please read the article and answer the question shortly

Alfaisal University
College of Business
Internal Controls, Audit & Fraud Prevention & Detection
(MBA556)
Article Questions
Article Discussions will take place: Weeks 4,5,8,9,15
All of the article readings and questions aim to assist you to develop your critical
thinking and writing skills. This will help you with the course assessments as well as
support you to develop your graduate skills further.
These readings will also help you achieve the intended outcomes. Much of the
feedback from your work will be self-generated i.e. by comparing your work with the
work of your peers, you will be able to see where your own work could be
strengthened. Also, in class discussions, you will be able to test your own opinions
against those of your colleagues.
Class discussions are not intended to be a part of the lecture. They are designed to
be facilitated by the instructor and highly interactive. There are no solutions. If you
do not prepare for the article questions, you will not get value out of the class
discussion. Also, you will reduce the value of the discussion to students who have
prepared because they will not receive your input.
Week
Week 4
Week 5
Unit
Unit 2: Ethics & The Audit Profession
Unit 3: Audit Evidence, Materiality & Audit Risk
Week 8
Unit 4: Substantive Testing
Week 9
Unit 5: Internal Controls & Fraud
Week 15
Unit 6: Fraud Detection and Mitigation
Unit 2
Question: How does the current regulatory framework for audit affect audit quality?
Before the lecture:
Read these recommended journals (see reading list):
 Bazerman, MH, Loewenstein, G and Moore, DA (2002), Why Good
Accountants Do Bad Audits, Harvard Business Review
 Bazerman, MH and Moore, DA (2011), Is it time for auditor independence
yet? Accounting Organisations and Society, 36:310 -312
and answer the following questions:
a) What are the authors’ key messages? (Put into your own words ie do not
copy and paste the authors’ words)
b) Do you agree or disagree and why?
c) What are the advantages and disadvantages of long association with an
audit?
d) Do you think that the current rules about partner and firm rotation are fit
for purpose?
e) Prepare at least a single A4 sheet of notes to inform the discussion in
class
During the lecture:
f) In small breakout groups, use your summary to discuss the following
questions:
i.What are the factors that cause auditors to behave unethically?
ii.How can unethical behaviour be reduced or eliminated?
iii. Do recent audit reforms help?
g) Appoint a spokesperson to provide feedback to the whole class.
Unit 3
Question: What is audit risk and how is it related to business risk?
Before the lecture:
a) Read these recommended papers (see reading list for links): De Martinis, K. and
Houghton, K. (2019); ‘The Business Risk Audit Approach and Audit Production
Efficiency’; Abacus; V.55, No.4, pp734-782 and ISA 315.
b) Write a short response to the question using what you have read.
During the lecture:
c) In small breakout groups discuss your responses and agree a group response to
the question. Agree who will give feedback to the whole class.
Unit 4
Part (a) Question: What factors influence professional scepticism?
Before the lecture:
a) Read Hurtt, K.R., Brown-Liburd, H., Earley, C., E. and Krishnamoorthy, G. (2013)
‘Research on Auditor Professional Skepticism: Literature Synthesis and
Opportunities for Future Research’, Auditing: A Journal of Practice & Theory
American Accounting Association Vol. 32, Supplement 1, p45-97.
b) The researchers have categorised the antecedents to sceptical judgement and
sceptical action. In your own words, summarise the antecedents.
c) Do you agree with the researcher’s categorisations and why do you agree or
disagree?
d) Rank the antecedents in order of importance.
During the lecture:
In small breakout groups, discuss the paper and the work you have prepared. Come
up with a group ranking of the antecedents. One person from each group will be
asked to present the group’s rankings and to explain why they have been ranked in
that order.
Part (b) Question: Have greater transparency and the disclosure of key judgements
in audit reports reduced or increased the risk exposure of auditors?
Before the lecture:
a) Read this recommended paper (see reading list for links): Brasel, K., Doxey,
M.M., Grenier, J.H., Reffett, A. (2016); ‘Risk Disclosure Preceding Negative
Outcomes: The Effects of Reporting Critical Audit Matters on Judgments of
Auditor Liability’; Current Issues in Auditing; V.10, No. 2, pp1-10
b) Summarise the paper in your own words.
During the lecture:
c) In small breakout groups discuss the paper and the reasons why auditors
expected their risk exposure to increase. Come to a group decision on whether
the paper says that CAMs increase or decrease the risk exposure of auditors.
Agree who will give feedback to the whole class.
Unit 6
ABACUS, Vol. 55, No. 4, 2019
doi: 10.1111/abac.12178
MICHAEL DE MARTINIS AND KEITH HOUGHTON
The Business Risk Audit Approach and
Audit Production Efﬁciency
Essentially, this study asks: Does the business risk audit (BRA) approach
increase audit production efﬁciency? To answer this question empirically,
direct and indirect tests are employed using proprietary, working paper
data from the larger clients of a major Australian public sector audit
provider and an efﬁciency frontier analytic methodology, data
envelopment analysis (DEA). Results based on this proprietary, audit
hours data for audit engagements carried out just after BRA approach
implementation show that they have high levels of production efﬁciency
and are risk-adjusted, with no signiﬁcant difference in production
efﬁciency between higher and lower business risk audit engagements.
Results based on audit fees data for audit engagements carried out shortly
before and after BRA approach implementation show that overall
production efﬁciency signiﬁcantly improves. Importantly, while this
improvement is signiﬁcant for lower-risk audit engagements, there is no
signiﬁcant improvement for higher-risk audit engagements. In the context
of this study’s research site, this is consistent with the BRA approach
addressing inefﬁciencies created when lower-risk audit engagements are
being over-audited. That is, the BRA approach can result in both riskadjusted and more efﬁciently produced audits. With the re-emergence of
the BRA approach in the literature and in practice, this study provides
empirical evidence to support the claim that this audit approach can lead
to ‘creating auditing efﬁciencies’ (Bell et al., 1997, p. 1).
Key words: Audit production efﬁciency; Auditee business risk (ABR);
Business risk audit (BRA) approach; Data envelopment analysis (DEA);
Public sector auditing.
With the re-emergence of the business risk audit (BRA) approach in practice, this
study examines its effect on audit production efﬁciency. Proponents of the BRA
approach claim that this methodology can lead to ‘creating auditing efﬁciencies’
(Bell et al., 1997, p. 1). An empirical examination of the validity of this claim
MICHAEL DE MARTINIS (michaeldemartinis1@gmail.com) is at the Melbourne Institute of Technology.
KEITH HOUGHTON is at the Australian National University and the University of New England. The
authors are grateful to the Editor-in-Chief, Professor Stewart Jones, the Associate Editor, Professor
David Lont, and the two anonymous reviewers for their insightful inputs into this paper. This paper has
also beneﬁted from comments received from a number of conference and research workshop
participants, including those of Dr Amir Moradi and Professor Christine Jubb. Importantly, without
access to the proprietary data, this research would not have been possible. The authors gratefully
acknowledge the assistance of the audit ﬁrm referred to in this research. While neither it nor its clients
can be identiﬁed, as agreed in the conﬁdentiality arrangements, its support is noteworthy. Additionally,
the authors gratefully acknowledge the time and effort given by a number of its executives and staff.
734
© 2019 Accounting Foundation, The University of Sydney
requires rare access and analysis of detailed and often highly conﬁdential data.
With such access and analysis of proprietary data, this study contributes to our
knowledge and understanding of audit practice through an empirical assessment of
a major utilization of the BRA approach. This contribution is particularly useful in
light of the re-emergence of the BRA approach in an era of audit methodological
developments responding to the rise of big data and data analytics. Others claim
that it is a methodology not well understood by regulators and standard setters
(Curtis et al., 2016). This may explain why this approach was discontinued
prematurely from practical application by certain audit ﬁrms, who had heavily
invested in it. It also provides further, strong motivation to examine empirically
the practical impact of the BRA approach.
The BRA approach refers to an audit methodology and practice adopted mainly
by large private sector accounting ﬁrms initially from the mid- to late 1980s. In this
approach, the auditor uses what is referred to as a ‘top-down approach’ (PCAOB,
2007, AS 5, paragraph 21), particularly when assessing auditee business risk (ABR).
ABR is deﬁned broadly by auditing pronouncements as the risk that an entity’s
business objectives will not be achieved (AS 12, ASA 315, ISA 315) (PCAOB, 2010;
AUASB, 2015a; IAASB 2015a). This assessment involves obtaining knowledge and
understanding of the auditee and the relevant environment, including industry,
regulatory and other external factors, business objectives, strategies, activities,
interactions, processes, and related risks, as well as entity-level controls.1
The literature advocates that the BRA approach can have a positive impact on
two aspects of audit quality: effectiveness, that is, forming and reporting
appropriate audit opinions (De Angelo, 1981), and performance, that is, efﬁciently
forming and reporting such audit opinions (e.g., Bell et al., 1997; SEC, 2007; Schultz
et al., 2010). The presumption is that, compared to traditional audit approaches,
which focus auditor effort on reductionist or component assessments of the risk of
material misstatements, the BRA approach, which focuses auditor attention on
assessments of complex, holistic, and interdependent ABRs, can improve the
likelihood of an effective and more efﬁciently produced audit. By enhancing the
auditor’s awareness of entity-level risks, such as ABR, this can facilitate greater
attentiveness to the risk of material misstatements (i.e., inherent risks and control
risks) in the auditee’s ﬁnancial statements and related disclosures (AS 5; Bell et al.,
1997). Such heightened awareness also can improve the auditor’s assessment of its
own business risk from being associated with that particular auditee (Stanley, 2011).
Bell et al. (1997, p. 1) make the case directly that ‘such approaches can serve the
auditor’s primary assurance goal by providing a greater power to detect material
misstatements while concurrently creating auditing efﬁciencies’. Put simply,
compared to traditional audit approaches, the BRA approach can result in
1
Alternative early forms of the BRA approach, including strategic systems auditing (Peecher et al.,
2007), include: KPMG’s Business Measurement Process (BMP) (e.g., Bell et al., 1997; Elliot et al.,
1999; Eilifsen et al., 2001; Bell et al., 2002, 2005; KPMG, 2002); Arthur Andersen’s Business Audit
(Arthur Andersen, 1998); Deloitte and Touche’s Risk Assessment Methodology (Armour, 2000);
PricewaterhouseCoopers’ PwC Audit Approach (Winograd et al., 2000); and Ernst and Young’s
Audit Innovation. See also Lemon et al. (2000) and Robson et al. (2007).
735
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
producing both what are referred to as ‘risk-adjusted’ audits (Mock and
Wright, 1999, p. 55) and efﬁcient audits. That is, lower-risk audit engagements
consuming fewer audit hours and resources than higher-risk audit
engagements. This pattern of resourcing means that audits are neither overaudited nor under-audited. Over-auditing is when unnecessary audit
procedures are carried out, because the risk of material misstatements at the
account balance has been incorrectly assessed as high. Under-auditing is when
required audit procedures are not carried out, because the risk of material
misstatements at the account balance has been incorrectly assessed as not
high. Under-auditing is more serious than over-auditing, or audit inefﬁciency,
because it adversely impacts that aspect of audit quality related to audit
effectiveness. Therefore, by focusing audit resource allocation to address overauditing, the BRA approach not only can directly improve production
efﬁciency but it may also facilitate improvements in audit quality related to
such audits being effectively produced, or not under-audited (e.g., Bell et al.,
1997; Peecher et al., 2007). Whether the BRA approach drives the production
of risk-adjusted and more efﬁcient audits are empirical questions that have
been largely unanswered.
This study differs from prior BRA approach literature in that it seeks to
answer these questions using internally generated (proprietary) data and a
direct, validated measurement of efﬁciency that informs how well an
organization uses its resources to achieve required outcomes. The literature so
far has been either largely theoretical or based on indirect measures of
efﬁciency, such as changes (reductions or increases) in audit hours or audit fees
charged per client assets (Curtis and Turley, 2007). In this study, audit
production efﬁciency is measured not by changes in audit hours, audit costs/
revenues, or audit fees, but by the extent to which audit resources (inputs) are
optimally transformed into outputs to produce desired audit outcomes. The
efﬁciency of a production process can be examined and measured using the
analytical technique of data envelopment analysis (DEA). DEA is a nonparametric optimization technique that computes and benchmarks the relative
efﬁciency of a production process for a given set of inputs and outputs that
capture the production factors and outcomes, respectively, and other production
assumptions. This performance measure is referred to as theta. The conceptual
research question (including borrowing the words of Bell et al., 1997, p. 1) is:
Does the BRA approach result in ‘creating auditing efﬁciencies’? Speciﬁcally,
this study examines whether implementing the BRA approach addresses the
over-auditing of lower-risk audit engagements and, therefore, improves audit
production efﬁciency through such risk-adjusted audits.
To address this research question, a general model of audit production
efﬁciency is presented, drawn from production efﬁciency models used in the social,
physical, and technological sciences. These models include theta as the dependent
variable, measuring the relative production efﬁciency of an outcome (e.g., in an
engineering context, the efﬁcient use of resources for building a bridge, with
inputs including the costs of labour and materials, and outputs including the
736
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
bridge’s size, carrying capacity, etc.). In an auditing context, this dependent
variable can be the relative production efﬁciency of completing an audit
engagement, with inputs including the cost of labour (e.g., directors, managers,
seniors, auditors) and outputs including the hours of different evidence-gathering
audit activities (e.g., planning, risk understanding, control and substantive testing).
The independent variables capture auditee characteristics, including auditorassessed risks, such as ABR, this study’s test variable for the BRA approach, audit
production characteristics, and audit team characteristics, including audit team
composition and audit staff experience.
The data used in this study are provided on a conﬁdential basis and include
many sensitive audit production-based variables. The data are supplied by a major
Australian public sector audit provider with auditees having varying business risk
proﬁles. These include government departments, quasi-government agencies and
authorities, local councils, and other public sector entities. Data include
information from the working papers of large (not randomly selected) audit
engagements carried out in ﬁnancial years shortly before and after this audit
provider implemented the BRA approach. This approach was adopted to help
achieve sought-after productivity gains, particularly via a reduction in the overauditing of its lower-risk audit engagements.
Results using audit hours-based data for a sample of 60 audit engagements
carried out immediately after the implementation of the BRA approach are as
follows. On audit production, results show that higher business risk audit
engagements require more audit hours and, consequently, are more costly to
produce, and are charged higher audit fees. In addition, they have audit teams
with more experienced audit directors. On audit production efﬁciency, results
show no signiﬁcant difference between higher and lower business risk audit
engagements. Other results show that the level of production efﬁciency is
relatively high, with most audits produced relatively efﬁciently. Therefore, in
the context of this study’s post-BRA approach implementation setting and
sample of audit engagements, these results reveal risk-adjusted and efﬁciently
produced audits, with no evidence of over-auditing of the lower-risk audit
engagements.
Also examined is whether there is a change in audit production efﬁciency for a
sample of 74 audit engagements carried out before and after the implementation
of the BRA approach. This is tested using publicly available audit fees data, as the
proprietary audit production data (audit labour hours and audit activity hours) are
not available for audit engagements carried out prior to the adoption of the BRA
approach. Results show that audit production efﬁciency signiﬁcantly improves in
the year after the implementation of the BRA approach. Further, this ﬁnding is
for the full sample and the sub-sample of lower-risk audit engagements, but not
for the sub-sample of higher-risk audit engagements. These results are consistent
with over-auditing of the lower-risk audit engagements before the implementation
of the BRA approach. In sum, the above results reveal that implementing the
BRA approach can improve audit production efﬁciency, particularly where there
are risk-adjusted audits.
737
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
BACKGROUND
The Emergence and Re-emergence of the BRA Approach
The origins of the BRA approach surfaced in the mid- to late 1980s when forms of
the BRA approach emerged ‘in parallel without much collaboration’ and
represented a ‘re-engineering’ of the audit process (Knechel, 2007, pp. 385, 393).
This progression is seen as a response to key changes to audit methodology and
practice alongside other forces for change during and beyond the 1990s (Power,
2007; Robson et al., 2007). These forces can be summarized as: (i) a focus on client
risk management (Knechel, 2007); (ii) spectacular corporate collapses
(e.g., Enron, Waste Management, World Com) and audit failures (e.g., Arthur
Andersen); (iii) an increasingly complex global business environment and its
impact on ﬁnancial reporting (Bell et al., 1997); and (iv) the emergence of a
market (demand and supply) for value-added assurance services (Bell et al., 1997;
Robson et al., 2007).
From a somewhat cynical perspective, Power (2007, p. 380) points out that
forms of the BRA approach were taken up in ‘a response to a perceived need to
restructure a service product in decline and with low internal prestige in the large
ﬁrms’. Here, the origin of the BRA approach is seen as ‘a signiﬁcant political
project’ (Knechel, 2007, p. 427, citing Covaleski et al., 2003) and ‘opportunistic
behaviour’ (Knechel, 2007, p. 427) stemming from changed and changing audit
market conditions that began in the mid-1980s. During the 1990s, there was a
decline in audit services’ proﬁtability attributable to rising costs (e.g., in training
and insurance) and competitive pressure on audit pricing (e.g., see Robson et al.,
2007, Tables 1–3). This meant that audit ﬁrms increased, or at least maintained,
total revenues through increased provision of lucrative consulting (non-audit)
services. In other words, in response to a highly competitive, undifferentiated,
and regulated market of traditional audit services, the BRA approach was ‘coconstructed’ (Robson et al., 2007, p. 423) and then legitimized as a gateway for
providing value-added assurance services from which the knowledge spillovers
could ﬂow into the traditional ﬁnancial report audit.
Another perspective on the naissance and establishment of the BRA approach
relates to the desire for audit ﬁrms not only to maintain total revenue streams, but
also to create, or at least squeeze, greater efﬁciency out of production processes
(Bell et al., 1997). This was particularly the case for the traditional ﬁnancial audit
services market, which, as noted above, was also struggling in terms of rising costs
and declining proﬁts. If, as claimed, improvements in ﬁnancial audit production
efﬁciency can be possible given the BRA approach’s ability to generate
knowledge spillovers into the traditional ﬁnancial report audit, ultimately this
makes cutting audit costs and increasing proﬁts more possible.
The BRA approach has continual interest from the research community (see,
e.g., Fukukawa et al., 2006, 2011; Abdullatif and Al-Khadash, 2010; Fukukawa and
Mock 2011; van Buuren et al., 2014, 2018; Curtis et al., 2016; Wright, 2016). It also
continues to have detractors and sceptics about its added value, mainly to the
small to mid-sized segments of the audit industry (Humphrey et al., 2004; Curtis
738
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
et al., 2016). The high-proﬁle audit failures around the turn of the century
(e.g., Enron) played a major role in motivating much of the early cynicism, if not,
criticism, of the BRA approach (see, e.g., Covaleski et al., 2003; Knechel, 2007;
Power, 2007; Robson et al., 2007). However, and probably because of a lack of
high-proﬁle audit failures since the early part of this century, the recent reinvigorated and strong (web-based) marketing campaigns of the Big 4 audit ﬁrms
suggest that the BRA approach has re-emerged bigger and better than before
(KPMG, 2016; Deloitte, 2017; Ernst and Young, 2017; PwC Australia, 2018).
Further, regulators and standard setters worldwide have not moved away from a
position of legitimizing the BRA approach. This is set through auditing standards
AS 5, ASA 315, and ISA 315, whose prior versions were released soon after the
collapse of Enron and the demise of Arthur Andersen. This legitimization
supports the view that the re-emergence of the BRA approach will further
entrench its place in contemporary audit methodology and practice, particularly of
the Big 4 audit ﬁrms. Most likely, the BRA approach will withstand the next audit
industry shock, and re-ignite discussion from defenders and supporters alike.
Operationalization of the BRA Approach: Its Application in Australasia
While the BRA approach was formally espoused in international and Australian
standards of auditing, there were challenges apparent after the Enron and other
corporate collapses. Australia had its own Enron equivalent with the collapse of
HIH Insurance—at the time, the country’s largest insurer in some markets. That
company was also audited by Arthur Andersen and, like Enron, there were public
claims that audit failure was a key feature of this collapse. This public disquiet
rendered the demise of the BRA approach—at least its high public proﬁle—in
both Australia’s private and public sector auditing environments. As noted above,
the rebirth of the BRA approach occurred in the early to mid-2010s, and today
the literature produced by the large professional accounting services highlights the
importance of planning assurance services around assessed client business risks.
In the late 1990s in Australasia, there was a growing movement to refashion
public sector auditing by enhancing competition and efﬁciency (English, 1997,
2003; Maddock et al., 1997; Houghton and Jubb, 1998; Guthrie and Parker, 1999;
Harris, 1999; Houghton et al., 2002; De Martinis and Clark, 2003; Chong et al.,
2009). For this study’s major public sector audit provider, the response to this
environment was a series of strategic and operational changes. The ethos of this
audit ﬁrm was to seek new approaches and methodologies to secure more efﬁcient
(and therefore less costly) audit production capabilities. At an operational level,
the aim was to achieve signiﬁcant productivity gains across its portfolio of
auditees, speciﬁcally, via a reduction in the over-auditing of its lower-risk audit
engagements.
A key component for achieving this productivity improvement was the decision
to adopt the BRA approach. Indeed, this audit ﬁrm can be seen as an early
adopter of this methodology. The ﬁrst year of implementation was 2000. The
implementation of this approach resulted in the abandonment of what is referred
to as the ‘same-as-last-year’ (SALY) approach to audit planning (Mock and
739
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
Wright, 1999). The SALY approach is characterized by the repeated use of
previous audit production characteristics and resourcing. This can result in underor over-auditing, or audits that are not risk-adjusted. The audit ﬁrm’s expectation
was that BRA-approach audit engagements could be planned and executed, with
the required quality and the necessary variation in audit effort, to appropriately
and efﬁciently reﬂect auditee characteristics, particularly the level of business risk.
Put more directly, wastage of audit resources could be removed under a BRA
approach.
RESEARCH FOCUS AND HYPOTHESIS DEVELOPMENT
As discussed above, the literature signals that the BRA approach can be useful in
producing auditing efﬁciencies. This is because the auditor concentrates on
holistic, complex ﬁnancial report-/entity-level risks. This awareness can enhance
audit procedure selection and result in audits resourced without wastage, or not
over-audited. Such enhancements can address other forms of audit inefﬁciencies,
including those stemming from audit cost pressures resulting from over-auditing
(Lemon et al., 2000). Further, such attention may lead to audit quality
improvements related to audits being effectively produced, or not under-audited
(e.g., Bell et al., 1997; Peecher et al., 2007). A practical example of this is when the
auditor better recognizes and appropriately chooses audit procedures that are
both more informative and diagnostic (Bell et al., 2008). The scope of this study is
the practical impact of the BRA approach on over-auditing.
Bell et al. (2008) provide some support for efﬁciency improvements by reporting
marginally lower total labour usage and audit fees than pre-BRA approach
benchmarks of these metrics for clients of a Big 4 audit ﬁrm. Their ﬁnding of an
overall decline in audit fees is generally consistent with results from more recent
studies examining changes in audit fees when the PCAOB replaced its ‘costly’ AS
2 (PCAOB, 2004), seen as the traditional audit approach, with the ‘less costly’ topdown-based AS 5, reﬂecting the BRA approach (see, e.g., Doogar et al., 2010;
Kinney and Shepardson, 2011; Krishnan et al., 2011). These studies measure audit
efﬁciency improvements based on evidence of audit fee reduction following
compliance with AS 2, and with AS 5 relative to AS 2. Other studies also show
that the BRA approach can improve audit efﬁciency and, in particular, audit
effectiveness, under certain circumstances (Erickson et al., 2000; Bell and
Solomon, 2002; Choy and King, 2005). However, these studies are distinct from
those that ﬁnd signiﬁcant audit fee increases after the implementation of AS
2 (see, e.g., Raghunandan and Rama, 2006; Foster et al., 2007; Hogan and Wilkins,
2008; Hoitash et al., 2008; Krishnan et al., 2008).
On the other hand, Knechel (2007, p. 401) suggests that such efﬁciency
improvements can be read simply as ‘less work’. Further, Bell et al. (2008) point
out that in other settings the requisite focus on key complex risks can lead to
inefﬁciencies, for example, in the form of increased labour allocation in total or at
any labour rank. That study’s multivariate results show a positive (no) association
740
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
between their composite client business risk measure and total audit hours of new
(continuing) clients. Curtis and Turley (2007) provide evidence of this form of
inefﬁciency in reporting higher total audit hours for a major accounting ﬁrm in the
early stages of implementing the BRA approach. Blokdijk et al. (2006) ﬁnd no
association between a BRA approach variable and total audit hours for the clients
of a Big 5 audit ﬁrm, and offer no direct observations on audit production
efﬁciency.
Apart from uncertainty as to whether the BRA approach can create auditing
efﬁciencies, these mixed results also bring into question the metrics used to make
such inferences about the efﬁciency of the BRA approach, that is, based on levels,
trends, and changes in total audit labour hours and other related metrics (audit
fees). As explained below, efﬁciency is a relative concept with metrics measuring
how well an organization uses its resources (inputs) to achieve desired outcomes.
Further, there is tension in the literature regarding whether the BRA approach
can produce higher-quality audits (e.g., Bell et al., 1997; Peecher et al., 2007), given
that ‘less work’ (Knechel, 2007, p. 401) can represent ‘an unsound departure from
traditional audit practices that can lead to under auditing’ (Bell et al., 2008,
p. 730). Others suggest that the BRA approach is simply a marketing mechanism
to generate non-audit service revenues amidst falling audit service revenues (see,
e.g., Knechel, 2007; Power, 2007; Robson et al., 2007).
On the positive side, and with a focus on audit quality, an increasing number of
experimental studies focus on auditor judgements and decision making in the
context of the BRA approach. Following the works of Knechel et al. (2010),
Schultz et al. (2010), and Kochetova-Kozloski and Messier (2011), KochetovaKozloski et al. (2013) examine and report evidence of a positive association
between linkages between business risks identiﬁed at the entity and process levels.
Schultz et al. (2010) provide evidence that this approach can enhance the auditor’s
attention to business risk because they are more likely to integrate evidence about
business risk directly into planning judgements about the risk of material
misstatements. O’Donnell and Schultz (2003) ﬁnd evidence that this approach can
improve auditor judgement by increasing the identiﬁcation of risk factors during
analytical procedures.
Collectively, and despite the above noted tension, these studies provide
compelling evidence that the BRA approach can raise auditing efﬁciencies
through enhanced auditor judgement and decision making about the selection of
audit procedures and the interpretation of results thereof. Ultimately, such
enhancements increase the likelihood of forming and reporting an appropriate
audit opinion (De Angelo, 1981). Accordingly, the BRA approach may improve
audit quality from both efﬁciency and effectiveness perspectives. As noted above,
this study focuses on the efﬁciency issue.
However, Ballou et al. (2004) and O’Donnell and Schultz (2005) show that
auditors who use the BRA and strategic systems auditing approaches to form
preliminary assessments of the risk of material misstatements are more likely to
de-sensitize later-stage auditor judgements, that is, at the assertion or account
balance level. This ﬁnding suggests that such de-sensitization can adversely impact
741
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
both audit efﬁciency (over-auditing) and audit effectiveness (under-auditing).
Bruynseels et al. (2011, p. 1) report that ‘audit ﬁrms that use a business risk audit
methodology are less likely to issue a going-concern opinion for a ﬁrm that
subsequently goes bankrupt if the client has undertaken operating initiatives to
mitigate ﬁnancial distress’.
In sum, primarily drawing from auditing standards (AS 5, AS 12, ASA 315, ISA
315), and studies supporting the BRA approach (see, e.g., Bell et al., 1997;
Peecher et al., 1997; Erickson et al., 2000; Bell and Solomon, 2002; Schultz et al.,
2010; Kochetova-Kozloski et al., 2013), the BRA approach is, ceteris paribus,
expected to generate auditing efﬁciencies. However, the tension and results from
other studies suggest the contrary (see, e.g., Ballou et al., 2004; O’Donnell and
Schultz, 2005; Blokdijk et al., 2006; Curtis and Turley, 2007; Knechel, 2007; Bell
et al., 2008; Bruynseels et al., 2011). Further, a common theme is that measures of
audit efﬁciency are largely, if not entirely, indirect. This combination of mixed
results and use of indirect metrics means that questions on the extent to which the
BRA approach can produce risk-adjusted and more efﬁcient audits remain largely
unanswered. This study seeks to answer these questions empirically using a
unique, working paper-sourced dataset. The operational research question of this
study is: In context of the BRA approach creating auditing efﬁciencies and within
a public sector audit environment in Australia, are lower-risk audit engagements
produced as efﬁciently as higher-risk audit engagements?
Speciﬁcally, the hypothesis tested is:
H1: When the BRA approach is used, there is no difference in the production
efﬁciency of audit engagements with higher or lower assessments of auditee
business risk.
MEASURING AUDIT PERFORMANCE
The research question and hypothesis both refer to audit production efﬁciency as
a measure of audit performance. To test the hypothesis this study uses audit
production data in the form of different audit team labour hours and audit activity
hours, and not a common proxy, such as audit fees charged. The following details
the construction of a valid and reliable method for measuring production
efﬁciency, widely adopted in a variety of scholarly contexts, including audit
production.
Productivity, Efﬁciency, and Performance
Before discussing and measuring audit performance in the context of audit
production efﬁciency, it is useful to explain in general terms the concepts of
productivity, efﬁciency, and performance. Productivity is an absolute concept
measured by the ratio of outputs to inputs, while efﬁciency is a relative concept
measured by comparing the actual ratio of outputs to inputs with the optimal ratio
742
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
of outputs to inputs. Performance can be measured across production outcomes
for several objectives, including over time for the same producer, relative to other
producers (i.e., benchmarking), or explaining planning deviations. Regardless of
the objective, such comparative performance analysis can be undertaken by
various parametric and non-parametric (linear programming) performance
evaluation methods. These include single (output/input) ratio analysis, leastsquares regression, total factor productivity, and frontier techniques such as
stochastic frontier analysis (SFA) and data envelopment analysis (DEA).
Measuring Performance Using DEA
DEA is extensively used in the scholarly literature to measure and benchmark the
relative efﬁciency of decision-making units (DMUs) (Farrell, 1957; Charnes et al.,
1978; Banker et al., 1984). As a non-parametric optimization technique, it extracts
efﬁcient DMUs from the sample of DMUs being examined. It does this via an
algorithm that calculates the relative efﬁciency of each DMU’s production process
(i.e., theta) in reference to a speciﬁed production framework and set of
assumptions. The production framework consists of the same set of production
factors (i.e., inputs) and outcomes (i.e., outputs) for the entire sample of DMUs.
Two key production assumptions are on the orientation of the model; input
minimization, that is, when the objective is to minimize the inputs used given the
output(s) achieved, and/or output maximization, that is, when the objective is to
maximize the output(s) achieved given the inputs used. The other key assumption
is on the characterization of the production process; constant returns to scale
(CRS) or variable returns to scale (VRS). While different algorithms can reﬂect
different production frameworks and assumptions, basically, they compare a
DMU’s actual ratio of outputs to inputs to the optimal ratio of outputs to inputs,
which is calculated in reference to the other sample DMUs.
The literature shows the use of DEA in a wide variety of settings to measure
the relative efﬁciencies of DMUs (see, e.g., Charnes et al., 1994; Cooper et al.,
2004, 2006, 2007.) These include government and public sector service providers,
banking (bank branches), transportation (e.g., highway patrols, airlines), health
care (e.g., hospital services), ﬁnancial markets (e.g., equity, mutual funds),
education (e.g., universities/higher education providers (Moradi-Motlagh et al.,
2016), schools and libraries, and ﬁrms (e.g., operational and managerial
efﬁciency). DEA helps answer questions such as: How many more outputs can be
produced and/or how many fewer inputs can be used in order to perform at best
practice? Given the nature of this study’s research question, DEA is an
appropriate analytical method.
Studies of Audit Production Efﬁciency Using DEA
DEA can examine audit performance by measuring the relative efﬁciency of an
audit production process in reference to a speciﬁed production framework
(i.e., inputs and outputs) and key production assumptions (e.g., input minimization
and/or output maximization, and CRS or VRS, where the latter assumes that audit
engagements exhibit scale effects in audit effort with increases in client
743
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
characteristics (i.e., size, complexity, risk) (Knechel et al., 2009; Gaeremynck et al.,
2016)). Table 1 provides a brief chronological review of seminal audit production
efﬁciency studies structured around these production speciﬁcations and
assumptions. This review highlights how such performance studies are relatively
scarce due to the unavailability and lack of necessary and often private data. It
also highlights how DEA can determine relative optimality in the production of
audit services, which then can be used to implement change in the carrying out of
audits (Gaeremynck et al., 2016).2
Table 1 shows that in terms of audit production frameworks, the inputs
predominantly reﬂect the different labour ranks of an audit team (e.g., partner/
director, manager, supervisor/senior, and auditor), measured in terms of effort
expended (hours), costs incurred, or revenue generated. While the outputs are
more varied, for example, reﬂecting measures of audit ﬁrm revenues, client
characteristics related to size and risks, and hours on various evidence-gathering
procedures. In terms of optimization assumptions, input orientation
(minimization) and output orientation (maximization) are equally applied.
However, on the characterization of the audit production function, VRS is
predominantly applied. In sum, this literature review shows that regardless of the
research setting, the DMU of interest (i.e., the unit of analysis such as audit ﬁrm
or audit engagement), the speciﬁed audit production framework, or the DEA
model assumptions, the average production efﬁciency of audits (theta) is around
80 to 85%.3
Of particular relevance to this study are the scarce audit production efﬁciency
studies whose unit of analysis is the audit engagement or audit team (Knechel
et al., 2009; Dopuch et al., 2003; Gaeremynck et al., 2016; Chang et al., 2018).
These studies highlight the nascent state of this literature, and how even within
this small number of studies there is diversity in their theoretical frameworks
and research settings (i.e., source and type of production data). On the
theoretical frameworks, that diversity is operationalized through different audit
2
There are other streams of related literature that broadly examine the public accounting industry
(including Big N audit ﬁrms) in terms of productivity (production function) (Banker et al., 2003;
Chang et al., 2015) and productivity growth (Lin et al., 2008; Chang et al., 2011). These studies are
distinct from those that examine audit performance, productivity, or production (via audit hours or
audit fees) at the audit ﬁrm (including ofﬁce/branch) level or the audit engagement level (see,
e.g., seminal studies by Simunic (1980), Ferris and Larcker (1983), and O’Keefe et al. (1994)).
However, none of these studies examine audit production efﬁciency from a benchmarking context
(e.g., using DEA) and, therefore, are beyond the scope of this study. For detailed reviews, see Hay
et al. (2006) and Causholli et al. (2010).
3
It is beyond the scope of this study to question the choice of inputs, outputs, or other DEA
assumption of the studies listed in Table 1. For example, the use of client characteristics as outputs
given the assumption of constant level of assurance by Dopuch et al. (2003) and Chang et al. (2018)
(Knechel et al., 2009; Gaeremynck et al., 2016). The rationale for this study’s DEA assumptions for
measuring audit production efﬁciency is provided below. Further, while Table 1 notes the studies’
DEA performance measure results, that is, the thetas, which for this study are THETA_HOURS and
THETA_FEES (see Table 3), it does not include results explaining the thetas, that is, where the
thetas are used as the dependent variable in subsequent regression analyses (see equation (1) and
Tables 7 and 10).
744
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
745
© 2019 Accounting Foundation, The University of Sydney
Dopuch et al. US, a Big 6 audit ﬁrm,
(2003)
247 audits, 198 public
ﬁrms and 49 private
ﬁrms; 1989
Banker et al. One international public
(2002)
accounting ﬁrm, ﬁve
ofﬁces; 1997
(60 months) and 1999
(60 months)
• No. of ﬁrm
Taiwan, 150 CPA ﬁrms;
1994
Cheng et al.
(2000)
Hours of different
labour ranks used,
including:
• partner
• manager
• senior
• staff
labour costs for
different levels of
accounting
professionals
• Monthly ofﬁce
operating costs and
ofﬁce expenses
• Monthly ofﬁce
employees
• Firm net ﬁxed
assets
Inputs
Research setting details
Study
Client characteristics for:
• size
• risk
revenue, computed
from annual fees
• Monthly ofﬁce
• Firm revenue
Outputs
Output
Output
Input
Key DEA result(s)
(Continues)
The mean efﬁciency score
is 0.722. The mean
efﬁciency score for the
non-efﬁcient ﬁrms is
0.614. Twenty-eight
percent of the ﬁrms are
efﬁcient.
Not determinable The mean (in)efﬁciency
score in 1997 is (1.085)
0.915, in 1999 it is
(1.051) 0.949. Efﬁciency
improves by about 3%
after investment and
implementation of
information technology
in 1998.
Constant
The mean efﬁciency score
is 0.88 and 85.8% are
efﬁcient. Public ﬁrms
are more efﬁcient than
private ﬁrms; the mean
efﬁciency scores are
0.952 and 0.876,
respectively. However,
more private ﬁrms are
efﬁcient; 88.8% versus
69.3% for the public
ﬁrms.
Variable
Orientation
Returns to scale
(input,
output,
(constant or
or both)
variable)
Audit production framework details and DEA assumptions
AUDIT PRODUCTION EFFICIENCY STUDIES USING DEA
TABLE 1
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
© 2019 Accounting Foundation, The University of Sydney
746
Research setting details
Inputs
Tsai et al.
(2008)
Taiwan, 120 audit ﬁrms,
including 10 merged
audit ﬁrms; 1997 to
2001
Service revenues from:
• accounting and
auditing
• tax
• management advisory
Outputs
Service revenues from: • Total annual salary
divided by total no. of
employees
• Total expenditure
divided by total
consultation, and
amount of book value
other
of ﬁxed assets
• No. of audit ﬁrm
branches
• audit
• tax
• management,
Banker et al. US, 100 of the largest Big • Professional labour
(2005)
5 and non-Big 5 audit
costs, including
ﬁrms; 1995 to 1999
salaries paid
different levels of
accounting
professionals
• Operating costs
• No. of other
employees
Study
Input
Output
Variable
Variable
Orientation
(input,
Returns to scale
output,
(constant or
or both)
variable)
Audit production framework details and DEA assumptions
CONTINUED
TABLE 1
(Continues)
Between 1995 and 1999
the mean efﬁciency
score declines from
0.884 to 0.864 (2.5%).
The Big 5 ﬁrms are
more efﬁcient than the
non-Big 5 ﬁrms both in
1995 (0.995 versus
0.875) and in 1999 (1.00
versus 0.852). These
results are despite
productivity growth
increasing by about
9.5% over the same
period.
Across the ﬁve years, the
mean (allocative)
efﬁciency scores for the
full sample, merged
audit ﬁrms, and nonmerged audit ﬁrms are
0.849, 0.839, and 0.946,
respectively.
Key DEA result(s)
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
747
© 2019 Accounting Foundation, The University of Sydney
Chang et al.
(2009b)
US, 62 of the 100 largest •
CPA ﬁrms, including
Big 4 and non-Big
4 accounting ﬁrms; pre
•
and post the enactment
of the SOX Act in 2002;
•
2000 to 2001 and 2003
to 2004
US, 56 of the 100 largest •
CPA ﬁrms, including
Big 4 and non-Big
4 accounting ﬁrms; pre
•
and post the enactment
of the SOX Act in 2002;
•
1996 to 1999 and 2003
to 2006
Chang et al.
(2009a)
No. of partners,
owners, and/or
shareholders
No. of other
professionals
No. of other
employees
No. of partners,
owners, and/or
shareholders
No. of other
professionals
No. of other
employees
Inputs
Research setting details
Study
Service revenues from:
• accounting and
auditing
• taxation
• management advisory
Service revenues from:
• accounting and
auditing
• taxation
• management advisory
Outputs
Output
Output
Constant
Constant
Orientation
Returns to scale
(input,
(constant or
output,
variable)
or both)
Audit production framework details and DEA assumptions
CONTINUED
TABLE 1
(Continues)
Between the pre- and
post-SOX period, mean
(in)efﬁciency score (in)
decreases from (1.297)
0.703 to (1.344) 0.656
(3.1%). The Big
4 ﬁrms’ efﬁciency score
increases slightly from
0.978 to 0.99 (1.1%),
whereas the non-Big
4 ﬁrms’ efﬁciency score
decreases from 0.682 to
0.630 (3.4%). Over the
same period,
productivity growth
increases by 16.8%.
Efﬁciency scores not
disclosed; only
efﬁciency score
changes, i.e., Malmquist
productivity index
decompositions.
Between the pre- and
post-SOX period,
productive efﬁciency
increases by 15.2%.
Non-Big 4 ﬁrms’
Key DEA result(s)
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
© 2019 Accounting Foundation, The University of Sydney
748
Taiwan, 173 medium-sized • No. of partners
audit ﬁrms; 2005
• No. of other
employees
• No. of branches
• Total expenditure
Lee (2009)
Inputs
Research setting details
Study
Service revenues from:
• attestation
• tax business
• management
consultancy
• corporate registration
and other and other
business services
Outputs
Output
Constant (for
overall
technical
efﬁciency).
Variable (for
pure technical
efﬁciency
and scale
efﬁciency)
Orientation
(input,
Returns to scale
output,
(constant or
or both)
variable)
Audit production framework details and DEA assumptions
CONTINUED
TABLE 1
(Continues)
efﬁciency score
improves by 16.2%. Big
4 ﬁrms’ efﬁciency score
improves by just under
1%. Over the same
period, productivity
growth increases by
14.5%.
The mean scores of
overall technical
efﬁciency, pure
technical efﬁciency, and
scale efﬁciency are
0.778, 0.863, and 0.902,
respectively. Larger
audit ﬁrms are
signiﬁcantly more
efﬁcient than smaller
audit ﬁrms. Audit ﬁrms
with higher revenues in
all businesses, more
branches, larger
number of total
employees and
partners, and higher
total expenditures have
higher efﬁciency scores.
Key DEA result(s)
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
749
© 2019 Accounting Foundation, The University of Sydney
US, a large international
audit ﬁrm, 307 audits;
1991
Knechel
et al.
(2009)
Total labour costs of
all audit personnel
used, including:
• partner
• manager
• supervisor
• staff
Inputs
Gaeremynck Belgium, ofﬁces of an
Categories of relative
et al.
international accounting
(not actual) auditor
(2016)
ﬁrm, 158 audit
costs by staff ranks
engagements, using
(adjusted for
three-stage DEA (Fried
environmental
et al., 2002); 2006 or
characteristics
2007
slacks), including:
• partner
• manager
• senior
• assistant
• expert
Research setting details
Study
Input
Level of assurance
Input
provided (audit
quality), proxied by the
level of planning
materiality
(in monetary terms,
euros), deﬁned as the
inverse of the level of
engagement materiality
Disaggregated labour
hours spent on up to
eight categories of
evidence-gathering
activities
Outputs
Variable
Constant and
Variable
Orientation
Returns to scale
(input,
output,
(constant or
or both)
variable)
Audit production framework details and DEA assumptions
CONTINUED
TABLE 1
(Continues)
Across the set of
alternative DEA model
speciﬁcations, the mean
efﬁciency score range is
0.8053 to 0.9558. Audits
are more efﬁcient for
clients that are larger,
have a December yearend, and highly
automated. Audits are
less efﬁcient when the
auditor relies on
internal control, tax
services are provided,
and the client has
subsidiaries.
Controlling for
uncontrollable
environmental (ﬁrm)
characteristics, i.e., the
stage three results, the
mean efﬁciency score is
0.72, with 66% efﬁcient.
These results show that
both controllable and
uncontrollable
characteristics of the
audit engagement
contribute to efﬁciency.
Key DEA result(s)
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
© 2019 Accounting Foundation, The University of Sydney
750
Australia, a major public
sector audit services
provider, after it
Branch ofﬁce of a Big
4 international public
accounting ﬁrm,
165 audit engagements
of its largest clients;
during a recent ﬁscal
year
Chang et al.
(2018)
This study
Research setting details
Study
Outputs
Total audit
Disaggregated hours of
engagement costs,
evidence-gathering
based on the labour
Staff hours of different Client characteristics for:
categories of audit
• size
professionals
• complexity
including:
• risk
• partners
• managers
• other
Inputs
Input
Input
Variable
Variable
Orientation
(input,
Returns to scale
output,
(constant or
or both)
variable)
Audit production framework details and DEA assumptions
CONTINUED
TABLE 1
(Continues)
The mean technical (in)
efﬁciency score is
(1.481) 0.519, with (102)
63 audit engagements
having technical (in)
efﬁciency (less than)
equal to one. The mean
allocative (in)efﬁciency
is (1.191) 0.809, with
(142) 23 audit
engagements having
allocative (in)efﬁciency
(less than) equal to
1. (Forty-one) 124 audit
engagements are both
technically and
allocatively (in)efﬁcient.
Both technical and
allocative (in)
efﬁciencies lead to
(lower) higher billing
realization rates.
The mean–
regular–efﬁciency score
is 0.957 and 55% are
Key DEA result(s)
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
751
© 2019 Accounting Foundation, The University of Sydney
Study
Australia, a major public
sector audit services
provider, before and
after it implemented
the BRA approach in
2000; 74 audit
engagements; 1999 and
2001
implemented the
business risk audit
(BRA) approach in
2000; 60 audit
engagements; 2000 and
2001
Research setting details
Outputs
Auditee characteristics Audit fees
for:
• size
• risk (including
auditee
business risk)
costs of audit team
audit activities,
members, including:
including:
• directors
• planning and analytical
• managers
• internal control
understanding
• seniors
• internal control testing
• auditors
• substantive and closing
Inputs
Input
Variable
Orientation
Returns to scale
(input,
output,
(constant or
or both)
variable)
Audit production framework details and DEA assumptions
CONTINUED
TABLE 1
(Continues)
efﬁcient. The
mean–super,
bilateral–efﬁciency
score is 1.13 and 68.3%
are efﬁcient. No
signiﬁcant difference in
audit production
efﬁciency between the
higher and lower
business risk audit
engagements; i.e., the
BRA approach can
produce risk-adjusted
audits, with no overauditing of lower-risk
audit engagements.
The mean–
regular–efﬁciency score
is 0.826 and 12.2% are
efﬁcient. After
implementation of the
BRA approach, the
mean–super,
bilateral–efﬁciency
score signiﬁcantly
increases from 0.834 to
Key DEA result(s)
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
© 2019 Accounting Foundation, The University of Sydney
752
Study
Research setting details
Inputs
Outputs
Orientation
(input,
Returns to scale
output,
(constant or
or both)
variable)
Audit production framework details and DEA assumptions
CONTINUED
TABLE 1
1.058 (26.9%). Further,
there is a signiﬁcantly
greater proportion of
efﬁcient audit
engagements, increasing
from 11.1% to 39.5%.
For the lower-risk audit
engagements, the
mean–super,
bilateral–efﬁciency
score signiﬁcantly
increases from 0.847 to
1.097 (29.5%).
Key DEA result(s)
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
production frameworks and assumptions used to derive the different thetas.
Further, there is variation in the speciﬁcation and estimation of models used to
explain the thetas, that is, where different sets of explanatory variables are used
to explain the thetas.
Dopuch et al. (2003) and Chang et al. (2018) consider audit labour rank hours as
the inputs and client characteristics (size, complexity, and risks) as the outputs.
Dopuch et al.’s (2003) model orientation is output maximization. The objective of
this speciﬁcation is to maximize the client characteristics given the audit hours,
that is, an efﬁcient audit engagement is one produced with the greatest client
characteristics (largest size, complexity, and risks) given the audit hours. Chang
et al.’s (2018) model orientation is input minimization. The (literal) objective of
this speciﬁcation is to minimize the audit hours given the client characteristics.
That is, an efﬁcient audit engagement is one produced with the least audit hours
given the client characteristics (size, complexity, and risks).
Knechel et al. (2009) have total labour costs of audit personnel used as the
inputs, and hours of different combinations of evidence-gathering procedures as
the outputs, with input minimization as the model orientation. The objective of
this speciﬁcation is to minimize the audit labour costs given the audit hours. That
is, an efﬁcient audit engagement is one produced with the least labour costs given
the audit activity hours. Gaeremynck et al. (2016) have relative (not actual) costs
of various staff ranks (adjusted for environmental characteristic) as the inputs. The
output is assurance provided, proxied by planning materiality (in monetary terms,
euros), deﬁned as the inverse of the level of engagement materiality, with input
minimization as the model orientation. The objective of this speciﬁcation is to
minimize the audit labour costs given the level of assurance (materiality). That is,
an efﬁcient audit engagement is one produced with the least labour costs given the
level of assurance (materiality), which monotonically increases (decreases) with
more audit effort.
RESEARCH METHOD
Data Supply and Compilation
Table 2 summarizes the sample compilation. Given the detailed data
requirements, audit ﬁrm executives assisted with the preparation, administration,
and piloting of the data collection instrument. Further, they advised that the
sample consisted of large auditees, that is, audit fees greater than $25,000. This
resulted in the initial selection of 47 auditees. Audit engagement team staff
extracted and compiled the required data from the working papers and other inhouse management reporting systems. To enhance accurate responses, each
completed data collection instrument was reviewed by the audit team leader or an
audit ﬁrm executive. Missing auditor effort (hours) data resulted in usable
responses for 60 audit engagements; 22 and 38 for ﬁnancial years ending 2000 and
2001, respectively.
753
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
TABLE 2
SAMPLE COMPILATION
No. of auditees as selected by audit ﬁrm executives
Less missing auditor effort (hours) data
Audit engagements with all necessary data
Consisting of:
Government departments
Water authorities
Local councils
Education institutions
Othersb
a
2000
2001
Total
47
25
22
47
9
38
94
34
60
10
1
3
2
9
22
7
6
3
4
15
38
17
7
6
6
26
60
a
Given the detailed data requirements, the sample population of auditees consists of large auditees,
that is, audit fess greater than $25,000.
Includes hospital networks, insurance entities, sporting/development agencies, gaming authorities,
utility agencies, and other agencies. For these categories, none had more than three auditees.
b
Data Items
Table 3 details the data labels, descriptions, and measurements. Panel A contains
the auditor effort and audit production (hours) variables. These include total audit
hours (TOTALHOURS) or ‘input intensity’ (Blokdijk et al., 2006, p. 42),
disaggregated across four labour ranks and four key audit activities. The four
labour ranks are: directors (DIR) (equivalent to partner in the for-proﬁt audit
sector, but not a residual equity owner), managers (MAN), audit seniors (SEN),
and auditors (AUD). The four key audit activities are: preliminary and planning
procedures (PPA), analytical procedures and internal control risk understanding
procedures (IRU), internal control risk testing procedures (IRT), and additional
direct testing procedures, including substantive, review, and closing (ADT). With
proprietary cost rates for each labour rank this allows for the calculation of total
audit engagement cost (TOTALCOSTS). As conﬁrmed by audit ﬁrm executives,
these labour cost rates, also referred to as billing or charge rates, reﬂect a costrecovery-based audit fee pricing strategy. As such, they are void of the proﬁt
mark-up component normally imbedded in the billing or charge rates of for-proﬁt
audit ﬁrms (i.e., non-Big/Big 4 audit ﬁrms).
Table 3, Panel B contains the audit team characteristics variables. These include
measures of ‘audit technology’ (ACTIVITYMIX) and ‘input quality’
(LABOURMIX). While ‘input intensity’ refers to an audit engagement’s total
audit hours (Blokdijk et al., 2006, p. 42), ‘audit technology’ refers to the
proportion or mix of audit hours across key audit activities or procedures
(Blokdijk et al., 2006, p. 42). ACTIVITYMIX differentiates between soft,
judgemental activities, as carried out in the planning and risk assessment phases of
the audit, and other less judgemental activities, conducted in the other audit
phases. ‘Input quality’ refers to the proportion or mix of total audit hours across
staff labour ranks (Blokdijk et al., 2006, p. 28). That is, a ‘richer’ mix of audit
754
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
TABLE 3
DATA LABELS, DESCRIPTIONS, AND MEASUREMENTS
InX is the natural logarithm, where X is the variable of interest. Monetary values ($) are in Australian
dollars.
Panel A: Auditor effort and audit production characteristics
DIR, MAN, SEN, AUD are reported audit labour hours at the director (DIR), manager (MAN), senior
(SEN), and auditor labour ranks (AUD), respectively.
PPA, IRU, IRT, ADT are reported audit activity hours for preliminary and planning procedures
(PPA), analytical procedures and internal control risk understanding procedures (IRU), internal
control risk-testing procedures (IRT), and additional direct testing procedures, including substantive,
review, and closing (ADT), respectively.
TOTALHOURS (‘input intensity’, Blokdijk et al. (2006, p. 42) is total reported audit hours, calculated
as the sum of all audit labour rank hours (DIR + MAN + SEN + AUD) or the sum of all audit
activity hours (PPA + IRU + IRT + ADT).
TOTALCOSTS is the total audit engagement costs. This is based on the sum of the four individual
labour rank costs, which is calculated as the product of a labour rank’s hours (DIR, MAN, SEN,
AUD) and the cost rate for that labour rank.
FEES is total audit fees.
TOTALHOURS_RESIDUAL is the unstandardized OLS residual from an auditor effort model where
InTOTALHOURS is the dependant variable and the independent variables include InASSETS,
ABR ≥ 5 = 1*InASSETS, and other auditee risk characteristics (untabulated).
TOTALCOSTS_RESIDUAL is the unstandardized OLS residual from an auditor effort model where
InTOTALCOSTS is the dependant variable and the independent variables include InASSETS,
ABR ≥ 5 = 1*InASSETS, and other auditee risk characteristics (untabulated).
FEES_RESIDUAL is the unstandardized OLS residual from an auditor effort model where InFEES is
the dependant variable and the independent variables include InASSETS, ABR ≥ 5 = 1*InASSETS,
and other auditee risk characteristics (untabulated).
Panel B: Audit team characteristics
ACTIVITYMIX (‘audit technology’ (Blokdijk et al., 2006, p. 42)) is the proportion of the two higher
judgemental audit activity hours (PPA and IRU) to total audit hours ([PPA + IRT]/
TOTALHOURS).
LABOURMIX (‘input quality’ (Blokdijk et al., 2006, p. 28)) is the proportion of the two higher-ranked
labour hours to total audit hours (DIR and MAN) to total audit hours ([DIR + MAN])/
TOTALHOURS).
TENURE_DIR is the number of years the director has been on the audit engagement.
TENURE_DIR ≥ 5 = 1 is a dichotomous variable and equals one if TENURE_DIR is equal to or
greater than ﬁve years, otherwise zero.
INDUSTRY_DIR is the number of years the director has been involved in audit engagements in the
same industry as the current engagement.
INDUSTRY_DIR > 10 = 1 is a dichotomous variable and equals one if INDUSTRY_DIR is greater
than 10 years, otherwise zero.
Panel C: Auditee characteristics
ABR is auditor-assessed auditee business risk (range is 1 = extremely low to 7 = extremely high).
ABR ≥ 5 = 1 is a dichotomous variable that equals one if ABR is equal to or greater than ﬁve, that is, a
higher-risk audit engagement (Group High), otherwise zero, that is, a lower-risk audit engagement
(Group Low).
ASSETS is auditee total assets.
REVENUE is auditee total revenue.
Panel D: Audit production efﬁciency characteristics
THETA_HOURS is a DEA-computed bilateral and super efﬁciency score (super theta) between lower
ABR audit engagements and higher ABR audit engagements. It is based on the Knechel et al. (2009)
‘modiﬁed’ audit production framework assumptions, that is, variable returns to scale (VRS), input
orientation (input minimization). The input is total audit engagement cost (in $) based on the labour
costs of audit team members, proxied by this study’s TOTALCOST, and the outputs are
disaggregated audit engagement hours of various evidence-gathering audit activities, equivalent to
this study’s PPA, ERA, IRT, and ADT.
(Continues)
755
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
TABLE 3
CONTINUED
THETA_HOURS ≥ 1 = 1 is a dichotomous variable and equals one if THETA_HOURS equals or is
greater than one, otherwise zero.
Panel E: Variables for audit production efﬁciency pre and post BRA approach implementation
THETA_FEES is a DEA-computed bilateral and super efﬁciency score (super theta) between audit
engagements carried out the year before (year 1999) and the year after (year 2001) the
implementation of the BRA approach. It is based on the Knechel et al. (2009) ‘modiﬁed’ audit
production framework assumptions, that is, variable returns to scale (VRS), input orientation (input
minimization). The inputs are auditee characteristics, proxied by size (ASSETS) and risks (including
ABR), and the output is audit fees charged (FEES).
THETA_FEES ≥ 1 = 1 is a dichotomous variable and equals one if THETA_FEES equals or is greater
than one, otherwise zero.
THETA_FEES_LOW is THETA_FEES for the sub-sample of the lower ABR audit engagements.
THETA_FEES ≥ 1 = 1_LOW is a dichotomous variable and equals one if THETA_FEES equals or is
greater than one, otherwise zero, for a lower ABR audit engagement.
THETA_FEES_HIGH is THETA_FEES for the sub-sample of the higher ABR audit engagements.
THETA_FEES ≥ 1 = 1_HIGH is a dichotomous variable and equals one if THETA_FEES equals or is
greater than one, otherwise zero, for a higher ABR audit engagement.
DIR_NEW is a dichotomous variable and equals one for audit engagements with a new (ﬁrst year)
director, otherwise zero.
YEAR_POST is a dichotomous variable and equals one for audit engagements carried out in the year
after the implementation of the BRA approach (year 2001, Group Post), otherwise zero, that is, for
audit engagements carried out before the implementation of the BRA approach (year 1999, Group
Pre).
hours is where there is a greater proportion of higher-ranked labour (directors
and managers) hours. Also included are audit team experience variables centred
on measures of director tenure in the audit engagement (TENURE_DIR) and in
the same industry as the audit engagement (INDUSTRY_DIR).
Table 3, Panel C contains the auditee characteristic variables, including the
auditor’s assessment of ABR (ABR). This test variable is a risk factor constructed
from a set of ﬁve individual auditor-assessed ABR items, as supported by working
paper references. The risk items are measured using semantic differential scales
consisting of a seven-point equal interval scale, with the bi-polar verbal anchors
reﬂecting low to high levels of the auditor’s risk assessments. The range of the
scale is one to seven; the lower the score, the lower the assessed risk; that is, one
equals ‘extremely low’, seven equals ‘extremely high’. The ABR score is calculated
by adding the scores of its constituent risk items, and then dividing this summated
score by ﬁve (the number of constituent items) to obtain an average summated
factor score. An advantage of this method is that it addresses any missing data at
the item level, thereby avoiding the need to either drop audit engagements with
missing item data or estimate missing values using statistical techniques.4
Next, and after consulting with audit ﬁrm executives and consistent with
auditing standards that espouse a dichotomous assessment of risks as either non4
See Appendix A for further details on how the ﬁve individual auditor assessed ABR items are
selected and associated with the BRA approach, including statistics on the reliability and validity of
this factor.
756
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
high or high (ASA 330, ISA 330) (AUASB, 2015b; IAASB, 2015b), ABR is split
into two groups of such risk levels. One group contains audit engagements with
ABR scores less than ﬁve, representing the lower- (non-high) risk audit
engagements (Group Low). These audit engagements are assigned a new score of
zero. The other group contains audit engagements with ABR scores equal to or
greater than ﬁve, representing the higher-risk audit engagements (Group High).
These audit engagements are assigned a new score of one. This process creates a
dichotomous measure of ABR, ABR ≥ 5 = 1.5
DEA Performance Measure of Audit Production Efﬁciency
Table 3, Panel D contains an audit production efﬁciency measure computed using
DEA and based on Knechel et al.’s (2009) ‘modiﬁed’ audit production framework
(THETA_HOURS). This framework is chosen because its analysis of audit
performance is at the same level as this study, that is, at the audit engagement/
team level, and it uses an hours-based DEA measure of audit production
efﬁciency. In addition, its research setting is a US Big N audit ﬁrm, which is where
this study’s data provider purchased the intellectual property and resources that
enabled the production of BRA approach audits. This framework considers
evidence collecting audit activity hours as the outputs and total audit engagement
costs, based on labour hours times cost rates, as the input. The objective is to
minimize the total audit engagement costs given the audit activity hours. In this
case, an efﬁcient audit engagement is one produced with the lowest level of total
labour costs given the audit activity hours. In sum, Knechel et al.’s (2009)
framework is adapted since its model assumptions and data closely match this
study’s objectives and data.
As detailed in Table 1, the audit production framework and assumptions for
THETA_HOURS are: (i) the input is total audit engagement costs (in $) based on
the labour costs of audit team members (TOTALCOSTS); (ii) the outputs are
disaggregated audit engagement hours of various evidence-gathering audit
activities, equivalent to this study’s PPA, ERA, IRT, and ADT; (iii) orientation is
input (input minimization); and (iv) returns to scale are an assumed variable
(VRS). DEA always assigns inefﬁcient DMU values (i.e., thetas) between zero and
less than one. Depending on the algorithm, efﬁcient DMUs are assigned values of
either one (regular theta analysis) or equal to or greater than one (super theta
analysis). THETA_HOURS is dichotomized accordingly, creating THETA_
HOURS ≥ 1 = 1.
This study uses super theta analysis since this allows a ranking of the efﬁcient
DMUs (Esmaeilzadeh and Hadi-Vencheh, 2015). This is particularly useful when
5
Recall from Table 2 that the audit ﬁrm’s portfolio consists of government, quasi-government, and
public sector auditees. These can be considered as having a range of overall business risk between
low to medium. Nonetheless, within this set of auditees and range of business risk, the auditor
separately assesses risks at the entity level, such as ABR. Accordingly, such assessments can vary
between our lower (Group Low) and higher (Group High) labels. For example, an audit
engagement can have a high risk assessment, even though it is part of a portfolio of auditees whose
overall business risk is likely to be, at most, moderate.
757
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
comparing efﬁciency between groups of DMUs, such as this study’s lower- and
higher-risk audit engagements (see below). Super theta analysis also can be more
useful than regular theta analysis when the sample size of DMUs is small (less than
30) and there are many relatively efﬁcient DMUs (Cooper et al., 2006), as is the
case for this study, that is, around 70% (see Table 4, Panel B). Further, super
thetas are useful for detecting outliers, which DEA results can be sensitive to and
should be omitted from subsequent re-computation and analyses (Banker and
Chang, 2006). (See footnote 6.)
A General Model of Audit Production Efﬁciency
Hypothesis 1 tests for an association between audit production efﬁciency and the
level of ABR. Results for both univariate and multivariate analyses are reported
in the next section. The univariate analysis tests for differences in the relative
audit production efﬁciency of the lower-risk audit engagements (Group Low) and
the higher-risk audit engagements (Group High). The multivariate (regression)
analysis augments this initial test of Hypothesis 1 by helping explain the variation
in an audit engagement’s relative audit production efﬁciency.
In order to carry out this multivariate test of Hypothesis 1, a general model of
audit production efﬁciency is speciﬁed and estimated containing independent
variables for ‘controllable’ and ‘uncontrollable’ features of an audit engagement
(Knechel et al., 2009). Drawing from the relevant literature, such a model
explaining audit production efﬁciency is presented in the form of equation (1).
Audit production efficiency = b0 + b1 *½Auditee characteristics
+ b2 *½Audit production characteristics
+ b3 *½Audit team characteristics
ð1Þ
Equation (1) includes ‘uncontrollable’ audit engagement-based independent
variables capturing Auditee characteristics for size (InASSETS) and risk
(ABR ≥ 5 = 1). Also included are two categories of ‘controllable’ variables
capturing audit production characteristics and audit team characteristics. Audit
production characteristics include one of the three (unstandardized) residual-based
variables (FEES_RESIDUAL, TOTALHOURS_RESIDUAL, or TOTALCOSTS_
RESIDUAL). These control for the expectation that audit engagements with audit
fees, audit hours, or audit costs higher or lower than expected may be more or less
efﬁcient as auditors adjust the output to reﬂect the unexpected input level
(Knechel et al., 2009). Audit team characteristics include the variables for ‘audit
technology’ (ACTIVITYMIX), ‘input quality’ (LABOURMIX), and audit team
experience, particularly of the team leader (TENURE_DIR ≥ 5 = 1 and
INDUSTRY_DIR > 10 = 1). These are included since different proportions or mix
of labour and activity hours, and auditor experience are expected to impact audit
production efﬁciency differently.
758
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
The dependent variable is the dichotomous DEA performance measure on
whether production of the audit engagement is relatively efﬁcient
(THETA_HOURS ≥ 1 = 1). Consistent with the wording of Hypothesis
1, directional effects for the coefﬁcients of the independent variables are not
hypothesized since their impact on audit production efﬁciency is not yet clearly
established in the scarce determinants of audit production efﬁciency literature (see
Table 1).
RESULTS
Descriptive Statistics
Table 4 presents descriptive statistics for the sample data and variables used in
estimating equation (1) models. Focusing on the column (1) results (means), Panel
A shows that for the sample audit engagements most labour hours are (junior)
auditors (AUD = 380) and most audit activity hours are additional direct testing
(i.e., substantive testing) (ADT = 663). The mean total audit hours
(TOTALHOURS) is 1,042 hours, while the means for total audit costs
(TOTALCOSTS) and total audit fees (FEES) are $108,000 and $109,000,
respectively. This small difference is expected since, as noted, the audit ﬁrm’s
labour cost rates, used to calculate TOTALCOSTS, exclude a proﬁt mark-up
component and its audit fee pricing strategy is based on costrecovery. However, in
a for-proﬁt (non-Big/Big 4) audit sector setting, the differences between audit fees,
total audit revenues (based on charge rates that include a proﬁt mark-up
component), and total audit costs (based on labour cost rates) may be signiﬁcant.
Table 4, Panel B shows that 68.3% of the sample audit engagements are
relatively efﬁcient (THETA_HOURS ≥ 1 = 1 equals 0.683), with a mean (median)
of THETA_HOURS equal to 1.13 (1.031). Panel C shows that the proportions of
the softer, judgemental planning hours (ACTIVITYMIX) and of the richer, highlevel auditor hours (LABOURMIX) are both around 27% of total hours used.
The mean director tenure (TENURE_DIR) on an audit engagement is
3.383 years, with 20% having been a director for a period equal to or greater than
ﬁve years (TENURE_DIR ≥ 5 = 1). The mean number of years that directors
have been involved in audit engagements in the same industry as the current
engagement (INDUSTRY_DIR) is 9.23 years, with 35% of directors having
worked in that audit engagement’s industry for greater than 10 years
(INDUSTRY_DIR > 10 = 1). Panel D shows the mean (median) ABR assessment
score (ABR) is 4.425 (4.50), being just over the mid-point of the one to seven
Likert scale used. The mean size of the sample audit engagements as measured by
either assets (ASSETS) or revenues (REVENUE) is around $1.2 million.
Univariate Results
Table 5 presents descriptive statistics and a comparison of sample data and
variables used in estimating equation (1) models for the lower-risk (Group Low)
and higher-risk (Group High) audit engagements. Panel A contains the dependent
759
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
TABLE 4
DESCRIPTIVE STATISTICS FOR SAMPLE DATA AND VARIABLES USED IN
EQUATION (1) MODELSA
1
2
3
4
5
6
7
Percentiles
Statistics
Panel A: Auditor effort and production
characteristics
DIR
MAN
SEN
AUD
PPA
IRU
IRT
ADT
TOTALHOURS
TOTALCOSTS ($000)
FEES ($000)
TOTALHOURS_RESIDUAL
TOTALCOSTS_RESIDUAL
FEES_RESIDUAL
Panel B: Audit production efﬁciency
characteristics
THETA_HOURS
THETA_HOURS ≥ 1 = 1
Panel C: Audit team characteristics
ACTIVITYMIX
LABOURMIX
TENURE_DIR
TENURE_DIR ≥ 5 = 1
INDUSTRY_DIR
INDUSTRY_DIR > 10 = 1
Panel D: Auditee characteristics
ABR
ASSETS ($m)
REVENUE ($m)
Mean Median
Std.
dev.
Min.
Max.
5
95
83
231
348
380
81
170
128
663
1,042
108
109
0
0
0
48
111
229
245
51
103
81
395
593
61
61
–0.090
–0.126
–0.150
86
269
302
315
83
165
150
629
911
99
94
0.619
0.646
0.615
4
12
6
0
4
0
0
13
92
9
26
–1.56
–1.56
–1.12
311
8
283
1,211
16
878
1,296
62
1,102
1,304
57
1,132
415
12
274
746
1
585
655
1
593
2,576
145 2,210
3,535
296 3,422
398
29
361
378
27
320
1.227 –0.792 1.047
1.276 –0.882 1.151
1.493 –0.904 0.999
1.130
0.683
1.031
1
0.357
0.469
0.745 2.874
0
1
0.843
0
1.534
1
0.266
0.270
3.383
0.200
9.230
0.350
0.228
0.260
2.000
0
9.000
0
0.134
0.108
3.345
0.403
6.557
0.481
0.079 0.859
0.036 0.512
0
16.00
0
1
0
26
0
1
0.125
0.096
1
0
1
0
0.485
0.442
10.95
1
20.95
1
4.425
1,245
1,185
4.50
414
267
0.825
2,025
2,551
2.30
6
2.705
3.60 9,207 33.58
15.6 12,755 21.1
5.50
6,448
7,617
a
See Table 3 for variable descriptions and measurements. For all variables, the number of sample audit
engagements (N) is 60.
variable (THETA_HOURS ≥ 1 = 1) and Panel B contains the independent
variables. Columns (9) and (10) report the Z and t statistics and the signiﬁcance
level for the test of hypothesis of whether the distribution and mean of the
variables, respectively, for the lower-risk and higher-risk audit engagements are
(not) signiﬁcantly different; that is, p values are less (greater) than 0.05 (0.10), two
tailed. Since the distribution of the DEA efﬁciency score (theta) is usually
unknown (Cooper et al., 2006), non-parametric statistics are appropriate.
Accordingly, the rank-sum-test developed by Wilcoxon-Mann-Whitney (Z) is used
to identify signiﬁcant differences between the efﬁciency scores of the two groups.
760
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
761
© 2019 Accounting Foundation, The University of Sydney
LABOURMIX
ACTIVITYMIX
FEES_RESIDUAL
FEES ($000)
TOTALCOSTS_RESIDUAL
TOTALCOSTS ($000)
TOTALHOURS_RESIDUAL
TOTALHOURS
REVENUE ($m)
ASSETS ($m)
Panel B: Independent variables
ABR
Panel A: Dependent variable
THETA_HOURS ≥ 1 = 1
Statistics
b
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
High
Low
Low
High
Group
1
4.105
5.306
1,048
1,789
631
2,709
881
1,486
0.005
–0.013
90.397
155.705
0.004
–0.012
94
152
0.001
–0.003
0.271
0.253
0.265
0.682
0.688
Mean
2
4.350
5.300
394
588
239
0 581
540
934
–0.084
–0.246
58.990
88.153
–0.126
–0.228
57
114
–0.177
0.206
0.223
0.261
0.251
0
0
Median
3
0.717
0.272
1,741
2,647
1,131
4,307
704
1,246
0.590
0.713
77.077
135.695
0.598
0.783
84
109
0.593
0.690
0.147
0.089
0.103
0.471
0.479
Std. dev.
4
2.300
5.000
3.6
32.3
15.6
33.9
92
298
–1.557
–0.800
9.000
25.87
–1.555
–0.965
26
37
–0.908
–1.123
0.079
0.125
0.036
0
1
Min.
5
4.900
6.000
8,141
9,207
6,782
12,755
3,146
3,535
1.220
1.227
361.6
397.6
1.153
1.276
378
336
1.493
1
0.859
0.442
0.461
0
1
Max.
6
8
0
0
5
0.442
0.584
0.997
297
1.103
252.9
1.045
2,599
2,417
5,524
4.900
1
95
Percentiles
2.475
5.000
17.6
32.3
18.2
33.9
289
298
–0.781
–0.800
29.42
25.87
–0.726
–0.965
26
37
–0.790
–1.123
0.121
0.125
0.087
7
0.967
Sig. (2-tailed)
10
0.588
–0.544
(Continues)
0.649
0.458
0.984
0.033
–2.186
0.020
0.932
0.086
0.023
–2.337
0.021
–2.364
0.925
0.004
–2.969
0.094
0.213
0.000
Sig. (2-tailed)
–1.260
–6.504
t-statistic
Difference in meansd
–0.041
Z-score c
9
DESCRIPTIVE STATISTICS AND COMPARISON OF SAMPLE DATA AND VARIABLES USED IN EQUATION (1) MODELS: LOWER
AND HIGHER ABR AUDIT ENGAGEMENTSA
TABLE 5
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
© 2019 Accounting Foundation, The University of Sydney
762
High
Low
High
Low
High
Low
High
Low
High
Group
b
0.283
3.773
2.313
0.250
0.063
10.360
6.130
0.432
0.125
Mean
2
0.277
2.000
2.000
0
0
10.000
4.000
0
0
Median
3
0.124
3.735
1.537
0.438
0.250
6.609
5.464
0.501
0.342
Std. dev.
4
0.095
0
1
0
0
0
1
0
0
Min.
5
0.512
16.00
7.000
1
1
26
2
1
1
Max.
6
7
0.095
1
1
0
0
1
1
0
0
5
1
23.70
1
14.00
95
Percentiles
8
2.260
2.293
–0.544
1.511
Z-score c
9
0.028
0.026
0.588
0.136
Sig. (2-tailed)
10
See Table 3 for variable descriptions and measurements.
Group Low is where ABR is less than ﬁve (out of seven on a Likert scale), that is, the lower-risk audit engagements. Group High is where ABR is equal
to or greater than ﬁve (out of seven on a Likert scale), that is, the higher-risk audit engagements. For all variables, the number of sample audit
engagements (N) for Group Low equals 44, while N for Group High equals 16.
c
The Wilcoxon-Mann-Whitney rank-sum-test statistic (Z score) is used to test whether differences between the lower-risk audit engagements (Group
Low) and the higher-risk audit engagements (Group High) are signiﬁcant (p < 0.05, two-tailed). This is because the theoretical distribution of the efﬁciency score (theta) in DEA is usually unknown (Cooper et al., 2006). d The t-statistic is used to test whether the difference between the means of the variables for the lower-risk audit engagements (Group Low) and the higher-risk audit engagements (Group High) is signiﬁcant (p < 0.05, two-tailed). b a INDUSTRY_DIR > 10 = 1
INDUSTRY_DIR
TENURE_DIR ≥ 5 = 1
TENURE_DIR
Statistics
1
CONTINUED
TABLE 5
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
This allows for this study’s initial test of Hypothesis 1. Table 6 presents correlation
(Pearson and Spearman rho) results for variables used in the regression analyses
of Hypothesis 1.
Table 5, Panel A reports the comparison of the DEA bilateral super thetas
between the lower-risk (Group Low) and the higher-risk (Group High) audit
engagements. These super thetas are based on inter-group comparisons. That is,
the DEA algorithm calculates an audit engagement’s efﬁciency score in one group
relative to all the audit engagements in the other group. Compared to the
alternative of computing efﬁciency scores relative to all the audit engagements in
both groups, this bilateral comparison provides stronger discrimination of the
relative efﬁciency between the groups (Cooper et al., 2006).
For THETA_HOURS ≥ 1 = 1, the column (2) results show that most audit
engagements are relatively efﬁcient regardless of their ABR assessment. That is,
the proportion of the higher-risk audit engagements that are relatively efﬁcient is
68.8%, compared to 68.2% for the lower-risk audit engagements. The column
(10) results show that the distributions of the efﬁciency scores for the lower-risk
and higher-risk audit engagements are not signiﬁcantly different (p value >0.10).
That is, one group does not outperform the other. Therefore, this result supports
Hypothesis 1.6
Table 5, Panel B reports the comparison of the audit engagement characteristic
variables between the lower-risk (Group Low) and higher-risk (Group High) audit
engagements. Focusing on the variables that are signiﬁcantly different (p values
≤0.05), the results in column (10) show that the higher-risk audit engagements are
likely to be larger in terms of revenues (REVENUE). Consistent with this result,
the higher-risk audit engagements are likely to consume more total audit hours
(TOTALHOURS), incur greater total audit labour cost (TOTALCOSTS), charge
higher audit fees (FEES), and have directors with more industry experience
(INDUSTRY_DIR).
6
This result is not evident for the non-dichotomized variable THETA_HOURS (i.e., mean = 1.130),
where the distributions of the efﬁciency scores for the lower-risk (Group Low) and higher-risk
(Group High) audit engagements are not signiﬁcantly different (i.e., 1.183 and 0.986, respectively,
p = 0.037, untabulated). However, to consider the impact of outliers on this univariate result, three
approaches are taken. First, audit engagements with DEA super thetas greater than three are
treated as outliers (Banker and Chang, 2006), and subsequently eliminated from further analyses.
Based on this outlier threshold, the number of sample audit engagements (N) remains at
60 (i.e., none is eliminated), since the maximum super theta score for either group is 2.874. Second,
no audit engagement is treated as an outlier and none is eliminated. This is effectively the same as
the ﬁrst approach, since no audit engagement is treated as an outlier (i.e., N remains at 60). Third,
audit engagements with DEA super thetas greater than two are treated as outliers. Results
(untabulated) under this more restrictive outlier threshold show that N drops to 58, the mean
(median) is equal to 1.084 (1.0), and now the comparison result indicates that the distributions of
the efﬁciency scores for the lower-risk (Group Low) and higher-risk (Group High) audit
engagements are not signiﬁcantly different (i.e., 1.108 and 1.022, respectively, p value >0.10),
therefore supporting Hypothesis 1.
763
© 2019 Accounting Foundation, The University of Sydney
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
BRA AND AUDIT EFFICIENCY
© 2019 Accounting Foundation, The University of Sydney
764
.072
–.009
–.067
–.072
–.207
–.284*
.061
0.000
.005
1
.100
1
.063
.140
.125
.161
–.120
.138
–.333**
.028
(2)
–.003
.000
1
.775**
.778**
–.342**
.148
.277*
–.017
.249
(3)
–.012
.000
.790**
1
.984**
–.253
.286*
.226
–.161
.142
(4)
–.011
.000
.788**
.991**
1
–.262*
.300*
.346**
–.112
.113
(5)
–.207
.191
–.313*
–.226
–.231
1
.157
–.106
.096
–.018
(6)
–.284*
–.069
.203
.317*
.321*
.157
1
.152
–.033
–.177
(7)
.071
.135
.261*
.224
.339**
–.131
.088
1
.050
–.351
(8)
–.060
–.202
–.002
–.250
–.221
.009
–.008
.066
1
.187
(9)
.005
.049
.201
.111
.063
–.018
–.177
–.333
.200
1
(10)
See Table 3 for variable descriptions and measurements. Pearson product moment correlations are above the diagonal and Spearman rho correlations
are below the diagonal. For all variables, the number of sample audit engagements (N) equals 60.
**Correlation is signiﬁcant at the 0.01 level (two-tailed).
*Correlation is signiﬁcant at the 0.05 level (two-tailed).
a
ABR ≥ 5 = 1 (1)
InASSETS (2)
FEES_RESIDUAL (3)
TOTALHOURS_RESIDUAL (4)
TOTALCOSTS_RESIDUAL (5)
TENURE_DIR ≥ 5 = 1 (6)
INDUSTRY_DIR > 10 = 1 (7)
LABOURMIX (8)
ACTIVITYMIX (9)
THETA_HOURS ≥ 1 = 1 (10)
(1)
CORRELATION COEFFICIENTS OF VARIABLES USED IN EQUATION (1) MODELS A
TABLE 6
14676281, 2019, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/abac.12178 by -member@gla.ac.uk, Wiley Online Library on [10/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ABACUS
Multivariate Results
Table 7 reports the effect of ABR assessments on audit production efﬁciency,
estimated using models in the form of equation (1) and ordinal regression using
the alternative link function of logit. This allows for this study’s second test of
Hypothesis 1. Models (1a) to (1c) use THETA_HOURS ≥ 1 = 1 as the ordinal
dependent variable and either one of the three residual-based variables
(FEES_RESIDUAL,
TOTALHOURS_RESIDUAL,
or
TOTALCOSTS_
RESIDUAL) capturing audit effort and production characteristics. Given the
signiﬁcantly high bivariate correlations between these three residual-based
variables (see Table 6), the three models of equation (1) in Table 7 are estimated
including only one of these variables. In addition to the elimination of audit
engagements with super thetas greater than three (as per the univariate analysis;
see footnote 6), audit engagements were identiﬁed as outliers if they had
regression residuals exceeding three standard deviation…

Turn in your highest-quality paper
Get a qualified writer to help you with

“ please read the article and answer the question shortly ”

Get high-quality paper

Guarantee! All work is written by expert writers!

Still stressed from student homework?

Get quality assistance from academic writers!

Order now