focused only on the manufacturing process and not on
accounting issues, such as costs and correlations among cost
drivers. Neither the scripts nor the slides contained any explicit
instruction on the accounting significance of characteristics of the ‘
production process. For example, the lesson includes the fact
that the number of print stations that need to be used depends on
the number of colors to be printed. The accounting implication
that colors would drive press setup costs was not explicitly stated.

In the second session, just prior to the performance task, all
subjects were given a training session to familiarize themselves
with the requirements of the task. The training session involved a
review of a sample cost driver selection problem using a non-
manufacturing setting. The correct answers were given and

reviewed. The sample case was provided to ensure that all

39

participants clearly understood the objective of the experimental
stimuli was to identify the most accurate, least costly set of
drivers. The training session took approximately fifteen minutes.

Next, the actual performance instrument, a case study, was
administered (See Appendix B). After completing the
performance instrument, participants were asked to complete exit
interview questions that included demographic and other
debriefing information. The case study and exit interview
questionnaire were self-paced. For most participants the second
session lasted less than 1 1/3 hour.

3.2 Participants

3.2.1 Sample Statistics Originally, a total of 180
undergraduate and graduate level students participated in this
experiment. All students had the same instructor and were
enrolled in a class that covered activity-based costing. To
ensure uniform coverage of the topic, the ABC instruction
included a handout that was covered in class.

All the participants were volunteers and expected to be
adequately motivated and give adequate attention to the
experimental tasks. To verify the assumption of adequate
motivation and attention to task, the experiment included five

straightforward cost computation questions. An example of this

40

type of question is discussed in section 3.3.1. A total of 11
participants were eliminated when they failed to correctly answer
at least four out of five of these questions.3

Although ABC was covered in class, some participants
may not have learned the basics due to poor attendance. To
verify that all participants had a working knowledge of the cost
driver heuristic described in section 2.2.1, five simple cost driver
questions were included in the experimental stimuli. These
questions, simplified versions of the performance instrument, are
described in section 3.3.2. A total of 15 participants were
eliminated when they failed to answer correctly at least four out of
the five of the simple ABC questions.

Finally, 11 participants were eliminated because they failed
to fully take part in the second part of the experiment making the
final sample size equal to 143, as shown below in Table 3.1.

Of the 143 participants, 10 were graduate students and 133
were undergraduate cost accounting students. The 143 students
remaining in the study were a fairly homogeneous group in terms
of motivation, basic ABC knowledge and familiarity with basic

management accounting concepts. Specifically, nearly all

 

3 Poor performance may have reflected factors other than low motivation, such as lack of
rudimentary accounting knowledge and weak computation skills. These particular individuals
tended to perform poorIy on all of the tasks.

41

participants remaining in the study received a perfect score on the
easy cost computation questions designed to test motivation and
the simple cost driver questions designed to test ABC knowledge.
In addition, in a self report, nearly all remaining participants
indicated that they were either familiar or very familiar with
general management accounting concepts and cost driver
selection.4 Of the 143 participants, 65 were completing their first
management accounting class, while 78 were completing their
second. The experiment was run during the last weeks of the

term in which the management accounting class was taken.

Table 3.1
Sample Statistics

Original participants 180
Lack of motivation (11)
Lack of ABC knowledge (15)
Lack of full participation L‘fl).
Total sample size 143

The participants were primarily composed of inexperienced
individuals. Only three individuals, all graduate students, had

accounting work experience of over one year. The three

 

4 A multiple choice test was also administered to measure basic competence in management
accounting concepts. For the participants remaining in the study, the test scores had no
explanatory value once the number of management accounting classes was considered.

42

experienced accountants were unfamiliar with printing prior to the
experiment.

3.2.2 Esk Incentivge Participants were paid $1.00 to
complete the experiment. In addition, monetary incentives were
used to motivate all participants. The monetary incentives were
based on performance on the case study with the incentive portion
of the payment ranging from zero to $17.00. In addition, a
$25.00 lottery was held for all participants at the end of school
term.

3.3 Task

Participants were presented with a case study for a package
printing plant that recently had purchased another plant. The
case materials clearly stated that both plants were identical in
terms of the manufacturing capacity. The only salient differences
between the plants related to product mix and the volume of
business handled at each location. Because of the acquisition,
the product mix produced at each plant was subject to change.
The problem facing the printer was to determine the adequacy of
the company’s single cost driver system given the impending
change in mix.

The participants were given 10 different problem sets and

asked to compute setup costs and select a cost driver for the

43

setup activity. The first five problem sets were relatively easy
with high correlation among all the cost drivers. Responses to
the first five problem sets served to filter out participants who
lacked motivation or basic ABC knowledge.

The second five problem sets were more difficult. For
hypotheses testing purposes, the responses to the difficult cost
driver problems provided the basis for computation of the
performance variables. In this study, skilled performance is
defined as selecting the most accurate, least costly system out of
five alternatives.

The instructions indicated that each problem set should be
answered independently of the others. The case materials were
developed such that every participant received the same set of
cost driver problems in the same order. The order went from a
set of five problems with high correlation among the drivers to a
set of five difficult ones with low correlation.5

Section 3.3.1 discusses the cost computation task. Section
3.3.2 covers the set of five straightforward problems used to test
basic ABC knowledge. Section 3.3.3 and section 3.3.4 present

detailed descriptions of the two performance variables based on

 

5 Progressing from high correlation to low correlation among the cost drivers was necessary
to portray how mix realistically changes in a package printing plant. According to industry
sources, product mix is likely to become more complex rather than less overtime.

44
responses to the second set of five cost driver problems. Section
3.3.3 discusses how skill at identifying an accurate driver is
measured. Section 3.3.4 discusses how skill at selecting the
least costly driver is measured.

3.3.1 Cost computation task The cost computation task
requires the computation of annual variable setup costs for the
newly acquired plant. Information was provided about the
resource usage for the typical job produced at both the original
and new acquired plants. The instructions explicitly state that
participant needs to follow these steps:

1. Compare the resource usage of the typical order for the
original plant to the typical order for the new facility.

2. Compute an accurate setup cost per order for the new plant.

3. Compute annual variable setup costs for the new plant using

the following formula:

Annual costs = setup cost per order * # of orders

To allow detail computation of setup costs, the narrative
accompanying the case materials states that setup costs are
composed of labor, ink waste and the cost of a plastic setup roll.
The narrative also includes all the information needed to compute

the labor rate per hour and the input prices of the two types of

45
indirect materials, ink and plastic. The setup labor rate in the
case materials computes to $60.00 per hour. The ink cost
computes to $3.00 per pound and the setup roll costs $25.00
each.

The resource consumption information illustrates that the
typical order produced in the original plant involves one setup,
uses 1.5 hours of labor, 27 pounds of ink and 1 setup roll. Based
on the input prices and the specifications of the typical order, the

cost per setup computes to $196.00, as shown below:

Variable cost per setup - 360'1.5 hrs + 83* 27 ink lbs. + $25 * 1 setup roll

$196.00 ' $90.00 + $81.00 +25.00

As previously discussed in section 3.2.1, the first five cost
computation problem sets were designed to test motivation and
attention to task. The only relevant change from one
straightforward question to another was a change in the number of
setups per order.

Problem set 2, reproduced in Table 3.2, is an example of

one of these questions.

46

Table 3.2
Straightforward cost problem

Typical specifications New Plant Original
# of orders annually 200 500

# of setups per order 3 1

# of setup rolls per setup 1 1

I! of colors in the design 3 3

# of setup labor hours per setup 1.5 1.5

# of ink lbs wasted per setup 27 27
Variable setup costs per order ? $196.00
Annual variable setup costs ? $98,000

Each participant was expected to use this information to
compute annual setup costs to be $117,600 ($196.00 cost per
setup * 3 setups *200 orders ). Of the 143 participants in the
final sample, all but 4 had perfect scores on the first five cost
computation questions. As noted previously, to be included in
the study, the participant had to answer at least four out of five of
these problems correct.

The last five cost computation problems were significantly
more difficult than the first five. To compute costs correctly, the
participant needed to recognize that the cost function had
changed due to a change in mix. Participants who continued to
compute costs based on $196.00 per setup were scored as having

failed at this task. This type of cost computation error is referred

47

to as the batch error in subsequent discussions. In contrast,
participants who computed costs using detailed information about
resource use were scored as having succeeded at the task.6

An example from the actual case study, Problem set 7, will
be presented to illustrate the correct solution and the magnitude
of the effect of the batch error on cost computation. The
information presented as part of Problem set 7 is shown in Table
3.3.

Upon review of the resource data, the successful participant
is expected to notice that the new plant uses more labor and ink
per setup than the original plant. In this type of situation, the
participant needs to apply the input prices to the quantities of
labor, ink and rolls used. The simplistic rule using $196.00 per
setup needs to be abandoned.

The correct solution and the batch error are illustrated in
Panels A and B respectively of Table 3.4. As shown in Panel A
of Table 3.4, the correct setup cost computes to be $253 per
setup, $1,012 per order and $303,600 annually. In contrast,
using $196.00 per setup results in computing costs to be

$235,200 annually.

 

6 The correlation between the batch error in cost computation discussed here and selecting
an inaccurate system, discussed in section 3.3.3, was close to one.

48

Table 3.3
Difficult Problem

Typical specifications New Plant Original
# of orders annually 300 500

# of setups per order 4 1

# of setup rolls per setup 1 1

# of colors in the design 4 3

# of setup labor hours per setup 2 1.5

# of ink lbs wasted per setup 36 27
Variable setup costs per order ? $196.00
Annual variable setup costs 7 $98,000

Each of the other four problem sets used in measuring the
cost computation variable followed the same pattern. Failure to
abandon the $196.00 per setup cost function results in
significantly distorted costs. For each problem set, the
magnitude of the batch error as a percent of cost is quite large.

As presented in Table 3.5, the absolute value of the batch
error ranges from a low of 23% of cost for problem set 7 to a high
of 139% for problem set 8. In a business dependent upon long
term contracts and competitive bidding, such large errors would

have serious business implications.

49

Table 3.4
Cost Computation: Difficult Problem

Panel A Correct Solution

2 Setup labor hours“ $60.00 = $120.00
36 lbs of ink * $3.00 per pound ‘= $108.00
1 setup roll * $25.00 per roll = $25.00
variable cost per setup = $253.00
* # of setups per order = 4
Variable setup cost per order= $1012.00
* # of orders annually 300

Mi #9181? {film}: .§§.§ii.9..§9.31 ............. ..535???527523?3’5".51"15?5.5575?'Ii'335?TSi}5ii.135?.5f22§23§5§§€§§§5§§§$§9 .359 9.11:1].1 ....

where variable cost per setup is:

Variable cost per setup - $60‘2.0 hours + 53*36 ink pounds + $25 * 1 setup roll
$253.00 - $120.00 + $108.00 +$25.00

Panel B Batch Error

variable cost per setup = $196.00

* # of setups per order = 4

Variable setup cost per order = $784.00

(1* # of orders annually 4 _ _ ._ 300 . ,,.
iiiiiiilgfianuai variable some cast 3235.200

 

50

Table 3.6
Magnitude of the batch error

 

 

 

 

 

 

 

 

 

 

 

Cost Computation Magnitude of

the batch error

Correct Batch error Dollar amt. % error

Problem set 6 $75,060 $106,840 $30,780 41%
Problem set 7 $303,600 $235,200 $68,400 23%

Problem set 8 $65,600 $156,800 $91,200 139%
Problem set 9 $173,600 $109,760 $63,840 37%
Problem set 10 $44,040 $23,520 $20,520 47%

 

 

 

3.3.2. gaaic ABC knowledge To test for basic ABC
knowledge, the first five driver selection questions were similar to
problems covered in the management accounting classes. All
participants were expected to easily recognize press setup as a
type of machine setup, a batch-level activity. Due to basic ABC
training, all participants were expected to know how to apply the
simple heuristic discussed in Chapter II and shown in Figure 2.1.

Because the simple heuristic does not specifically deal with
tracking costs, there was a concern that some participants might
select number of setups as a cost driver when an equally
accurate, less costly driver was available.

To determine whether participants noticed the tracking

51

costs, one of the five straightforward problem sets assumed that
all cost drivers were perfectly correlated. This problem set was
added to determine if participants would always select a system
that uses a batch-level driver even when a less costly, equally
accurate single cost driver system would suffice. Virtually all of
the participants answered this question correctly by
recommending that the company continue with its single cost
driver system.

The next four problems required a batch-level cost driver,
number of press setups. Problem set 2, shown in Table 3.2 in
section 3.3.1, is an example of one of these problems. The
tracking cost information illustrates that number of press setups is
less costly to track than setup labor, ink use or number of colors.
As a consequence, the participant was expected to recognize that
a system that tracks data by number of setups would provide
accurate information with the lowest tracking cost. Nearly all of
the 143 participants received a perfect score on these questions.

3.3.3 Dependent variable one: Accuracy Whereas the first
five cost driver problems were simple and meant to test basic ABC
knowledge, the remaining five problems were more difficult and
served as the basis for computation of the performance variables.

In the discussions that follow, the performance variables

52
were defined solely in terms of the difficult cost driver questions
rather than incorporating the cost computation responses into the
results. Since the correlation between making the batch error
and selecting an inaccurate system was close to one, the results
were essentially the same regardless of the definition of accuracy
used.7

Therefore, for hypotheses testing purposes, the accuracy
variable is defined solely in terms of selecting an accurate system
design out of five possible alternatives. The alternatives
included two highly inaccurate options and three designs that
would permit accurate costs.

The five alternatives included one volume-level driver,
number of orders, and one batch-level driver, number of setups.
The three accurate drivers were setup labor hours, ink pounds
wasted and number of colors in the design. Resource
consumption patterns demonstrated that three cost drivers, labor,
ink, and colors were perfectly correlated.

It is significant to note that the participant does not have to
compute costs in order to selecting an accurate system design.

As shown in figure 3.1, the participant only needs to focus on the

 

7 In computing costs, many participants omitted one element of cost. This omission error
was unrelated to skill at selecting the least costly, most accurate system design. Although
not presented here, statistical analysis of the omission error provides some evidence that use
of information requires skills other than those required for system design.

53

resource consumption patterns. By comparing the relative
resource usage between the old and new plant, the participant is
expected to realize that resource usage is not a flat amount per
setup.

The same resource patterns shown in Table 3.3 in section
3.3.1 will be used to illustrate skill at selecting an accurate
system.

The skilled performer is expected to notice that setup labor
and ink costs are not a uniform amount per setup. As shown in
the shaded area in Table 3.6, the new plant uses 11I3 as much
labor and ink per setup as the old plant.

Each of the other four problem sets followed the same
pattern as shown above. In each of these situations, continued
use of a simplistic heuristic would result in recommending a
system with highly distorted costs. In scoring the accuracy
variable, the participant’s solution to each of the problems was
examined. If the participant recommended any one of the three
accurate drivers, the answer was considered correct. If number
of setups was selected, the answer was scored as wrong. As
expected, none of the 143 participants selected a single cost

driver system as the solution to these problems.

Figure 3.1
Skilled performer: Accuracy

Compare the resource usage of the typical order for the
original plant to the typical order for the new facility.

Notice that setup labor and ink usage are not a flat amount
per batch.

Abandon the simplistic heuristic.

Select any one of the accurate system alternatives,

 

   

colors, labor or ink.

 

55

Table 3.6
Ratio of Resource Usage

Typical specifications New Original Ratio of
Plant resource
“539° , .
# of setup rolls per setup 1 1 1t01
# of colors in the design 4 3 jii’ii3i‘lf'l53gitioflg}3:}
# of setup labor hrs per setup 2 1.5 {1121;1337151115,g}
# of ink lbs wasted per setup 36 27 1113to‘i ,:

3.3.4 Dependent variable two: Lrackinflost As indicated
in section 2.3.2, only the responses of the participants who
selected one of the three accurate cost drivers are included in the
analysis of the second dependent variable, tracking cost.

In addition to the resource consumption information, the
participant was given tracking cost information for each of the five
cost driver alternatives. The relative tracking costs are
presented in Table 3.7. The tracking costs of the five systems
remained the same for all the problem sets. This was necessary
to have a realistic situation portrayed in the case. Discussions
with a package printing controller and an industry expert
confirmed that the ranking of the tracking costs used in the study

was realistic for their industry.

56

Table 3.7
Costs of Tracking

# Alternative System Design Annual Tracking
Cost

1 # of orders (the current system) $0

2 # of setups per order $1,000

3 # of colors in the design 8. # of setups $2,500

4 # of setup hours and # of setups $3,000

5 # of ink lbs 8. # of setups $10,000

As depicted in figure 3.2, the skilled performer notices the
link between colors and resource use and that the color driver is
least costly. The unskilled performer, shown in figure 3.3,
focuses only on the labor and ink correlation ignoring the color
driver entirely. The variable, tracking cost, was scored correct if
the color driver was chosen, wrong otherwise.

3.3.5 Pretests The case materials used in this study were
developed with the assistance of a controller of a midwest
package printing company and an industry expert. The controller
was especially helpful in reviewing the reasonableness of the
resource consumption patterns shown in the problem sets and the
input prices used. Earlier versions of the case study were tested

with both undergraduate and graduate level students.

57

Figure 3.2
Skilled performer: Tracking cost

Notice that colors, setup labor and ink waste are
correlated.

Notice that using colors as the driver is less costly than
using either setup labor hours or ink waste.

Select the system that uses colors as the cost driver.

 

   

58

Figure 3.3
Unskilled performer:Tracking cost

Notice that setup labor and ink waste are correlated.
Notice that setup labor is less costly than ink waste.

Select the system that uses setup labor as the cost driver.

 
 

 
 

 

59

In addition to 180 participants discussed in section 3.2.1,
nine individuals were identified prior to the experiment as having
printing industry experience. Of these nine experienced
individuals, six were very familiar with multi-color printing and
three were only familiar with single color printing. One individual
was a consultant to the printing industry. All nine selected an
accurate cost driver. All but one experienced participant
selected number of colors as the least costly, most accurate
driver. The one individual who failed to select colors as the cost
driver was one of the three unfamiliar with multi-color printing.

Other than not being randomized to either the control or
treatment condition, the nine experienced participants were
treated the same as the other volunteers. Their responses,
however, were not included in the statistical analyses for
hypotheses testing. Instead, these participants attended the
control condition and helped to further validate the case study
materials. Prior to taking part in the experiment, none of the
experienced individuals were aware that the project related to the
printing industry.

3.3.6 Ability variable Ability is measured using the
embedded figures test, an instrument which has been validated in

previous research and found to have a reliability in excess of .80.

60

The theory of field independence, discussed in Chapter II,
provides the theoretical justification for use of the embedded
figures test to measure ability.

The embedded figures instrument has three timed sections,
two of which are scored. The first section is unscored and has a
two minute time limit. The participant’s solutions to the first
section are reviewed to ensure familiarity with the requirements of
the test before proceeding to the scored sections. The second
and third sections of the test include nine questions each and
have time limits of five minutes each. Each question requires the
participant isolate a simple geometric figure, such as a cube, in a
more complex figure. The questions in sections two and three are
significantly more difficult than those in the unscored first section.
In this study, FD1 refers to the score on section two of the
embedded figures test. FDZ refers to the score on section three.
The responses to FD1 and F02 were tested for reliability prior to
hypothesis testing and found to be adequate. The sum of both
sections, FD1 and F02, is referred to as FDTOT in the test of
hypotheses in the next chapter.

3.3.7 Manipulation checks As a manipulation check on the
independent variable, industry training, participants were asked to

rate their familiarity with multi-color printing on the post

61

experiment questionnaire. In addition, after the experiment
participants were asked to list the major factors that drive
changes in the complexity of the press setup activity. The
individuals with industry training were expected to be more
familiar with printing and mention colors more frequently than
those in the control condition.

3.4 Method of Analysis

3.4.1 Confirmatorv factor analysis. Prior to testing the
research hypotheses, confirmatory factor analysis was conducted
to ensure that the five accuracy responses would form one
unidimensional performance measure while the five tracking cost
responses would form another.

3.4.2 ANOVA Analysis of variance and graphing were
employed to test for interactions. As shown in section 4.4.1, the
only significant interaction that occurred related to the two
accuracy hypotheses with the management accounting knowledge
variable.

3.4.3 Regression. Logistic regression and ordinary least
squares(OLS) regression were used to test the hypotheses. In
OLS the objective is to find the coefficients that result in the
smallest sums of squared distances between the observed and

predicted values of the dependent variable. Hence, the method

62

is called least-squares. Ordinary leased squares methods are
quite useful but have certain assumptions that need to be
considered prior to hypothesis testing. As a consequence,
before using OLS, the distributional properties of the accuracy
and cost driver selection variables were examined.

In this study, the experimental stimuli had the deliberate
effect of classifying participants into one of two mutually exclusive
categories, those who performed the task well and those that did
not. Essentially, the performance variables are distributed in a
pattern similar to a dichotomous variable. This was expected.8

Since the responses tended to fall into one of two
categories as described above, one of the key assumptions of
OLS, normally distributed errors, was violated. To deal with this
problem, logistic regression was run using SPSS. Logistic
regression does not require as many distributional assumptions as
OLS.

Measurement error in the variables was also a concern. To
deal with this type of error, linear regression was run twice. The
first regression was done assuming perfect measures. A second
regression was computed using the same variables but with a

corrected correlation matrix. Both OLS regressions and the

 

8 In pilot studies it was found that substantially increasing the number of problem sets did not
change the distributional properties of the variables.

63

logistic regressions are discussed in the next chapter.
3.5 Summary

This chapter contained an overview of the experimental
setting, participants, decision task and methods of analysis used

in the study. The experimental results follow in the next chapter.

CHAPTER IV
DATA ANALYSIS
4.0 Overview

This chapter contains the experimental results of the study.
As discussed previously, the experiment was conducted to
examine the effects of industry knowledge on cost driver
selection. To accomplish this goal, a performance instrument was
developed with objective performance criterion.

Confirmatory factor analysis of the performance variables,
accuracy and tracking cost, was conducted prior to hypotheses
testing. A discussion of the confirmatory factor analysis is
presented in section 1. Section 2 reviews the results of
manipulation checks on the independent variable, industry
knowledge. Section 3 presents descriptive statistics and an
examination of the distributional properties of key variables.
Section 4 describes the results of the statistical analysis for the
accuracy variable. Section 5 discusses the results of the
statistical analysis for tracking cost. The final section of this
chapter is a summary of overall results.

In the discussions that follow the term, treatment group, is
used to describe those subjects who took part in the industry
training session. The term, control group, is used to describe all

others. On the various charts and tables, Industry = 0 will

65

designate the control group, while Industry = 1 will designate the
treatment group. Similarly, low management accounting
knowledge will be depicted as Genknow = 0, and high
management accounting knowledge will be referred to as
Genknow = 1. For example, a label that combines Genknow = O
and Industry = 1 will refer to low management accounting
knowledge participants who took part in the industry training
session.
4.1 Confirmatory Factor Analysis

The confirmatory factor analysis evaluated the accuracy
items, tracking cost items and the ability scores in terms of both
internal and external consistency. The concept of internal
consistency indicates that all the items in a given scale measure
the same underlying construct. External consistency (also called
parallelism), on the other hand, requires that all the items within a
scale relate to items in other scales in a similar fashion. Item by
item correlations are typically examined to evaluate both internal
and external consistency.

To test for the internal consistency of the accuracy
responses, the correlations among the five accuracy questions
were examined. Although the correlations among the five items

were very high, three of the five questions had close to a perfect

66

correlation. As a result, the two questions with slightly lower
correlations were eliminated. A high correlation among the
accuracy responses was expected since participants tended to
use the same strategy or heuristic in answering each of the
questions. As shown in section 4.3.2, the responses for
remaining questions formed a bimodal distribution. Participants
tended to consistently succeed or consistently fail.

Once the internal consistency of the accuracy scale had
been established, the issue of external consistency needed to be
evaluated. As stated in section 3.4.1, a key expectation of this
study was that the accuracy responses would form one internally
consistent scale while the tracking cost responses would form
another. Determining that the responses formed two separate
scales was a significant goal of this research. As stated in
chapter I, the accuracy part of the task and the tracking cost
questions were expected to correspond to different costs
associated with sub-optimal systems design decisions. The
accuracy aspect of the experimental task was intended to tap into
skill at differentiating between accurate and highly distorted cost
systems. In contrast, the tracking cost part of the task was
intended to measure another aspect of performance, skill at

differentiating between systems of equal accuracy but varying

67

operating costs.

To evaluate external consistency, the correlations between
each accuracy item, each tracking cost item and the two ability
scores were examined. If externally consistent, one would expect
correlations of a similar magnitude between each of the accuracy
items and each of the tracking cost and ability items.

Examination of item by item correlations indicated that external
consistency was strong. The correlations between each accuracy
item and the items in the other two scales were of a similar
magnitude. In addition, the correlation between each accuracy
response and each outside item was much lower than the
correlations among the accuracy items themselves. Since it
appears to be both internally and externally consistent, the three
item accuracy scale was used in evaluating the hypotheses.

The same approach was taken to evaluate the tracking cost
responses. This evaluation resulted in the elimination of one
cost driver item. Similar to the accuracy responses, the
correlation among the four remaining items was close to 1.
Therefore, the four item tracking cost scale was used to test the

hypotheses.9

 

9 For the accuracy and tracking cost responses, the high correlation among the items
precluded the use of standard confirmatory factor analysis programs and made their use
unnecessary. Nevertheless, in the interest of completeness, the hypotheses were tested with
and without the eliminated items. The results were unchanged.

68

Several confirmatory factor analysis programs were used to
compute the reliability of the two ability items, F01 and F02.
According to the programs, the two item ability measure has a
standard score coefficient alpha of .88. As a result, the scores

on F01 and F02 were summed and used in the remaining

evaluation of the test hypotheses. The summed ability measure

will be referred to as FDTOT.
4.2 Manipulation checks

This next section includes an examination of manipulation

checks on the training session.
4.2.1 Manipulation check: Familiarity with printing To
check on the effectiveness of the manipulation of industry

knowledge, the exit interview questionnaire included the following

question about perceived familiarity with multi-color printing.

 

At this point, how familiar do you feel with commercial multicolor

pflnfing?

(a) very unfamiliar
(b) unfamiliar

(c) familiar

(d) very familiar

 

 

69

The results of this question indicate that the control and the
treatment groups were quite different in terms of perceived
familiarity with printing. Table 4.1 indicates that only 17
participants out of 70 in the control group felt familiar with
multicolor printing. This contrasts with the treatment group who
seemed to feel more familiar with multicolor printing. As noted in
Table 4.1, 58 out of 73 participants in the industry trained group

felt familiar or very familiar with multicolor printing.

Table 4.1
Familiarity with printing

 

 

 

 

 

Frequency
Response Industry = O lndustg = 1
Unfamiliar or very unfamiliar 53 15
Familiar and very familiar 17 58
Total 70 73

 

 

 

 

The means were computed to be 2.0143 for the control
group and 2.8767 for the treatment group. As shown on Table

4.2, the t test indicates that the difference was significant.

70

Table 4.2
t test: Familiarity

t-tests for Independent Samples of INDUSTR!

Nudber

of Cases Mean SD 83 of Mean

Variable
“““““““\\\\“V\“\\\\\\\\\\\\\\\\\\fl“\\\\\\\\\\\\\\\\“““\\\\“\\\\\\\\“““\\\\“\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\“\\“\\““
FAMILIAR

INDUSTR! 0 70 2.0143 .732 .088
INDUSTR! 1 73 2.6767 .622 .073

\\““\\‘\\\\\\\“\\\\\\\\“\\“\\““““\\\\““““\\\\\\““\\\\\\\\\\\\\\\\\\“\\\\\\\\\\\\“\\\\\\“\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\““\\

Mean Difference = -.8624

Levene's Test for Equality of Variances: F= 1.251 P= .265

t—test for Equality of Means 958
variances t-value df 2-Tail Big 88 of Diff CI for Diff
\nnunnuunuunnusuu“nu“\uununnnuuun\“u\\\\\\\\\\\\\n\\\\\\u\\\“\\\u\\u\\n\\u\\u\u“u\nuunuuuuuuununuut
Equal -7.60 141 .000 .113 (-1.087, -.638)

135.44 .000 .114 (-1.088, -.637)

unequal -7.57
W\\\\\\\\\\\\“\\\\\\\\\\\\\\\\“\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\“\\\\“\\“\\\\“\\\\\\\\\\

71

4.2.2. Manipulation check: knowlegge of co|g_r_s In addition

to familiarity with printing, the training session was also expected

to increase awareness of the causal link between colors and

production complexity. In particular, it was expected that the

industry training session would provide knowledge about the

prominence of colors as a characteristic of product that

significantly affects press setup.

To help to identify the kind of knowledge that was conveyed
by the training session, after the experiment, the participants
were asked to list the factors that drive the press setup activity.
Industry trained participants were expected to mention colors in

the design more frequently than the control group.

Review of the responses indicated that the control and

treatment groups differed in their assessment of the link between

colors and production complexity. As shown in Table 4.3, only 5

out of 70 participants without industry training mentioned colors in

contrast to the majority of the trained participants.

The results of the t test and the Mann-Whitney U test,
shown in Panel A and B respectively of Table 4.4, indicate that

the treatment and the control group differed in how frequently

they mentioned colors.

72

 

 

 

 

 

 

Table 4.3
Colors
Frequency
Response Industry = 0 Industry = 1
Do not mention colors 65 10
Mention colors 5 63
Total 70 73

 

 

 

 

The study had predicted that there would be no difference
between high and low management accounting knowledge groups
related to acquisition of industry knowledge. The industry
training session was expected to allow both high and low
management accounting knowledge participants to easily acquire
the knowledge that colors affects press setup. Any difference
between high and low management accounting knowledge was
expected to relate to the application rather than acquisition of
industry knowledge. The t test and Mann-Whitney U test, shown
in Table 4.5, suggests that there was no difference between
management accounting knowledge groups in terms of knowledge

of colors.

73

Table 4.4
t test: Colors

PANEL A: t: tests for Independent Samples of INDUSTRY

Number
Variable of Cases Mean SD SE of Mean
\1“!\“‘\\\“\H\\H\\\\\\\\\\\\\\\\\\\\\\\\\“\\“WW“\\\\\\\\“\\“\\\\“\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\|\\\\\\\\\\\\\\“\\\\\\
COLORS
INDUSTRY 0 70 .0714 .259 .031
INDUSTRY 1 73 .8630 .346 .041
\nnnuuuuuunuuu‘vnu‘u\uunun‘\H\\\\\\\\uunusnuuunu\uuuuuu\\u\\u\\unuuuuuunnn\\ununnuu‘
Mean Difference - -.7916
Levene's Test for Equality of Variances: F= 6.818 P= .010
t-test for Equality of Means 95%
Variances t-value df 2-Tail Sig SE of Diff CI for Diff
“fl““\\‘\\\“\\\\\\\\“\\\\\\\\\\\\\\\|\\\\“\\\\\\\\“\\\\\\\\\\“|\\\\\“\\\\\\“\\“““\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\‘\\\\\\\
Equal -15.42 141 .000 .051 (-.893, -.690)
Unequal -15.51 133.30 .000 .051 (-.893, -.691)

“““‘\\\“\\‘\\\\\\\“\\\\“\\“\\““\\\\\\“\\\\“\‘\“\\\\\\\\‘\\\“\\\\“\\\\“\\\\\\\|\\\\\\\\\\“\\“\\“\\“\\\\“\\“\\\\\\\\““\\\\““\\“\\“““

PANEL 8: - - Mann-Whitney U - Wilcoxon Rank Sum W Test

canons
by INDUSTRY

Mean Rank Sum of Ranks Cases
43.11 3017.5 70 INDUSTRY -- 0
99. 71 7278.5 73 INDUSTRY = 1
143 Total
U H z 2-Tailed P

532.5 3017.5 -9.4421 .0000

 

' u "var-11¢“

 

74

Table 4.5
t test: Colors by Genknow

PANEL A: t-tests for Independent Samples of GENKNOW

Nunber
of Cases Mean SD SE of Mean

Variable
1H1“I!“ItHHHHIIHHHHIHM“\\ \\ \\\\ \\ \\“\\“\\‘\“\\\\“|\\\“\\ \\ \\\\ \\ \\ \\\\ \\ \\ \\\\ “ \\\\\\\\\\\\“ \\ \\\\\\\\\\\\“\\\\\| \\ \\

COLORS
amour O 65 .4923 .504 .062
78 .4615 .502 .057

I"““NN|\““I\“‘\““‘\“““““\\\\“\\\\“\\\\“\\\\“\\“\\ “nu“ \\ \\\\\\\\\\\\\\“\\\\“ \\\\ \\“ \\\\\\“\\\\\\“\\\\\\ “““““

Mean Difference - .0308

Levene's Test for Equality of Variances: F= .340 P= .561

95%

t-test for Equality of Means
CI for Diff

a: 2-Tail Sig as of our

Variances t-value

‘1“IIfl“|\“‘1fl“fl““\\“\\\\“\\“\\“\\“\\\\\\\\\\\\\\\\““\\\\“\\\\\\\\\\“\\“\\\\\\\\\\\\““I\““\\\\“\\“\\\\\\\\\\\\“\\“\\\\‘\““\\“\\\\

Equal .36 141 .716 .084 (-.136, .198)
136.19 .716 .084 (-.136, .198)

unequal .36
unuunnunuunnuuu“uu“n““\n‘\\\\“u\unuuuuuuuuu‘“uuu“nu“u“\\u“\\\\\\\\\\\\\\\\u“nunnnuunuunnnuuu

PANEL 3: Mann-Whitney U - Wilcoxon Rank Sum W Test

corona
by amen

Mean Rank Sun of Ranks Cases
73.20 4758.0 65 GENKNOW = 0
71 . 00 5538 . 0 78 GENKNOW = 1
143 Total
U W 2 2-Tailed P
-— . 3656 .7147

2457.0 5538.0

75

Based on the manipulation checks, the treatment group
seemed to be more familiar with the production process than the
control group. In addition, industry training seemed to have
made colors more salient for the treatment group. In this case,
industry training, not management accounting knowledge, appears
to make the difference in acquiring industry knowledge.

4.3 An analysis of the properties of the key variables.

Prior to hypothesis testing, frequency information and
descriptive statistics of the key variables are presented. Since
the two hypotheses related to accuracy require splitting the
sample into low and high management accounting knowledge
groups, some of the data are presented for the sample split into
applicable sub-groups.

The first section, section 4.3.1, presents descriptive
statistics for the one measured independent variable, ability.
After the descriptive statistics on ability are shown, a series oft
tests examine the equivalence of various groups in terms of
ability.

Section 4.3.2, covers the dependent variable, accuracy.

The final section, section 4.3.3, covers tracking cost.

76
4.3.1 Ability A histogram with descriptive statistics for the

ability variable is presented in Figure 4.1. The ability scores
ranged from a low of 1 to a high of 18. The mean is 11.112, the

median is 12 and the mode is 12.

Figure 4.1
Ability Histogram

 

 

A series of t tests were also run to test for differences
between the control group and treatment groups related to ability.
As depicted in table 4.6, the mean score for ability for the control
group was 11.1429, while the mean score was 11.0822 for the
treatment group. The t test results indicate that the means for the

control and treatment groups were not significantly different.

77

TABLE 4.6
t test of Ability: Full sample
N=143

t-tests for Independent Samples of INDUSTRI

number
Variable of Cases Mean SD SE of Mean

““““\\\\\\\\\\“\\“““\\\\\\\\\\\\\\\\\\\\\\\\\\‘\\\“\\\\\\\\“\\“\\\\\\“\\\\\\\\\\\\\\\\\\\\\\\\“\\“\\\\\\\\\\“““\\\\\\“\\““

FDTOT

INDUSTR! 0 70 11.1429 4.305 .515
INDUSTRX 1 73 11.0822 4.657 .545

““\\\\“\\\\““\\“““\\“\\\\I\“““““\\““\\““““‘\“\\“““\\\\“\\\\“\\“‘\““\\“\\““““““““\\\\\\\\\\\\\\\\““““

Mean Difference a .0607

Levene's Test for Equality of Variances: F= .486 P= .487

t-test for Equality of Means 95%
Variances t-value df 2-Tail 819 SE of Diff CI for Diff
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\““\\“\\\\\\“\\\\\I\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\i\\\\\\\\\\\\\““\\\\\\\\\\\\\\\\\\“\\“\\‘\\\\\\\“I\“\\“
Equal .08 141 .936 .751 (-1.424, 1.545)
unequal .08 140.81 .936 .750 (-1.421, 1.542)

\\\\\\\\\\\\\\“\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\“\\\\\\\\\\\\“\\\\““““\\\\\\\\\\\\\\“\\“\\\\“\\\\\\“\\\\“\\“\\‘\“\\“\\“\\‘\\\\\

78

Since hypothesis testing of the tracking cost variable was
conducted with a reduced sample, a second t test was run. The
second t test, shown in Table 4.7, indicates that the treatment and
control groups were not different in terms of ability.

4.3.2 Accuracy A histogram with descriptive statistics for
the accuracy variable is presented in Figure 4.2. The mean is
2.4 and the standard deviation is 1.17. As expected, the

distribution appears to be bimodal.

Figure 4.2
Accuracy Histogram

 

 

 

 

1.0 2.0 3.0

79

TABLE 4.7
t test of Ability: Reduced sample
N=1 1 6

t-tests for Independent Samples of INDUSTR!

Number
Variable of Cases Mean SD SE of Mean
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\“\\\\\\\\\\\\1‘\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\““\\“““\\\\\\“\\
FDTOT
INDUSTR! 0 48 12.3125 3.855 .556
INDUSTR! 1 68 11.4118 4.565 .554

““I““““““““\\\\“\\\\\\‘\\\\\\\\\\\\\““\\““\\““““\\\\“\\\\\\“\\“““\\““\|\\\\\\\\‘\\\““\\“\\\\\\\\\\\\\\“\\\\\\\\

Mean Difference - .9007

Levene's Test for Equality of Variances: F= 2.489 P= .117

t-test for Equality of Means 958
Variances t-value df 2-Tail Sig SE of Diff CI for Diff
“\\V\\\\\““91““\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\‘\\\\\\\\\\\\\\\\\\“\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\I\\\\\“““\\“\\\\\\\\\\“\\\\\\\\“\\“\\\\\\\\\\““
Equal 1.11 114 .267 .808 (-.700, 2.502)
Unequal 1.15 110.30 .254 .785 (-.655, 2.456)

“\\“““\\“\\\\“\\\\“““\\““\\“\\““\\““\\\\““fl“\\\\\\\I“‘\\\\\\\\\“\\\I\\\\\\\\\\\\\\\\“\\\\\\\\\\\\\\\\“\\\\\\\\““\\““\\‘\\\\\“““

80

Frequency information for the accuracy variable is
presented in Table 4.8. Panel A of Table 4.8 splits the whole
sample of 143 into treatment and control groups. Panel B splits
the sample into management accounting knowledge groups and
Panel C shows the low management accounting knowledge group
split by treatment.

Panel A indicates that most of the participants were able to
select an accurate cost system with 116 out 143 succeeding at the
task. Of the 27 participants who failed at the task, 22 were from
the control group and 5 were from the treatment group.

Panel B of Table 4.8 provides additional insight by splitting
the information into management accounting knowledge
categories. It is clear from this data that industry training was
unnecessary for the high management accounting knowledge
group. According to Panel B, 26 of the 27 low scorers were low
management accounting knowledge participants (Genknow = 0).

Panel C presents the data for the low management
accounting knowledge group(Genknow = 0) split by industry. The
low management accounting knowledge control group (Genknow =
0; Industry = 0) had more difficulty with the accuracy task than the

corresponding treatment group (Genknow = 0; Industry = 1).

81

Table 4.8

Frequencies: Accuracy

PANEL A: BY INDUSTRY

 

 

 

 

 

WHOLE INDUSTRY INDUSTRY
SAMPLE I0 I 1
1 or less 27 22 5
2 or more 116 48 68
TOTAL 143 70 73

 

 

 

 

 

PANEL B: BY MANAGEMENT ACCOUNTING KNOWLEDGE GROUP

 

 

 

 

 

WHOLE GENKNOWI GENKNOW
SAMPLE 0 = 1
1 or less 27 26 1
2 or more 116 39 77
TOTAL 143 65 78

 

 

 

 

 

PANEL C: LOW MANAGEMENT ACCOUNTING KNOWLEDGE BY
INDUSTRY

 

 

 

 

 

Industry - 0 Industry - 1
Freq Percent Freq Percent
1 or less 21 66% 5 15%
2 or more 11 34% 28 85%
TOTAL 32 100% 33 100%

 

 

 

 

 

 

82
4.3.3 Tracking cost In the analyses that follow, the 27

participants who failed to select an accurate system were
excluded. Therefore, the analyses for the tracking cost portion Of
the task will be based on the reduced sample size Of116.

A histogram with descriptive statistics for the tracking cost
responses is presented in Figure 4.3. The mean is 2.2, and the
standard deviation is 1.94. As expected, the distribution

appears to be bimodal.

Figure 4.3
Tracking cost Histogram

   

 

83

Using a format similar to that Of the accuracy variable,
frequency information is presented on Table 4.9. Panel A
presents the responses split by control and treatment group. The
frequency data clearly shows the positive effect Of industry
knowledge. Of the 62 participants who succeeded at the task, 52
were from the treatment group, while only 10 were from the
control group.

Panel B shows the frequencies split by management
accounting category. While the results were not as dramatic as
the industry training variable, management accounting knowledge
also seems to positively affect performance. In particular, the
results indicate that 45 out Of the 62 high scoring participants
were from the high management accounting knowledge group in
contrast tO 17 from the low management accounting knowledge
group.

Panel C, the low management accounting knowledge table,
shows that one member of the control group succeeded in
comparison to a majority Of the treatment group.

Panel 0, the high management accounting knowledge group,
shows that 9 out Of 37 Of the control group selected the least

costly driver in contrast tO 36 out Of 40 Of the treatment group.

84

Table 4.9

Frequencies: Tracking cost

PANEL A: BY INDUSTRY

 

 

 

 

 

 

 

REDUCED INDUSTRY INDUSTRY
SAMPLE =0 8 1
2 or less 54 38 16
3 or more 62 10 52
TOTAL 116 48 68

 

 

PANEL B: BY MANAGEMENT ACCOUNTING KNOWLEDGE GROUP

 

 

 

 

 

 

 

REDUCED GENKNOWI GENKNOW
SAMPLE 0 = 1
2 or less 54 22 32
3 or more 62 17 45
TOTAL 116 39 77

 

 

PANEL C: LOW MANAGEMENT ACCOUNTING KNOWLEDGE‘BY

 

 

 

 

INDUSTRY
I Industry .0 Industry .1 I
I2 or less 10 12 I
I3 or more 1 16 I
[TOTAL 11 28 |

 

 

 

PANEL D: HIGH MANAGEMENT ACCOUNTING KNOWLEDGE BY INDUSTRY

 

 

 

 

I Industry =0 Industry -1 I
I 2 or less 28 4 I
I 3 or more 9 36 I
| TOTAL 37 40 |

 

 

 

 

 

85

4.4 H1 and H2: Determinants of skill at accuracy.

H1 predicted that, for the low management accounting group,
industry training would have a positive effect on skill at selecting
an accurate cost system. H; predicted the positive effect of
ability for the same group of participants.

Before proceeding to the test of the hypotheses, two steps
were taken to assess the possibility of an interaction. If an
interaction occurs, the sample would need to be split. The first
step involved plotting the mean values of each dependent
variable by independent variable categories. ANOVA was then
run to statistically test for the significance of any interaction.

Both steps, graphing and ANOVA, require separating
participants into mutually exclusive categories. The categories
related to the treatment and management accounting knowledge
variables were straightforward. The classification of ability,
however, was more complex.

To classify participants as either high or low in ability, the
median score on FDTOT of 12 was used as a cut-off. All
participants who scored 12 or greater on the FDTOT measure
were classified as high ability. Those who scored less than 12

were classified as low ability.1o On all the figures and tables that

 

1° Various cut-offs were used. If there was any indication of an interaction, the applicable
groups were split and separate analyses were performed for hypothesis testing purposes.

86
follow the dichotomized ability variable is labeled as FDDIC12 to

differentiate it from the ability variable, FDTOT.

Although scores on the accuracy variable could range from
zero to three, the interpretation of the results is most meaningfully
viewed in terms of a binary result, success or failure at selecting
an accurate system. By design, the experiment involved
predicting whether success would occur given different types of
knowledge and ability.

A dichotomous dependent variable creates problems for the
testing of the hypotheses. The assumptions necessary for
ordinary least squares regression analyses are violated. To
deal with the problems associated with a binary result, a variety of
statistical techniques and approaches were utilized to test the
hypotheses. In particular, after ANOVA was run to test for
interactions, logistic regression was run with a binary version of
the dependent variable.

Measurement error in the ability variable is another problem
that may mitigate the results. To gauge the effect of
measurement error, ordinary least squares regression was run
twice. The first ordinary least squares regression assumed
perfect measures. The second regression was computed with the

correlation matrix corrected for measurement error in the ability

87

variable.

Both hypotheses, industry training and ability, are tested in
sections 4.4.1 through 4.4.3. The first section uses ANOVA and
graphs to test for interactions. The logistics regression portion
of the analysis is presented next in section 4.4.2. Ordinary least
squares is covered in section 4.4.3.

4.4.1 Potential intergcfiin effects: accuracy As Figure 4.4
indicates, industry training provides no additional benefit for the

high management accounting knowledge group.

Figure 4.4
Genknow versus Industry: Accuracy

3.5 ‘

3.0:

 

2.5:
2.0:

1.5‘

O
O
O
O
O
O
o
a
r
O
f
'
fl
0
C
O
O
O
G
O
C .
O
I
O
'
’
O
O
G
O
a
O
’
O
’
O
'
O
r
’0
C
‘

‘

I

9
A

GENKNOW

 

Moan ACCTOT
h

I

 

INDUSTRY

88
The ANOVA presented in Table 4.10 shows one significant

and one marginally significant interaction. Both interactions
occur with the management accounting variable. This interaction
was expected due to the lack of variance for the accuracy variable
with the high management accounting group. Therefore, the
analyses that follow for the accuracy variable focus on
participants with low levels of management accounting knowledge.

Figure 4.5 plots the means for accuracy by industry training
and ability for the low management accounting group. The graph
indicates that both industry training and ability effect
performance. In addition, the graph makes it appear that the
slope of the ability line for the control group is steeper than the
slope for the treatment group. Different slopes suggest the
possibility of an interaction.

To investigate the possibility of an interaction between
industry training and ability for the low management accounting
group, ANOVA was rerun. The ANOVA, presented in Table 4.11,
shows main effects for industry training and ability. The

interaction between ability and industry training is not significant.

89

Table 4.10
ANOVA of AccuracyzFull sample

Tests of Significance for ACCURACY using UNIQUE sums of squares

Source of variation 88 DF MB F Sig of F
WITHIN+RESIDUAL 102.65, 135 .76

FDDIC12 4.79 1 4.79 6.29 .013
GENKNOW 38.87 1 38.87 51.12 .000
INDUSTRY 19.84 1 19.84 26.09 .000
FDDIClZ BY GENKNOW 2.46 1 2.46 3.24 .074
FDDIC12 BY INDUSTR! 1.26 1 1.26 1.65 .201
GENKNOW'BY INDUSTRI 14.71 1 14.71 19.34 .000
FDDIClZ BY GENKNOH'BY .25 1 .25 .33 .566
INDUSTR!

(model) 90.60 7 12.94 17.02 .000
(Total) 193.24 142 1.36

R-Squared - .469

Adjusted R-Squared = .441

90

Table 4.11
ANOVA of Accuracy: Genknow = 0

Tests of Significance for ACCURACY using UNIQUE sums of squares

Source of Variation SS DE MS F Sig of F
WITHIN+RESIDUAL 94 . 29 61 1 . 55

INDUSTRY 32 . 11 1 32 . 11 20 . 77 . 000
FDDIC12 6. 59 1 6. 59 4 . 27 . 043
INDUSTRY BY FDDIC12 1 . 23 1 1 . 23 . 80 . 376
(Model) 43.49 3 14.50 9.38 .000
(Total) 137 . 78 64 2 . 15

R-Squared - . 316

Adjusted R-Squared - .282

91

Figure 4.5
Industry versus Ability: Accuracy

"'l

3.0
2.5
2.0
1 .3

1.04 .............
""""" INDUSTRY

0

Mean ACCTOT

 

I

 

.00 1.00

FDDIC12

4.4.2 l___o_gistic reorgssigg. To accommodate logistic
regression, a binary version of the dependent variable was
created by collapsing the accuracy variable into the two mutually
exclusive categories. Participants who scored at least two out of
three questions correctly were classified as having succeeded at
the task. Those who scored one or less were classified as
having failed. The variable, accdic, is used to differentiate it
from the dependent variable, accuracy.

Logistic regression was computed with industry training,

92

ability and gender being entered into the model as potential
predictors.11 As Panel A of Table 4.12 shows, industry training
and ability were the only predictors to enter the equation.
Gender was insignificant.

The classification table presented in Panel B of Table 4.12
indicates that the model correctly predicted the success or failure
of 52 participants out of 65. The off-diagonal counts of 5 and 8
are observations that were not correctly classified by the model.
Overall, the interpretation of this observation is that the model
allowed the correct classification of 80% of the participants.

Another way to assess goodness of fit is to review the Model
Chi-Square statistics shown in Panel C of Table 4.12. The model
Chi-Square tests the null hypothesis that the coefficients of
industry training and ability are zero. The small significance
value indicates that the null hypothesis should be rejected. For
the low management accounting knowledge group, industry
training and ability appear to affect performance.

4.4.3 OLS regression The results of the OLS regression,
as shown in Table 4.13, are consistent with the logistic

regression analysis in that industry training and ability are

 

1‘ Prior accounting research (Awathi and Pratt, 1990) found a gender effect. Gender was
entered in all regressions to eliminate this factor as a potential explanatory predictor.

93

Table 4.12
Logistic regression of Accdic

PANEL A: LOGISTIC RBGRRBSION EQUATION

---------------------- Variables in the Equation -----------------------

Variable B S.!. Wald d1 Sig R. Exp(8)
INDUSTR! 2.7040 .6870 15.4926 1 .0001 .3927 14.9395
FDTOT .1913 .0821 5.4336 1 .0198 .1981 1.2109
Constant -2.6190 .9662 7.3478 1 .0067

 

- Model if Term.Removed ------------------
Based on Conditional Parameter Estimates

Term Log Significance
Removed Likelihood -2 Log LR df of Log LR

INDUSTR! -42.335 21.540 1 .0000
FDTOT -34.712 6.295 1 .0121

--------------- Variables not in the Equation -----------------

Residual Chi Square .560 with 1 df Sig = .4543
Variable Score df Sig R
GENDER .5599 1 .4543 .0000

PANEL B: Classification Table for ACCDIC

Predicted
.00 1.00 Percent Correct
0 " 1
Observed .“uuuuunguuuuuuu.
.00 o " 18 " 8 " 69.23%
.uuuuuuu.ununuuu.
1.00 1 " 5 " 34 " 87.18%
.\\\\““\\“\1.\\\\\\\\\\\\\\.

Overall 80.00%

PANEL C: Chi-Square Statistics
Chi-Square df Significance

Model Chi-Square 24.362 2 .0000
Improvement 6.125 1 .0133

94

Table 4.13
OLS regression of Accuracy

Multiple R . 57611
R Square .33190
Adjusted R Square .29905
Standard Error 1.22844

Analysis of variance

DF Sum.of Squares Mban Square
Regression 3 45.73109 15.24370
Residual 61 92.05352 1.50907
F - 10.10136 Signif F - .0000
------------------ variables in the Equation ------------------
variable B SE B Beta T Sig T
INDUSTR! 1.498535 .306590 .514567 4.888 .0000
FDTOT .100201 .038921 .275847 2.575 .0125
GENDER -.166515 .322232 -.055641 -.517 .6072
(Constant) .148995 .437109 .341 .7344

95

significant. Gender is not significant. The conclusion is
unchanged when measurement error in ability is considered. As
shown in Table 4.14, industry training and ability are the only
significant variables.

In conclusion, a shown in figure 4.6, the results supported
H1, the positive effect of industry training and Hz, the positive
effect of ability for those with low levels of management
accounting knowledge.

The study had predicted that industry training would provide
the treatment group with knowledge about the complexity of the
production process in a multi-color printing plant. Based on the
results, such knowledge appeared to help the treatment group
recognize the effect of a change in product mix on production
complexity.

Ability also seemed to affect performance for the low
management accounting group. Since the slides and scripts
provided no explicit accounting information, the successful
participant needed to make the connection between a change in
production complexity and the need to abandon the simplistic cost
driver heuristic shown in figure 2.1. According to the
manipulation check discussed in section 4.2.2, most of the

treatment group recognized the relationship between colors and

96

Table 4.14
Corrected OLS regression of Accuracy

Multiple R . 58471
R Square .34189
Adjusted R Square .30952
Standard Error 1.21923

Analysis of variance

DF Sum of Squares Mean Square
Regression 3 47.10688 15.70229
Residual 61 90.67773 1.48652
F - 10.56312 Signif F = .0000

------------------ Variables in the Equation ------------------

variable B SE B Beta T Sig T
INDUSTRY 1.498946 .304287 .514708 4.926 .0000
FDTOT .107024 .038684 .294630 2.767 .0075
GENDER -.184096 .320275 -.061516 -.575 .5675

(Constant) .089522 .433512 .207 .8371

97

Figure 4.6
Accuracy Hypotheses

 
   
 

Industry Training

Accuracy

Ability H2

98

production complexity. Nevertheless, certain of the lower ability
participants appeared to have difficulty applying that knowledge
to the task at hand. It is conceivable that, for the low
management accounting group, the task was somewhat novel and
difficult. Given a novel task, a positive effect of ability would be
expected and would be consistent with previous behavioral
research. Auditing behavioral research (Bonner and Lewis, 1990;
Libby and Tan ,1994) suggests that novel tasks require ability.

In contrast, neither industry training nor ability seem to be
needed if management accounting knowledge is high. One
explanation of this result focuses on the effect of strong data
analysis skills on performance. During debriefing many of the
high management accounting participants cited recent training in
regression as a key factor in their success. Previous regression
training may have sensitized participants to importance of
carefully examining patterns of resource usage. Knowledge of
data analysis techniques may be critical to application of
substantive knowledge about the production process to system
design decisions. As such, the results provide some preliminary
evidence that superior knowledge of data analysis techniques may
be able to substitute for both industry training and ability. In this

study, high levels of management accounting knowledge,

99

specifically familiarity with regression, seemed to allow
participants to know when to abandon the simple cost driver
selection rule. 12

4.5 H3 and H4: determinants of skill at noticing tracking cost

H3 predicted that industry training would positively affect the
tracking cost portion of the problem while H4 predicted the
positive effect of management accounting knowledge. Similar to
the results with accuracy, the first section of hypothesis testing
includes an analysis using graphs and ANOVA to test for
interaction effects. Both hypotheses were tested using logistic
regression and OLS regression.

To make the tracking cost variable suitable for logistic
regression, the responses were dichotomized into those who got
three out of four cost driver questions correct and those who did
not. The binary cost driver variable is shown as trackdic to
differentiate it from the tracking cost variable. In dichotomizing
the tracking cost variable, two borderline cases were noted in
which the participants scored exactly two out of the four questions
correctly. It could be argued that these two cases are difficult to
classify as either successes or failures. Therefore, the statistical

tests were run with and without these cases. The results were

 

‘2 The recent training in regression involved a class project unrelated to printing and activity-
based costing.

100
not substantially changed with the exclusion of the borderline
cases.

Each test was run first with all 116 participants and then
with the borderline cases eliminated. Section 4.5.1 covers the
ANOVA results. Section 4.5.2 covers the logistic regression
results. Section 4.5.3 through section 4.5.5 cover OLS
regression. Section 4.5.3 shows regression results assuming
perfect measures. Section 4.5.4. shows the results with the
borderline cases eliminated. Section 4.5.5 presents the linear
regression results when the two borderline cases are eliminated
and the resulting correlation matrix is corrected for measurement
error in ability.

4.5.1 PMtigl interafiction effecgz trackim cost A series
of five graphs were developed to examine the potential for two-
way and three-way interactions. The two way interactions are
examined in figure 4.7 through figure 4.9. The three way
interactions are considered in figures 4.10 and 4.11.

Figure 4.7 , the management accounting knowledge versus
industry training graph, indicates that industry knowledge and
management accounting knowledge have positive effects on

performance. There is no indication of an interaction.

101

Figure 4.7
Genknow versus Industryz'l'racking cost

4.01'
3.5‘
3.0
2.5‘

2.0

‘
0
CI

‘
s
O

...................... INDUSTRY

 

Mean TRACKING

0.0

 

The graph of industry training versus ability, Figure 4.8,
shows a positive effect for industry training. The ability line is

slightly positive for both the treatment and control groups. There

is no indication of an interaction.

102

Figure 4.8
Industry versus Ability:Tracking cost

4.0 ‘I
3.5:
3.0‘
2.5‘
2.0.
1.5‘

1.0: _____________
........................... INDUSTRY
'5‘ --.o
0.0 - - 1
.oo 1.00

 

Mean TRACKING

FDDIC12

Figure 4.9, the management accounting knowledge and
ability graph, shows a positive management accounting knowledge
effect and little or no effect for ability. The ANOVA results,
shown in Table 4.15 and Table 4.16, indicate that the interaction
between management accounting knowledge and ability is not

significant.

103

Figure 4.9
Genknow by Abilitszracking cost

4.0‘ *
3.6l
3.01
2.5‘ "‘

200‘

 

 

o 1.5
E
x
‘5 ‘-°‘ GENKNOW
.—

,5. 0
§ .
2 M .00 1.00

FDDIC12

The next two charts were developed to examine the
possibility of a three way interaction. To accomplish this, the
sample was first split into high and low management accounting
knowledge groups and the means for each management

accounting knowledge group were plotted by industry training and

ability.

TABLE 4.15
ANOVA of Tracking cost

104

Tests of Significance for tracking cost using UNIQUE sums

Source of variation

WITHIN+RESIDUAL
INDUSTRY

GENKNOW

FDDIC12

INDUSTRY BY GENKNOW
INDUSTRY 3! FDDICIZ
GENKNOW'BY FDDIC12
INDUSTRI BY GENKNOW
BY FDDIClZ

(model)
(Total)

Rquuared =
Adjusted R-Squared -

268.
110.
15.
2.
1.

14
76
53
83
36

.24
.14
.51

166.
434.

.383
.343

41
55

DE

108

HI‘I‘F‘P‘HI‘

115

2.
110.
15.
.83

2

1.
.24
.14
.51

MS
48
76
53

36

.77
.78

of squares
F Sig of F

.61
.26
.14
.55
.10
.06
.20

.57

.000
.014
.288
.461
.757
.812
.652

.000

105

TABLE 4.16
ANOVA of Trackdic

Tests of Significance for Trackdic using UNIQUE sums of squares

Source of Variation SS DF MS F Sig of F
WITHIN+RESIDUAL 18 . 08 108 . l7

INDUSTRY 6. 91 1 6. 91 41.28 .000
GENKNOW 1.23 1 1.23 7.36 .008
FDDIC12 . 08 1 . 08 . 48 . 489
INDUSTRY BY GENKNOW . 11 1 . 11 . 68 . 410
INDUSTRY BY FDDIC12 . 01 1 . 01 . 04 . 834
GENKNOW BY FDDIC12 . 01 1 . 01 . 05 . 827
INDUSTRY BY GENKNOW .04 1 .04 .22 .639
BY FDDIC12

(Model) 10.78 7 1.54 9.20 .000
(Total) 28.86 115 .25

R-Squared = . 374

Adjusted R—Squared = .333

106

Figure 4.10, the high management accounting knowledge
graph shows a large industry training effect with the ability line

showing virtually no effect on performance.

Figure 4.10
Industry versus Ability: Tracking cost
Genknow :3 1

4.01
3.5
3.0

.‘N
.90

‘

e

O
A

INDUSTRY

s
A

Mean TRACKING

.0
o

 

 

FDDIC12

Figure 4.11, the low management accounting knowledge
graph, also shows a large industry effect. Although the ability
line for the control group appears slightly steeper than the ability
line for the treatment group, the ANOVA results indicate that the

interaction is not significant.

107

Figure 4.11
Industry by Ability: Tracking cost
Genknow = 0

4.0
3.6
3.0
2.51 4
2.0<
1.51

1.01

 

Mean TRACKING

 

_. mousrav
.sl ................ o
0.0.; . I, -------- 1
.oo 1.00
FDDIC12

4.5.2 Logistic Regression Industry training, management
accounting knowledge, ability and gender were entered into the
model as potential predictors of success. As the results in Panel
A of Table 4.17 indicate, industry training and management
accounting knowledge were the only predictors to enter the

equation. Ability and gender were not significant.

108

Table 4.17
Logistic regression of trackdic: N=116

PANEL A: LOGISTIC REGRESSION EQUATION:

---------------------- Variables in the Equation -----------------------

Variable B S.E. Wald df Sig R. Exp(B)
INDUSTRY 3.1732 .5734 30.6246 1 .0000 .4226 23.8847
GENRNOI 1.7499 .5759 9.2336 1 .0024 .2125 5.7541
Constant -2.8290 .6440 19.2958 1 .0000

----------------- Model if Term Removed -—----------------
Based on Conditional Parameter Estimates

Term Log Significance
Removed Likelihood -2 Log LR df of Log LR

INDUSTRY -80.348 48.390 1 .0000
GENRNON -61.967 11.627 1 .0007

--------------- Variables not in the Equation -----------------

Residual Chi Square 1.539 with 2 df Sig I .4633
Variable Score df Sig R
FDTOT 1.5268 1 .2166 .0000
GENDER. .0250 1 .8744 .0000

PANEL B: Classification Table

Predicted
.00 1.00 Percent Correct
0 " 1
ob.°md .Ilnnflflnn.fl"""""".
.00 0 " 38 " 16 " 70.37%
.flIl"""""."""flnfl".
1.00 1 " 10 " 52 " 83.87%
.nnnnrrrrn.nnnnrrnn.

Overall 77.59%

PANEL C: Chi-Square Statistics

Chi-Square df Significance

Model Chi-Square 47.951 2 .0000
Improvement 11.021 1 .0009

109

In terms of goodness of fit, the classification table
presented in Panel B of Table 4.17, indicates that the model
allowed correct classification of 77.59% of the participants.

The Chi-square statistics presented in Panel C show that
the model containing the industry training and management
accounting knowledge variables is significant.

Logistic regression was rerun with the two borderline cases
eliminated. The results, shown in Table 4.18, were not
substantially different.

4.5.3 OLS regression. OLS regression, shown in Table
4.19, assumes perfect measurement and indicates that industry
and management accounting knowledge are the only significant
variables. Ability and gender were not significant at conventional
levels. OLS Regression was also run using the binary
dependent variable, trackdic. The results, shown in Table 4.20,
are not substantially different.

4.5.4 OLS remession- withogt the borderline cases The
regression program was run again without the borderline cases.
The results are shown in Table 4.21 and Table 4.22. Consistent
with the logistic regression, industry training and management
accounting knowledge were the only two variables entering the

regression equation.

110

Table 4.18
Logistic regression of trackdic: N=1 14

PANEL A: LOGISTIC REGRESSION:

---------------------- Variables in the Equation -----------------------

Variable B S.E. Wald df Sig R. Exp(B)
INDUSTRY 3.1594 .5725 30.4548 1 .0000 .4255 23.5569
GENKNON 1.6836 .5780 8.4832 1 .0036 .2031 5.3847
Constant -2.7340 .6428 18.0890 1 .0000

----------------- Model if Term Removed ------------------
Based on Conditional Parameter Estimates

Term Log Significance
Removed Likelihood -2 Log LR df of Log LR

INDUSTRY ~78.897 47.887 1 .0000
GENKNOI -60.223 10.539 1 .0012

-- - variables not in the Equation -----------------

 

Residual Chi Square 1.911 with 2 df Sig = .3847
Variable Score df Sig R
FDTOT 1.9054 1 .1675 .0000
GENDER. .0145 1 .9040 .0000

PANEL B: Classification Table

Predicted
.00 1.00 Percent Correct
0 " 1
ob.°md .“\\““\\“\\.“‘\“““\\“.
.00 0 " 37 " 15 ” 71.15%
.uuunuuu.uuuuuuu.
1.00 1 ” 10 " 52 " 83.87%

.““\\““““.\\\\““““\|.

Overall 78.07%

PANEL C: Chi-Square Statistics
Chi-Square df Significance

Model Chi-Square 47.252 2 .0000
Improvement 10.005 1 .0016

111

Table 4.19

OLS regression of Tracking cost: N=116

Mean Square

Multiple R . 61979
R Square .38414
Adjusted R Square .36194
Standard Error 1.55275
Analysis of variance

DF Sum.of Squares
Regression 4 166.92775
Residual 111 267.62398
F - 17.30878 Signif F a .0000
------------------ variables in the Equation ------------------
variable B SE B Beta
INDUSTRY 2.421341 .298897 .616143
GENKNOW .857610 .318923 .209323
FDTOT .052600 .034676 .116115
GENDER .088670 .292889 .022687
(Constant) -.474352 .509692

41.73194

2.41103
T Sig T
8.101 .0000
2.689 .0083
1.517 .1321
.303 .7627
—.931 .3540

112

Table 4.20

OLS regression of Trackdic: N=116

Mean Square

HNQ

2.68781
.16316

T S

.867
.981
.204
.117

ig T

.0000
.0035
.2311
.9067

Multiple R . 61033
R Square .37250
Adjusted R Square .34989
Standard Error .40393
Analysis of variance

DE Sum.of Squares
Regression 4 10.75125
Residual 111 18.11082
F a 16.47341 Signif F = .0000
------------------ variables in the Equation ------------------
variable B SE B Beta
INDUSTRY .611731 .077755 .604008
GENRNOW .247308 .082964 .234220
FDTOT .010862 .009020 .093038
GENDER .008946 .076192 .008882
(Constant) -.120137 .132591

.906

.3669

multiple R
R Square

Adjusted.R Square

Standard Error

113

TabIe 4.21

OLS regression of Tracking cost: N=114

.62115
.38583
.36329
1.56467

Analysis of variance

Mean Square

41.90977

2.44818
T Sig T
8.052 .0000
2.598 .0107
1.473 .1435
.232 .8168
-.886 .3777

DE‘ Sum.of Squares
Regression 4 167.63907
Residual 109 266.85216
F - 17.11871 Signif F I .0000
------------------ variables in the Equation ------------------
variable E SE B Beta
INDUSTRY 2.439374 .302966 .615066
GENRNON .840855 .323674 .203038
EDTOT .051754 .035125 .113511
SENDER .069023 .297237 .017503
(Constant) -.455895 .514656

114

Table 4.22
OLS regression of Trackdic: N=114

Multiple R . 61436
R Square .37744
Adjusted R Square .35460
Standard Error .40190
Analysis of variance

DF Sum.of Squares Mean Square
Regression 4 10.67431 2.66858
Residual 109 17.60639 .16153
F = 16.52098 Signif F = .0000
------------------ variables in the Equation ------------------
variable B SE B Beta T Sig T
INDUSTRY .612732 .077820 .605563 7.874 .0000
GENKNOW .233090 .083140 .220610 2.804 .0060
FDTOT .012108 .009022 .104091 1.342 .1824
GENDER .006744 .076349 .006703 .088 .9298
(Constant) -.116657 .132196 -.882 .3795

115

4.5.5 _O_LS regression---Correcteg for mea_§_quement error
The regressions were rerun with the correlation matrix corrected
for measurement error in ability. As shown in Table 4.23,
industry training and management accounting knowledge are the
only variables that enter the equation. One more regression was
run without the borderline cases and with the correlation matrix
corrected for measurement error in ability. The results shown in
Table 4.24 are not substantially different from the previous
regressions. Accounting for measurement error and exclusion of
the borderline cases did not change the conclusions regarding
determinants of performance.

In conclusion, as shown in figure 4.12, the results supported
H3, the positive effects of industry training, and H4, the positive
effects of management accounting knowledge. Both industry
training and management accounting knowledge seem to affect
skill at selecting the least costly driver.

To succeed, the participant needed to notice that resources
were used in direct proportion to the number of colors and that
the color driver was least costly to track. The positive effect
shown for industry training suggests that knowledge of the
production process helps participants identify the number of

colors as a major factor affecting press setup complexity.

P"

116

Table 4.23
Corrected OLS regression of Tracking cost: N=116

Multiple R . 62225
R Square .38720
Adjusted R Square .36511

Standard Error 1.54889

Analysis of variance

DE Sum.of Squares Mean Square
Regression 4 168.25718 42.06429
Residual 111 266.29455 2.39905
F a 17.53373 Signif F 8 .0000
------------------ Variables in the Equation ------------------
variable B SE B Beta T Sig T
INDUSTRY 2.427020 .298248 .617588 8.138 .0000
GENENOW .841504 .318741 .205392 2.640 .0095
FDTOT .058715 .034679 .129614 1.693 .0932
SENEER .087991 .292153 .022514 .301 .7638
(Constant) -.538762 .508465 -1.060 .2916

117

Table 4.24
Corrected OLS regression of Tracking cost: N=114

Multiple R . 62320
R Square .38838
Adjusted R Square .36593
Standard Error 1.56142
Analysis of variance

DF Sum.of Squares IMean Square
Regression 4 168.74676 42.18669
Residual 109 265.74446 2.43802
F - 17.30365 Signif F = .0000
------------------ variables in the Equation ------------------
variable B SE B Beta T Sig T
INDUSTRY 2.444379 .302428 .616328 8.083 .0000
GENRNON .824697 .323845 .199136 2.547 .0123
FDTOT .057079 .035167 .125189 1.623 .1075
GENDER .068820 .296614 .017451 .232 .8170
(Constant) -.510471 .513374 -.994 .3223

118

Figure 4.12
Tracking cost Hypotheses

 
   
  
   
 
   
 

H3
Industry Training

Tracking cost

» Management Accounting

119

Though knowledge of the link between color and the
complexity of press setup seems beneficial, familiarity with
management accounting also has an effect on performance. This
result was expected since the industry training scripts and slides
contain no explicit information about the accounting implications
of the relationship between colors and resource usage. Nor did
the scripts use terms such as cost driver or correlation in
describing the production process. For those reasons,
management accounting knowledge may have helped the
participant recognize the accounting significance of the
relationship between resource use and colors. In this study, both
industry training and management accounting knowledge
increased the participant’s likelihood of success.

4.6 Summary

Sections 1 discussed the confirmatory factor analysis for
both dependent variables. As a result of the confirmatory factor
analysis, a three item scale was used for the accuracy variable
and a four item scale was used for the tracking cost variable.

Before testing the hypotheses, manipulation checks were
examined. The manipulation of industry knowledge appeared to
work well for the both high and low management accounting

knowledge participants. Industry training allowed easy

120

acquisition of knowledge about the link between colors and
production complexity.

The experimental results supported both H1, the positive
effect of industry training, and Hz, the positive effect of ability for
the low management accounting group. The positive effect of
ability suggests industry training alone does not guarantee
success. This result may indicate that ability is critical to
application of industry knowledge when management accounting
knowledge is low. Among those with low levels of management
accounting knowledge, lower ability participants appeared to have
difficulty in applying industry knowledge to the accuracy task. In
contrast, nearly all of the high management accounting knowledge
participants were able to distinguish between an accurate and
distorted cost system. Due to recent training in regression, the
high management accounting group may have been more aware of
the importance of examining resource patterns than their low
management accounting counterparts. As a consequence,
industry training and ability had no effect on performance for the
high management accounting group.

H3, the positive effects of Industry training, and H.., the
positive effects of management accounting knowledge, were

supported. Industry training appears to help participants identify

121

color as a factor affecting production complexity. Management
accounting knowledge appears to help participants recognize the
accounting implications of the relationship between colors and
resource use.

The chapter that follows summarizes the study and provides

suggestions for future extensions to this work.

CHAPTER V
CONCLUSION
5.0 Overview

In this chapter, a summary of the research results is
presented, including a discussion of the limitations, contributions
and future extensions of the research. A summary of the
research results is presented in the first section. Limitations are
discussed in Section 2. Contributions and future extensions in
systems are presented in Section 3. Contributions related to
education are discussed in Section 4 and future extensions
related to economic implications are shown in the final section.
5.1 Summary of Results.

In the current study, superior management accounting
knowledge allowed high and low ability participants to identify
inaccurate systems without the benefit of specific industry
training. In contrast, given the same task, ability and industry
training showed a positive effect on performance for low
management accounting knowledge participants. For the low
management accounting knowledge group, industry training and
ability seemed to mitigate a tendency to rely on an inappropriate
heuristic for cost driver selection. The results suggests that low

management accounting knowledge participants have difficulty

122

123

identifying when a simple heuristic does not apply. One
preliminary implication of this finding is that superior levels of
management accounting knowledge may allow participants to
differentiate between significant and superficial aspects of a
given problem. Competence at differentiating between relevant
and irrelevant facts may assist the skilled performer at
recognizing when simple decision rules needs to be abandoned.

In terms of the tracking cost part of task, both types of
knowledge, industry specific and general management accounting,
positively affected performance. Whereas the high management
accounting knowledge participant brings superior command of
data manipulation techniques to the task, the industry trained
participant brings knowledge of the production process. Both
types of knowledge appear to aid the participant in noticing the
high correlation among competing cost drivers.
5.2. Limitations

The stated objective in the performance instrument was to
find the most accurate driver subject to minimizing tracking costs.
Using an accuracy criterion is consistent with previous analytical
work on cost driver optimization (Babad and Balachandran, 1993).
By adopting an accuracy criterion, the experiment did not ask

participants to explicitly quantify the cost of making incorrect

 

124

decisions. As a consequence, participants were not making a
tradeoff between the opportunity cost of incorrect decisions and
tracking costs. This simplification of the task was done for
reasons of experimental control and to avoid an extremely long
experimental session for the participants. In an earlier version of
the instrument, the correct cost driver answer required the
participant to explicitly compute the opportunity cost of an
incorrect decision in terms of a single decision and single
decision maker. The overwhelming response from pretest
participants was that the single decision single decision maker
setting was unrealistic. Using a single decision setting was also
judged unrealistic for the institutional setting chosen for this
study, package printing. In package printing, the practical uses
of cost information include a variety of decisions including
contract negotiation, capital investment decisions and long term
strategic planning issues. Adding multiple decisions to the
experimental stimuli would have required a significant increase in
the length and complexity of the experiment. In addition,
participants may have different attitudes about the costs
associated with making incorrect decisions. This difference in
attitude would be expected to confound the results making it

impossible to classify a response as right or wrong. Therefore,

 

125

the final version of the experimental stimuli asks the subject to
select the most accurate set of cost drivers subject to minimizing
tracking costs.

The study also assumes perfect correlations between the
number of colors in the design and the components of variable
setup cost. In reality, some variance from planned resource
usage would be expected. In this particular industry, however,
there is an extremely high correlation between the between
number of colors and setup resource usage. In the course of
validating the case materials, a package plant controller reviewed
all the resource and cost information and judged them to be
realistic. Beyond that, the study had six participants who had
multi-color printing experience.13 One of these individuals was a
trained accountant who works as a consultant to the printing
industry. All six of the experienced individuals answered the
accuracy and tracking cost aspects of the task correctly without
the benefit of the industry training session. During debriefing, all
six individuals indicated that the materials appeared realistic
based on their personal experiences in the field.

Another potential limitation is that this study focused on cost

driver selection in one major industry. In addition, the scope of

 

‘3 The six individuals were not part of the statistical results. Please refer to section 3.3.3.

126

the industry knowledge manipulation was deliberately limited for
reasons of experimental control. Therefore, the results may not
be applicable to other tasks and other institutional settings.
Nevertheless, demonstrating the effect of a specific element of
knowledge on performance is a logical starting point for a
research agenda that examines the relationship between
technology and knowledge in an ABC environment. If knowledge
reduces task complexity in cost driver selection, similar
knowledge effects may be applicable in other tasks and settings.
In fact, Libby and Luft (1993) make the following suggestion to

accounting behavioral researchers outside the audit area:

"We recommend a similar approach to that taken in the

audit literature, beginning with an analysis of key

attributes of the settings and task requirements. (p.40)"
5.3 Contributions and Future Extensions: Systems

By providing insight about individual differences in

performance, this study has two major implications for systems
development. First, the research provides preliminary evidence
on the effect of different types of knowledge on cost driver

selection. The results suggest that both management accounting

knowledge and industry training contribute to simplification of the

127

process of selecting and evaluating cost drivers. The preliminary
evidence about different types of knowledge may be helpful in
developing guidelines for selection of team members for cost
system development projects. Including individuals with different
perspectives may enhance the productivity of an ABC
development team. It is conceivable that a team comprised of
individuals with different cost perspectives may have synergies
that would facilitate the design of ABC systems. An extension of
this current research could examine the effects of management
accounting and industry specific knowledge in a group decision
making setting.

Second, a better understanding of knowledge effects is also
expected to help systems designers build effective cost systems.
By increasing awareness of knowledge effects, this study is
expected to highlight the need to consider individual differences
when designing computer-based decision aids. The user
interfaces and decision aids are likely to be quite different
depending upon the expertise level assumed for the user of the
system. Future research is expected to use the results of this
study to plan experiments that examine the relationship between

knowledge and specific decision aids.

128

5.4 Contributions: Education

By examining the relationship between performance and
individual differences, this research has implications for designing
training and instructional materials. The research results may be
helpful in designing learning experiences that allow efficient
acquisition of knowledge in formal educational settings. This
study demonstrated that a short presentation on real world
production processes can dramatically improve participants'
performance.
5.5 Future extensions --- Economic Implications

In this study, knowledgeable individuals were able to
recognize situations in which use of a simplistic heuristic would
result in a system with distorted costs. Skill at recognizing when
to abandon a simple decision rule may be a valuable skill in
dealing with a less than accurate cost accounting system.
Surveys of practice have consistently shown that decision makers
rely heavily on costs to set prices (Cornick, et al.1988,
Govindarajan and Anthony, 1983). By recognizing distortions in
cost, the knowledgeable system user may be able to minimize the
opportunity cost of relying on an inaccurate cost system. The

well-trained user, as a component of the management accounting

129

system, may be able to compensate for lack of accuracy in the
accounting system. The quality of the decisions made may
depend not only on the formal accounting system but also on the
type and level of expertise of the systems user.

One extension of this research would examine the economic
implications of transferring knowledge from human experts to the
formal control system. Jensen and Meckling (1992) indicate that
control systems are a means of dealing with diverse knowledge
and decision rights within a decentralized institutional setting.

According to Jensen and Meckling, effective firms carefully
consider the costs and benefits of transferring knowledge and
decision rights. Firms that are adopting ABC may have found it
impractical to continue to rely on human expertise. The decision
to replace the cost system may reflect a decision to transfer some

of the human expertise to the control system.

APPENDIX A

APPENDIX A

THE INDUSTRY TRAINING SESSION

The training session was derived from training materials developed and
used by the industry trade association (Flexographic Technical Association,
1991; Flexographic Technical Association, 1986 ; Flexographic Technical
Association and Graphics Arts Technical Foundation, 1982). According to the
industryexpert, the material includes a sufficient amount of basic information to
provide an understanding of the fundamentals of the production process to
non-technical employees. Typically, the materials are used for in-plant
training of salesman, accountants, marketing managers, personnel managers
and other non-technical employees.

The training materials were reviewed by two committee members with
extensive management accounting experience. To keep the treatment to a
reasonable length, some material was eliminated as part of this review process.
For example, flexographic printing can accommodate two types of design, line
art and continuous-tone art. The industry training materials include a sizable
amount of information about both types of design. The case materials in this
study, on the other hand, only involve one type of design, line art. Therefore,
the decision was made to exclude the material related to continuous-tone art

for purposes of this study.

130

131

Before a final decision was made about which material to use in the
study, the industry expert was again consulted. The industry expert, who had
administered the training materials countless times, had specific suggestions
given the scope and objective of this study. These suggestions were
incorporated into the treatment used in this study. The final training package
used in this study included 30 audio visual slides and scripts covering the
following topics:

1. History of printing

2. Pre-Press activities

3. Printing Press Equipment

4. Press setup

The first series of slides provide an overview of the history of printing.
The history lesson includes a review of products that use packages that are
printed with flexography. The products shown in the slides include many
common, everyday products. A key characteristic of the products shown in
the slides is the use of packaging with multiple colors in the design.

After the history slides are shown, the next group of slides cover key
activities done prior to press setup. The descriptions of these activities imply
that there is a relationship between the complexity of the job and the number of
colors in the design.

Two additional slides show other products that make use of

flexography in their packaging. The scripts related to these slides discuss the

132

fact that many companies feel that colorful packaging helps to sell their
product.

The next series of slides and scripts cover the components of a typical
printing press. The central impression press, the same equipment used by the
company in the case materials, is highlighted in this part of the lesson. These
slides also introduce the significance of the print station in flexographic
printing. Modern flexographic presses have up to eight print stations which
allow printing up to eight colors with one pass through the printing press.

Press setup is covered in two slides. The first slide shows a central
impression press with the print stations around a metal cylinder. The next
slide summarizes the activities that go into setting up a central impression
press for a multi-color job. The accompanying script states that the number of
print stations that need to be used on any given job depends on the number of
colors to be printed. The lesson concludes with more examples of products

that use printed packaging.

APPENDIX B

APPENDIX B

CASE STUDY

Mike Thompson's printing press setup cosh.
Background

Mike Thompson is the owner of a printing business that supplies printed bags to
food businesses. Until recently. all of Mike's business was printed in one plant located in
the town of Webfield. Last week, Mike finalized the acquisition of another printing plant
in the nearby town of Scotsport '

Mfire plans to judiciously schedule production at both plants to maximize
profitability. As a result, some of the jobs that are currently produced at Webfield may
eventually be produced at Scotsport and visa versa.

Because of the acquisition of Scotsport. Mike hired you as a consultant. He wants
you to make a recommendation about the adequacy of his current record keeping
practices given the acquisition of the Scotsport business. Your first priority is to examine
the record keeping practices surrounding press setup. Mike wants accurate
information about variable setup costs but he doesn't want to spend any more than
necessary tracking the data.

What is Involved In the Press setup activity?

Because of the large size of the press machinery. each printing press requires
two press operators be involved in setting up the press.

Setting up the press requires the press operators place the printing plate
cylinders for that particular job onto the printing press. The printing plate cylinder holds
the image to be printed for the given production run. Then the operators place ink into an
ink pan. Once these steps are done. the printed images are aligned with each other.
Lastly, the press machinery is started and the settings between and among the various
component parts of the press are made.

Current Record Keeping Practices

Currently. the Webfield plant does not track any information about variable setup
costs on a per order basis. Instead. management keeps track of total variable setup
costs on an aggregate basis. Last year. total variable setup costs were 898.000 for 500
customer orders. Total variable setup costs consists of three types of the costs: (1)
$45,000 in setup labor (2) $40,500 In ink wasted during setup and (3) $12,500 for plastic
film setup rolls. A total of 750 setup labor hours. 13,500 pounds of ink and 500 setup
rolls were used in setting up the 500 customer orders last year.

The practice of relying on averages has led to the use of a 'Rule of Thumb' that
estimates variable setup cost at fiat amount of $196.00 per order. The “Rule of Thumb'
of $196.00 per order is very accurate for estimating variable setup costs for the
Webfield business. Although each customer orders their bags to their own design
specifications. the setup costs of any two of these orders tends to be very similar.

(Continue to Next Page)

133

134

The Typical Webfield order

The typical order currently produced at the Webfield plant has the following
specifications:

 

 

 

Type of design line art
# of setups per order 1

# of setup rolls per setup 1

it of colors in the design 3

I of setup labor hours per setup 1.5
8 of ink lbs wasted per setup. 27

 

As the above specifications show. printing for the typical Webfield customer
requires only one press setup per order. This is because these customers are willing to
take shipment of a whole order of bags at once. This has allowed Mike to print each
customer's individual order in one single press run. As a consequence. the cost per
order and the cost per setup are both $196.00 for these customers.($196.00 per
setup X 1 setup per order 8 $196.00 per order). Mike is confident that the variable setup
cost for the these orders will continue to be $196.00 per order in the future.

The Scotsport Plant

Mike is much less certain about using $196.00 per order to estimate the setup
costs for the Scotsport customers. Although the newly acquired Scotsport plant is
virtually identical to the Webfield plant in production capacity. the typical orders
produced at each individual plant may be different Mike wants you to compare the
typical Webfield order to the typical Scotsport order and compute an estimate of the
annual variable setup cost for the Scotsport customers.

In computing variable costs. Mike assures you that the setup labor rate per hour.
the ink price per pound and cost per setup roll for the Webfield and Scotsport plants are
the same and are expected to remain so in the future. There are only two salient
differences between the two plants: (1) the specifications of the typical order currently
produced at each location and (2) the total number of customer orders each plant
currently produces.

Afier you compute an annual variable cost amount. Mike also wants you to
recommend whether or not he needs to keep track of more detailed information (called a
cost driver) about each order. He wants the cheapest system that provides him with
accurate costs.

To help you in the cost driver part of the assignment. Mike reduced the number of
cost driver alternatives you need to evaluate to the five listed in Table 1 on the next
page. The five alternatives along with the annual tracking cost of each are shown on the
next page. One of the five alternatives is to maintain the current system and continue to
use the $196.00 per order to compute variable setup costs for the Webfield and Scotport
business. One of the advantages of the current system is that it doesn't cost anything to
maintain. This advantage is lost. however. if the using the current system would result in

135

inaccurate variable setup costs. Mike works on long-term contracts and needs accurate
cost information.
Table 1

I of orders current

I order
I of colors In the I. I of
I I of hours a I of
5 I of Ink lbs wasted a I of 1

 

Alternatives 2 - 5 shown above require that MikeJteep track of additional
information over and above the number of orders.

Requlred:

You will be given 10 different problem sets (numbered 1-10) and asked to
complete the same two questions shown for each. Answer each of the problem sets
INDEPENDENTLY of the others.

The two questions are as follows:

a. What is the annual variable setup cost for the Scotsport customers?(Five multiple
choice answers will be provided for each of these questions.)

In completing this question. you should first determine the accumte cost per
order for the Scotsport business. This requires that you compare the specifications of
the Webfield and Scotsport orders. Once you have computed the accurate setup cost
per order for the Scotsport business. use the following formula to compute annual
variable costs.

Annual costs (Scotsport business) I- Scotsport cost per order ° I or Scotsport orders.

b. Which of the following alternatives will allow computation of accurate variable setup
costs per order at the lowest tracking cost?

(1) I of orders (the current system)

(2) I of setups per order

(3) both I of setups per order and the I of colors in the design

(4) both I of setups per order and the I of setup hours per setup
(5) both I of setups per order and the I of ink lbs wasted per setup.

For this question. assume that accurate costs per order are needed and select
the system that is accurate and the least costly.

PLEASE SHOW YOUR WORK
(Continue to Next Page)

136

Problem Set 1

Given the background information. supplemental information and the following additional
information about the Scotsport customers. answer questions a and b below:

Estimated Annual Order Quantity and Variable Setup cost
Webfield vs. Scotsport Customers

 

Webfleld Scotsport |

_—

I et orders gr year 500 IWI
Variable Setup cost per year 890,000 7 |

 

 

 

 

 

 

Specifications of the The Typical Order
Webfleld versus Scotsport

 

608“

a. What is the annual variable setup cost for the Scotsport customers?

(1) $196.00

(2) $98,000

(3) $186,200

(4) 588.200

(5) none of the above.

b. Which of the following altematlves will allow computation of accurate variable setup
costs per order at the lowest tracking cost?

(1) I of orders (the current system)

(2) I of setups per order

(3) both I of setups per order and the I of colors in the design

(4) both I of setups per order and the I of setup hours per setup
(5) both I of setups per order and the I of ink lbs wasted per setup.

137

Problem Set 2

Given the background information. supplemental information and the following additional
information about the Scotsport customers. answer questions a and b below:

Estimated Annual Order Quantity and Variable Setup cost
Webfleld vs. Scotsport Customers

Webfleld Sooner; I
I of orders per year 800

Variable Setup cost per year 890.000 T!

 

 

 

 

 

 

 

 

Specifications of the The Typical Order
Webfield versus Scotsport

COO”

 

a. What is the annual variable setup cost for the Scotsport customers?

(1) 8117.600

(2) 8102.600 .

(3) 8294.000

(4) 839.200

(5) none of the above.

b. Which of the following alternatives will allow computation of accurate variable setup
costs per order at the lowest trscklng cost?

(1) I of orders (the current system)

(2) I of setups per order

(3) both I of setups per order and the I of colors In the design

(4) both I of setups per order and the I of setup hours per setup
(5) both I of setups per order and the I of ink lbs wasted per setup.

138

Problem Set 3

Given the background information. supplemental information and the following additional
information about the Scotsport customers. answer questions a and b below:

Estimated Annual Order Quantity and Variable Setup cost
Webfleld vs. Scotsport Customers

 

Webfleld Scots

I of orders per year 500 50
Variable Setup cost per year 898.000 ? l

 

 

 

 

 

 

Specifications of the The Typical Order
Webfleld versus Scotsport

 

costs

a. What is the annual variable setup cost for the Scotsport customers?

(1) 89.800

(2) 851.300

(3) 858.800

(4) 8588.000

(5) none of the above.

b. Which of the following alternatives will allow computation of accurate variable setup
costs per order at the lowest tracking cost?

(1) I of orders (the current system)

(2) I of setups per order

(3) both I of setups per order and the I of colors in the design

(4) both I of setups per order and the I of setup hours per setup
(5) both I of setups per order and the I of ink lbs wasted per setup.

139

Problem Set 4

Given the background information. supplemental information and the following additional
information about the Scotsport customers. answer questions a and b below:

Estimated Annual Order Quantity and Variable Setup cost
Webfleld vs. Scotsport Customers

 

Webfleld Scotsport
I of orders gr year 800 300
Variable Setup cost per year 898,000 7

 

 

 

 

 

 

 

Specifications of the The Typical Order
Webfleld versus Scotsport

CO.”

 

a. What is the annual variable setup cost for the Scotsport customers?

(1) 8392.000

(2) 8235.200

(3) 858.800

(4) 8205.200

(5) none of the above.

b. Which of the following altematives will allow computation of accurate variable setup
costs per order at the lowest tracklng cost?

(1) I of orders (the current system)

(2) I of setups per order

(3) both I of setups per order and the I of colors in the design
(4)bothIofsetupsperorderandtheIofsetuphours persetup
(5) both I of setups per order and the I of ink lbs wasted per setup.

140

Problem Set 5

Given the background information. supplemental information and the following additional
information about the Scotsport customers. answer questions a and b below:

Estimated Annual Order Quantity and Variable Setup cost
Webfleld vs. Scotsport Customers

 

Webfleld Scotsport
I of orders per year 800 100
Variable Setup cost per year 998.000 ?

 

 

 

 

 

 

 

Specifications of the The Typical Order
Webfleld versus Scotsport

CO."

 

a. What is the annual variable setup cost for the Scotsport customers?

(1) 819.600

(2) 8490.000

(3) 885.500

(4) 898.000

(5) none of the above.

b. Which of the following alternatives will allow computation of accurate variable setup
costs per order at the lowest tracking cost?

(1) I of orders (the current system)

(2) I of setups per order

(3) both I of setups per order and the I of colors in the design

(4) both I of setups per order and the I of setup hours per setup
(5) both I of setups per order and the I of ink lbs wasted per setup.

141

Problem Set 6

Given the background information. supplemental information and the following additional
information about the Scotsport customers. answer questions a and b below:

Estimated Annual Order Quantity and Variable Setup cost
Webfleld vs. Scotsport Customers

 

 

 

 

 

Webfleld _ Scotsport |
I of orders per year 500 90 I
Variable Setup cost per year 899,000 ? |

 

 

Specifications of the The Typical Order
Webfleld versus Scotsport

 

60888

a. What is the annual variable setup cost for the Scotsport customers?

(1) 875.060

(2) 861.560

(3) 817.640

(4) 8105.840

(5) none of the above.

b. Which of the following alternatives will allow computation of accurate variable setup
costs per order at the lowest tracking cost?

(1) I of orders (the current system)

(2) I of setups per order

(3) both I of setups per order and the I of colors in the design

(4) both I of setups per order and the I of setup hours per setup'
(5) both I of setups per order and the I of ink lbs wasted per setup.

142

Problem Set 7

Given the background information. supplemental information and the following additional
information about the Scotsport customers. answer questions a and b below.

Estimated Annual Order Quantity and Variable Setup cost
Webfield vs. Scotsport Customers

 

Webfield Scotsport fl

r

I of orders per year 800 300
Variable Setup cost per year 899.000 7

 

 

 

 

 

 

Specifications of the The Typical Order
Webfleld versus Scotsport

€083

 

a. What is the annual variable setup cost for the Scotsport customers?

(1) 858.800

(2) 8303.600

(3) 8273.600

(4) 8235.200

(5) none of the above.

b. Which of the following alternatives will allow computation of accurate variable setup
costs per order at the lowest tracking cost?

(1) I of orders (the current system)

(2) I of setups per order

(3) both I of setups per order and the I of colors in the design

(4) both I of setups per order and the I of setup hours per setup
(5) both I of setups per order and the I of ink lbs wasted per setup.

143

Problem Set 8

Given the background information. supplemental information and the following additional
information about the Scotsport customers. answer questions a and b below:

Esdmated Annual Order Quantity and Variable Setup cost
Webfleld vs. Scotsport Customers

Webfleld Scotsport
I of orders per year 500 400

Variable Setup cost per year 899,000 7 ]

 

 

 

 

 

 

 

Specifications of the The Typical Order
Webfleld versus Scotsport

 

COS”

a. What is the annual variable setup cost for the Scotsport customers?

(1) 8156.800

(2) 878.400

(3) 865.600

(4) 845.600

(5) none of the above.

b. Which of the following alternatives will allow computation of accurate variable setup
costs per order at the lowest tracking cost?

(1) I of orders (the current system)

(2) I of setups per order

(3) both I of setups per order and the I of colors in the design

(4) both I of setups per order and the I of setup hours per setup
(5) both I of setups per order and the I of ink lbs wasted per setup.

144

Problem Set 9

Given the background information. supplemental information and the following additional
information about the Scotsport customers. answer questions a and b below:

Estimated Annual Order Quantity and Variable Setup cost
Webfleld vs. Scotsport Customers

 

 

 

 

 

 

Webfleld Scotsport J
I of orders per year 500 80 |
Variable Setup cost per year 899.000 7 |

 

Specifications of the The Typical Order
Webfleld versus Scotsport

 

608”

a. What is the annual variable setup cost for the Scotsport customers?

(1) 8173.600

(2) 815.680

(3) 8109.760

(4) 8159.600

(5) none of the above.

b. Which of the following altematives will allow computation of accurate variable setup
costs per order at the lowest tracking cost?

( 1) I of orders (the current system)

(2) I of setups per order

(3) both I of setups per order and the I of colors in the design

(4) both I of setups per order and the I of setup hours per setup
(5) both I of setups per order and the I of ink lbs wasted per setup.

145

Problem Set 10

Given the background information. supplemental information and the following additional
information about the Scotsport customers. answer questions a and b below:

Estimated Annual Order Quantity and Variable Setup cost
Webfield vs. Scotsport Customers

 

Webfleld Scotsport
I of orders per year 500 80
Variable Setup cost per year 898.000 ?

 

 

 

 

 

 

 

Specifications of the The Typical Order
Webfield versus Scotsport

costs

 

a. What is the annual variable setup cost for the Scotsport customers?

(1) 841.040
(2)81 1.760
(3)844.040
(4) 823.520
(5) none of the above.

b. Which of the following alternatives will allow computation of accurate variable setup
costs per order at the lowest tracking cost?

(1) I of orders (the current system)

(2) I of setups per order

(3) both I of setups per order and the I of colors in the design

(4) both I of setups per order and the I of setup hours per setup
(5) both I of setups per order and the I of ink lbs wasted per setup.

REFERENCES

REFERENCES

Ackerman,P.L.1988. Determinants of individual differences during skill
acquisition. Journal of Experimental Psychology, General.
117:288-318.

Akin, O. 1980. Models of Architectural Knowledge .London: Pion.

Awathi, V.. and J. Pratt. 1990. The effects of monetary incentives on effort and
decision performance: The role of cognitive characteristics. The
Accounting Review 65 (October):797—81 1.

Babad, Y. and B.Balachandran. 1993. Cost driver optimization in Activity-
Based Costing. The Accounting Review 68 (July):563-575.

Banker, R. and H. Johnston. 1993. An empirical study of cost drivers in the
U.S.Airline Industry. The Accounting Review 68 (July):576-601.

and G. Potter. Economic Implications of single cost driver
systems. Journal of Management Accounting Research. Vol.5 (Fall):15-
31.

Bedard, J. and S. Biggs, 1991. Pattern recognition. hypothesis generation and
auditor performance in an analytical task.. The Accounting Review.
66 (July):622-642

Benbasat, Land A.S.Dexter. 1979. Value and events approaches to accounting:
An experimental evaluation. The Accounting Review..54 (October):
735-49

Bonner, S.E. 1994. A Model of the effects of Audit Task Complexity. Working
Paper. University of Southern California.

Bonner. S.E. and N. Pennington. 1991. Cognitive processes and knowledge as
determinants of auditor expertise. Journal of Accounting Literature: 17-
38.

and B. Lewis. 1990. Determinants of auditor expertise.Joumal of
Accounting Research. (Supplement): 1 -20.

146

147

Charness. N. 1979. Components of skill in bridge. Canadian Journal of
Psychology. (March): 1-16.

Chase, W.G. and HA. Simon. 1973. The mind's eye in chess. In Visual
Information Processing. New York: Academic Press.

Chi. M. R.Glaser and E.Reese.1982. Expertise in Problem Solving. In
Advances in the psychology of human intelligence, edited by
R.Sternberg, Vol. 1.Hi|lsdale, NJ.Lawrence Erlbaum Associates

Comick. M. and W. Cooper. and S. Wilson. 1988. How do companies analyze
overhead? Management Accounting. (June): 41 -43.

Datar, S.M, S. Kekre. S. T. Mukhopadhyay, and K. Srinivasan. 1993.
Simultaneous estimation of cost drivers. The Accounting Review
68(July):602-614.

Datar. SM. and M. Gupta. 1994. Aggregation, specification and measurement
errors in product costing. The Accounting Review 69. (October):567-591.

Demski.J. 1980./nformation Analysis 2nd Ed Reading MA: Addison-Wesley.

Demski, J. and G. Feltham. 1972. Forecast evaluation. The Accounting Review
47(July):533-548.

Demski. J. and G. Feltham. 1976. Cost Determination: A Conceptual Approach.
Ames, IA: Iowa State University Press.

Dewan RM. and RP. Magee. 1992. Product costing systems: Information,
economics, and implementation. Working paper, Northwestern
University, Evanston, ll.

Doktor, R. H. and W. F. Hamilton. 1973. Cognitive style and the acceptance of
management science recommendations. Management Science.19
(April) 884-94.

Dopuch.N. 1993. A Perspective on Cost Drivers. The Accounting Review. 68
(July):615—620.

Egan. D and B. J. Schwartz. 1979. Chunking in recall of symbolic drawings.
Memory and Cognition. 7: 149-158

148

Engle. R.W. and L. Bukstel. 1978. Memory processes among bridge players of
differing expertise. American Journal of Psychology.(December): 763-
789.

Feltham. G. 1968. The Value of Information. The Accounting Review (October):
684- 696.

Feltham, G. 1977. Cost aggregation: An information economic analysis. Journal
of Accounting Research (Spring):42-70.

Feltham. G. and J. Demski. 1970. The use of models in information evaluation.
The Accounting Review 45 (October):623-640.

Foster. G. and M. Gupta. 1990.Manufacturing overhead cost driver analysis.
Journal of Accounting and Economics 12:309-337.

Foundation of Flexographic Technical Association. 1991. FIex/Sys Technical
Training System. FTA. Ronkonkoma,New York.

. 1986. The Beginning Flexographer. FTA. Ronkonkoma, New York..

. and Graphics Arts Technical Foundation. 1982. The Flexographic
Pn'nting Process. FTA. Huntington Station. New York.

Govindarajan, V. and R. Anthony. 1983. How firms use cost data in price
decisions. Management Accounting (July): 30-36.

Gui. F. A. 1984. The joint and moderating role of personality and cognitive style
on decision making. The Accounting Review 59 (April): 264-77.

and O. Zaid. 1981. Field dependence and accountants'confidence in
decisions. Psychological Reports (December): 949-50.

Gupta. M. 1993. Heterogeneity issues in aggregated costing systems. Journal
of Management Accounting Research. (Fall):180-212.

Hammond. KR. 1986. A theoretically based review of theory and research in
judgment and decision making. Research report .No.260. Center for
Research on Judgment and Policy, University of Colorado.

Hunter. J.E. 1986. "Cognitive ability. cognitive aptitude, job knowledge. and job
performance. Journal of Vocational Behavior 29: 340-362.

 

149

1983A Causal analysis of cognitive ability.job knowledge.job
performance, and supervisory ratings, In Performance measurement
and theory, edited by F.Landy. S. Zedeck and J. Cleveland,Hi|ldale. NJ.
Lawrence Erlbaum Associates.

and D. W. Gerbing. 1982. Unidimensional measurement, second
order factor analysis. and causal models. Research in Organizational
Behavior 4: 267-320.

Hwang, Y., J.H. Evans Ill and V. Hedge. 1993. Product cost bias and selection
of an allocation base. Journal of Management Accounting .Research
Fall: 13-242.

Jensen. MC. and W.H. Meckling.1992. Specific and general knowledge,and
organizational structure. From Contract Economics by Werin, Lars
and Hans Wijkander. Blackwell Publishers. Oxford: 251 -274.

Joseph. GM. and V.L. Patel. 1986. Specificity of expertise in clinical reasoning.
In Proceedings of the eighth annual conference of cognitive science
society. Hillsdale. NJ. Lawrence Erlbaum Associates.

Kaplan, RS. 1986. Accounting langhe obsolescence of cost accounting
systems. California Management Review (Winter):175—199.

Lesgold, A. H. Rubinson. P. Feltovich. R. Glaser. D. Klopfer. and Y.Wang.
1988. Expertise in complex skill: Diagnosing X-Ray pictures. In The
Nature of Expertise, edited by M. Chi et al.Hillsdale, NY: Lawrence
Erblaum Associates.

Lesgold. AM. 1984. Acquiring Expertise. In Tutorials in Learning and
Memory'Essays in Honor of Gordon Bower, edited by JR. Anderson
and SM. Kosslyn. New York: W.H. Freeman and Co.

Libby. R. and H. Tan. 1994. Modeling the determinants of audit
expertise.Working Paper, Cornell University.

, 1994. The role of knowledge and memory in audit judgment. In
Judgment and Decision Making Research in Accounting and Auditing.
ed. R.H. Ashton and A..H.Ashton.

 

150

and J. Luft. 1993. Determinants of judgment performance in
accounting settingszAbility, knowledge. motivation and environment.
Accounting. Organizations, and Society, Vol. 18 No. 5: 425-450.

Lusk, E. J. 1973. Cognitive aspects of annual reports: Field
independence/dependence. Journal of Accounting Research 11
(Supplement): 191 -203.

McCarthy. W. 1982. The REA accounting model:A generalized framework for
accounting systems in a shared data environment. The Accounting
Review (July):554-578.

McKeithen, K.B. J.S. Reitman, H.H. Reuter. and SC. Hirtle. 1981. Knowledge
organization and skill differences in computer programmers. Cognitive
Psychology 13, 307-325.

Naylor. J. C. and EA. Schenck. 1968. The influence of cue redundancy upon
human inference process for tasks of varying degree of
predictability. Organizational Behavior and Human Performance.47-61..

Newell. A. and HA Simon. 1972. Human problem solving. Englewood Cliff,
NJzPrentice Hall.

Simon, HA. 1979. Information processing models of cognition. Annual Review
of Psychology. :363-396.

Soloway. E.. B. Adelson and K Ehrlich. 1988. Knowledge and processes in the
comprehension of computer programs. In Nature of Expertise, edited by
M.Chi et al. Hillsdale. NY: Lawrence Erblaum Associates.

nICHIan sran UNIV. LIBRARIES
IIIIIWIIIIIIIIIIIIWIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII
31293016885471