Description
In two studies, practising auditors’ responses to hypothetical audit conflict scenarios were used to test
hypotheses that moral reasoning development and belief in a just world influence resistance to client
management power. Based on a mixed factorial ANOVA design, results confirmed that auditor independence
is interactively determined.
Pergamon Accounttng, Organizaffons and Society, Vol. 20, No. 7/S, pp. 701-720, 1995
Copyright 0 1995 Elsevier Science Ltd
Printed in Great Britain. All rights reserved.
0361-3682/ 95 $9.50+0.00
03613682(95)00018-fii
THE EFFECT OF CLIENT MANAGEMENT BARGAINING POWER, MORAL
REASONING DEVELOPMENT, AND BELIEF IN A JUST WORLD ON AUDITOR
INDEPENDENCE*
CAROLYN A. WINDSOR
Gr@tb University (Nathan Campus)
and
NEAL M. ASHKANASY
The University of Queensland
Abstract
In two studies, practising auditors’ responses to hypothetical audit conflict scenarios were used to test
hypotheses that moral reasoning development and belief in a just world influence resistance to client
management power. Based on a mixed factorial ANOVA design, results confirmed that auditor indepen-
dence is interactively determined. Three styles of auditor decision-making emerged: “autonomous”,
“accommodating” and “pragmatic”. Autonomous auditors were responsive to personal beliefs, such
that those with strong just world beliefs were more likely to resist client management power. Pragmatic
auditors were responsive to client management power, irrespective of beliefs. Accommodating auditors
responded both to personal beliefs and client management power; they comprised the least resistant
group, especially when they believed in an unjust world.
The ethical dilemmas faced by authors who
may be forced to compromise their indepen-
dence when pressured by client management
power provided the focus of the research
described in this paper. These issues are cur-
rently attracting attention in the academic lit-
erature (e.g. Ponemon & Gabhart, 1990) and
carry implications for audit practitioners (see
Ashkanasy & Windsor, 1994; Godsell, 1993).
The paper extends Ponemon and Gabhart’s
(1990) study of auditor independence judg-
ments, by including economic and personal
belief variables in the context of an audit con-
flict with client management. We argue that
the process of auditor independence is com-
plex and that no one factor can provide an
instant remedy.
In this paper, the concept of auditor inde-
pendence is first developed, based on a com-
plex decision-making process, which incor-
porates Kohlberg’s (1969) moral reasoning
development construct, and Lemer’s (1980)
concept of belief in justice. In order to over-
come concerns about the internal and external
validity of results based on hypotheses invol-
ving high-level interactions (Keppel, 1989), a
two-stage mixed factorial repeated-measures
design was employed in an examination of
l The authors acknowledge with thanks Ferdinand Gul, Errol Iselin, and Lawrence Ponemon for their contributions and
encouragement. We also thank the anonymous reviewers for their comments on an earlier draft of this paper which was
presented (in poster form) at the American Accounting Association ABO Research Conference, San Antonio, 10-l 1 March
1994, and the anonymous reviewers of the present version of this paper. The support and willing participation of public
accounting firms is acknowledged with thanks.
701
702 C. A. WINDSOR and N. M. ASHKANASY
auditors’ independence judgments in the face
of client management bargaining power.
Audi tor i ndependence
Auditors are required to remain independent
of the client but at the same time depend upon
the client for their livelihood through fees
(Mautz & Sharaf, 1961). An audit conflict may
develop, however, when auditors request dis-
closure of information that client management
may not wish to reveal in financial reports to
the public. The conflict becomes an ethical
dilemma when the auditor is confronted with
the decision to compromise independence and
integrity for economic gain. In the present
study, evidence about the existence of audit
dilemmas was collected through interviews
with senior audit partners from first-tier multi-
national public accounting firms. The partners
maintained that auditors have to balance pro-
fessional ethics with the economic practical-
ities of operating a viable business. Yet,
philosophically, auditors are supposed to serve
society, rather than to focus on economic com-
pensation (Gas, 1992; Mautz & Sharaf, 1961).
Gaa (1992) has argued further that there is a
“social contract” between the auditing profes-
sion and the rest of society which establishes
reasons for auditors to act ethically. Thus, audi-
tors are confronted with balancing “I” versus
“we” considerations (Etzioni, 1988). “I” repre-
sents the self interested, pragmatic considera-
tions of personal and economic gain; “we”, on
the other hand, encompasses professional and
ethical reponsibilities to remain independent.
The concept of auditor independence is inte-
gral to professional codes of ethics and stan-
dards which guide auditors in practice (Mautz
& Sharaf, 1961). Auditors must maintain inde-
pendence, defined by Mautz and Sharaf (1961)
as a “lack of bias in forming delicate judg-
ments” (p. 206). Auditor independence should
therefore be characterized by integrity and an
objective approach to professional work,
described by Mautz & Sharaf (1961) as “a state
of mind”. The present paper focuses specifi-
cally on independence, which is based on the
view that auditors are social human beings,
subject to a full range of emotions, beliefs,
and prejudices (Gas, 1992). This is in contrast
to the objective view, where the obligation is
imposed on auditors “to be fair, intellectually
honest, and free of conflicts of interest” (IFAC,
1990, Part A, para 1.1).
A key component in auditor independence is
the nature of the auditor-client relationship
(Goldman & Barlev, 1974). Research in this
area, however, has emphasized issues related
to firms switching auditors (e.g. Eichenscher
& Shields, 1983; Craswell, 1988) and issues
of individual auditor independence from audi-
tor and third party perceptions (e.g. Gul, 1991;
Knapp, 1985, 1987; Pany & Reekers, 1988;
Shockley, 1981). Until recently, however, little
research has focused specifically on auditors’
cognitions and psychological aspects of inde-
pendence (Bamber, 1993). Shaub et al . (1993)
have argued this is because the accounting pro-
fession has tended to rely on authoritative
codes of ethical conduct and practice stan-
dards. Recently, a number of researchers
(e.g. Lampe & Finn, 1992; Ponemon, 1992;
Ponemon & Gabhart, 1990, Shaub et al .,
1993) have begun to look at cognitive pro-
cesses as a means of seeking a deeper under-
standing of auditor independence decision-
making. For example, Ponemon and Gabhart
(1990) examined the impact of moral reason-
ing development on auditor decision-making,
and proposed that three levels of auditor inde-
pendence could be identified, corresponding
to Kohlberg’s (1969) levels of moral reasoning
development. Cushing (1990), however, has
argued that the cognitive perspective, taken
in isolation, may be overly restrictive, and
needs to be incorporated into more compre-
hensive models of decision-making. He sug-
gests that Ponemon and Gabhart’s (1990)
model could be extended by including the
effect of economic variables in a materiality
judgment situation. We argue further that per-
sonal beliefs about justice also play an impor-
tant role in the way people process important
information (Lerner, 1991). Lerner’s (1980,
1981) concept of belief in justice is relevant
in the auditor independence context because
AUDITOR INDEPENDENCE 703
justice is an inseparable component of integrity
(Karniol & Miller, 1981). The aim of the pre-
sent research, therefore, was to extend our
understanding of auditor independence judg-
ments by an examination of the combined
impact of psychological factors and client eco-
nomic power.
Complex decision-making
Mautz and Sharaf (1961) described auditor
independence as a “state of mind”, but this
simple definition does not reveal the complex-
ities of the auditor’s decision-making process.
Shaub et al. (1993) for example, adapted Hunt
and Vitell’s (1986) model of general theory of
marketing ethics, and showed that auditors’
sensitivity to ethical dilemmas depends upon
their attitudes to moral rules. Lampe and Finn
(1992) developed a “five-element” decision-
making model by adapting Jones’ (1991) issue
contingent model. The fiveelement model
combines three cognitive developmentalist
stage models and was shown to describe ethi-
cal decision-making better than a “Code-
implied” model. Bonner and Pennington
(1991) also argued for adoption of models of
auditor decision-making based on cognitive
audit task analysis.
The cognitive-based models (as described by
Lampe & Finn, 1992) however, do not take
into account the wide range of factors poten-
tially affecting auditor decision-making. For
example, Gaa (1992) criticized Lampe and
Finn’s (1992) five-element model for its failure
to address the full range of complexities of
auditor independence judgments, and suggests
a more complicated explanation based on Tre-
vino’s (1986) interactionist model. In effect,
Gaa (1992) calls for a more sophisticated
theoretical model of decision-making which
goes beyond cognitions. The interactionist
approach focuses on combinations of personal
and situational variables (rather than on single
variables acting alone, see Trevino, 1986) and
thus provides a more sophisticated and realistic
indication of human decision-making processes
in real situations (which are described in terms
of high-order interactions, see Keppel, 1989).
However, while the Trevino (1986) model
provides an appropriate theoretical basis, it
gives little guidance in respect of empirical
testing, and therefore needed to be adapted
for use in the present study. This was accom-
plished using a model developed by Ashkanasy
(1989) in the context of evaluative decision-
making. This model (see also Ashkanasy,
1991) is based on a general information proces-
sing framework in which cognitions are trig-
gered by contextual information, and are
affected by deeply embedded psychological
beliefs when an immediate decision is difficult
to achieve. In essence, the model posits that
impressionistic and cognitive processes are
influenced by personal beliefs, which are in
turn reflected in interactions between situa-
tional contingencies and cognitive processes.
The model is also consistent with findings by
Brief et al. (1991) that personal values interact
with perceptions of accountability and explicit-
ness to determine responses to ethical dilem-
mas in managerial decision-making.
By substituting the variables from Trevino’s
(1986) conceptual model of ethical decision-
making for those in the original Ashkanasy
(1989) model, it was possible to derive a testa-
ble model in the present instance. In this case,
the three variables were: (1) economic dimen-
sions of client bargaining power; (2) moral rea-
soning development (Kohlberg, 1969); and (3)
belief in a just world (Lerner, 1980, 1981). We
expected to find that auditors’ decisions in
response to client management economic bar-
gaining power would be interactively influ-
enced by cognitive processes involving moral
reasoning and beliefs.
Client management bargaining power
Client management has an inherent bargain-
ing advantage because auditors depend upon
client fees for their livelihood (Mautz &
Sharaf, 1961). At the same time, auditors may
need to qualify their client’s financial report in
a manner which could harm the client manage-
ment’s interests (Goldman & Barlev, 1974).
Further, Hill and Jones (1992) have shown
that client management has the bargaining
704 C. A. WINDSOR and N. M. ASHKANASY
advantage by virtue of corporate management’s
position at the nexus of implicit and explicit
contracts that constitute the firm. Antle and
Nalebuff (1991) have argued also that client
management has superior knowledge of the
firm’s income. Using a game-theory model of
client-auditor negotiations of financial state-
ments, they demonstrated that client manage-
ment always has the negotiating edge.
Auditor independence is further threatened
by the very nature of the relationship between
client management and the auditor, which is
fostered by professional practice standards.
These standards require auditors to have a
sound knowledge of their clients’ businesses
and industries, hence setting the scene for a
close professional relationship, and likely to
result in reduced independence. The auditor-
client management relationship is strengthened
over time as the auditor develops specialized
expertise (which would lose its value if
applied to another relationship, see Levinthal
& Fichman, 1988).
auditor’s dependence on client fees as a poten-
tial economic factor that could affect auditor
independence. This is supported by Gul
(1991) who found that, when fees are a signif-
icant proportion of the auditors’ total revenue,
third party perceptions of the auditors’ ability
to withstand management power are adversely
affected. Gul (1991) explained this finding in
terms of Knapp’s (1985) observation that audi-
tors are more responsive to client management
power when the client provides a significant
proportion of the auditor’s income.
In the present study, we have conceptua-
lized client bargaining power in terms of situa-
tional factors that exert pressure on auditors,
and therefore influence their independence
behaviour. The initial interviews with senior
audit partners indicated that the three primary
dimensions of client management power were
(1) client financial condition, (2) size of client
fees, and (3) whether or not the client calls
tenders for audit work.
Tendering. When client firms invite prospec-
tive auditors to tender for provision of audit
services at a competitive cost, there is a risk
that auditors may experience difficulty in
remaining independent. This is because client
management can overtly or covertly threaten to
replace the auditor. Craswell (1992) for exam-
ple, showed that the introduction of tendering
led to an increase in the frequency of auditor
switches and fee reductions. Hence, client
management can use tendering to place more
economic pressure on the auditor, thus impair-
ing independence.
Moral reasoning development
Financial condition of the client. Auditors
dealing with clients in poor financial condition
are likely to face a higher level of legal liability
than if the client was in good financial condi-
tion (Farmer et al., 1987; Palmrose, 1987). In
this respect, Knapp (1985) found also that
financial statement users perceive that manage-
ment is more likely to obtain its preferred reso-
lution in a conflict when the client is in a good,
rather than poor financial condition. It follows,
therefore, that auditors would be likely to exer-
cise greater caution when dealing with clients
who are in poor financial condition.
The present study utilized Kohlberg’s (1969)
moral reasoning construct. Although this con-
struct has been subject to some controversy
(see Modgil & Modgil, 1986) it has been used
successfully as a cognitive measure in a number
of studies in accounting and organizational set-
tings over recent years (Gaa, 1992; Ponemon,
1992; Ponemon & Gabhart, 1990; Trevino,
1986). Gaa (1992) argued that the construct is
a measure of “moral expertise” in the specific
context of auditor judgments. It is based on a
fundamental assumption that moral reasoning
reflects an underlying organization of thinking
which becomes more complex and sophisti-
cated with the individual’s moral development
(Kohlberg, 1969). Rest (1986) argued further
that the word “moral” as used here involves
social interaction, and emphasized its applic-
ability to concepts of justice or fairness.
Size of the client fee. Mautz and Sharaf Kohlberg’s theory includes three broad
(1961) drew attention to the problem of the levels of moral reasoning development, each
AUDITOR INDEPENDENCE 705
composed of two stages. Moral reasoning
development involves the individual’s transi-
tion from stage to stage in an invariant, irrever-
sible sequence. At the preconventional level,
stages one and two, an individual is concerned
with concrete consequences, particularly
reward and punishment affecting individual
immediate interests. At the conventional level,
or stages three and four, “right” is conforming
to the expectations of acceptable behaviour of
significant others; for example, larger society, a
peer group or family.
At the post-conventional level, or stages five
and six, “right” is determined by universal
values or principles; for example, saving
human life. When laws violate these princi-
ples, the individual may act in defiance of the
law. The individual at this level sees beyond
social norms, laws, and authority of groups
and indivdiuals regarding judgments of right
and wrong.
Ponemon and Gabhart (1990) used Kohl-
berg’s moral reasoning construct in a study of
auditor independence under varying condi-
tions of internal sanction. They found that
moral reasoning development affected
response to penalty sanctions; auditors with
low (pre-conventional) levels of moral reason-
ing development were more sensitive to pen-
alty sanctions than auditors with higher levels
of moral development. These results were
based on Rest’s (1979) “P-score”, which is a
continuous measure, rather than a stage mea-
sure. Ponemon and Gabhart (1990) maintain
that P-scores are strongly associated with Kohl-
berg’s stages (see Colby & Kohlberg, 1987)
and constitute a satisfactory operationalization
of the Kohlberg stage concept. In the present
study, we anticipated that higher levels of
moral reasoning (measured using P-scores)
would be associated with greater resistance to
client management bargaining power.
Personal beliefs in a just world
Finally, personal beliefs were operationalized
in the present study using a measure of belief in
a just world. This construct is appropriate to
test auditor independence because justice and
integrity are intimately intertwined (Karniol &
Miller, 1981). Lerner (1991) argues that the just
world motive is a tendency to believe the
world as operating in a consistent, just manner
where the good are rewarded and the bad are
punished. The belief in justice operates at a
preconscious level and is revealed indirectly
in the person’s reactions to a given event (Ler-
ner, 1980, 1991). Research has convinced Ler-
ner and his colleagues that the commitment to
justice in people’s lives is not simply a principle
governing how to view their situation, but as
the central organizing principle in how they
acquire and allocate desired resources (Ler-
ner, 1991). At a conscious level, people with
just world beliefs appear to others and them-
selves to have a realistic view of injustices, but
plan their lives as if they live in a just world as
means of bringing order to their lives and cop-
ing with life’s tragedies. Thus, individuals with
just world beliefs are committed to a personal
contract that presupposes people deserve their
outcomes and receive, in the long run, what
they are entitled to. Hence, this personal con-
tract becomes a catalyst for the individual to
relinquish immediate outcomes for long-term
rewards. In contrast, people with unjust world
beliefs believe that people are not rewarded or
punished in accordance with their actions, and
feel that they are victims. These people there-
fore behave as if they have little control over
life’s events.
To measure justice beliefs, we used the belief
in a just/ unjust world dimension identified by
Collins (1974). Lerner (1980) has suggested
this is an appropriate measure of beliefs in
justice. It is a dimension of Rotter’s (1966)
locus of control scale, which measures a
personality construct derived from social
learning theory (Rotter, 1954). Research has
shown that Rotter’s (1966) locus of control
construct is multidimensional (e.g. Collins,
1974) and subsequent studies have specifically
replicated Collins’ belief in a just world dimen-
sion (Ashkanasy, 1985; Zuckerman & Gerbasi,
1977a,b; Zuckerman et al. 1977). Zuckerman
and colleagues have provided evidence that
706 C. A. WINDSOR and N. M. ASHKANASY
the unjust/ just world dimension is reliable,
valid, and independent of the other locus of
control dimensions.
Hypotheses
Based on the Ashkanasy (1989) and Trevino
(1986) models, we have argued that auditor
independence is affected by moral reasoning
development and personal beliefs embedded
in the preconscious and activated by ethical
dilemmas. Thus, when auditors are pressured
by client management bargaining power, the
extent to which personal beliefs affect their
independence is determined by the auditor’s
level of moral reasoning development. We
expected to find in this study that auditors
with highly developed moral reasoning would
be influenced by their personal beliefs about a
just world, while those with a lower level
would be influenced primarily by client man-
agement bargaining power. Auditors between
these extremes were expected to exhibit char-
acteristics of both of these groups. In essence,
a three-way interaction was predicted, demon-
strating that personal beliefs and moral reason-
ing interact with client management power in
complex decision-making situations. The nat-
ure of this interaction is illustrated in Fig. 1.’
Specific hypotheses are:
(31)
-
High moral reasoning auditors’ resistance to client
management power is affected by just world
beliefs. They are less likely to resist when they
believe in an unjust world than when they believe
in a just world.
Low moral reasoning auditors are affected by client
management power, but not by just world beliefs.
Overall, they are less likely than high moral reason-
ing auditors to resist client management power.
High moral reasonmg
(Post-conventional) auditors
Low moral reasoning
(Pie-conventlonat) auditors
Mid moral reasoning
(Conventional) auditors i
Just world
beliefs
Unjust world
beliefs
Hypothesized pattern of auditor resistance to client
management power.
Analysis method
Mid-moral reasoning auditors are affected by client
management power and just world beliefs, They
are less likely than high moral reasoning auditors
to resist client management power. Further, they
are less likely to resist when they believe in an
unjust world than when they believe in a just
world.
The purpose of the present study was not
to deal just with auditors’ responses to eco-
nomic factors (e.g. GUI, 1991; Knapp, 1985).
Nor did we intend to address the effect of
psychological factors alone, as was done by
Ponemon and Gabhart (1990). Rather, the
focus in the present paper is on the i nterac-
ti on between client economic factors, moral
reasoning development, and beliefs in a just
world.
To test such interactive hypotheses, a mixed
factorial ANOVA research design with three
repeated measures and two between-group
’ The vertical axis in this figure is auditor resistance to client pressure (represented by the economic variables: client
financial condition, size of fees, tendering). As such it represents difference scores across levels of economic variables. The
graph therefore depicts a three-way interaction between economic variables (ordinal difference scores), just w or l d beliefs
(abscissor), and moral reasoning development group.
AUDITOR INDEPENDENCE 707
independent variables was appropriate.2 The
three repeated measures were the economic
factors identified by the senior audit partners:
(1) client financial condition; (2) size of fees;
and (3) tendering. The two between-group
independent variables were (1) level of moral
reasoning and (2) belief in a just world. The
dependent variable was a measure of the
extent to which the auditor would acquiesce
to the wishes of client management.
A problem in studies of high-order interac-
tions is that such interactions may be unstable
(Keppel, 1989). This research was therefore
carried out as two independent studies: a first
or initial study; and a second or validation study
based on a larger sample to increase the power
of analysis. While differences in the results of
the initial and validation studies were
expected, we hoped to be able to demonstrate
that the underlying theory would apply in both
instances.
INITIAL STUDY
Method
Experi mental i nstrument. This consisted of
a questionnaire booklet in three parts: (1) the
ethical dilemma; (2) Collins’ (1974) locus of
control (LOC) scale; and (3) Rest’s (1979) DIT
scale.3 The moral reasoning development and
just world belief measures were administered
after the ethical dilemma so that the scenarios
used in the DIT and the LOC measure did not
bias respondents’ perceptions of the audit con-
flict (see Trevino & Youngblood, 1990).
Ethi cal di l emma. Cushing (1990) suggested
using an audit conflict involving disclosure of
an accounting transaction that would have a
material effect on the client’s financial state-
ments. In the present study, this situation was
presented in short case-study form to test audi-
tor independence under pressure of client man-
agement power using three repeated measures.
The conflict provided in the booklet concerned
the materiaIity of certain unrecorded liabilities
discovered by the firm’s auditors. These liabil-
ities consisted primarily of expenses incurred
in the previous year which were neither paid
nor recorded until the following year. Respon-
dents were asked to assume that they were the
auditors of the hypothetical firm where client
management disagreed with the auditor that
the amount was material, with 8% materiality.
The figure of 8% materiality level was recom-
mended by the senior audit partners as a way to
create an element of realistic ambiguity, and to
ensure that the decision-making task required
professional judgment.*
The aim of the dilemma was to place audi-
tors in a conflict situation with client manage-
ment, whose economic bargaining power was
represented by eight scenarios. These were
varied by presenting three statements: client
financial condition (good or bad); size of fees
(large or small); and external audit tendered
(yes or no). Respondents were requested to
make a subjective judgment about the
dilemma using a seven-point Likert scale,
where 1 indicated a very low likelihood that
they would acquiesce to client management’s
‘ This approach was adopted in preference to regression for three reasons. The first is that Kohlberg’s theory implies
discrete groups, leading explicitly to group-specific hypotheses (see also Ponemon & Gabhart, 1990). The second reason is
that hypotheses are framed in terms of high-order interactions, which are difficult to interpret in regression models. Finally,
predictions relating to levels of moral reasoning are nonlinear, and therefore are inherently not amenable to analysis by
techniques based on linear regression. ANOVA focuses only on between-group differences and does not assume linearity
across factorial groupings (Keppel, 1989).
3 Copies of the instrument may be obtained by writing to C. A. Windsor, Faculty of Commerce and Administration, Griffith
University, Nathan, Queensland, 4111, Australia.
’ The materiality level was mid-way between the immateriality threshold limit of 5% and the materiality threshold limit of
10% (see AAS5, ASCPA, 1994).
705 C. A. WINDSOR and N. M. ASHKANASY
view, and 7 indicated a very high likelihood of
acquiescing. 5
The scenarios were generated by computer
to ensure that they were presented to respon-
dents in random order. Respondents were
requested to complete each of the scenarios
in sequence, without consulting colleagues.
The same instrument was subsequently tested
on a sample of 160 third-year auditing students.
In this instance, no significant effects of the
three economic variables were detected. This
suggests that responses to the scenarios by the
practising auditors were derived from the audi-
tors’ experience in dealing with client manage-
ment, rather than an artifact of the measure.
This finding is similar to that reported by
Farmer et al. (1987), who reported that practi-
cing auditors’ independence judgments were
affected by organizational acculturation.
Belief scale. This was the just world subscale
of Collins’ (1974) 46-item Likert-scale version
of Rotter’s (1966) internal-external control
questionnaire. The dimensional structure of
this version of Rotter’s scale was verified by
Ashkanasy (1985), and the just world subscale
has been separately validated by Zuckerman
and Gerbasi (1977a,b). Responses to each of
the items were based on a seven-point Likert
scale, where 1 indicated strong disagreement,
and 7 indicated strong agreement.
Defining issues test (DI D. The short (3-
story) version of Rest’s (1979) scale was used.
This is based on Kohlberg’s (1969) theory of
moral reasoning development, and consists of
a number of brief stories that present moral
dilemmas in everyday life. The DIT has been
extensively validated, and was described by
McCrae (1985) as a paradigm of measurement
instruments (see Gaa, 1992). The 3-story ver-
sion of Rest’s scale was used in preference to
the Gstory version because of the time con-
straints on professional auditors. Rest (1979)
cites Cronbach alpha reliability for the 3-story
DIT typically around 0.7. Respondents answer
questions concerning the importance of con-
sidering various aspects of each story using
five-point response scales, where 1 indicates
least importance, and 5 indicates greatest
importance. Scoring of responses is based on
a formula set out by Rest (1979).
Procedure. Prior to administering the
research instruments, a three-stage strategy
was used to strengthen internal validity and
to overcome data collection problems. First,
preliminary interviews were conducted with
six senior audit partners (one from each of
the big six audit firms) to assess the nature of
audit conflict problems, and to gain the part-
ners’ co-operation. In the second stage, a draft
case study was circulated to the audit partners
to enlist their help in developing realistic sce-
narios, which appropriately represent client
management power. The advice and com-
ments provided by the senior audit partners
were used to compile the final versions of the
instrument. In the third stage, the audit part-
ners were briefed on how to distribute and
collect the questionnaire booklets so as to mini-
mize internal validity problems such as history
and maturation. The co-opted partners were
instructed not to participate in the experi-
ment, nor to advise participating auditors
about the substance of the questionnaires.
The questionnaire booklets were distributed
to experienced auditors using secretarial ser-
vices provided by the participating firms,
under supervision of the co-opted audit part-
ners. The secretarial staff offered the booklets
to participating auditors immediately on their
return from audit assignments, and provided no
indication that the study was dealing with ethi-
cal issues. Those agreeing to participate were
asked to complete the booklets within 30 min-
utes, and to return them (in sealed envelopes)
to the person who gave them out. Three weeks
after distribution to the firms, the completed
instrument (still in sealed envelopes) were
delivered by the secretarial staff to the
’ Use of a 7-point Likert-type format was justified in the present study
scale in a study of perceptions of auditor independence.
on the of
Knapp (19551, who used a -/-point
AUDITOR INDEPENDENCE 709
researchers. Participating auditors were guar-
anteed full anonymity and confidentially.
Neither the researchers nor the senior partners
knew who participated in the study.
Following the analyses, feedback interviews
were conducted with senior audit partners.
Results of the study were also made generally
available through preliminary papers presented
at accounting association conferences.’
Sample
The study sample comprised sixty-one
respondents, consisting of experienced audi-
tors drawn from a cross-section of large multi-
national chartered firms located in a single
metropolitan centre (location 1). Care was
taken in the distribution of questionnaires to
ensure that there was no dominance by a par-
ticular firm. Fifty (89%) of the auditors origin-
ally contacted responded. The completed
responses came from twelve senior audit part-
ners, twenty-five senior managers/ managers,
eight assistant managers/ supervisors, and five
audit seniors. There were fifteen females and
thirty-five males. The respondents had a mean
age of 33 years and mean auditing experience
of 12 years. (One respondent incorrectly com-
pleted the DIT and was dropped from the
analysis.)
Results
Rest’s (1979) DIT measures level of moral
reasoning development using scores of Kohl-
berg’s stages 5 and 6. This is called the mean
P-score; it ranges from 0 to 70, and indicates
level of moral reasoning development. The
higher the P-score, the more highly developed
is the respondent’s moral reasoning.’
Rest (1979) recommended that the P-scores
can be used to categorize the auditors into
high, mid, and low groups, corresponding,
respectively, to Kohlberg’s (1969) post-conven-
tional, conventional, and preconventional
levels of moral reasoning development.* Based
on a large normative sample, Rest determined
that P-scores of 27 and 41 were the appropriate
cut-off scores to define the three groups. These
values were adopted in the present study,
where the overall mean P-score was 30.6 (s.d.
=
16.1).9 The auditors were thus divided into
low (n =24, mean =17.42, s.d. =9.51), mid (n
= 11, mean =34.42, s.d. =3.93) and high (n =
14, mean =53.27, s.d. =8.73) moral reasoning
development groups.
Calculation of scores on Collins’ (1974) just/
unjust world dimension was done using Col-
lins’ original factor structure, which included
11 items, giving a range of 11 to 77. The
mean score on this was 40.2 (s.d. = 7.02)
with Cronbach alpha = 0.73. The auditors
were divided into believers and non-believers
in a just world on the basis of the mean score.
Mean scores for the two groups were 35.41
(just world beliefs: n = 29, s.d. = 4.08) and
46.91 (unjust world beliefs: n =21, s.d. =4.15).
The dependent variable in the analysis was a
measure of auditors’ acquiescence to the cli-
ent’s demand concerning materiality conflict.
6 Preliminary results of the present study were given in an unpublished paper presented at the Annual Confwence of&e
Accounti ng Associ ati on of Austral i a and New Zeal and, Darwin (July 1993).
’ A second score, termed the M-score (for meaningless) is calculated as means of identifying unreliable respondents. Rest
(1979) recommends that respondents whose M-score is 4 or more be eliminated from the sample. In the present study,
however, no respondents were eliminated in this fashion.
s Ponemon and Gabhart (1990) used a simple median split, and therefore did not tap into the three-level definition of moral
reasoning development given by Kohlberg.
9 This is substantially lower than Rest’s (1979) general community norms which were all greater than 40, and is also less
than Gaa’s (1992) list of DIT scores obtained across a range of studies involving accountants. It is, however, similar to the
mean score of 32.6 obtained by Ponemon and Gabhart (1990), and suggests that, although on the low side, P-scores for the
present sample fitted within the mid-group limits defined by Rest (1979).
710 C. A. WINDSOR and N. M. ASHKANASY
TABLE 1. Mean and standard deviations on acquiescing scores for treatment levels of the dependent variables in the initial
studv
Treatment levels 1 2
Financial conditions (Good) 3.31 (1.47) (Poor) 2.34 (1.61)
Size of audit fees
(Large)
2.97 (1.64) (Small) 2.68 (1.57)
Tendering
(Yes)
2.84 (1.62)
(No)
2.81 (1.61)
TABLE 2. Mean and standard deviations on acquiescing scores for the interaction in the initial study, showing simpleeffect
F-ratios
Level of moral develonment
Beliefs about a
iust world
Financial condition
Good Poor
High
F(FC)a, d.f. = 1,12
F(FC X BJW)b, d.f. = 1,12
Mid (tender condition)
F(FC), d.f. = 1,9
F(FC X BJW), d.f. = 1,9
Mid (No-tender condition)
F(FC), d.f. = I,9
F(FC X BJW), d.f. = 1,9
Low
F(FC), d.f. = 1,22
F(FC X BJW), d.f. = 1,22
Just 2.94 (1.13)
Unjust 3.47 (1.64)
Just
Unjust
Just
Unjust
Just
Unjust
2.70 (1.48)
4.17 (1.63)
3.00 (1.58)
3.50 (1.26)
3.67 (1.61)
3.11 (1.23)
2.42 (2.31)
3.19 (2.42)
0.82
0.08
1.70 (0.45)
1.83 (0.68)
19.03”’
3.14’
1.80 (0.45)
2.00 (0.89)
15.16***
0.18
2.61 (1.55)
2.11 (1.16)
1 j.38”
0.05
*p < 0.10, **p < 0.02, ***p < 0.01.
‘F(FC) = Simple effect Fratio for Financial Condition main effect.
‘F(FC X UJW) = Simple effect F-ratio for Financial Condition X Unjust World Belief interaction
Results of the ANOVA indicated main effects on
acquiescence for financial condition, (F =
21.85, d.f. = 1,43, p < 0.01) and for size of
fee (F = 16.23, d.f. = 1,43, p < 0.01). There
was no effect for tendering, however (F =
0.79, d.f. = 1,43). Mean acquiescing scores for
each level of the economic variables are shown
in Table 1.
There was a 4-way interaction between audit
conflict variables of tendering and client tinan-
cial condition, belief in a just world variable
and moral reasoning development (F = 6.50,
d.f. = 2,43, p < 0.01). lo Mean scores on acquies-
cence for the groups involved in this interac-
tion are given in Table 2, together with simple
effect statistics for each level of moral reason-
ing. The interaction between moral reasoning
score, just world beliefs, and client manage-
ment power (financial condition and tender)
is illustrated in Fig. 2.” The vertical axis in
I” Bartlett tests showed no violation of the heterogeneity of variance assumption.
t’ Respondents actually completed all 46 items of the Collins locus of control (LOC) scale. Separate ANOVAs were done
using the LOC total score as the independent variable to check if the effects reported in this paper were a result of just
world beliefs, or an artifact of LOC. These analyses resulted in no LOC effects, supporting our contention that the results
stem from belief in a just/unjust world.
AUDITOR INDEPENDENCE 711
High moral reasoning
(Autonomous) auditors
Low moral reasoning
(Pragmatic) auditors
+
Mid modal reasoning
(Accommodating) auditors
: - No interaction with tender
i - - - - Interaction with tender
i
:
1
Just world Unlust world
beliefs beliefs
‘Scale = 4 + (Ap - Ag). where Ap = acquiescence score for poor financial
condition; and Ag = acquescence score for good financial condltion
Fig. 2. Auditor resistance to client management power
(financial size condition) in the initial study.
this figure has been resealed to indicate audi-
tor resistance to client management power
(financial condition), calculated as 4 + (Ar -
As>, where A, = acquiescence score for poor
financial condition; and A, = acquiescence
score for good financial condition. The four-
way interaction involving the tender condition
was present in respect of the responses of the
mid-moral reasoning group only (simple effect
F =7.21, d.f. = 1,9,p
In two studies, practising auditors’ responses to hypothetical audit conflict scenarios were used to test
hypotheses that moral reasoning development and belief in a just world influence resistance to client
management power. Based on a mixed factorial ANOVA design, results confirmed that auditor independence
is interactively determined.
Pergamon Accounttng, Organizaffons and Society, Vol. 20, No. 7/S, pp. 701-720, 1995
Copyright 0 1995 Elsevier Science Ltd
Printed in Great Britain. All rights reserved.
0361-3682/ 95 $9.50+0.00
03613682(95)00018-fii
THE EFFECT OF CLIENT MANAGEMENT BARGAINING POWER, MORAL
REASONING DEVELOPMENT, AND BELIEF IN A JUST WORLD ON AUDITOR
INDEPENDENCE*
CAROLYN A. WINDSOR
Gr@tb University (Nathan Campus)
and
NEAL M. ASHKANASY
The University of Queensland
Abstract
In two studies, practising auditors’ responses to hypothetical audit conflict scenarios were used to test
hypotheses that moral reasoning development and belief in a just world influence resistance to client
management power. Based on a mixed factorial ANOVA design, results confirmed that auditor indepen-
dence is interactively determined. Three styles of auditor decision-making emerged: “autonomous”,
“accommodating” and “pragmatic”. Autonomous auditors were responsive to personal beliefs, such
that those with strong just world beliefs were more likely to resist client management power. Pragmatic
auditors were responsive to client management power, irrespective of beliefs. Accommodating auditors
responded both to personal beliefs and client management power; they comprised the least resistant
group, especially when they believed in an unjust world.
The ethical dilemmas faced by authors who
may be forced to compromise their indepen-
dence when pressured by client management
power provided the focus of the research
described in this paper. These issues are cur-
rently attracting attention in the academic lit-
erature (e.g. Ponemon & Gabhart, 1990) and
carry implications for audit practitioners (see
Ashkanasy & Windsor, 1994; Godsell, 1993).
The paper extends Ponemon and Gabhart’s
(1990) study of auditor independence judg-
ments, by including economic and personal
belief variables in the context of an audit con-
flict with client management. We argue that
the process of auditor independence is com-
plex and that no one factor can provide an
instant remedy.
In this paper, the concept of auditor inde-
pendence is first developed, based on a com-
plex decision-making process, which incor-
porates Kohlberg’s (1969) moral reasoning
development construct, and Lemer’s (1980)
concept of belief in justice. In order to over-
come concerns about the internal and external
validity of results based on hypotheses invol-
ving high-level interactions (Keppel, 1989), a
two-stage mixed factorial repeated-measures
design was employed in an examination of
l The authors acknowledge with thanks Ferdinand Gul, Errol Iselin, and Lawrence Ponemon for their contributions and
encouragement. We also thank the anonymous reviewers for their comments on an earlier draft of this paper which was
presented (in poster form) at the American Accounting Association ABO Research Conference, San Antonio, 10-l 1 March
1994, and the anonymous reviewers of the present version of this paper. The support and willing participation of public
accounting firms is acknowledged with thanks.
701
702 C. A. WINDSOR and N. M. ASHKANASY
auditors’ independence judgments in the face
of client management bargaining power.
Audi tor i ndependence
Auditors are required to remain independent
of the client but at the same time depend upon
the client for their livelihood through fees
(Mautz & Sharaf, 1961). An audit conflict may
develop, however, when auditors request dis-
closure of information that client management
may not wish to reveal in financial reports to
the public. The conflict becomes an ethical
dilemma when the auditor is confronted with
the decision to compromise independence and
integrity for economic gain. In the present
study, evidence about the existence of audit
dilemmas was collected through interviews
with senior audit partners from first-tier multi-
national public accounting firms. The partners
maintained that auditors have to balance pro-
fessional ethics with the economic practical-
ities of operating a viable business. Yet,
philosophically, auditors are supposed to serve
society, rather than to focus on economic com-
pensation (Gas, 1992; Mautz & Sharaf, 1961).
Gaa (1992) has argued further that there is a
“social contract” between the auditing profes-
sion and the rest of society which establishes
reasons for auditors to act ethically. Thus, audi-
tors are confronted with balancing “I” versus
“we” considerations (Etzioni, 1988). “I” repre-
sents the self interested, pragmatic considera-
tions of personal and economic gain; “we”, on
the other hand, encompasses professional and
ethical reponsibilities to remain independent.
The concept of auditor independence is inte-
gral to professional codes of ethics and stan-
dards which guide auditors in practice (Mautz
& Sharaf, 1961). Auditors must maintain inde-
pendence, defined by Mautz and Sharaf (1961)
as a “lack of bias in forming delicate judg-
ments” (p. 206). Auditor independence should
therefore be characterized by integrity and an
objective approach to professional work,
described by Mautz & Sharaf (1961) as “a state
of mind”. The present paper focuses specifi-
cally on independence, which is based on the
view that auditors are social human beings,
subject to a full range of emotions, beliefs,
and prejudices (Gas, 1992). This is in contrast
to the objective view, where the obligation is
imposed on auditors “to be fair, intellectually
honest, and free of conflicts of interest” (IFAC,
1990, Part A, para 1.1).
A key component in auditor independence is
the nature of the auditor-client relationship
(Goldman & Barlev, 1974). Research in this
area, however, has emphasized issues related
to firms switching auditors (e.g. Eichenscher
& Shields, 1983; Craswell, 1988) and issues
of individual auditor independence from audi-
tor and third party perceptions (e.g. Gul, 1991;
Knapp, 1985, 1987; Pany & Reekers, 1988;
Shockley, 1981). Until recently, however, little
research has focused specifically on auditors’
cognitions and psychological aspects of inde-
pendence (Bamber, 1993). Shaub et al . (1993)
have argued this is because the accounting pro-
fession has tended to rely on authoritative
codes of ethical conduct and practice stan-
dards. Recently, a number of researchers
(e.g. Lampe & Finn, 1992; Ponemon, 1992;
Ponemon & Gabhart, 1990, Shaub et al .,
1993) have begun to look at cognitive pro-
cesses as a means of seeking a deeper under-
standing of auditor independence decision-
making. For example, Ponemon and Gabhart
(1990) examined the impact of moral reason-
ing development on auditor decision-making,
and proposed that three levels of auditor inde-
pendence could be identified, corresponding
to Kohlberg’s (1969) levels of moral reasoning
development. Cushing (1990), however, has
argued that the cognitive perspective, taken
in isolation, may be overly restrictive, and
needs to be incorporated into more compre-
hensive models of decision-making. He sug-
gests that Ponemon and Gabhart’s (1990)
model could be extended by including the
effect of economic variables in a materiality
judgment situation. We argue further that per-
sonal beliefs about justice also play an impor-
tant role in the way people process important
information (Lerner, 1991). Lerner’s (1980,
1981) concept of belief in justice is relevant
in the auditor independence context because
AUDITOR INDEPENDENCE 703
justice is an inseparable component of integrity
(Karniol & Miller, 1981). The aim of the pre-
sent research, therefore, was to extend our
understanding of auditor independence judg-
ments by an examination of the combined
impact of psychological factors and client eco-
nomic power.
Complex decision-making
Mautz and Sharaf (1961) described auditor
independence as a “state of mind”, but this
simple definition does not reveal the complex-
ities of the auditor’s decision-making process.
Shaub et al. (1993) for example, adapted Hunt
and Vitell’s (1986) model of general theory of
marketing ethics, and showed that auditors’
sensitivity to ethical dilemmas depends upon
their attitudes to moral rules. Lampe and Finn
(1992) developed a “five-element” decision-
making model by adapting Jones’ (1991) issue
contingent model. The fiveelement model
combines three cognitive developmentalist
stage models and was shown to describe ethi-
cal decision-making better than a “Code-
implied” model. Bonner and Pennington
(1991) also argued for adoption of models of
auditor decision-making based on cognitive
audit task analysis.
The cognitive-based models (as described by
Lampe & Finn, 1992) however, do not take
into account the wide range of factors poten-
tially affecting auditor decision-making. For
example, Gaa (1992) criticized Lampe and
Finn’s (1992) five-element model for its failure
to address the full range of complexities of
auditor independence judgments, and suggests
a more complicated explanation based on Tre-
vino’s (1986) interactionist model. In effect,
Gaa (1992) calls for a more sophisticated
theoretical model of decision-making which
goes beyond cognitions. The interactionist
approach focuses on combinations of personal
and situational variables (rather than on single
variables acting alone, see Trevino, 1986) and
thus provides a more sophisticated and realistic
indication of human decision-making processes
in real situations (which are described in terms
of high-order interactions, see Keppel, 1989).
However, while the Trevino (1986) model
provides an appropriate theoretical basis, it
gives little guidance in respect of empirical
testing, and therefore needed to be adapted
for use in the present study. This was accom-
plished using a model developed by Ashkanasy
(1989) in the context of evaluative decision-
making. This model (see also Ashkanasy,
1991) is based on a general information proces-
sing framework in which cognitions are trig-
gered by contextual information, and are
affected by deeply embedded psychological
beliefs when an immediate decision is difficult
to achieve. In essence, the model posits that
impressionistic and cognitive processes are
influenced by personal beliefs, which are in
turn reflected in interactions between situa-
tional contingencies and cognitive processes.
The model is also consistent with findings by
Brief et al. (1991) that personal values interact
with perceptions of accountability and explicit-
ness to determine responses to ethical dilem-
mas in managerial decision-making.
By substituting the variables from Trevino’s
(1986) conceptual model of ethical decision-
making for those in the original Ashkanasy
(1989) model, it was possible to derive a testa-
ble model in the present instance. In this case,
the three variables were: (1) economic dimen-
sions of client bargaining power; (2) moral rea-
soning development (Kohlberg, 1969); and (3)
belief in a just world (Lerner, 1980, 1981). We
expected to find that auditors’ decisions in
response to client management economic bar-
gaining power would be interactively influ-
enced by cognitive processes involving moral
reasoning and beliefs.
Client management bargaining power
Client management has an inherent bargain-
ing advantage because auditors depend upon
client fees for their livelihood (Mautz &
Sharaf, 1961). At the same time, auditors may
need to qualify their client’s financial report in
a manner which could harm the client manage-
ment’s interests (Goldman & Barlev, 1974).
Further, Hill and Jones (1992) have shown
that client management has the bargaining
704 C. A. WINDSOR and N. M. ASHKANASY
advantage by virtue of corporate management’s
position at the nexus of implicit and explicit
contracts that constitute the firm. Antle and
Nalebuff (1991) have argued also that client
management has superior knowledge of the
firm’s income. Using a game-theory model of
client-auditor negotiations of financial state-
ments, they demonstrated that client manage-
ment always has the negotiating edge.
Auditor independence is further threatened
by the very nature of the relationship between
client management and the auditor, which is
fostered by professional practice standards.
These standards require auditors to have a
sound knowledge of their clients’ businesses
and industries, hence setting the scene for a
close professional relationship, and likely to
result in reduced independence. The auditor-
client management relationship is strengthened
over time as the auditor develops specialized
expertise (which would lose its value if
applied to another relationship, see Levinthal
& Fichman, 1988).
auditor’s dependence on client fees as a poten-
tial economic factor that could affect auditor
independence. This is supported by Gul
(1991) who found that, when fees are a signif-
icant proportion of the auditors’ total revenue,
third party perceptions of the auditors’ ability
to withstand management power are adversely
affected. Gul (1991) explained this finding in
terms of Knapp’s (1985) observation that audi-
tors are more responsive to client management
power when the client provides a significant
proportion of the auditor’s income.
In the present study, we have conceptua-
lized client bargaining power in terms of situa-
tional factors that exert pressure on auditors,
and therefore influence their independence
behaviour. The initial interviews with senior
audit partners indicated that the three primary
dimensions of client management power were
(1) client financial condition, (2) size of client
fees, and (3) whether or not the client calls
tenders for audit work.
Tendering. When client firms invite prospec-
tive auditors to tender for provision of audit
services at a competitive cost, there is a risk
that auditors may experience difficulty in
remaining independent. This is because client
management can overtly or covertly threaten to
replace the auditor. Craswell (1992) for exam-
ple, showed that the introduction of tendering
led to an increase in the frequency of auditor
switches and fee reductions. Hence, client
management can use tendering to place more
economic pressure on the auditor, thus impair-
ing independence.
Moral reasoning development
Financial condition of the client. Auditors
dealing with clients in poor financial condition
are likely to face a higher level of legal liability
than if the client was in good financial condi-
tion (Farmer et al., 1987; Palmrose, 1987). In
this respect, Knapp (1985) found also that
financial statement users perceive that manage-
ment is more likely to obtain its preferred reso-
lution in a conflict when the client is in a good,
rather than poor financial condition. It follows,
therefore, that auditors would be likely to exer-
cise greater caution when dealing with clients
who are in poor financial condition.
The present study utilized Kohlberg’s (1969)
moral reasoning construct. Although this con-
struct has been subject to some controversy
(see Modgil & Modgil, 1986) it has been used
successfully as a cognitive measure in a number
of studies in accounting and organizational set-
tings over recent years (Gaa, 1992; Ponemon,
1992; Ponemon & Gabhart, 1990; Trevino,
1986). Gaa (1992) argued that the construct is
a measure of “moral expertise” in the specific
context of auditor judgments. It is based on a
fundamental assumption that moral reasoning
reflects an underlying organization of thinking
which becomes more complex and sophisti-
cated with the individual’s moral development
(Kohlberg, 1969). Rest (1986) argued further
that the word “moral” as used here involves
social interaction, and emphasized its applic-
ability to concepts of justice or fairness.
Size of the client fee. Mautz and Sharaf Kohlberg’s theory includes three broad
(1961) drew attention to the problem of the levels of moral reasoning development, each
AUDITOR INDEPENDENCE 705
composed of two stages. Moral reasoning
development involves the individual’s transi-
tion from stage to stage in an invariant, irrever-
sible sequence. At the preconventional level,
stages one and two, an individual is concerned
with concrete consequences, particularly
reward and punishment affecting individual
immediate interests. At the conventional level,
or stages three and four, “right” is conforming
to the expectations of acceptable behaviour of
significant others; for example, larger society, a
peer group or family.
At the post-conventional level, or stages five
and six, “right” is determined by universal
values or principles; for example, saving
human life. When laws violate these princi-
ples, the individual may act in defiance of the
law. The individual at this level sees beyond
social norms, laws, and authority of groups
and indivdiuals regarding judgments of right
and wrong.
Ponemon and Gabhart (1990) used Kohl-
berg’s moral reasoning construct in a study of
auditor independence under varying condi-
tions of internal sanction. They found that
moral reasoning development affected
response to penalty sanctions; auditors with
low (pre-conventional) levels of moral reason-
ing development were more sensitive to pen-
alty sanctions than auditors with higher levels
of moral development. These results were
based on Rest’s (1979) “P-score”, which is a
continuous measure, rather than a stage mea-
sure. Ponemon and Gabhart (1990) maintain
that P-scores are strongly associated with Kohl-
berg’s stages (see Colby & Kohlberg, 1987)
and constitute a satisfactory operationalization
of the Kohlberg stage concept. In the present
study, we anticipated that higher levels of
moral reasoning (measured using P-scores)
would be associated with greater resistance to
client management bargaining power.
Personal beliefs in a just world
Finally, personal beliefs were operationalized
in the present study using a measure of belief in
a just world. This construct is appropriate to
test auditor independence because justice and
integrity are intimately intertwined (Karniol &
Miller, 1981). Lerner (1991) argues that the just
world motive is a tendency to believe the
world as operating in a consistent, just manner
where the good are rewarded and the bad are
punished. The belief in justice operates at a
preconscious level and is revealed indirectly
in the person’s reactions to a given event (Ler-
ner, 1980, 1991). Research has convinced Ler-
ner and his colleagues that the commitment to
justice in people’s lives is not simply a principle
governing how to view their situation, but as
the central organizing principle in how they
acquire and allocate desired resources (Ler-
ner, 1991). At a conscious level, people with
just world beliefs appear to others and them-
selves to have a realistic view of injustices, but
plan their lives as if they live in a just world as
means of bringing order to their lives and cop-
ing with life’s tragedies. Thus, individuals with
just world beliefs are committed to a personal
contract that presupposes people deserve their
outcomes and receive, in the long run, what
they are entitled to. Hence, this personal con-
tract becomes a catalyst for the individual to
relinquish immediate outcomes for long-term
rewards. In contrast, people with unjust world
beliefs believe that people are not rewarded or
punished in accordance with their actions, and
feel that they are victims. These people there-
fore behave as if they have little control over
life’s events.
To measure justice beliefs, we used the belief
in a just/ unjust world dimension identified by
Collins (1974). Lerner (1980) has suggested
this is an appropriate measure of beliefs in
justice. It is a dimension of Rotter’s (1966)
locus of control scale, which measures a
personality construct derived from social
learning theory (Rotter, 1954). Research has
shown that Rotter’s (1966) locus of control
construct is multidimensional (e.g. Collins,
1974) and subsequent studies have specifically
replicated Collins’ belief in a just world dimen-
sion (Ashkanasy, 1985; Zuckerman & Gerbasi,
1977a,b; Zuckerman et al. 1977). Zuckerman
and colleagues have provided evidence that
706 C. A. WINDSOR and N. M. ASHKANASY
the unjust/ just world dimension is reliable,
valid, and independent of the other locus of
control dimensions.
Hypotheses
Based on the Ashkanasy (1989) and Trevino
(1986) models, we have argued that auditor
independence is affected by moral reasoning
development and personal beliefs embedded
in the preconscious and activated by ethical
dilemmas. Thus, when auditors are pressured
by client management bargaining power, the
extent to which personal beliefs affect their
independence is determined by the auditor’s
level of moral reasoning development. We
expected to find in this study that auditors
with highly developed moral reasoning would
be influenced by their personal beliefs about a
just world, while those with a lower level
would be influenced primarily by client man-
agement bargaining power. Auditors between
these extremes were expected to exhibit char-
acteristics of both of these groups. In essence,
a three-way interaction was predicted, demon-
strating that personal beliefs and moral reason-
ing interact with client management power in
complex decision-making situations. The nat-
ure of this interaction is illustrated in Fig. 1.’
Specific hypotheses are:
(31)
-
High moral reasoning auditors’ resistance to client
management power is affected by just world
beliefs. They are less likely to resist when they
believe in an unjust world than when they believe
in a just world.
Low moral reasoning auditors are affected by client
management power, but not by just world beliefs.
Overall, they are less likely than high moral reason-
ing auditors to resist client management power.
High moral reasonmg
(Post-conventional) auditors
Low moral reasoning
(Pie-conventlonat) auditors
Mid moral reasoning
(Conventional) auditors i
Just world
beliefs
Unjust world
beliefs
Hypothesized pattern of auditor resistance to client
management power.
Analysis method
Mid-moral reasoning auditors are affected by client
management power and just world beliefs, They
are less likely than high moral reasoning auditors
to resist client management power. Further, they
are less likely to resist when they believe in an
unjust world than when they believe in a just
world.
The purpose of the present study was not
to deal just with auditors’ responses to eco-
nomic factors (e.g. GUI, 1991; Knapp, 1985).
Nor did we intend to address the effect of
psychological factors alone, as was done by
Ponemon and Gabhart (1990). Rather, the
focus in the present paper is on the i nterac-
ti on between client economic factors, moral
reasoning development, and beliefs in a just
world.
To test such interactive hypotheses, a mixed
factorial ANOVA research design with three
repeated measures and two between-group
’ The vertical axis in this figure is auditor resistance to client pressure (represented by the economic variables: client
financial condition, size of fees, tendering). As such it represents difference scores across levels of economic variables. The
graph therefore depicts a three-way interaction between economic variables (ordinal difference scores), just w or l d beliefs
(abscissor), and moral reasoning development group.
AUDITOR INDEPENDENCE 707
independent variables was appropriate.2 The
three repeated measures were the economic
factors identified by the senior audit partners:
(1) client financial condition; (2) size of fees;
and (3) tendering. The two between-group
independent variables were (1) level of moral
reasoning and (2) belief in a just world. The
dependent variable was a measure of the
extent to which the auditor would acquiesce
to the wishes of client management.
A problem in studies of high-order interac-
tions is that such interactions may be unstable
(Keppel, 1989). This research was therefore
carried out as two independent studies: a first
or initial study; and a second or validation study
based on a larger sample to increase the power
of analysis. While differences in the results of
the initial and validation studies were
expected, we hoped to be able to demonstrate
that the underlying theory would apply in both
instances.
INITIAL STUDY
Method
Experi mental i nstrument. This consisted of
a questionnaire booklet in three parts: (1) the
ethical dilemma; (2) Collins’ (1974) locus of
control (LOC) scale; and (3) Rest’s (1979) DIT
scale.3 The moral reasoning development and
just world belief measures were administered
after the ethical dilemma so that the scenarios
used in the DIT and the LOC measure did not
bias respondents’ perceptions of the audit con-
flict (see Trevino & Youngblood, 1990).
Ethi cal di l emma. Cushing (1990) suggested
using an audit conflict involving disclosure of
an accounting transaction that would have a
material effect on the client’s financial state-
ments. In the present study, this situation was
presented in short case-study form to test audi-
tor independence under pressure of client man-
agement power using three repeated measures.
The conflict provided in the booklet concerned
the materiaIity of certain unrecorded liabilities
discovered by the firm’s auditors. These liabil-
ities consisted primarily of expenses incurred
in the previous year which were neither paid
nor recorded until the following year. Respon-
dents were asked to assume that they were the
auditors of the hypothetical firm where client
management disagreed with the auditor that
the amount was material, with 8% materiality.
The figure of 8% materiality level was recom-
mended by the senior audit partners as a way to
create an element of realistic ambiguity, and to
ensure that the decision-making task required
professional judgment.*
The aim of the dilemma was to place audi-
tors in a conflict situation with client manage-
ment, whose economic bargaining power was
represented by eight scenarios. These were
varied by presenting three statements: client
financial condition (good or bad); size of fees
(large or small); and external audit tendered
(yes or no). Respondents were requested to
make a subjective judgment about the
dilemma using a seven-point Likert scale,
where 1 indicated a very low likelihood that
they would acquiesce to client management’s
‘ This approach was adopted in preference to regression for three reasons. The first is that Kohlberg’s theory implies
discrete groups, leading explicitly to group-specific hypotheses (see also Ponemon & Gabhart, 1990). The second reason is
that hypotheses are framed in terms of high-order interactions, which are difficult to interpret in regression models. Finally,
predictions relating to levels of moral reasoning are nonlinear, and therefore are inherently not amenable to analysis by
techniques based on linear regression. ANOVA focuses only on between-group differences and does not assume linearity
across factorial groupings (Keppel, 1989).
3 Copies of the instrument may be obtained by writing to C. A. Windsor, Faculty of Commerce and Administration, Griffith
University, Nathan, Queensland, 4111, Australia.
’ The materiality level was mid-way between the immateriality threshold limit of 5% and the materiality threshold limit of
10% (see AAS5, ASCPA, 1994).
705 C. A. WINDSOR and N. M. ASHKANASY
view, and 7 indicated a very high likelihood of
acquiescing. 5
The scenarios were generated by computer
to ensure that they were presented to respon-
dents in random order. Respondents were
requested to complete each of the scenarios
in sequence, without consulting colleagues.
The same instrument was subsequently tested
on a sample of 160 third-year auditing students.
In this instance, no significant effects of the
three economic variables were detected. This
suggests that responses to the scenarios by the
practising auditors were derived from the audi-
tors’ experience in dealing with client manage-
ment, rather than an artifact of the measure.
This finding is similar to that reported by
Farmer et al. (1987), who reported that practi-
cing auditors’ independence judgments were
affected by organizational acculturation.
Belief scale. This was the just world subscale
of Collins’ (1974) 46-item Likert-scale version
of Rotter’s (1966) internal-external control
questionnaire. The dimensional structure of
this version of Rotter’s scale was verified by
Ashkanasy (1985), and the just world subscale
has been separately validated by Zuckerman
and Gerbasi (1977a,b). Responses to each of
the items were based on a seven-point Likert
scale, where 1 indicated strong disagreement,
and 7 indicated strong agreement.
Defining issues test (DI D. The short (3-
story) version of Rest’s (1979) scale was used.
This is based on Kohlberg’s (1969) theory of
moral reasoning development, and consists of
a number of brief stories that present moral
dilemmas in everyday life. The DIT has been
extensively validated, and was described by
McCrae (1985) as a paradigm of measurement
instruments (see Gaa, 1992). The 3-story ver-
sion of Rest’s scale was used in preference to
the Gstory version because of the time con-
straints on professional auditors. Rest (1979)
cites Cronbach alpha reliability for the 3-story
DIT typically around 0.7. Respondents answer
questions concerning the importance of con-
sidering various aspects of each story using
five-point response scales, where 1 indicates
least importance, and 5 indicates greatest
importance. Scoring of responses is based on
a formula set out by Rest (1979).
Procedure. Prior to administering the
research instruments, a three-stage strategy
was used to strengthen internal validity and
to overcome data collection problems. First,
preliminary interviews were conducted with
six senior audit partners (one from each of
the big six audit firms) to assess the nature of
audit conflict problems, and to gain the part-
ners’ co-operation. In the second stage, a draft
case study was circulated to the audit partners
to enlist their help in developing realistic sce-
narios, which appropriately represent client
management power. The advice and com-
ments provided by the senior audit partners
were used to compile the final versions of the
instrument. In the third stage, the audit part-
ners were briefed on how to distribute and
collect the questionnaire booklets so as to mini-
mize internal validity problems such as history
and maturation. The co-opted partners were
instructed not to participate in the experi-
ment, nor to advise participating auditors
about the substance of the questionnaires.
The questionnaire booklets were distributed
to experienced auditors using secretarial ser-
vices provided by the participating firms,
under supervision of the co-opted audit part-
ners. The secretarial staff offered the booklets
to participating auditors immediately on their
return from audit assignments, and provided no
indication that the study was dealing with ethi-
cal issues. Those agreeing to participate were
asked to complete the booklets within 30 min-
utes, and to return them (in sealed envelopes)
to the person who gave them out. Three weeks
after distribution to the firms, the completed
instrument (still in sealed envelopes) were
delivered by the secretarial staff to the
’ Use of a 7-point Likert-type format was justified in the present study
scale in a study of perceptions of auditor independence.
on the of
Knapp (19551, who used a -/-point
AUDITOR INDEPENDENCE 709
researchers. Participating auditors were guar-
anteed full anonymity and confidentially.
Neither the researchers nor the senior partners
knew who participated in the study.
Following the analyses, feedback interviews
were conducted with senior audit partners.
Results of the study were also made generally
available through preliminary papers presented
at accounting association conferences.’
Sample
The study sample comprised sixty-one
respondents, consisting of experienced audi-
tors drawn from a cross-section of large multi-
national chartered firms located in a single
metropolitan centre (location 1). Care was
taken in the distribution of questionnaires to
ensure that there was no dominance by a par-
ticular firm. Fifty (89%) of the auditors origin-
ally contacted responded. The completed
responses came from twelve senior audit part-
ners, twenty-five senior managers/ managers,
eight assistant managers/ supervisors, and five
audit seniors. There were fifteen females and
thirty-five males. The respondents had a mean
age of 33 years and mean auditing experience
of 12 years. (One respondent incorrectly com-
pleted the DIT and was dropped from the
analysis.)
Results
Rest’s (1979) DIT measures level of moral
reasoning development using scores of Kohl-
berg’s stages 5 and 6. This is called the mean
P-score; it ranges from 0 to 70, and indicates
level of moral reasoning development. The
higher the P-score, the more highly developed
is the respondent’s moral reasoning.’
Rest (1979) recommended that the P-scores
can be used to categorize the auditors into
high, mid, and low groups, corresponding,
respectively, to Kohlberg’s (1969) post-conven-
tional, conventional, and preconventional
levels of moral reasoning development.* Based
on a large normative sample, Rest determined
that P-scores of 27 and 41 were the appropriate
cut-off scores to define the three groups. These
values were adopted in the present study,
where the overall mean P-score was 30.6 (s.d.
=
16.1).9 The auditors were thus divided into
low (n =24, mean =17.42, s.d. =9.51), mid (n
= 11, mean =34.42, s.d. =3.93) and high (n =
14, mean =53.27, s.d. =8.73) moral reasoning
development groups.
Calculation of scores on Collins’ (1974) just/
unjust world dimension was done using Col-
lins’ original factor structure, which included
11 items, giving a range of 11 to 77. The
mean score on this was 40.2 (s.d. = 7.02)
with Cronbach alpha = 0.73. The auditors
were divided into believers and non-believers
in a just world on the basis of the mean score.
Mean scores for the two groups were 35.41
(just world beliefs: n = 29, s.d. = 4.08) and
46.91 (unjust world beliefs: n =21, s.d. =4.15).
The dependent variable in the analysis was a
measure of auditors’ acquiescence to the cli-
ent’s demand concerning materiality conflict.
6 Preliminary results of the present study were given in an unpublished paper presented at the Annual Confwence of&e
Accounti ng Associ ati on of Austral i a and New Zeal and, Darwin (July 1993).
’ A second score, termed the M-score (for meaningless) is calculated as means of identifying unreliable respondents. Rest
(1979) recommends that respondents whose M-score is 4 or more be eliminated from the sample. In the present study,
however, no respondents were eliminated in this fashion.
s Ponemon and Gabhart (1990) used a simple median split, and therefore did not tap into the three-level definition of moral
reasoning development given by Kohlberg.
9 This is substantially lower than Rest’s (1979) general community norms which were all greater than 40, and is also less
than Gaa’s (1992) list of DIT scores obtained across a range of studies involving accountants. It is, however, similar to the
mean score of 32.6 obtained by Ponemon and Gabhart (1990), and suggests that, although on the low side, P-scores for the
present sample fitted within the mid-group limits defined by Rest (1979).
710 C. A. WINDSOR and N. M. ASHKANASY
TABLE 1. Mean and standard deviations on acquiescing scores for treatment levels of the dependent variables in the initial
studv
Treatment levels 1 2
Financial conditions (Good) 3.31 (1.47) (Poor) 2.34 (1.61)
Size of audit fees
(Large)
2.97 (1.64) (Small) 2.68 (1.57)
Tendering
(Yes)
2.84 (1.62)
(No)
2.81 (1.61)
TABLE 2. Mean and standard deviations on acquiescing scores for the interaction in the initial study, showing simpleeffect
F-ratios
Level of moral develonment
Beliefs about a
iust world
Financial condition
Good Poor
High
F(FC)a, d.f. = 1,12
F(FC X BJW)b, d.f. = 1,12
Mid (tender condition)
F(FC), d.f. = 1,9
F(FC X BJW), d.f. = 1,9
Mid (No-tender condition)
F(FC), d.f. = I,9
F(FC X BJW), d.f. = 1,9
Low
F(FC), d.f. = 1,22
F(FC X BJW), d.f. = 1,22
Just 2.94 (1.13)
Unjust 3.47 (1.64)
Just
Unjust
Just
Unjust
Just
Unjust
2.70 (1.48)
4.17 (1.63)
3.00 (1.58)
3.50 (1.26)
3.67 (1.61)
3.11 (1.23)
2.42 (2.31)
3.19 (2.42)
0.82
0.08
1.70 (0.45)
1.83 (0.68)
19.03”’
3.14’
1.80 (0.45)
2.00 (0.89)
15.16***
0.18
2.61 (1.55)
2.11 (1.16)
1 j.38”
0.05
*p < 0.10, **p < 0.02, ***p < 0.01.
‘F(FC) = Simple effect Fratio for Financial Condition main effect.
‘F(FC X UJW) = Simple effect F-ratio for Financial Condition X Unjust World Belief interaction
Results of the ANOVA indicated main effects on
acquiescence for financial condition, (F =
21.85, d.f. = 1,43, p < 0.01) and for size of
fee (F = 16.23, d.f. = 1,43, p < 0.01). There
was no effect for tendering, however (F =
0.79, d.f. = 1,43). Mean acquiescing scores for
each level of the economic variables are shown
in Table 1.
There was a 4-way interaction between audit
conflict variables of tendering and client tinan-
cial condition, belief in a just world variable
and moral reasoning development (F = 6.50,
d.f. = 2,43, p < 0.01). lo Mean scores on acquies-
cence for the groups involved in this interac-
tion are given in Table 2, together with simple
effect statistics for each level of moral reason-
ing. The interaction between moral reasoning
score, just world beliefs, and client manage-
ment power (financial condition and tender)
is illustrated in Fig. 2.” The vertical axis in
I” Bartlett tests showed no violation of the heterogeneity of variance assumption.
t’ Respondents actually completed all 46 items of the Collins locus of control (LOC) scale. Separate ANOVAs were done
using the LOC total score as the independent variable to check if the effects reported in this paper were a result of just
world beliefs, or an artifact of LOC. These analyses resulted in no LOC effects, supporting our contention that the results
stem from belief in a just/unjust world.
AUDITOR INDEPENDENCE 711
High moral reasoning
(Autonomous) auditors
Low moral reasoning
(Pragmatic) auditors
+
Mid modal reasoning
(Accommodating) auditors
: - No interaction with tender
i - - - - Interaction with tender
i
:
1
Just world Unlust world
beliefs beliefs
‘Scale = 4 + (Ap - Ag). where Ap = acquiescence score for poor financial
condition; and Ag = acquescence score for good financial condltion
Fig. 2. Auditor resistance to client management power
(financial size condition) in the initial study.
this figure has been resealed to indicate audi-
tor resistance to client management power
(financial condition), calculated as 4 + (Ar -
As>, where A, = acquiescence score for poor
financial condition; and A, = acquiescence
score for good financial condition. The four-
way interaction involving the tender condition
was present in respect of the responses of the
mid-moral reasoning group only (simple effect
F =7.21, d.f. = 1,9,p