You are on page 1of 28

Online 

Library: digital copies
Copyright Notice
Staff and students of the University of London are reminded that copyright subsists 
in this extract and the work from which it was taken. This Digital Copy has been 
made under the terms of a CLA licence which allows Course Users to:

‐ access and download a copy;
‐ print out a copy.

This Digital Copy and any digital or printed copy supplied under the terms of this 
Licence  are  for  use  in  connection  with  this  Course  of  Study.  They  should  NOT  be 
downloaded  or  printed  by  anyone  other  than  a  student  enrolled  on  the  named 
course.

All copies (including electronic copies) shall include this Copyright Notice and shall 
be destroyed and/or deleted if and when required by the University.

Except as provided for by copyright law, no further copying, storage or distribution 
(including by e‐mail) is permitted without the consent of the copyright holder.

The author (which term includes artists and other visual creators) has moral rights in 
the work and neither staff nor students may cause, or permit, the distortion, 
mutilation or other modification of  the work, or any other derogatory treatment of 
it, which would be prejudicial to the honour or reputation of the author.

Name of Designated Person authorising scanning:
Publishing Manager, University of London International Programmes.

Course of study Auditing and assurance AC3093


Extract title: Perceived auditor independence and audit firm fees
Title author: Holland, K. and Lane, J

Publication Year, Volume, Issue: 2012 42(2)


Page extent: 115‐141
Source title: Accounting and Business Research
ISBN/ISSN: 0001‐4788
Accounting and Business Research
Vol. 42, No. 2, June 2012, 115–141

Perceived auditor independence and audit


firm fees

KEVIN HOLLANDa∗ and JENNIFER LANEb


Downloaded by [The British Library] at 03:43 09 October 2017

a
Southampton Management School, University of Southampton, Southampton SO17 1BJ, UK;
b
School of Management and Business, Aberystwyth University, Aberystwyth SY23 3DD, UK

Regulations requiring the disclosure of fees paid to an auditor for audit and non-audit services
(NAS) respond to concerns that such payments are potentially detrimental to auditors’ actual or
perceived independence. Although empirical studies have failed to produce unequivocal
evidence of detrimental effects on auditor independence, the actions of regulators, audit
firms and companies are consistent with the belief that economic bonding generated by fees
can impair perceived levels of auditor independence.
Using a sample of UK companies over a six year period to March 2006, we study perceived
impairment of auditor independence by examining the relationship between levels of total
relative fees (combined audit and NAS fees payable by a company to its auditor as a
proportion of the audit firm’s UK income) and market value. This paper’s methodological
innovation is its use of a valuation framework in this setting. A further contribution lies in
dropping the assumption of linearity found in most prior empirical studies. We provide
evidence that shareholders perceive a threat to auditor independence only at high total
relative fee levels. At lower levels, total relative fees are positively related to company
value. These results suggest that disclosure of NAS and audit fees are of relevance to
investors, as is information about auditor income. Our results support the view that
regulation by reference to the threshold at which total relative fees are perceived negatively
is more consistent with investor preferences than prohibition of the supply of NAS by
auditors to their audit clients.
Keywords: auditor independence; audit fees; non-audit services; company valuation

1. Introduction
Regulations requiring the disclosure of fees paid to an auditor for both audit and non-audit ser-
vices (NAS) respond to concerns that such contracts are potentially detrimental to auditors’
actual or perceived independence (Young 1989). The sources of this potential detriment can be
conceptualised in two ways (Beattie and Fearnley 2002, Francis 2006). Change in the role of
the auditor may modify the auditor’s perspective from independent outsider to internal adviser
lacking scepticism. Alternatively, economic bonding between the two parties may arise from


Corresponding author. Email: k.holland@soton.ac.uk

ISSN 0001-4788 print/ISSN 2159-4260 online


# 2012 Taylor & Francis
http://dx.doi.org/10.1080/00014788.2012.628157
http://www.tandfonline.com

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


116 K. Holland and J. Lane

increasing auditor reliance on fee income received from an audited company. We explore the
second of these perspectives, and do so by examining shareholder perceptions expressed in
UK market reaction over a six-year period ending in March 2006.
Regulation relating specifically to the provision of NAS initially took the form of requiring
disclosure of the level of NAS supplied by a company’s auditors. This requirement, first intro-
duced in the USA in 1978 (Scheiner 1984), and revoked only two years later, was subsequently
implemented in the UK in 1992 (Ezzamel et al. 1996). In the environment of heightened indepen-
dence concerns after the collapse of Enron in 2002, the US Sarbanes-Oxley Act (2002)1 reintro-
duced disclosure requirements and prohibited the provision of specific auditor-supplied NAS. The
EU issued a recommendation in 2002 strongly discouraging provision of some types of NAS.2
This recommendation was followed in 2006 by a new harmonising directive. This directive
referred in its explanatory preamble to the potential threat to independence posed by such ser-
vices, but required relatively limited disclosure rather than prohibition.3 In the UK, the principles
set out by the ICAEW in 2002 were largely compatible with the EU guidance (Beattie et al. 2009)
Downloaded by [The British Library] at 03:43 09 October 2017

while the 2005 Companies Regulations substantially extended existing statutory disclosure
requirements.4Meanwhile the UK Audit Practice Board (APB) updated its Ethical Standards in
2004 to regulate provision of NAS and place a cap on the level of total audit and NAS fees
that audit firms may receive from an auditee.5 These actions demonstrate a greater regulatory
focus on NAS in the UK during this period; regulators certainly claimed strong action had
been taken (APB 2009). However, insofar as restriction on provision of NAS was enacted in
the UK, arguably it resulted from a continuation of a principles-based approach developed
from earlier disclosure requirements and professional guidance.
In the same period audit firms also behaved in a manner consistent with awareness of potential
threats to auditor independence, and hence to reputation, resulting from provision of NAS.
Following the Enron audit failure, the major audit firms publicly signalled a change in behaviour
through a series of announcements suggesting that they would separate their consulting and audit
business, e.g. PricewaterhouseCoopers (Accountancy, 2002). However, audit firms did not with-
draw from NAS provision altogether. Although revenues from NAS sources declined from 25%
in 2001 they still represented 19% of revenue at the end of our six-year period (POB 2009).
Whilst the above audit firm and regulatory actions are consistent with a relationship between
NAS and actual or perceived impairment of auditor independence, no such consensus exists in the
academic literature (Schneider et al. 2006). This is true of both the literature focused on actual
independence, proxied mainly by means of earnings quality, and that focused on perceived inde-
pendence, where the perceptions of investors, lenders and others have been examined by means of
a range of methodologies. In part this lack of consensus is due to choice of research design,
especially in modelling economic bonding. For instance, while much research has focused on
the relationship between company NAS fees and audit fees (e.g. Frankel et al. 2002, Davis
and Hollie 2008) others (e.g. Ashbaugh et al. 2003, Larcker and Richardson 2004, Khurana
and Raman 2006) have employed both total fees and relative fee measures. We examine the argu-
ment that both audit and NAS fees have the potential to create economic dependence on the part
of the auditor, that potential being greater where such fees are economically significant to the
auditor.6 Hence fuller modelling of economic bonding may be achieved by relating these fees
to a measure of the fee income of the audit firm. The only such measure reliably available to inves-
tors in the UK during the period studied is total audit firm fee income. Thus, we define total rela-
tive fee as the combined audit and NAS fees payable by a company to its auditor, as a proportion
of the audit firm’s UK income.
This study makes three contributions to the literature examining perceived auditor indepen-
dence. The main methodological contribution is the use of a valuation relevance approach that
is new to this setting, which we discuss in more detail in section three. In summary, the inherent

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Accounting and Business Research 117

limitation of the event study approach widely used in prior literature is the identification of the
unexpected level of NAS, and the need to isolate the effect of NAS announcements from other
contemporaneously announced information. The valuation model employed in this paper also
avoids problematic aspects of valuation-related ex ante cost of equity capital models, which
owing to reliance upon analyst forecasts of expected earnings, are particularly marked in the
UK market. We study impairment primarily by examining the relationship between total relative
fees and company market values.
A second contribution is to test the possibility that any relationship between total relative fees
and perceived auditor independence may be non-linear. Although non-linear relationships have
been found in related topics, such as auditor tenure (Boone et al. 2008) prior empirical studies
of this area have assumed linearity (Francis 2006). Simunic (1984) suggests that management
trade off the benefits of increased levels of audit and NAS against the reduction in company
value that results from shareholders’ demand for compensation for the risk of loss of auditor inde-
pendence. This trade-off implies a non-linear relationship between auditor fees and market value,
Downloaded by [The British Library] at 03:43 09 October 2017

and leads us to expect the benefits of audit and NAS to increase with modest increases in levels of
provision, but to be outweighed at higher levels by impairment concerns. In dropping the assump-
tion of linearity we build on the contribution of Francis (2006) who speculates that impairment
may occur only at high levels of NAS.
Thirdly, the specification of our sample frame, the six-year period ending 2006, allows the
stability of any relationship to be analysed over a relatively long period of time. This period is
timed to capture reaction on the part of investors to a stream of international audit scandals,
and specifically permits analysis both immediately prior to and after the collapse of Enron and
WorldCom. Assuming that value-enhancing aspects of audit and NAS did not change during
this period, any change in the relationship between total relative fees and company valuation
could be interpreted as indicating changing investor concerns over auditors’ economic depen-
dence on their audit clients.
Furthermore, our sample allows us to study the perceptions of investors in the UK market.
Much of the prior literature discussed above is based on US data, yet it cannot be assumed
that research findings translate identically to different institutional or regulatory settings
(Francis 2006, Gwilliam 1987). Over the period of our study, the changes in UK disclosure
requirements were more evolutionary than the fundamental shift seen in US requirements. The
corporate governance framework, revised in 2003 (FRC 2003), also retained the relatively
flexible, principles-led approach developed over the preceding decade.
Prior to our sample period, audit firms in the UK had transformed themselves into multi-
disciplinary professional service organisations, so much so that the mean NAS to audit fee
ratio among FTSE 100 companies peaked at 300% in 2001 (Beattie et al. 2009). This sub-
sequently declined steadily to below 90% at the end of our sample period in 2006. By examining
UK data over this time frame, this paper contributes to addressing the shortcomings of the existing
research base available to policy makers, especially with regard to expanding the research base
relevant to the UK setting (APB 2009). As a result, it informs UK policy setting and contributes
to wider understanding of the extent of variation in perception across institutional and regulatory
settings.
Our results can be summarised as follows. In general, only at relatively high levels of total
relative fees do adverse effects on investors’ perception of independence outweigh beneficial
valuation effects. Specifically, we find a concave relationship between total relative fees and
market values, which is observed over the six-year period as a whole and appears to be driven
by relative NAS fees. These results remain robust to the exclusion of Andersen-audited compa-
nies and to alternative model specifications. However, the relationship is not consistently
observed when each year is examined separately. As a result we find no evidence that investor

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


118 K. Holland and J. Lane

perception was driven by the Enron collapse, and no pattern that can unequivocally be attributed
to specific regulatory change. Overall, our results suggest a sophisticated response on the part of
shareholders, which we interpret as lending support to regulatory demands for disclosure of fee
levels, without which shareholders would be unable to evaluate the degree of risk and hence
would be poorly protected.
The remainder of the paper is organised as follows: section 2 summarises the relevant auditor
independence literature; section 3 explains our methodology with reference to relevant literature
and describes our data sources; section 4 both reports our initial results and discusses our sub-
sequent robustness and sensitivity analyses. In section 5 we conclude with a summary of our find-
ings and the policy implications of the study.

2. Prior literature
Downloaded by [The British Library] at 03:43 09 October 2017

2.1. Auditor independence


As a key attribute of audit quality, auditor independence is essential to the generation of confi-
dence in information reported in financial statements. Low quality or lack of confidence in
reported information can be expected to undermine its value and affect investors’ perceptions
of riskiness of assets. It has long been appreciated that there are risks to independence inherent
in the auditor’s role (Firth 2002, Francis 2006), an agency issue highlighted by corporate govern-
ance guidelines such as the Combined Code (FRC 2003). Meanwhile, a substantial body of lit-
erature addresses the potential for impairment of independence as a result of joint provision of
audit and NAS. The causes of this potential detriment can be conceptualised in two ways:
either in relation to change in the role of the auditor, modifying the auditor’s perspective from
independent outsider to internal adviser, or in terms of economic bonding between the two
parties arising from increasing reliance on fee income (Beattie and Fearnley 2002, Francis
2006). In either case prior literature examines both actual and perceived independence. Studies
of actual independence examine proxies for specific aspects of the audit process, seeking evidence
of impairment of auditor independence. Overall, empirical studies of this topic have failed to
produce unequivocal evidence of any detrimental effects of NAS on auditor independence, but
equally have failed to rule out such an effect (Schneider et al. 2006, Francis 2006).

2.2. Perceived independence


Prior research has addressed perceptions of independence through a wide range of research meth-
odologies, encompassing qualitative and experimental research as well as market-based studies.
Much of the qualitative and experimental research pre-dates the expansion of NAS provision and
audit firm concentration of the 1990s (Francis 2006). However, subsequent studies suggest con-
sistent independence concerns over high NAS among US loan officers (Lowe and Pany 1995,
Bartlett 1997) and their UK equivalents (Beattie et al. 1999). Recent qualitative research suggests
that independence concerns arising from NAS provision, and from economic dependence in
general, were common among UK investors in the period that we study (Dart, 2011).7 Meanwhile
experimental studies support these findings of independence concerns among users (Dopuch et al.
2003, Davis and Hollie 2008). Experimental research also suggests that disclosure of NAS
reduces the accuracy of investor perceptions of auditor independence, leading investors in an
experimental market to perceive impairment where none exists (Davis and Hollie 2008).
If investor perceptions are indeed inconsistent with evidence of the degree of actual impair-
ment, as Davis and Hollie (2008) suggest, there is no reason to expect perception studies of
natural markets to reflect the inconclusive findings of studies of actual independence. Market-

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Accounting and Business Research 119

based perception studies make use of a variety of approaches. Frankel et al. (2002) and Ashbaugh
et al. (2003) use an event study methodology to study NAS announcements. The two papers
report conflicting results, with Ashbaugh et al. (2003) concluding that the negative market reac-
tion documented by the former is sensitive to research design. However, studies examining earn-
ings response coefficients have been more consistent, finding lower coefficients for companies
with higher levels of NAS fee payments (Krishnan et al. 2005, Francis and Ke 2006). A
similar picture emerges from studies, using US data, of the relationship between auditor payments
and ex ante cost of capital, showing a negative association between NAS and both ex ante cost of
debt (Dhaliwal et al. 2008, Amir et al. 2010) and ex ante cost of equity capital (Khurana and
Raman 2006). The latter found a similarly negative association using both NAS and total pay-
ments to auditors. Such findings are consistent with perception on the part of investors that
auditor independence may be impaired by joint provision of audit and NAS.
Overall, while academic evidence regarding impairment of actual independence remains
mixed, there is more evidence pointing to perception among investors of potential impairment
Downloaded by [The British Library] at 03:43 09 October 2017

of auditor independence. However, this pattern is established largely on the basis of US data.
By contrast the UK context is not extensively explored in prior literature, particularly with
respect to market-based empirical research. There is no reason to assume that research findings
translate identically to different institutional or regulatory settings (Francis 2006, Gwilliam
1987). Dart’s (2011) qualitative study suggests that while independence concerns are common
among UK investors, they are not universal, and the strength and impact of those concerns
over time remains unknown.

3. Hypothesis development and research design


Francis (2006) observes that although overall the literature suggests that NAS can adversely affect
the appearance of auditor independence, there has been a pronounced tendency to examine NAS
in isolation from their broader contracting environment, and in particular from audit fees. Francis
also speculates that impairment may occur only at high levels of NAS. We address each of these
issues in more detail below.

3.1. Measures of economic bonding


Exploration of the perceived impact of economic bonding requires examination of the compo-
sition of the fees on which auditors may become dependent. It has long been argued that as
NAS fees may generate both knowledge spillovers (Simunic 1984) and economic rents for the
auditor, they are a potential source of economic bonding. Indeed, much of the audit independence
literature focuses on NAS as the sole or primary source of detriment (Francis 2006). Srinidhi and
Gul (2007) argue that the dual effects of higher regulation and competition in the audit market
reduce the scope for rents in audit fees. Similarly, audit fees may be seen as loss-leaders, with
the consequence that rents must be attached to the NAS fees that the audit attracts. By contrast,
Francis (2006) argues that NAS are not unique in their scope for generating dependence: all
auditor fees pose a risk of dependence owing to the inherent contracting environment of the
audit. Hence the risk to audit independence arising from audit fees alone has long been a
concern of the profession. Francis particularly advocates examination of total fee dependence
on the client, as a contribution to fuller examination of the institutional context and incentives
linked to the audit contracting environment. This view is echoed by Khurana and Raman
(2006) who argue that it may now be more appropriate to focus on total fees rather than NAS
alone. Their argument is in part supported by anecdotal evidence of rising US audit fees
post-SOX, which calls into question the assumption that audit fees carry no rents. In the

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


120 K. Holland and J. Lane

absence of certain knowledge of the price elasticity of demand for audit, investors may interpret
higher audit fees as generating economic rents. Alternatively, the same pattern of higher audit fees
may be perceived as reflecting more effort and thus higher quality audit, consistent with auditors’
concern to protect reputational capital (Reynolds and Francis 2001). Ghosh and Pawlewicz (2009)
suggest this may be true of the changing US pattern. They interpret increased mean levels of audit
fees paid by US companies as the consequence of increased audit work required to comply with
the Sarbanes-Oxley Act and increased litigation risk. Falling levels of NAS are interpreted as a
direct consequence of the restrictions imposed by the Act.
The changes in regulation in the UK in the period studied are less abrupt than those imposed in
the US. These are characterised by a sequence of regulatory actions over a four-year period,
suggesting that differing patterns of perception change are feasible. Interpretation of changing
audit fee patterns can also be expected to be more complex towards the end of our sample
period, owing to the potential impact on audit effort of the move to IFRS reporting from
January 2005. It may be that uncertainty on the part of investors about how to interpret com-
Downloaded by [The British Library] at 03:43 09 October 2017

ponents of auditor fees increases their reliance on total fees. Consequently we test in this paper
the argument that in the contracting environment that exists between a company and its
auditor, all fees have the potential to create economic dependence on the part of the auditor.
Thus, we base our primary measure on total fees. However, given the conflicting views explored
here, we also examine audit and NAS fees separately.
Whichever fee basis is used, the potential for economic bonding will be greater where the rel-
evant fees are economically significant to the auditor, yet the commonly used ratio of NAS to total
fees does not reflect the effects of scale. To capture these effects of scale, our primary measure
comprises the total annual fee payments made by a company to its auditors divided by the
annual UK income of the audit firm (Ashbaugh et al. 2003, Larcker and Richardson 2004,
Khurana and Raman 2006) which we refer to as the total relative fee. In the partnership structure
of the auditor, location-specific (office or city level) data would better capture the potential econ-
omic importance to the auditor, since it would allow for differing office size. This measure is used
in some US studies (e.g. Khurana and Raman 2006), but cannot be constructed from contempora-
neous publicly available UK data. For this reason we measure total audit firm income at the
national (UK) level.

3.2. Non-linearity
Francis (2006) notes that prior research has tended to rely on linear models and speculates that
impairment may take place only at high fee levels. In relying on linear models, researchers
have focused on the potential for negative spillovers between NAS and audit, in which the audi-
tor’s independence may be impaired by economic bonding. However, interaction between nega-
tive and positive spillovers cannot be addressed in this way, nor can any potential bonding effect
arising from variation in audit fees. Simunic (1984) draws attention to the possibility of positive
spillovers between NAS and audit. Provision of NAS may be perceived as improving the auditor’s
knowledge of the company, thus generating improved quality or cost benefits. Simunic suggests
that the process by which companies determine the levels of audit and NAS to purchase involves a
trade-off. Management consider the benefits of increased levels of audit and NAS against a
reduction in company value resulting from shareholder demands for compensation for loss of
auditor independence. This implies a non-linear relationship.
Despite the widespread focus on linear modelling, non-linear relationships have been
suggested by a few studies of an experimental nature and in parallel fields of enquiry. Davis
and Hollie (2008) find a non-linear (stepped) association in an experimental market characterised
by stepped proportions of NAS and restricted information. This has not been replicated in

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Accounting and Business Research 121

empirical studies of natural markets using similar NAS measures (that is, NAS levels or as a NAS
ratio). In a study of smaller UK private companies, Dedman et al. (2009) document a positive
relationship between NAS and voluntary audit, which they interpret as consistent with positive
spillover effects, eventually becoming negative at very high NAS levels. This non-linear relation-
ship relates to management rather than investor perceptions. Looking at another aspect of audit
quality, Boone et al. (2008) find a non-linear relationship between ex ante equity risk premium
and audit firm tenure, consistent with investor concerns over impaired independence eventually
dominating the positive effects of auditor learning.
We expect the relationship with total relative fees to be complex and characterised by trade-
offs. Modest increases in relative fees could be associated with higher market values if the
increases are interpreted as representing ‘more audit’, i.e. achieving a higher level of assurance,
with consequent reduction in internal agency costs. Likewise, the potential for positive spillover
effects means that some benefits may be expected from modestly increasing quantities of NAS
expertise (Simunic 1984).8 We expect concerns over auditor impairment resulting from economic
Downloaded by [The British Library] at 03:43 09 October 2017

bonding to be closely related to materiality (represented by relative fee income) such that at low
levels it may be negligible or outweighed by the positive effects just described. However, at
higher levels of relative fee income, impairment concerns can be expected to dominate the posi-
tive audit fee and NAS fee effects (Francis 2006). Consistent with the above discussion on the role
of economic bonding we predict a concave relationship between total relative fee levels and
market value. We formalise this in our first hypothesis as follows:

Hypothesis 1:. There is a concave relationship between market value of equity and total relative fees.
Beyond the implied turning point the positive benefits associated with increased audit fees and non-
audit services are outweighed by concerns over impairment of auditor independence. Prior to the
turning point the former effect dominates the latter.

Our discussion above notes conflicting views expressed in recent literature as to the respective
roles of audit and NAS fees in economic bonding, and for this reason we also examine each sep-
arately. Hence we express our second hypothesis as follows:

Hypothesis 2:. The relationship between market value of equity and relative audit fees differs from that
between market value of equity and relative non-audit services fees.

It is possible that specific external events may affect investors’ perceptions sufficiently to lead to
change over the six-year period. If, for example, the Enron collapse was significant in heightening
investors’ concerns we would expect to see a related change in perception in the second year of
our sample, 2001/02. Regulatory response perceived as a significant change in approach or
robustness could have an impact on investor confidence in the final years of the sample period.
While we argue that UK regulation was more evolutionary than the comparable US responses,
this may not have been apparent to investors at the time. In general, therefore, we hypothesise
that:

Hypothesis 3:. The relationship between market value of equity and relative fees varies over the six-
year sample period.

3.3. Choice of model


The relationship between fees paid by a company and its market value is examined using a stan-
dard valuation model. This approach avoids a number of problematic issues in using event study

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


122 K. Holland and J. Lane

methodology, for instance, identifying relevant announcement dates, determining existing expec-
tations as to the level of fees and controlling for contemporaneously announced information. It
also avoids the more problematic aspects of valuation-related exante cost of equity capital
models that rely upon analyst forecasts of expected earnings. While the argument in favour of
devising ex ante (rather than ex post) measures has considerable merit, we have a number of con-
cerns about the suitability of extant methods for this study, which we discuss below.
In examining investor perceptions of auditor independence, Khurana and Raman (2006)
employ analyst forecasts as a proxy for market expectations. They then derive the cost of
equity implied by the current share price after making an assumption about the sustainability
of the forecast growth of earnings per share (EPS). This procedure requires the forecast EPSt1
to be positive, and imposes a growth requirement, EPSt2 . EPSt1, that risks severe sample
bias. This risk is exacerbated by the exclusion of companies not followed by an analyst or
where a forecast is not made, which is more common in the UK than the US. In particular this
can occur with respect to small capitalisation stocks (Collett 2004). If these smaller companies
Downloaded by [The British Library] at 03:43 09 October 2017

are (as intuitively expected) associated with lower relative fees, exclusion could systematically
overstate the perception of impairment.
More fundamentally, we are also sceptical that the main condition can be met: the role of the
forecast is to proxy for investors’ expectations, but its validity for this purpose is not directly tes-
table. There is substantial evidence that UK analyst forecasts are biased (Capstaff et al. 2001,
Hodgkinson 2001) and have little predictive value when made more than twelve months
before the year end (Capstaff et al. 2001). This limitation is exacerbated by the significant
number of companies for which only a single analyst forecast is available (Collett 2004),
thereby losing the benefit of a consensus forecast with its potentially desirable netting out prop-
erties. Hence we concur with the conclusion of Easton and Sommers (2007, p. 1013) with respect
to bias in analysts’ forecasts, that ‘extant measures of implied expected rate of return should be
used with considerable caution’.9
A standard valuation model avoids a number of problematic issues found in competing
methods as discussed above. We employ the model used by O’Hanlon and Taylor (2007),
which is similar to Horton (2008), with the addition of a control variable for research and devel-
opment expenditure to avoid a potentially omitted variable bias. Green et al. (1996) document
expenditure on research and development as being value relevant. As levels of research and devel-
opment and either audit or NAS fees could be related, it is a potentially omitted variable. Accord-
ingly, we include the annual level of research and development expenditure as a separate
independent variable in the model. The basic model we employ (Model 1) is as follows:

MVEi,t BVi,t NIi,t +RDi,t NIi,t∗ LOSSSDUMi,t RDi,t


= a + b1 + b2 + b3 + b4
BVi,t−1 BVi,t−1 BVi,t−1 BVi,t−1 BVi,t−1

15 
20
+ b5 TRELFEEi,t + b6 TRELFEE2i,t + bi INDDUMSi,t + bi YEARDUMSi +1i (1)
i−7 i−16

where for company i at balance sheet t:

MVE ¼ the market value of equity measured six months after the
balance sheet date;
BV ¼ the book value of equity;
NI ¼ net income after tax available for equity holders;
RD ¼ research and development expenditure written off in the year;

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Accounting and Business Research 123

LOSSDUM ¼ coded 1(0) if company reports a loss(profit) for the period;


TRELFEE ¼ the total relative fee defined as the sum of audit and NAS fees
payable by a company to its auditor, divided by annual total
fee income of the audit firm involved;
TRELFEE2 ¼ square of TRELFEE designed to capture a non-linear relation-
ship between market value and total relative fee.

Consistent with hypothesis 1 we expect a positive coefficient for TRELFEE and a negative coef-
ficient for TRELFEE2.
We also include standard year and industry dummies (YEARDUMS and INDDUMS)
(O’Hanlon and Taylor 2007). We deflate the dependent and the independent control variables
by opening book value in order to control for potential scale effects (Barth and Kallapur
1996). We subsequently report robustness tests of alternative specifications.10
Downloaded by [The British Library] at 03:43 09 October 2017

In testing hypothesis 2 we amend Model 1 above by replacing the variables TRELFEE and
TRELFEE2 with two sets of variables to allow differences in the relationship between market
value and the form of the service provided i.e., audit and NAS. Specifically, we include as vari-
ables: RELAF, the relative audit fee defined as the company’s audit fee divided by the audit firm’s
annual total fee income; RELNAS, the relative NAS fee defined as the company’s NAS fee
divided by the audit firm’s annual total fee income; and their respective quadratic terms
RELAF2 and RELNAS2, defined as RELAF∗ RELAF and RELNAS∗ RELNAS respectively. Con-
sistent with hypothesis 2 we would expect significant differences between coefficient estimates
of: RELAF and RELNAS; RELAF2 and RELNAS2; and between RELAF + RELAF2 and
RELNAS + RELNAS2. Model 2 is as follows:

MVEi,t BVi,t NIi,t + RDi,t NIi,t∗ LOSSSDUMi,t RDi,t


= a + b1 + b2 + b3 + b4
BVi,t−1 BVi,t−1 BVi,t−1 BVi,t−1 BVi,t−1
+ b5 RELAFi,t + b6 RELAF2i,t + b7 RELNASi,t + b8 RELNAS2i,t

17 
22
+ bi INDDUMSi + bi YEARDUMSi + 1i (2)
i=9 i=18

3.4. Sample
The sample is drawn from those companies included in the Financial Times - (London) Stock
Exchange (FTSE) Industrials Index 1999– 2006. For inclusion a company must satisfy the fol-
lowing criteria: (1) have been listed throughout the period; (2) have a balance sheet date of
either 31 December or 31 March that remains materially unchanged throughout the sample
period;11 (3) have been audited throughout the period by a Big 4(5) audit firm and (4) have a posi-
tive net book value at each balance sheet date. These conditions provide sufficient homogeneity to
allow us to consider whether investor attitudes changed during the six-year period. The compo-
sition of the sample is set out in panel A of Table 1, along with a description of how the year ends
are summarised by year in panel B.
With the exception of NAS fee data, which was collected from FAME database (Bureau van
Dijk), all company-specific data was obtained via Datastream (Thomson Reuters).12 Audit firm
total fee data was collected from the annual Accountancy magazine annual fee income table
(Accountancy, various years).13

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


124 K. Holland and J. Lane

Table 1. Composition of sample and classification of year ends.


Panel A Composition of sample
Number of FT-SE Industrials 580
Less: Financials (excluding property 57
development)
Non-Financials 523
Analysis of year ends
No %
31 March year end 115 22.0% 115
31 December year end 232 44.3% 232
Other year ends 176 33.7%
Total 523 100%
Mar Dec 347
Missing data (43)
Downloaded by [The British Library] at 03:43 09 October 2017

Change of year end (45)


Non-Big 5 auditor (35)
Useable sample 224
Number of year ends (6 ∗ 224) 1,344
Negative book value (71)
Useable company year ends 1,273
Outliers (1% trimming of distribution (116)
by year and variable)
Sample 1,157
Panel B Year designation and year ends
Year Year end (MV measured + 6 months) Pre/post Enron
2000/01 31 December 2000 and 31 March 2001 Pre Enron
2001/02 31 December 2001 and 31 March 2002 Post Enron
2002/03 31 December 2002 and 31 March 2003
2003/04 31 December 2003 and 31 March 2004
2004/05 31 December 2004 and 31 March 2005
2005/06 31 December 2005 and 31 March 2006

Arguably, the relationship between market value and fee income may be different for newly
listed companies with a shorter trading record, and for companies that lose their quoted status
through either ‘failure’ or take over. Imposing the first filter increases the homogeneity of the
sample year on year.14 Minimising the variation in balance sheet date within each year
reduces the potential bias that could arise from changes in market perceptions over a longer
time interval. December and March year ends are the most frequently occurring and therefore
provide an adequate sample size: see Table 1 and Barron (1986). The period under review was
characterised by heightened independence concerns following the collapse of Enron. An
additional advantage of the December and March year-end specification is that it allows a
specific analysis to be performed on both a pre- and post- Enron sub-sample. The Big 4(5)
auditor requirement isolates the results from the effects of variations in audit firm quality
that have been documented in earlier studies of the UK audit market, e.g. Chan et al. (1993)
and Holland and Horton (1993). As a final filter, ‘outliers’ are identified and deleted on an
annual basis from each of the dependent and continuous independent variables using a cut-
off value of 1% at each end of the distribution to identify extreme observations. Table 1
(panel A) shows how the final sample of 1,157 company year ends was achieved using the
various filters.

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Accounting and Business Research 125

4. Results
4.1. Summary and descriptive statistics
Variable definitions are given in Table 2 while summary statistics and correlations are given in
Tables 3 and 4. The sample company year ends represent a wide cross-section of quoted compa-
nies in terms of market capitalisation, Table 3 (panel A), with an average market capitalisation of
approximately £1219 million and ranging from a minimum value of £2.77 million to a maximum
of £33,421 million.15 Total fee payments to auditors for all services range from £11,000 to £32.5
million, with a mean of approximately £1.37 million. The mean and maximum total relative fees
represent 0.1% and 2.7% respectively of the averaged auditors’ total fee.16
The descriptive statistics in Table 3 (panel B) suggest a pattern of change in fee components in
the sample. Mean audit fees show a fairly consistent and dispersed increase over the six-year
period, while mean NAS fees show consistent decline over the first five years.17 This pattern is
consistent with the declining UK NAS ratio noted by Abidin et al. (2010) for the same period,
Downloaded by [The British Library] at 03:43 09 October 2017

and with the changing composition observed in the US following the introduction of the
Sarbanes-Oxley Act (Ghosh and Pawlewicz 2009). While the increase in mean audit fees in
2004/05 and 2005/06 might be attributable to increased audit effort associated with the
demands of transition to IFRS reporting, this is not a convincing explanation for increases in
earlier years. We analyse the annual fee pattern further in section 4.5.
The correlation matrix in Table 4 indicates significant positive correlations between the
dependent variable and, in turn, each of the control variables. The use of a quadratic term
could lead to an unacceptably high level of multicollinearity. Indeed there is a high significant
correlation between the variables TRELFEE and TRELFEE2. Although its value of 0.855 is
below 0.9, the level which can indicate serious levels of multicollinearity (Hair et al. 1998),
we examine the issue of multicollinearity in more detail in sections 4.2 and 4.5 below. In
summary, the analysis indicates that the data set does not exhibit excessive levels of multicolli-
nearity and that the coefficient estimates and reported standard errors are reliable in this respect.

4.2. Model 1
Table 5 reports the results of several estimations. Column 1 is based on Model 1 above which is
referred to as the ‘full model’ to distinguish it from the ‘restricted model’ discussed below, in

Table 2. Variable definitions.


Variable abbreviation Description
MVE Market value of equity six months after year end
BV Book value of equity
NI Net profit after tax attributable to ordinary shareholders
RD Research and development expenditure expensed
LOSS Dummy variable coded 1 ¼ if NI , 0 otherwise ¼ 0
SALES Sales for period
AF Audit fee paid by company
NAS Non-audit service fees paid by company
TRELFEE Total relative fees paid by company ¼ (AF + NAS)/AFFEE
where AFFEE ¼ audit firm annual total fee income
TRELFEE2 TRELFEE ∗ TRELFEE
RELAF Relative audit fee paid by company ¼ AF/AFFEE
RELAF2 RELAF ∗ RELAF
RELNAS Relative non- audit service fee paid by company ¼ RELNAS/AFFEE
RELNAS2 RELNAS ∗ RELNAS

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


126
Table 3. Descriptive statistics.
Panel A
£000s Mean Min Max Standard deviation
Downloaded by [The British Library] at 03:43 09 October 2017

MVE 1,219,421.0 2,770.0 33,420,500.0 3,141,555.0


BV 483,324.1 1,209.0 15,000,000.0 1,272,148.0
NI 52,894.5 26,470,000.0 2,793,000.0 346,515.7
AF + NAS 1,367.3 11.3 32,487.0 2,701.7
TRELFEEt 0.0012 0.0001 0.0286 0.0022
Deflated variables:
MVEt/BVt21 3.1837 0.1839 32.8011 3.0570

K. Holland and J. Lane


BVt/BVt21 1.1212 0.1242 9.7943 0.5882
NIt/BVt21 0.1031 21.0241 1.6251 0.2223
(n ¼ 1157)
Where MVE ¼ market value of equity six months after year end; BV ¼ book value equity; NI ¼ profit after tax
attributable to ordinary shareholders; (AF) ¼ audit fee paid by company; NAS¼ non-audit services fee paid by
company; and TRELFEE (total relative fee) ¼ (AF + NAS)/AFFEE where AFFEE ¼ audit firm annual total
fee income.

Panel B
Fee components by year: £000s
AF NAS AF + NAS
Year Mean Standard deviation Mean Standard deviation Mean Standard deviation SNAS/SAF
2000/01 509.00 851.11 914.02 2,247.78 1,423.02 2,822.68 1.796
2001/02 491.84 799.69 836.22 2,057.81 1,328.07 2,676.24 1.700
2002/03 636.22 1,658.86 741.08 1,969.93 1,377.30 3,242.95 1.165
2003/04 641.82 1,162.96 605.22 1,377.96 1,247.05 2,438.89 0.943
2004/05 752.10 1,388.92 591.64 1,024.51 1,343.74 2,310.55 0.787
2005/06 769.30 1,335.29 722.56 1,494.05 1,491.87 2,632.98 0.939

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Downloaded by [The British Library] at 03:43 09 October 2017

Table 4. Pearson correlation matrix.


(n ¼ 1,157) MVEt/BVt21 BVt/BVt21 (NIt + RDt)/BVt21 RDt/BVt21 TRELFEEt TRELFEEt2 RELAFt2 RELNASt RELNASt2

Accounting and Business Research


RELAFt
MVEt/BVt21 1.000 0.479∗∗∗ 0.548∗∗∗ 0.4024∗∗∗ 0.070∗∗ 0.021 0.024 20.021 0.094∗∗∗ 0.068∗∗
BVt/BVt21 1.000 0.334∗∗∗ 0.155∗∗∗ 20.008 20.009 20.018 20.012 0.002 0.005
(NIt + RDt)/BVt21 1.000 0.425∗∗∗ 0.093∗∗∗ 0.081∗∗∗ 0.68∗∗ 0.012 0.094∗∗∗ 0.11∗∗∗
RDt/BVt21 1.00 20.002 0.027 20.022 20.010 0.016 0.054
TRELFEEt 1.00 0.855∗∗∗ 0.861∗∗∗ 0.537∗∗∗ 0.900∗∗∗ 0.729∗∗∗
TRELFEEt2 1.000 0.829∗∗∗ 0.786∗∗∗ 0.690∗∗∗ 0.704∗∗∗
RELAFt 1.000 0.786∗∗∗ 0.552∗∗∗ 0.372∗∗∗
RELAFt2 1.000 0.205∗∗∗ 0.139∗∗∗
RELNASt 1.000 0.875∗∗∗
RELNASt2 1.000
Notes: (a) Variable definitions as follows: MVE ¼ market value of equity; BV ¼ book value of equity; NI¼ profit after tax attributable to ordinary shareholders; RD ¼ current year
expensed research and development expenditure; TRELFEE (total relative fee) ¼ (AF + NAS)/AFFEE where AF ¼ audit fee paid by company, NAS¼ non-audit services fee paid
by company and AFFEE ¼ audit firm annual total fee income; TRELFEE2 ¼ TRELFEE∗ TRELFEE; RELAF (relative audit fee) ¼ AF/AFFEE; RELNAS (relative non audit
services fees) ¼ NAS/AFFEE; RELRAF2 ¼ RELAF∗ RELAF and RELNAS2 ¼ RELNAS∗ RELNAS. (b) ∗ , ∗∗ , ∗∗∗ indicate significance at the 5%, 2.5% and 1% level respectively.

127

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


128 K. Holland and J. Lane

which we omit the quadratic term. The two fee variables TRELFEE and TRELFEE2 are of the
expected signs and both are significant at the 2.5% level, with control variables also exhibiting
their expected signs. This is a necessary condition for the presence of a concave relationship.
We also confirm, using the U-test (Lind and Mehlum 2010), that a further condition is satisfied
i.e. that the relationship between market value and TRELFEE is increasing (decreasing) at low
(high) values of TRELFEE.18
The relative values of the fee variable coefficients imply that independence concerns outweigh
the beneficial effects of audit and NAS only when a company’s total relative fee exceeds 0.9% of
the auditor’s fee income (Table 5). In monetary terms this turning point occurs when total fees are
in the region of £10.7 million.19 As an economically ‘reasonable’ figure this supports the validity
of the model and estimation process̄.
The value of TRELFEE at the turning point falls within the 98% percentile of its distribution.
As an informal test of the robustness of the concave relationship we exclude the 21 observations
with values of TRELFEE in excess of the turning point. In the re-estimated model a concave
Downloaded by [The British Library] at 03:43 09 October 2017

relationship is again present although with a different, lower, turning point reflecting the
revised sample composition. This process can be repeated once more before the concave relation-
ship becomes insignificant and is replaced by a significant positive linear relationship. The results
of these informal tests are consistent with the formal U-test reported above.
Existence of the two compensating effects is confirmed when the above model is estimated
after omitting the quadratic term. In this restricted model the single remaining fee variable
TRELFEE is not statistically significant at the 5% level; this restricted model is reported in
column 2 of Table 5.
The resulting reduction in the adjusted R2 is statistically significant at 5% level,20 and indi-
cates that the addition of the quadratic term increases the explanatory power of the model.
Further evidence that inclusion of the quadratic term does not induce an unacceptable level of
multicollinearity is obtained from an analysis of condition indices (Belsey et al. 1980)21 and
inspection of VIF statistics.22

4.3. Robustness tests


4.3.1. Impact of controlling for scale effects
The above results are based on book value-deflated models and are insensitive to the exclusion of
‘influential observations’, i.e. those with a critical DFITS statistic (Belsey et al. 1980).23 Conse-
quently, the full sample of observations has been included. Although the overall inferences are
retained, the results are sensitive to the treatment of influential observations when alternative
bases for controlling for scale effects have been used. Model 1, when deflated by number of
shares, results in positive and negative coefficients for TRELFEE and TRELFEE2 respectively
which are significant at the 10% and 2.5% levels when influential observations are included.
When these observations are excluded the coefficient signs remain the same and the significance
level in both cases is at the 1% level. When deflated by sales, the coefficients remain positive and
negative as above but only show reasonable levels of significance (at 10% and 5% respectively)
after the exclusion of influential observations. An alternative approach to controlling for size
effects is to estimate the model undeflated with the inclusion of a size control (Barth and Kallapur
1996). This approach, reported in Table 5 (column 5), results in positive and negative coefficients
for TRELFEE and TRELFEE2, significant at the 1% and 2.5% levels respectively.
Audit fee studies have consistently found that company size is the most significant determi-
nant of audit and NAS fees (Hay et al. 2006). Even though we deflate the dependent and the
control variables to remove any size effect, if the process is imperfect there is a potential issue

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Table 5. Model 1 regressions.
1 2 3 4 5
Variables Model 1 Model 1 Model 1 + ‘book value’ Model 1 + ‘sales size’ Undeflated Model 1
Downloaded by [The British Library] at 03:43 09 October 2017

full restricted size control control variables undeflated


BVt/BVt21 1.621 1.620 1.649 1.612 BVt1 0.001
4.56 ∗ ∗ ∗ 4.57 ∗ ∗ ∗ 4.62 ∗ ∗ ∗ 4.50 ∗ ∗ ∗ 3.93 ∗ ∗ ∗
(Nit + RDt)/BVt21 5.460 5.492 5.375 5.569 (Nit + RDt) 0.006
6.08 ∗ ∗ ∗ 6.09 ∗ ∗ ∗ 6.06 ∗ ∗ ∗ 6.06 ∗ ∗ ∗ 5.07 ∗ ∗ ∗
(Nit + RDt)∗ LOSSDUM 25.914 26.044 25.605 25.933 (Nit + RDt)∗

Accounting and Business Research


0.005
25.14 ∗ ∗ ∗ 25.23 ∗ ∗ ∗ 24.92 ∗ ∗ ∗ 25.20 ∗ ∗ ∗ LOSS 24.54 ∗ ∗ ∗
RDt/BVt21 6.219 6.198 5.967 6.133 RDt 0.002
4.44 ∗ ∗ 4.42 ∗ ∗ ∗ 4.25 ∗ ∗ ∗ 4.34 ∗ ∗ ∗ 0.51
BVt 22.91e207
22.96 ∗ ∗ ∗
SALESt 28.28e208
1.45
TRELFEEt 174.459 37.709 332.670 260.367 436,226.9
1.97 ∗ ∗ 0.82 3.00 ∗ ∗ ∗ 2.46 ∗ ∗ ∗ 2.81 ∗ ∗ ∗
TRELFEEt 2
29,674.014 213,244.38 211,217.85 22.23e20.7
21.99 ∗ ∗ 22.83 ∗ ∗ ∗ 2.49 ∗ ∗ ∗ 22.33 ∗ ∗ ∗
CONSTANT 0.646 0.770 0.604 0.597 194.164
1.24 1.47 1.18 1.15 0.39
Adjusted R2 50.56% 50.25% 51.19% 50.73% 82.73%
F 63.21∗∗∗ 65.88∗∗∗ 61.61∗∗∗ 60.51∗∗∗ 292.39∗∗∗
(18,218) (17,218) (19,218) (19,218) (18,218)
N 1,157 1,157 1,157 1,157 1,157
Partial F 3.97 (1,218)∗
Value of TRELFEE at 0.009 n/a 0.012 0.012 0.009
turning point
Notes: (a) Coefficient estimates and standard errors for industry dummies (INDDUMS) and year dummies (YEARDUMS) are not reported. (b) The reported t-statistics in italics are
Rogers (1993) company cluster adjusted. (c) ∗ , ∗∗ , ∗∗∗ indicate significance at the 5%, 2.5% and 1% level respectively. (d) The x coordinate of the turning point of a quadratic of the
general form y ¼ ax2 + bx + c is given by: x ¼ 2b/2a.
129

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


130 K. Holland and J. Lane

of endogeneity of the fee variable through an omitted variable (Larcker and Rusticus 2010).24 The
two most significant size proxies identified in the audit fee literature are book value and sales (Hay
et al. 2006): when we include these in turn as a proxy for an omitted size variable, the qualitative
results that we obtain are consistent with the original models, see columns 3 and 4, Table 5.25

4.3.2 Profit and loss-making status


We consider whether the relationship between fees and market value is conditional upon a com-
pany’s profit or loss status. The financial reporting implications of reduced auditor independence
may include increased earnings management, the direction of which could vary depending on
whether the company is in a profit- or loss-making position (Hayn 1995). Consequently, we esti-
mate the model after having separated the sample into sub-samples of profit-reporting year ends
and loss-reporting year ends, see columns 1 and 2 of Table 6.26
The higher level of significance associated with the two fee variables in the sample of loss-
Downloaded by [The British Library] at 03:43 09 October 2017

making companies, together with the lower associated turning point, suggest that at a given
level of fees, investors are more cautious in valuing loss-making companies than they are
profit-making companies. When alternate deflators are used, the same general result of a differ-
ential response holds when deflating either by the number of shares or by sales. In both cases
the implied turning point is lower for loss-making companies than profit-making companies,
although the pattern of higher significance of the two fee variables for the loss-making sub-
sample is not observed. When deflated by the number of shares, the statistical significances of
the TRELFEE and TRELFEE2 coefficients are at the same (1%) level for both the profit- and
loss-making sub-samples. When estimated using the sales deflated model, the significance of
the TRELFEE and TRELFEE2 coefficients is higher (5%) for the profit sub-sample than for
the loss-making companies (10%).

4.3.3 Andersen audits and auditor change


Two of the largest audit failures during the sample period, Enron and WorldCom, were audited by
Andersen. To assess whether the above results are conditional or dependent upon the inclusion of
Andersen audited companies, we re-estimate the models firstly excluding all year ends in which
Andersen acted as the auditor, and secondly excluding all companies that were audited by Ander-
sen at any time during the sample period. Under both procedures the qualitative interpretation of
the above results does not change; see columns 3 and 4 respectively in Table 6.27

4.3.4 Earnings expectations


We do not control for variation in earnings expectations using analysts’ forecasts for a number of
practical and theoretical reasons, as discussed in section 3.3. To assess the effect of relaxing the
assumption of cross-sectional equality of expectations, we re-estimate Model 1 with, alternatively,
the inclusion of an ex ante or ex post earnings growth variable to proxy for variation in earnings
expectations (reported in Table 6 columns 5 and 6 respectively).28 Neither changes the interpret-
ation of the results reported earlier.

4.3.5. Research and development expenditure


Although we have controlled for cross-sectional variation in research and development expendi-
ture, it is still possible that the fee variables are related to research and development through a
relationship with audit costs or demand for NAS. Consequently, we re-estimate the model

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Table 6. Model 1 additional regressions.
1 2 3 4 5 6
Variables Model 1 Profit Model I Loss Model I Excluding Model I Excluding Model 1 + Ex-ante Model 1 + Ex-
Downloaded by [The British Library] at 03:43 09 October 2017

makers makers Andersen audits Andersen (former) clients growth post growth
BVt/BVt21 1.139 2.521 1.856 1.845 1.650 1.621
5.32 ∗ ∗ ∗ 3.407 ∗ ∗ ∗ 3.65 ∗ ∗ ∗ 3.43 ∗ ∗ ∗ 4.62 ∗ ∗ ∗ 4.55 ∗ ∗ ∗
(Nit + RDt)/BVt21 7.795 21.559 5.346 5.240 5.452 5.460
6.29 ∗ ∗ ∗ 21.34 5.55 ∗ ∗ ∗ 5.00 ∗ ∗ ∗ 6.06 ∗ ∗ ∗ 6.05 ∗ ∗ ∗

Accounting and Business Research


(Nit + RDt)∗ LOSSDUM 25.888 25.742 25.831 25.918
24.92∗∗∗ 24.17 ∗ ∗ ∗ 25.04 ∗ ∗ ∗ 25.10 ∗ ∗ ∗
RDt/BVt21 1.124 8.103 6.169 6.117 6.184 6.221
0.50 5.14 ∗ ∗ ∗ 4.51 ∗ ∗ ∗ 4.14 ∗ ∗ ∗ 4.41 ∗ ∗ ∗ 4.40 ∗ ∗ ∗
(NIt+1- NIt)/| NIt | 0.171
2.11 ∗
(NIt— Nit21)/|Nit21| 0.001
0.05
TRELFEEt 191.549 397.099 174.534 181.310 173.220 174.412
1.87 ∗ 2.30 ∗ ∗ 1.98 ∗ ∗ 1.94 ∗ 1.97 ∗ ∗ 1.97 ∗ ∗
TRELFEEt 2
210,363.62 238,732.55 29,668.06 29,874.82 29,569.29 29,671.22
21.93 ∗ 22.03 ∗ ∗ 22.00∗∗ 21.98 ∗ ∗ 21.99 ∗ ∗ 21.99 ∗ ∗
CONSTANT 0.743 20.555 0.834 0.342 0.598 0.646
1.56 20.06 1.58 0.45 1.14 1.24
Adjusted R2 49.51% 61.55% 50.06% 51.27% 50.51%
F 50.24∗∗∗ 23.32∗∗∗ 59.39∗∗∗ 49.68∗∗∗ 61.81 60.00∗∗∗
(17,209) (18,223) (18,217) (18,180) (19,218) (19,218)
N 905 252 1108 942 1,157 1,157
Value of TRELFEE at 0.009 0.005 0.009 0.009 0.009 0.009
turning point
Notes: (a) Coefficient estimates and standard errors for industry dummies (INDDUMS) and year dummies (YEARDUMS) are not reported. (b) The reported t-statistics in italics are
Rogers (1993) company cluster adjusted. (c) ∗ , ∗∗ , ∗∗∗ indicate significance at the 5%, 2.5% and 1% level respectively.(d) The x coordinate of the turning point of a quadratic of the
general form y ¼ ax2 + bx + c is given by: x ¼ 2b/2a. 131

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


132 K. Holland and J. Lane

having excluded the 580 year ends with research and development expenditure; the results based
on the remaining sample of 577 year ends are qualitatively identical to initial results in Model 1.29

4.4. Model 2: NAS and audit fees


We discussed in section 3.1 conflicting interpretations of the relative influence on perceived inde-
pendence of audit fees and NAS fees. To assess these we estimate Model 2, with separate relative
fee variables for audit (RELAF) and NAS fee (RELNAS) and their related quadratic terms
RELAF2 and RELNAS2. These variables are used in place of the combined fee variables
TRELFEE and TRELFEE2 employed in Model 1. The results, which are reported in column 1
of Table 7 (panel A), suggest that the quadratic relationship found with respect to total relative
fees is driven by NAS.
The RELNAS and RELNAS2 coefficients are positive and negative respectively, with signifi-
cance at the 1% and 2.5% levels in turn, whilst the RELAF and RELAF2 audit fee variables are
Downloaded by [The British Library] at 03:43 09 October 2017

not significant at normal levels.30 The results of testing hypothesis 2 are shown in panel B and
indicate significant differences between the coefficients of RELAF and RELNAS and between
the coefficients RELAF2 and RELNAS2. Further, when added together to obtain their net
effect, the combined coefficient of RELAF and RELAF2 is significantly different from the com-
bined coefficient of RELNAS and RELNAS2. These results are consistent with investors inter-
preting non-audit services (but not audit fees) as carrying economic rents, an argument
exemplified by Srinidhi and Gul (2007). Since we find no evidence to suggest that investors per-
ceive the rising audit fees noted in section 4.1 as indicating attempts on the part of audit firms to
increase the profitability of audit, we speculate that those fee increases are attributed to ‘more
audit’.
To provide comparability with earlier studies, e.g. Frankel et al. (2002), we examine the effect
of using, as an alternate fee variable measure, the ratio of NAS fees to audit fees NAS/AF,
together with its related quadratic (NAS/AF)2. As we discuss in section 3, this measure does
not capture scale effects and so does not provide a clear measure of economic bonding.
However, it may be value relevant if investors perceive a ‘high’ value of the measure as over-
reliance by a company on its auditors for advice on managerial issues. When this ratio is used
as a fee variable, the coefficient, and that of the related quadratic term, are not statistically signifi-
cant at accepted levels. Columns 2 and 3 of Table 7 report the model of Frankel et al. in full and
restricted forms respectively.
These results, combined with the results based on the relative fee variables used in Model 2,
reported in column 1 of Table 7, suggest that the finding of a non-linear relationship between fees
and value is driven by concerns over auditor independence rather than a lack of management
capability.

4.5 Annual estimations


We also estimate Models 1 and 2 on a year-by-year basis, thereby allowing the fee variable coef-
ficients to vary by year. It is conceivable that investors’ attitudes to fee levels and auditor inde-
pendence could vary over time. This is particularly the case here, where the sample period
includes one year of observations prior to both the first disclosures concerning Enron and World-
Com and the subsequent responses of accounting firms. The annual regressions are more sensitive
to the treatment of influential observations. We report results based on the sample before the
exclusion of influential observations, in order to maintain identical sample composition
between the annual and pooled regressions as reported in Table 5.

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Accounting and Business Research 133

Table 7. Model 2 and ‘Frankel et al.’ model regressions.


1 2 3
Variables: Model 2 Frankel et al Frankel et al
model - full model - restricted
Panel A
BVt/BVt21 1.615 1.613 1.61
4.53 ∗ ∗ ∗ 4.54 ∗ ∗ ∗ 4.56 ∗ ∗ ∗
(Nit + RDt)/BVt21 5.516 5.555 5.552
6.09 ∗ ∗ ∗ 6.10 ∗ ∗ ∗ 6.08 ∗ ∗ ∗
(Nit + RDt)∗ LOSS 25.914 26.163 26.155
25.16 ∗ ∗ ∗ 25.28 ∗ ∗ ∗ 25.27∗∗∗
RDt/BVt21 6.177 6.206 6.210
4.40 ∗ ∗ ∗ 4.40 ∗ ∗ ∗ 4.39 ∗ ∗ ∗
RELAFt ¼ AFt/AFFEEt 2194.323
21.32
Downloaded by [The British Library] at 03:43 09 October 2017

RELAFt2 ¼ RELAFt ∗ RELAFt 4,051.018


0.70
RELNASt ¼ NASt/AFFEEt 447.220
2.52 ∗ ∗ ∗
RELNASt2 ¼ RELNASt∗ RELNASt 230,975.33
22.25 ∗ ∗
NASt/AFt 20.016 20.005
20.15 20.09
(NASt/AFt)2 0.001
0.19
CONSTANT 0.667 1.27 0.841 0.834
1.60 1.58
Adjusted R2 50.63% 50.14% 50.18%
F 57.46∗∗∗ 62.18∗∗∗ 65.69∗∗∗
(20,218) (18,218) (17,218)
N 1,157 1,157 1,157
Partial F 0.04
(1,218)
Value of RELNAS at turning point 0.007
Panel B
H2(1)o: bRELAF ¼ bRELNAS 4.92∗∗
(1,218)
H2(2)o: bRELAF2 ¼ bRELNAS2 4.86∗∗
(1,218)
H2(3)o: bRELAF + bRELAF2 ¼ bRELNAS + 4.81∗∗
bRELNAS2 (1,218)
Notes: (a) Coefficient estimates and standard errors for industry dummies (INDDUMS) and year dummies
(YEARDUMS) are not reported. (b) The reported t-statistics in italics are Rogers (1993) company cluster adjusted.
(c) ∗ , ∗∗ , ∗∗∗ indicate significance at the 5%, 2.5% and 1% level respectively. (d) The x coordinate of the turning
point of a quadratic of the general form y ¼ ax2 + bx + c is given by: x ¼ 2b/2a.

4.5.1 Total relative fees


In Table 8 (panel A) a summary of the annual estimates of Model 1 showing TRELFEE and
TRELFEE2 are reported.
The fee variables display the hypothesised signs in all six annual regressions. In each of the
years 2000/01, 2002/03 and 2003/04 both coefficients are significant at the 5% level or better.
The lack of significance in the other years is unlikely to be due to excessive multicollinearity.31

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


134 K. Holland and J. Lane

In none of the six years is there a significant linear relationship observed between market value
and TRELFEE when estimating the restricted form of Model 1 (Table 8 panel B). The critical or
turning point level of relative fees fluctuates over the six years, from a minimum of 0.06% in
2003/04 to a maximum of 0.12% in 2002/03. Immediately pre Enron it was 0.08%. Although
there is variation in the turning points, significant differences in the coefficient estimates
between years, which would be indicative of significant changes in shareholder attitudes, are
absent in all but two cases.32

4.5.2 Total relative fees disaggregated


The observed variation may be due to changes in the composition of fee paid by companies to
their auditor. Though Model 2 detects a significant relationship in the case of NAS, and none
in the case of audit fees, it is instructive to consider whether this result holds in each of the indi-
vidual years. This is particularly relevant in light of the change in relative proportions as reported
Downloaded by [The British Library] at 03:43 09 October 2017

in Table 3 (panel B) above. In Table 8 (panel C) we report further disaggregated annual results
based on separate testing of the audit and NAS fees; these are the annual equivalents of the
pooled Model 2 reported in column 1 Table 7. The audit fee-related variables RELAF and
RELAF2 do not exhibit a systematic pattern and, with the exception of 2003/04, are consistent
with the insignificant results observed in the pooled Model 2 reported in Table 7. The results
for the NAS variables RELNAS and RELNAS2 are more consistent with the pooled results
reported in Table 7 in displaying the same combination of positive and negative coefficient for
RELNAS and RELNAS2 in five of the six years. However, the coefficients are statistically sig-
nificant only in 2002/03.33 Analysed annually, neither NAS nor audit fees fully explain the
pooled results reported in Tables 5 and 7, which support the view that NAS fees dominate inves-
tors’ perceptions of independence.
The lack of a systematic annual relationship, particularly with respect to the disaggregated
annual results reported here, means that we are unable to confidently ascribe the observed vari-
ations to any specific external events, whether audit failure such as Enron and WorldCom or regu-
latory change. There is no evidence of a negative relationship following these audit failures.
Indeed the highest turning point in total relative fees is observed later, in 2002/3 (Table 8
panel A). By then regulators and the profession in the UK had signalled willingness to act on
NAS and broader corporate governance, while the US had already done so. That is not itself per-
suasive evidence of the impact of individual regulatory responses, since no significant relation-
ship is observed in three of the five remaining years (table 8 panel A). We speculate that
subsequent lack of significance may be related to perception on the part of investors that the
change that took place in the UK corporate governance and regulatory framework in the final
years of this period was incremental in nature.

5. Conclusions
This study makes three specific contributions to the literature examining the relationship between
auditor fees and perceived auditor independence and in doing so provides evidence from the UK
markets that we expect to have broader significance. Firstly, by adopting a valuation relevance
approach we are able to examine the impact of investors’ perceptions in a UK setting that is
unfavourable to the use of ex ante models reliant on analysts’ forecasts, while overcoming
some of the difficulties inherent in an event study. Our second contribution is to provide empirical
evidence that the relationship between auditor fees and perceived auditor independence is non-
linear, whereas prior empirical studies of this topic have generally assumed a linear relationship.
The turning point in this non-linear relationship occurs when total fees are in the region of £10.7

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Table 8. Models 1 and 2 annual regressions.

Period 2000/01 2001/02 2002/03 2003/04 2004/05 2005/06


Downloaded by [The British Library] at 03:43 09 October 2017

Panel A
TRELFEEt ¼ (AFt + NASt )/AFFEEt 435.780 290.853 201.600 433.410 288.366 39.493
2.46 ∗ ∗ 1.35% 2.37 ∗ ∗ 2.86 ∗ ∗ ∗ 1.62% 0.27

TRELFEEt2 ¼ TRELFEEt TRELFEEt 227,304.89 227,853.25 28,670.942 237,887.05 227,768.41 29,251.015
22.08 ∗ 21.40 23.33 ∗ ∗ ∗ 23.56∗∗∗ 21.75% 20.62
Adjusted R2 63.70% 39.63% 55.83% 38.75% 31.71% 62.07%

Accounting and Business Research


F 21.94 8.49 16.53 9.81 7.30 24.29
Value of TRELFEE at turning point 0.008 n/a 0.012 0.006 n/a n/a
N 190 196 198 196 191 186
Partial f test 4.34 1.96 11.07∗∗∗ 12.67∗ ∗∗ 3.06 0.38
(1,16) (1,16) (1,16) (1,16) (1,15) (1,14)
Panel B
TRELFEEt ¼ (AFt + NASt)/AFFEEt 66.900 67.001 8.6434 69.430 58.038 247.544
0.63 1.11 0.32 0.61 1.02 21.46
Panel C
RELAFt ¼ AFt/AFFEEt 2588.09 287.87 2110.85 1,209.41 217.05 231.860
21.14 0.61 20.60 3.47 ∗ ∗ ∗ 0.45 20.19

RELAFt2 ¼ RELAFt RELAFt 18,216.39 271,992.35 1,749.13 233,0092.8 268,381.12 256,146.48
0.48 21.42 0.26 22.88 ∗ ∗ 21.09 22.41 ∗ ∗
RELNASt ¼ NASt/AFFEEt 770.90 319.71 461.19 2193.01 495.95 298.59
2.14∗ 0.69 3.60 ∗ ∗ ∗ 20.49 0.47 1.01
RELNASt2 ¼ RELNASt∗ RELNASt 245,789.7 239,648.99 231,981.9 98,163.78 240,597.48 215,978.27
1.40 20.79 24.47 ∗ ∗ ∗ 1.03 20.19 20.50
Adjusted R2 60.55% 34.56% 52.45% 44.85% 37.06% 62.21%
F 19.13 ∗ ∗ ∗ 7.44 ∗ ∗ ∗ 14.58 ∗ ∗ ∗ 9.10 ∗ ∗ ∗ 6.40 ∗ ∗ ∗ 21.30 ∗ ∗ ∗
(16,173) (16,179) (16,181) (16,179) (16,174) (15,170)
Value of RELAF at turning point n/a n/a n/a 0.002 n/a n/a
Value of RELNAS at turning point n/a n/a 0.007 n/a n/a n/a

Notes: (a) Coefficient estimates and standard errors for industry dummies (INDDUMS) and year dummies (YEARDUMS) are not reported. (b) The reported t-statistics in italics are
Rogers (1993) industry cluster adjusted. (c) ∗ , ∗∗ , ∗∗∗ indicate significance at the 5%, 2.5% and 1% level respectively.(d) The x coordinate of the turning point of a quadratic of the general
form y ¼ ax2 + bx + c is given by: x ¼ 2b/2a.

135

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


136 K. Holland and J. Lane

million, equivalent to 0.9% of auditor fee income. This finding is consistent across both profit-
and loss-reporting companies, and is robust to the use of alternative model specifications. This
confirms our hypothesis that shareholders perceive a threat to auditor independence only at
high levels of total relative fees, and is consistent with the argument that the risk of economic
bonding drives investor concerns over fee levels. At lower levels, relative fees are positively
related to company value, which is consistent with both the observed levels of NAS purchases
made by companies on a regular annual basis, and reduced agency costs from increases in the
level of audit performed. Further analysis suggests that relative NAS fees are the source of the
concave relationship. We also examine the effect of the widely used NAS ratio (NAS fees to
audit fees) since investors may perceive the ratio as a measure of auditee dependence rather
than as a measure of auditor dependence. However, we find this ratio to have no significance,
and interpret this as further strengthening our conclusion that our main result of a non-linear
relationship reflects concerns over economic bonding rather than management capability.
Thirdly, our six-year sample period, and its specification, allows the stability of any relation-
Downloaded by [The British Library] at 03:43 09 October 2017

ship to be analysed over a relatively long period of time. We find the existence of the above non-
linear relationship to lack consistent significance year by year over the sample period, although
the results show the same sign in each year.34 We find no clear significance of any kind in the
sample year immediately following the first public disclosure relating to Enron, although the
two following years both display significant non-linear relationships. Owing to the lack of con-
sistency and robustness of the annual results we cannot attribute our overall results to any specific
external events, whether corporate failure or regulatory change. Our annual results imply that any
event study of this topic is likely to suffer from the effect of fluctuation in the expression of market
perceptions. We suggest that studies of the relationship between fees and perceived auditor inde-
pendence need to be designed and interpreted to take account of both changing fee composition
and potential variation in investor responses over time. Similarly, impact of incremental regulat-
ory change may be cumulative or gradual and not clearly reflected in event studies.
The policy implications of this paper lie in its confirmation that separate disclosure of NAS
fees in the UK are of relevance to investors. This reveals a sophisticated response on their part
to disclosed information even where, as here, that information is itself relatively crude. Hence
it supports the argument for retaining disclosure. Secondly, at the lower relative levels that rep-
resent the majority of the companies studied here, investors’ concerns over impaired auditor inde-
pendence are outweighed by positive evaluation of the benefits. Our interpretation of this result is
twofold. Firstly, it gives some support to calls for oversight or regulation to be based on threshold
relative fees. Secondly, it suggests that investors may not perceive prohibition of NAS supplied by
auditors to be in their interests in the majority of companies examined, since even for those com-
panies with relatively high levels of auditor fees, an element of price protection is evident, as
implied by the concave relationship.
Further research is required to fully explore the relationship between investor perceptions of
economic bonding and corporate governance in the UK context. This relationship has received
more attention in the US context (Larcker and Richardson 2004, Khurana and Raman 2006).
Though these studies do not suggest that perceptions of NAS are conditional upon recognised
corporate governance factors, their findings may not be applicable to different regulatory
frameworks. We speculate that in a weak internal governance environment, investors may be
more concerned about NAS levels, reflected in lower turning points. Conversely, increased
audit fees may be interpreted as improving the overall corporate governance of the company.
By contrast, a relatively strong governance environment may result in either downward pressure
on NAS fees, dampening the non-linear relationship, or enhanced investor confidence reflected in
a higher turning point.

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Accounting and Business Research 137

Finally, further research is needed to establish whether the current disclosure requirements
provide investors with all of the necessary information with which to make the required judge-
ments, in particular with respect to categorisation of expenditure and audit partners. An important
component of the total relative fee takes the form of auditor fee income, hence any attempt to
address the adequacy and costliness of information available to investors needs to consider infor-
mation provided by both a company and its auditor.

Acknowledgements
We appreciate helpful comments from Martin Broad, Mark Clatworthy, Stella Fearnley, David Gwilliam,
Deborah Page, Mike Page, Mike Peel, Megan Williams and seminar participants at University of Wales
Accounting and Finance Colloquium; Southampton Management School, University of Southampton; Not-
tingham University Business School; University of Portsmouth, 2009 National Auditing Conference and
2009 American Accounting Association Annual Meeting. In particular we would like to thank the editor
and two reviewers for their detailed comments and suggestions.
Downloaded by [The British Library] at 03:43 09 October 2017

Notes
1. Sarbanes-Oxley Act (2002). Public Company Accounting Reform and Investor Protection Act of
2002. Washington D.C., US Congress. Pub L 107-204 116 Stat: 745.
2. Commission Recommendation 2002/590/EC: Statutory auditors’ independence in the EU: a set of
fundamental principles [online]. Available from: http://eur-lex.europa.eu [Accessed 13 August
2011]. Using a principles-based approach, it discourages the following services: accounting services,
IT, valuation, internal audit, legal representation, and senior management recruitment.
3. Directive 2006/43/EC: Preamble 11 and Article 49.1 respectively. This directive updated the ‘Eighth
Directive’ i.e. Directive 78/660/EEC in requiring separate disclosure of other assurance, taxation and
other NAS. Available from http://eur-lex.europa.eu [Accessed 6 June 2009].
4. Companies (Disclosure of Auditor Remuneration) Regulations, 2005. SI 2005/2417. This introduced
a requirement for more detailed disclosure of fees other than audit in the annual financial statements of
companies which do not qualify as ‘small’ or ‘medium’ sized. The nine categories of fees other than
audit include all categories recommended by 2002/590/EC (fn1).
5. Audit Practice Board, Ethical Standards 4 and 5 (2004) respectively. ES5 restricts NAS, particularly
relating to listed companies, by reference to materiality and the level of safeguards required. In
certain cases the standard in effect prohibits provision to listed companies. Its provisions are consistent
with but more extensive than 2002/590/EC (fn1). ES4 23-26 prohibits audit firms from taking on a
client where the total fees received from that client would regularly exceed 10% of firm income. Pre-
vious guidance recommended the same cap.
6. At company level, until the 2005 regulations came into force (cf footnote 4) companies were required
to disclose total NAS (along with audit fees) but a detailed breakdown was not required.
7. 43% of both institutional and small investors in Dart’s study, based on 2005 data, expressed concern
about the risk to independence posed by NAS provision. The level of concern also varied with respect
to the type of service (e.g. internal audit, tax).
8. Arguably, different types of NAS may generate different patterns of perception, if known to investors.
However, disclosure of this data was not required until 2005 cf footnote 4.
9. However a potential disadvantage in not utilising analysts’ forecasts is loss of information concerning
future earnings growth. In subsequent sensitivity analysis we include proxies for earnings growth. We
thank a reviewer for raising this point.
10. Specifically, we deflate by sales and number of shares in turn, exclude all year-ends with research and
development expenditure, exclude Andersen audited observations, incorporate a size variable, include
an earnings expectation control, and exclude influential observations. Overall the results are qualitat-
ively unchanged from those reported on the above model.
11. This restriction results in the time ‘clustering’ with the potential of significant cross correlation in the
error terms. We follow Petersen (2009) and report Rogers (1993) cluster-adjusted standard errors by
company. The inferences based on this method are consistent with estimations using White (1980)
adjusted standard errors. Further, potential bias could arise from excluding non-December and
March year-ends if choice of year-end is influenced by industry membership, yet we are not aware

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


138 K. Holland and J. Lane

of any evidence pointing to such an association. In the reported models we also control for industry. We
thank a reviewer for raising these two points.
12. Audit fees were cross-checked between FAME and Datastream databases. Where a difference
occurred, the audit fee and NAS fee were obtained from the company’s financial statements.
13. PricewaterhouseCoopers (PwC) disclosed fee income for the first time in 2002, which corresponds to
the third year of data analysed in this study. For the two earlier years PwC’s fee income was proxied by
that of the second largest firm, KPMG. In the absence of publicly available information on the level of
NAS services supplied by PwC to non-auditees it is not possible to construct a contemporaneous esti-
mate of PwC’s fee income for these two years.
14. To the extent that firms are delisted because of poor performance we can attempt a control for any
potential survivorship bias. By treating the occurrence of a loss as evidence of poor performance
and excluding such observations, the estimation can be made on a sub-sample of persistent profit
makers. Such an analysis is reported in the section on additional tests. However, if survivorship
bias can arise from either poor or high performance, then it is unclear what bias would be introduced
to the results from not including delisted companies.
15. These figures are comparable with those in a recent UK-based study by O’Hanlon and Taylor (2007).
16. The weighted average fee income for a ‘Big 4/5’ audit firm in the six-year sample period is
Downloaded by [The British Library] at 03:43 09 October 2017

£1,186.966M. It is the sum of the weighted total fee income per audit firm where the weighting
factor is the relative frequency of audit firms’ clients in the sample.
17. Audit fees increased at a greater rate than total Big 4/5 audit firm revenues, which increased by 31% in
the same period.
18. The test statistic U ¼ 1.88 is significant at the 5% level.
19. Calculated as: 0.009∗ £1,186.966M (the weighted average total fee income per audit firm, see note 16).
20. Partial F –test 3.97, reported in Table 5 column 2.
21. In the book value deflated model (1) and the alternate models deflated by number of shares or sales, the
maximum condition indices are 15.28, 13.63 and 18.88 respectively. As all three values are under the
‘threshold’ level of 30 (Hair et al. 1998) we conclude that the degree of multicollinearity in the data set
is acceptable. Notwithstanding the insignificant condition indices, the maximum condition index
explains a significant proportion (.0.9) of the variance of only one variable which again indicates
no significant multicollinearity (Hair et al. 1998).
22. At 5.62 the maximum VIF is below the usual threshold value of 10 (Hair et al. 1998).
23. The critical DFIT value is where abs (DFIT) . 2∗ (P/N)2 where P and N are the number of variables
and observations respectively. Note that this procedure is different from the exclusion of outliers
referred to in the specification of the sample.
24. As expected the correlation coefficients between the book value and the fee variables are all significant
at the 1% level. However, none of the correlation coefficients are significant (1%) when the process is
repeated with the deflated book value variable. This suggests the process of deflating is effective.
25. Larcker and Rusticus (2010) discuss a second potential source of endogeneity, i.e. simultaneity, when
at least one of the explanatory variables is partially determined by the dependent variable. As a con-
sequence, any attempt to ascribe causality to the observed relationship between the variables concerned
would be thwarted. In the relationship between company market value and fee variables reported in
this paper we are not aware of any findings that suggest that fee levels are determined by company
market value in a manner that would explain the observed quadratic relationship.
26. Estimating the model over a sample consisting entirely of profit makers has the advantage also of
removing loss makers, whose inclusion in earnings based valuation models raises conceptual issues
(Hayn 1995).
27. In addition to forty-four changes of audit firm following the collapse of Andersen, thirty-one further
audit firm changes occurred. Excluding all seventy-five companies, or controlling for the year of
change with a dummy variable, does not change the qualitative interpretation of the above results.
Results not tabulated, but available from authors.
28. The ex ante variable is defined as (NIt+1 2NIt)/|NIt|, the ex post variable is defined as (NIt 2 Nit21)/
|Nit21| where NI ¼ net income after tax available to shareholders.
29. Results not tabulated, but available from authors.
30. This result is consistent with the positive correlation between the level of TRELFEE and the ratio of
NAS to AF. (Pearson correlation coefficient ¼ 0.219 significant at 1% level).
31. Over the six annual regressions the largest condition index is 19.58, in 2003/04, which is below the
‘threshold’ level of 30 (Hair et al. 1998).

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Accounting and Business Research 139

32. Testing the difference in the value of the TRELFEE (TRELFEE2) coefficients between years reveals
only two significant differences, the first in the comparison between 2003/04 and 2005/06 in the case
of TRELFEE, and the second between 2002/03 and 2003/04 in the case of TRELFEE2.
33. The lack of significance is unlikely to be due to excessive multicollinearity. Over the six annual
regressions the largest condition index is 21.47, in year 4, which is below the “threshold” level of
30 (Hair et al. 1998).
34. If coefficients, because they are being generated by a random process, are insignificantly different from
zero, one would expect to see an equal number of positive and negative signed coefficients i.e. ran-
domly distributed around a population mean of zero. However, observing a disproportionate
number of positive or negative coefficients, irrespective of their statistical significance, is consistent
with the presence of a systematic or non-random effect. The statistical significance of the five pairs
of positive signed coefficients arising in the six annual regressions is estimated using the binomial dis-
tribution at 10%.

References
Downloaded by [The British Library] at 03:43 09 October 2017

Abidin, S. Beattie, V. and Goodacre, A., 2010. Audit market structure, fees and choice in a period of struc-
tural change: evidence from the UK 1998–2003. British accounting review, 42 (3), 187–206.
Accountancy, 2002. Marking the break – PwC confirms IPO for consulting arm [online]. CCH. Available
from: http://www.accountancymagazine.com [Accessed 13 August 2011].
Accountancy, various years. Top 60 UK firms [online]. CCH. Available from: http://www.
accountancymagazine.com [Accessed 13 August 2011].
Amir, E., Guan, Y. and Livne, G., 2010. Auditor independence and the cost of capital before and after
Sarbanes-Oxley: the case of newly issued public debt. European accounting review, 19 (4), 633–664.
APB, 2009. Consultation on audit firms providing non-audit services to listed companies that they audit,
Auditing Practices Board. London: Financial Reporting Council.
Ashbaugh, H.R., LaFond, M. and Mayhew, B.W., 2003. Do non-audit services compromise auditor indepen-
dence? Accounting review, 78 (3), 611–39.
Barron, M., 1986. Year-end heterogeneity in calculations of industry and economy averages of accounting
numbers. Accounting and business research, 16 (6 –4), 275–284.
Barth, M. and Kallapur, S., 1996. The effects of cross-sectional scale differences on regression results in
empirical accounting research. Contemporary accounting research, 13 (2), 527–567.
Bartlett, R.W., 1997. Auditor independence: five scenarios involving potential conflicts of interest. Research
on accounting ethics, 3, 245–277.
Beattie, V. and Fearnley, S., 2002. Auditor independence and non-audit services:a literature review. London:
The Institute of Chartered Accountants in England and Wales.
Beattie, V., Fearnley, S. and Brandt, R., 1999. Perceptions of auditor independence: U.K. evidence. Journal
of international accounting, auditing & taxation, 8 (1), 67–107.
Beattie, V., Fearnley, S. and Hines, T., 2009. The impact of changes to the non-audit services regime on
finance directors, audit committee chairs and audit partners of UK listed companies. London: The
Institute of Chartered Accountants in England and Wales.
Belsey, D.A., Kuh, E. and Welsch, R.E., 1980. Regression diagnostics. New York: John Wiley & Sons.
Boone, J., Khurana, I. and Raman, K., 2008. Audit firm tenure and the equity risk premium. Journal of
accounting, auditing and finance, 23 (1), 115–140.
Capstaff, J., Paudyal, K. and Rees, W., 2001. A comparative analysis of earnings forecasts in Europe.
Journal of business finance & accounting, 28 (5 and 6), 531–562.
Chan, P.W., Ezzamel, M. and Gwilliam, D.R., 1993. Determinants of audit fees for quoted UK companies.
Journal of business finance and accounting, 20 (6), 765–86.
Collett, N., 2004. Reactions of the London Stock Exchange to company trading statement announcements.
Journal of business finance & accounting, 31 (1 and 2), 3–35.
Dart, E., 2011. UK investors’ perceptions of auditor independence. British accounting review, 43 (3), 155 –
250.
Davis, S.M. and Hollie, D., 2008. The impact of non-audit service fee levels on investors’ perceptions of
auditor independence. Behavioural research in accounting, 20 (1), 31–44.
Dedman, E., Kausar, A. and Lennox, C., 2009. Non-audit services and the demand for voluntary audits
[online]. Available from: http://ssrn.com/abstract=1324082 [Accessed 9 August 2011].
Dhaliwal, D.S., et al., 2008. Auditor fees and cost of debt. Journal of accounting, auditing & finance, 23 (1),
122–145.

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


140 K. Holland and J. Lane

Dopuch, N., et al., 2003. Independence in appearance and in fact: an experimental investigation.
Contemporary accounting research, 20 (1), 79–119.
Easton, P.D. and Sommers, G.A., 2007. Effect of analysts optimism on estimates of the expected rate of
return implied by earnings forecasts. Journal of accounting research, 45 (5), 983–1015.
Ezzamel, M., Gwilliam, D.R. and Holland, K.M., 1996. Some empirical evidence from publicly quoted UK
companies on the relationship between the pricing of audit and non-audit services. Accounting and
business research, 27 (1), 3–16.
Firth, M., 2002. Auditor provided consultancy services and their association with audit fees and audit
opinions. Journal of business finance and accounting, 29 (5 and 6), 661–693.
Francis, J.R., 2006. Are auditors compromised by nonaudit services? Assessing the evidence. Contemporary
accounting research, 23 (3), 747–760.
Francis, J.R. and Ke, B., 2006. Disclosure of fees paid to auditors and the market valuation of earnings sur-
prises. Review of accounting studies, 11 (4), 1380–6653.
Frankel, R.M., Johnson, M.F. and Nelson, K.F., 2002. The relation between auditors fees for nonaudit ser-
vices and earnings management. Accounting review, 77 (Supplement), 71–105.
FRC, 2003. The combined code on corporate governance. London: Financial Reporting Council.
Ghosh, A. and Pawlewicz, R., 2009. The impact of regulation on auditor fees: evidence from the Sarbanes-
Downloaded by [The British Library] at 03:43 09 October 2017

Oxley Act. Auditing: a journal of practice & theory, 28 (2), 171–197.


Green, J.P., Stark, A.W. and Thomas, H.M., 1996. UK evidence on the market valuation of research and
development expenditures. Journal of business finance and accounting, 23 (2), 191–216.
Gwilliam, D.R., 1987. A survey of auditing research. London: Prentice-Hall/ICAEW.
Hair, J.F., et al., 1998. Multivariate data analysis. 5th ed. Prentice-Hall: Upper Saddle River.
Hay, D., Knechel, W.R. and Wong., N., 2006. Audit fees: a meta-analysis of the effect of demand and supply
attributes. Contemporary accounting research, 23 (1), 141–91.
Hayn, C., 1995. The information content of losses. Journal of accounting and economics, 20 (2), 125–133.
Hodgkinson, L., 2001. Analysts forecasts and agency problems. Journal of business finance and accounting,
28 (7 and 8), 943–961.
Holland, K.M. and Horton, J., 1993. Initial public offerings on the unlisted securities market: the impact of
professional advisers. Accounting and business research, 24 (93), 19–34.
Horton, J., 2008. The value relevance of realistic reporting: evidence from UK life insurers. Accounting and
business research, 37 (3), 175–197.
Khurana, I.K. and Raman, K.K., 2006. Do investors care about the auditors economic dependence on the
client? Contemporary accounting research, 23 (4), 977–1016.
Krishnan, J., Heibatollah, S. and Zhang, Y., 2005. Does the provision of nonaudit services affect investor
perceptions of auditor independence? Auditing: a journal of practice & theory, 24 (2), 111–35.
Larcker, D.F. and Richardson, S.A., 2004. Fees paid to audit firms, accruals choices, and corporate govern-
ance. Journal of accounting research, 42 (3), 625–658.
Larcker, D.F. and Rusticus, T.O., 2010. On the use of instrumental variables in accounting research. Journal
of accounting and economics, 49 (3), 186–205.
Lind, J.T. and Mehlum, H., 2010. With or without U? The appropriate test for a U-shaped relationship.
Oxford bulletin of economics and statistics, 72 (1), 109–118.
Lowe, D.J. and Pany, K., 1995. CPA performance of consulting engagements with audit clients: effect on
financial statement users perceptions and decisions. Auditing: a journal of practice & theory, 14 (2),
35–53.
O’Hanlon, J. and Taylor, P., 2007. The value relevance of disclosures of liabilities of equity-accounted inves-
tees: UK evidence. Accounting and business research, 37 (4), 267–284.
Petersen, M.A., 2009. Estimating standard errors in finance panel data sets: comparing approaches. Review of
financial studies, 22 (1), 435–480.
POB, 2009. Key facts and trends in the accountancy profession. London: Professional Oversight Board,
Financial Reporting Council.
Reynolds, J.K. and Francis, J., 2001. Does size matter? The influence of large clients on office-level auditor
reporting decisions. Journal of accounting and economics, 30 (3), 375–400.
Rogers, W., 1993. Regression standard errors in clustered samples. Stata Technical Bulletin, 13, 19–23.
Reprinted in Stata Technical Bulletin Reprints, 3, 88–94.
Scheiner, J.H., 1984. An empirical assessment of the impact of SEC nonaudit service disclosure require-
ments on independent auditors and their clients. Journal of accounting research, 22 (2), 789–797.
Schneider, A., Church, B.K. and Ely, K.M., 2006. Non-audit services and auditor independence: a review of
the literature. Journal of accounting literature, 25, 169–211.

Supplied by the British Library 09 Oct 2017, 11:44 (BST)


Accounting and Business Research 141

Simunic, D.A., 1984. Auditing, consulting, and auditor independence. Journal of accounting research, 22
(2), 679–02.
Srinidhi, B.N. and Gul, F.A., 2007. The differential effects of auditors nonaudit and audit fees on accrual
quality. Contemporary accounting research, 24 (2), 595–625.
White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroske-
dasticty. Econometrica, 48 (4), 817–838.
Young, Rt. Hon. Lord., 1989. HL Deb (1998–89) 16 January 1989, 503 c 10.
Downloaded by [The British Library] at 03:43 09 October 2017

Supplied by the British Library 09 Oct 2017, 11:44 (BST)

You might also like