Perceptions of auditors independence

Perceptions of Auditors' Independence: An Empirical Analysis
Author(s): Randolph A. Shockley
Source: The Accounting Review, Vol. 56, No. 4 (Oct., 1981), pp. 785-800
Published by: American Accounting Association
Stable URL: .
Accessed: 09/06/2014 18:57
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
/>
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact

.

American Accounting Association is collaborating with JSTOR to digitize, preserve and extend access to The
Accounting Review.



This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


THE ACCOUNTING
Vol. LVI, No. 4
October 1981

REVIEW

Perceptions

An

of

Auditors'

Empirical

Independence:
Analysis

RandolphA. Shockley
ABSTRACT: Recent challenges of the audit independence assumption have impelled
the accounting profession to consider ways of improving the credibility of audit reports.
This study examines the perceived effects of competition, MAS, audit-firm size, and tenure
on the risk that audit independence may become impaired. Factorial analysis of variance
techniques are used to analyze judgments obtained in an experimental task from four
subject groups: (a) Big Eight partners, (2) partners from local and regional CPA firms,
(3) commercial loan officers, and (4) financial analysts. Results indicate that audit firms
operating in highly competitive environments, firms providing MAS, and smaller audit
firms are perceived as having a higher risk of losing independence. An audit firm's tenure
with a given client is not significant. Though perceptual differences exist between the
groups, an overall analysis ranks competition as the most important factor, followed by
audit-firm size and MAS.

value of auditing services depends upon the fundamental assumption that certified public
accountants are independent of their
clients. Experience has shown, however,
that independence in fact is insufficient to
support the credibility of this assumption.

Concern about a potential lack of independence in auditors exists within
government [U.S. Senate, "Metcalf Staff
Report," 1976; U.S. SEC Report to
Congress, 19791 and in the financial
community [AICPA, "Cohen Commission Report," 1978, p. 94]. This concern
threatens the viability of the auditor's
role in society, for credibility depends
ultimately on the perception rather than
on the fact of independence.'
The publication of the Metcalf Staff
Report on "The Accounting Establishment" [U.S. Senate, 1976] spurred increased interest in the possibility of
nonindependence. In its own report
[19771, the subcommittee clearly stated
its position on independence:
T

Because of the economic and social environment which now exists, the subcommittee
believes it is timely for the public and the
accounting profession to reassess the role
which independent auditors should play in
making the Nation's economic system function effectively.... Above all, the auditor's
essential qualities of independence and professionalism must be strengthened and adapted
to the present environment [p. 5].

HE

' The importance of the appearance of independence
is well established in the AICPA Code of Professional
Ethics, ET section 52 [1978].
This paper is based on my dissertation at the University of North Carolina. I would like to thank the following faculty members there for their comments and

assistance: Howard 0. Rockness, Lewis F. Davidson,
Edward J. Blocher, William D. Perreault, Jr., Dannie J.
Moffie, and Harold Q. Langenderfer. The comments of
Carl S. Warren are also gratefully acknowledged.

Randolph A. Shockley is Assistant
Professor of Accounting, University of
Georgia.
Manuscript received January 1980.
Revisions received August and November 1980.
AacceptedJanuary 1981.

785

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


The Accounting Review, October 1981

786
As concern about independence has
increased, the profession has come under
increasing pressure to take positive steps
to minimize any potential for nonindependence. The current controversy over
the advisability of restrictions on management advisory services (MAS) is one
obvious example [Burton, 1980, p. 48].
However, empirical evidence useful in
evaluating the potential impact of such
steps is limited. As a result, the accounting profession has been obliged to rely

too heavily on conjecture and assertion
in its efforts to establish effective policy.
The objective of this study is to provide
additional empirical evidence relevant to
perceptions of auditors' independence.
Specifically, the study examines the effects
of four variables on the perceptions of
four subject groups with respect to audit
independence. The four variables are (1)
competition for audit clients, (2) whether
management advisory services (MAS)
are provided to audit clients, (3) the size
of the audit firm, and (4) the tenure of an
audit firm with a given client. The four
groups are (1) partners from Big Eight
auditing firms, (2) partners from local or
regional auditing firms, (3) bank commercial loan officers, and (4) financial
analysts. It is hypothesized that accountants and financial-statement users perceive a higher risk of a loss of independence for CPA firms which operate in
high-competition environments, render
MAS, are smaller in size, or have longer
periods of tenure. Interactions between
certain of these factors are also hypothesized. The hypotheses are examined by
using factorial analysis of variance tech-

acterizedby threeselectioncriteria.First,
they arepresentin all auditengagements;
i.e., they are characteristicsof the audit
environment. Second, their effects on
independencehypotheticallymay be influencedby the institutionof appropriate
policy. Third, they may be associated

with audit firms rather than with individuals. Alternative hypotheses associated with each variableappearbelow:

niques.

Two additional hypotheses specify
interactionsbetween (1) size and MAS,
and (2) size and competition.
H5: The perceivedrisk of an impairment of independencedue to the
renderingof MAS is greaterfor

EXPERIMENTALVARIABLESAND
HYPOTHESES

The independent variables were chosen
from a larger set identified by a review of
the accounting literature. They are char-

Hi: CPA firms operating in an
environmentcharacterizedby a
highlevelof competitionfor audit
clients are perceivedas having a
greaterrisk of losing their audit
independencethan are CPAs operatingin a low-competitionenvironment.
H2: CPA firms rendering management advisory services to their
audit clients are perceived as
having a greater risk of losing
theirauditindependencethan are
CPAsnot renderingsuchservices.
H3: Large CPA firms (those which
perform SEC audits) are perceived as having a smallerrisk of

losing their audit independence
than are smaller firms (those
which do not perform SEC audits).
H4: CPA firmswhichhave performed
a given client'saudit for a period
exceeding five consecutive years
are perceivedas having a greater
risk of losing their audit independence than are firms which
have performed the audit for
five yearsor less.

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


Shockley

787

small audit firms than for larger
firms.
H6: The perceived risk of an impairment of independence due to
competition for audit clients is
greater for small audit firms than
for larger firms.

of this argument by testing the effect of
competition on perceptions of independence. For experimental purposes, it
implicitly defines the two levels of competition in a particular firm's geographic
area as "high" or "low.'"

The rationale for these six hypotheses
is developed in the following discussion
of the four exogenous variables.

MAS has been the most prevalent
topic for independence-related research.2
Most empirical work on the issue has
been survey research, the most widely
known and quoted group being the
studies of Schulte [1965], Briloff [1966],
Titard [1971], and Hartley and Ross
[1972]. In general, these studies report
conflicting results; their data can be used
to support both sides of the MAS
question.
For example, certain empirical results
of Schulte [1965, p. 590] and Hartley and
Ross [1972, p. 44] may be interpreted as
support for the position that rendering
MAS increases an audit firm's independence. These results are consistent with
an argument presented by Goldman and
Barlev [1974, p. 715]. Goldman and
Barlev's analysis suggests that the offering of nonroutine services makes the
auditor more nearly unique to the client,
thereby increasing the auditor's value to
and relative power over the client. Consequently, the audit firm is better able to
resist client pressure and so is more likely
to maintain its independence.
However, evidence and arguments opposing the provision of MAS to audit

clients are more common. Schulte [1965,
p. 593] lists four arguments which are
representative. First, CPAs providing
consulting services may become advocates of the client. making it difficult to

Competition

Two points of view exist with respect to
the desirability of competition in the
auditing profession. For example, the
Metcalf subcommittee's staff believed
that too little competition exists, allowing
the Big Eight accounting firms too much
influence in the establishment of accounting principles and policy [U.S. Senate,
1976, pp. 35-46]. On the other hand, the
Cohen Commission states that "It is not
lack of competition, however, but possible excessive competition that appears to
present a problem to the public accounting profession today" [1978, p. 109].
The Cohen Commission's chief concern was competition's effect on the
extent and quality of audit services provided. However, an audit firm which
allows competition, and implicitly its
fee, to affect the nature of its audit may
also be dangerously close to nonindependence. As competition for audit clients increases, clients' opportunities and
incentives to replace incumbent auditors
also increase. Reasons for the change
may range from minimization of audit
fees to a search for a more compliant
auditor. Regardless, auditors' dependence on their clients may increase if they
believe that other auditing firms would be
happy to accept the engagement should a

client become displeased [Beams and
Killough, 1970, p. 17].
Hypothesis 1 examines the plausibility

Management Advisory Services

2 The MAS-independence controversy is now generally referred to as the "scope of services" issue. For an
excellent history and overview of the issue, see the Public
Oversight Board's Scope of Services by CPA Firms
11979].

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


788

The Accounting Review, October 1981

Consequently, it would be difficult to
interpret results based on a comprehensive definition of MAS. In this study,
MAS are defined for the subjects as the
"design and installation of financial and
cost accounting systems, budget and
inventory control systems, and other
accounting-related information systems." While this definition prevents
generalization to all consulting services
provided by CPAs, it should include the
majority of services typically provided.

dependent on a given client than is a
smaller firm because the associated fees
usually constitute a smaller proportion
of the audit firm's total resources [Mautz
and Sharaf, 1961, p. 213 ]. Second, certain
characteristics inherent in small audit
practices may increase the danger of
impairment, e.g., the nature of the typical
small firm client or the tendency toward
a more personal mode of service and
close relationships with the client.
Hypothesis 3 is consistent with the
direction of effect implicit in these arguments. The operational definition of
large and small firms is consistent with
the two practice divisions of the AICPA:
large firms are those which perform
audits of SEC-registered corporations,
while small firms do not perform such
audits.
Audit firm size may also modify the
effects of other variables in this study.
Such an interaction with the MAS
variable may occur in at least two ways.
First, the impact of losing a client and its
MAS revenue would not be as great for a
large firm as for a smaller firm [Hartley
and Ross, 1972, p. 48]. Second, large
CPA firms usually have separate MAS
departments. Separation of auditors and
consultants would act to reduce any

negative effect of MAS on audit independence. In small firms, however, the
auditor and the consultant are often the
same person. Empirical support for this
hypothesis is also available [Schulte,
1965, p. 587; Titard, 1971, pp. 51-52;
Hartley and Ross, 1972, p. 48]. These
arguments are reflected in Hypothesis 5.
Audit-firm size may also interact with

Size3
Larger audit firms are often considered
less subject to a loss of independence
than are smaller firms. Arguments to this
effect commonly follow one of two lines.
First, a large audit firm tends to be less

3 Size of the audit firm may be
interpreted as a
surrogate for numerous other variables, possibly including the relative proportion of a CPA firm's total
fees obtained from a given client, the size of the client,
the presence or absence of specialization of the MAS
function within the firm, or the proportion of total fees
from a given client represented by MAS fees from that
client.

remain truly independent. Second, the
audit firm may develop a stake in the
client, for its prestige as a successful
adviser depends upon the client's success. Third, consultants may, in effect,
become decision makers, thus placing

the audit firm in the position of auditing
its own decisions. Finally, CPAs may
develop too close a relationship with
management during an MAS engagement. In addition, Hartley and Ross
suggest that an audit firm's financial
dependence on a client may increase as
the size of the MAS fee increases [1972,
p. 44]. Research findings supporting
such negative relations between MAS
and independence have been reported by
Schulte [1965, p. 590], Briloff [1966, p.
492], Titard [1971, p. 51], and Hartley
and Ross [1972, p. 44].
The empirical evidence against MAS
appears more convincing than the evidence for it-hence the direction of
hypothesis 2. Note, however, that the
MAS issue is often not considered in
terms of such a simple dichotomy. The
question has often focused on the advisability of providing specific services.

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


Shockley

789

the level of competition in the audit
environment. Because of its smaller

revenue base, a small firm may not be
able to withstand the effects of competition as well as a larger firm (Hypothesis
6). It is usually at a disadvantage because
of (1) the prestige associated with larger
firms, and (2) the superior financial resources available to larger firms [AICPA,
Cohen Commission Report, 1978, p.
1141.
Tenure

An audit firm's tenure, the length of
time it has been filling the audit needs of a
given client, has been cited as having an
impact on the risk of a loss of independence:
Long association between a corporation and
an accounting firm may lead to such close
identification of the accounting firm with the
interests of its client's management that truly
independent action by the accounting firm
becomes difficult [U.S. Senate, 1976, p. 211.
In a great many . .. cases, however, the
greatest threat to [the auditor's] independence is a slow, gradual, almost casual erosion
of his "honest disinterestedness" [Mautz &
Sharaf, 1961, p. 208].

Complacency, lack of innovation, less
rigorous audit procedures, and a learned
confidence in the client may arise after
long association with a client. The Metcalf subcommittee's staff considered this
danger serious enough to recommend
mandatory rotation of auditors as a

possible remedy [U.S. Senate, 1976, p.
211.
Hypothesis 4 divides possible levels of
tenure into two categories: (1) five or
fewer years, and (2) more than five years.
This division is consistent with the membership requirements of the AICPA's
SEC Practice Section, which provide
that one partner may not remain in
charge of an audit for more than five
years. It is doubtful that perceptions of

an audit report's credibility are influenced by the tenure of an individual
(unless he is a sole practitioner). In most
cases, the financial statement user's
knowledge of the auditor is limited to the
name of the auditing firm on its report.
This study therefore adopts a similar fiveyear criterion.
While the direction of Hypothesis 4
agrees with the preceding arguments,
there is a potential for a causal effect in
the opposite direction, i.e., that long
association may increase independence.
Opponents of mandatory rotation point
out that a CPA firm gains a deeper
familiarity and insight into the client's
operations through audit repetition, thus
facilitating more efficient, less costly
audit service than is possible for a
"fresh" auditor. Because the client is
likely to consider the "tenured" auditor

to be more valuable, the audit firm
becomes less dependent on the client and
better able to resist client pressure.
METHODOLOGY

Essentially, this study analyzes the
perceptions of four respondent groups
with regard to the effects of the four independent variables on the risk that an
audit firm's independence might become
impaired. The data are analyzed by
factorial analysis of variance techniques.
Sample
The study compares the perceptions of
CPAs and two groups of financialstatement users. Financial analysts were
chosen as a surrogate group for investors.
Bank commercial loan officers were
chosen to represent creditors. The CPA
category was divided into two groups:
partners from large national (Big Eight)
CPA firms and partners from local or
regional firms. This segregation was made
because it seemed likely that the perceptual models of the two groups are

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


The AccountingReview,October1981

790

TABLE I

MAILED, SAMPLINGRATE, AND RESPONSERATE
SUMMARYOF QUESTIONNAIRES

Group

Questionnaires
Mailed

Sampling
Rate

Questionnaires
Returned

Response
Rate

Big Eightpartners
OtherCPAs
Loan officers
Financialanalysts
Overall

77
69
67
64
277

1.00
.10
.54
.58
na

48
44
45
39
176

0.623
.638
.672
.609
.635

and response-ratecolumns
Note: Questionnairesmailed exclude those undeliverable.Questionnaires-returned
excludeunusableresponses.

significantly different [Briloff, 1966, p.
492].
All of the subjects were selected from
relevant populations in North Carolina.
The survey instrument was mailed to all
77 partners of Big Eight firms residing in
the state. Local and regional CPAs were

randomly selected from the Directory
of Members of the North Carolina
Association of Certified Public Accountants. Loan officers were selected from
the North Carolina membership of the
Virginia-Carolina Chapter of Robert
Morris Associates. Finally, financial analysts were randomly selected from the
roster of the North Carolina Society of
Financial Analysts.
Table 1 provides information on sampling and return rates. Ten of the 287
questionnaires mailed were undeliverable
and are not included in the table. Of the
questionnaires returned, another ten
were unusable due to missing data or
ambiguity in group classification. Respondents failing to return the questionnaire in four weeks were mailed a second
request along with a copy of the original
questionnaire.4
Data and Instrumentation
The instrument used to collect the data
is illustrated in Appendix A. Essentially,
it presented the subjects with a scaling
task. Each subject was required to make a
subjective estimate [Torgerson, 1958, pp.

61-93] of the relative strengths of 16
scenarios along a numerical scale from
one to seven. The scale represented the
risk that an audit firm's independence
may become impaired, the dependent
variable. The scenario depicted a set of
circumstances describing a hypothetical

audit firm, each scenario representing
one of 16 possible combinations of levels
of the four experimental variables.5 Each
subject was asked to rate each scenario
entirely on the basis of the four pieces of
information provided therein.
It is assumed that the data are measured at the interval level.6 The dependent variable scale is anchored at two
points, defining the unit and origin of the
scale judgments. The number 7 was
defined as "very high risk," while number
1 was defined as "very low risk." Re4 Approximately 74 percent of the subjects responding
did so before the second mailing.
5 The independent variables were restricted to two
levels for theoretical as well as pragmatic reasons.
Levels for MAS and size have counterparts in the auditing environment, i.e., the pro-con arguments for MAS
and the usual distinction between Big Eight and "other"
firms. Additional levels for tenure or competition would
greatly increase the size of the experimental task, potentially lowering the response rate and effort invested
by the subjects, thereby lowering data quality. In addition, it was uncertain whether the respondents could
usefully distinguish between additional levels.
6 Analysis of the data suggests that this is a reasonable
assumption. Following the method suggested by Grove
and Savich [1979, p. 5311, correlational analysis was
conducted using both interval and ordinal assumptions.
Little difference was found between the two.

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions

Shockley

791

spondents had all 16 scenarios available
simultaneously and were encouraged to
alter any response if they should change
their minds while completing the questionnaire.7 To support the assumption of
equal intervals, captions were omitted
from the intermediate scale values to
avoid bias due to differing interpretations
of the captions.
Possible bias due to order and learning
effects inherent in the instrument was
anticipated. To minimize the potential
for bias, the order of scenarios and variables within scenarios was randomized
and printed separately for each subject.8
Experimental Design

Analysis of variance was chosen as the
most appropriate methodology for the
study because of ANOVA's unique ability to focus on the hypothesized interactions and its utility in dealing with categorical variables. The experimental task
called for repeated judgments by individual subjects across the 16 scenarios.
Hence, the experimental design was a
repeated-measures block design with one
grouping factor and four trial factors
[Winer, 1971, pp. 514-603]. The four
grouping factor levels corresponded to
the four categories of subjects. Each trial
factor had two levels corresponding to

the levels defined implicitly in hypotheses
I through 4.
Analysis of the data was conducted at
the overall level and within each subject
group. At the overall level, each of the
four experimental factors, their interactions, and the grouping factor were
tested for significance. Because of the
expected perceptual differences between
groups, data for each group were also
analyzed separately. The formal hypotheses for each group would be similar to
those presented for the overall analysis.
The separate analyses allowed computation of a more precise set of descriptive

statistics for each group, yielding the
relative weights ascribed to the factors by
the different groups.
The w2 statistic was computed for significant effects in all of the analyses. This
statistic provided a measure of the weight
ascribed to each effect by estimating the
proportion of variance which it explained
[Hays, 1973, pp. 484 488]. Analysis of
variance often finds trivial effects to be
significant because the F statistic is
partly a function of sample size. The wt2
statistic provides additional insight into
an effect's significance. For examples of
its use in the accounting literature, see
Ashton [1974], Joyce [1976], Hofstedt
and Hughes [1977], and Schultz and
Gustavson [1978].

RESULTS

Overall Analysis
The overall analysis of variance is
summarized in Table 2. Tests of the main
effects in the ANOVA model are equivalent to tests of Hypotheses 1 through 4.
Tests of the first-order interactions (1)
MAS and size of the audit firm, and (2)
competition and size are equivalent to
tests of Hypotheses 5 and 6. Interactions
between the grouping factor and independent variables indicate differences between groups in the perceived significance of the independent variables.
Three of the four main effects were
significant at the .01 level. This result
indicates that, overall, the subjects perceive (1) that higher levels of competition
increase the risk that independence may
be impaired, (2) that CPA firms which
' This technique, described as the "multiple stimulus
procedure" by Torgerson [1958, p. 66], helps to minimize bias due to modification of the psychological scale
as a subject progresses through the stimulus series.
' The printing of the randomized instruments was accomplished by a computerized randomization procedure
in combination with a text editing program.

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


The AccountingReview,October1981

792
TABLE2

ANALYSISOF VARIANCEFORALL SUBJECTRESPONSES

Main Effects
C Competition
M MAS
S Size
T Tenure
G Groups
Interactions
CM
CS
CT
CG
MS
MT
MG
ST
SG
TG

SS

df

MS

F

p

647.9
234.7
544.4
8.6
454.9

1
i
1
1
3

647.9
234.7
544.4
8.6
151.6

134.2*
85.6*
132.2*
3.9
13.8*

.000
.000
.000
.050
.000

.5
6.6
.0
13.6
1.9
.7
45.4
.1
181.2
4.1

1
1
1
3
1
1
3
1
3
3

.5
6.6
.0
4.5
1.9
.7
15.1

.1
60.4
1.4

1.1
11.3*
.0
.9
3.4
1.3
5.5*
.2
14.7*
.6

.302
.001
.986
.423
.067
.262
.001
.687
.000
.601

Note: Repeated-measures designs use the mean square associated with the specific treatment-subject interaction as
the F ratio denominator. Though not shown above, each error term has 172 degrees of freedom. None of the second or
higher order interactions are significant for a=.01.
*p<.O0

provide MAS to audit clients are more
likely to lose independence than those
which do not, and (3) that smaller firms
are more likely to lose their independence
than are larger firms. The effect of the
tenure factor (Hypothesis 4) was not
significant.
Thus, the analysis supports Hypotheses 1, 2, and 3. The competition-size
interaction is also significant, supporting
Hypothesis 6. That is, the subjects perceive competition as a greater problem
for small firms than for larger firms. The
hypothesized MAS-size interaction (Hypothesis 5) is rejected. However, the
practical impact of even the competitionsize interaction is questionable. Table 3,
which reports the W2 statistic for each
significant effect in an absolute and
relative sense, shows that the variance
explained by this effect is negligible. Its
statistical significance arises mainly because of the relatively large number of
observations analyzed. Similarly, while
tenure may be considered significant at

a= .05, the variance it explains is also

negligible (<.000).
Table 3 also shows that the grouping
factor and its two interactions account
for 31.7 percent of the variance explained.
The significance of the grouping factor
implies that the scale judgments of the

subject groups differ in an absolute sense.
The mean response for all scenarios
across all subjects within each group is
2.57, 3.08, 3.54, and 3.53 for the Big
Eight CPAs, other CPAs, bank loan
officers, and financial analysts, respectively.9 The group factor interactions
imply that perceptions between groups
differ significantly at least for the MAS
and size factors.
These differences in group judgments
could restrict the utility of the overall
9 The significance of the grouping factor arises primarily from the judgments of the Big Eight subjects.
Pairwise comparisons of these means via the NewmanKeuls test indicates that their judgments are significantly
different (a =.01) from the other three groups. Other
comparisons are not significant.

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


793

Shockley
TABLE3
RESPONSES
FORALLSUBJECT
EFFECTS
BYSIGNIFICANT
EXPLAINED
VARIANCE

Relative
co2

co 2

134.2
85.6
132.2

0.087
.031
.073

0.311
.111
.261

Group

13.8

.060

.214

Group-MAS interaction
Group-Size interaction
Coanpetition-Size interaction

5.5

14.7
11.3

.005
.024

.018
.085

0.280

1.000

F

Effect

Competition
Management advisory services
Size of audit firm

TOTAL

TABLE4
EFFECTS
BYGROUP
BYSIGNIFICANT
EXPLAINED
VARIANCE

Big Eight Partners

Other CPA Partners

W222

Competition
MAS

0.062
.010

0.218
.035

Size of firm
Tenure
CS interaction
MS interaction

.205

.722

Total

wo2s

co2

w2

0.138
.052

.018
.007

0.284

1.000

0.190

Relative
co2

co 2

0.726
.274

0.067
.009

0.298
.040

0.117
.092

0.480
.377

-

.148
.001

.658
.004

.035

.143

1.000

0.225

1.000

0.244

1.000

-

.005
.002

Financial Analysts

Relative

Relative

Relative
Effect

Commercial Loan
Officers

Note: Effects for which entries are included were significant at the .05 level. The relative co2is an estimate of the
proportion of variance explained by a specific effect relative to the total variance explained by all significant effects.

results. Additional insight into the relevance of the experimental factors to each
group can be obtained by analyzing each
group's responses separately, thereby
eliminating the confounding effect of the
grouping factor.
Within GroupAnalysis
Tables 4 and 5 summarize the results
of the ANOVA's performed on each
group. Table 4 presents the w2 statistic
for all significant effects along with the
relative variance explained. Table 5 reports the marginal means for each factor
by group.

The within-group analyses are generally consistent with the overall analysis.

With respect to direction, the main
effects for all groups (whether significant
or not) are consistent with the hypotheses. However, two differences become
apparent in the more detailed analysis.
First, while tenure was not significant in
the overall analysis, it was significant to
the loan officers. The variance explained,
however, is so low that it should probably
be discounted. Second, the small CPAs
did not find the size factor significant.
The latter result, along with the relatively high weight placed on the size

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


The Accounting Review, October 1981

794
TABLE5

MARGINALMEANS FORFACTORLEVELSBY GROUP

Big Eight
Partners
Eflect
Competition (Low, high)
MAS (No, yes)
Size (SEC, nonSEC)
Tenure (<5, >5)

Commercial
Loan Officers

Other CPA
Partners

Financial
Analysts

1

2

1

2

/

2

1

2

2.17
2.41
1.84
2.54

2.99
2.74
3.31
2.61

2.51
2.73
2.97
2.99

3.65
3.43
3.19
3.17

3.12
3.37
2.92
3.44

3.96
3.71
4.16
3.64

3.00
3.05
3.23
3.52

4.05
4.00
3.82
3.53

TABLE6
OF MEAN SCENARIOJUDGMENTSACROSSGROUPS: BETWEENGROUP CONSENSUS
RANK-ORDERCORRELATION

A.
B.
C.
D.

Big Eight partners
Other CPA partners
Commercial loan officers
Financial analysts

A

B

C

D

1.000
.633

.957
.717

1.000
.728
.930

1.000
.815

1.000

factor by the Big Eight partners, is
perhaps not surprising. Given a presumption in the profession consistent with
Hypothesis 3, i.e., that substandard
audits or lack of professionalism is more
likely in smaller firms, this is the result
which might be expected. Possibly it
reflects a defensive attitude on the part of
the small CPAs or a "superior" attitude
among the Big Eight CPAs. However, it
is not possible to explain the response
within the scope of this study.
Interaction effects are insignificant for
all groups except the Big Eight partners,
who found the MAS-size and competition-size interactions significant. However, the relative variance explained is
again very small for both. The analysis
suggests that, for practical purposes,
interaction effects were ignored by the
respondents. This is consistent with the

results of similar studies [Libby and
Lewis, 1977, pp. 246, 2521.
There are, however, large differences
in weights ascribed to competition and

MAS by the two accountant groups. The
smaller CPAs weighted competition
much more heavily than did the Big
Eight subjects. This may indicate that
competition does, in fact, have a larger
impact on partners in smaller firms. Such
an interaction would not necessarily be
reflected in the subjects' perceptual models of independence, for most subjects
will not have experienced the effects of
competition from the perspective of
different-sized firms. A similar argument
can be made for the MAS variable, which
also is weighted more heavily by the
smaller CPAs. The difference may result
because segregation of the MAS and
audit functions is less distinct in smaller
firms. Again, a subject without experience in both categories might not incorporate the interaction in his perceptual
model. Thus, for practical purposes, the
hypothesized perceptual interactions can
be rejected or discounted. However, the
results do not suggest that the perceptions of accountants are unaffected by

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions

Shockley

795

interaction of these factors.
A priori, one might expect a similarity
of response between the two accountant
groups and between the nonaccountant
groups, suggesting basic differences in
the perceptual models between accountants and third parties. Instead, Table 4
suggests a similarity in judgment between
(1) the Big Eight partners and loan
officers, and (2) the local and regional
CPAs and financial analysts. To provide
an index of similarity between aggregate
group perceptual models, Table 6 reports
Spearman correlation coefficients computed on the mean judgments for each
scenario determined for each group. The
mean scenario judgments are presented in
Appendix B.
The entries in Table 6 represent an
index of similarity, or consensus, of the
rankings which the subjects assigned to
the scenarios as a group. Consistent with
Table 4, the highest correlation between
group judgments is that between the Big
Eight partners and commercial loan
officers (.957). Next highest is that between the small-firm CPAs and financial
analysts. The lowest correlation related

the two accountant groups.
Aggregating individual judgments to
facilitate group comparisons eliminates
the variance arising from individual
differences. Thus, the Spearman correlations are best interpreted as rough
indices of consensus only in an aggregate
sense. The impact of individual differences is discussed below.
Individual Differences

For each group, the variance explained
by the experimental effects appears to be
relatively low, though it is consistent with
the findings of similar research [Hofstedt
and Hughes, 1977, p. 389; Joyce, 1976, p.
32]. The large remainder, associated with
experimental error, is indicative of a
relatively'high level of individual differ-

ences, and suggests further investigation.l0
Analyses of variancewere performed
on the judgments of each individual
[Kirk, 1968,pp. 227-229] and the associated w? computedfor significanteffects.
The a? statisticsummedacrossall significant effects for each individual yielded
an estimateof the proportionof variance
which the model explainedfor that individual. This sum, averagedacross individuals within each group, was 69.6,
56.4, 64, and 71 percent for the Big
Eight CPAs, local and regional CPAs,
bankloan officers,and financialanalysts,
respectively.11
Thus,theaveragingacross

individuals inherent in the larger
ANOVA's resulted in large losses of
explanatory power due to individual
differences.
These differencesmay arise from (1)
differences in the factors considered
significant,(2) differencesin the weights
ascribedto factors, or (3) differencesin
the perceived direction of the factors'
effects. The latter cause is the most
critical to the evaluation of the larger
ANOVA's. Table 7 presentsa frequency
tabulationof the significanteffects consistent or inconsistent with the stated
hypotheses,as indicatedby theindividual
ANOVA's. Note that in most cases the
'0 Low variance explained is generally synonymous
with low within-group consensus. One index of withingroup consensus is the average correlation between the
judgments of each possible pair of individuals in the
group [Ashton, 1974; Schultz and Gustavson, 1978;
Joyce, 19761. For these data, the Spearman correlations
are .58, .33, .28, and .26 for the Big Eight CPAs, other
CPAs, bank loan officers, and financial analysts, respectively. These levels do not appear to be unusually low
when compared with previous research. In a classic
study of gastroenterologists' diagnoses of the benignancy
or malignancy of gastric ulcers, Hoffman et al. [1968]
reported a level of consensus of .38. Joyce [1976, p. 42],
using CPAs as subjects, found a consensus level of .373
in a study of audit program planning. Ashton [1974], in
a study of internal control quality in payroll systems,
reported a higher level of .70 [p. 151].

" For individuals, the variance explained ranged
from zero (no factors significant) to 100 percent.

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


The AccountingReview,October 1981

796
TABLE

7

FREQUENCY OF SIGNIFICANT EFFECTS FOR INDIVIDUALS CONSISTENT AND INCONSISTENT

WITH EXPERIMENTAL
HYPOTHESES

MAS

Tenure

Size

12
1

8
3

35
0

2
0

9
0

I

20
1

11
4

9
5

1
2

1
2

25
5

15
6

7
2

26
2

1
2

3
2

25
4

18
1

10
8

17
2

0
0

3
0

108
I1

65
9

36
17

87
9

4
4

16
4

Competition
Big Eight partners (n= 48)
Consistent
Inconsistent
Local and regional CPAs (n 44)
Consistent
Inconsistent
Commercial

loan officers

28

1
30

(n -45)

Consistent
Inconsistent
Financial analysts (n 39)
Consistent
Inconsistent
Total (n= 176)
Consistent
Inconsistent

CompetitionMAS-Size
Size
Interaction Interaction

direction of effect coincides with the
hypotheses.
Because of the presenceof significant
individual differences,it is prudent to
interpretthe W2 statisticsin Tables 3 and
4 only as indices of the relative weight
ascribedto the factorsby groups of subjects. However,analysisof the individual
ANOVA's shows that the instrument

was more successful in capturing the
judgmentsof individualsthan the overall
c2 statisticsmight suggest. Table 7 suggests a high level of consensusabout the
direction of effects despite differencesin
weightsascribedto them.Thus, the more
detailed analysis reduces somewhat the
interpretational difficulties usually associated with individualdifferences.
LIMITATIONS

The evaluationof the precedingresults
and following implicationsrequiresconsiderationof the study's limitations. Of
these, the most obvious is external
validity. Generalization of results to
populations outside the groups partici-

pating in the experiment is tenuous due
due to the geographic and qualitative
restrictions inherent in the sample. Also,
the degree to which the experimental
task simulates a "real" environment is
unmeasured. Subjects may not respond
in an experimental setting in the same
fashion as in more realistic contexts. It is
possible that the mere administration of
the instrument may have affected the
results.
Similarly, the specification of the
ANOVA model used here is restrictive.
The inclusion, exclusion, or replacement
of independent variables could lead to

different results. Redefinition of the
variables or their levels could also significantly affect the findings. For example, in this study the problem of definition is particularly relevant to the MAS
variable.
Finally, while measures were taken to
insure high response rates and data
quality, it is impossible to predict the
judgments of nonrespondents. However,
a method similar to that used by Buzby

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


Shockley

797

[1974, p. 426] was used to search for
response bias. For each group, the data
for each of the 16 scenarios were averaged
for responses received after the second
mailing and compared to an equal number of the earliest responses. Of the
resulting 64 t tests, nine (14 percent) were
significant at a =.2. This approximates
what one would expect from chance.
None were significant beyond oc=.08.
IMPLICATIONS
At a general level, the results are consistent with earlier surveys [Briloff, 1966;
Hartley and Ross, 1972; Lavin, 1976]
which also found differences between

different groups' perceptions of independence. It is imperative that the profession recognize these differences when
responding to current pressures for policy
and structural changes in the accounting
profession. Problems of credibility arise
from outside the profession. Therefore,
the auditor as policy maker cannot
afford a biased perspective; he must be
constantly aware of the perceptions of
those who depend on his service and how
these perceptions differ from his own.
This study suggests that such differences exist between categories of accountants as well as between accountants
and third parties. It is likely that these
differences are reflected in attitudes and
decisions about independence in practice.
Whether or not such differences are undesirable depends on their cause. If they
reflect characteristics of the practices
arising from the application of identical
standards in different audit environments, they are desirable. If they reflect
different standards, they are not. The task
facing the profession is to identify the
causes of the differences.
The significance of competition to the
judgments of all four groups (Table 4)
supports the Cohen Commission's position that the problem facing the profes-

sion is too much rather than too little
competition. From the standpoint of
independence, the stance of the Metcalf
subcommittee's staff in favor of increasing competition through legislation seems
ill-advised. Competition among auditors

is likely to increase now that constraints
on advertising and solicitation have been
removed. The increased potential for
damage to the profession's credibility
justifies careful consideration of any
policy designed to insulate the auditors
from inappropriate influence by manment. An existing example of this type of
policy, long endorsed by the AICPA, is
the use of an audit committee. Strengthening this position by making the audit
committee a prerequisite for an audit is a
policy step which should be considered
further.
The MAS variable also significantly
affects the judgments of all four groups.
By definition, the study restricts MAS to
"accounting related" services; hence,
generalization to the full range of services
provided in practice is unsupportable.
However, the data do support the general
findings of previous empirical research
that MAS, even when restricted to accounting related services, may impair
perceptions of independence [Titard,
1971]. At a minimum, the profession
must continue to emphasize to its practitioners the importance of objectively
evaluating engagements involving MAS
from the perspective of appearance as
well as fact. Replication of the current
study with an extension in the definition
of MAS could significantly extend its
usefulness.

One common defense of MAS argues
that a separation of the consulting and
audit function, as is common in larger
firms, minimizes any detrimental effect of
MAS. The findings do not support this
argument. The MAS-size interaction was
not important to third parties at any

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


The AccountingReview,October 1981

798
level of analysis. Thus, the separation of
the two functions is not likely to improve
third parties' perceptions

of indepen-

dence.
Another significant implication of the
findings is that third parties may view
small CPA firms as less independent than
larger firms. This is a severe problem, for
the majority of the nation's CPAs are, of
course, "small" practitioners. Certainly,
this differential in perceptions is undesirable. However, the scope of this
experiment did not allow identification of

its causes. Additional investigation is
needed to establish these causes and to
examine possible remedies for the problem.
The nonsignificance of the tenure factor implies that any policy action taken
to reduce the average tenure of auditors
may have little positive effect on perceptions of auditors' independence. However, this result must be interpreted with
caution. Individual analyses indicated
that of the 36 subjects (30 percent) who
found the tenure factor significant, 17
weighted it in the direction opposite to
the hypothesized negative effect. Thus,
the net effect of the tenure factor may
include significant offsetting effects.
These offsetting effects were not totally
unexpected; a rationale was presented
earlier. However, their presence suggests
that policy approaches directly affecting
tenure, e.g., mandatory rotation of audit
firms, are probably too simplistic. It is
possible that carefully selected alternative policy steps could have a net positive
effect. For example, stricter in-house
review procedures for CPA firms, made
known to third parties, might reduce the
effect's negative component without destroying tenure's positive aspects. Research is needed to measure the potential
effects of various alternatives on perceptions of financial statement users.
The impact on policy of the high level

of individual differences characterizing
the study also must be evaluated. At the
least, policies consistent with the effects

identified in the overall analysis will
please more people than they will displease. The individual analyses reveal
very few cases in which the direction of
effect for individuals is inconsistent with
the overall analysis. While the weights
assigned to factors varied considerably,
the subjects were highly consistent in
judging the direction of effect. Therefore,
individual differences would not be expected to seriously undermine the effectiveness of related policy actions.
Finally, in evaluating a study of perceptions, it is prudent to consider that
financial-statement users may recognize
the risk of nonindependence but accept
it as a necessary part of the audit function.
If so, at what point does the risk become
unacceptable? Research is needed to
relate perceptions of the risk directly to
third parties' reliance on financial statements.
SUMMARY

This study examined the perceptions
of certified public accountants and financial statement users with respect to the
independence of public accounting firms.
The subjects considered the impact of
four variables on the risk that the independence of an audit firm might become
impaired: competition for audit clients,
the provision of MAS, the size of the
audit firm, and the audit firm's tenure.
Firms operating in highly competitive
environments, firms providing MAS, and
smaller audit firms were perceived as

having a higher risk of impairment of
independence. Tenure was not found to
have a significant impact on perceptions
of independence. For all respondents, the
competition variable was ranked most
important, followed by the size and MAS
factors. Analyses for each group, how-

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


Shockley

799

ever, indicated that significant differences
existed in the weights ascribed to the
variables by different groups.
A relatively high level of individual
differences was found in the weights
ascribed to the experimental variables.
Most subjects, however, agreed about the
direction of the main effects.
Conclusions about the relative importance of these three variables in realworld contexts must be considered tentative. However, this study indicates that
each variable has a definite potential impact on the credibility of the independence assumption, and that the direction
of impact can be predicted. Future policy
decisions designed to improve perceptions of independence should therefore
take these effects into account.
APPENDIXA

INSTRUMENT

The instrument contained 16 scenarios
of the form illustrated below. Together,
they represented all possible combina-

tions of the four independentvariables.
The scale refersto the risk that the independenceof the CPA firmdepictedin the
scenariomay becomeimpaired.Subjects
were asked to base their responsesonly
on the four pieces of information contained in the scenario;all other possible
relevant factors were to be considered
constant.
ScenarioJ.
Levelof competition...........
High
Size of auditfirm..............
Large
Auditor'stenure(years). ............Morethan five
Are MAS rendered?...........
No
Very
Low
Risk

1

3

2

4

5

6

Very
High
Risk

7

(The actual instrument contained full
instructionsand definitions.Additional
questions were included primarily to
collect demographicdata and to identify
unanticipatedproblemswith the instrument or design. None was found. Additionaldataareavailablefromtheauthor.)

APPRNDIX B
RANK-ORDERINGOF SCENARIOSBY GROUP

SCENARIO
Comp.

MAS

Size

Tenure

high
high
high
high
low
low
low
low
high
high
high
high
low
low
low
low

yes
yes
no
no
yes
yes
no
no
yes
yes
no
no

yes
yes
no
no

small
small
small
small
small
small
small
small
large
large
large
large
large
large
large
large

> 5

<5
>5
<5
>5
<5
>5

?5
>5
<5
>5
?5
>5
<5
>5
?5

Big
Eight
partners
16
15
13
14
12
11
10
9
8
6.5
6.5
5
4
3
1
2

Commercial
loan
officers
16
14
15
13
12
10
7

9
11
8
6
5
4
3
2
1

Other
CPA
partners
16
15
10
11.5
9
6.5

4
2.5
14
13
11.5
8
6.5
5
2.5
1

Financial
analysts
15
16
12
11
10
9
4
3
14
13
6
8
7
5
1
2

Note: Scenarios with the same mean judgment score are assigned the mean of the two ranks which they otherwise
would have taken. Lower ranks correspond with lower perceived risk of impaired independence.

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions


800

The AccountingReview,October1981
REFERENCES

American Institute of Certified Public Accountants, Commission on Auditors' Responsibilities, Report,
Conclusions, and Recommendations(AICPA, 1978).
American Institute of Certified Public Accountants, Ethics, Bylaws, Quality Control, Vol. 2, Professional
Standards (Commerce Clearing House, 1978).
Ashton, R. H. (1974), "An Experimental Study of Internal Control Judgments," Journal of Accounting
Research (Spring 1974), pp. 143-157.
Beams, F. A. and L. N. Killough (1970), "Audit Independence-An Extension of the Concept," The
National Public Accountant (December 1970), pp. 12-19, 26-28.
Briloff, A. J. (1966), "Old Myths and New Realities in Accountancy,"

THE ACCOUNTING REVIEW (July

1966), pp. 484-95.
Burton, J. C. (1980), "A Critical Look at Professionalism and Scope of Services," Journal of Accountancy
(April 1980), pp. 48-56.
Buzby, S. L. (1974), "Selected Items of Information and Their Disclosure in Annual Reports," THE
ACCOUNTING REVIEW (July, 1974), pp. 423-435.
Goldman, A. and B. Barlev (1974), "The Auditor-Firm Conflict of Interests: Its Implications for Independence," THE ACCOUNTING REVIEW (October 1974), pp. 707-18.

Grove, H. D. and R. S. Savich (1979), "Attitude Research in Accounting: A Model for Reliability and
Validity Considerations," THE ACCOUNTING REVIEW (July 1979), pp. 522-537.
Hartley, R. V. and T. L. Ross (1972), "MAS and Audit Independence: An Image Problem," Journal of
Accountancy (November 1972), pp. 42-52.
Hays, W. L. (1973), Statistics for the Social Sciences (Holt, Rinehart, and Winston, Inc., 1973).
Hoffman, P. J., P. Slovic, and L. G. Rorer (1968), "An Analysis-of-Variance Model for the Assessment of
Configural Cue Utilization in Clinical Judgment," Psychological Bulletin (May 1968), pp. 338-49.
Hofstedt, T. R. and G. D. Hughes (1977), "An Experimental Study of the Judgment Element in Disclosure
Decisions," THE ACCOUNTING REVIEW (April 1977), pp. 379-395.
Joyce, E. J. (1976), "Expert Judgment in Audit Program Planning," Journal of Accounting Research
(Supplement, 1976), pp. 29-60.
Kirk, R. E. (1968), Experimental Design. Proceduresfor the Behavioral Sciences (Brooke/Cole Publishing
Co., 1968).
Lavin, D. (1976), "Perceptions of the Independence of the Auditor," THE ACCOUNTING REVIEW (January
1976), pp. 41-50.
Libby, R. and B. L. Lewis (1977), "Human Information Processing Research in Accounting: The State
of the Art," Accounting, Organizations and Society, Vol. 2, No. 3 (1977), pp. 245-68.
Mautz, R. K. and H. A. Sharaf (1961), The Philosophy of Auditing (American Accounting Association,
1961).
Public Oversight Board, Scope of Services by CPA Firms, published for the Public Oversight Board of the
SEC Practice Section of the Division for CPA Firms in the American Institute of Certified Public
Accountants (AICPA, 1979).
Schulte, A. A. Jr. (1965), "Compatibility of Management Consulting and Auditing," THE ACCOUNTING
REVIEW (July 1965), pp. 587-93.

Schultz, J. J. Jr. and S. G. Gustavson (1978), "Actuaries' Perceptions of Variables Affecting the Independent Auditor's Legal Liability," THE ACCOUNTING REVIEW (July 1978), pp. 626-41.
Titard, P. L. (1971), "Independence and MAS-Opinions of Financial Statement Users," Journal o]
Accountancy (July 1971), pp. 47-52.
Torgerson, W. S. (1958), Theory and Methods of Scaling (John Wiley & Sons, 1958).
U. S. Securities and Exchange Commission (1979), Securities and Exchange Commission Report to

Congress on the Accounting Protession and the Commission's Oversight Role (U. S. Government Printing
Office, July 1979).
U. S. Senate, Subcommittee on Reports, Accounting, and Management of the Committee on Government
Operations, The Accounting Establishment. A Staff Study (U. S. Government Printing Office, 1976).
U. S. Senate, Subcommittee on Reports, Accounting and Management of the Committee on Governmental Affairs, Improvingthe Accountability of Publicly Owned Corporationsand Their Auditors (U. S.
Government Printing Office, 1977).
Winer, B. J. (1971), Statistical Principles in Experimental Design (McGraw-Hill, 1971).

This content downloaded from 188.72.96.138 on Mon, 9 Jun 2014 18:57:47 PM
All use subject to JSTOR Terms and Conditions