A. prepare a summary of each paper (including the Libby box summary)

Focus on summarizing the most salient points of the article. Copy/paste is okay as long as what’s in there makes complete sense. After that make sure the summaries address the following questions (I will send them).

Avoid too much about methods/stats

B .and make PowerPoint slides for each study.

Three Articles in the attachments

1. Beaver.68

2. Birnberg.11.

3. Kothari.01

BEHAVIORAL RESEARCH IN ACCOUNTING
Vol. 23, No. 1
2011
pp. 1–43
American Accounting Association
DOI: 10.2308/bria.2011.23.1.1
A Proposed Framework for Behavioral
Accounting Research
Jacob G. Birnberg
University of Pittsburgh
ABSTRACT: Behavioral accounting research 共BAR兲 is richer today, in the topics covered, the methods used, and the range of sub-areas of accounting in which it is performed, than ever before. This paper offers a framework within which BAR literature can
be viewed as a whole rather than in segments, such as by accounting sub-areas or by
research method. The framework classifies BAR by the focus of the research: the
individual, group, organization, or the society within which accounting exists. The purpose of the framework is to help researchers in BAR to appreciate the insights to their
research questions that can be found in BAR using another research method or studying a similar issue in another sub-area of accounting. Existing research in each of these
four areas is discussed to illustrate the usefulness of the framework. In addition, behavioral research in other disciplines that could impact BAR and areas of potential
future research are discussed.
Keywords: behavioral accounting research.
INTRODUCTION
n the 20 or so years since Birnberg and Shields 共1989兲 reviewed behavioral accounting research 共BAR兲, the area of applied behavioral research in general and BAR in particular has
burgeoned. The BAR literature has grown in breadth, depth, and complexity. This change
reflects an important trend in BAR: the reference disciplines and the object of accounting and
nonaccounting behavioral researchers have broadened.
The behavioral decision-making and cognitive psychology literatures that stimulated a significant portion of the emerging BAR research up to the late 1980s continue to have a significant
influence on BAR 共e.g., Camerer 2001兲. In addition, the role of behavioral research has grown in
other social science disciplines. Experimental economics has moved into the mainstream 共e.g.,
McCaffery and Slemrod 2006兲. This literature has had an impact on BAR 共Moser 1998; Callahan
et al. 2006兲. Legal researchers, heavily influenced by the writings of Kahneman and Tversky 共e.g.,
Kahneman and Tversky 1979兲, have begun to actively pursue behavioral issues 共see Sunstein
2000兲. A strong behavioral school even has developed within finance 共e.g., Thaler 1993; Barberis
and Thaler 2003兲. Medical researchers have joined with behavioral researchers to investigate
I
The author thanks the two reviewers for their insightful comments, the editor for the paper, Bryan Church, not only for all
his help, but also for his patience, and numerous colleagues for their help along the way.
A dagger 共†兲 at the end of select references indicates a review of the literature or a paper that includes an extensive set of
references.
Published Online: February 2011
1
2
Birnberg
issues such as how individuals react to prospective changes in the state of their health 共Udel et al.
2005兲. Even philosophy has developed a set of experimental researchers 共Knobe 2003; Appiah
2007兲 and a journal. Emerging methods for researching old questions are altering the form of
behavioral research, such as neuroeconomics 共Knudsen et al. 2007兲. These new tools permit
researchers to go beyond the observed behaviors of the decision makers and penetrate the “black
box”: that is, observe the brain’s activity during decision making. Finally, these new behavioral
researchers include economic modelers who have developed richer models of economic decision
makers 共“economic man”兲 intended to explain behaviors such as cooperation 共e.g., Rabin 1993,
1998兲, and empiricists who have utilized aggregated data to test these models 共e.g., La Porta et al.
1997; Ittner 2007兲.
The burgeoning of BAR and the expansion of disciplines that in one form or another have
added “behavioral” as an adjective to one of their sub-disciplines has enriched the extant research
on which BAR can draw 共e.g., Dickhaut et al. 2003; Hannan 2005兲. However, the increased
interest and diversity of methods used to research behavioral issues also leads to a blurring of the
definition of “behavioral research” in general and the boundaries of BAR in particular. What was
relatively clear 20 years ago is less clear today. The proliferation of research methods has meant
that BAR is more than laboratory experiments, surveys, and the occasional field study. A variety
of archival databases have been used to investigate essentially behavioral issues 共Banker et al.
2000b; Ittner 2007兲. Even efficient markets researchers, who would not be considered part of the
BAR community, are identifying and researching issues that clearly are intended to understand
individual investors’ behaviors: most notably, anomalous behavior relative to the predictions of the
efficient market 共Sloan 1996兲.
This blurring of boundaries between research thrusts has led to an often unrecognized degree
of commonality across BAR thrusts. While this has obvious potential benefits that will be discussed latter, it means the boundaries used in this paper necessarily are arbitrary and subjective. In
general, the questions studied and the papers cited will be related to the actual behavior of people,
whether it is as individuals or collectivities of varying degrees of size or complexity 共e.g., groups
or organizations兲, as they interact with each other and/or their environment. The test used in this
paper is analogous to one offered as an operational definition of obscenity: We know BAR when
we see it. At the margin different people will draw the line in different places. However, there is
little disagreement in the core of the research.
Given the growth in BAR, any attempt to provide a detailed review of BAR in general would
lead to a paper far beyond one this author could be expected to competently produce. Moreover,
recently a significant number of specialized reviews have been published offering the potentially
interested reader a wide variety of in-depth studies of BAR by both research topic 共e.g., auditing,
management accounting兲 and research method 共e.g., laboratory experiments, field research兲. These
reviews are cited in this paper where appropriate and review papers, or those with particularly
useful reviews of the literature, are identified in the reference section of this paper.
What would appear to be needed at this point in time is a framework within which the reader
can integrate the diverse studies making up BAR. To do this, I will present a framework that
focuses on the reference group of the studies, highlighting examples of research conducted in each
focal domain using different research methods and from different accounting sub-fields within
BAR. This approach not only is more parsimonious, but also permits the highlighting of a critical
facet of any research: complementarities of BAR across accounting sub-fields and methods. For
example, a paper dealing with audit teams may inform researchers interested in teams in management accounting, and a field study may provide a laboratory researcher with the insight needed to
design a better experiment.
Behavioral Research In Accounting
American Accounting Association
Volume 23, Number 1, 2011
A Proposed Framework for Behavioral Accounting Research
3
The paper consists of six sections. The first provides an overview of the paper and the
framework used. The second through fifth sections discuss each of the broad categories of studies
in the framework. The final section offers a brief summary of the paper.
ORGANIZATION AND SCOPE OF THE REVIEW
The approach used in this paper to categorize BAR is the behavioral unit that is the object of
the research. Does the research study the behavior of an individual, group, etc.? Organizing studies
in this manner highlights the similarities across otherwise diverse studies and is intended to
facilitate intellectual exchange among accounting researchers. To do this, I must necessarily restrict the depth of the review in any section to accommodate the desired breadth of coverage. The
framework is described in the next section. Like BAR, the boundaries between these categories at
times are subjective. For example, a paper may cover issues appropriate for understanding both
groups and organizations 共Anderson et al. 2002兲.
Framework
I have elected to view the extant BAR by what I have labeled its “focus.” I define focus as the
unit used to analyze the research question共s兲. The units range from the study of individuals to the
study of the environment that acts upon accounting or that accounting helps to shape. The four
categories used in this review were selected because they define distinct sets of research
questions.1 The categories include:
• individuals,
• small groups,
• organizations, and
• environmental conditions.
Because a study’s classification is determined by the set of individuals it considers in the
research question共s兲 and/or the analysis, the categories can be viewed as constituting a series of
concentric circles, with the innermost circles representing the more micro studies. The outer
“rings” represent more macro studies reflecting the broader focus of the research question共s兲. The
environmental conditions category can be interpreted as the “world” within which all other events
occur. Two important points should be noted. First, within the categories, particularly the individual category, there may be sub-categories. Second, studies from one category may inform
studies in another, likely adjacent category.
Definition and Discussion of the Categories
Individuals. These studies focus on the characteristics of a single actor and/or that actor’s
response to a particular accounting data set, accounting-related stimulus, or accounting-related
setting. It is by far the most active of the BAR categories discussed in this paper and can be
viewed as consisting of its own sub-categories. One line of individual research can be characterized by a concern with how individuals solve problems. I label these “pure choice” studies because
they focus on how well any actor can solve a problem without consideration being given to the
behavior of other actor共s兲. Recently, many of these studies have investigated the manner in which
the economic model 共“economic man”兲 in some significant way does not fit the behavior we
observe.
The second line of research explicitly considers the role of strategic behavior in the actor’s
1
This organization is similar to Hopwood’s 共1976, 5兲 Figure 1.1 describing the social context of accounting. He had four
categories: individual needs, group pressures and control, organizational structures and control strategies, and the social
economic environment. The organization used here differs from Hopwood’s by recognizing differences within the group
pressures and control categories between individuals and groups. This reflects changes in BAR over the decades.
Behavioral Research In Accounting
Volume 23, Number 1, 2011
American Accounting Association
4
Birnberg
decision. In these studies the actor explicitly 共should兲 consider the behavior of a second actor who
actually is present in the setting. These studies would include negotiation 共e.g., Fisher et al. 2000兲
or “cheap talk” 共e.g., Zhang 2008兲. I label these “strategic studies.”
Groups. Research classified as covering groups includes those studies where the relevant unit
of analysis consists of a small number of individuals. Typically, the members will be viewed by
the organization as affiliated 共i.e., as acting in concert in some significant way兲. Thus, what
differentiates group research from research studying participants individually or strategically interacting in dyads is the affiliation of the members. The actors are assumed to be in the same unit
at the time of the study. This would exclude studies such as those where the individuals are located
in different levels in a hierarchy. It is distinguished from research on organizations on two dimensions. One is pragmatic. Groups are small enough to permit the researcher to study the interaction
among the multiple participants. As the size of the group increases, researchers find it more
difficult to create and/or analyze the interactions 共process兲 and the focus of the research shifts from
the members of the group/organization to the organization itself. The other distinction is the focus
of the research. While group research is concerned with the activities of the group’s members,
organization research is concerned with the role of policy or the effect of characteristics of the
organization or its environment on the organization’s accounting policy or the organization as a
whole. This reflects a higher level of aggregation where the behavior of the individuals is lost. For
practical purposes the upper limit of group research usually is relatively small, typically four.
Organizations. As noted above, the focus of this research is on the characteristics of the unit.
The entity studied may be described by the legal boundaries of a firm or a division within a larger
entity. The research question often is the role played by structural characteristics such as task
complexity or the organization’s accounting system design. These studies move us farther away
from the characteristics of the individual discussed in the two previous categories. It identifies the
individuals/groups that compose the organization by the roles they occupy rather than by focusing
on the characteristics/actions of the individuals who occupy them.
Environmental conditions. These studies examine the role of accounting in society. Studies
included in this category reflect the interaction between accounting and society: that is, the broader
world of which accounting is a part. The interaction can take the form of the external forces that
shape accounting, as well as studies of the role accounting has played in shaping the world in
which we live. The former may be closely related to BAR studies in organizations. For example,
Prime Minister Margaret Thatcher’s intention to privatize British Rail affected the relative roles of
accounting and engineering within the organization 共Dent 1991兲, or the potential impact of the
whistleblower provisions of Sarbanes-Oxley 共Hunton and Rose 2010; DeZoort et al. 2008兲. How
the institution of standards for outputs led to the establishing of standard sizes for clothing 共Jeacle
2003a兲 is an example of how developments in accounting 共standard costs兲 can lead to changes in
the environment 共standard sizes兲.
INDIVIDUALS
The earliest BAR studies across all accounting areas were of this type and it continues to be
the dominant form of BAR. Shields 共2007兲 reported that 90 percent of the papers published in
BRIA from 2004 to 2007 studied the behavior of the individual. As noted earlier, studies of the
individual are of two types: individual choice studies and strategic studies. While the two share a
common core of issues such as the selection of participants and the research methods utilized, they
are significantly different in many other ways. Thus, this section of the paper is organized in a
slightly different manner than those discussing the other elements of the framework. The first
sub-section discusses issues common to both. The second sub-section discusses elements specific
to individual choice studies, and the third sub-section does the same for strategic choice studies.
Behavioral Research In Accounting
American Accounting Association
Volume 23, Number 1, 2011
A Proposed Framework for Behavioral Accounting Research
5
Common Issues
The two types of individual choice studies share many common features. These include the
research method selected and the choice of participants. Each of these is discussed below. The
section also discusses differences between the traditional economic model of self-interested behavior and recent findings in the areas of interpersonal utility, trust, and cooperation found in this
research.
Research Methods
The individual choice studies consist predominately of experiments, though some utilize
surveys 共Shields 2007兲. Experiments are particularly appropriate when the relevant dimensions of
the decision environment in which the decision maker interacts with the stimulus and makes the
decision are well known. Experiments have been used in BAR to examine a wide variety of
questions, including internal policies, external policies, tax reporting policies, incentive systems,
various types of resource allocation decisions, ethical issues, and various types of reports. The
responses measured have varied from objective outcomes such as investment decisions 共Libby and
Tan 1999兲 to more subjective perceptions such as fairness 共Evans et al. 2005兲 or trust 共Coletti et
al. 2005兲. Overall, studies of this type are the predominant form of research in BAR, particularly
North American BAR, and can be found across a wide variety of topics, accounting sub-areas, and
settings.
Individual choice studies also utilize surveys 共e.g., Chalos and Poon 2000; Clinton and
Hunton 2001兲 and archival data 共e.g., Banker et al. 2000a兲. Archival studies often reflect a naturally occurring experiment that permits the researcher to study behavior before and after the
change 共“stimulus”兲 has taken place.
Participants
A significant shift has taken place in the nature of the participants used in experimental
studies. Participants in the early studies most often were students 共undergraduate business majors
and/or M.B.A. students兲. BAR studies of the individual over the past two decades, however, have
required and utilized professionals as participants to a far greater degree. This is a significant
difference from the disciplines from which BAR draws its theories 共e.g., psychology兲, where the
generic participant remains the norm. This reflects the differences in the two groups’ reference
populations for external validity. The use of professionals as participants became necessary when
BAR shifted from its initial focus of “how participants respond while playing a particular role” to
“whether the skills accumulated by professionals insulate them from the negative effects of heuristics and biases when performing complex tasks” 共e.g., Libby and Trotman 1993; Kennedy
1993兲. Students cannot simulate that accumulated experience or professional knowledge, nor can
a mundane experimental task provide insight into the professional’s work.
The use of professional participants in BAR implicitly assumes that the professional’s behavior in an experimental setting accurately reflects their behavior “on the job.” Fehr and Leibbrandt
共2008兲 address this issue. They examine the cooperating behavior of fishermen both in a laboratory trust experiment and their level of cooperation to avoid over-fishing a given area. They find
that the participants’ behavior in the experimental setting accurately predicted their work behavior.
The broadening of the issues covered by BAR has expanded the type of professional participants required. The revival of interest in financial BAR now requires participants possessing
accounting expertise. BAR investigating proposed changes in the accounting rules requires
sophisticated/expert participants to test the validity of the hypotheses and enhance the study’s
external validity 共e.g., Hirst and Hopkins 1998兲. This also is true of BAR investigating anomalies
Behavioral Research In Accounting
Volume 23, Number 1, 2011
American Accounting Association
6
Birnberg
found in archival financial accounting research to generate BAR hypotheses 共e.g., Joe 2003兲,2 as
well as studies of the behavior of information providers in financial markets, such as security
analysts 共e.g., Libby and Tan 1999兲. However, many financial accounting-oriented BAR studies
continue to utilize M.B.A. students as surrogates for the “naïve investor” 共e.g., Tan and Tan 2009兲.
For a review of these studies and a discussion of the issues, see Libby et al. 共2002兲 and Koonce
and Mercer 共2005兲.
An exception to the use of professionals as participants is found in experiments in management accounting. What we have learned from use of auditors and investors as participants would
suggest that manager participants likely would exhibit many of the same cognitive biases as
student participants 共e.g., Kennedy 1993; Gilad and Kliger 2008兲. However, this comparability
may not carry over to activities such as budgeting behavior and negotiation. See Vance et al.
共2008兲 for an auditing example.
Some researchers utilizing student participants attempt to compensate for the participants’
lack of expertise by measuring participants’ task-specific knowledge 共e.g., so many courses in
accounting or years of work experience兲. They also use measures of the participants’ general
problem-solving ability, such as SAT or GMAT scores or responses to selected questions from
tests of that type 共e.g., Dearman and Shields 2005兲. For a nonaccounting study, see Burks et al.
共2008兲. These measures typically are used to identify potentially relevant differences among inexperienced participants 共i.e., students兲. However, Dearman and Shields 共2005兲 use their problemsolving ability measure as an independent variable to explain why some participants exhibit
nonfixated behavior while others did.
One topic related to the selection of participants in which BAR has shown less interest than
others of decision making-oriented research is gender differences. Non-BAR strongly suggests
that this may be an issue. These studies have reported significant gender-related differences in
areas such as risk taking 共e.g., Jacobsen et al. 2007; Huang and Kisgen 2008兲, competition 共e.g.,
Gupta et al. 2005兲, and negotiation behavior 共e.g., Bowles et al. 2007兲. All these areas can be
important in BAR.
Those BAR studies reporting the presence 共or absence兲 of gender-related differences in observed behavior have utilized these data in one of two ways. One uses the participant’s gender as
an independent variable 共e.g., Johnson et al. 1998兲. These studies investigate the conditions under
which the participant’s gender could affect behavior. If gender differences exist, randomization
may obscure their effect共s兲. Other studies check for gender differences to be sure that they do not
confound the experiment’s results 共e.g., Booker et al. 2007; Fleischman et al. 2007兲. Future BAR
may show greater awareness of the issue since SSRN in June 2009 established an ARN for
“Demographics, Gender, and Diversity Accounting Abstracts.”
Because of the limited research, it is an open question whether gender is as relevant an issue
when professional participants are used as it is in other studies. Do their professional training and
experiences override any gender issues? Two studies suggest that the differences may persist. Chin
and Chi 共2008兲, using archival data from Taiwanese audits, found that female auditors are more
risk-averse and more ethical in evaluating clients’ accruals. A survey of U.S., German, Italian, and
Thai fund managers 共Beckmann and Menkhoff 2008兲 found what they describe as the “expected
gender differences”: female respondents are more risk-averse and exhibit greater aversion to
competition.
Noneconomic Dimensions Affecting the Individual
In what could be labeled “post-modern” BAR, a line of research focuses on the appropriateness of two assumptions in the traditional economic model. One is that self-interest is the sole
2
In an interesting twist, Allee et al. 共2007兲 used archival financial accounting data to provide convergent validity for BAR
hypotheses.
Behavioral Research In Accounting
American Accounting Association
Volume 23, Number 1, 2011
A Proposed Framework for Behavioral Accounting Research
7
motivator of choice; the other is the use of monetary outcomes as the sole basis for measuring the
utility of an outcome. While it is possible to integrate these arguments into the utility function
共e.g., Birnberg and Snodgrass 1988; Luft 1997; Casadesus-Masanell 2004兲, BAR tends to view
these dimensions as if they are constraints on the individual’s wealth-maximizing behavior.
Typically, BAR studies of this type bring together literature from psychology and experimental economics. They stress that rather than behave in a self-interested manner, individuals conform
to certain social norms such as fairness, equity, trust, honesty, or a willingness to cooperate. For a
discussion of these issues, see Camerer 共2001兲, Rabin 共1993, 1998兲, Fehr and Gaechter 共2000兲,
Fehr and Schmidt 共1999兲, Moser 共1998兲, Evans et al. 共2001兲, Evans et al. 共2005兲, and Dawes and
Thaler 共1988兲. Another dimension related to fairness and equity but not explicitly discussed in
BAR is egalitarianism 共Dawes et al. 2007兲. Overall, these studies are important for BAR for two
reasons. First, they show how little it takes for the participants to exhibit non-self-interested
behavior. Second, they show the importance of the individual’s perception of equal/fair treatment
relative to his or her peers and how they respond to a lack of perceived equity/fairness. Trust is of
interest to behavioral researchers of all types 共Rousseau et al. 1998; Sapienza et al. 2007兲. In BAR,
Rose 共2007兲 examined how management’s financial reporting behavior affected the investors’
willingness to trust them. Evans et al. 共2001兲 focus on the individual in a management accounting
environment and show that individuals will behave honestly in a setting where their dishonest
behavior would not be detected, thereby violating the self-interest assumption. As a possible
explanation of this type of behavior, Rutledge and Karim 共1999兲 found that those participants who
did not exploit their asymmetric information in a principal-agent setting scored higher on ethical
development than those who did. Their research and many other papers suggest that non-totally
self-interested behavior is the norm or “default” behavior for many individuals and in many
settings, rather than the self-interested behavior postulated in traditional economic theory. A possible explanation for this behavior is their perception of whether they were treated fairly 共e.g.,
Greenberg 1990; Hannan 2005兲.
These findings can lead to interesting research on the individual’s response to their absence of
fairness. Remindful of Lucy van Pelt and Charlie Brown’s ongoing “relationship” over his kicking
the football, Bohnet and Zeckhauser 共2004兲 report that decision makers exhibit an aversion to
betrayal and take actions to avoid it. Wang 共2007兲 examines the symmetry between the punishment for dishonesty and the reward for honesty. She finds that honesty is rewarded more generously than dishonesty is punished. Issues of this type can be related to resource allocation in
managerial accounting and client behavior in auditing. In both cases, the research question would
involve identifying which behaviors lead to trust 共or distrust兲 between the parties. What causes an
auditor to trust one client more than another? What causes a superior 共manager or auditor兲 to trust
a particular subordinate?
Any trust-oriented research raises 共at least兲 two questions related to experimental design. One
is the importance of the experiment’s context 共degree of realism兲 and the choice of participants
共students or professionals兲 used in the study. The other is the importance of the presence or
absence of the interaction with a real person when the participant is told of the existence of
another participant. The latter issue is discussed under strategic choice situations.
Culture and Its Impact on Decision Makers
BAR studies dealing with social norms and potentially differing values across cultures ask
whether differences in culture result in different decisions/behaviors. For the most part, these
studies have utilized the framework of Hofstede 共1980兲. However, it is important to be aware that
some issues have been raised about the appropriateness of his categories 共e.g., Baskerville 2003;
McSweeney 2002兲. Because of the readily apparent cultural differences, the greatest portion of
this research has compared Asian and North American workers 共e.g., Birnberg and Snodgrass
Behavioral Research In Accounting
Volume 23, Number 1, 2011
American Accounting Association
8
Birnberg
1988; Chow et al. 1999兲. Thus far, the studies are inconclusive. While some of the studies have
found differences consistent with their predictions 共e.g., Kachelmeier and Shehata 1997兲, others
have not 共Birnberg et al. 2008兲. In an interesting archival financial accounting study related to
BAR, Doupnik 共2008兲 finds inter-country differences in earnings management after allowing for
differences for legal regimes.
The potential role of national cultures is becoming more important as BAR internationalizes
and research findings reported by researchers from many different countries appear in journals and
SSRN. This raises the following question. Are research findings from one country universally
applicable or should we be concerned and replicate them before we accept their universality?
As management systems and styles “internationalize” in large, industrialized economies, it
may mitigate concerns over cross-cultural differences. However, this homogeneity may not be
present in small-scale economies. In contrast to results reported in some BAR, Henrich and the
Cross Cultural Ultimatum Game Research Group conducted an extensive study across 15 smallscale economies. Their study is important because they examine behavior among economies
where the variation in economic development is far greater than those typically studied by BAR.
Using the dictator game and a social dilemma game, as well as the ultimatum game, they report
that the “textbook economic model” failed to predict the observed behavior. Their results are
reported in various forms 共Henrich et al. 2005, 2001兲, as well as in Henrich’s 共2007兲 plenary
address at the AAA’s 2007 annual meeting. They conclude behavior in the experiments is generally consistent with economic patterns of everyday life in these societies. Henrich et al. 共2001,
73–74兲 report that, “The higher the degree of market integration 关in their society兴 and the higher
the payoffs to cooperation 关in their society兴, the greater the level of cooperation in experimental
games.”
Summary
While the methods used to study individual behavior have not changed significantly since
Birnberg and Shields 共1989兲, BAR has paralleled the trend found in experimental economics. A
significant portion of BAR now focuses on factors that influence decision makers in directions at
odds with the self-interest and wealth-maximizing assumptions. These noneconomic dimensions
include trusting behavior, cooperation, and the expectation of a fair share of any rewards. In
certain settings this can lead to greater monetary returns to the decision maker. However, they also
can expose the decision maker to greater risk. Other characteristics of the “work environment,”
such as the national/local culture, also can affect the expectations and behavior of the decision
maker. It has been suggested that certain of the cultural differences observed in individuals may be
based on different market conditions among countries.
Individual Choice Studies
There are a variety of reasons for the popularity of individual-focused research in BAR. The
first is simplicity. Considering the individual investor, auditor, etc., in isolation lends simplicity to
both the study’s research model and its design. It also simplifies the analysis and interpretation of
the results. The second is parsimony. It takes the fewest number of participants to achieve the
desired number of observations per cell. This is especially important when the participants are
professionals. The third reflects the models generated in the disciplines on which BAR has drawn
most heavily 共economics and psychology兲. Both contain a significant literature relating to how the
individual makes a decision. Sociology and organization theory consider the “group” to be the
smallest unit and have been drawn on by BAR to a significantly lesser extent.
Individual choice studies in BAR can be divided into two types, depending on the type of
variable investigated. One group of studies is interested in better understanding the impact of
Behavioral Research In Accounting
American Accounting Association
Volume 23, Number 1, 2011
A Proposed Framework for Behavioral Accounting Research
9
elements of the setting within which the individual acts on the individual. The other is concerned
with the appropriateness of rational wealth-maximizing characterization of the decision maker.
Factors Related to the Task Setting
Four elements of the task setting are of particular interest in individual BAR. These are
incentives, participation, accountability, and systems interface. The first two are the focus of a
significant portion of BAR; the latter two, much less.
Incentives. Chow 共1983兲 initiated experimental research on the role of incentives in BAR.
This line of BAR literature typically uses the principal-agent model to generate hypotheses. For a
survey of the economic models of incentives, see Prendergast 共1999兲. In general, the studies report
that incentives matter and the nature of the incentive system impacts an agent’s behavior 共e.g.,
Bonner et al. 2000; Towry 2003; Sprinkle et al. 2008兲.
Participation. Participation is, essentially, concerned with the honesty of communication
within the organizational hierarchy. Early BAR investigated how accurately the workers/agents
would communicate their private information. Would they use it to create slack? Generally, the
answer was yes 共e.g., Young 1985; Shields and Shields 1988兲.3 However, as discussed subsequently, later research recognized the strategic nature of the interaction between the subordinate
and the superior and modeled participation as a negotiation process.
Accountability. Given the function of accounting, it is surprising that the formal development of accountability was in psychology 共see Lerner et al. 1998 for a review兲 despite the obvious
link to management accounting research; that is, the effect of evaluation on individual behavior
共e.g., Argyris 1952; Prakash and Rappaport 1977兲. The notion of evaluation in BAR is not limited
to management accounting. When the superior in an audit team examines the work of a subordinate or a client examines the work of a tax professional, an “evaluation” is taking place. The
difference between the evaluation literature and BAR on accountability is reflected in the breadth
of the questions they ask. The evaluation literature focuses on how the accounting system 共e.g., the
performance indicator兲 affects the extent and direction of the effort provided by the “workers”
共Prakash and Rappaport 1977兲. Accountability BAR not only asks for what the worker feels
accountable, but also asks to whom the “worker” feels accountable when facing conflicting demands 共e.g., Johnson and Kaplan 1991; Messier and Quilliam 1992兲, or how elements present in
the accountability setting 共e.g., a need to justify one’s actions兲 affect the worker’s behavior 共Ahrens 1996兲.
Miller et al. 共2006兲 recognized that there is an element of mutual accountability in the evaluation process. The superior likely has a prior relationship with the subordinate and in many
instances must “justify” any evaluation he/she makes. Their study focuses on the reviewer in an
audit setting. While the study only examines one party to the dyad, their findings suggest that
factors such as familiarity between the two parties can affect the reviewer’s assessment. There
may be limitations on the ability to perform these experiments with professional participants in
dyads because of the potential impact on the participants’ post-experimental relations.
Systems interface. Information systems in BAR essentially are viewed as decision aids. They
are discussed under various labels, such as decision support systems 共DSS兲 and knowledge based
systems 共KBS兲. The DSS typically is used in the management information systems literature to
describe an information system intended to support a specific decision and is closest to the term
decision aid 共DA兲, which typically is used in auditing to describe what may or may not be a
computerized calculating system. In contrast, the KBS refers to a database collected for a specific
3
Those familiar with the dictator game discussed below will recognize that Young’s 共1985兲 task is essentially the use of
a dictator game to simulate participation.
Behavioral Research In Accounting
Volume 23, Number 1, 2011
American Accounting Association
10
Birnberg
area of inquiry 共e.g., XBRL兲.
The simpler of the two is the DSS. Two broad questions are researched under DSS. How well
are the systems utilized by those for whom they are intended? And, what characteristics of the
DSS facilitate or inhibit their utilization? Specific issues researched under the former include not
only whether the DSS improves decisions, but whether the potential users utilize them and
whether the system can be used to facilitate learning. They differ from the individual choice BAR
studies discussed earlier 共i.e., that examined how the individual responds to specific outputs of the
system兲. Those studies typically are linked to cognitive issues and the use of accounting data 共e.g.,
Lipe and Salterio 2000; Dearman and Shields 2005兲. The papers discussed in this section are
concerned with the utilization of a DSS as a DA designed to assist an individual perform a specific
task. In general, they report that the DSS is not always utilized 共e.g., Whitecotton 1996; Eining et
al. 1997兲.
Whitecotton 共1996兲 found that auditors’ reliance on the DA was inversely related to their
confidence in their own judgment. Obviously, this raises two questions. Is the auditor’s confidence
appropriate? And, how do those using the DA perform relative to the best auditors? Rose and
Wolfe 共2000兲 shed some light on the second question. Using student participants and a tax calculation task, they report participants who performed the calculation using “pencil and paper” rather
than the DA outperformed the best DA-assisted group by 22 percent, but required 112 percent
more effort 共Rose and Wolfe 2000, 297; also see Glover et al. 1997; Borthick et al. 2006兲. It is
important to learn whether the results can be replicated with professionals because it is likely that
their judgment is superior to that of the students.
Arnold et al. 共2006兲 studied the type of data from the KBS used by 共relative兲 novices 共senior/
staff auditors兲 and 共relative兲 experts 共partner/manager兲. The two groups differed on several dimensions. Novices chose feedforward explanations, while the experts chose feedback. Arnold et al.
共2006兲 report that the greater the experts’ reliance on feedback explanations from the KBS, the
greater their adherence to the KBS’ recommendation.
There also are interactive systems intended to facilitate access to larger databases. These DSS
are intended to improve the quality of decision making or assist in training. The issues considered
revolve around the usefulness of the database. In BAR, the issue typically can be framed in terms
of the behavioral characteristics of the user and the usefulness to the user of the DSS. The XBRL
is an example of such a system. It is intended to enhance the user’s ability to obtain and understand financial data about the firm. Hodge et al. 共2004兲 found that nonprofessional users of
financial statements were better able to ascertain the impact of differing reporting methods for
stock options between firms using the XBRL than without it. However, like Rose and Wolfe
共2000兲, they reported that many of their participants did not utilize XBRL. Other BAR has as its
purpose examining the use of DSS as a tool for training/educating novices.
Alternative modes of communicating information, such as graphs, frequently are used in
reports. For example, nonnumerical formats are regularly used in corporations’ annual reports,
internal reports, and our research. This issue initially was asked by MIS researchers in the 1970s
共Dickson et al. 1977兲 and subsequently extended 共e.g., Vessey 1994兲. Despite the extensive use of
pie charts and graphs in internal and external reports, there is little research in BAR on this topic
共for an exception, see Amer 2005兲. In marketing, MacKay and Villarreal 共2007兲 found that the
recipient’s ability to take advantage of the simpler nature of nonnumerical data is likely to vary
among individuals. An interesting example of earlier research in this area, using faces to communicate financial data, was reported by Moriarity 共1979兲.
Noneconomic Dimensions Affecting the Individual
The above dimensions of the task are essentially elements of the task setting in which the
individual makes a decision. They typically are set by the organization or environment within
Behavioral Research In Accounting
American Accounting Association
Volume 23, Number 1, 2011
A Proposed Framework for Behavioral Accounting Research
11
which the decision maker is operating. The decision maker also brings certain characteristics such
as trust and fairness to the setting. These characteristics may be 共relatively兲 stable for any decision
maker 共e.g., desire to be treated fairly兲, or they may vary with the situation 共e.g., the decision
maker’s mood兲. In this section, these characteristics as they relate to individual choice are discussed.
Ethics. Closely related to the study of norms is the study of ethical behavior. The former
often is researched in the context of what others expect the actor to do, while ethical behavior
typically refers to the actor’s behavior. Noreen 共1988兲 offers a theoretical link between ethics and
agency theory. He argues that parties to the contract could be expected to follow social norms.
Early BAR on ethics focused on the participants’ moral development 共e.g., Ponemon 1990兲. These
studies are concerned with two issues. How developed is the moral reasoning of particular
individuals/groups? And, how does a given level of ethical development affect participants’ onthe-job behavior? These two questions can easily be adapted for BAR in any of the accounting
sub-areas. The broader issue is how significant the ethical issue is in that sub-area. Auditing
researchers have led the way in considering the role of ethics in BAR. For reviews, see Louwers
et al. 共1997兲 and Jones et al. 共2003兲.
Like the cross-culture research described earlier, the ethics-based research has been characterized by issues over how to measure the level of ethical development/behavior of the participants. This is not surprising since, like culture, the level of an individual’s ethical development is
not observable 共as distinct from actions兲. For a discussion of the different approaches, see Cohen
et al. 共1996兲.
In a post-Enron world, BAR in both auditing and management may find the issue of increased
importance. The problem facing the researcher is likely to be one of access. To minimize the
degree of intrusiveness and obtain responses, this research typically relies on surveys or cases to
elicit responses. There also appears to be a reluctance to publish these papers in the mainstream
accounting journals. A significant number of BAR studies have been published in The Journal of
Business Ethics 共e.g., Arnold et al. 2007; Emerson et al. 2007兲.
Two tax-oriented ethics studies suggest possible studies for management accounting behavioral researchers. Fleischman et al. 共2007兲 demonstrate the linkage across the various aspects of
individual-focused research. The paper examines the evaluation by managers in a case concerning
the ethical behavior of a spouse in the context of a tax setting 共innocent spouse rule兲. The paper
explores the potential existence of the innocent spouse rule as a norm and the extent to which
research in ethics by behavioral scientists can explain it. Similar studies might be conducted in
management accounting. They could relate the participant’s response to the firing of an innocent
manager and, for example, the participant’s predicted subsequent job behavior. This behavior
relates to the issue of perceived fairness discussed earlier. In the area of financial accounting, Rose
共2007兲 related how what could be labeled 共un兲ethical reporting by management leads to 共dis兲trust
on the part of investors.
Cruz et al. 共2000兲 report that tax professionals’ willingness to resist the client’s desire for
aggressive tax reporting is positively correlated with professionals’ score on the Multidimensional
Ethics Scale. This raises the question of how a subordinate might respond to a superior’s efforts
for a more favorable set of budget estimates. Would a measure of ethical development predict the
likelihood of cooperation? In an experiment in financial reporting, Vance et al. 共2008兲 hypothesized and found that the better the superior-subordinate relationship, the less likely the subordinate was to resist the superior’s request for aggressive financial reporting.
Two sets of BAR studies have extended early BAR on ethics in interesting ways. They
examine the impact of the individual’s environment on the individual’s ethical behavior. Booth
and Schulz 共2004兲 examine the impact of the organization’s ethical climate on the individual’s
behavior. In a laboratory study, they find that holding the participant’s level of ethical development
Behavioral Research In Accounting
Volume 23, Number 1, 2011
American Accounting Association
12
Birnberg
constant, the behavior of the participant moves in the direction of the organization’s ethical
climate. There is no reason to believe that similar results would not be found in the effect of the
permissiveness of audit firms on auditor behavior.
Spicer et al. 共2004兲 and Bailey and Spicer 共2007兲 linked cross-cultural research and ethics.
Earlier studies had reported ethical differences among auditors in different countries 共e.g., Patel et
al. 2003; Arnold et al. 2007兲. Spicer et al. 共2004兲 and Bailey and Spicer 共2007兲 researched the
ethical norms of a culture on individuals raised in a different culture. In their studies, they studied
U.S. expatriates in Russia involved in the Russian business community. They report convergence
in ethical attitudes and intended behaviors between U.S. expatriate and Russian respondents to
their ethics survey. The U.S. expatriates in their study responded more like their Russian counterparts than U.S. nationals in the U.S. The respondents also expressed similar attitudes toward
organizational practices that violated the ethical standards or “hyper-norms.” The U.S. expatriate
respondents who were highly integrated into the Russian community expressed ethical attitudes
similar to those of Russian respondents under conditions of “local 共Russian兲 norms.” In both cases,
the ethical attitudes of Russians and Americans converge despite the differences that might have
been expected to arise due to their respective national identities.
Mood. Recently, psychologists, experimental economists, and accountants have begun to
examine the role of the decision maker’s emotional state 共affect兲 on the decision process. These
studies could be important if different mood states affect the decision maker’s perceptions and
decisions. While mood could affect strategic interactions, the research undertaken in BAR thus far
has focused on the individual decision maker.
The rationale underlying studies of this type is that mood affects the nature of the prior
experiences retrieved from memory. Positive mood states lead to retrieving positive outcomes in
comparable situations and vice versa. Wright and Bower 共1992兲, in a BAR-related study, reported
the effect of decision makers’ emotional state 共happy, neutral, or sad兲 on their perception of the
degree of riskiness of a decision and probability of success. As they conjectured, the subjective
probability estimate is influenced by the decision maker’s mood. “Happy” decision makers give
higher probabilities for the outcome of positive events and lower probabilities for the outcome of
negative events. They report the opposite results for “sad” decision makers.
In an accounting context, Moreno et al. 共2002兲 and Kida et al. 共2001兲 report similar results.
Consistent with these results, Chung et al. 共2007兲 studied auditors making inventory valuation
decisions and find that mood state affects the degree of conservatism in the auditor’s inventory
valuation. Auditors in a positive mood are less conservative than those in a negative mood.
Moreno and Bhattacharjee 共2008兲, in a single-party study 共the other party did not actually exist兲,
report that knowledge of the other party’s emotional state affects bargaining behavior. For a
discussion of the literature arguing that emotion can enhance the individual’s ability to make
rational choices, see Ackert et al. 共2003兲.
Psychologists and experimental economists have studied other emotional states that could be
of interest to accountants. Lerner and Keltner 共2000, 2001兲 report that fearful participants make
more pessimistic estimates and more risk-averse choices, while anger leads participants to make
more optimistic risk estimates and risk-seeking choices. Interestingly, the responses of angry
participants more closely resembled those of happy participants than those of fearful participants.
For reviews, see Lerner et al. 共2004兲 and Pham 共2007兲.
An interesting issue raised by these studies is whether the effect of these emotions is to make
people overly optimistic/pessimistic. We cannot conclude one way or the other without having
some baseline measure of the probability. What should the individuals believe the probability to
be? Since the participants disagree, we can assume that their emotional state has led at least one
of the groups to be incorrect, but that does not preclude the possibility that they both may be in
error. Ideally, further research will be undertaken in this area where there is a known “correct”
Behavioral Research In Accounting
American Accounting Association
Volume 23, Number 1, 2011
A Proposed Framework for Behavioral Accounting Research
13
answer. A topic that conceivably could be related to the issue of optimism/pessimism is the effect
of regret in decision making. It has been shown to have an impact in many nonbusiness decision
settings 共e.g., Gilbert et al. 2004兲.
A paper by Libby et al. 共2008兲 suggests that optimism/pessimism is not always the “irrational”
result of the decision maker’s emotional state. They report that in some circumstances optimism/
pessimism may be the result of the incentives. If analysts desire good relations with management,
they report that, all else being held constant, the optimism/pessimism of sell-side analysts is a
deliberate act and not based on an emotion or trait.
Two recent studies suggest the possibility of yet another emotion that could be affecting
worker behavior—guilt/guilt aversion. These studies also illustrate how labels potentially can
serve to separate like ideas. Schnedler and Vadovic 共2007兲 hypothesize and find that guilt aversion
motivated participants to exert effort beyond the minimum required by the control system. One
might conjecture that this merely renames the concept embodied in “gift exchange” 共e.g., Hannan
2005兲. Staffiero 共2006兲 used guilt to describe the behavior of individual members of Japanese
work groups. The workers felt guilt when they made insufficient contributions to their work group.
In contrast, Birnberg and Snodgrass 共1988兲 offer a more positive explanation of this behavior,
suggesting that the outcomes to other members of the group may have a positive utility to an
individual member. Failure to achieve the group’s goal results in lowered utility because of the
loss to others as well as to oneself.
Fairness. While the perception of fairness has primarily been researched in strategic settings,
the perceived fairness of the accounting system affects the behavior of the individual in individual
choice settings as well. Libby 共2001兲 and Hufnagel and Birnberg 共1994兲 found that the participants
were sensitive to the perceived unfairness of the accounting system 共procedural fairness兲 even
when they were not adversely affected by the rule or system.
Physiological Measures and BAR
Behavioral accounting researchers have tried a variety of methods to understand decision
processes. The methods utilized are relatively non-intrusive, but provide greater insight than
observing an outcome/response in an experimental setting. These approaches include think-aloud
protocols 共e.g., Bedard and Biggs 1991兲 and data boards 共e.g., Shields 1980兲. These approaches
yielded insights into “cognitive flow” or the decision process being followed. However, both of
these methods directly involve the participant and are limited to reporting the decision maker’s
conscious behavior. The methods discussed in this section measure the same behaviors discussed
earlier, but use methods intended to measure physiological changes.
Hunton and McEwen 共1997兲 utilized an eye movement retinal imaging computer to study the
information search strategy of financial analysts. Unlike protocol analysis that relies on selfreporting and data boards that report only choices, they were able to track the search strategies of
the analysts in a less obtrusive but more detailed manner. They were able to observe data scanned
but not reported 共protocols兲 or chosen 共data boards兲 by the participants. Consistent with data board
research, they found that the more accurate analysts used a directed rather than a sequential search
strategy. Their search appeared to be motivated by hypotheses generated by the process.4
In finance, Lo and Repin 共2002兲 used more traditional methods 共electro-dermal and pulse rate
measures兲 to measure the emotional state 共level of excitement兲 of ten stock traders while they were
actually trading. Lo and Repin 共2002兲 found significant differences between periods when significant market events were and were not taking place. They argue this suggests that emotion is a
4
For a discussion of the use of eye movements in marketing research where they have been used more often, see Zaltman
共1997兲.
Behavioral Research In Accounting
Volume 23, Number 1, 2011
American Accounting Association
14
Birnberg
relevant component of the traders’ decisions. Their data suggest that the response varies with
experience, but the sample is too small to draw any statistically significant conclusions.
Neuroeconomics and Neuroaccounting
Recently researchers studying decision making have taken a new approach. Working with
neuroscientists, they have gone one step deeper inside the “black box” that is the decision maker.
Using various devices, they observe the patterns of brain activation as individuals make choices
共e.g., McCabe et al. 2001; Camerer et al. 2005; Knudsen et al. 2007兲. Given the neuroscientists’
knowledge of the function of the brain centers, conclusions can be drawn about what underlies the
observed behavior. By moving one step closer to the decision maker’s cognitive activity, the role
of the stimulus and the response changes in an interesting way. The decision, typically considered
the response in BAR studies, now is the stimulus and the brain center activation is the response.
This is in contrast to traditional research in BAR where researchers observed behavior and inferred
the underlying cognitive processes or extracted them from protocols.
Thus far, little research of this type has been undertaken by behavioral accounting researchers
except for John Dickhaut 共e.g., Dickhaut et al. 2003; Smith and Dickhaut 2005; Rustichini et al.
2005; Dickhaut 2009兲. Dickhaut and his colleagues have papers 共Dickhaut 2009; Dickhaut et al.
2009a, 2009b兲 using neuroscience to study the evolution of recordkeeping 共i.e., accounting兲.
However, none of these papers provide the type of systematic review of the possible link between
neuroscience and BAR that can be found for finance in Sapra and Zak 共2008兲, who offer neuroscience explanations for observed behaviors in financial decision making where data from neuroscience and neuroeconomics are available.
While potentially quite insightful, there are at least three reasons why research of this type
will progress more slowly than other types of BAR. First, it requires cooperation with a researcher
possessing access to machines to perform the scans and skilled in reading brain scans. Second, it
would appear that research of this type is quite expensive. Third, explaining the findings to other
BAR researchers may be difficult. Moreover, the results may not eliminate the issue of “hardwired” versus “learned” behavior as the explanation for the response.
An example of neuroeconomic research’s potential relevance to BAR can be illustrated using
the findings of Luft 共1994兲 and Hannan et al. 共2005兲. Luft 共1994兲 found that participants in her
study preferred a bonus to a penalty pay scheme even though the payoffs from the two systems
were equivalent. Hannan et al. 共2005兲 found that the participants in the penalty condition exerted
more effort. Given that neuroscientists have shown that different brain centers are used to measure
pleasure 共reward兲 and pain 共penalty兲 共Delgado et al. 2000兲, this raises the question of whether the
preference for a bonus scheme reflects differences between the pleasure and pain brain centers
共“hardwired” neuroscience explanation兲 or whether it is the approval implied by the “reward” and
disapproval associated with a “penalty” 共a social psychology issue of intrinsic reward兲. Barnea et
al. 共2009兲, using Swedish data on twins to study investing behavior, suggest that there is both a
genetic and a learned component.
A series of neuroscience studies may provide some insight into what is happening. Using the
ultimatum game, Tabibnia et al. 共2008兲 report MRIs of the brain that suggest similar results to
those above for fair and unfair behavior. Their design utilized an individual choice study using
only participants who receive the “offer” 共ultimatum兲. First, the results suggest that the 共recipient兲
participants differ in what they believe to be a fair offer. Second, those who judge the offer to be
“unfair” show different patterns of brain activity than those who consider the offer to be “fair.”
Finally, participants who accept an unfair offer had different patterns in their MRIs than those who
reject unfair offers 共Tabibnia et al. 2008兲.
A study by Harbaugh et al. 共2007兲 that relates brain activity to altruism in decision makers
also illustrates the potential link of neuroscience to BAR. They studied the brain scans of 19
Behavioral Research In Accounting
American Accounting Association
Volume 23, Number 1, 2011
A Proposed Framework for Behavioral Accounting Research
15
female students who were asked to make a decision allocating $100 between a food bank and
themselves. The brain scans of the “altruistic” 共gave more兲 and “selfish” 共gave less兲 participants
show that the altruistic participants exhibit greater activity in the part of the brain that reflects
pleasure than do the selfish participants. The altruistic participants show significant activity in that
part of the brain even when they were required to contribute a fixed portion of the $100 to the
charity. Studies of this type suggest that there is a physiological basis for the altruistic behavior
that is observed in the real world. It does not explain if the behavior is inherent and hardwired
共Hsu et al. 2008兲 or related to interacting with people and learned 共Andreoni 1990兲. The authors
suggest that they believe their results also would apply to male participants had they been included
in the study.
Zak and his colleagues introduced a line of neuroeconomic research that approaches the
“black box” of human cognitive processes in a different way. They argued that the observed
behavior, in this case trust, is based on the brain’s response to a particular hormone. Trusting
participants exhibit higher levels of the relevant hormone than nontrusting 共i.e., economically
rational兲 participants. This work is summarized in Zak 共2008兲. Kuhnen and Chiao 共2009兲 show
that there also appears to be a genetic basis for the differences in the amount of dopamine and
serotonin. In their study these differences, like those reported by Zak 共2008兲, are associated with
different patterns of behavior.
Summary: Individual Choice Studies
Overall, research focused on the individual’s decision-making behavior has played an important role in BAR historically. The predominance of individual-focused research, particularly
among North American and many Australian researchers, is easily observed by examining a recent
issue of BRIA 共2007兲. It contained 13 papers. All of these papers could be classified as focused on
the individual even though they may describe in the scenario the existence of another/other
hypothetical person共s兲 or have a scripted confederate role-play the “other person.” Equally important is the diversity in topics/areas in which the research is located. Three were related to auditing.
Four dealt with aspects of management accounting. Three were related to financial reporting/
decision making. There was one in tax ethics, one in cross-cultural ethics, and one related to
education. While this admittedly is a convenience sample, the results are similar to Shields 共2007兲.
They likely are representative of current BAR in North America. A very different view of BAR in
Europe would result from examining an issue共s兲 of AOS or other European-based accounting
journals.
This emphasis on individual-focused research is likely to continue to be true of BAR in North
America for several reasons. Many BAR questions focus on the behavior of individuals acting
alone. For example, some of the studies involve one individual’s processing data provided by
another individual or a system 共e.g., Fedor and Ramsey 2007兲. Others continue to be concerned
with the cognitive processes of individuals 共e.g., Joe 2003兲. Still others involve norms, ethics, and
culture, which typically have been studied by examining the behavior of the individual in isolation. Finally, the individual also may be the easiest approach for researchers.
Individual choice studies do not exist in isolation from the other categories of BAR discussed
in this paper. As the research on strategic choice and group-focused behavior shows, understanding the behavior of individuals often is the basis for hypotheses about behavior in dyads and
groups. Behavior such as honesty 共Evans et al. 2001; Cohen et al. 2007兲, that has been exhibited
in studies in which the individual does not actually interact with another participant, can lead to
predictions of behavior in dyads and groups that differ from those of classical economics. This is
particularly true because many of the individually focused studies are studies isolating one member of a network of individuals. This is readily apparent in the next section in the discussion of
participation.
Behavioral Research In Accounting
Volume 23, Number 1, 2011
American Accounting Association
16
Birnberg
There also are limitations in studying the individual in isolation. In part, this results from the
movement in organizations to make groups and teams the decision-making unit. In addition, a
certain amount of the richness found in the decision-making situation may be lost when BAR
isolates the individual from his or her environment.
Strategic Choice Studies
Studies that explicitly consider the participants’ strategic behavior are relatively new in BAR,
though strategic behavior often was implicit and important in earlier BAR. How managers behave
in a participative management setting is an example of a strategic setting. Moving from an
individual choice study where the actor’s behavior is “inward facing” to one where another actor’s
behavior explicitly must be considered introduces the strategic dimension to BAR. In contrast to
the individual choice studies, in the strategic behavior studies the decision maker must consider
the choices made 共or to be made兲 by an actual rather than a hypothetical fellow participant. For
example, in a management accounting study, the “strategy” to which the participant responds
could be the choice of budget level set by another participant acting as “management.” While an
individual choice study informs us how the manager/agent responds to a given budget level, we do
not learn which budget level the owner/principal would choose to offer to motivate the manager/
agent. In an individual choice study, the researcher may set the independent variable 共e.g., the
budget兲 at levels different from those a manager actually would choose.
A significant amount of experimental economics research uses experimental dyads 共see Roth
1995兲. In BAR, strategic choice studies recognize the limitations in studying the individual in
isolation from the environment and the importance in many settings of the behavior of the “other”
party on the individual. Some argue that it is important actually to have the “other party” exist
whenever the instructions indicate he/she does. Experimental economists argue that it is required
for one of two reasons. The first is “maintaining the integrity of the participant pool.” Experimental economists often utilize the same pool of participants in different studies. In some studies, the
participant’s experience in a prior study even is a criterion for selection. They argue it is important
the participants believe what they are told. If the post experimental debriefing informs them that
something was not really the case, they may speculate in future studies about the true nature of the
study. The other reason relates to the richness of the experimental setting. Unless the experimenter
has insight into how the other party will behave from prior field or laboratory research, including
the actual behavior of a participant will increase both the potential insights from and the validity
of the study. See Calegari et al. 共1998兲 for an example of this issue.5
Negotiation Studies
The negotiation process is ubiquitous in the business setting. For a review, see Tsay and
Bazerman 共2009兲. Audit firms negotiate with clients over changes in financial statements and
accounting methods 共McCracken et al. 2010兲, firms negotiate with suppliers when they establish
operationally intimate relationships 共JIT兲, and sub-units within the organization negotiate transfer
prices and/or quantities. While the surface characteristics of the situations are different, many of
the behaviors may be the same 共e.g., the strategies adopted by the parties兲. They may differ on
information asymmetry, division of payoffs, and relative power. The degree of information asymmetry would be expected to affect negotiation, as could the incentives of the parties. For example,
in budgeting negotiations the parties typically are playing a zero-sum game. The slack absorbed by
the worker reduces the manager’s/principal’s profit by a like amount. In other cases, such as the
5
For a discussion of this literature from an auditing perspective but germane to all BAR, see Hooks and Schultz 共1996兲
and the symposium in Auditing 共e.g., Dopuch 1992 and Gibbins 1992兲. For the contrary view from psychology, see
Kelman 共1967兲.
Behavioral Research In Accounting
American Accounting Association
Volume 23, Number 1, 2011
A Proposed Framework for Behavioral Accounting Research
17
audit or transfer price settings, the negotiation game being played need not be a zero-sum game.
Rather, a small concession by one party may be significant to the other. Such an asymmetry in
payoffs should affect the negotiation process. Negotiation studies also can be characterized based
on the relative power of the participants: those where the parties have equal power and those
where one party has an advantage.
The significance of the strategic interaction is of particular importance for BAR because of
the importance of performance as a “response.” An example of how individual choice literature
and strategic choice settings are related can be found in Fisher et al.’s 共2000兲 study of participation
utilizing interacting dyads. In the framework utilized in this paper, this represents a paradigm shift.
Early BAR into participative budgeting focused on how the “worker” would behave. Would the
workers take advantage of their private information to create slack? Young 共1985兲 even had his
participants meet with a “supervisor” played by the experimenter or a colleague. However, the
“supervisor” did nothing more than accept the worker-participant’s budget. Thus, Young’s 共1985兲
study essentially is an individual choice study. While social pressure was present 共the design
forced the worker-participant to face a supervisor兲, it omitted any negotiation over the acceptability of the worker’s proposed budget. The explicit power in the situation was vested with the
worker. In reality, the budget-setting process is quite different. In the natural setting, the supervisor
also has significant power. Thus, while Young 共1985兲 reported how the worker would act in
isolation, important aspects of participation are better captured as a dyad that permits strategic
interaction.
A second area of negotiation studies where the use of dyads is present is in the transfer price
literature. Like the participation studies, they are outcome-oriented. In an early study, DeJong et
al. 共1989兲 test the efficacy of various transfer pricing rules. Haka et al. 共2000兲 vary the precision
of the accounting data the manager possesses. The participants receiving the less precise information negotiated strategically. They tried to achieve the best price at the risk of failing to reach an
agreement. In contrast, the participants with more precise data used the negotiation process to
communicate information to the other party about his or her position in an attempt to reach a more
informed decision. Chalos and Haka 共1990兲 and Ghosh 共2000兲 also studied the negotiation process
in the transfer price setting in laboratory experiments. Ghosh 共2000兲 observed that when the
incentive system is consistent with the sourcing of the input, the systems are perceived as fairer
and the participants behaved in a less exploitive manner. Also see Luft and Libby 共1997兲.
How humans negotiate and what motivates them to behave in a particular way is a question
of interest to all BAR. Findings in one area have implications for the others. Calegari et al. 共1998兲
report two interesting results concerning dyads using an auditing-based task. One relates to the
outcome of the negotiation process, the other to method. In their study, M.B.A. students, participating in the experiments as “auditors” and “clients,” exhibited two types of behavior: competitive
pairs and cooperative pairs. The competitive pairs behave as Calegari et al.’s 共1998兲 economicbased hypotheses predict. However, the cooperative pairs exhibit what Calegari et al. 共1998兲
describe as signaling and cooperative behavior. What causes the pairs to behave differently is an
unanswered question that should interest BAR.
Calegari et al. 共1998兲 also reported an interesting methodological result. The outcomes from
a human-computer dyad were different from those of the human-human pairs. Obviously, the
computer was not programmed to respond to cues/signals, such as willingness to cooperate, that
the human partner might send. This reinforces the concern about the limits in utilizing the individual choice style of research when the “other party” has an opportunity to act/interact strategically. This is especially true where the set of actions includes choices that could facilitate reaching
a noncompetitive, but mutually beneficial, conclusion.
There are, however, settings when studying dyads in a laboratory may not be practical or even
feasible. This would be especially true in cases such as Calegari et al. 共1998兲, where students may
Behavioral Research In Accounting
Volume 23, Number 1, 2011
American Accounting Association
18
Birnberg
not be suitable surrogates for professionals. This raises the issue of external validity. Researchers
have tried to resolve this problem in an audit setting by studying the negotiation process using
professionals as participants in individual choice studies that “simulate” interacting dyads. For
example, Favere-Marchesi 共2006兲 studied the initial negotiation postures of auditors and clients
over a proposed change in the financials, giving the same case study separately to each type of
participant. They conclude that ex ante the clients have a better understanding of the auditors’
initial position than the auditors do of the clients’. In a related study, Tan and Trotman 共2007兲
proposed and tested a model of when in the negotiation process auditors should make concessions
to clients. Their experiment uses financial officers as clients and a computer simulation as the
auditor who negotiates with the client 共via email兲. They report the clients’ responses and the
clients’ strategies in responding to the simulated auditor. However, their findings should be viewed
in light of Calegari et al. 共1998兲. How this initial difference and differing strategies would play out
during negotiations between financial officers and actual auditors remains an open question.
Because of the potential problems involved in using actual auditors and their clients, it is unlikely
to be studied in an experimental setting using professionals as participants in both roles. We may
need to rely on archival research to understand the behavior of these dyads 共e.g., Nelson et al.
2002兲.
Settings with explicitly unequal power. Other papers have utilized dyads in negotiation/
bargaining studies where the parties possess unequal power. These studies usually investigate the
presence or absence of the norm of fairness in economic man rather than negotiation in a specific
setting. They typically utilize either the ultimatum or the dictator game 共Roth 1995兲. In the dictator
game, one person 共the dictator兲 is given an endowment to allocate between self and another party
共the recipient兲. The recipient must accept the dictator’s allocation. These studies utilize actual
rather than simulated recipients. Because the recipient is passive in the experimental setting, the
use of a dyad would appear to be intended to meet the criterion of not misleading the participants.6
In contrast, in the ultimatum game, the first party’s 共the “proposer”兲 situation is identical to that of
the dictator except that the recipient now may accept or reject the proposer’s offer. If accepted, the
proposer’s offer determines each party’s payoff. However, if rejected, both parties receive nothing.
The results of studies using both games tend to support a norm of fair treatment expected by
the responders and recognized by the dictator/proposer 共Roth et al. 1991; Berg et al. 1995兲. In both
the dictator and ultimatum games, the first party makes an offer approaching, on average, 40
percent of the endowment 共Roth 1995兲. This result appears to reflect the recognition by many of
the participants of a norm that sets the “fair” allocation of the endowment.
Cheap talk research in dyads. The typical “cheap talk” study also reflects a setting where
the strategic interaction is germane to the study 共e.g., Kachelmeier et al. 1994; Rankin et al. 2003兲.
How will the party receiving the nonbinding message react to it? Obviously, such a study could be
done using the individual receiving the message as the focus. However, such a study would lose
the behavior of the participant who is allowed to make the cheap talk commitment. That individual’s behavior also is of interest to the researcher. Thus, it is preferable for the study to use a dyad
共potential sender and receiver兲 rather than only a receiver. In general, research has found that the
cheap talk often is viewed by the recipient as if it is a binding commitment 共e.g., Kachelmeier et
al. 1994; Zhang 2008兲. Cheap talk studies can be conducted in any setting in accounting where the
context permits one party to communicate with and make a nonbinding pre-commitment to another party that, if true, should affect the other party’s behavior.
6
There are, of course, designs where the recipient-participant could be needed later for another experiment. For example,
the recipient-participant in the early rounds could, in the later rounds or in another experiment, play the role of the
dictator. The researcher could study the interaction between the amount offered to a participant and the amount subsequently offered by that participant when acting as the dictator.
Behavioral Research In Accounting
American Accounting Association
Volume 23, Number 1, 2011
A Proposed Framework for Behavioral Accounting Research
19
Effect of non-negotiating third party. The work of Fehr and Gaechter 共2000兲 and Zhang
共2008兲 provide insight into why it is beneficial for the researcher to include all the potential parties
in a study. Fehr and Gaechter 共2000兲 report that a third party, who only observes unfair behavior,
is willing to incur a cost to punish the unfair participant. Zhang 共2008兲, in a BAR study, provides
an interesting twist on the strategic interaction present in dyads. The dyad about which she
hypothesizes consists of two managers 共agents兲 who report to the same owner 共principal兲. She
examined the truthfulness and whistle-blowing behavior of two agents. Each agent’s cost is common knowledge to the two agents, but asymmetrical information to the principal. Essentially, her
findings show that the strategic behavior of the members of the dyad 共the agents兲 depends on the
endogenous behavior 共fairness兲 of the third party 共the principal兲. The actual presence of the third
party in the study had two benefits. First, it enhances the internal validity of the study. Second, it
ensures that the principal’s behavior in the experiment actually reflects how the principal would
act. In this case, the principal offers a lower wage because of concerns over being cheated by the
agents. This insight, in turn, can serve as a basis for future BAR on the principal’s behavior in this
setting.
Reputation. We all utilize information on another’s past behavior 共i.e., reputation兲 in making
choices. Similarly, managers must rely on the reputation of other managers in making investment
decisions, and investors, analysts, and auditors rely on managers’ reputations in their interactions
with firms. However, there is limited research on the role of reputation in the willingness of one
party to trust another.7 This reflects the design of experiments. Most studies, such as those described in the previous sections, use a “turnpike” approach. The participants are anonymously
paired and typically do not “play” the same participant more than once. This is intended to
eliminate reputation as a factor in decision making and as a potential confound. Thus, the question
of the reputation of individual players must be set aside.
But what is known is that when players interact over time, expectations and reputations are
formed and, moreover, the quality of decision making may improve relative to the turnpike design
共Schwartz and Young 2002兲. Duffy et al. 共2009兲 provide further insight into reputations. Participants may not always recognize the value of acquiring information about the other participant’s
behavior, a form of reputation. They reported that participants who initially received costless
feedback about the behavior of others utilized the feedback/reputation-related information. However, those participants who did not receive feedback information until later in the experiment did
not utilize the information to the same degree. In addition, they report that when a nominal cost is
attached to the feedback, participants did not buy the information even though it was quite
profitable to do so.
Note that in the studies discussed above, reputation is very stylized: it takes the form of very
specific information. This encapsulates the idea of reputation in the laboratory. However, in the
“real world,” the information that goes into forming a reputation may be subjective and imprecise.
Given the role that reputation can play in business settings, there is room for additional research
in this area.
Summary: Strategic Choice Studies
The study of dyads is at the intersection of individual and group BAR. It offers valuable
insights into the individual’s strategic behavior and is important for three reasons. One is that
strategic behavior is integral to many business activities. A second is that participants act differently when the other party is present rather than hypothetical 共e.g., Calegari et al. 1998兲. Finally,
7
Archival markets research has concerned itself with audit firm reputation, particularly in the wake of Arthur Andersen
共e.g., Barton 2005兲. There has been very limited BAR in this area 共Mayhew 2001, Mayhew et al. 2001兲. BAR has
operationalized the audit firm as an individual in experimental markets studies.
Behavioral Research In Accounting
Volume 23, Number 1, 2011
American Accounting Association
20
Birnberg
and perhaps most importantly, the use of dyads permits the researcher to study both sides of the
strategic interaction and do so over a series of iterations between members of the dyad. The dyad
may be composed of peers as in Zhang 共2008兲 and Towry 共2003兲, or be hierarchical as in the
studies of budget negotiation 共Fisher et al. 2000兲 and the dictator and ultimatum games.
BAR research undertaken thus far suggests that the presence of a “real person” with whom the
participant interacts affects their behavior 共Calegari et al. 1998兲. BAR using dyads could be useful
in developing a better understanding of how managers and workers, as well as auditors and tax
professionals/payers, behave in various settings, in addition to insights into the negation process.
It also could reveal how “soft behavioral constraints” such as norms can affect behavior.
The nature of the interaction can vary, as can the mechanism used to achieve it. As even the
ultimatum game shows, both parties possess some power 共i.e., the ability to affect the behavior of
the other兲, albeit in some cases a very “soft” power. The study of how they use this power and how
the parties interact 共their strategies兲 is what makes the study of dyads interesting. It is important to
note that the results discussed above and elsewhere often run counter to the simplistic notion of the
self-interested, wealth-maximizing “economic person.”
Because dyads can be viewed as a subset of group behavior, studying dyads yields potentially
valuable insights into group behavior. However, there are obvious limitations. The greater level of
complexity facing the individual members of a group increases with the number of members
interacting. Thus, many of the laboratory studies reported below under group-focused BAR limit
the strategic choices available to the interacting parties. As useful as data gleaned from the study
of dyads may be, to better understand the group phenomenon in question researchers have turned
to alternative research methods relying on naturally occurring events 共fieldwork, archival data,
surveys, and interviews兲.
The ability to undertake research on dyads and observe the strategic interaction of the parties
may not be as easy as the BAR focusing on the individual. Dyad research at least doubles the
number of participants required with a comparable increase in the cost of the experiment. It also
can require a high degree of coordination. The participants must be available at the same time and,
typically, in the same place. This suggests that research of this type is likely to take place in a
laboratory or through fieldwork. The former is likely to mean student participants; the latter,
professionals performing their job in their natural environment. This would appear to limit the
amount of work of this sort that will be undertaken using nonstudent participants.
GROUPS
The label “group” in this context is used to include a variety of organizational structures.
Group is defined as any collection of individuals greater than two and typically no more than four
in laboratory studies. Rarely is it more than five members. This definition is admittedly arbitrary,
but consistent with the literature in the area. The above definition does not specify a particular
organizational structure共s兲 for a group. Thus, group as defined for this section includes not only
peer groups, but also teams and hierarchical groups.
Psychology research on group decision making initially focused on the quality and nature of
the individual versus group decisions. Which makes the better decision? Which makes the riskier
decisions? For a review, see Sutton and Hayne 共1997兲 and Daroca 共1984兲. Sociology was interested in the development of networks 共e.g., Homans 1951兲 and the affect of context variables on
group behavior 共e.g., Dalton 1959兲. For a review of sociology based studies, see Miller 共2007兲.
More recent studies have focused on the nature of the group processes. How does the composition
of the group 共e.g., temporary or permanent兲 affect its decision? What is the effect of changes in
group membership? How does the decision rule used by/imposed on the group affect their decision?
Behavioral Research In Accounting
American Accounting Association
Volume 23, Number 1, 2011
A Proposed Framework for Behavioral Accounting Research
21
BAR on groups has addressed five broad categories: 共1兲 individual versus group performance,
共2兲 group decision processes, 共3兲 the role of technical and accounting systems in group decisions,
共4兲 the role of incentives, and 共5兲 the role of a group’s characteristics in its performance. Many
studies have asked questions that relate to more than one of the above categories. BAR group
research has utilized the full range of research methods including experiments 共e.g., Young et al.
1993兲, surveys 共e.g., Chalos and Poon 2000兲, protocols 共e.g., Bedard et al. 1998兲, video 共Walker
and Aritz 2006兲, and field research 共e.g., Anderson et al. 2002兲.
Participants
The type of participants used in group research has varied depending primarily on the subcategory of BAR being studied. As is described below, auditing studies have used auditors as
participants whenever possible. Recently, studies have again begun to use students. This reflects
both the declining availability of auditors as participants and the belief that student participants
possess the appropriate knowledge, skill, and experience for many group tasks.
In contrast, the study of groups in other areas, particularly management accounting, has used
a more diverse set of participants. Laboratory studies typically have used students, albeit often
with significant work experience 共e.g., Daroca 1984; Rowe 2004兲. Managerial accounting researchers have studied “real people” in field studies 共e.g., Anderson et al. 2002兲, and surveys of
managers reporting on “on the job” experiences 共e.g., Chalos and Poon 2000兲.
Group Decisions and Processes
It is interesting to note that the much of the early research on groups in BAR was in auditing
共Schultz and Reckers 1981; Reckers and Schultz 1982; Trotman et al. 1983兲. This likely reflected
the overall level of BAR interest in auditing during this period, as well as the absence of teamwork
in U.S. firms at that time. The findings of the auditing BAR studies generally are consistent with
earlier non-BAR group research. For example, Schultz and Reckers 共1981兲 report that decisionmaking groups exhibited higher confidence and less variability than individuals. In a topic more
closely related to accounting than generic group research, Reckers and Schultz 共1982兲 report that
groups adhere to the accounting rules more closely than individuals. Indeed, because groups 共audit
teams兲 are the way audits are performed, the use of groups in auditing has been a continuing area
of BAR in auditing 共e.g., Solomon 1987; Reckers and Schultz 1993兲.
In management accounting, Daroca 共1984兲 studied participation in a group setting. He reported that, as Becker and Green 共1962兲 conjectured, participation could result in group polarization against management, leading to negative rather than positive “gains” from participation.
These findings, like those of Zhang 共2008兲 and Greenberg 共1990兲, indicate that group involvement
may have negative outcomes for the organization if the leader’s style is perceived negatively by
the group.
Unlike the typical generic group study that focused solely on the group’s output/decision,
Bedard et al. 共1998兲 studied group processes as well as the efficacy of groups versus individuals.
They utilized protocols developed from audio tapes to examine communication among group
members and identify what type of interactions characterized successful and unsuccessful groups.
Because their sample was of necessity small, the findings must be viewed tentatively. However,
they raised an important issue by delving into what makes groups effective. Accordingly, they
studied process as well as outcomes. Bedard et al. 共1998兲 also investigated how the voting rule,
formal or informal, affects group behavior. Given the range of possible rules 共e.g., unanimity,
majority rule, and leader with a veto兲, it is reasonable to expect that the voting rule could affect the
group’s behavior and output/decision 共Birnberg et al. 1970兲. This issue is relevant to any group
decision-making setting within accounting. Given the size of Bedard et al.’s 共1998兲 sample, it is
hard to draw a conclusion.
Behavioral Research In Accounting
Volume 23, Number 1, 2011
American Accounting Association
22
Birnberg
Chalos and Poon 共2000兲 used the group setting to study participation. They used a survey of
and interviews with 177 managers comprising 55 budget teams in a single firm to collect data on
the effect of group process on perceived quality of group decisions. They studied how the presence
of participation in the group’s capital budgeting process affects information sharing, budget emphasis, and self-reported measures of performance. They report that participation positively affects
the perception of the amount of performance information available, amount of information sharing, and the reported importance of the budget process. However, it is central to note that the
researchers did not observe the groups in action.
Role of Decision Support Systems
Just as an individual’s decision making can be affected by the use of a decision support
system 共DSS兲, group decision making can be altered by a DSS. Murthy and Kerr 共2003兲 and Kerr
and Murthy 共2004兲 investigated the impact of different types of computer-mediated communication 共CMC兲 in different task settings on the quality of the group’s decision. Typically, the conditions compared are face-to-face communication and computer-based system共s兲. The findings indicate that face-to-face groups outperformed CMC groups when problem solving was the measure
of performance 共see also Rowe 2004兲. Interestingly, both CMC and face-to-face were equally
effective in generating ideas, though performance appears to be sensitive to the task setting and the
type of CMC. Kerr and Murthy 共2004兲 report that a bulletin board form of CMC outperforms
“chat rooms” and face-to-face communication in a decision setting that requires the participants to
exchange uniquely held information to reach a successful conclusion.
One caveat in evaluating the above findings is the use of student participants. Ho 共1999兲 used
audit partners, managers, and seniors to study the role of computerized decision support system
relative to face-to-face communication in a going-concern evaluation. Her study reports that
groups of both types considered evidence that individuals did not. She reports when comparing the
two groups that CMC groups had greater agreement on the going-concern assessment than did
face-to-face groups and had greater satisfaction with their evaluation. A possible explanation is
that the “impersonal” CMC setting may neutralize the ability of an influential/powerful individual共s兲 in the group to exert undue influence in the group’s decision.
Carpenter 共2007兲 also used auditors in a group research study examining the recommendation
of SAS No. 99, which requires the use of groups formally in the audit process through “brainstorming” sessions. Because the brainstorming literature in psychology using students had not
uniformly reported the synergistic behavior expected from group discussion 共Dennis and Valacich
1993兲, she studied the process in an audit setting using auditors as participants. She hypothesized
that brainstorming groups would perform better than individuals or nominal groups in part because
the group members were professionals “doing their job” in the experiment rather than student
participants performing a mundane task. Her results support the benefits of brainstorming. Hoffman and Zimbelman 共2009兲 extended Carpenter 共2007兲. They used brainstorming to improve the
audit program. In their study, a panel of experts brainstorm potential modifications in the fraud
detection program for a case study. They report that auditors subsequently given the case and the
modified program performed better than those who were not.
Based on the findings of Ho 共1999兲, Carpenter 共2007兲, and Hoffman and Zimbelman 共2009兲,
it would appear that the findings of generic group research may not always apply to BAR. The
good news for BAR is that it opens a wide variety of questions. The bad news is the ability to
secure access to professionals functioning in a group setting. In its early stages, research on
computer-aided group decision making may need to rely on field and archival data.
Role of Incentive Systems
As in other areas, the role of incentive systems has been very important in group research.
Management accounting group research recognizes the conflict between group incentives and
Behavioral Research In Accounting
American Accounting Association
Volume 23, Number 1, 2011
A Proposed Framework for Behavioral Accounting Research
23
individual incentives. When the contribution of the individual is identifiable, an individual-based
incentive system prevents the free rider problem, where the individual makes a minimal contribution to the group effort and secures a disproportionately large reward. However, when the only
observable measure is the group outcome, the manager is limited on what he/she can base the
payoff 共e.g., Drake et al. 1999兲.
Researchers have attempted to ascertain ways in which free rider behavior in groups might be
mitigated. Towry 共2003兲 argues that information on peers’ performance that is unknown to the
principal/manager is observable by group members. Thus, the group members are capable of
mutual monitoring. She reports that the greater the members’ group identity, the more effective
their ability to monitor each other’s behavior 共mutual monitoring兲 and the greater degree of
coordination they can achieve.
Rowe 共2004兲, in a study that examined both systems and incentives, takes a slightly different
approach to resolving the free rider problem. Rather than using monitoring, he used the “information system” to inform the group’s members that free riding in his task actually was suboptimal behavior for both the free rider and the group. Rowe 共2004兲 modeled the free rider
problem as a public goods dilemma. Each member of a four-person group decided how much of
their endowment to contribute anonymously to a common pool. The amount in the pool was
tripled and divided equally among the group’s four members without regard to how much each
had contributed to the pool. Obviously, the self-interested strategy is to contribute nothing and
share whatever is in the tripled common pool. If all four members of the group follow a free rider
strategy, they would be no worse off than they were at the beginning of the experiment. At the
other extreme, if all members of the group contributed their entire endowment to the pool, everyone would be significantly better off. Rowe 共2004兲 found that the group members contribute 共and
therefore receive兲 more when the information system informs an intact group of the benefits of
contributing. This occurs even though the information system did not provide any new information
and there is no communication among group members.
The Impact of Extra-Group Factors
While all of the above studies were laboratory-based, BAR also has examined the behavior of
real groups. This permits the researcher to observe the effect of the setting in which the decision
takes place. Rowe et al. 共2008兲 and Anderson et al. 共2002兲 used two different research methods to
study the decision processes of groups in their natural setting. Both papers study 共among other
things兲 group conflict, the sharing of horizontal asymmetric information, and the potential role of
consultants. Rowe et al. 共2008兲 report the results of a longitudinal, participant-observer field study
of a particular cross-functional group within a division of a firm. The group was formed by
management to try to reduce costs. Initially, each group member behaved in a self-interested
fashion to retain slack and benefit their particular function in the organization. The outcome of the
study shows that consultants, by redesigning the information system, were able to mitigate selfinterested behavior 共conflict兲 and replace it with more group-oriented behaviors. Rowe 共2004兲8
tested this finding in a laboratory setting.
Post-decision, Anderson et al. 共2002兲 used a survey supplemented with interviews of group
members within a single firm. They examined the effects of a large number of variables 共e.g.,
conflict resolution, group size, presence of consultants, importance of decision�…