The effect of auditor changes on audit f
NORTH - HOLLAND
RESEARCH
NOTE
The Effect of Auditor Changes on Audit
Fees, Audit Hours, and Audit Quality
Donald R. Deis, Jr. and Gary Giroux
Our study investigates the relationship of independent auditor fees, hours and
quality, with particular emphasis on initial audits. Using audits of Texas independent school districts, empirical results indicated specific low bailing relationships.
First year independent audits had statistically significant lower audit fees. Despite
the lower fees, quality was higher and more audit hours were utilized.
1. Introduction
Our study investigates the effect of independent auditor changes on fees,
independent audit effort (as measured by audit hours), and independent
audit quality. The empirical analysis is based on audits conducted by local
and regional public accounting firms of independent school districts (ISDs)
in Texas. We found that initial audits were associated with lower fees,
higher quality, and higher audit hours at the .10 level (one tailed). The
results also indicated that both audit fees and audit hours were significantly related to audit quality in a manner consistent with the explanation
that audits are product differentiated in the market.
The empirical findings presented in our paper support previous research. Our paper confirms that both audit fees and audit hours are
influenced by audit quality. The evidence suggests that independent audi-
Address correspondence to" Professor Gary Giroux, Department of Accounting, Texas A & M
University, College Station, 'IX 77843.
Journal of Accounting and Public Policy, 15, 55-76 (1996)
© 1996 Elsevier Science Inc.
655 Avenue of the Americas, New York, NY 10010
0278-4254/96/$15.00
SSDI 0278-4254(95)00041-0
56
D.R. Deis, Jr. and G. Giroux
tors do not sacrifice quality to obtain a new audit client, though lower
audit fees were detected among initial audit engagements. Audit hours and
audit quality were higher among initial audits, confirming a conclusion
reported by Raman and Wilson (1992, p. 292) in their study of a sample of
U.S. cities over four years (1984-1987).
No previous study has compared audit fees and audit hours with both
audit service quality and independent auditor changes in a single analysis.
Initial audits (commonly associated with low bailing) may result in lower
fees but the relationship to audit effort has not been the subject of a great
deal of research (see Giroux et al. 1995; Davis et al. 1993; Palmrose 1989
for research on the topic) and never in combination with a measure of
audit service quality. If audit effort declines with audit fees on initial
audits, audit quality also should decline. DeAngelo (1981a, pp. 114,
118-120; 1981b, pp. 187-194), however, suggests that audit effort will be
maintained so that incumbent auditors can generate quasi-rents in future
years.
We investigate other issues associated with the independent audit market. Our study examines other aspects of audit economics (e.g., brand
name, specialization, and product costs) in the small CPA firm sector. 1 We
also extend Palmrose's (1986) and (1989) studies of the determinants of
audit hours to the public sector.
Further, we build upon the work of Deis and Giroux (1992) and Giroux
et al. (1995). The former study evaluated the potential determinants of
actual independent audit quality using a metric of audit quality based on
nineteen features of the independent auditor's work as documented in
working paper reviews conducted by state auditors at the Texas Education
Agency. The latter study compared the audit fees, auditor hours, and audit
quality of independent audits conducted by audit firms participating in a
voluntary peer review program to those not participating.
2. P r i o r R e s e a r c h o n Initial A u d i t s
Conventional wisdom suggests that when public accounting firms offer
price discounts to capture new audit clients, audit quality may be compromised. 2 A competing explanation advanced by DeAngelo (1981a, pp. 114,
118-120; 1981b, pp. 187-194), however, contends that low balling (i.e.,
bidding below the avoidable costs for the first year audit) represents an
i The National Commission on F r a u d u l e n t Financial R e p o r t i n g ( N C F F R 1987, p. 25) indicated
that t h r e e - q u a r t e r s of all Securities and E x c h a n g e Commission actions against i n d e p e n d e n t public
a c c o u n t a n t s involved n o n n a t i o n a l firms a n d often involved s u b s t a n d a r d audit work. Specific weaknesses include lack of specialized training a n d p o o r supervision of audit staff.
2 For example, the C o h e n Commission ( A I C P A 1978, pp. 111-112) a n d the Metcalf C o m m i t t e e
(U.S. Senate 1977, p. 45) have discussed the negative effects of price competition on audit quality.
Recently, the Dingell C o m m i t t e e p r o p o s e d the Financial F r a u d Detection a n d Disclosure A c t (U.S.
H o u s e 1992, pp. 1 - 2 ) to regulate the public a c c o u n t i n g profession citing that profession's pricing
practices as c o n t r i b u t i n g to the Savings a n d L o a n crisis.
The Effect of Auditor Changes
57
initial investment by the public accounting firm to obtain future profits in
repeat engagements created by the auditor's learning curve and the high
cost of changing auditors (DeAngelo 1981a, p. 120). Accordingly, such
practice does not (by itself) pose a threat to audit quality.
The effect of low bailing, however, is problematic barring measurement
of audit quality. If public accounting firms low ball to obtain future profits,
audit quality could be threatened by excessive price discounts. That is, low
bailing could be associated with reduced effort and lower audit quality
(AICPA 1978, pp. 111-112).
Research thus far has consistently demonstrated that independent auditor changes are associated with decreased audit fees (DeAngelo 1981a, p.
124 and 1981b, p. 190). Despite lower fees for initial audit engagements,
theory posits that product quality will be maintained (DeAngelo 1981a, p.
124 and 1981b, p. 190; Klein and Leffler 1981, p. 622). Empirical validation
of this theory, however, has been slow in developing due to lack of
adequate information about auditor effort and audit quality. Most independent audit fee studies do not have data on audit hours or engagement
specific measures of audit quality. 3
Palmrose (1986) used audit hours to support the argument that (then)
Big Eight audit firms conducted high quality audits. She (1986, p. 106)
found Big Eight audit firms exhibited both higher audit fees and audit
hours. From this evidence, Palmrose (1986, p. 108) concluded that Big
Eight audit firms commanded higher fees not because of monopoly power
but, rather, because they conducted higher quality audits. In a later study,
Palmrose (1989, p. 496) indicates that audit hours declined with audit firm
tenure while audit fees remained stable. Hence, profit margins may have
improved over time because audit effort decreased at a rate faster than
audit fees. As Palmrose (1989, p. 498) noted, this finding is consistent with
DeAngelo's (1981a) arguments about quasi-rents accruing to incumbent
audit firms. Nonetheless, audit hours may not fully reflect variations in
audit quality. Audit hours, Palmrose (1989, p. 498) cautioned, do not
reflect differences in the mix of audit firm personnel or level of effort.
Recent studies have utilized engagement specific measures of audit
quality based on government oversight agency reviews of public accounting
firm working papers and audit reports. Copley and Doucet (1993b, p. 32)
found the probability of a substandard audit to increase with the length of
auditor tenure. Similarly, O'Keefe et al. (1994, p. 52) found initial engagements associated with significantly fewer violations of generally accepted
3 In a study of p e r c e p t i o n of a u d i t o r s ' i n d e p e n d e n c e , Shockley r e p o r t e d "subjects perceive
competition as a g r e a t e r problem for small [audit] firms than for larger [audit] firms" (1981, p. 792).
W h e n asked, b a n k loan officers r e g a r d e d audit firm t e n u r e as a potential t h r e a t to a u d i t o r
i n d e p e n d e n c e (Shockley 1981, p. 788). Both of these results suggest problems for the g o v e r n m e n t a l
audit m a r k e t in which small public a c c o u n t i n g firms d o m i n a t e a n d lengthy i n d e p e n d e n t audit
firm-client relationships are c o m m o n place.
58
D.R. Deis, Jr. and G. Giroux
accounting standards (GAAS) reporting standards than repeat engagements. Also, Deis and Giroux (1992a, p. 476) and Banker et al. (1992,
p. 505) reported a positive relationship between audit hours and audit
quality.
Giroux et al. (1995) considered the effects of peer review on audit
economic relationships. That study was a public sector extension of a
private sector study that found no fee premia associated with audit firms
participating in a voluntary peer review program (Francis et al., 1990,
p. 375). The lack of fee premia for these firms lead Francis et al. to
conclude that there were economic disincentives to join a peer review
program (1990, p. 376). Hence, if peer review is desirable, a mandatory
program is necessary given their findings.
In contrast, a similarly motivated study in the public sector found fee
premia associated with peer reviewed audit firms and, accordingly, apparent economic incentives to voluntarily participate in a peer review program
(Giroux et al. 1995, pp. 78-79). On a fee-per-hour basis, however, the fee
premia disappeared implying that higher overall fees were associated with
increased audit effort (Giroux et al. 1995, pp. 75-76). It is important to
note that Giroux et al. (1995) did not test the relationship of fee and hours
to audit quality nor did they specifically test the interrelationships between
auditor changes and audit quality on hours and fees. Rather, Giroux et al.
(1995) addressed the contention that if peer review programs serve some
public interest, they must be mandatory due to the lack of economic
incentives to participate on a voluntary basis.
Though not central to the purpose of their study, certain findings
reported by Giroux et al. (1995) are pertinent to the present study. They
(1995, pp. 74-76) found that initial audits were associated with lower audit
fees, higher audit hours, and--naturally--lower fees per hour. Though
they did not incorporate audit quality in their fee, hours, or fee-per-hour
models, their analysis of variance model of audit quality (Giroux et al.
1995, pp. 76-78) indicated that audit quality tends to decline as independent auditor tenure lengthens (similar to Deis and Giroux 1992, pp.
475-476). In particular, multiple comparison procedures revealed that
initial audits were of significantly higher quality than audits conducted by
independent auditors with twenty or more years of tenure but not those
audits in 2 to 4 years or 5 to 19 years of tenure (1995, p. 78). While higher
audit hours among initial audits is encouraging, it does not guarantee that
sufficient audit quality is maintained, Given the inability to distinguish
between initial independent audits and those with less than twenty years of
tenure, it appears that quality is probably adequate. A more thorough
analysis than that provided by Giroux et al. (1995) is required, however,
before one can conclude that audit quality does not decline in initial audit
engagements.
The Effect of Auditor Changes
59
3. Sample
Data from 232 working paper reviews--also called quality control reviews
(QCRs) of public accounting firm audits of Texas ISDs were obtained
through an analysis of the Texas Education Agency (TEA) files. The TEA
regulates Texas school districts. Each QCR culminates in a letter of
findings to the public accounting firm by the agency's audit director. These
letters served as the basis for a comprehensive, direct measure of audit
quality. Additional information used in our study came from the QCR
files, the audit reports, and other reports issued by the TEA. This is the
same data base used by Deis and Giroux (1992a), Deis and Giroux (1994),
and Giroux et al. (1995).
Independent public accounting firms were limited to one observation in
the analysis. In other words, the sample was determined by ISDs having
QCRs. Of 308 QCRs conducted by TEA, 232 were used in this analysis.
QCRs conducted on 1983 and 1984 independent audits were not utilized
for lack of fee and hour data. Audits conducted by Big Eight, non-Big
Eight national, and regional public accounting firms were eliminated (10
public accounting firms in total were eliminated).4
4. The Empirical Models
Audit contracting characteristics affect audit fees and audit hours differently (Palmrose 1989, p. 494), hence two empirical models were used to
evaluate the public sector audit market. The first model was developed
independently, based, in part, on previous audit fee studies such as Rubin
(1988, p. 230). Also, three of the variables SIZE, CLIENTS, and HOURS
--were used in Deis and Giroux 1992a, see pp. 467-468. The variables
YR1, YR2, CLIENTS, and SIZE were used in Giroux et al. (1995, pp.
72-76). The model is used to separately consider each dependent variable.
The independent variables are described below.
In(FEE), In(HOURS)
= a 0 + Blln(SIZE) + B21n(PCI)
+ B3OFFICES + B4CLIENTS + BsCAFR + B6IAC + B7OPIN
+ BsYEAR1 + B9YEAR2 + B101n(QUALSCORE) + e
(1)
4 For additional information regarding the agency's process and the m a n n e r in which the QCRs
were coded, see Deis et al. (1990, pp. 37-38), Deis and Giroux (1992, pp. 468-469), Deis and Giroux
(1994, p. 219), or Giroux et al. (1995, pp. 68-69). (The results did not change when Big Eight and
national CPA firms were included.) Most other audit fee studies have included a d u m m y variable to
distinguish auditor size category. In our current study, the Q C R s for large public accounting firms
were associated with very large ISDs which have different characteristics from small ISDs. Also, the
n u m b e r of offices for the public accounting firm ( O F F I C E S ) is similar to brand n a m e in interpretation.
60
D.R. Deis, Jr. and G. Giroux
where the dependent variables are the natural logs of actual audit fees and
"actual audit hours" (see Deis and Giroux 1992a, p. 468; Giroux et al.
1995, pp. 72-73), and the explanatory variables are (note that the sources
of data are indicated):
Auditee characteristics:
S I Z E = in terms of students attending school in a school district;
see Deis and Giroux (1992a, p. 467) for a description of this
variable ( T E A Files);
PCI = natural log of per capita income in the school district
according to the 1980 U.S. Census
(U.S. Bureau of the Census tape STF-3F(1984));
Brand n a m e and experience:
O F F I C E S = n u m b e r of offices for the public accounting firm
(firm letterhead on auditor's opinion);
C L I E N T S = In of n u m b e r of school districts audited; see Deis and
Giroux (1992a, p. 467) for a description
of this variable ( T E A files);
Audit production costs:
C A F R = 1, if the audit engagement was a comprehensive annual financial
report ( C A F R ) , 0 otherwise (from the reports themselves);
(the audit can be conducted on a C A F R or general
purpose financial statements (GPFS) basis);
I A C = 1, if the auditor reported a material weakness in internal
control, 0 otherwise (from C A F R ) ;
O P I N = 1, if the auditor qualified the independent auditor's report,
0 otherwise (from C A F R ) ;
Initial audits:
Y E A R 1 = 1, if the audit was an initial engagement,
0 otherwise ( T E A files);
Y E A R 2 -- 1, if the audit was a second year engagement,
0 otherwise ( T E A files); and
The Effect of Auditor Changes
61
Product Differentiation:
Q U A L S C O R E = a metric (score) of 19 possible audit quality deficiencies
discovered in the QCR weighted by their importance
to the referral decision. As described below in Section
4.5, this measure differs from Deis and Giroux
(1992a, pp. 468-469), but is the same as the one used
by Giroux et al. (1995, p. 70).
As OFFICES and CLIENTS both represent potential ex ante measures
of audit quality and Q U A L S C O R E measures ex post quality, the models
were estimated with and without Q U A L S C O R E for comparative purposes. 5
To test the effects of initial engagements on audit quality the following
model was estimated:
l n ( Q U A L S C O R E ) = a 0 - B l l n ( S I Z E ) + B 2 ( W E A L T H ) + B3CLIENTS
+ B4OFFICES + B 5P E E R + B6TIME
+ B71n(HOURS) + BsYEAR1 + B9YEAR2
(2)
The quality model is a modified version of Deis and Giroux (1992a, pp.
467-468), substituting YEAR1 and Y E A R 2 for the continuous tenure
variable, adding OFFICES, and omitting three insignificant variables in
that study (see Deis and Giroux 1992a, pp. 467-471 for a description of the
variables). The variables SIZE, CLIENTS, OFFICES, H O U R S , YEAR1,
and YEAR2, are described earlier in the paper. The "assessed property tax
value per student" ( W E A L T H ) "proxies for the financial health" of the
ISD (Deis and Giroux 1992, p. 471). For those independent audit firms
voluntarily participating in a peer review program, the indicator variable
P E E R takes on a value of "1" and "0" otherwise (Deis and Giroux 1992,
pp. 467 and 470; Giroux et al. 1995, pp. 76-77). The timeliness of the
independent audit (TIME) is expressed as the percentage of the 120 day
period following the fiscal year end (i.e., August 31) allowed by Texas state
law to complete the audit (Deis and Giroux 1992, p. 471).
As previously discussed, Giroux et al. (1995, pp. 76-78) analyzed audit
quality using analysis of variance (ANOVA) and multiple comparison tests.
In that analysis audit quality was significantly higher for peer reviewed
ISDs, first year audits relative to those with audit tenure of 20 or more
years, when the audit firm had five or more ISD clients, and when the
audit reports were submitted on a timely basis (Giroux et al. 1995, pp.
77-78).
5This point was made by an anonymous reviewer. Moreover, Ward et al. (1994, pp. 406-407)
found fee premia associated with industry expertise.
62
D.R. Deis, Jr. and G. Giroux
4.1 Auditee Characteristics
Larger public sector organizations generally require more independent
audit time because they enter into more financial transactions (Simunic
1980, pp. 172-173; Rubin 1988, p. 221) or because taxpayers will demand
more information (Wallace 1986, p. 40). Therefore, the size of the government unit being audited should be positively related to both the price of
the audit engagement (FEE) and audit hours (HOURS). The natural logs
of average daily attendance (SIZE), audit fees (FEE), and audit hours
(HOURS) have been used. These transformations provide a better fit to
ordinary least squares (OLS) assumptions and are consistent with past
private sector and public sector research (Baber et al. 1987, p. 298;
Palmrose 1989, pp. 491-492; Roberts and Glezen 1990, pp. 130-132;
Rubin 1988, p. 225).
Natural log of per capital income (PCI) in the ISDs proxies for underlying financial strength (see, for example, Baber et al. 1987, p. 298). Following Baber et al. (1987, p. 298) an inverse relationship is expected for PCI.
4.2 Brand Name and Industry Experience
Although client management can observe the audit production process
(albeit imperfectly), it is costly for users (e.g., citizens, regulators, investors
and creditors) to observe the performance of the independent auditor.
Users therefore rely on the independent auditor's brand name or reputation as an industry specialist as surrogate measures of audit service
credibility (DeAngelo 1981b, pp. 187-192; Klein and Leffler 1981, p. 620;
Simunic and Stein 1986, p. 7).
Although brand name (e.g., Big Eight and Non-big Eight) generally
served as a proxy for auditor quality in prior research (e.g., Rubin 1988, p.
20), a few studies have used the concepts within a single segment of audit
producers. O'Keefe and Westort (1992, pp. 56-57) used the number of
CPAs employed by small public accounting firms to proxy for brand name
in a study of compliance with reporting standards. Shockley and Holt
(1983, p. 556) found that users perceived industry market shares to
differentiate levels of specialization among Big Eight auditors.
In our present study, the number of offices maintained by each public
accounting firm (OFFICES) proxies for brand name and the number of
ISD clients (CLIENTS) proxies for industry experience. Following Palmrose (1989, p. 494) a positive relationship is expected between OFFICES
and audit fees and audit hours. Conversely, a negative association is
expected between industry experience on audit fees and audit hours. ISD
audits are highly regulated, including federal single audit requirements
and additional TEA rules (Deis et al. 1990, pp. 36-37). Firms specializing
The Effect of Auditor Changes
63
in ISD audits are posited to be more efficient in conducting the audit and,
hence, this efficiency will be reflected in fewer audit hours. Giroux et al.
(1995, p. 75) using the same data found a positive significant relation
between audit hours and the number of clients. However, they (1995, p. 75)
did not find a significant relationship between industry experience (clients)
and audit fees or audit fees-per hour. Assuming a competitive market,
specialists are also expected to share some of the benefits of their lower
production costs in the form of lower audit fees with their clients. 6
4.3 Audit Production Costs
Several variables were included in the fee and hours model (model 1) to
capture the complexity of the audit and factors associated with audit
production costs as identified by prior research. According to generally
accepted accounting principles (AICPA 1994, pp. 143-145), Texas ISDs
are not required to issue a CAFR to meet TEA report requirements. Thus,
when the audit engagement is on a CAFR, the audit contract is expected
to reflect the auditor's additional reporting responsibilities concomitantly
with a higher audit fee and more audit hours (Rubin 1988, pp. 222-223).
Similar to Roberts and Glezen (1990, p. 131) and Raman and Wilson
(1992, p. 288), two report variables were evaluated: 1) qualified audit
opinions (OPIN), 7 and 2) the report on internal accounting controls (IAC).
Traditionally, both private and public sector audit fee studies include an
indicator variable for modified (e.g., qualified) audit opinions as a measure
of the auditor's risk of loss arising from various auditing problems (Simunic
1980, pp. 173-174). By virtue of the expanded reporting requirements
under The Single Audit Act of 1984, independent auditors of public sector
entities must disclose material weaknesses in the client's internal control
environment in a report on internal controls. Roberts and Glezen claim
that this report reveals the "quality of the client's internal financial
6 0 ' K e e f e et al. (1994, p. 65) r e p o r t e d a significant negative correlation between audit fees a n d
the audit firm's industry specific knowledge (i.e., the n u m b e r of school district audits) in their study
of violations o f generally a c c e p t e d a c c o u n t i n g principles r e p o r t i n g s t a n d a r d s d e t e c t e d in 935
California 1986 school district audits. W a r d et al. (1994, p. 407), in contrast, f o u n d higher audit fees
associated with a single regional audit firm responsible for 21% of the s a m p l e d Michigan municipalities. As o u r sample was c o m p r i s e d o f o n e audit p e r firm, we w e r e u n a b l e to test for a similar effect.
Moreover, with nearly 1,200 Texas ISDs it is unlikely for a n y firm to attain a 2 0 % s h a r e of the
market.
7 O u r study focuses on r e p o r t e d b r e a c h e s in the c u r r e n t y e a r which is consistent with earlier
studies (e.g., B a b e r et al. 1987, p. 303; R o b e r t s a n d Glezen, 1990, p. 140; a n d R u b i n 1988, p. 223).
A n alternative is to use the p r i o r y e a r r e p o r t s on the a s s u m p t i o n t h a t r e p o r t e d b r e a c h e s f r o m the
previous y e a r w o u l d have i n c r e a s e d audit risks, driving u p b o t h fees a n d hours. Predictions are the
s a m e in either case, as a r e results.
64
D.R. Deis, Jr. and G. Giroux
reporting system" (1990, p. 135). Since both OPIN and IAC capture
potentially time consuming aspects of the audit engagement, a positive
relationship is predicted for both report-related variables on audit fees and
audit hours.
4.4 Initial Audits
Several recent studies detected lower audit fees in the initial and second
engagement years (Baber et al. 1987, p. 303; Ettredge and Greenberg 1990,
p. 208; Roberts et al. 1990, p. 227; Simon and Francis 1988, p. 261; Turpen
1990, p. 68). Giroux et al. (1995, p. 74) found that Iow-baUing existed
during the first year of an audit, but not the second year. The same
categorical variable used by Giroux et al. (1995, p. 73) were included here
to evaluate initial year (YEAR1) and second year (YEAR2) competitive
effects on audit fees and audit hours.
4.5 Product Differentiation
Audit quality is the key feature of the product differentiation explanation
for observed variations in audit pricing. If independent auditors can
convince clients that their product differs, then audit fees can vary (YardIcy et al. 1992, p. 165-171). Public accounting firms are assumed to plan a
level of audit quality they intend to supply and this information is incorporated in audit fees and audit hours. (Simunic and Stein 1987, pp. 1-4).
QCR results were used in our study to measure audit quality ex post,
anticipating that this reflects the independent auditor's ex ante plan
regarding the level of audit quality to be supplied. As Copley et al. (1994,
p. 248) asserted, a positive association exists between the planned level of
audit assurance and the propensity to be in compliance with professional
standards. Public accounting firms planning on lower quality levels will (on
average) exhibit lower frequencies of compliance with professional standards than firms planning on higher quality levels. Hence, ex post measures of compliance with professional standards are sufficiently correlated
with audit quality to reasonably proxy for ex ante planned quality.
We performed a content analysis of 232 letters of findings in the QCR
files and found 19 categories of findings common to many audits (Deis and
Giroux 1992a, p. 468; Deis and Giroux 1994, p. 219; Giroux et al. 1995,
p. 67). TEA's "director of audits ranked the importance of each category,
from 1 (most important) to 19 (least important)," in considering whether or
not the public accounting firm should be referred to the Texas State Board
of Public Accountancy (Deis and Giroux 1992a, p. 468; also see Deis and
Giroux 1994, pp. 219-220 and Giroux et al. 1995, pp. 67-68).
A metric for audit quality was built by multiplying each category by its
(negative) rank. The least severe category (substantive tests: minor) was
The Effect of Auditor Changes
65
weighted - 1; the most severe (audit program: major) by - 19. The weights
were summed, resulting in the metric Q U A L S C O R E . 8 Both audit fees
(FEE) and audit hours ( H O U R S ) are projected to be directly related to
Q U A L S C O R E . This measure is similar to Deis and Giroux (1992a, pp.
468-469). They divided the deficiency code by its assigned rank. Under
their measure, a zero was the highest possible value and a higher quality
audit had a lower quality score. Q U A L S C O R E is the same measure as
" S C O R E " , the quality measure used by Giroux et al. (1995, p. 70-71) in
their A N O V A model.
The analysis of audit quality ( Q U A L S C O R E ) on audit fees and audit
hours raises certain econometric issues. It is evident from previous studies
that audit hours and audit fees affect audit quality (Deis and Giroux
1992a, p. 476; O'I(eefe and Westort 1992, p. 65; O'Keefe et al. 1994,
p. 261; Copley et al. 1994, pp. 253-254). Our current study proposes to
reverse the order and investigate the effects of audit quality on audit fees
and audit hours. Therefore, QUAI_SCORE may be an endogenous variable which is not independent of the error term in the proposed empirical
m o d e l s - - t h u s violating OLS assumptions and suggesting simultaneous
equation estimation (SEE) would be appropriate (Maddala 1988, p. 293).
However, use of such an approach is inefficient when the error terms are
independent. Fortunately, Hausman (1978) provided a specification test to
determine whether OLS or SEE is most appropriate. Hausman's specification error test is asymptotically equivalent to a test of the covariance
between the error terms produced from OLS and SEE. The test statistic is
(Maddala 1988, pp. 435-439):
M = qZ/var(q)
(3)
where
q = B 1 - Bo
and
var(q) = var(B 1) - var(B 0)
using B 1 indicates estimators from SEE and B 2 estimators are from OLS.
The test statistics, m, has X z distribution with k degrees of freedom (k
number of parameters). The calculated m statistics for our study were
0.825 for the audit hours model and 0.693 for the audit fee model. Neither
of these m statistics were significant ( p > 0.10, two tailed) ( X0.90
2 d.f.
8An alternative m e a s u r e of quality is a categorical variable, w h e r e 1 = a report referred to the
Texas State Board of Public A c c o u n t a n c y - - 3 2 or about 14% were referred (see Deis et al. 1990, pp.
3 7 - 3 8 for a description of the process; also, see Deis and Giroux 1992a, p. 472). When this d u m m y
variable replaced Q U A L I T Y in the regression runs, results indicated that referred audits have both
lower fees (significant at .01, one-tail test) and lower hours (significant at .0001). O t h e r quality
metrics produced similar results in regression runs, suggesting the results are robust regardless of
the m a n n e r in which audit deficiencies are indexed.
66
D.R. Deis, Jr. and G. Giroux
15 = 22.31). Hence, Q U A L S C O R E and the error terms in the models are
not correlated and OLS estimations are justified (Maddala 1988, 437). 9
5. Empirical Results
Descriptive statistics appear in Table 1 and a correlation matrix for all
variables appears in Table 2. Thirty-two independent auditors in our
sample were referred to the Texas State Board of Public Accountancy for
particularly low quality audits (see Deis and Giroux 1992a, p. 472). These
audits were considered so deficient relative to professional standards, that
actions by the Texas State Board of Public Accountancy were deemed
necessary. Board action can include imposing additional professional education, other remedial actions, or punitive actions. The Board can revoke a
CPA's license in an extreme case (Deis et al. 1990, pp. 37-38). Most of the
correlations among the independent variables were low, with the largest
correlation between SIZE and CLIENTS at .32 (Table 2).
Tables 3 and 4 present the results of the audit fee and audit hours
models. Several diagnostic procedures were used to assure that the results
were not unduly influenced by outliers, multicollinearity, or heteroskedasticity. 1° The full model estimating In (FEES) was significant at .0001
( F = 41.26, two tailed test) and has comparable explanatory power (adjusted R 2 = 0.64) to Roberts and Glezen's (1990, p. 142) model of 356
Texas ISDs (adjusted R 2 = 0.66) and Giroux et al. (1995, p. 75) using the
same data (R 2 = .599). The full model estimating audit hours was also
significant at .001 ( F = 22.63, two tailed test) and explains about half the
variation in audit hours (R 2 = 0.47), compared to Giroux et al. (1995, p.
75) where R 2 = .427. As results for the reduced models are similar to
those of the full models, the discussion to follow will address the results in
the full models.
Most of the variables were significant in the expected direction to both
audit fees and audit hours. As found in previous studies (e.g., Rubin 1988,
p. 230; Giroux et al. (1995, p. 75)), large clients (SIZE) had higher audit
fees and more audit hours. ISD financial strength (PCI) was negatively and
significantly associated with both audit hours and fees, confirming Baber et
al. (1987, p. 303). The larger local CPA firms (OFFICES) received fee
9 We are grateful to R. Carter Hill for help in conducting the H a u s m a n test. The S H A Z A M
statistical package was used to conduct this test.
10 No important violations were detected. Plots of the residuals against the predicted value in
each model depict a cloud-like pattern. Univariate statistics on the residual terms from each model
indicated a m e a n of zero ( p = 1.0) and the assumption of normality, using the Kolmogorov-Smirnov
test, could not be rejected at the 0.15 level. Variance inflation factors (VIF) were low, ranging from
1.03 to 1.17 for each model. Condition indexes were 1.64 or less in both models, considerably less
than the lowest threshold of 15 suggested by Belsley et al. (1980, p. 157) over which multicollinearity
becomes a problem. Based on these diagnostic procedures, the results of both regression models
can be interpreted without concern that they are influenced by violations in OLS assumptions.
The Effect of Auditor Changes
67
Table 1. Descriptive Statistics for a Sample of 232 Quality Control Reviews of
Independent Audits of Texas Independent School Districts Conducted
from 1984-1989
Continuous Variables:
Audit fee (FEE)
Audit hours ( H O U R S )
Average daily attendance (SIZE)
Per capita income (PCI)
No. of firms offices (OFFICES)
No. of ISD clients (CLIENTS)
Q C R weighted metric ( Q U A L S C O R E )
Mean
Minimum
Maximum
$10,416
294
3,616
$6,229
1.3
3.3
- 36.9
800
38
36
2,642
1
1
- 157
50,000
1,994
44,776
12,827
8
30
0
Categorical Variables:
Frequency
% (Rounded)
No. of comprehensive annual financial
reports (CAFR)
No. of internal control breaches (IAC)
No. of auditor reported breaches (OPIN)
No. of initial audit engagements
(YEAR1)
No. of 2nd year engagements (YEAR2)
25
11
86
11
22
37
5
9
20
9
premiums indicative of a brand name within that segment of the audit
market. Audit hours were significant and positively related with OFFICES.
Audit hours were negative and significantly associated with the number
of ISD clients audited by the audit firm, the same as found by Giroux et al.
(1995, p. 75). CLIENTS also was negative and significantly associated with
Table 2. Spearman Rank Order Correlations
FEE
HOUR
SIZE
PCI
OFFICES
CLIE
CAFR
IAC
OPIN
YR1
YR2
QUAL
HOURS
.76*
.75*
.62*
-.00
-.01
.15"**
.12
.24**
.10
.29**
.24**
-.01
.01
.07
.09
-.09
.05
-.06
-.03
.27**
.33*
SIZE
.14"**
.11
.32*
.28**
-.13
-.00
-.04
-.06
.15"**
* Significant at .0001.
** Significant at .01.
*** Significant at .10.
(All significance tests are two-tailed.)
n = 232
PCI O F F I C E S C L I E N T C A F R IAC O P I N Y R 1 YR2
-.08
.02
.08
-.24**
-.10
-.03
-.06
-.07
.24**
--.06
-.02
.08
.08
--.02
.09
--.05
--.04
-.02
-.07
.15"**
.17"**
-.15
- . 0 1 .21"*
- . 1 1 .09
-.00
.00 - . 1 0
- . 0 6 .02
.00.10
.03 .12 .02
68
D.R. Deis, Jr. and G. Giroux
T a b l e 3. E m p i r i c a l
Model Results Ordinary
L o g o f A u d i t F e e is t h e D e p e n d e n t
Least Squares Regression: Natural
Variable (Model
1)
n = 232
Description
Predicted
Sign
Full M o d e l
Parameter
Estimate
(t-Value)
Reduced Model:
Parameter
Estimate
(t-Value)
9.590
9.310
0.386
(17.36)*
- 0.360
( - 2.97) **
0.389
(17.18)*
- 0.380
( - 3.08)* *
0.053
(1.83)***
-0.015
( - 1.57)***
0.072
(2.48)**
-0.010
( - 1.25)
Intercept
Auditee Characteristics:
A v e r a g e daily a t t e n d a n c e
ln(SIZE)
P e r capital i n c o m e I n ( P C I )
+
-
Brand Name & Specialization:
No. of firm O F F I C E S
+
No. of I S D C L I E N T S
-
Audit Production Costs:
CAFR engagement
+
Internal accounting
controls ( I A C )
Qualified opinion (OPIN)
+
0.228
(2.31)***
0.109
(1.73)***
0.201
(1.45)***
+
0.239
(2.38)**
0.124
(1.94)***
0.172
(1.22)
Inhial Audits:
Initial e n g a g e m e n t
( Y E A R 1)
Second year engagement
( Y E A R 2)
-
((-
0.150
1.49)***
0.023
0.22)
- 0.120
(1.18)
- 0.013
( - 0.12)
Product Differentiation:
In ( Q U A L S C O R E )
Adj. R 2
F statistic (two-tailed test)
+
0.117
(3.08)**
0.635
41.26"
0.622
43.15"
* Significant at .0001.
* * Significant at .01.
*** Significant at .10.
(All significance tests are one-tailed--except when noted to be two-tailed.)
audit fees in the full model but not in the reduced model with quality
omitted. The reduced model result is similar to the fee model result
reported by Giroux et al. (1995, p. 75) which suggests that quality is an
important omitted variable in that study in regards to industry experience
(CLIENTS). As expected, audit efficiency was a benefit produced from
investments in industry expertise, with economies passed on to clients as
The Effect of Auditor Changes
69
Table 4. Empirical Model Results Ordinary Least Squares Regressiop: Natural
Log of Audit Hours is the Dependent Variable (Model 1)
n = 232
Description
Predicted
Sign
Intercept
Full Model
Parameter
Estimate
(t-Value)
Reduced Model:
Parameter
Estimate
(t-Value)
6.457
6.023
Audited Characteristics:
Average daily attendance
In(SIZE)
Per capita income ln(PC1)
+
-
0.308
(11.74)*
- 0.322
( - 2.26)***
0,312
(11,54)*
- 0,35
( - 2,39)**
0.057
(1.67)***
-0.021
(-2.26)***
0,086
(2.48)**
-0,017
(-1.82)***
0.273
(2.35)**
0.131
(1.76)***
0.275
(1.69)***
0.290
(2.42)**
0.154
(2.01)***
0,231
(1.37)***
0.159
(1.34)***
0.028
(0.23)
0.204
(1.68)***
0.044
(0.35)
0.179
(4.00)*
0.472
21.63"
0.436
20.84*
Brand Name & Specialization:
No. of firm O F F I C E S
+
No. of ISD CLIENTS
-
Audit Production Costs:
CAFR engagement
+
Internal accounting
controls (IAC)
Qualified opinion (OPIN)
+
+
Initial Audits:
Initial engagement
(YEAR 1)
Second year engagement
( Y E A R 2)
+
+
Product Differentiation:
ln(QUALSCORE)
Adj. R 2
F statistic (two-tailed test)
+
* Significant at.0001.
** Significant at .01.
*** Significant at .10.
(Allsignificance tests are one-tailed--except when noted to be two-tailed.)
lower costs, t~ Moreover, both Deis and Giroux (1992a, p. 476), O'Keefe et
al. (1994, p. 261), and Giroux et al. (1995, p. 77) found higher audit quality
associated with increases in industry-specific knowledge (i.e., the number
11 Both OFFICES and CLIENTS are potential ex ante proxies for audit quality demanded. Both
are significantly correlated with QUALSCORE. Additional analysis (using Duncan's Multiple
Range Test) indicated that auditors with five or more ISD clients had significantly higher quality
than those with fewer clients (Huck et al. 1974, pp. 68-69). Offices was not significant using
Duncan's Multiple Range Test.
70
D.R. Deis, Jr. and G. Giroux
of ISD clients). Hence, industry-specific knowledge is an important attribute for consideration in audit procurement as it is accompanied by
lower audit costs and higher audit quality in this segment.
Similar to related studies (e.g., Rubin 1988, p. 224), variables identified
as components of audit production costs were significant to both audit
hours and audit fees. CAFR engagements entail more audit effort which is
reflected in both higher fees (Rubin 1988, p. 230) and additional audit
hours. Similar to prior public sector studies (O'Keefe et al. 1994, p. 44;
Raman and Wilson 1992, p. 288; Roberts and Glezen 1990, p. 135), higher
audit fees were associated with reports of weaknesses in internal accounting controls (IAC) and qualified audit opinions (OPIN). Similar to private
sector studies (Davis et al. 1993, p. 143; Palmrose 1986, p. 106 and 1989,
p. 494), public sector report modifications (OPIN and IAC) were also
associated with higher audit hours.
The link between report modifications and audit hours is clear--more
time is expended when difficulties are encountered during the engagement. The interpretation between report modifications and audit fees,
however, is somewhat problematic. It is unlikely that independent auditors
can freely adjust the price of public sector audits to accommodate unanticipated audit hours. Nonetheless, this result (i.e., higher audit fees associated with current year report modifications) has been found in public
sector studies (Rubin 1988, p. 230; Roberts and Glezen 1990, p. 140). As it
is our understanding that there are a large number of fixed fee engagements in the public sector, these findings imply that the independent
auditor develops an ex ante expectation that reporting problems may exist
and plans the audit accordingly.
Significantly lower audit fees were detected in 22 initial engagements
(YEAR1) in the sample. Second year engagement (YEAR2) audit fees
were not significantly different in the model (model 1--see Table 3).
Initial year engagements were positive and significantly associated with
audit hours. The same results were found in Giroux et al. (1995, p. 75).
This evidence serves to further confirm the practice of low bailing by
auditors to obtain new clients. These results generally support DeAngelo
(1981a, p. 192), with lower fees in the initial year only.
Comparing the full model result for initial engagements on audit fees to
those in the reduced model (Table 3) reveals that initial engagements are
significant only in the full model. A similar comparison of full and reduced
audit hour models (Table 4) indicates that initial engagements are significant in both models. These reduced model results for initial engagements
coincide with those reported by Palmrose (1989) who did not have a direct
audit quality measure. She (1989, pp. 496 and 498) reported insignificant
results for initial engagements in her fee model but significantly less audit
hours as tenure lengthens. After considering audit quality variations (i.e.,
in the full models), we see that initial audit fees are in fact lower while
The Effect of Auditor Changes
71
audit hours are higher. This is additional evidence to strengthen Palmrose's
suggestion that profit margins improve over the auditor's tenure as audit
hours decrease (1989, p. 498). Our study also suggests that without controlling for audit quality cross-sectional audit fee studies may not detect price
cutting of initial engagements.
Q U A L S C O R E was a significant determinant of both audit fees ( p =
0.001) and audit hours ( p = 0.0001)--see Tables 3 and 4 (one-tailed tests).
Audits with high quality had higher fees and more audit effort (hours). A t
test of Q U A L S C O R E when the sample was split after the second year of
auditor tenure confirms this relationship. Using a t test in a separate
analysis we found that audits conducted in the first two years of tenure had
significantly higher audit quality than audits conducted by auditors with
three or more years of tenure ( p = 0.001, two tailed test). First and second
year audits had a similar SIZE variable (t statistic = -0.11); hence, the
associations between YEAR1 and Y E A R 2 on F E E and H O U R and
Q U A L S C O R E (in the t test) were not driven by audited size. 12' 13 Moreover, as shown in Table 5, the multivariant test for the effects of initial
audit engagements on audit quality also demonstrates higher quality on
initial audit engagements (t = 1.37, p = .09, one-tailed test).
6. Conclusions
For nearly twenty years widespread concerns have been expressed that
aggressive price competition within the market for audit services erodes
the quality of work provided by auditors. Based partially on these concerns, federal and state agency oversight of the audit profession has
increased dramatically and continues to expand. A central tenet behind
this heightened regulation is the purported detrimental relationship between 'low-ball' audit pricing and audit quality. Under this scenario, when
the initial year audit is conducted at a substantial price discount (i.e., a
'low-ball' fee) audit quality suffers because audit firms (1) expend less
effort so as to maintain profit margins in light of the fee reductions or (2)
they fail to report problems in order to protect "receivables" in future
audit engagements. The 'low-bailing' part of this scenario has been well
documented in academic studies and by the business media. Heretofore
research on the critical link between 'low-bailing' and audit quality,
12As an alternative test, n u m b e r of years of audit tenure ( T E N U R E ) replaced Y E A R 1 and
Y E A R 2 and a t e n u r e / q u a l i t y interaction term added in additional regression runs. Regression
results indicate that T E N U R E is significant and positive for fees, but insignificant for hours. T h e
interaction t e r m was significant for hours only. T h e s e results reinforce the importance of the
relationship between audit tenure and both fees and hours.
13 To d e t e r m i n e if the r e f e r r e d audits were biasing the results, the regression models were rerun
excluding the 32 r e f e r r e d audits. No differences were detected. Thus, the subset of " a c c e p t a b l e "
audits can be interpreted as product differentiated.
72
D.R. Deis, Jr. and G. Giroux
Table 5. Empirical Model Results Ordinary Least Squares Regression: Natural
Log
of A u d i t Q u a l i t y is the D e p e n d e n t V a r i a b l e ( M o d e l
2)
n = 232
Description
Predicted
Sign
Intercept
Parameter
(t-Value)
-- 3 . 2 9 8
Ln(SIZE)
-
WEALTH
-
- 0.124
( - 2.67)**
- 0.091
( - 1.32)* **
CLIENTS
+
0.018
(1.32)***
OFFICES
+
0.121
(2.55)**
PEER
+
0.353
(2.96)**
TIME
-
- 0.588
( - 2.66)**
HOURS
+
0.368
(4.18)*
YEAR1
+
0.227
(1.37)***
YEAR2
+
0.072
(0.42)
Adj. R 2
F
statistic (two-tailed test)
0.175
6.43*
* Significant at .0001.
** Significant at .01.
*** Significant at .10.
(All significance tests are one-tailed--except when noted to be two-tailed.)
however, has been inconclusive since audit quality was unobservable.
Ironically, increased oversight of the audit profession has, for the first
time, provided researchers with direct measures of audit quality and,
therefore, an opportunity to address the twenty year old question: D o
auditors lower audit quality on initial year 'low-ball' engagements?
DeAngelo's (1981a and 1981b) seminal theoretical work on audit quality
has been the basis for a myriad of audit market studies; yet, her primary
conclusions have been difficult to test empirically for the data limitation
reason just cited. Using state regulator QCR data, Deis and Giroux (1992)
developed a direct, continuous measurement of actual audit quality to
address a series of questions stemming from DeAngelo's propositions. In
our initial study (Deis and Giroux 1992, p. 476) we confirmed two elements
of audit quality envisioned by DeAngelo (1981a and 1981b). First, we
detected improved audit quality associated with two measures of the audit
firm's investment in reputation (industry expertise and voluntary peer
The Effect of Auditor Changes
73
review section). This was termed by DeAngelo as the "collateral" effect
(1981b, pp. 190-191). Second, we found audit quality to decline with the
length of auditor tenure (defined as the number of years of audit tenure).
While our 1992 paper did not explicitly test the 'low-bailing' issue, this
second finding is consistent with both DeAngelo's explanation of opportunistic behavior among incumbent auditors (1981b, p. 189) and the
"complacency" effect described by Mautz and Sharaf (1961, p. 208).
The positive association between audit quality and voluntary membership in AICPA's peer review program detected in our 1992 paper was
expounded upon in our 1995 paper (Giroux et al. 1995, p. 67). As an
observable quality surrogate, membership in peer review programs should,
under DeAngelo's theory, signal an audit firm's investment in reputation
and, hence, be assoc.iated with audit fee premia. That is precisely what we
found (1995, p. 75). This finding is of interest for two reasons: (1) the
existence of market incentives to voluntarily join a peer review program
calls into question mandatory peer review participation required by the
United States General Accounting Office's 1988 revision to Governmental
Auditing Standards and (2) the results conflict with a similar study of
private sector audits (Francis et al. 1990, p. 374). Audit firm tenure was
also explored further in this study. Evidence of low-bailing was found in
the audit fee regression model (1995, p. 75) and a categorical tenure
variable was negatively related to audit quality in an analysis of variance
(ANOVA) model (1995, p. 77). We did not directly test the 'low-ball' to
audit quality issue in this paper, however, since peer review membership
was the focal point of both the regression and ANOVA models. Hence,
DeAngelo's propositions on 'low-bailing' were not conclusively evaluated
in the 1992 or the 1995 papers.
Building on our earlier two papers, the primary contribution of this
study is that the effects of 'low-bailing' audit fees is tested in light of a
direct audit quality measure. We found lower fees among first year audits
without concomitant audit quality deterioration--as DeAngelo predicted.
Accordingly, as DeAngelo concludes (1981a, p. 126), regulation aimed at
curbing the 'low-bailing' practice should be reconsidered given the lack of
evidence that a price-for-quality tradeoff exists. In addition, since the
results for other variables in the study's empirical models were consistent
with related public sector literature (Baber et al. 1987, p. 303; Roberts and
Glezen, 1990, p. 140; Rubin 1988, p. 230; Ward et al. 1994, pp. 406-407),
previous studies lacking a direct audit quality measure need not be
reestimated with one. Moreover, given the similarities between private
sector and public sector audit economic studies (Rubin 1988, pp. 234-235)
it is plausible that our findings and conclusions are applicable to the
private sector audit market. Hence, conclusions in previous studies regarding: auditor selection, non-aud
RESEARCH
NOTE
The Effect of Auditor Changes on Audit
Fees, Audit Hours, and Audit Quality
Donald R. Deis, Jr. and Gary Giroux
Our study investigates the relationship of independent auditor fees, hours and
quality, with particular emphasis on initial audits. Using audits of Texas independent school districts, empirical results indicated specific low bailing relationships.
First year independent audits had statistically significant lower audit fees. Despite
the lower fees, quality was higher and more audit hours were utilized.
1. Introduction
Our study investigates the effect of independent auditor changes on fees,
independent audit effort (as measured by audit hours), and independent
audit quality. The empirical analysis is based on audits conducted by local
and regional public accounting firms of independent school districts (ISDs)
in Texas. We found that initial audits were associated with lower fees,
higher quality, and higher audit hours at the .10 level (one tailed). The
results also indicated that both audit fees and audit hours were significantly related to audit quality in a manner consistent with the explanation
that audits are product differentiated in the market.
The empirical findings presented in our paper support previous research. Our paper confirms that both audit fees and audit hours are
influenced by audit quality. The evidence suggests that independent audi-
Address correspondence to" Professor Gary Giroux, Department of Accounting, Texas A & M
University, College Station, 'IX 77843.
Journal of Accounting and Public Policy, 15, 55-76 (1996)
© 1996 Elsevier Science Inc.
655 Avenue of the Americas, New York, NY 10010
0278-4254/96/$15.00
SSDI 0278-4254(95)00041-0
56
D.R. Deis, Jr. and G. Giroux
tors do not sacrifice quality to obtain a new audit client, though lower
audit fees were detected among initial audit engagements. Audit hours and
audit quality were higher among initial audits, confirming a conclusion
reported by Raman and Wilson (1992, p. 292) in their study of a sample of
U.S. cities over four years (1984-1987).
No previous study has compared audit fees and audit hours with both
audit service quality and independent auditor changes in a single analysis.
Initial audits (commonly associated with low bailing) may result in lower
fees but the relationship to audit effort has not been the subject of a great
deal of research (see Giroux et al. 1995; Davis et al. 1993; Palmrose 1989
for research on the topic) and never in combination with a measure of
audit service quality. If audit effort declines with audit fees on initial
audits, audit quality also should decline. DeAngelo (1981a, pp. 114,
118-120; 1981b, pp. 187-194), however, suggests that audit effort will be
maintained so that incumbent auditors can generate quasi-rents in future
years.
We investigate other issues associated with the independent audit market. Our study examines other aspects of audit economics (e.g., brand
name, specialization, and product costs) in the small CPA firm sector. 1 We
also extend Palmrose's (1986) and (1989) studies of the determinants of
audit hours to the public sector.
Further, we build upon the work of Deis and Giroux (1992) and Giroux
et al. (1995). The former study evaluated the potential determinants of
actual independent audit quality using a metric of audit quality based on
nineteen features of the independent auditor's work as documented in
working paper reviews conducted by state auditors at the Texas Education
Agency. The latter study compared the audit fees, auditor hours, and audit
quality of independent audits conducted by audit firms participating in a
voluntary peer review program to those not participating.
2. P r i o r R e s e a r c h o n Initial A u d i t s
Conventional wisdom suggests that when public accounting firms offer
price discounts to capture new audit clients, audit quality may be compromised. 2 A competing explanation advanced by DeAngelo (1981a, pp. 114,
118-120; 1981b, pp. 187-194), however, contends that low balling (i.e.,
bidding below the avoidable costs for the first year audit) represents an
i The National Commission on F r a u d u l e n t Financial R e p o r t i n g ( N C F F R 1987, p. 25) indicated
that t h r e e - q u a r t e r s of all Securities and E x c h a n g e Commission actions against i n d e p e n d e n t public
a c c o u n t a n t s involved n o n n a t i o n a l firms a n d often involved s u b s t a n d a r d audit work. Specific weaknesses include lack of specialized training a n d p o o r supervision of audit staff.
2 For example, the C o h e n Commission ( A I C P A 1978, pp. 111-112) a n d the Metcalf C o m m i t t e e
(U.S. Senate 1977, p. 45) have discussed the negative effects of price competition on audit quality.
Recently, the Dingell C o m m i t t e e p r o p o s e d the Financial F r a u d Detection a n d Disclosure A c t (U.S.
H o u s e 1992, pp. 1 - 2 ) to regulate the public a c c o u n t i n g profession citing that profession's pricing
practices as c o n t r i b u t i n g to the Savings a n d L o a n crisis.
The Effect of Auditor Changes
57
initial investment by the public accounting firm to obtain future profits in
repeat engagements created by the auditor's learning curve and the high
cost of changing auditors (DeAngelo 1981a, p. 120). Accordingly, such
practice does not (by itself) pose a threat to audit quality.
The effect of low bailing, however, is problematic barring measurement
of audit quality. If public accounting firms low ball to obtain future profits,
audit quality could be threatened by excessive price discounts. That is, low
bailing could be associated with reduced effort and lower audit quality
(AICPA 1978, pp. 111-112).
Research thus far has consistently demonstrated that independent auditor changes are associated with decreased audit fees (DeAngelo 1981a, p.
124 and 1981b, p. 190). Despite lower fees for initial audit engagements,
theory posits that product quality will be maintained (DeAngelo 1981a, p.
124 and 1981b, p. 190; Klein and Leffler 1981, p. 622). Empirical validation
of this theory, however, has been slow in developing due to lack of
adequate information about auditor effort and audit quality. Most independent audit fee studies do not have data on audit hours or engagement
specific measures of audit quality. 3
Palmrose (1986) used audit hours to support the argument that (then)
Big Eight audit firms conducted high quality audits. She (1986, p. 106)
found Big Eight audit firms exhibited both higher audit fees and audit
hours. From this evidence, Palmrose (1986, p. 108) concluded that Big
Eight audit firms commanded higher fees not because of monopoly power
but, rather, because they conducted higher quality audits. In a later study,
Palmrose (1989, p. 496) indicates that audit hours declined with audit firm
tenure while audit fees remained stable. Hence, profit margins may have
improved over time because audit effort decreased at a rate faster than
audit fees. As Palmrose (1989, p. 498) noted, this finding is consistent with
DeAngelo's (1981a) arguments about quasi-rents accruing to incumbent
audit firms. Nonetheless, audit hours may not fully reflect variations in
audit quality. Audit hours, Palmrose (1989, p. 498) cautioned, do not
reflect differences in the mix of audit firm personnel or level of effort.
Recent studies have utilized engagement specific measures of audit
quality based on government oversight agency reviews of public accounting
firm working papers and audit reports. Copley and Doucet (1993b, p. 32)
found the probability of a substandard audit to increase with the length of
auditor tenure. Similarly, O'Keefe et al. (1994, p. 52) found initial engagements associated with significantly fewer violations of generally accepted
3 In a study of p e r c e p t i o n of a u d i t o r s ' i n d e p e n d e n c e , Shockley r e p o r t e d "subjects perceive
competition as a g r e a t e r problem for small [audit] firms than for larger [audit] firms" (1981, p. 792).
W h e n asked, b a n k loan officers r e g a r d e d audit firm t e n u r e as a potential t h r e a t to a u d i t o r
i n d e p e n d e n c e (Shockley 1981, p. 788). Both of these results suggest problems for the g o v e r n m e n t a l
audit m a r k e t in which small public a c c o u n t i n g firms d o m i n a t e a n d lengthy i n d e p e n d e n t audit
firm-client relationships are c o m m o n place.
58
D.R. Deis, Jr. and G. Giroux
accounting standards (GAAS) reporting standards than repeat engagements. Also, Deis and Giroux (1992a, p. 476) and Banker et al. (1992,
p. 505) reported a positive relationship between audit hours and audit
quality.
Giroux et al. (1995) considered the effects of peer review on audit
economic relationships. That study was a public sector extension of a
private sector study that found no fee premia associated with audit firms
participating in a voluntary peer review program (Francis et al., 1990,
p. 375). The lack of fee premia for these firms lead Francis et al. to
conclude that there were economic disincentives to join a peer review
program (1990, p. 376). Hence, if peer review is desirable, a mandatory
program is necessary given their findings.
In contrast, a similarly motivated study in the public sector found fee
premia associated with peer reviewed audit firms and, accordingly, apparent economic incentives to voluntarily participate in a peer review program
(Giroux et al. 1995, pp. 78-79). On a fee-per-hour basis, however, the fee
premia disappeared implying that higher overall fees were associated with
increased audit effort (Giroux et al. 1995, pp. 75-76). It is important to
note that Giroux et al. (1995) did not test the relationship of fee and hours
to audit quality nor did they specifically test the interrelationships between
auditor changes and audit quality on hours and fees. Rather, Giroux et al.
(1995) addressed the contention that if peer review programs serve some
public interest, they must be mandatory due to the lack of economic
incentives to participate on a voluntary basis.
Though not central to the purpose of their study, certain findings
reported by Giroux et al. (1995) are pertinent to the present study. They
(1995, pp. 74-76) found that initial audits were associated with lower audit
fees, higher audit hours, and--naturally--lower fees per hour. Though
they did not incorporate audit quality in their fee, hours, or fee-per-hour
models, their analysis of variance model of audit quality (Giroux et al.
1995, pp. 76-78) indicated that audit quality tends to decline as independent auditor tenure lengthens (similar to Deis and Giroux 1992, pp.
475-476). In particular, multiple comparison procedures revealed that
initial audits were of significantly higher quality than audits conducted by
independent auditors with twenty or more years of tenure but not those
audits in 2 to 4 years or 5 to 19 years of tenure (1995, p. 78). While higher
audit hours among initial audits is encouraging, it does not guarantee that
sufficient audit quality is maintained, Given the inability to distinguish
between initial independent audits and those with less than twenty years of
tenure, it appears that quality is probably adequate. A more thorough
analysis than that provided by Giroux et al. (1995) is required, however,
before one can conclude that audit quality does not decline in initial audit
engagements.
The Effect of Auditor Changes
59
3. Sample
Data from 232 working paper reviews--also called quality control reviews
(QCRs) of public accounting firm audits of Texas ISDs were obtained
through an analysis of the Texas Education Agency (TEA) files. The TEA
regulates Texas school districts. Each QCR culminates in a letter of
findings to the public accounting firm by the agency's audit director. These
letters served as the basis for a comprehensive, direct measure of audit
quality. Additional information used in our study came from the QCR
files, the audit reports, and other reports issued by the TEA. This is the
same data base used by Deis and Giroux (1992a), Deis and Giroux (1994),
and Giroux et al. (1995).
Independent public accounting firms were limited to one observation in
the analysis. In other words, the sample was determined by ISDs having
QCRs. Of 308 QCRs conducted by TEA, 232 were used in this analysis.
QCRs conducted on 1983 and 1984 independent audits were not utilized
for lack of fee and hour data. Audits conducted by Big Eight, non-Big
Eight national, and regional public accounting firms were eliminated (10
public accounting firms in total were eliminated).4
4. The Empirical Models
Audit contracting characteristics affect audit fees and audit hours differently (Palmrose 1989, p. 494), hence two empirical models were used to
evaluate the public sector audit market. The first model was developed
independently, based, in part, on previous audit fee studies such as Rubin
(1988, p. 230). Also, three of the variables SIZE, CLIENTS, and HOURS
--were used in Deis and Giroux 1992a, see pp. 467-468. The variables
YR1, YR2, CLIENTS, and SIZE were used in Giroux et al. (1995, pp.
72-76). The model is used to separately consider each dependent variable.
The independent variables are described below.
In(FEE), In(HOURS)
= a 0 + Blln(SIZE) + B21n(PCI)
+ B3OFFICES + B4CLIENTS + BsCAFR + B6IAC + B7OPIN
+ BsYEAR1 + B9YEAR2 + B101n(QUALSCORE) + e
(1)
4 For additional information regarding the agency's process and the m a n n e r in which the QCRs
were coded, see Deis et al. (1990, pp. 37-38), Deis and Giroux (1992, pp. 468-469), Deis and Giroux
(1994, p. 219), or Giroux et al. (1995, pp. 68-69). (The results did not change when Big Eight and
national CPA firms were included.) Most other audit fee studies have included a d u m m y variable to
distinguish auditor size category. In our current study, the Q C R s for large public accounting firms
were associated with very large ISDs which have different characteristics from small ISDs. Also, the
n u m b e r of offices for the public accounting firm ( O F F I C E S ) is similar to brand n a m e in interpretation.
60
D.R. Deis, Jr. and G. Giroux
where the dependent variables are the natural logs of actual audit fees and
"actual audit hours" (see Deis and Giroux 1992a, p. 468; Giroux et al.
1995, pp. 72-73), and the explanatory variables are (note that the sources
of data are indicated):
Auditee characteristics:
S I Z E = in terms of students attending school in a school district;
see Deis and Giroux (1992a, p. 467) for a description of this
variable ( T E A Files);
PCI = natural log of per capita income in the school district
according to the 1980 U.S. Census
(U.S. Bureau of the Census tape STF-3F(1984));
Brand n a m e and experience:
O F F I C E S = n u m b e r of offices for the public accounting firm
(firm letterhead on auditor's opinion);
C L I E N T S = In of n u m b e r of school districts audited; see Deis and
Giroux (1992a, p. 467) for a description
of this variable ( T E A files);
Audit production costs:
C A F R = 1, if the audit engagement was a comprehensive annual financial
report ( C A F R ) , 0 otherwise (from the reports themselves);
(the audit can be conducted on a C A F R or general
purpose financial statements (GPFS) basis);
I A C = 1, if the auditor reported a material weakness in internal
control, 0 otherwise (from C A F R ) ;
O P I N = 1, if the auditor qualified the independent auditor's report,
0 otherwise (from C A F R ) ;
Initial audits:
Y E A R 1 = 1, if the audit was an initial engagement,
0 otherwise ( T E A files);
Y E A R 2 -- 1, if the audit was a second year engagement,
0 otherwise ( T E A files); and
The Effect of Auditor Changes
61
Product Differentiation:
Q U A L S C O R E = a metric (score) of 19 possible audit quality deficiencies
discovered in the QCR weighted by their importance
to the referral decision. As described below in Section
4.5, this measure differs from Deis and Giroux
(1992a, pp. 468-469), but is the same as the one used
by Giroux et al. (1995, p. 70).
As OFFICES and CLIENTS both represent potential ex ante measures
of audit quality and Q U A L S C O R E measures ex post quality, the models
were estimated with and without Q U A L S C O R E for comparative purposes. 5
To test the effects of initial engagements on audit quality the following
model was estimated:
l n ( Q U A L S C O R E ) = a 0 - B l l n ( S I Z E ) + B 2 ( W E A L T H ) + B3CLIENTS
+ B4OFFICES + B 5P E E R + B6TIME
+ B71n(HOURS) + BsYEAR1 + B9YEAR2
(2)
The quality model is a modified version of Deis and Giroux (1992a, pp.
467-468), substituting YEAR1 and Y E A R 2 for the continuous tenure
variable, adding OFFICES, and omitting three insignificant variables in
that study (see Deis and Giroux 1992a, pp. 467-471 for a description of the
variables). The variables SIZE, CLIENTS, OFFICES, H O U R S , YEAR1,
and YEAR2, are described earlier in the paper. The "assessed property tax
value per student" ( W E A L T H ) "proxies for the financial health" of the
ISD (Deis and Giroux 1992, p. 471). For those independent audit firms
voluntarily participating in a peer review program, the indicator variable
P E E R takes on a value of "1" and "0" otherwise (Deis and Giroux 1992,
pp. 467 and 470; Giroux et al. 1995, pp. 76-77). The timeliness of the
independent audit (TIME) is expressed as the percentage of the 120 day
period following the fiscal year end (i.e., August 31) allowed by Texas state
law to complete the audit (Deis and Giroux 1992, p. 471).
As previously discussed, Giroux et al. (1995, pp. 76-78) analyzed audit
quality using analysis of variance (ANOVA) and multiple comparison tests.
In that analysis audit quality was significantly higher for peer reviewed
ISDs, first year audits relative to those with audit tenure of 20 or more
years, when the audit firm had five or more ISD clients, and when the
audit reports were submitted on a timely basis (Giroux et al. 1995, pp.
77-78).
5This point was made by an anonymous reviewer. Moreover, Ward et al. (1994, pp. 406-407)
found fee premia associated with industry expertise.
62
D.R. Deis, Jr. and G. Giroux
4.1 Auditee Characteristics
Larger public sector organizations generally require more independent
audit time because they enter into more financial transactions (Simunic
1980, pp. 172-173; Rubin 1988, p. 221) or because taxpayers will demand
more information (Wallace 1986, p. 40). Therefore, the size of the government unit being audited should be positively related to both the price of
the audit engagement (FEE) and audit hours (HOURS). The natural logs
of average daily attendance (SIZE), audit fees (FEE), and audit hours
(HOURS) have been used. These transformations provide a better fit to
ordinary least squares (OLS) assumptions and are consistent with past
private sector and public sector research (Baber et al. 1987, p. 298;
Palmrose 1989, pp. 491-492; Roberts and Glezen 1990, pp. 130-132;
Rubin 1988, p. 225).
Natural log of per capital income (PCI) in the ISDs proxies for underlying financial strength (see, for example, Baber et al. 1987, p. 298). Following Baber et al. (1987, p. 298) an inverse relationship is expected for PCI.
4.2 Brand Name and Industry Experience
Although client management can observe the audit production process
(albeit imperfectly), it is costly for users (e.g., citizens, regulators, investors
and creditors) to observe the performance of the independent auditor.
Users therefore rely on the independent auditor's brand name or reputation as an industry specialist as surrogate measures of audit service
credibility (DeAngelo 1981b, pp. 187-192; Klein and Leffler 1981, p. 620;
Simunic and Stein 1986, p. 7).
Although brand name (e.g., Big Eight and Non-big Eight) generally
served as a proxy for auditor quality in prior research (e.g., Rubin 1988, p.
20), a few studies have used the concepts within a single segment of audit
producers. O'Keefe and Westort (1992, pp. 56-57) used the number of
CPAs employed by small public accounting firms to proxy for brand name
in a study of compliance with reporting standards. Shockley and Holt
(1983, p. 556) found that users perceived industry market shares to
differentiate levels of specialization among Big Eight auditors.
In our present study, the number of offices maintained by each public
accounting firm (OFFICES) proxies for brand name and the number of
ISD clients (CLIENTS) proxies for industry experience. Following Palmrose (1989, p. 494) a positive relationship is expected between OFFICES
and audit fees and audit hours. Conversely, a negative association is
expected between industry experience on audit fees and audit hours. ISD
audits are highly regulated, including federal single audit requirements
and additional TEA rules (Deis et al. 1990, pp. 36-37). Firms specializing
The Effect of Auditor Changes
63
in ISD audits are posited to be more efficient in conducting the audit and,
hence, this efficiency will be reflected in fewer audit hours. Giroux et al.
(1995, p. 75) using the same data found a positive significant relation
between audit hours and the number of clients. However, they (1995, p. 75)
did not find a significant relationship between industry experience (clients)
and audit fees or audit fees-per hour. Assuming a competitive market,
specialists are also expected to share some of the benefits of their lower
production costs in the form of lower audit fees with their clients. 6
4.3 Audit Production Costs
Several variables were included in the fee and hours model (model 1) to
capture the complexity of the audit and factors associated with audit
production costs as identified by prior research. According to generally
accepted accounting principles (AICPA 1994, pp. 143-145), Texas ISDs
are not required to issue a CAFR to meet TEA report requirements. Thus,
when the audit engagement is on a CAFR, the audit contract is expected
to reflect the auditor's additional reporting responsibilities concomitantly
with a higher audit fee and more audit hours (Rubin 1988, pp. 222-223).
Similar to Roberts and Glezen (1990, p. 131) and Raman and Wilson
(1992, p. 288), two report variables were evaluated: 1) qualified audit
opinions (OPIN), 7 and 2) the report on internal accounting controls (IAC).
Traditionally, both private and public sector audit fee studies include an
indicator variable for modified (e.g., qualified) audit opinions as a measure
of the auditor's risk of loss arising from various auditing problems (Simunic
1980, pp. 173-174). By virtue of the expanded reporting requirements
under The Single Audit Act of 1984, independent auditors of public sector
entities must disclose material weaknesses in the client's internal control
environment in a report on internal controls. Roberts and Glezen claim
that this report reveals the "quality of the client's internal financial
6 0 ' K e e f e et al. (1994, p. 65) r e p o r t e d a significant negative correlation between audit fees a n d
the audit firm's industry specific knowledge (i.e., the n u m b e r of school district audits) in their study
of violations o f generally a c c e p t e d a c c o u n t i n g principles r e p o r t i n g s t a n d a r d s d e t e c t e d in 935
California 1986 school district audits. W a r d et al. (1994, p. 407), in contrast, f o u n d higher audit fees
associated with a single regional audit firm responsible for 21% of the s a m p l e d Michigan municipalities. As o u r sample was c o m p r i s e d o f o n e audit p e r firm, we w e r e u n a b l e to test for a similar effect.
Moreover, with nearly 1,200 Texas ISDs it is unlikely for a n y firm to attain a 2 0 % s h a r e of the
market.
7 O u r study focuses on r e p o r t e d b r e a c h e s in the c u r r e n t y e a r which is consistent with earlier
studies (e.g., B a b e r et al. 1987, p. 303; R o b e r t s a n d Glezen, 1990, p. 140; a n d R u b i n 1988, p. 223).
A n alternative is to use the p r i o r y e a r r e p o r t s on the a s s u m p t i o n t h a t r e p o r t e d b r e a c h e s f r o m the
previous y e a r w o u l d have i n c r e a s e d audit risks, driving u p b o t h fees a n d hours. Predictions are the
s a m e in either case, as a r e results.
64
D.R. Deis, Jr. and G. Giroux
reporting system" (1990, p. 135). Since both OPIN and IAC capture
potentially time consuming aspects of the audit engagement, a positive
relationship is predicted for both report-related variables on audit fees and
audit hours.
4.4 Initial Audits
Several recent studies detected lower audit fees in the initial and second
engagement years (Baber et al. 1987, p. 303; Ettredge and Greenberg 1990,
p. 208; Roberts et al. 1990, p. 227; Simon and Francis 1988, p. 261; Turpen
1990, p. 68). Giroux et al. (1995, p. 74) found that Iow-baUing existed
during the first year of an audit, but not the second year. The same
categorical variable used by Giroux et al. (1995, p. 73) were included here
to evaluate initial year (YEAR1) and second year (YEAR2) competitive
effects on audit fees and audit hours.
4.5 Product Differentiation
Audit quality is the key feature of the product differentiation explanation
for observed variations in audit pricing. If independent auditors can
convince clients that their product differs, then audit fees can vary (YardIcy et al. 1992, p. 165-171). Public accounting firms are assumed to plan a
level of audit quality they intend to supply and this information is incorporated in audit fees and audit hours. (Simunic and Stein 1987, pp. 1-4).
QCR results were used in our study to measure audit quality ex post,
anticipating that this reflects the independent auditor's ex ante plan
regarding the level of audit quality to be supplied. As Copley et al. (1994,
p. 248) asserted, a positive association exists between the planned level of
audit assurance and the propensity to be in compliance with professional
standards. Public accounting firms planning on lower quality levels will (on
average) exhibit lower frequencies of compliance with professional standards than firms planning on higher quality levels. Hence, ex post measures of compliance with professional standards are sufficiently correlated
with audit quality to reasonably proxy for ex ante planned quality.
We performed a content analysis of 232 letters of findings in the QCR
files and found 19 categories of findings common to many audits (Deis and
Giroux 1992a, p. 468; Deis and Giroux 1994, p. 219; Giroux et al. 1995,
p. 67). TEA's "director of audits ranked the importance of each category,
from 1 (most important) to 19 (least important)," in considering whether or
not the public accounting firm should be referred to the Texas State Board
of Public Accountancy (Deis and Giroux 1992a, p. 468; also see Deis and
Giroux 1994, pp. 219-220 and Giroux et al. 1995, pp. 67-68).
A metric for audit quality was built by multiplying each category by its
(negative) rank. The least severe category (substantive tests: minor) was
The Effect of Auditor Changes
65
weighted - 1; the most severe (audit program: major) by - 19. The weights
were summed, resulting in the metric Q U A L S C O R E . 8 Both audit fees
(FEE) and audit hours ( H O U R S ) are projected to be directly related to
Q U A L S C O R E . This measure is similar to Deis and Giroux (1992a, pp.
468-469). They divided the deficiency code by its assigned rank. Under
their measure, a zero was the highest possible value and a higher quality
audit had a lower quality score. Q U A L S C O R E is the same measure as
" S C O R E " , the quality measure used by Giroux et al. (1995, p. 70-71) in
their A N O V A model.
The analysis of audit quality ( Q U A L S C O R E ) on audit fees and audit
hours raises certain econometric issues. It is evident from previous studies
that audit hours and audit fees affect audit quality (Deis and Giroux
1992a, p. 476; O'I(eefe and Westort 1992, p. 65; O'Keefe et al. 1994,
p. 261; Copley et al. 1994, pp. 253-254). Our current study proposes to
reverse the order and investigate the effects of audit quality on audit fees
and audit hours. Therefore, QUAI_SCORE may be an endogenous variable which is not independent of the error term in the proposed empirical
m o d e l s - - t h u s violating OLS assumptions and suggesting simultaneous
equation estimation (SEE) would be appropriate (Maddala 1988, p. 293).
However, use of such an approach is inefficient when the error terms are
independent. Fortunately, Hausman (1978) provided a specification test to
determine whether OLS or SEE is most appropriate. Hausman's specification error test is asymptotically equivalent to a test of the covariance
between the error terms produced from OLS and SEE. The test statistic is
(Maddala 1988, pp. 435-439):
M = qZ/var(q)
(3)
where
q = B 1 - Bo
and
var(q) = var(B 1) - var(B 0)
using B 1 indicates estimators from SEE and B 2 estimators are from OLS.
The test statistics, m, has X z distribution with k degrees of freedom (k
number of parameters). The calculated m statistics for our study were
0.825 for the audit hours model and 0.693 for the audit fee model. Neither
of these m statistics were significant ( p > 0.10, two tailed) ( X0.90
2 d.f.
8An alternative m e a s u r e of quality is a categorical variable, w h e r e 1 = a report referred to the
Texas State Board of Public A c c o u n t a n c y - - 3 2 or about 14% were referred (see Deis et al. 1990, pp.
3 7 - 3 8 for a description of the process; also, see Deis and Giroux 1992a, p. 472). When this d u m m y
variable replaced Q U A L I T Y in the regression runs, results indicated that referred audits have both
lower fees (significant at .01, one-tail test) and lower hours (significant at .0001). O t h e r quality
metrics produced similar results in regression runs, suggesting the results are robust regardless of
the m a n n e r in which audit deficiencies are indexed.
66
D.R. Deis, Jr. and G. Giroux
15 = 22.31). Hence, Q U A L S C O R E and the error terms in the models are
not correlated and OLS estimations are justified (Maddala 1988, 437). 9
5. Empirical Results
Descriptive statistics appear in Table 1 and a correlation matrix for all
variables appears in Table 2. Thirty-two independent auditors in our
sample were referred to the Texas State Board of Public Accountancy for
particularly low quality audits (see Deis and Giroux 1992a, p. 472). These
audits were considered so deficient relative to professional standards, that
actions by the Texas State Board of Public Accountancy were deemed
necessary. Board action can include imposing additional professional education, other remedial actions, or punitive actions. The Board can revoke a
CPA's license in an extreme case (Deis et al. 1990, pp. 37-38). Most of the
correlations among the independent variables were low, with the largest
correlation between SIZE and CLIENTS at .32 (Table 2).
Tables 3 and 4 present the results of the audit fee and audit hours
models. Several diagnostic procedures were used to assure that the results
were not unduly influenced by outliers, multicollinearity, or heteroskedasticity. 1° The full model estimating In (FEES) was significant at .0001
( F = 41.26, two tailed test) and has comparable explanatory power (adjusted R 2 = 0.64) to Roberts and Glezen's (1990, p. 142) model of 356
Texas ISDs (adjusted R 2 = 0.66) and Giroux et al. (1995, p. 75) using the
same data (R 2 = .599). The full model estimating audit hours was also
significant at .001 ( F = 22.63, two tailed test) and explains about half the
variation in audit hours (R 2 = 0.47), compared to Giroux et al. (1995, p.
75) where R 2 = .427. As results for the reduced models are similar to
those of the full models, the discussion to follow will address the results in
the full models.
Most of the variables were significant in the expected direction to both
audit fees and audit hours. As found in previous studies (e.g., Rubin 1988,
p. 230; Giroux et al. (1995, p. 75)), large clients (SIZE) had higher audit
fees and more audit hours. ISD financial strength (PCI) was negatively and
significantly associated with both audit hours and fees, confirming Baber et
al. (1987, p. 303). The larger local CPA firms (OFFICES) received fee
9 We are grateful to R. Carter Hill for help in conducting the H a u s m a n test. The S H A Z A M
statistical package was used to conduct this test.
10 No important violations were detected. Plots of the residuals against the predicted value in
each model depict a cloud-like pattern. Univariate statistics on the residual terms from each model
indicated a m e a n of zero ( p = 1.0) and the assumption of normality, using the Kolmogorov-Smirnov
test, could not be rejected at the 0.15 level. Variance inflation factors (VIF) were low, ranging from
1.03 to 1.17 for each model. Condition indexes were 1.64 or less in both models, considerably less
than the lowest threshold of 15 suggested by Belsley et al. (1980, p. 157) over which multicollinearity
becomes a problem. Based on these diagnostic procedures, the results of both regression models
can be interpreted without concern that they are influenced by violations in OLS assumptions.
The Effect of Auditor Changes
67
Table 1. Descriptive Statistics for a Sample of 232 Quality Control Reviews of
Independent Audits of Texas Independent School Districts Conducted
from 1984-1989
Continuous Variables:
Audit fee (FEE)
Audit hours ( H O U R S )
Average daily attendance (SIZE)
Per capita income (PCI)
No. of firms offices (OFFICES)
No. of ISD clients (CLIENTS)
Q C R weighted metric ( Q U A L S C O R E )
Mean
Minimum
Maximum
$10,416
294
3,616
$6,229
1.3
3.3
- 36.9
800
38
36
2,642
1
1
- 157
50,000
1,994
44,776
12,827
8
30
0
Categorical Variables:
Frequency
% (Rounded)
No. of comprehensive annual financial
reports (CAFR)
No. of internal control breaches (IAC)
No. of auditor reported breaches (OPIN)
No. of initial audit engagements
(YEAR1)
No. of 2nd year engagements (YEAR2)
25
11
86
11
22
37
5
9
20
9
premiums indicative of a brand name within that segment of the audit
market. Audit hours were significant and positively related with OFFICES.
Audit hours were negative and significantly associated with the number
of ISD clients audited by the audit firm, the same as found by Giroux et al.
(1995, p. 75). CLIENTS also was negative and significantly associated with
Table 2. Spearman Rank Order Correlations
FEE
HOUR
SIZE
PCI
OFFICES
CLIE
CAFR
IAC
OPIN
YR1
YR2
QUAL
HOURS
.76*
.75*
.62*
-.00
-.01
.15"**
.12
.24**
.10
.29**
.24**
-.01
.01
.07
.09
-.09
.05
-.06
-.03
.27**
.33*
SIZE
.14"**
.11
.32*
.28**
-.13
-.00
-.04
-.06
.15"**
* Significant at .0001.
** Significant at .01.
*** Significant at .10.
(All significance tests are two-tailed.)
n = 232
PCI O F F I C E S C L I E N T C A F R IAC O P I N Y R 1 YR2
-.08
.02
.08
-.24**
-.10
-.03
-.06
-.07
.24**
--.06
-.02
.08
.08
--.02
.09
--.05
--.04
-.02
-.07
.15"**
.17"**
-.15
- . 0 1 .21"*
- . 1 1 .09
-.00
.00 - . 1 0
- . 0 6 .02
.00.10
.03 .12 .02
68
D.R. Deis, Jr. and G. Giroux
T a b l e 3. E m p i r i c a l
Model Results Ordinary
L o g o f A u d i t F e e is t h e D e p e n d e n t
Least Squares Regression: Natural
Variable (Model
1)
n = 232
Description
Predicted
Sign
Full M o d e l
Parameter
Estimate
(t-Value)
Reduced Model:
Parameter
Estimate
(t-Value)
9.590
9.310
0.386
(17.36)*
- 0.360
( - 2.97) **
0.389
(17.18)*
- 0.380
( - 3.08)* *
0.053
(1.83)***
-0.015
( - 1.57)***
0.072
(2.48)**
-0.010
( - 1.25)
Intercept
Auditee Characteristics:
A v e r a g e daily a t t e n d a n c e
ln(SIZE)
P e r capital i n c o m e I n ( P C I )
+
-
Brand Name & Specialization:
No. of firm O F F I C E S
+
No. of I S D C L I E N T S
-
Audit Production Costs:
CAFR engagement
+
Internal accounting
controls ( I A C )
Qualified opinion (OPIN)
+
0.228
(2.31)***
0.109
(1.73)***
0.201
(1.45)***
+
0.239
(2.38)**
0.124
(1.94)***
0.172
(1.22)
Inhial Audits:
Initial e n g a g e m e n t
( Y E A R 1)
Second year engagement
( Y E A R 2)
-
((-
0.150
1.49)***
0.023
0.22)
- 0.120
(1.18)
- 0.013
( - 0.12)
Product Differentiation:
In ( Q U A L S C O R E )
Adj. R 2
F statistic (two-tailed test)
+
0.117
(3.08)**
0.635
41.26"
0.622
43.15"
* Significant at .0001.
* * Significant at .01.
*** Significant at .10.
(All significance tests are one-tailed--except when noted to be two-tailed.)
audit fees in the full model but not in the reduced model with quality
omitted. The reduced model result is similar to the fee model result
reported by Giroux et al. (1995, p. 75) which suggests that quality is an
important omitted variable in that study in regards to industry experience
(CLIENTS). As expected, audit efficiency was a benefit produced from
investments in industry expertise, with economies passed on to clients as
The Effect of Auditor Changes
69
Table 4. Empirical Model Results Ordinary Least Squares Regressiop: Natural
Log of Audit Hours is the Dependent Variable (Model 1)
n = 232
Description
Predicted
Sign
Intercept
Full Model
Parameter
Estimate
(t-Value)
Reduced Model:
Parameter
Estimate
(t-Value)
6.457
6.023
Audited Characteristics:
Average daily attendance
In(SIZE)
Per capita income ln(PC1)
+
-
0.308
(11.74)*
- 0.322
( - 2.26)***
0,312
(11,54)*
- 0,35
( - 2,39)**
0.057
(1.67)***
-0.021
(-2.26)***
0,086
(2.48)**
-0,017
(-1.82)***
0.273
(2.35)**
0.131
(1.76)***
0.275
(1.69)***
0.290
(2.42)**
0.154
(2.01)***
0,231
(1.37)***
0.159
(1.34)***
0.028
(0.23)
0.204
(1.68)***
0.044
(0.35)
0.179
(4.00)*
0.472
21.63"
0.436
20.84*
Brand Name & Specialization:
No. of firm O F F I C E S
+
No. of ISD CLIENTS
-
Audit Production Costs:
CAFR engagement
+
Internal accounting
controls (IAC)
Qualified opinion (OPIN)
+
+
Initial Audits:
Initial engagement
(YEAR 1)
Second year engagement
( Y E A R 2)
+
+
Product Differentiation:
ln(QUALSCORE)
Adj. R 2
F statistic (two-tailed test)
+
* Significant at.0001.
** Significant at .01.
*** Significant at .10.
(Allsignificance tests are one-tailed--except when noted to be two-tailed.)
lower costs, t~ Moreover, both Deis and Giroux (1992a, p. 476), O'Keefe et
al. (1994, p. 261), and Giroux et al. (1995, p. 77) found higher audit quality
associated with increases in industry-specific knowledge (i.e., the number
11 Both OFFICES and CLIENTS are potential ex ante proxies for audit quality demanded. Both
are significantly correlated with QUALSCORE. Additional analysis (using Duncan's Multiple
Range Test) indicated that auditors with five or more ISD clients had significantly higher quality
than those with fewer clients (Huck et al. 1974, pp. 68-69). Offices was not significant using
Duncan's Multiple Range Test.
70
D.R. Deis, Jr. and G. Giroux
of ISD clients). Hence, industry-specific knowledge is an important attribute for consideration in audit procurement as it is accompanied by
lower audit costs and higher audit quality in this segment.
Similar to related studies (e.g., Rubin 1988, p. 224), variables identified
as components of audit production costs were significant to both audit
hours and audit fees. CAFR engagements entail more audit effort which is
reflected in both higher fees (Rubin 1988, p. 230) and additional audit
hours. Similar to prior public sector studies (O'Keefe et al. 1994, p. 44;
Raman and Wilson 1992, p. 288; Roberts and Glezen 1990, p. 135), higher
audit fees were associated with reports of weaknesses in internal accounting controls (IAC) and qualified audit opinions (OPIN). Similar to private
sector studies (Davis et al. 1993, p. 143; Palmrose 1986, p. 106 and 1989,
p. 494), public sector report modifications (OPIN and IAC) were also
associated with higher audit hours.
The link between report modifications and audit hours is clear--more
time is expended when difficulties are encountered during the engagement. The interpretation between report modifications and audit fees,
however, is somewhat problematic. It is unlikely that independent auditors
can freely adjust the price of public sector audits to accommodate unanticipated audit hours. Nonetheless, this result (i.e., higher audit fees associated with current year report modifications) has been found in public
sector studies (Rubin 1988, p. 230; Roberts and Glezen 1990, p. 140). As it
is our understanding that there are a large number of fixed fee engagements in the public sector, these findings imply that the independent
auditor develops an ex ante expectation that reporting problems may exist
and plans the audit accordingly.
Significantly lower audit fees were detected in 22 initial engagements
(YEAR1) in the sample. Second year engagement (YEAR2) audit fees
were not significantly different in the model (model 1--see Table 3).
Initial year engagements were positive and significantly associated with
audit hours. The same results were found in Giroux et al. (1995, p. 75).
This evidence serves to further confirm the practice of low bailing by
auditors to obtain new clients. These results generally support DeAngelo
(1981a, p. 192), with lower fees in the initial year only.
Comparing the full model result for initial engagements on audit fees to
those in the reduced model (Table 3) reveals that initial engagements are
significant only in the full model. A similar comparison of full and reduced
audit hour models (Table 4) indicates that initial engagements are significant in both models. These reduced model results for initial engagements
coincide with those reported by Palmrose (1989) who did not have a direct
audit quality measure. She (1989, pp. 496 and 498) reported insignificant
results for initial engagements in her fee model but significantly less audit
hours as tenure lengthens. After considering audit quality variations (i.e.,
in the full models), we see that initial audit fees are in fact lower while
The Effect of Auditor Changes
71
audit hours are higher. This is additional evidence to strengthen Palmrose's
suggestion that profit margins improve over the auditor's tenure as audit
hours decrease (1989, p. 498). Our study also suggests that without controlling for audit quality cross-sectional audit fee studies may not detect price
cutting of initial engagements.
Q U A L S C O R E was a significant determinant of both audit fees ( p =
0.001) and audit hours ( p = 0.0001)--see Tables 3 and 4 (one-tailed tests).
Audits with high quality had higher fees and more audit effort (hours). A t
test of Q U A L S C O R E when the sample was split after the second year of
auditor tenure confirms this relationship. Using a t test in a separate
analysis we found that audits conducted in the first two years of tenure had
significantly higher audit quality than audits conducted by auditors with
three or more years of tenure ( p = 0.001, two tailed test). First and second
year audits had a similar SIZE variable (t statistic = -0.11); hence, the
associations between YEAR1 and Y E A R 2 on F E E and H O U R and
Q U A L S C O R E (in the t test) were not driven by audited size. 12' 13 Moreover, as shown in Table 5, the multivariant test for the effects of initial
audit engagements on audit quality also demonstrates higher quality on
initial audit engagements (t = 1.37, p = .09, one-tailed test).
6. Conclusions
For nearly twenty years widespread concerns have been expressed that
aggressive price competition within the market for audit services erodes
the quality of work provided by auditors. Based partially on these concerns, federal and state agency oversight of the audit profession has
increased dramatically and continues to expand. A central tenet behind
this heightened regulation is the purported detrimental relationship between 'low-ball' audit pricing and audit quality. Under this scenario, when
the initial year audit is conducted at a substantial price discount (i.e., a
'low-ball' fee) audit quality suffers because audit firms (1) expend less
effort so as to maintain profit margins in light of the fee reductions or (2)
they fail to report problems in order to protect "receivables" in future
audit engagements. The 'low-bailing' part of this scenario has been well
documented in academic studies and by the business media. Heretofore
research on the critical link between 'low-bailing' and audit quality,
12As an alternative test, n u m b e r of years of audit tenure ( T E N U R E ) replaced Y E A R 1 and
Y E A R 2 and a t e n u r e / q u a l i t y interaction term added in additional regression runs. Regression
results indicate that T E N U R E is significant and positive for fees, but insignificant for hours. T h e
interaction t e r m was significant for hours only. T h e s e results reinforce the importance of the
relationship between audit tenure and both fees and hours.
13 To d e t e r m i n e if the r e f e r r e d audits were biasing the results, the regression models were rerun
excluding the 32 r e f e r r e d audits. No differences were detected. Thus, the subset of " a c c e p t a b l e "
audits can be interpreted as product differentiated.
72
D.R. Deis, Jr. and G. Giroux
Table 5. Empirical Model Results Ordinary Least Squares Regression: Natural
Log
of A u d i t Q u a l i t y is the D e p e n d e n t V a r i a b l e ( M o d e l
2)
n = 232
Description
Predicted
Sign
Intercept
Parameter
(t-Value)
-- 3 . 2 9 8
Ln(SIZE)
-
WEALTH
-
- 0.124
( - 2.67)**
- 0.091
( - 1.32)* **
CLIENTS
+
0.018
(1.32)***
OFFICES
+
0.121
(2.55)**
PEER
+
0.353
(2.96)**
TIME
-
- 0.588
( - 2.66)**
HOURS
+
0.368
(4.18)*
YEAR1
+
0.227
(1.37)***
YEAR2
+
0.072
(0.42)
Adj. R 2
F
statistic (two-tailed test)
0.175
6.43*
* Significant at .0001.
** Significant at .01.
*** Significant at .10.
(All significance tests are one-tailed--except when noted to be two-tailed.)
however, has been inconclusive since audit quality was unobservable.
Ironically, increased oversight of the audit profession has, for the first
time, provided researchers with direct measures of audit quality and,
therefore, an opportunity to address the twenty year old question: D o
auditors lower audit quality on initial year 'low-ball' engagements?
DeAngelo's (1981a and 1981b) seminal theoretical work on audit quality
has been the basis for a myriad of audit market studies; yet, her primary
conclusions have been difficult to test empirically for the data limitation
reason just cited. Using state regulator QCR data, Deis and Giroux (1992)
developed a direct, continuous measurement of actual audit quality to
address a series of questions stemming from DeAngelo's propositions. In
our initial study (Deis and Giroux 1992, p. 476) we confirmed two elements
of audit quality envisioned by DeAngelo (1981a and 1981b). First, we
detected improved audit quality associated with two measures of the audit
firm's investment in reputation (industry expertise and voluntary peer
The Effect of Auditor Changes
73
review section). This was termed by DeAngelo as the "collateral" effect
(1981b, pp. 190-191). Second, we found audit quality to decline with the
length of auditor tenure (defined as the number of years of audit tenure).
While our 1992 paper did not explicitly test the 'low-bailing' issue, this
second finding is consistent with both DeAngelo's explanation of opportunistic behavior among incumbent auditors (1981b, p. 189) and the
"complacency" effect described by Mautz and Sharaf (1961, p. 208).
The positive association between audit quality and voluntary membership in AICPA's peer review program detected in our 1992 paper was
expounded upon in our 1995 paper (Giroux et al. 1995, p. 67). As an
observable quality surrogate, membership in peer review programs should,
under DeAngelo's theory, signal an audit firm's investment in reputation
and, hence, be assoc.iated with audit fee premia. That is precisely what we
found (1995, p. 75). This finding is of interest for two reasons: (1) the
existence of market incentives to voluntarily join a peer review program
calls into question mandatory peer review participation required by the
United States General Accounting Office's 1988 revision to Governmental
Auditing Standards and (2) the results conflict with a similar study of
private sector audits (Francis et al. 1990, p. 374). Audit firm tenure was
also explored further in this study. Evidence of low-bailing was found in
the audit fee regression model (1995, p. 75) and a categorical tenure
variable was negatively related to audit quality in an analysis of variance
(ANOVA) model (1995, p. 77). We did not directly test the 'low-ball' to
audit quality issue in this paper, however, since peer review membership
was the focal point of both the regression and ANOVA models. Hence,
DeAngelo's propositions on 'low-bailing' were not conclusively evaluated
in the 1992 or the 1995 papers.
Building on our earlier two papers, the primary contribution of this
study is that the effects of 'low-bailing' audit fees is tested in light of a
direct audit quality measure. We found lower fees among first year audits
without concomitant audit quality deterioration--as DeAngelo predicted.
Accordingly, as DeAngelo concludes (1981a, p. 126), regulation aimed at
curbing the 'low-bailing' practice should be reconsidered given the lack of
evidence that a price-for-quality tradeoff exists. In addition, since the
results for other variables in the study's empirical models were consistent
with related public sector literature (Baber et al. 1987, p. 303; Roberts and
Glezen, 1990, p. 140; Rubin 1988, p. 230; Ward et al. 1994, pp. 406-407),
previous studies lacking a direct audit quality measure need not be
reestimated with one. Moreover, given the similarities between private
sector and public sector audit economic studies (Rubin 1988, pp. 234-235)
it is plausible that our findings and conclusions are applicable to the
private sector audit market. Hence, conclusions in previous studies regarding: auditor selection, non-aud