Description
We are interested in understanding how agency conflicts in private firms arise through
ownership structures and family relationships. Specifically, we analyze auditors’ increase
of effort and firms’ choice of auditors in situations with higher level of agency conflicts.
For a large sample of private firms, we use unique and confidential data (obtained through
special permission by the government) to measure direct and ultimate ownership for each
shareholder as well as extended family relationships (based on marriage and blood lines,
going back four generations and extending out to fourth cousin) among all shareholders,
board members, and CEOs. We first find that audit fees, our proxy for audit effort, vary
as hypothesized with firm-level characteristics related to ownership structures and family
relationships. Second, we find evidence that firms in higher agency cost settings respond by
having their financial statements audited by a higher-quality auditor (i.e., a Big 4 firm).
Agency con?icts and auditing in private ?rms
Ole-Kristian Hope
a,?
, John Christian Langli
b
, Wayne B. Thomas
c
a
Rotman School of Management, University of Toronto, Canada
b
Department of Accounting, Auditing, and Law, BI Norwegian Business School, Norway
c
Michael F. Price College of Business, University of Oklahoma, United States
a b s t r a c t
We are interested in understanding how agency con?icts in private ?rms arise through
ownership structures and family relationships. Speci?cally, we analyze auditors’ increase
of effort and ?rms’ choice of auditors in situations with higher level of agency con?icts.
For a large sample of private ?rms, we use unique and con?dential data (obtained through
special permission by the government) to measure direct and ultimate ownership for each
shareholder as well as extended family relationships (based on marriage and blood lines,
going back four generations and extending out to fourth cousin) among all shareholders,
board members, and CEOs. We ?rst ?nd that audit fees, our proxy for audit effort, vary
as hypothesized with ?rm-level characteristics related to ownership structures and family
relationships. Second, we ?nd evidence that ?rms in higher agency cost settings respond by
having their ?nancial statements audited by a higher-quality auditor (i.e., a Big 4 ?rm).
However, for CEO family-related settings (i.e., where the CEO is related to the major share-
holder or as the number of board members related to the CEO increases), we ?nd no evi-
dence of a greater demand for a Big 4 auditor.
Ó 2012 Elsevier Ltd. All rights reserved.
Introduction
In this study, we seek to understand how ownership
structures and family relationships in?uence agency costs
in private ?rms. We do this by observing two aspects re-
lated to auditing. First, in higher agency cost settings, audi-
tors are more likely to supply greater effort to prevent
misstatement associated with moral hazard and adverse
selection problems. We examine how auditors adjust their
level of effort when auditing ?nancial accounting informa-
tion. Second, a subset of ?rms in higher agency cost settings
likely have a greater demand to choose a higher-quality
auditor to provide a credible signal of their commitment
to higher-quality reporting. To test this, we examine the
extent to which ?rms with various characteristics hire a
Big 4 auditor.
Our examinations draw on very detailed data on ulti-
mate ownership and extended family relationships pro-
vided by the Norwegian government. We ?nd that audit
fees (i.e., our proxy for auditor effort) increase with ex-
pected agency costs.
1
Audit fees relate negatively to owner-
ship concentration and to the extent of ownership by the
second-largest shareholder. Concentrated ownership in-
creases the likelihood that a large shareholder closely mon-
itors managerial actions, and an in?uential second
shareholder monitors potential expropriation by the largest
shareholder. Audit fees also relate negatively to the portion
of shares held by the CEO, consistent with ownership align-
ing the incentives of the CEO and other stakeholders. Audit
fees are positively associated with family relationships be-
tween the CEO and the major shareholder (consistent with
these family relationships indicating reduced monitoring).
0361-3682/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.aos.2012.06.002
?
Corresponding author.
E-mail addresses: [email protected] (O.-K. Hope), john.c.
[email protected] (J.C. Langli), [email protected] (W.B. Thomas).
1
The fee (or effort) regression includes controls for 24 client-?rm
characteristics, ?ve audit-?rm variables, as well as year and industry ?xed
effects.
Accounting, Organizations and Society 37 (2012) 500–517
Contents lists available at SciVerse ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
With respect to board independence, we ?nd that audit
fees decline as the number of board members related to
the largest shareholder increases, consistent with fewer
agency con?icts between owners and the board. In con-
trast, as the number of board members related to the
CEO increases, audit fees increase, suggesting less board
independence and more agency con?icts.
For our tests of demand for Big 4 auditor, we report two
interesting sets of results. First, for agency settings that are
not CEO family-related, we observe results consistent with
those obtained for our auditor effort tests. Speci?cally, the
propensity to hire a Big 4 auditor increases as ownership
concentration decreases, ownership of the second largest
owner decreases, and the major shareholder’s family in?u-
ence on the board decreases. These results are consistent
with the demand for a Big 4 auditor being greater in higher
agency cost settings. In these settings, a higher-quality
auditor plays a stronger role in reducing agency costs by
sending a more credible signal of managers’ commitment
to higher-quality reporting. We do not ?nd signi?cant evi-
dence of a relation between hiring a Big 4 auditor and the
fraction of shares owned by the CEO for our main tests, but
we do in sensitivity tests.
We ?nd no association between the choice to hire a
Big 4 auditor and CEO family-related agency variables.
Speci?cally, we ?nd no signi?cant evidence that the
demand for a Big 4 auditor is affected when a family rela-
tionship exists between the CEO and the major share-
holder or as the number of board members related to
the CEO increases. One explanation for the lack of signif-
icance could be that while some CEOs in family-related
agency settings may wish to signal more credible report-
ing by hiring a Big 4 auditor, other CEOs in these settings
may feel a Big 4 auditor is either unnecessary given close
family ties or unwanted because of the gains from
extracting private bene?ts which could be reduced by a
Big 4 audit.
2
Our research is motivated by the need to understand
agency con?icts facing private ?rms. Private ?rms make
up a signi?cant portion of the economic activity in Norway
and nearly all other countries, yet prior research focuses
primarily on public ?rms. Given the sometimes vast differ-
ences between public and private ?rms (e.g., Ball & Shi-
vakumar, 2005; Beatty, Ke, & Petroni, 2002; Chaney, Jeter,
& Shivakumar, 2004), it is not apparent without testing
that results for public ?rms will generalize to private ?rms.
Thus, private ?rms offer an economically important sam-
ple worth testing. While the bene?ts to understanding
agency con?icts accrue directly to the ?rm’s investors, they
will also be important to many other stakeholders (e.g.,
creditors, employees, suppliers, and customers), regulatory
bodies supervising auditors and ?rms’ ?nancial reporting,
and society in general.
A sample of private (as opposed to public) ?rms may
also offer a stronger test of agency con?icts related to own-
ership structure and family relationships. As we discuss in
more detail in Hypotheses, prior research sometimes pro-
vides con?icting evidence or con?icting predictions for
the impact of ownership structures and family relation-
ships on agency con?icts. Our study provides a potentially
strong setting for testing agency con?icts because private
?rms exhibit heterogeneous ownership characteristics
and family relationships. Public ?rms are more homoge-
neous, including wide-spread ownership, relatively low
CEO ownership, and fewer family ties between managers
and shareholders and between managers and board mem-
bers. Private ?rms offer interesting ownership structures
that potentially increase our understanding of the relation
between agency con?icts and the supply of auditor effort.
For example, private ?rms show considerable variation in
ownership percentages by second largest shareholders.
This allows us to provide a meaningful test of the impact
of agency con?icts among shareholders (i.e., monitoring
of largest shareholders by second largest shareholders).
Private ?rms also show greater variation in their choice
of auditor (only 18.1% choose a Big 4 auditor). Nearly all
public ?rms opt for a Big 4 auditor, limiting the ability to
empirically test signaling through demand for a high-qual-
ity auditor.
Related to tests of the supply of auditor effort, a single-
country setting (Norway) controls for cross-country varia-
tion in audit practices and fees and the strength of legal
institutions. Cross-country differences could easily con-
found inferences. Norway also offers an environment
where the impact of litigation on audit fees is relatively
limited (Hope & Langli, 2010). This increases our ability
to make more reliable inferences from using audit fees to
measure auditor effort, and adds to calls for research to
better understand the role of ?rmgovernance in explaining
audit fees (Hay, Knechel, & Wong, 2006).
Finally, given the unique data we use in this study, we
are able to measure attributes of ownership structure
and family relationships that have been dif?cult to mea-
sure in the past. Speci?cally, for all private limited liability
?rms we have detailed information available to compute
both direct and ultimate ownership for each owner, board
member, and CEO.
3
In addition, we have detailed data on
family relationships among all owners, board members,
board chairs, and CEOs (based on both marriage and blood
lines, going back four generations and extending out to
fourth cousin). To our knowledge, no prior study has been
able to test the effects of family relationship using such de-
tailed data. These data, based on merging databases using
social security numbers, are obtained through special
2
As per Norway’s Companies Act (§ 7-1), the annual meeting (or General
Assembly) of the shareholders elects the auditor. While the selection of the
auditor is technically the responsibility of the General Assembly (or in
practice the board), it is almost certainly the case that the CEO plays a
signi?cant role in our setting. There is a large literature on how CEOs
in?uence the selection of board members and more generally exercise
‘‘power’’ over the board (see, e.g., Bebchuk & Fried 2003). Even for large
public ?rms in the US and UK the CEO has an impact on which auditor is
selected (e.g., Beattie & Fearnley 1995; Carcello, Neal, Palmrose, & Scholz
2011; Firth 1999). For our sample of smaller private ?rms, the CEO is
expected to have a greater in?uence on auditor selection because the
boards are typically smaller, CEO ownership is common (the CEO owns
shares in 78% of our sample ?rms), and the largest owner often is related to
the CEO.
3
For example, suppose an investor owns 30% of ?rm A and30% of ?rm B,
and ?rm A in turn owns 30% of ?rm B. The investor’s direct ownership in
?rm B is 30%, while her ultimate ownership is 39%. Note that our data also
account for cross-holdings.
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 501
permission from the Norwegian government. Having these
data allows us to simultaneously test multiple sources of
agency con?icts in a single test. In contrast, prior studies, be-
cause of data limitations, have been able to focus only on a
particular test variable (and thus leave out potentially
important ownership and family details). The full model bet-
ter controls for multiple agency settings. We show that the
ability to simultaneously test the association of these vari-
ables on auditor effort and ?rms’ selection of the auditor
can affect conclusions in some cases.
We next brie?y discuss the importance of accounting
and auditing, the role of auditing in private ?rms, and
the importance of auditing in the private ?rm market.
Hypotheses provides hypotheses of the expected impact
that agency con?icts have on the supply of auditor effort
and the demand for a Big 4 auditor. Research design intro-
duces our research design. Data on ownership and family
relationships provides details on the data underlying our
study. We present empirical results in Sample and results,
and Concluding remarks concludes.
Background, related research, and institutional setting
Importance of auditing in private ?rm setting
An external audit plays a role in corporate governance
by providing an independent assessment of the accuracy
and fairness with which ?nancial statements represent
the results of operations in accordance with generally ac-
cepted accounting principles. While researchers commonly
consider the role of auditing for public ?rms, audits for pri-
vate ?rms can also play a role. Private ?rms typically dis-
close less non-accounting information, potentially
increasing the importance of ?nancial accounting informa-
tion to external providers of capital in monitoring manage-
rial activities. In addition, managerial activities of public
?rms are partially constrained by market-based mecha-
nisms. For example, public ?rms are more susceptible to
takeovers, indicating that such mechanisms help control
for agency con?icts (Lennox, 2005). In the absence of mar-
ket-based measures of ?rm-value (and other sources of
information such as ?nancial analysts), high-quality
reporting may be particularly relevant for evaluation of
managerial performance and to support personnel and
compensation decisions (Indjejikian & Matejka, 2009),
resulting in a demand for high-quality audits (Chaney et
al., 2004).
4
Brief review of prior research
Although there is a huge auditing literature, there is rel-
atively limited prior evidence on auditing issues in private
?rms.
5
Some prior studies investigate private ?rms’ choice to
have an audit. Carey, Simnett, and Tanewski (2000) ?nd that
nonfamily involvement correlates positively with the volun-
tary demand for auditing in Australian family businesses.
Blackwell, Noland, and Winters (1998) ?nd that private US
?rms that elect to have their ?nancial statements audited
pay signi?cantly lower interest rates than nonaudited pri-
vate ?rms. Similarly, Allee and Yohn (2009) use National
Survey of Small Business Finances data from 2003 to exam-
ine the ?nancial reporting practices of small privately held
U.S. businesses. They ?nd that ?rms with audited ?nancial
statements enjoy greater access to credit. Using a large sam-
ple of private ?rms from the World Bank Enterprise Surveys,
Hope, Thomas, and Vyas (2011) show that ?rms which have
their ?nancial statements reviewed by an external auditor
experience easier access to external ?nancing and obtain
those funds at lower costs.
In a recent study, Lennox and Pittman (2011) make use
of a natural experiment in the United Kingdom. Starting in
2004, an external audit is no longer required for private
UK companies. Lennox and Pittman focus on the ?rms
that are audited under both regimes (i.e., the ?rms that
reveal their preference to be audited). They ?nd that these
companies attract upgrades to their credit ratings. In con-
trast, companies that no longer submit to an audit suffer
credit rating downgrades. These ?ndings provide further
support for the usefulness of auditing in a private ?rm
setting.
There is also research on how audit fees relate to own-
ership characteristics, primarily for publicly traded compa-
nies. For example, Chan, Ezzamel, and Gwilliam(1993) ?nd
that ownership control, de?ned in their study as ‘‘directors’
bene?cial and non-bene?cial shareholdings and all dis-
closed shareholdings in excess of 5%,’’ is negatively related
to levels of audit fees for their full sample of 280 publicly
traded U.K. companies from 1987. In contrast, they ?nd
no signi?cant effect for smaller auditees (i.e., ?rms that
are more similar to the private ?rms in our sample).
Mitra, Hossain, and Deis (2007) examine 358 NYSE-
listed manufacturing and retail companies that were au-
dited by Big 5 auditors in year 2000 (i.e., ?rms that are very
different from the small private ?rms in our sample). They
document that audit fees are positively associated with dif-
fused institutional stock ownership (i.e., having less than
5% individual shareholding) and negatively associated with
institutional blockholder ownership (i.e., having 5% or
more individual shareholding).
4
According to Van Tendeloo and Vanstraelen (2008), a choice to contract
for high-quality auditing (e.g., proxied by choosing a Big 4 auditing ?rm)
could signal ?nancial reporting quality and perhaps deter a rigorous tax
audit in the private ?rm market. They further argue that private ?rms may
also want to convince suppliers, clients, or employees of the credibility of
their ?nancial statements. This may be especially important in an
environment like Norway where ?nancial statements of all limited liability
companies (public and private) are publicly available. To illustrate, most
information from the income statements and the balance sheet, in addition
to names and shareholdings of owners, CEOs, and board members, is freely
available through websites (e.g., www.proff.no). Information on family
relationships between owners, CEOs, and board members, however, is not
publicity available.
5
The usefulness of accounting information for private ?rms has been
shown in recent studies including Chen, Hope, Li, and Wang (2011) and
Indjejikian and Matejka (2009). Chen et al. ?nd that emerging market ?rms
with higher ?nancial reporting quality exhibit greater investment ef?-
ciency. Indjejikian and Matejka further show that accounting information is
used in compensation contracting by US private ?rms. Studies that
compare the ?nancial reporting quality of public and private ?rms include
Ball and Shivakumar (2005), Burgstahler, Hail, and Leuz (2006), Asker,
Farre-Mensa, and Ljungqvist (2011), and Hope, Thomas, and Vyas (2012).
502 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
Hay, Knechel, and Ling (2008) show that an indicator for
whether there is a single shareholder who controls 20% or
more of the company’s stock is positively related to audit
fees for New Zealand public companies in 1995 (sample
size: 130). However, there is no signi?cant correlation in
2005 (sample size: 83).
As discussed, almost all prior research is on public, not
private ?rms. In addition, given that all private limited lia-
bility companies in Norway are audited during our sample
period, voluntary demand for audits per se is not an issue
for our study. The choice we examine is the demand for a
high-quality auditor (Big 4). Equally important, no prior
study has employed detailed data on family relationships
among shareholders, board members, and CEOs to mea-
sure agency con?icts, and to our knowledge few prior
studies have as detailed measures of ownership as this
study.
6
Finally, in contrast to most prior research, we exam-
ine the effects of several agency cost related factors simul-
taneously. In sum, although there is suf?cient extant
research ensuring that we have theoretical support for
our hypotheses, we believe our study ?lls a void by exam-
ining a large sample of private ?rms and testing for poten-
tially important effects for which prior research has had
limited data.
Institutional setting
In Norway, external auditing of all private limited liabil-
ity companies is mandated by the government. In the rest
of Europe, only private companies that meet certain size
criteria are required to have their ?nancial statements au-
dited.
7
The statutory auditor is expected to provide different
stakeholders of the company assurance concerning the accu-
racy of the ?nancial statements, the non-existence of ?nan-
cial statement fraud, and the going concern status (e.g., Van
Tendeloo & Vanstraelen, 2008).
Norwegian auditing standards follow International
Standards of Auditing (ISA). The ISAs require auditors to as-
sess several agency con?icts when assessing audit risk. For
example, from ISA 315, an auditor assesses risk of material
misstatement by understanding the entity’s ownership
and governance structures and the way that the entity is
structured (paragraph 11). This could include noting
whether the client ?rm has an owner-manager, more than
one owner, dispersed ownership, etc.
8
In addition, IAS 315
suggests that assessing risk involves understanding the rela-
tions between owners and other people, which likely in-
cludes family relationships (paragraph A23). Referring to
the audits of smaller ?rms (which are much more likely to
be private), IAS 315 states that the presence of an owner-
manager may mitigate certain risks arising from a lack of
segregation of duties (paragraph A76). We believe that our
investigation of the relation between audit and ownership
characteristics and family relationships provides a unique
setting for understanding the interplay between agency con-
?icts and auditing in the private ?rm segment of the
economy.
Hypotheses
The framework for our tests can be understood by con-
sidering that there is a principal (i.e., the party for whom
?nancial statements are prepared such as owners, credi-
tors, regulators, suppliers) and two agents (the manager
and the auditor). We observe the actions of these two
agents – the manager’s demand for a Big 4 auditor and
the auditor’s supply of effort – to understand the impact
of ownership characteristics and family relationships on
agency costs.
In the ?rst principal-agent setting, the manager under-
stands that her potentially suboptimal actions are unob-
servable to the principal, and therefore the principal will
impose a monetary penalty on the manager. To avoid this
penalty in high agency cost settings, the manager is will-
ing to hire a higher-quality auditor to provide a more
credible signal that ?nancial statements are free of mate-
rial misstatement and that the manager has committed to
refrain from siphoning private bene?ts. Previous research
suggests that larger auditors provide higher quality audits
(e.g., DeAngelo, 1981; Palmrose, 1988) and that Big N
auditors in particular provide higher-quality outcomes in
a variety of settings (e.g., Becker, DeFond, Jiambalvo, &
Subramanyam, 1998; DeFond, 1992; Mansi, Maxwell, &
Miller, 2004). These arguments lead to the expectation
that ?rms in higher agency cost settings are more likely
to demand a Big 4 auditor. This is the typical principal-
agent framework put forth in most auditor selection
studies.
9
The second agency setting involves the auditor acting as
an agent, having a preference for compensation and a dis-
utility for effort. In other words, the optimal action for the
auditor is to supply the minimum amount of effort needed
to ‘‘obtain reasonable assurance about whether the ?nan-
cial statements as a whole are free from material misstate-
ment, whether due to fraud or error’’ (ISA, 2010, paragraph
5). Doing so maximizes the auditor’s utility. The higher the
audit effort is, the less likely type II errors occur in audit re-
ports (e.g., Dye, 1993; Laux & Newman, 2010). Consistent
with prior research (e.g., Davis, Ricchiute, & Trompeter,
1993; Whisenant, Sankaraguruswamy, & Raghunandan,
2003), we measure auditor effort using audit fees. Due to
transparency of the audit market and the large number
of auditors and audit ?rms in Norway (Financial Supervi-
sory Authority of Norway, Annual Report, 2007), the mar-
6
Lennox (2005) investigates the relation between management owner-
ship and audit ?rm size among private UK companies.
7
Norway and Sweden were the two last countries in Europe that
abandoned the requirement that all limited liability companies should
disclose audited ?nancial statements. Starting November 10, 2010 in
Sweden and January 1, 2011 in Norway, the (very) smallest ?rms may
decide not to engage an auditor.
8
Standard auditing textbooks provide further support. For example,
Arens, Elder, and Beasley (2012) note that the distribution of ownership has
the potential to affect audit risk.
9
The obvious counterargument is that hiring a Big 4 auditor is more
costly. For example, Chaney et al. (2004) conclude that UK private ?rms
choose auditors that minimize their audit fees. However, they do not
examine whether audit fees vary with agency con?icts as in our study. We
contend that ?rms in higher agency cost settings are more willing to hire a
Big 4 auditor (and incur the higher audit cost) because of the bene?ts
received from signaling more credible reporting.
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 503
ket for auditing services is highly competitive. Audit fees
should therefore re?ect the effort assessed by the auditor
in assuring the accuracy of ?nancial statements.
10
In the next section, we develop our six hypotheses. The
?rst three focus on ownership characteristics – ownership
concentration among all shareholders, ownership by sec-
ond largest shareholder, and fraction of shares held by
the CEO. The second three hypotheses focus on family rela-
tionships – family ties between the CEO and largest own-
ing family, percentage of board members related to
largest owning family, and family ties between the CEO
and board members. For each of these six settings, we con-
sider whether agency concerns affect managers’ decision
to hire a Big 4 auditor or auditors’ supply of effort.
Ownership structures
Ownership concentration
Prior research provides evidence that managers, when
left unmonitored, are more likely to manage earnings,
commit fraud, or make suboptimal investment decisions
(e.g., Biddle & Hillary, 2006; Hope & Thomas, 2008). Thus,
shareholder monitoring is an important mechanism by
which agency costs can be reduced. However, while all
shareholders have the responsibility to monitor manage-
rial activities, the bene?ts of doing so by any individual
shareholder are proportional to the percentage of shares
owned. As the percentage of ownership by individual
shareholders increases (i.e., concentration increases), the
more willing individual shareholders are to incur neces-
sary monitoring costs.
Monitoring by a large shareholder could take many
forms. Perhaps the most commonly discussed means of
monitoring discussed in the literature involves a large
shareholder having a seat on the board. Several studies
show in a variety of contexts the board’s role in monitoring
managers (e.g., Adams, Hermalin, & Weisbach, 2010;
Beasley, 1996; Dechow, Sloan, & Sweeney, 1996; Fama,
1980; Fama & Jensen, 1983; Byrd and Hickman, 1992;
Anderson, Mansi, & Reed, 2004; Laksmana, 2008, to name
just a few).
11
Other forms of direct monitoring would be a
large shareholder actively participating in the ?rm’s opera-
tions or having routine meetings with managers. As the pro-
portion of ownership increases, the more bene?cial it is for
large shareholders to engage in these types of costly direct
monitoring activities. Large shareholders can also serve to
block business decisions that may be considered suboptimal
(e.g., aggressive expansion through negative net present va-
lue projects). Doing so involves an investment in time and
expertise by the shareholder to understand the conse-
quences of major business decisions. Large shareholders
are also likely to have more control over the ?rm’s dividend
(or capital distribution) policy, as a way to further discipline
managers’ actions.
When ownership is widely dispersed, it is economically
less feasible for any individual shareholder to incur signif-
icant monitoring costs, because she will receive only a
small portion of bene?ts. This is the typical ‘‘vertical
agency cost’’ (e.g., Gogineni, Linn, & Yadav, 2010) argument
(i.e., con?icts between managers and owners) and leads to
the prediction that agency costs are expected to be lower
as ownership concentration increases.
12
When agency costs
are lower, we expect that auditors supply less effort and
there is less demand for a Big 4 auditor.
Hyphothesis 1a. As ownership concentration increases,
audit fees decrease.
Hyphothesis 1b. As ownership concentration increases,
choice of Big 4 auditor decreases.
Second largest shareholder
While the previous discussion explains the need for
shareholders to monitor managers, the literature also
establishes the need for shareholders to monitor one an-
other. For example, controlling shareholders have the abil-
ity to exploit minority shareholders in closely-held
corporations (e.g., Burkart, Gromb, & Panunzi, 1997,
1998; Laeven & Levine, 2008; Nagar, Petroni, & Wolfenson,
2011). Such exploitation can include higher compensation
to controlling shareholders, misappropriation of assets,
and dilution of minority shareholders’ interests through
the issuance of stock or dividends (Gogineni, Linn, & Yadav,
2010). As the ownership stake of a second shareholder in-
creases, so does her ability and willingness to effectively
monitor the largest shareholder. The monitoring activities
by the second largest shareholder would be similar to
those used by the largest shareholder to monitor managers
(see discussion above in Ownership concentration).
Pagano and Roell (1998) specify conditions under which
large shareholders monitor each other, reducing expropri-
ation and improving ?rm performance. Their theoretical
model predicts that expropriation of minority shareholders
is likely to be less severe when the ownership stake of non-
controlling shareholders is more concentrated, as such
concentration makes it easier and more effective to moni-
tor the controlling shareholder (see also Bloch & Hege,
2001; Gogineni et al., 2010; Volpin, 2002). This is the
10
Some prior research in other settings suggests that it is important to
distinguish the component of audit fees that also re?ects compensation for
litigation risk. We do not view this distinction important for our tests. In
addition, both the litigation risk and reputation risk of auditors are
relatively low for private ?rms in Norway. For a detailed discussion of this
issue, see Hope and Langli (2010), who examine all court cases and other
legal proceedings against auditors over a 60-year period and conclude that
auditors face much lower litigation risk in Norway than in other more
litigious environments. It is important to note that, even in an environment
with low litigation, there is still a role for agency costs. For example, debt
and equity ?nancing will be more costly when agency costs are high. To the
extent that companies in settings with high agency problems are able to
signal more credible reporting, ?nancing will be less costly and more
accessible. Firms can signal this credibility with an audit. As another
example, suppliers may also be concerned about the viability of the ?rm
when deciding whether to enter long-term contracts. To the extent that the
supplier can adequately rely on ?nancial reports to assess the ?rm’s long-
run viability and stability, they are more willing to enter those contracts,
increasing the operating ef?ciency and pro?tability of the ?rm (e.g., Dou,
Hope, & Thomas, 2012).
11
See Bhagat and Black (1999), Hermalin and Weisbach (2003), and
Adams et al. (2010) for surveys on corporate boards.
12
An alternative prediction is that greater ownership concentration leads
to entrenchment, resulting in higher agency costs.
504 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
typical ‘‘horizontal agency cost’’ (e.g., Gogineni et al., 2010)
argument (i.e., con?icts between majority and minority
shareholders) and leads to the prediction that as owner-
ship by the second largest shareholder increases, agency
costs decrease.
Hyphothesis 2a. As ownership of the second largest
owner increases, audit fees decrease.
Hyphothesis 2b. As ownership of the second largest
owner increases, choice of Big 4 auditor decreases.
CEO ownership
A dominant belief in the literature is that as CEO own-
ership increases, her incentives align more with those of
other shareholders, reducing the agency problem that
arises from separation of ownership and control (e.g.,
Jensen & Meckling, 1976; Tirole, 2006). This is known as
the alignment effect. In our sample, CEO ownership is high
compared with prior studies (e.g., mean CEO ownership is
49% of the shares in our study). We predict that as CEO
ownership increases, agency costs decline and therefore
auditors supply less effort. Consistent with the belief that
CEO ownership reduces agency costs, we also predict that
?rms with high CEO ownership are less likely to employ
a Big 4 auditor.
Hyphothesis 3a. As CEO ownership increases, audit fees
decrease.
Hyphothesis 3b. As CEO ownership increases, choice of
Big 4 auditor decreases.
Family relationships
CEO and major shareholder
Major shareholders are often family members of the CEO
for private ?rms. There are interesting competing hypothe-
ses when the CEO is related to the major shareholder. Be-
cause of the family relationship, these shareholders no
longer act as independent monitors in disciplining CEOs’
decisions. In addition, family-controlled ?rms are likely to
suffer fromgreater horizontal agency costs. It may be easier
for major shareholders, who are family members of the
CEO, to extract private bene?ts from minority shareholders
or other stakeholders (Anderson & Reeb, 2004; DeAngelo &
DeAngelo, 2000; Morck, Shleifer, & Vishny, 1988). The rea-
son it may be easier to extract these bene?ts is that major
family owners typically have strong in?uence over choos-
ing members of the board (Johannisson & Huse, 2000). Con-
sequently, the monitoring effectiveness of the board may
be impaired when its composition is determined primarily
by the CEO’s family. These arguments would support the
idea that agency costs will increase when there is a family
relation between the CEO and major shareholder. In this
case, auditors are expected to supply more effort, and ?rms
are expected to demand Big 4 auditors.
An alternative view is that family member CEOs are less
likely to act in ways that opportunistically harm other
family members. That is, installing a family member as
the CEO could be a mechanism through which family-
owned companies can increase their monitoring of man-
agement and reduce the need for external monitoring. If
this effect dominates, the agency costs are smaller when
the CEO is a family member because familial ties are likely
to create closer alignment of the CEO’s preferences with
those of family owners.
The demand for a Big 4 in the presence of CEO-major
shareholder family relationship also presents interesting
counter-arguments. One the one hand, demand for a Big
4 auditor could increase to the extent that this family rela-
tionship increases agency costs. Agency costs would be
higher for reasons discussed above (e.g., lack of indepen-
dent monitoring by major shareholders and expropriation
of minority shareholders). In this setting, CEOs potentially
bene?t by signaling their commitment to higher-quality
reporting.
On the other hand, demand for a Big 4 could decrease
under at least two conditions. First, while hiring a Big 4
auditor has commitment value for the ?rm (and the
CEO), the CEO in a family-relationship setting may wish
to reduce audit cost by not hiring a Big 4 auditor. Major
shareholders’ family relationship with the CEO may negate
the need for costly independent veri?cation by Big 4 audi-
tors. The saved resources by using a less costly auditor in-
crease ?rm value, which is in the best interest of both the
CEO and major shareholders (who are in the same family).
Second, the CEO also has incentives not to demand a high-
er-quality auditor when this means that her ability (or her
family’s ability, including the major shareholder’s ability)
to extract private bene?ts from the ?rm would be limited
through such a hire. Presumably, Big 4 auditors would
have a greater ability to limit these private bene?ts.
Thus, for this CEO family-related agency setting, the de-
mand for a Big 4 auditor re?ects the trade-off between the
bene?ts from signaling higher-quality reporting versus the
costs of additional audit fees and reduced consumption of
private bene?ts. Because of competing arguments, we state
our fourth hypothesis as two-sided:
Hyphothesis 4a. When a family relationship exists
between the major shareholder and the CEO, audit fees
are affected.
Hyphothesis 4b. When a family relationship exists
between the major shareholder and the CEO, choice of
Big 4 auditor is affected.
Board independence
Boards are meant to protect shareholders’ assets and
the interests of the company’s other stakeholders (e.g.,
creditors and employees). In this sense, boards are directed
to monitor the activities of managers. An extensive litera-
ture exists which supports the notion that more indepen-
dent boards more effectively monitor managers’
activities. Firms with more independent boards commit
less ?nancial statement fraud (Beasley, 1996) and have less
earnings management or provide fewer discretionary
accruals (Dechow et al., 1996; Jaggi, Leung, & Gul, 2009;
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 505
Peasnell, Pope, & Young, 2000; Xie, Davidson, & DaDalt,
2003).
However, consistent with our two-sided predictions for
H
4a
, as the proportion of board members from the CEO’s
family increases, audit effort could either increase or de-
crease. With respect to the demand for a Big 4 auditor,
the relation is also not obvious because the CEO faces con-
?icting incentives. The additional (potentially unneces-
sary) costs associated with hiring a Big 4 auditor and the
reduced ability of the CEO or her family board members
to privately bene?t from these family relationships reduce
the expected demand for a Big 4 auditor in this higher
agency cost setting. As a result, the demand for a Big 4
auditor may be confounded in the presence of family rela-
tionships between the CEO and board members.
We propose two hypotheses related to board indepen-
dence. First, we predict that family relationships between
major shareholders and the board imply that owners have
insiders to monitor managers, which makes audited infor-
mation less important. In other words, fewer agency con-
?icts reduce the need for auditing. This setting would
result in the reduced supply of auditor effort and less de-
mand for a Big 4 auditor.
Second, if there are family ties between the CEO and
board members, given that there are competing arguments
as to whether the board will act more or less indepen-
dently of the CEO, our hypothesis is two-sided. We sum-
marize these hypotheses as:
Hyphothesis 5a. As the proportion of board members
from the largest owning family increases, audit fees
decrease.
Hyphothesis 5b. As the proportion of board members
from the largest owning family increases, choice of Big 4
auditor decreases.
Hyphothesis 6a. As the proportion of board members
from the CEO’s family increases, audit fees are affected.
Hyphothesis 6b. As the proportion of board members
from the CEO’s family increases, choice of Big 4 auditor is
affected.
Research design
For our auditor effort tests, we use model (1) to test our
six hypotheses of the relation between agency costs and
the supply of auditor effort after controlling for numerous
?rm and audit characteristics that explain normal audit
fees.
13
Speci?cally, we regress the log of raw audit fees
(LNFEE) on our six test variables as well as 24 client-?rm
characteristics, ?ve audit-?rm variables, and year and indus-
try ?xed effects (?rm subscripts omitted).
LNFEE
t
¼ a
0
þa
1
OWNER CONC
t
þa
2
OWNER SECOND
t
þa
3
OWNER CEOt
þa
4
FAM CEO=OWNER
t
þa
5
FAM BOARD=OWNER
t
þa
6
FAM BOARD=CEO
t
þa
7
LNSALES
t
þa
8
LNEMPLOY
t
þa
9
LEV
t
þa
10
CHLEVt
þa
11
INVREC
t
þa12GROWTH
t
þa
13
ROA
t
þa
14
INCPIC
t
þa
15
DECPIC
t
þa
16
INTANG
t
þa
17
UNITEMS
t
þa
18
NOINDt
þa
19
FOREIGN
t
þa
20
LNSUB
t
þa
21
FYE
t
þa
22
INVESTMENTS
t
þa
23
ACQUISITION
t
þa
24
LOSS
t
þa
25
CURRATIO
t
þa
26
BIG4
t
þa
27
GAAP
t
þa
28
CHGAAP
t
þa
29
INDSPEC
t
þa
30
CHAUDITOR t þa
31
LASTYRAUDIT
t
þa
32
MODOPINIONS
t
þa
32
CHCEO
t
þa
33
CITY
t
þRa
34;c
COUNTY
c
þRa
35;i
IND
i
þRa
36;y
YR
y
þv
t
ð1Þ
Appendix lists all variables employed and their de?ni-
tions. a
1
through a
6
test hypotheses H
1a
through H
6a
, while
the remaining variables serve as controls. We measure
ownership concentration (OWNER_CONC) using the Her?n-
dahl index and expect that as ownership concentration in-
creases, agency costs decrease (a
1
< 0). To test the second
hypothesis, we use ultimate ownership of the second larg-
est owner (OWNER_SECOND), and expect that as ownership
increases, agency costs decrease (a
2
< 0).
14
For the third
hypothesis, we use the fraction of shares owned by the
CEO as our primary measure (OWNER_CEO) and predict a
negative effect of CEO ownership on agency costs (a
3
< 0).
In Section Non-linearity of CEO ownership, we test whether
this relation is non-linear.
For our fourth hypothesis, we use an indicator variable
equal to one if the CEO is a member of the largest owning
family (FAM_CEO/OWNER). We predict that family owner-
ship has an effect on agency costs (a
4
–0). We refrain from
making a directional hypothesis due to the opposing forces
that are at work. For the ?fth hypothesis, we measure fam-
ily in?uence of the largest shareholder on the board as the
proportion of board members related to the largest share-
holder (FAM_BOARD/OWNER). We expect a negative rela-
tion (a
5
< 0). Finally, for our sixth hypothesis the test
variable is the proportion of board members related to
the CEO (FAM_BOARD/CEO). FAM_BOARD/CEO can thus be
interpreted as a measure of the strength of family ties be-
tween the CEO and the board.
15
As with our fourth hypoth-
eses, we refrain from a directional hypothesis (i.e., a
6
–0).
13
This approach is similar to ?rst estimating abnormal audit fees as the
residuals from a regression of total audit fees on the control variables and
then using the residuals as the dependent variable in a regression on our
test variables. No inferences are affected if we use that approach instead.
14
We provide a caveat that there is a possible negative mechanical
relation between OWNER_SECOND and OWNER_CONC (see correlation in
Table 2 below). When excluding either of these variables from the model,
inferences remain the same for the demand model (discussed next). For the
effort model, we ?nd that OWNER_CONC becomes insigni?cant when
OWNER_SECOND is dropped from the test.
15
Note that FAM_BOARD/OWNER and FAM_BOARD/CEO measure the
degree of family in?uence rather than just the existence of such in?uence.
506 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
Our choice of control variables is guided by prior re-
search (e.g., DeFond, Raghunandan, & Subramanyam,
2002; Frankel, Johnson, & Nelson, 2002; Choi, Kim, & Zang,
2010; Hope & Langli, 2010) and we additionally employ
other variables relevant for our private ?rm setting. There-
fore, model (1) examines whether our ownership charac-
teristics and family relationships explain audit fees
beyond an extensive set of factors previously identi?ed in
the literature.
We include the natural log of sales (LNSALES), natural log
of number of employees (LNEMPLOY), leverage (LEV) and
change in leverage (CHLEV), proportion of inventory and
receivables to total assets (INVREC), sales growth
(GROWTH), return on assets (ROA), increases or decreases
in paid-in capital (INCPIC and DECPIC), proportion of intan-
gible assets to total assets (INTANG), the existence of special
or unusual items (UNITEMS), the number of industries in
which the ?rm operates (NOIND), the existence of foreign
operations (FOREIGN), natural logarithm of the number of
subsidiaries (LNSUB), a ?scal year-end other than December
31 (FYE), long- and short-term investments in securities
scaled by sales (INVESTMENTS), acquisitions during the year
(ACQUISITIONS), negative earnings (LOSS), current ratio
(CURRATIO), Big 4 audit ?rm (BIG4), use of regular GAAP
(GAAP; as opposed to simpli?ed GAAP for smaller enter-
prises), changed to/from regular GAAP from/to simpli?ed
GAAP during the year (CHGAAP), auditor industry special-
ization (INDSPEC), change of auditor during the year (CHAU-
DITOR), an indicator for the last year of the incumbent
auditor (LASTYRAUDIT), audit report modi?cations
(MODOPINIONS), change in CEO during the year (CHCEO),
and whether the client resides in a municipality of?cially
classi?ed as a city or not (CITY). In addition, we include
?xed effects for the county in which the client ?rm is lo-
cated (COUNTY), two-digit industry (IND), and year (YR).
To test our six hypotheses of the demand for a Big 4
auditor, we employ the following logit regression model
with control variables (?rm subscript omitted):
BIG4
t
¼ b
0
þ b
1
OWNER CONC
t
þ b
2
OWNER SECOND
t
þ b
3
OWNER CEO
t
þ b
4
FAM CEO=OWNER
t
þ b
5
FAM BOARD=OWNER
t
þ b
6
FAM
B
OARD=CEO
t
þ b
7
LNTA
t
þ b
8
LEV
t
þ b
9
UNSECURED
t
þ b
10
LOSS
t
þ b
11
FOREIGN
t
þ b
12
EXANTEFIN
t
þ b
13
INCPIC
t
þ b
14
ROA
t
þ b
15
CITY
t
þRb
16;c
COUNTY
c
þRb
17;i
IND
i
þRb
18;y
YR
y
þg
t
ð2Þ
b
1
through b
6
test hypotheses H
1b
through H
6b
, while
the remaining variables serve as controls. BIG4 is an indica-
tor variable for whether the ?rm employs a Big 4 auditing
?rm or a forerunner. We expect that our agency variables
relate to the demand for a Big 4 auditor in the same way
that they relate to auditor effort. Speci?cally, our indicator
variable for a Big 4 auditor is expected to relate negatively
to OWNER_CONC, OWNER_SECOND, OWNER_CEO, and
FAM_BOARD/OWNER. For our two CEO family-related vari-
ables (FAM_CEO/OWNER and FAM_BOARD/CEO), we provide
expectations consistent with those for auditor effort. That
is, family relationships between the CEO and either the
major shareholder (FAM_CEO/OWNER) or board members
(FAM_BOARD/CEO) may decrease or increase the demand
for Big 4 auditors.
We base the control variables on prior research that has
examined the choice of Big 4 auditors (e.g., Khurana &
Raman, 2004; Knechel, Niemi, & Sundgren, 2008). In
particular, we control for ?rm size (LNTA), leverage (LEV),
unsecured debt (UNSECURED), loss ?rms (LOSS), the per-
centage of foreign subsidiaries (FOREIGN), operating cash
?ow less net investments in tangible and intangible ?xed
assets scaled by current assets (EXANTEFIN), increase in
paid-in capital (INCPIC), return on assets (ROA), and
whether the client ?rm is located in a city or not (CITY).
Finally, the logit regression models include county, indus-
try, and year ?xed effects. As an alternative speci?cation,
we additionally include controls for four alternative
ownership characteristics ( INSTITUTIONAL, STATE, INTER-
NATIONAL, and INDUSTRIAL).
Data on ownership and family relationships
We obtain our data from two sources. First, ?rm-spe-
ci?c information, such as the ?rm’s ?nancial information,
auditor, CEO, board members, and owners, comes from
Experian AS. Experian collects information from the
Brønnøysund Register Center (BRC), which is an adminis-
trative agency responsible for a number of national control
and registration schemes for business and industry. These
data are publicly available. Second, information on family
relationships comes from the National Register Of?ce
(NRO). In contrast to BRC data, NRO data are generally
not publicly available. Speci?cally, the social security num-
bers of any person having roles as owners, CEOs, or board
members are not available unless special permissions are
obtained. Family relationships among individuals are also
not publicly available. We gained access to these data
through a contract with the Centre for Corporate Gover-
nance Research (CCGR) at the Norwegian Business School.
CCGR obtained permissions to gather and merge data from
the two sources in accordance with con?dentiality and se-
crecy rules set forth by the Norwegian Data Inspectorate,
an independent administrative body that is set up to en-
sure that private and public storage of data are in accor-
dance with the Norwegian Personal Data Act.
In Norway, all limited liability ?rms, independent of size
and listing status, must send audited ?nancial statements
to the Center for Annual Accounts (CAA), which is part of
the BRC. According to the Accounting Act (paragraph 7–
26), limited liability ?rms that do not qualify as ‘‘small
enterprises’’ are required to disclose its 20 largest share-
holders and their shareholdings in the notes as long as indi-
vidual shareholdings exceed 1%. Firms that qualify as small
enterprises must disclose names and shareholdings of the
ten largest shareholders as long as the shareholder owns
5% or more. In addition, the Accounting Act requires disclo-
sure of names and shareholdings of all CEOs and members
of the Board and the Corporate Assembly. Experian collects
ownership information from the notes and to the extent
possible looks up the social security number of each owner.
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 507
The names of CEOs and board members are collected from
the Central Coordinating Register for Legal Entities (CCRLE),
which is also a part BRC. Data that identify CEOs and board
members (fromCCRLE) are merged with data on ownership
(from CAA) using the social security number for each indi-
vidual that occurs in the two databases.
16
The notes to the accounts specify direct ownership for
each ?rm. Utilizing the fact that direct ownership is avail-
able for all limited liability ?rms, data on ultimate owner-
ship for each ?rm has been calculated by tracing
ownership structures. Information on families’ ultimate
ownership and family in?uence on the CEO and boards
has been constructed by checking for relationships by
blood and/or marriage for any pair of persons that occurs
for owners, board members, CEOs, and Chairman using
data from NRO. Family relationships may cover up to four
generations. We assigned persons to the same family if
they (as of year-end 2007) are related through their par-
ents, grandparents, or by marriage. A single person without
relatives constitutes a family of size one.
Sample and results
Sample and descriptive statistics
Our sample selection procedure is explained in Table 1.
Our sample period spans 2000 through 2007. We include
only ?rms that are not publicly traded on a stock exchange.
We exclude ?rms with sales or total assets less than NOK
800,000 and ?rms with missing information concerning
audit fees or with fees below NOK 3000. We further elimi-
nate ?nancial ?rms, ?rms with missing information,
subsidiaries, and ?rm controlled by unknown owners or
the State.
17
Because the data provider did not provide a split
between audit and non-audit fees for years 2003 through
2005, our sample covers the 5-year period from 2000–2002
and 2006–2007. We have a ?nal sample of 185,109 ?rm-year
observations. Our tests are based on pooled samples, and we
use robust standard errors clustered at the ?rm level.
18
Panel A of Table 2 provides descriptive statistics. As can
be expected for a sample of private ?rms, ?rms are small
on average (mean SALES of NOK 16 million) but with a
large standard deviation. Audit fees (FEE) have a mean of
NOK 21,452, and 18.1% of ?rms employ Big 4 auditors. As
expected, ownership concentration (OWNER_CONC) is
quite high for private ?rms, with a mean Her?ndahl index
of 0.605. CEO ownership (OWNER_CEO) is also quite high
(mean of 49.2%). On average, the CEO comes from the larg-
est family (FAM_CEO/OWNER) in 70.6% of the cases, 71.2%
of board members come from the largest family
(FAM_BOARD/OWNER), and 15.1% of the board members
are family members of the CEO (FAM_BOARD/CEO). Panel
B presents Spearman correlation coef?cients among the
dependent variables, test variables, and primary control
variables.
19
As expected, most of the variables are signi?-
cantly correlated. However, unreported variance in?ation
factors and condition indices give no indication that multi-
collinearity is a serious issue in our multivariate analyses.
20
Important for our tests, and consistent with our unique data
on ownership and family relationships, we are interested in
testing the incremental effect of one agency characteristic
while holding the others constant.
Tests of hypotheses
Auditor effort tests
We report the results of tests of whether auditors’ fees
(i.e., auditors’ effort) vary with agency con?icts in private
?rms in Table 3. Column 1 reports our primary speci?ca-
tion with all test variables included simultaneously and
measured as continuous variables (except FAM_CEO/OWN-
ER which is an indicator variable). The regression has an
adjusted R
2
of 52.8% and we note that most of the control
variables are signi?cant in the predicted direction. For
example, audit fees are higher for ?rms that are larger,
have more ?nancial leverage, have unusual items, are more
diversi?ed (in terms of industries and subsidiaries), have
greater foreign operations, engage in acquisitions, and em-
ploy a Big 4 auditor.
21
The estimated coef?cient on OWNER_CONC is negative
and signi?cant, consistent with our prediction in H
1a
that
greater ownership concentration leads to easier monitor-
ing of managers, translating into less effort by external
auditors. Next, the coef?cient on OWNER_SECOND is nega-
tive and signi?cant, a ?nding which is in line with H
2a
and
recent research in ?nance suggesting that a strong second
shareholder can mitigate potential exploitation of minority
shareholders by the controlling shareholder. We further
?nd a negative and signi?cant coef?cient on OWNER_CEO,
consistent with H
3a
that CEO ownership aligns managers’
interests with those of the ?rms’ other stakeholders.
The next three variables, FAM_CEO/OWNER, FAM_
BOARD/OWNER, and FAM_BOARD/CEO, measure differences
in monitoring and independence. Whereas FAM_CEO/OWN-
ER captures the family relationship between the CEO and
major shareholders, the other two measure the degree of
16
For any pair of individuals, our data source has classi?ed the pair in
terms of a basic family relationship type. These types are parent,
grandparent, great grandparent, great-great grandparent, and marriage.
Using these basic types, we have generated 18 types of family relationships
between two individuals, including relationships like brother, sister, cousin,
uncle/aunt, and nephew/niece.
17
To mitigate the potential in?uence of outliers in such a large sample,
we remove observations with (absolute) studentized residuals exceeding
two. No inferences are affected by this choice. Similarly, no conclusions
change if we instead winsorize all continuous variables at the 1st and 99th
percentiles.
18
Results are consistent when employing the Fama–MacBeth approach.
19
Note that correlation coef?cients are provided for descriptive purposes
only and do not constitute tests of our hypotheses. Speci?cally, we are
testing our agency cost proxies against a proxy for abnormal audit fees;
hence it is essential to control for the normal factors that explain fees (i.e.,
risk and effort variables). It is also important to control for other agency-
related variables, as we do in our multiple regression tests.
20
Speci?cally, no variance in?ation factors are above 3.04 in any of the
speci?cations tabulated in Table 3 and Table 4.
21
The estimated coef?cient on INDSPEC is negative. The Pearson corre-
lation between audit fees and INDSPEC is signi?cantly positive (0.109) and
the negative regression coef?cient is likely caused by correlation with the
?rm size controls and BIG4 (i.e., INDSPEC is positive and signi?cant when
the model is estimated on a subset of ?rms with non-Big 4 auditors). In
addition, CHLEV has the opposite sign of expectations. No inferences are
affected if we exclude that (or any other) control variable from the test.
508 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
board independence based on family relationships. We
?nd a positive and signi?cant coef?cient for FAM_CEO/
OWNER. In other words, we document that when the CEO
is also a member of the largest family, auditors increase
their effort. Thus, our empirical ?ndings support the idea
that, because of the family relationship, these shareholders
no longer act as independent monitors in disciplining
CEOs’ decisions, thus increasing agency costs.
We further ?nd empirical support for H
5a
. Auditor effort
is decreasing in the proportion of board members related
to the largest family owner (FAM_BOARD/OWNER), consis-
tent with fewer agency con?icts between family owners
and the board. Finally, the empirical results indicate that
audit effort is increasing in the proportion of board mem-
bers related to the CEO (FAM_BOARD/CEO). This ?nding is
consistent with the loss of board independence when fam-
ily relationships exist between the CEO and board mem-
bers. Thus, to control for the increase in agency con?icts,
auditors exert more effort.
22
In the second column of Table 3, we report results of
regressions in which we replace continuous test variables
with indicator variables equal to one when equal to or
above the mean, zero otherwise. Results are similar for ?ve
the six tests. For FAM_CEO/OWNER (H
4a
), the coef?cients
changes from signi?cantly positive to only marginally sig-
ni?cant (t = 1.56).
In the next six columns, we provide results with one
test variable included at a time. We note two differences
from the full model results: OWNER_SECOND is no longer
signi?cant and FAM_CEO/OWNER is now signi?cantly neg-
ative (as opposed to signi?cantly positive). These differ-
ences highlight the importance of testing for multiple
agency settings simultaneously. As we discussed previ-
ously, prior studies have been limited to looking at one
or a few agency variables because of data limitations. We
are able to provide detailed tests of six agency settings.
Thus, we base our primary conclusions on the full model,
which we believe better controls for multiple agency
settings.
Demand for high-quality auditor tests
In Table 4, we report results of the analyses that mea-
sure the demand for higher-quality auditing by ?rms/own-
ers to reduce agency con?icts. As predicted in H
1b
, H
2b
, and
H
5b
, we observe negative and signi?cant coef?cients on the
three variables that capture non-CEO-related agency con-
?icts: OWNER_CONC, OWNER_SECOND, and FAM_BOARD/
OWNER. These results are consistent with those obtained
when testing audit fees (or auditor effort). In contrast,
the coef?cient for OWNER_CEO is negative, as predicted
in H
3b
, but not signi?cant.
For our two CEO family-related variables, FAM_CEO/
OWNER (H
4b
) and FAM_BOARD/CEO (H
6b
), we ?nd no sig-
ni?cant evidence that the demand for a Big 4 auditor is af-
fected when a family relationship exists between the CEO
and major shareholder or as the number of board mem-
bers related to the CEO increases. One interpretation of
these insigni?cant coef?cients is that, while hiring a high-
er-quality auditor may bring value to the ?rm as a whole
(in part by reducing agency con?icts), CEOs in these fam-
Table 1
Sample selection criteria.
2000 2001 2002 2003 2004 2005 2006 2007 Total
Private and public limited liability companies 136,140 138,745 141,146 141,991 144,426 157,710 180,709 192,011 1232,878
Exclusion criteria:
Public limited liability companies 582 571 529 492 463 459 471 384 3951
Operating revenue less than 800,000 NOK 55,180 56,092 57,081 56,142 55,151 64,325 81,038 87,257 512,266
Total assets less than 800,000 NOK 14,218 13,747 13,836 14,286 14,427 14,598 13,933 13,652 112,697
Missing information on audit fees, audit fees less than
3000 NOK, or unknown auditor
2465 2239 2034 71,071 74,385 78,328 13,904 2831 247,257
Missing information (e.g. industry af?liation, founding
date, prior year ?nancial statement) or ?rm using IFRS
6419 7400 6862 0 0 0 9252 11,690 41,623
Subsidiaries 11,088 12,131 12,199 0 0 0 16,274 25,242 76,934
Information on family ownership not available 7439 6134 7058 0 0 0 8421 8635 37,687
Aggregate ownership less than 50% 238 381 420 0 0 0 266 383 1688
Controlled by unspeci?ed owners or by the State 1298 724 678 0 0 0 781 693 4174
Observations with absolute value of studentized residuals
greater than two
1898 2008 2079 0 0 0 1694 1814 9493
Sample size 35,315 37,318 38,370 0 0 0 34,675 39,431 185,109
Note that the data provider does not provide a breakdown of audit remuneration into audit fees and non-audit fees for the years 2003, 2004, and 2005.
22
To assess the economic signi?cance of our results, we conduct two
tests. First, we consider the effect of a change from the ?rst to the third
quartile of the distribution of our test variables (holding other variables
constant). The effect for each translates into approximately NOK 1000 or
NOK 6000 in total. While this does not appear to be a large amount, note
again that the sample ?rms in our study are very small compared with
those examined in prior research. For example, using data from Statistics
Norway, the aggregate effect is equivalent to 18.8 man-hours based on
average salaries for full-time employed ‘‘senior of?cials and managers in
professional, scienti?c, and technical activities,’’ which includes auditors.
Compared with all full-time employed persons in Norway (in the profes-
sional, scienti?c, and technical activities), the aggregate effect is equivalent
to 34.9 (27.3) man-hours. In addition, we also compare these effects
directly with the average audit fees in our sample. The average effect for
our test variables is 4.7% and in aggregate the effect is 28.5% of average fees.
Second, we compare the effects of the test variables with those of the
control variables. In order to conduct such a comparison, we ?rst
standardize the estimated coef?cients so that they are all directly
comparable. Consistent with many prior studies, ?rm size is clearly the
most important determinant of fees as well as the demand for a Big 4
auditor. However, for the fee regressions, in untabulated analyses we
observe that OWNER_SECOND has an equal or greater effect than 12 of the
other control variables. The other test variables also fare quite well in such
a comparison. Overall we conclude that, although ?rm size is economically
more signi?cant than our test variables, the test variables have a non-trivial
effect on both audit effort and audit demand.
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 509
Table 2
Descriptive statistics.
Mean St. dev. 5% 25% Median 75% 95%
Panel A: Distribution of variables (N = 185,109)
Corporate governance
OWNER_CONC 0.605 0.296 0.194 0.355 0.500 1 1
OWNER_SECOND 0.219 0.184 0 0 0.226 0.35 0.5
OWNER_CEO 0.492 0.370 0 0.143 0.5 0.923 1
FAM_CEO/OWNER 0.706 0.455 0 0 1 1 1
FAM_BOARD/OWNER 0.712 0.341 0 0.4 1 1 1
FAM_BOARD/CEO 0.151 0.268 0 0 0 0.333 0.667
INSTITUTIONS 0.012 0.108 0 0 0 0 0
STATE 0.003 0.053 0 0 0 0 0
INTERNATIONAL 0.010 0.101 0 0 0 0 0
INDUSTRIAL 0.011 0.106 0 0 0 0 0
Financial variables
ACQUISITIONS 0.047 0.213 0 0 0 0 0
CHLEV À0.007 0.143 À0.221 À0.055 0 0.019 0.242
CURRATIO 1.679 1.727 0.401 0.968 1.246 1.772 4.067
DECPIC 0.012 0.111 0 0 0 0 0
EXANTEFIN 0.237 0.686 À0.871 À0.241 0.357 0.591 1.222
FEE 21.452 17.930 7 12 17 25 49
GROWTH 0.279 1.080 À0.307 À0.038 0.069 0.236 1.150
INCPIC 0.066 0.248 0 0 0 0 1
INTANG 0.020 0.066 0 0 0 0.006 0.109
INVESTMENTS 0.326 0.784 0.006 0.036 0.111 0.272 1.215
INVREC 0.181 0.170 0 0.061 0.147 0.249 0.487
LEV 0.267 0.286 0 0 0.180 0.457 0.832
LNFEE 2.868 0.602 1.946 2.485 2.833 3.219 3.892
LOSS 0.214 0.410 0 0 0 0 1
ROA 0.160 0.212 À0.131 0.042 0.124 0.256 0.561
SALES 16,242.9 182,098.9 1096.0 2523.0 5020.0 11,186.0 43,862.0
TA 13,860.4 133,659.0 935.0 1621.0 3139.0 7185.0 32,575.0
UNITEMS 0.036 0.186 0 0 0 0 0
UNSECURED À0.272 0.286 À0.884 À0.452 À0.153 À0.037 0.000
Other variables
AGE 12.814 12.100 2 5 10 16 33
EMPLOY 9.434 61.611 0 1.605 3.669 7.648 26.717
BIG4 0.181 0.385 0 0 0 0 1
CHAUDITOR 0.083 0.276 0 0 0 0 1
CHCEO 0.051 0.219 0 0 0 0 1
CHGAAP 0.006 0.077 0 0 0 0 0
CITY 0.558 0.497 0 0 1 1 1
FOREIGN 0.015 0.120 0 0 0 0 0
FYE 0.001 0.028 0 0 0 0 0
GAAP 0.036 0.187 0 0 0 0 0
INDSPEC 0.027 0.064 0 0.000 0.001 0.005 0.175
LASTYRAUDIT 0.084 0.277 0 0 0 0 1
MODOPINONS 0.287 0669 0 0 0 0 2
NOIND 1.514 0.843 1 1 1 2 3
SUB 0.267 1.507 0 0 0 0 1
Variables LNFEE 1 2 3 4 5 6 7 8 9 10 11 12 13
Panel B: Spearman correlation coef?cients (N = 185,109)
[1] BIG4 0.092
*
[2] OWNER_CONC À0.121
*
À0.050
*
[3] OWNER_SECOND 0.015
*
À0.030
*
À0.572
*
[4] OWNER_CEO À0.084
*
À0.052
*
0.589
*
À0.369
*
[5] FAM_CEO/OWNER À0.044
*
À0.042
*
0.286
*
À0.135
*
0.663
*
[6] FAM_BOARD/OWNER À0.171
*
À0.062
*
0.493
*
À0.227
*
0.376
*
0.416
*
[7] FAM_BOARD/CEO 0.029
*
0.002 À0.107
*
0.138
*
À0.031
*
0.342
*
0.273
*
[8] GROWTH À0.053
*
À0.006
*
À0.023
*
0.012
*
À0.015
*
À0.022
*
À0.039
*
À0.038
*
[9] LEV À0.001 0.015
*
À0.012
*
À0.002 À0.042
*
À0.012
*
À0.001 À0.040
*
À0.015
*
[10] LOSS 0.013
*
0.010
*
À0.011
*
À0.016
*
À0.022
*
À0.021
*
À0.041
*
À0.065
*
0.037
*
0.251
*
[11] ROA À0.101
*
À0.032
*
0.044
*
0.026
*
0.054
*
0.029
*
0.065
*
0.305
*
À0.107
*
À0.397
*
À0.273
*
[12] SALES 0.609
*
0.082
*
À0.130
*
0.035
*
À0.068
*
À0.038
*
À0.189
*
À0.129
*
0.171
*
À0.087
*
À0.063
*
0.017
*
[13] UNSECURED 0.062
*
À0.011
*
0.009
*
0.010
*
0.043
*
0.004 À0.045
*
À0.104
*
0.317
*
À0.524
*
À0.062
*
0.190
*
0.130
*
[14] EMPLOY 0.549
*
0.051
*
À0.149
*
0.051
*
À0.056
*
À0.030
*
À0.220
*
À0.173
*
0.184
*
À0.122
*
À0.050
*
À0.028
*
0.772
*
0.082
*
The table reports Spearman correlation coef?cients between dependent variables and tests and main control variables. The sample is described in Table 1
and variables are de?ned in the Appendix.
*
Signi?cant correlation coef?cient at p < 0.01.
510 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
ily-relationship settings may perceive that the additional
costs of hiring a Big 4 auditor is not necessary. Major
shareholders and board members could believe that inde-
pendent veri?cation of a CEO family member is not war-
ranted, allowing ?rm value to increase through reduced
audit fees. In addition, the CEO may have personal incen-
tives not to demand a Big 4 auditor to the extent that
these external monitors make it more dif?cult for the
CEO (or the CEO’s family) to extract resources from the
?rm. Thus, we attribute the weaker relation between
BIG4 and our CEO family-related agency variables to the
trade-off that exists when deciding the net bene?ts to hir-
ing a Big 4 auditor. These trade-offs are less apparent
when CEO family-related agency settings are not consid-
ered. We conclude that ?rms demand greater audit quality
in the presence of agency con?icts (that are not CEO fam-
ily-related).
23,24
We then provide results using indicator variables for
the test variables (measured as above) in Column 2. Results
are similar to those using continuous measures, except
that OWNER_CEO changes from insigni?cantly negative
(t = À1.26) to signi?cantly negative (t = À3.45), consistent
with H
3b
. The remaining columns tabulate results with test
variables included one at a time. Whereas FAM_CEO/OWN-
ER and FAM_BOARD/CEO were insigni?cant in the full mod-
el, they are now signi?cantly negative. These differences
from the full model again highlight an advantage of con-
trolling for multiple agency variables to provide more com-
plete inferences.
Additional analyses and sensitivity tests (untabulated)
Additional controls for ?rm size
In our tabulated analyses, we include two explicit con-
trols for client ?rm size (LNSALES and LNEMPLOY). We also
include additional variables that relate to size, including
BIG4, FOREIGN, LNSUB, ACQUISITIONS, and CITY. However,
given the potential importance of ?rm size as a control var-
iable, we conduct a number of additional tests. First, we re-
place LNEMPLOY (or LNSALES) with log of total assets in the
audit fee model. Second, to test for non-linear size effects,
we add square root terms of both LNSALES and LNEMPLOY
to the model. Third, we add the square of both the size
variables. Fourth, we include interaction terms between
the size controls and all variables included in the model.
Fifth, as a related test, we either break our size controls
into quintiles and introduce them separately as controls,
or we include these as both main and interacting effects
in the model. Finally, we exclude the smallest 10% of ?rms
(based on sales). In all these tests, results are consistent
with those tabulated.
Big 6 instead of Big 4
As the Big 4 audit ?rms have a relatively small market
share among private ?rms (18.1% in our sample), as an
additional test we replace Big 4 with Big 6. Speci?cally,
in addition to the international Big 4 ?rms we also include
two other large audit ?rms: BDO Noraudit and Inter Revis-
jon. Combined, the Big 6 ?rms have a market share of
29.9%. In fact, BDO Noraudit and Inter Revisjon are the
two largest audit ?rms in terms of number of private client
?rms in Norway, with Ernst and Young and PWC close
behind.
We ?rst repeat the auditor effort tests of Table 3 by
replacing BIG4 with BIG6 in the audit fee model. No infer-
ences are affected.
25
More importantly, we redo the auditor
demand test in Table 4 using Big 6 as the dependent vari-
able. Again, results are quite similar.
26
Non-linearity of CEO ownership
To explore the possibility that the effect of CEO owner-
ship is non-linear, we add the square root of CEO owner-
ship to the models. We ?rst note that adding this
variable does not affect the inferences for the other vari-
ables and further observe that the square root variable is
negative and marginally statistically signi?cant for the ef-
fort regression. Thus, there is some support of a non-linear
relation for CEO ownership.
Controls for CEO compensation
Executives can extract rents from the company by also
receiving excessive compensation. We test if our results
are robust to controlling for CEO compensation by includ-
ing either raw CEO compensation or CEO compensation
scaled by ?rm size. No inferences are affected in these
tests.
Time period effects
In the aftermath of well-known international account-
ing scandals and the introduction of SOX in the US, changes
were made to Norwegian accounting and auditing regula-
tions that were aimed at increasing accounting and audit
quality. The changes took effect between the two periods
covered in our sample. Even though we control for time-
period effects through year ?xed effects, we separately
examine the time periods 2000–2002 and 2006–2007 in
order to assess if the relations have changed. We ?nd al-
23
Landsman, Nelson, and Rountree (2009) model the ‘‘match’’ (or
‘‘misalignment’’) between the audit ?rm and its client and use this binary
variable in their empirical tests. Our regressions include proxies for all the
variables included in their model (i.e., LNSALES; LNEMPLOY; ACQUISITIONS;
CHLEV; INCPIC; DECPIC; GROWTH). No inferences are affected if we replace
the individual control variables with a binary variable computed following
Landsman et al.’s approach.
24
The change in the probability of a ?rm choosing a Big 4 auditor
increases by 55.5% (i.e., from 0.076 to 0.13) when all six test variables
change from the third to ?rst quartile (holding all other variables constant
at their mean values). The percentage changes in the probability of a ?rm
choosing a Big 4 auditor vary between 2.4% and 24.5% for the six variables.
We further observe that, using standardized coef?cients, OWNER_SECOND
has a greater effect than all non-size control variables. Again, the other test
variables continue do compare well.
25
As an alternative to controlling for Big 4 in the audit fee model, we run
the auditor effort tests separately for Big 4 and non-Big 4 client ?rms.
Although this is not one of our hypotheses, it is interesting to note that four
of the six estimated coef?cients are larger in magnitude for the Big 4 than
for the non-Big 4 subsample. This observation is consistent with Big 4 ?rms
being more sophisticated than other audit ?rms and thus responding more
strongly to variations in agency con?icts.
26
Speci?cally, in this speci?cation OWNER_CONC, OWNER_SECOND, and
FAM_BOARD/OWNER remain signi?cant in the hypothesized direction, while
OWNER_CEO is now signi?cantly negative (as predicted in H
3b
) and
FAM_CEO/OWNER also becomes signi?cantly negative (as discussed in H
4b
).
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 511
Table 3
OLS regression of audit fees on test variables and control variables.
Predicted sign Model
1 2 3 4 5 6 7 8
OWNER_CONC (H
1a
) À À0.032
***
À0.024
***
À0.039
***
(À4.16) (À5.22) (À8.37)
OWNER_SECOND (H
2a
) À À0.070
***
À0.032
***
0.004
(À7.15) (À7.56) (0.63)
OWNER_CEO (H
3a
) À À0.030
***
À0.023
***
À0.034
***
(À5.20) (À6.82) (À9.01)
FAM_CEO/OWNER (H
4a
) +/À 0.012
***
0.007
*
À0.009
***
(3.11) (1.79) (À3.16)
FAM_BOARD/OWNER (H
5a
) À À0.038
***
À0.021
***
À0.039
***
(À7.83) (À6.83) (À9.52)
FAM_BOARD/CEO (H
6a
) +/À 0.028
***
0.017
***
0.020
***
(4.91) (5.24) (3.98)
LNSALES + 0.234
***
0.235
***
0.236
***
0.237
***
0.236
***
0.236
***
0.235
***
0.237
***
(97.16) (97.36) (97.93) (98.34) (97.86) (98.14) (97.70) (98.38)
LNEMPLOY + 0.150
***
0.150
***
0.151
***
0.153
***
0.152
***
0.153
***
0.151
***
0.152
***
(53.56) (53.70) (54.12) (54.59) (54.47) (54.59) (53.81) (54.58)
LEV + 0.078
***
0.078
***
0.078
***
0.078
***
0.077
***
0.078
***
0.078
***
0.077
***
(13.67) (13.60) (13.63) (13.62) (13.56) (13.62) (13.62) (13.57)
CHLEV + À0.064
***
À0.064
***
À0.066
***
À0.066
***
À0.065
***
À0.066
***
À0.065
***
À0.066
***
(À8.66) (À8.66) (À8.88) (À8.90) (À8.83) (À8.90) (À8.81) (À8.87)
INVREC + 0.261
***
0.262
***
0.263
***
0.265
***
0.264
***
0.265
***
0.262
***
0.265
***
(32.36) (32.44) (32.61) (32.82) (32.74) (32.81) (32.44) (32.84)
GROWTH + À0.049
***
À0.049
***
À0.049
***
À0.049
***
À0.049
***
À0.049
***
À0.049
***
À0.049
***
(À43.35) (À43.34) (À43.22) (À43.27) (À43.34) (À43.31) (À43.39) (À43.22)
ROA À À0.153
***
À0.154
***
À0.160
***
À0.162
***
À0.159
***
À0.161
***
À0.157
***
À0.161
***
(À23.62) (À23.85) (À24.75) (À25.10) (À24.67) (À24.96) (À24.37) (À25.06)
INCPIC + 0.066
***
0.067
***
0.069
***
0.071
***
0.069
***
0.070
***
0.068
***
0.071
***
(16.01) (16.21) (16.69) (17.20) (16.83) (17.09) (16.53) (17.26)
DECPIC + 0.100
***
0.101
***
0.105
***
0.107
***
0.105
***
0.106
***
0.103
***
0.107
***
(10.67) (10.73) (11.14) (11.35) (11.18) (11.30) (10.95) (11.37)
INTANG + 0.104
***
0.106
***
0.110
***
0.114
***
0.112
***
0.113
***
0.109
***
0.114
***
(4.89) (4.97) (5.18) (5.35) (5.24) (5.32) (5.14) (5.37)
UNITEMS + 0.072
***
0.072
***
0.072
***
0.072
***
0.072
***
0.072
***
0.072
***
0.072
***
(12.22) (12.25) (12.22) (12.21) (12.24) (12.23) (12.23) (12.22)
NOIND + 0.032
***
0.032
***
0.032
***
0.032
***
0.032
***
0.032
***
0.033
***
0.032
***
(18.76) (18.66) (18.90) (18.75) (18.82) (18.80) (18.96) (18.65)
FOREIGN + 0.111
***
0.112
***
0.113
***
0.114
***
0.114
***
0.114
***
0.112
***
0.115
***
(8.43) (8.51) (8.57) (8.66) (8.62) (8.63) (8.48) (8.70)
LNSUB + 0.046
***
0.046
***
0.046
***
0.045
***
0.045
***
0.045
***
0.046
***
0.045
***
(9.20) (9.10) (9.18) (8.99) (8.99) (8.99) (9.20) (8.96)
FYE + 0.117
**
0.117
**
0.128
***
0.128
***
0.120
**
0.128
***
0.118
**
0.130
***
(2.43) (2.45) (2.66) (2.66) (2.51) (2.66) (2.47) (2.69)
INVESTMENTS + 0.066
***
0.066
***
0.067
***
0.067
***
0.067
***
0.067
***
0.067
***
0.067
***
(31.22) (31.17) (31.63) (31.45) (31.28) (31.45) (31.74) (31.31)
ACQUISITIONS + 0.065
***
0.066
***
0.066
***
0.066
***
0.066
***
0.066
***
0.066
***
0.067
***
(11.36) (11.43) (11.53) (11.57) (11.57) (11.57) (11.42) (11.58)
LOSS + 0.045
***
0.045
***
0.046
***
0.047
***
0.046
***
0.046
***
0.045
***
0.047
***
(16.30) (16.42) (16.86) (17.05) (16.85) (16.97) (16.52) (17.11)
CURRATIO + 0.001
*
0.001
*
0.002
**
0.002
**
0.002
**
0.002
**
0.002
**
0.002
**
(1.86) (1.92) (2.16) (2.31) (2.16) (2.28) (2.23) (2.25)
BIG4 + 0.116
***
0.116
***
0.118
***
0.118
***
0.117
***
0.118
***
0.117
***
0.118
***
(21.19) (21.21) (21.50) (21.61) (21.46) (21.57) (21.41) (21.61)
GAAP ? À0.074
***
À0.074
***
À0.074
***
À0.075
***
À0.076
***
À0.075
***
À0.073
***
À0.075
***
(À7.13) (À7.15) (À7.13) (À7.20) (À7.27) (À7.22) (À7.05) (À7.20)
CHGAAP + 0.083
***
0.083
***
0.085
***
0.086
***
0.085
***
0.086
***
0.084
***
0.086
***
(5.65) (5.66) (5.75) (5.82) (5.75) (5.80) (5.72) (5.82)
INDSPEC + À0.390
***
À0.390
***
À0.390
***
À0.394
***
À0.391
***
À0.394
***
À0.392
***
À0.393
***
(À11.67) (À11.64) (À11.64) (À11.70) (À11.64) (À11.69) (À11.67) (À11.67)
CHUDITOR + 0.006
*
0.006
*
0.006
*
0.006
*
0.006
*
0.006
*
0.006
*
0.006
*
(1.67) (1.71) (1.70) (1.67) (1.65) (1.68) (1.67) (1.67)
LASTYRAUDIT +/À 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001
(0.30) (0.31) (0.24) (0.23) (0.28) (0.24) (0.24) (0.23)
MODOPINIONS + 0.033
***
0.033
***
0.033
***
0.032
***
0.032
***
0.032
***
0.033
***
0.032
***
(18.69) (18.61) (18.28) (17.63) (18.02) (17.70) (18.19) (17.67)
CHCEO +/À 0.021
***
0.023
***
0.030
***
0.033
***
0.023
***
0.030
***
0.027
***
0.033
***
(4.61) (5.10) (6.53) (7.17) (4.99) (6.49) (6.03) (7.17)
512 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
most identical results across the two periods. The only dif-
ference is that OWNER_CONC is not signi?cant in the earlier
time period for the audit fee test (H
1a
). Overall, our robust-
ness tests corroborate the main ?ndings tabulated in the
paper.
Concluding remarks
This study provides empirical evidence of agency con-
?icts associated with ownership characteristics and family
relationships for private ?rms in the Norwegian market.
Using a detailed data set obtained by special permission
from the government, we are able to measure ultimate
ownership concentration, ultimate ownership of the sec-
ond largest shareholder and of the CEO, and family rela-
tionships among owners, board members, and CEOs for
all Norwegian private limited liability ?rms. These data al-
low us to simultaneously test multiple sources of agency
con?icts.
We ?rst test for the extent of agency con?icts using
audit fees. Controlling for a large number of client-?rm
and audit-?rm characteristics, we predict and ?nd that
ownership characteristics and family relationships explain
audit fees. Speci?cally, we ?nd that audit fees decrease
(i.e., audit effort decreases) as ownership concentration in-
creases and as the proportion of shares held by the second
largest shareholder increases. The ?rst result is consistent
with greater ownership concentration alleviating agency
costs between shareholders and the managers (i.e., vertical
agency costs). Larger shareholders can more ef?ciently
monitor managers. The second result is consistent with re-
duced agency costs between controlling shareholders and
minority shareholders (i.e., horizontal agency costs). A sec-
ond large shareholder serves as a monitor of a controlling
shareholder who potentially has the ability to extract pri-
vate bene?ts from minority shareholders.
Related to the CEO, we ?nd that as CEO ownership in-
creases, audit fees decrease. This is consistent with in-
creased ownership aligning the incentives of the CEO
with those of the ?rms’ other stakeholders. In contrast,
holding CEO ownership constant, we ?nd that when the
CEO is a member of the largest owning family, audit fees
increase (i.e., auditor effort increases). This ?nding sug-
gests that shareholders are less likely to act as independent
monitors of the CEO when a family relationship exists,
increasing the probability of misappropriation by the CEO
or extraction of private bene?ts by controlling owners.
For board independence, we ?nd that audit fees de-
crease as the proportion of board members from the larg-
est owning family increases. This ?nding re?ects the
likelihood that those inside the ?rm (board members) are
more likely to act as effective monitors on behalf of those
outside the ?rm (controlling owners) when a family rela-
tion exists. We also ?nd that boards seem to lose their
independence as the proportion of board members from
the CEO’s family increases. This result follows from the
natural expectation that family relationships with the
CEO impair the oversight responsibility of the board.
As our second test, we consider each of our agency set-
tings and the ?rm’s demand for a Big 4 auditor. Consistent
with results found for the auditor effort tests, the demand
for a Big 4 auditor decreases with ownership concentra-
tion, level of ownership by the second largest owner, and
family relationships between the board and the largest
owner. We ?nd no relation between demand for a Big 4
auditor and CEO ownership for our main tests but we do
in sensitivity tests. Finally, for our two CEO family-related
agency cost settings, we observe no impact on the demand
for a Big 4 auditor when the CEO is related to the major
shareholder or as the number of board members related
to the CEO increases. These insigni?cant relations likely re-
?ect a trade-off between the bene?ts of more credible
reporting from using a Big 4 auditor versus the potential
costs of increased fees associated with a Big 4 auditor
and the reduced ability of the CEO (or the CEO’s family)
to extract resources from the ?rm.
Table 3 (continued)
Predicted sign Model
1 2 3 4 5 6 7 8
CITY + 0.064
***
0.065
***
0.065
***
0.064
***
0.064
***
0.064
***
0.064
***
0.064
***
(18.73) (18.76) (18.77) (18.63) (18.66) (18.60) (18.51) (18.73)
Constant 0.376
***
0.349
***
0.317
***
0.282
***
0.311
***
0.292
***
0.327
***
0.278
***
(14.61) (13.86) (12.74) (11.53) (12.60) (11.86) (13.12) (11.38)
County ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Industry ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Year ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Observations 185,109 185,109 185,109 185,109 185,109 185,109 185,109 185,109
Adjusted R
2
0.529 0.529 0.528 0.528 0.528 0.528 0.528 0.528
This table presents OLSÀregression coef?cients (t-values in parentheses) for the impact of ownership characteristics and family relationships on audit fees.
The dependent variable is LNFEE, the natural log of audit fees in NOK 1000. The sample is described in Table 1 and variables are de?ned in the Appendix
except for the test variables in Model 2 which are measured as indicator variables de?ned as 1 if the test variable is greater than or equal to the mean and 0
otherwise. County, industry, and year ?xed effects are included but omitted from table to conserve space. All regressions employ robust standard errors
clustered at the ?rm level.
*
Statistical signi?cance at the 0.10 level (two-sided).
**
Statistical signi?cance at the 0.05 level (two-sided).
***
Statistical signi?cance at the 0.01 level (two-sided).
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 513
Our research makes a contribution by providing a de-
tailed investigation of the agency settings for private ?rms.
Several differences exist between private and public ?rms,
and results for one group are not necessarily expected to
be the same as those for the other. In addition, in aggregate
private ?rms provide the majority of economic activity
throughout the world. We obtain a unique database for
private ?rms in Norway to provide tests of the impact of
ownership characteristics and family relationships on
agency con?icts. This database entails combining owner-
ship characteristics and family relationships using social
security numbers, available only through special permis-
sion from the government.
There is ample opportunity for future research to further
explore the issues we examine in this study. Perhaps the
most obvious suggestion would be to test for economic
consequences of variations in the agency con?icts and/or
audit variables that we consider. Another possibility is to
make further use of the register data available through link-
ing databases using social security numbers. For example,
auditor independence issues could be examined by consid-
ering auditors’ personal income and wealth. Other research
could make use of interesting aspects of alternative institu-
tional environments, or use research methods other than
archival data to explore agency con?icts in private ?rms.
Acknowledgements
We appreciate valuable comments from Ahmed
Abdel-Meguid, Erlend Kvaal, Alastair Lawrence, Gilad
Livne, Salvatore Miglietta, Mark Peecher (Editor), Ann
Vanstraelen, Joost van Buuren, two anonymous reviewers,
Table 4
Logit regressions of choice of Big 4 auditor on ownership, CEO, and board characteristics, as well as control variables.
Predicted sign Model
1 2 3 4 5 6 7 8
OWNER_CONC (H
1b
) À À0.227
***
À0.096
***
À0.078
**
(À4.16) (À2.89) (À2.22)
OWNER_SECOND (H
2b
) À À0.753
***
À0.258
***
À0.357
***
(À10.06) (À8.70) (À6.58)
OWNER_CEO (H
3b
) À À0.052 À0.087
***
À0.106
***
(À1.26) (À3.45) (À3.81)
FAM_CEO/OWNER (H
4b
) +/À À0.027 À0.023 À0.109
***
(À0.93) (À0.88) (À5.27)
FAM_BOARD/OWNER (H
5b
) À À0.219
***
À0.139
***
À0.248
***
(À6.04) (À5.96) (À8.35)
FAM_BOARD/CEO (H
6b
) +/À 0.027 0.010 À0.119
***
(0.61) (0.39) (À3.06)
LNTA + 0.335
***
0.339
***
0.357
***
0.360
***
0.355
***
0.356
***
0.347
***
0.361
***
(34.58) (35.12) (37.49) (38.21) (37.18) (37.69) (36.53) (38.32)
LEV + À0.442
***
À0.450
***
À0.467
***
À0.465
***
À0.466
***
À0.468
***
À0.466
***
À0.473
***
(À9.92) (À10.11) (À10.49) (À10.47) (À10.48) (À10.54) (À10.46) (À10.63)
UNSECURED + 0.361
***
0.365
***
0.381
***
0.384
***
0.380
***
0.377
***
0.361
***
0.378
***
(7.84) (7.92) (8.29) (8.37) (8.28) (8.21) (7.82) (8.21)
LOSS ? 0.130
***
0.133
***
0.148
***
0.147
***
0.147
***
0.146
***
0.140
***
0.148
***
(7.24) (7.39) (8.27) (8.19) (8.17) (8.14) (7.76) (8.26)
FOREIGN + 0.319
***
0.321
***
0.340
***
0.338
***
0.340
***
0.338
***
0.327
***
0.337
***
(4.23) (4.26) (4.52) (4.48) (4.52) (4.49) (4.33) (4.47)
EXANTEFIN + 0.036
***
0.037
***
0.037
***
0.038
***
0.037
***
0.037
***
0.037
***
0.038
***
(3.37) (3.43) (3.48) (3.53) (3.44) (3.44) (3.47) (3.53)
INCPIC + À0.017 À0.011 0.012 0.016 0.011 0.009 À0.002 0.014
(À0.64) (À0.41) (0.44) (0.60) (0.41) (0.36) (À0.07) (0.54)
ROA À À0.374
***
À0.389
***
À0.443
***
À0.439
***
À0.438
***
À0.439
***
À0.414
***
À0.449
***
(À8.39) (À8.73) (À9.94) (À9.89) (À9.85) (À9.89) (À9.30) (À10.11)
CITY + 0.081
***
0.081
***
0.084
***
0.082
***
0.083
***
0.082
***
0.080
***
0.081
***
(3.21) (3.21) (3.31) (3.23) (3.28) (3.23) (3.15) (3.20)
Constant À4.300
***
À4.523
***
À4.945
***
À4.945
***
À4.917
***
À4.902
***
À4.731
***
À5.002
***
(À25.84) (À28.01) (À31.07) (À31.71) (À31.17) (À31.24) (À29.79) (À32.16)
Count ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Industry ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Year ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Observations 185,109 185,109 185,109 185,109 185,109 185,109 185,109 185,109
Pseudo R
2
0.101 0.101 0.099 0.099 0.099 0.099 0.099 0.099
This table presents logistic regression coef?cients (z-values in parentheses) for the impact of ownership characteristics and family relationships on the
choice of Big 4 auditors. The sample is described in Table 1 and variables are de?ned in the Appendix except for the test variables in Model 2 which are
measured as indicator variables de?ned as 1 if the test variable is greater than or equal to mean and 0 otherwise. County, industry, and year ?xed effects are
included in the models but omitted from the table in order to conserve space. Robust standard errors clustered at the ?rm level are employed.
Ã
Statistical signi?cance at the 0.10 level (two-sided).
**
Statistical signi?cance at the 0.05 level (two-sided).
***
Statistical signi?cance at the 0.01 level (two-sided).
514 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
and participants at the 5th European Auditing Research
Network (EARNet) Symposium (Valencia), the European
Accounting Association Annual Meeting (Istanbul), the
16th Annual International Symposium on Audit Research
(Singapore), the American Accounting Association Annual
Meeting (San Francisco), the 3rd International EIASM
Workshop on Audit Quality (Bellagio), the 5th CCGR Work-
shop (Oslo), the 1st Nordic Conference on Financial
Accounting (Copenhagen), and the 6th European Auditing
Research Network (EARNet) Symposium (Bergen). We also
greatly appreciate a grant from the Centre for Corporate
Governance Research (BI Norwegian Business School),
and the support from its director, Professor Øyvind Bøhren.
Hope further acknowledges the support of the Deloitte
Professorship.
Appendix A. Appendix: De?nitions of variables
Test variables
OWNER_CONC Her?ndahl ownership
concentration ratio in year t based
on ultimate ownership
OWNER_SECOND Fraction of shares held by the
second largest owner in year t using
ultimate ownership
OWNER_CEO Fraction of shares held by the CEO in
year t, based on ultimate ownership
FAM_CEO/
OWNER
1 if the CEO is related to the largest
ultimate family owners through
blood or marriage
FAM_BOARD/
OWNER
Family in?uence on the board in
year t computed as the number of
board members from the largest
owning family divided by the total
number of board members
FAM_BOARD/
CEO
Family relationships between the
CEO and board members, de?ned as
the number of family members of
the CEO on the board divided by the
total number of board members
Other variables
ACQUISITIONS 1 if the ?rm has increased long-
term investments in other
companies from t À 1 to t, 0
otherwise
AGE Firm’s age in year t measured as
number of years since incorporation
BIG4 1 if the auditing ?rm is one of the
Big 4 auditing ?rms or their
forerunners in year t, 0 otherwise
CHAUDITOR 1 if the ?rm changed auditor in year
t, 0 otherwise
CHCEO 1 if the ?rm appointed a new CEO in
year t, 0 otherwise
CHGAAP 1 if the ?rm has changed from
simpli?ed GAAP to regular GAAP or
from regular GAAP to simpli?ed
GAAP during year t
CHLEV Change in interest bearing debt
from t À 1 to t = LEV
t
À LEV
tÀ1
CITY 1 if ?rm is located in a city, 0
otherwise
COUNTY
c
1 if ?rm is located in county c, 0
otherwise (c = 0, 1, . . ., 19)
CURRATIO Current ratio at the end of year t
DECPIC 1 if the ?rm decreased share capital
from t À 1 to t, 0 otherwise
EMPLOY Number of employees in year t is
estimated as sum wages in year t
divided by the average salary for full
time male and female workers in
year t if sum wages >0, 0 otherwise.
Data on average salary is taken from
Statistics Norway
EXANTEFIN Operating cash ?ow less net
investments in tangible and
intangible ?xed assets scaled by
current assets in year t. As in
Knechel et al. (2008), we use a
square root transformation of the
ratio multiplied by À1 if the
unadjusted ratio was negative.
Operating cash ?ow is estimated as
operating income + depreciation +
impairment À change in accounts
receivables À change in inventory +
change in accounts payable À
change in taxes payable. Net
investments in intangible and
tangible ?xed assets is estimated as
change in intangible and ?xed
assets during the year plus
depreciation and impairment since
cash ?ow statements are not
available
FEE Audit fee in NOK 1000, computed as
the total fee paid to the auditor for
auditing services in year t.
FOREIGN Percentage of foreign subsidiaries in
year t = the number of foreign
subsidiaries in year t
Ã
100/total
number of subsidiaries in year t
FYE 1 for ?scal years ending other than
December 31 in year t, 0 otherwise
GAAP 1 if ?rm uses regular GAAP in year t
and 0 otherwise
GROWTH Change in sales in year
t = (SALES
t
À SALES
tÀ1
)/SALES
tÀ1
INCPIC 1 if the ?rm increased share capital
from t À 1 to t, 0 otherwise
IND
i
1 if ?rm belongs to industry i, 0
otherwise. Two-digit industry codes
are used to classify ?rms into
industries. Firms in industries with
30 or less observations in a given
year are reclassi?ed to industry 99
(continued on next page)
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 515
INDUSTRIAL 1 if industrial investors own 20% or
more of the ?rm based on ultimate
ownership, 0 otherwise
INSTITUTIONAL 1 if institutional investors own 20%
or more of the ?rm based on
ultimate ownership, 0 otherwise
INTANG Intangible assets at the end of year t
scaled by SALES
t
INTERNATIONAL 1 if international investors own 20%
or more of the ?rm based on
ultimate ownership, 0 otherwise
INVESTMENTS (Long and short terminvestments in
securities + bank deposits + cash)
t
scaled by SALES
t
INVREC Inventory and accounts receivable
at the end of year t scaled by SALES
t
LASTYRAUDIT 1 in year t if the ?rm changed
auditor in year t + 1, 0 otherwise
LEV Long- and short-term interest
bearing debt/total assets, both at
the end of year t. Short-term
interest bearing debt = total short
term debt À accounts
payable À dividends À taxes
payable À VAT and social service
taxes À other short term debt
LNEMPLOY Natural logarithm of 1+ number of
employees = ln(1 + EMPLOY
t
)
LNFEE Natural logarithm of total fees paid
to the auditor for auditing services
in year t = ln(FEE
t
)
LNSALES Natural logarithm of total revenue
from operations = ln(SALES
t
)
LNSUB Natural logarithm of 1 + number of
subsidiaries in year t
LNTA Natural logarithm of total
assets = ln(TA
t
)
LOSS 1 if net income after taxes before
extraordinary item and taxes on
extraordinary item
We are interested in understanding how agency conflicts in private firms arise through
ownership structures and family relationships. Specifically, we analyze auditors’ increase
of effort and firms’ choice of auditors in situations with higher level of agency conflicts.
For a large sample of private firms, we use unique and confidential data (obtained through
special permission by the government) to measure direct and ultimate ownership for each
shareholder as well as extended family relationships (based on marriage and blood lines,
going back four generations and extending out to fourth cousin) among all shareholders,
board members, and CEOs. We first find that audit fees, our proxy for audit effort, vary
as hypothesized with firm-level characteristics related to ownership structures and family
relationships. Second, we find evidence that firms in higher agency cost settings respond by
having their financial statements audited by a higher-quality auditor (i.e., a Big 4 firm).
Agency con?icts and auditing in private ?rms
Ole-Kristian Hope
a,?
, John Christian Langli
b
, Wayne B. Thomas
c
a
Rotman School of Management, University of Toronto, Canada
b
Department of Accounting, Auditing, and Law, BI Norwegian Business School, Norway
c
Michael F. Price College of Business, University of Oklahoma, United States
a b s t r a c t
We are interested in understanding how agency con?icts in private ?rms arise through
ownership structures and family relationships. Speci?cally, we analyze auditors’ increase
of effort and ?rms’ choice of auditors in situations with higher level of agency con?icts.
For a large sample of private ?rms, we use unique and con?dential data (obtained through
special permission by the government) to measure direct and ultimate ownership for each
shareholder as well as extended family relationships (based on marriage and blood lines,
going back four generations and extending out to fourth cousin) among all shareholders,
board members, and CEOs. We ?rst ?nd that audit fees, our proxy for audit effort, vary
as hypothesized with ?rm-level characteristics related to ownership structures and family
relationships. Second, we ?nd evidence that ?rms in higher agency cost settings respond by
having their ?nancial statements audited by a higher-quality auditor (i.e., a Big 4 ?rm).
However, for CEO family-related settings (i.e., where the CEO is related to the major share-
holder or as the number of board members related to the CEO increases), we ?nd no evi-
dence of a greater demand for a Big 4 auditor.
Ó 2012 Elsevier Ltd. All rights reserved.
Introduction
In this study, we seek to understand how ownership
structures and family relationships in?uence agency costs
in private ?rms. We do this by observing two aspects re-
lated to auditing. First, in higher agency cost settings, audi-
tors are more likely to supply greater effort to prevent
misstatement associated with moral hazard and adverse
selection problems. We examine how auditors adjust their
level of effort when auditing ?nancial accounting informa-
tion. Second, a subset of ?rms in higher agency cost settings
likely have a greater demand to choose a higher-quality
auditor to provide a credible signal of their commitment
to higher-quality reporting. To test this, we examine the
extent to which ?rms with various characteristics hire a
Big 4 auditor.
Our examinations draw on very detailed data on ulti-
mate ownership and extended family relationships pro-
vided by the Norwegian government. We ?nd that audit
fees (i.e., our proxy for auditor effort) increase with ex-
pected agency costs.
1
Audit fees relate negatively to owner-
ship concentration and to the extent of ownership by the
second-largest shareholder. Concentrated ownership in-
creases the likelihood that a large shareholder closely mon-
itors managerial actions, and an in?uential second
shareholder monitors potential expropriation by the largest
shareholder. Audit fees also relate negatively to the portion
of shares held by the CEO, consistent with ownership align-
ing the incentives of the CEO and other stakeholders. Audit
fees are positively associated with family relationships be-
tween the CEO and the major shareholder (consistent with
these family relationships indicating reduced monitoring).
0361-3682/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.aos.2012.06.002
?
Corresponding author.
E-mail addresses: [email protected] (O.-K. Hope), john.c.
[email protected] (J.C. Langli), [email protected] (W.B. Thomas).
1
The fee (or effort) regression includes controls for 24 client-?rm
characteristics, ?ve audit-?rm variables, as well as year and industry ?xed
effects.
Accounting, Organizations and Society 37 (2012) 500–517
Contents lists available at SciVerse ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
With respect to board independence, we ?nd that audit
fees decline as the number of board members related to
the largest shareholder increases, consistent with fewer
agency con?icts between owners and the board. In con-
trast, as the number of board members related to the
CEO increases, audit fees increase, suggesting less board
independence and more agency con?icts.
For our tests of demand for Big 4 auditor, we report two
interesting sets of results. First, for agency settings that are
not CEO family-related, we observe results consistent with
those obtained for our auditor effort tests. Speci?cally, the
propensity to hire a Big 4 auditor increases as ownership
concentration decreases, ownership of the second largest
owner decreases, and the major shareholder’s family in?u-
ence on the board decreases. These results are consistent
with the demand for a Big 4 auditor being greater in higher
agency cost settings. In these settings, a higher-quality
auditor plays a stronger role in reducing agency costs by
sending a more credible signal of managers’ commitment
to higher-quality reporting. We do not ?nd signi?cant evi-
dence of a relation between hiring a Big 4 auditor and the
fraction of shares owned by the CEO for our main tests, but
we do in sensitivity tests.
We ?nd no association between the choice to hire a
Big 4 auditor and CEO family-related agency variables.
Speci?cally, we ?nd no signi?cant evidence that the
demand for a Big 4 auditor is affected when a family rela-
tionship exists between the CEO and the major share-
holder or as the number of board members related to
the CEO increases. One explanation for the lack of signif-
icance could be that while some CEOs in family-related
agency settings may wish to signal more credible report-
ing by hiring a Big 4 auditor, other CEOs in these settings
may feel a Big 4 auditor is either unnecessary given close
family ties or unwanted because of the gains from
extracting private bene?ts which could be reduced by a
Big 4 audit.
2
Our research is motivated by the need to understand
agency con?icts facing private ?rms. Private ?rms make
up a signi?cant portion of the economic activity in Norway
and nearly all other countries, yet prior research focuses
primarily on public ?rms. Given the sometimes vast differ-
ences between public and private ?rms (e.g., Ball & Shi-
vakumar, 2005; Beatty, Ke, & Petroni, 2002; Chaney, Jeter,
& Shivakumar, 2004), it is not apparent without testing
that results for public ?rms will generalize to private ?rms.
Thus, private ?rms offer an economically important sam-
ple worth testing. While the bene?ts to understanding
agency con?icts accrue directly to the ?rm’s investors, they
will also be important to many other stakeholders (e.g.,
creditors, employees, suppliers, and customers), regulatory
bodies supervising auditors and ?rms’ ?nancial reporting,
and society in general.
A sample of private (as opposed to public) ?rms may
also offer a stronger test of agency con?icts related to own-
ership structure and family relationships. As we discuss in
more detail in Hypotheses, prior research sometimes pro-
vides con?icting evidence or con?icting predictions for
the impact of ownership structures and family relation-
ships on agency con?icts. Our study provides a potentially
strong setting for testing agency con?icts because private
?rms exhibit heterogeneous ownership characteristics
and family relationships. Public ?rms are more homoge-
neous, including wide-spread ownership, relatively low
CEO ownership, and fewer family ties between managers
and shareholders and between managers and board mem-
bers. Private ?rms offer interesting ownership structures
that potentially increase our understanding of the relation
between agency con?icts and the supply of auditor effort.
For example, private ?rms show considerable variation in
ownership percentages by second largest shareholders.
This allows us to provide a meaningful test of the impact
of agency con?icts among shareholders (i.e., monitoring
of largest shareholders by second largest shareholders).
Private ?rms also show greater variation in their choice
of auditor (only 18.1% choose a Big 4 auditor). Nearly all
public ?rms opt for a Big 4 auditor, limiting the ability to
empirically test signaling through demand for a high-qual-
ity auditor.
Related to tests of the supply of auditor effort, a single-
country setting (Norway) controls for cross-country varia-
tion in audit practices and fees and the strength of legal
institutions. Cross-country differences could easily con-
found inferences. Norway also offers an environment
where the impact of litigation on audit fees is relatively
limited (Hope & Langli, 2010). This increases our ability
to make more reliable inferences from using audit fees to
measure auditor effort, and adds to calls for research to
better understand the role of ?rmgovernance in explaining
audit fees (Hay, Knechel, & Wong, 2006).
Finally, given the unique data we use in this study, we
are able to measure attributes of ownership structure
and family relationships that have been dif?cult to mea-
sure in the past. Speci?cally, for all private limited liability
?rms we have detailed information available to compute
both direct and ultimate ownership for each owner, board
member, and CEO.
3
In addition, we have detailed data on
family relationships among all owners, board members,
board chairs, and CEOs (based on both marriage and blood
lines, going back four generations and extending out to
fourth cousin). To our knowledge, no prior study has been
able to test the effects of family relationship using such de-
tailed data. These data, based on merging databases using
social security numbers, are obtained through special
2
As per Norway’s Companies Act (§ 7-1), the annual meeting (or General
Assembly) of the shareholders elects the auditor. While the selection of the
auditor is technically the responsibility of the General Assembly (or in
practice the board), it is almost certainly the case that the CEO plays a
signi?cant role in our setting. There is a large literature on how CEOs
in?uence the selection of board members and more generally exercise
‘‘power’’ over the board (see, e.g., Bebchuk & Fried 2003). Even for large
public ?rms in the US and UK the CEO has an impact on which auditor is
selected (e.g., Beattie & Fearnley 1995; Carcello, Neal, Palmrose, & Scholz
2011; Firth 1999). For our sample of smaller private ?rms, the CEO is
expected to have a greater in?uence on auditor selection because the
boards are typically smaller, CEO ownership is common (the CEO owns
shares in 78% of our sample ?rms), and the largest owner often is related to
the CEO.
3
For example, suppose an investor owns 30% of ?rm A and30% of ?rm B,
and ?rm A in turn owns 30% of ?rm B. The investor’s direct ownership in
?rm B is 30%, while her ultimate ownership is 39%. Note that our data also
account for cross-holdings.
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 501
permission from the Norwegian government. Having these
data allows us to simultaneously test multiple sources of
agency con?icts in a single test. In contrast, prior studies, be-
cause of data limitations, have been able to focus only on a
particular test variable (and thus leave out potentially
important ownership and family details). The full model bet-
ter controls for multiple agency settings. We show that the
ability to simultaneously test the association of these vari-
ables on auditor effort and ?rms’ selection of the auditor
can affect conclusions in some cases.
We next brie?y discuss the importance of accounting
and auditing, the role of auditing in private ?rms, and
the importance of auditing in the private ?rm market.
Hypotheses provides hypotheses of the expected impact
that agency con?icts have on the supply of auditor effort
and the demand for a Big 4 auditor. Research design intro-
duces our research design. Data on ownership and family
relationships provides details on the data underlying our
study. We present empirical results in Sample and results,
and Concluding remarks concludes.
Background, related research, and institutional setting
Importance of auditing in private ?rm setting
An external audit plays a role in corporate governance
by providing an independent assessment of the accuracy
and fairness with which ?nancial statements represent
the results of operations in accordance with generally ac-
cepted accounting principles. While researchers commonly
consider the role of auditing for public ?rms, audits for pri-
vate ?rms can also play a role. Private ?rms typically dis-
close less non-accounting information, potentially
increasing the importance of ?nancial accounting informa-
tion to external providers of capital in monitoring manage-
rial activities. In addition, managerial activities of public
?rms are partially constrained by market-based mecha-
nisms. For example, public ?rms are more susceptible to
takeovers, indicating that such mechanisms help control
for agency con?icts (Lennox, 2005). In the absence of mar-
ket-based measures of ?rm-value (and other sources of
information such as ?nancial analysts), high-quality
reporting may be particularly relevant for evaluation of
managerial performance and to support personnel and
compensation decisions (Indjejikian & Matejka, 2009),
resulting in a demand for high-quality audits (Chaney et
al., 2004).
4
Brief review of prior research
Although there is a huge auditing literature, there is rel-
atively limited prior evidence on auditing issues in private
?rms.
5
Some prior studies investigate private ?rms’ choice to
have an audit. Carey, Simnett, and Tanewski (2000) ?nd that
nonfamily involvement correlates positively with the volun-
tary demand for auditing in Australian family businesses.
Blackwell, Noland, and Winters (1998) ?nd that private US
?rms that elect to have their ?nancial statements audited
pay signi?cantly lower interest rates than nonaudited pri-
vate ?rms. Similarly, Allee and Yohn (2009) use National
Survey of Small Business Finances data from 2003 to exam-
ine the ?nancial reporting practices of small privately held
U.S. businesses. They ?nd that ?rms with audited ?nancial
statements enjoy greater access to credit. Using a large sam-
ple of private ?rms from the World Bank Enterprise Surveys,
Hope, Thomas, and Vyas (2011) show that ?rms which have
their ?nancial statements reviewed by an external auditor
experience easier access to external ?nancing and obtain
those funds at lower costs.
In a recent study, Lennox and Pittman (2011) make use
of a natural experiment in the United Kingdom. Starting in
2004, an external audit is no longer required for private
UK companies. Lennox and Pittman focus on the ?rms
that are audited under both regimes (i.e., the ?rms that
reveal their preference to be audited). They ?nd that these
companies attract upgrades to their credit ratings. In con-
trast, companies that no longer submit to an audit suffer
credit rating downgrades. These ?ndings provide further
support for the usefulness of auditing in a private ?rm
setting.
There is also research on how audit fees relate to own-
ership characteristics, primarily for publicly traded compa-
nies. For example, Chan, Ezzamel, and Gwilliam(1993) ?nd
that ownership control, de?ned in their study as ‘‘directors’
bene?cial and non-bene?cial shareholdings and all dis-
closed shareholdings in excess of 5%,’’ is negatively related
to levels of audit fees for their full sample of 280 publicly
traded U.K. companies from 1987. In contrast, they ?nd
no signi?cant effect for smaller auditees (i.e., ?rms that
are more similar to the private ?rms in our sample).
Mitra, Hossain, and Deis (2007) examine 358 NYSE-
listed manufacturing and retail companies that were au-
dited by Big 5 auditors in year 2000 (i.e., ?rms that are very
different from the small private ?rms in our sample). They
document that audit fees are positively associated with dif-
fused institutional stock ownership (i.e., having less than
5% individual shareholding) and negatively associated with
institutional blockholder ownership (i.e., having 5% or
more individual shareholding).
4
According to Van Tendeloo and Vanstraelen (2008), a choice to contract
for high-quality auditing (e.g., proxied by choosing a Big 4 auditing ?rm)
could signal ?nancial reporting quality and perhaps deter a rigorous tax
audit in the private ?rm market. They further argue that private ?rms may
also want to convince suppliers, clients, or employees of the credibility of
their ?nancial statements. This may be especially important in an
environment like Norway where ?nancial statements of all limited liability
companies (public and private) are publicly available. To illustrate, most
information from the income statements and the balance sheet, in addition
to names and shareholdings of owners, CEOs, and board members, is freely
available through websites (e.g., www.proff.no). Information on family
relationships between owners, CEOs, and board members, however, is not
publicity available.
5
The usefulness of accounting information for private ?rms has been
shown in recent studies including Chen, Hope, Li, and Wang (2011) and
Indjejikian and Matejka (2009). Chen et al. ?nd that emerging market ?rms
with higher ?nancial reporting quality exhibit greater investment ef?-
ciency. Indjejikian and Matejka further show that accounting information is
used in compensation contracting by US private ?rms. Studies that
compare the ?nancial reporting quality of public and private ?rms include
Ball and Shivakumar (2005), Burgstahler, Hail, and Leuz (2006), Asker,
Farre-Mensa, and Ljungqvist (2011), and Hope, Thomas, and Vyas (2012).
502 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
Hay, Knechel, and Ling (2008) show that an indicator for
whether there is a single shareholder who controls 20% or
more of the company’s stock is positively related to audit
fees for New Zealand public companies in 1995 (sample
size: 130). However, there is no signi?cant correlation in
2005 (sample size: 83).
As discussed, almost all prior research is on public, not
private ?rms. In addition, given that all private limited lia-
bility companies in Norway are audited during our sample
period, voluntary demand for audits per se is not an issue
for our study. The choice we examine is the demand for a
high-quality auditor (Big 4). Equally important, no prior
study has employed detailed data on family relationships
among shareholders, board members, and CEOs to mea-
sure agency con?icts, and to our knowledge few prior
studies have as detailed measures of ownership as this
study.
6
Finally, in contrast to most prior research, we exam-
ine the effects of several agency cost related factors simul-
taneously. In sum, although there is suf?cient extant
research ensuring that we have theoretical support for
our hypotheses, we believe our study ?lls a void by exam-
ining a large sample of private ?rms and testing for poten-
tially important effects for which prior research has had
limited data.
Institutional setting
In Norway, external auditing of all private limited liabil-
ity companies is mandated by the government. In the rest
of Europe, only private companies that meet certain size
criteria are required to have their ?nancial statements au-
dited.
7
The statutory auditor is expected to provide different
stakeholders of the company assurance concerning the accu-
racy of the ?nancial statements, the non-existence of ?nan-
cial statement fraud, and the going concern status (e.g., Van
Tendeloo & Vanstraelen, 2008).
Norwegian auditing standards follow International
Standards of Auditing (ISA). The ISAs require auditors to as-
sess several agency con?icts when assessing audit risk. For
example, from ISA 315, an auditor assesses risk of material
misstatement by understanding the entity’s ownership
and governance structures and the way that the entity is
structured (paragraph 11). This could include noting
whether the client ?rm has an owner-manager, more than
one owner, dispersed ownership, etc.
8
In addition, IAS 315
suggests that assessing risk involves understanding the rela-
tions between owners and other people, which likely in-
cludes family relationships (paragraph A23). Referring to
the audits of smaller ?rms (which are much more likely to
be private), IAS 315 states that the presence of an owner-
manager may mitigate certain risks arising from a lack of
segregation of duties (paragraph A76). We believe that our
investigation of the relation between audit and ownership
characteristics and family relationships provides a unique
setting for understanding the interplay between agency con-
?icts and auditing in the private ?rm segment of the
economy.
Hypotheses
The framework for our tests can be understood by con-
sidering that there is a principal (i.e., the party for whom
?nancial statements are prepared such as owners, credi-
tors, regulators, suppliers) and two agents (the manager
and the auditor). We observe the actions of these two
agents – the manager’s demand for a Big 4 auditor and
the auditor’s supply of effort – to understand the impact
of ownership characteristics and family relationships on
agency costs.
In the ?rst principal-agent setting, the manager under-
stands that her potentially suboptimal actions are unob-
servable to the principal, and therefore the principal will
impose a monetary penalty on the manager. To avoid this
penalty in high agency cost settings, the manager is will-
ing to hire a higher-quality auditor to provide a more
credible signal that ?nancial statements are free of mate-
rial misstatement and that the manager has committed to
refrain from siphoning private bene?ts. Previous research
suggests that larger auditors provide higher quality audits
(e.g., DeAngelo, 1981; Palmrose, 1988) and that Big N
auditors in particular provide higher-quality outcomes in
a variety of settings (e.g., Becker, DeFond, Jiambalvo, &
Subramanyam, 1998; DeFond, 1992; Mansi, Maxwell, &
Miller, 2004). These arguments lead to the expectation
that ?rms in higher agency cost settings are more likely
to demand a Big 4 auditor. This is the typical principal-
agent framework put forth in most auditor selection
studies.
9
The second agency setting involves the auditor acting as
an agent, having a preference for compensation and a dis-
utility for effort. In other words, the optimal action for the
auditor is to supply the minimum amount of effort needed
to ‘‘obtain reasonable assurance about whether the ?nan-
cial statements as a whole are free from material misstate-
ment, whether due to fraud or error’’ (ISA, 2010, paragraph
5). Doing so maximizes the auditor’s utility. The higher the
audit effort is, the less likely type II errors occur in audit re-
ports (e.g., Dye, 1993; Laux & Newman, 2010). Consistent
with prior research (e.g., Davis, Ricchiute, & Trompeter,
1993; Whisenant, Sankaraguruswamy, & Raghunandan,
2003), we measure auditor effort using audit fees. Due to
transparency of the audit market and the large number
of auditors and audit ?rms in Norway (Financial Supervi-
sory Authority of Norway, Annual Report, 2007), the mar-
6
Lennox (2005) investigates the relation between management owner-
ship and audit ?rm size among private UK companies.
7
Norway and Sweden were the two last countries in Europe that
abandoned the requirement that all limited liability companies should
disclose audited ?nancial statements. Starting November 10, 2010 in
Sweden and January 1, 2011 in Norway, the (very) smallest ?rms may
decide not to engage an auditor.
8
Standard auditing textbooks provide further support. For example,
Arens, Elder, and Beasley (2012) note that the distribution of ownership has
the potential to affect audit risk.
9
The obvious counterargument is that hiring a Big 4 auditor is more
costly. For example, Chaney et al. (2004) conclude that UK private ?rms
choose auditors that minimize their audit fees. However, they do not
examine whether audit fees vary with agency con?icts as in our study. We
contend that ?rms in higher agency cost settings are more willing to hire a
Big 4 auditor (and incur the higher audit cost) because of the bene?ts
received from signaling more credible reporting.
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 503
ket for auditing services is highly competitive. Audit fees
should therefore re?ect the effort assessed by the auditor
in assuring the accuracy of ?nancial statements.
10
In the next section, we develop our six hypotheses. The
?rst three focus on ownership characteristics – ownership
concentration among all shareholders, ownership by sec-
ond largest shareholder, and fraction of shares held by
the CEO. The second three hypotheses focus on family rela-
tionships – family ties between the CEO and largest own-
ing family, percentage of board members related to
largest owning family, and family ties between the CEO
and board members. For each of these six settings, we con-
sider whether agency concerns affect managers’ decision
to hire a Big 4 auditor or auditors’ supply of effort.
Ownership structures
Ownership concentration
Prior research provides evidence that managers, when
left unmonitored, are more likely to manage earnings,
commit fraud, or make suboptimal investment decisions
(e.g., Biddle & Hillary, 2006; Hope & Thomas, 2008). Thus,
shareholder monitoring is an important mechanism by
which agency costs can be reduced. However, while all
shareholders have the responsibility to monitor manage-
rial activities, the bene?ts of doing so by any individual
shareholder are proportional to the percentage of shares
owned. As the percentage of ownership by individual
shareholders increases (i.e., concentration increases), the
more willing individual shareholders are to incur neces-
sary monitoring costs.
Monitoring by a large shareholder could take many
forms. Perhaps the most commonly discussed means of
monitoring discussed in the literature involves a large
shareholder having a seat on the board. Several studies
show in a variety of contexts the board’s role in monitoring
managers (e.g., Adams, Hermalin, & Weisbach, 2010;
Beasley, 1996; Dechow, Sloan, & Sweeney, 1996; Fama,
1980; Fama & Jensen, 1983; Byrd and Hickman, 1992;
Anderson, Mansi, & Reed, 2004; Laksmana, 2008, to name
just a few).
11
Other forms of direct monitoring would be a
large shareholder actively participating in the ?rm’s opera-
tions or having routine meetings with managers. As the pro-
portion of ownership increases, the more bene?cial it is for
large shareholders to engage in these types of costly direct
monitoring activities. Large shareholders can also serve to
block business decisions that may be considered suboptimal
(e.g., aggressive expansion through negative net present va-
lue projects). Doing so involves an investment in time and
expertise by the shareholder to understand the conse-
quences of major business decisions. Large shareholders
are also likely to have more control over the ?rm’s dividend
(or capital distribution) policy, as a way to further discipline
managers’ actions.
When ownership is widely dispersed, it is economically
less feasible for any individual shareholder to incur signif-
icant monitoring costs, because she will receive only a
small portion of bene?ts. This is the typical ‘‘vertical
agency cost’’ (e.g., Gogineni, Linn, & Yadav, 2010) argument
(i.e., con?icts between managers and owners) and leads to
the prediction that agency costs are expected to be lower
as ownership concentration increases.
12
When agency costs
are lower, we expect that auditors supply less effort and
there is less demand for a Big 4 auditor.
Hyphothesis 1a. As ownership concentration increases,
audit fees decrease.
Hyphothesis 1b. As ownership concentration increases,
choice of Big 4 auditor decreases.
Second largest shareholder
While the previous discussion explains the need for
shareholders to monitor managers, the literature also
establishes the need for shareholders to monitor one an-
other. For example, controlling shareholders have the abil-
ity to exploit minority shareholders in closely-held
corporations (e.g., Burkart, Gromb, & Panunzi, 1997,
1998; Laeven & Levine, 2008; Nagar, Petroni, & Wolfenson,
2011). Such exploitation can include higher compensation
to controlling shareholders, misappropriation of assets,
and dilution of minority shareholders’ interests through
the issuance of stock or dividends (Gogineni, Linn, & Yadav,
2010). As the ownership stake of a second shareholder in-
creases, so does her ability and willingness to effectively
monitor the largest shareholder. The monitoring activities
by the second largest shareholder would be similar to
those used by the largest shareholder to monitor managers
(see discussion above in Ownership concentration).
Pagano and Roell (1998) specify conditions under which
large shareholders monitor each other, reducing expropri-
ation and improving ?rm performance. Their theoretical
model predicts that expropriation of minority shareholders
is likely to be less severe when the ownership stake of non-
controlling shareholders is more concentrated, as such
concentration makes it easier and more effective to moni-
tor the controlling shareholder (see also Bloch & Hege,
2001; Gogineni et al., 2010; Volpin, 2002). This is the
10
Some prior research in other settings suggests that it is important to
distinguish the component of audit fees that also re?ects compensation for
litigation risk. We do not view this distinction important for our tests. In
addition, both the litigation risk and reputation risk of auditors are
relatively low for private ?rms in Norway. For a detailed discussion of this
issue, see Hope and Langli (2010), who examine all court cases and other
legal proceedings against auditors over a 60-year period and conclude that
auditors face much lower litigation risk in Norway than in other more
litigious environments. It is important to note that, even in an environment
with low litigation, there is still a role for agency costs. For example, debt
and equity ?nancing will be more costly when agency costs are high. To the
extent that companies in settings with high agency problems are able to
signal more credible reporting, ?nancing will be less costly and more
accessible. Firms can signal this credibility with an audit. As another
example, suppliers may also be concerned about the viability of the ?rm
when deciding whether to enter long-term contracts. To the extent that the
supplier can adequately rely on ?nancial reports to assess the ?rm’s long-
run viability and stability, they are more willing to enter those contracts,
increasing the operating ef?ciency and pro?tability of the ?rm (e.g., Dou,
Hope, & Thomas, 2012).
11
See Bhagat and Black (1999), Hermalin and Weisbach (2003), and
Adams et al. (2010) for surveys on corporate boards.
12
An alternative prediction is that greater ownership concentration leads
to entrenchment, resulting in higher agency costs.
504 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
typical ‘‘horizontal agency cost’’ (e.g., Gogineni et al., 2010)
argument (i.e., con?icts between majority and minority
shareholders) and leads to the prediction that as owner-
ship by the second largest shareholder increases, agency
costs decrease.
Hyphothesis 2a. As ownership of the second largest
owner increases, audit fees decrease.
Hyphothesis 2b. As ownership of the second largest
owner increases, choice of Big 4 auditor decreases.
CEO ownership
A dominant belief in the literature is that as CEO own-
ership increases, her incentives align more with those of
other shareholders, reducing the agency problem that
arises from separation of ownership and control (e.g.,
Jensen & Meckling, 1976; Tirole, 2006). This is known as
the alignment effect. In our sample, CEO ownership is high
compared with prior studies (e.g., mean CEO ownership is
49% of the shares in our study). We predict that as CEO
ownership increases, agency costs decline and therefore
auditors supply less effort. Consistent with the belief that
CEO ownership reduces agency costs, we also predict that
?rms with high CEO ownership are less likely to employ
a Big 4 auditor.
Hyphothesis 3a. As CEO ownership increases, audit fees
decrease.
Hyphothesis 3b. As CEO ownership increases, choice of
Big 4 auditor decreases.
Family relationships
CEO and major shareholder
Major shareholders are often family members of the CEO
for private ?rms. There are interesting competing hypothe-
ses when the CEO is related to the major shareholder. Be-
cause of the family relationship, these shareholders no
longer act as independent monitors in disciplining CEOs’
decisions. In addition, family-controlled ?rms are likely to
suffer fromgreater horizontal agency costs. It may be easier
for major shareholders, who are family members of the
CEO, to extract private bene?ts from minority shareholders
or other stakeholders (Anderson & Reeb, 2004; DeAngelo &
DeAngelo, 2000; Morck, Shleifer, & Vishny, 1988). The rea-
son it may be easier to extract these bene?ts is that major
family owners typically have strong in?uence over choos-
ing members of the board (Johannisson & Huse, 2000). Con-
sequently, the monitoring effectiveness of the board may
be impaired when its composition is determined primarily
by the CEO’s family. These arguments would support the
idea that agency costs will increase when there is a family
relation between the CEO and major shareholder. In this
case, auditors are expected to supply more effort, and ?rms
are expected to demand Big 4 auditors.
An alternative view is that family member CEOs are less
likely to act in ways that opportunistically harm other
family members. That is, installing a family member as
the CEO could be a mechanism through which family-
owned companies can increase their monitoring of man-
agement and reduce the need for external monitoring. If
this effect dominates, the agency costs are smaller when
the CEO is a family member because familial ties are likely
to create closer alignment of the CEO’s preferences with
those of family owners.
The demand for a Big 4 in the presence of CEO-major
shareholder family relationship also presents interesting
counter-arguments. One the one hand, demand for a Big
4 auditor could increase to the extent that this family rela-
tionship increases agency costs. Agency costs would be
higher for reasons discussed above (e.g., lack of indepen-
dent monitoring by major shareholders and expropriation
of minority shareholders). In this setting, CEOs potentially
bene?t by signaling their commitment to higher-quality
reporting.
On the other hand, demand for a Big 4 could decrease
under at least two conditions. First, while hiring a Big 4
auditor has commitment value for the ?rm (and the
CEO), the CEO in a family-relationship setting may wish
to reduce audit cost by not hiring a Big 4 auditor. Major
shareholders’ family relationship with the CEO may negate
the need for costly independent veri?cation by Big 4 audi-
tors. The saved resources by using a less costly auditor in-
crease ?rm value, which is in the best interest of both the
CEO and major shareholders (who are in the same family).
Second, the CEO also has incentives not to demand a high-
er-quality auditor when this means that her ability (or her
family’s ability, including the major shareholder’s ability)
to extract private bene?ts from the ?rm would be limited
through such a hire. Presumably, Big 4 auditors would
have a greater ability to limit these private bene?ts.
Thus, for this CEO family-related agency setting, the de-
mand for a Big 4 auditor re?ects the trade-off between the
bene?ts from signaling higher-quality reporting versus the
costs of additional audit fees and reduced consumption of
private bene?ts. Because of competing arguments, we state
our fourth hypothesis as two-sided:
Hyphothesis 4a. When a family relationship exists
between the major shareholder and the CEO, audit fees
are affected.
Hyphothesis 4b. When a family relationship exists
between the major shareholder and the CEO, choice of
Big 4 auditor is affected.
Board independence
Boards are meant to protect shareholders’ assets and
the interests of the company’s other stakeholders (e.g.,
creditors and employees). In this sense, boards are directed
to monitor the activities of managers. An extensive litera-
ture exists which supports the notion that more indepen-
dent boards more effectively monitor managers’
activities. Firms with more independent boards commit
less ?nancial statement fraud (Beasley, 1996) and have less
earnings management or provide fewer discretionary
accruals (Dechow et al., 1996; Jaggi, Leung, & Gul, 2009;
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 505
Peasnell, Pope, & Young, 2000; Xie, Davidson, & DaDalt,
2003).
However, consistent with our two-sided predictions for
H
4a
, as the proportion of board members from the CEO’s
family increases, audit effort could either increase or de-
crease. With respect to the demand for a Big 4 auditor,
the relation is also not obvious because the CEO faces con-
?icting incentives. The additional (potentially unneces-
sary) costs associated with hiring a Big 4 auditor and the
reduced ability of the CEO or her family board members
to privately bene?t from these family relationships reduce
the expected demand for a Big 4 auditor in this higher
agency cost setting. As a result, the demand for a Big 4
auditor may be confounded in the presence of family rela-
tionships between the CEO and board members.
We propose two hypotheses related to board indepen-
dence. First, we predict that family relationships between
major shareholders and the board imply that owners have
insiders to monitor managers, which makes audited infor-
mation less important. In other words, fewer agency con-
?icts reduce the need for auditing. This setting would
result in the reduced supply of auditor effort and less de-
mand for a Big 4 auditor.
Second, if there are family ties between the CEO and
board members, given that there are competing arguments
as to whether the board will act more or less indepen-
dently of the CEO, our hypothesis is two-sided. We sum-
marize these hypotheses as:
Hyphothesis 5a. As the proportion of board members
from the largest owning family increases, audit fees
decrease.
Hyphothesis 5b. As the proportion of board members
from the largest owning family increases, choice of Big 4
auditor decreases.
Hyphothesis 6a. As the proportion of board members
from the CEO’s family increases, audit fees are affected.
Hyphothesis 6b. As the proportion of board members
from the CEO’s family increases, choice of Big 4 auditor is
affected.
Research design
For our auditor effort tests, we use model (1) to test our
six hypotheses of the relation between agency costs and
the supply of auditor effort after controlling for numerous
?rm and audit characteristics that explain normal audit
fees.
13
Speci?cally, we regress the log of raw audit fees
(LNFEE) on our six test variables as well as 24 client-?rm
characteristics, ?ve audit-?rm variables, and year and indus-
try ?xed effects (?rm subscripts omitted).
LNFEE
t
¼ a
0
þa
1
OWNER CONC
t
þa
2
OWNER SECOND
t
þa
3
OWNER CEOt
þa
4
FAM CEO=OWNER
t
þa
5
FAM BOARD=OWNER
t
þa
6
FAM BOARD=CEO
t
þa
7
LNSALES
t
þa
8
LNEMPLOY
t
þa
9
LEV
t
þa
10
CHLEVt
þa
11
INVREC
t
þa12GROWTH
t
þa
13
ROA
t
þa
14
INCPIC
t
þa
15
DECPIC
t
þa
16
INTANG
t
þa
17
UNITEMS
t
þa
18
NOINDt
þa
19
FOREIGN
t
þa
20
LNSUB
t
þa
21
FYE
t
þa
22
INVESTMENTS
t
þa
23
ACQUISITION
t
þa
24
LOSS
t
þa
25
CURRATIO
t
þa
26
BIG4
t
þa
27
GAAP
t
þa
28
CHGAAP
t
þa
29
INDSPEC
t
þa
30
CHAUDITOR t þa
31
LASTYRAUDIT
t
þa
32
MODOPINIONS
t
þa
32
CHCEO
t
þa
33
CITY
t
þRa
34;c
COUNTY
c
þRa
35;i
IND
i
þRa
36;y
YR
y
þv
t
ð1Þ
Appendix lists all variables employed and their de?ni-
tions. a
1
through a
6
test hypotheses H
1a
through H
6a
, while
the remaining variables serve as controls. We measure
ownership concentration (OWNER_CONC) using the Her?n-
dahl index and expect that as ownership concentration in-
creases, agency costs decrease (a
1
< 0). To test the second
hypothesis, we use ultimate ownership of the second larg-
est owner (OWNER_SECOND), and expect that as ownership
increases, agency costs decrease (a
2
< 0).
14
For the third
hypothesis, we use the fraction of shares owned by the
CEO as our primary measure (OWNER_CEO) and predict a
negative effect of CEO ownership on agency costs (a
3
< 0).
In Section Non-linearity of CEO ownership, we test whether
this relation is non-linear.
For our fourth hypothesis, we use an indicator variable
equal to one if the CEO is a member of the largest owning
family (FAM_CEO/OWNER). We predict that family owner-
ship has an effect on agency costs (a
4
–0). We refrain from
making a directional hypothesis due to the opposing forces
that are at work. For the ?fth hypothesis, we measure fam-
ily in?uence of the largest shareholder on the board as the
proportion of board members related to the largest share-
holder (FAM_BOARD/OWNER). We expect a negative rela-
tion (a
5
< 0). Finally, for our sixth hypothesis the test
variable is the proportion of board members related to
the CEO (FAM_BOARD/CEO). FAM_BOARD/CEO can thus be
interpreted as a measure of the strength of family ties be-
tween the CEO and the board.
15
As with our fourth hypoth-
eses, we refrain from a directional hypothesis (i.e., a
6
–0).
13
This approach is similar to ?rst estimating abnormal audit fees as the
residuals from a regression of total audit fees on the control variables and
then using the residuals as the dependent variable in a regression on our
test variables. No inferences are affected if we use that approach instead.
14
We provide a caveat that there is a possible negative mechanical
relation between OWNER_SECOND and OWNER_CONC (see correlation in
Table 2 below). When excluding either of these variables from the model,
inferences remain the same for the demand model (discussed next). For the
effort model, we ?nd that OWNER_CONC becomes insigni?cant when
OWNER_SECOND is dropped from the test.
15
Note that FAM_BOARD/OWNER and FAM_BOARD/CEO measure the
degree of family in?uence rather than just the existence of such in?uence.
506 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
Our choice of control variables is guided by prior re-
search (e.g., DeFond, Raghunandan, & Subramanyam,
2002; Frankel, Johnson, & Nelson, 2002; Choi, Kim, & Zang,
2010; Hope & Langli, 2010) and we additionally employ
other variables relevant for our private ?rm setting. There-
fore, model (1) examines whether our ownership charac-
teristics and family relationships explain audit fees
beyond an extensive set of factors previously identi?ed in
the literature.
We include the natural log of sales (LNSALES), natural log
of number of employees (LNEMPLOY), leverage (LEV) and
change in leverage (CHLEV), proportion of inventory and
receivables to total assets (INVREC), sales growth
(GROWTH), return on assets (ROA), increases or decreases
in paid-in capital (INCPIC and DECPIC), proportion of intan-
gible assets to total assets (INTANG), the existence of special
or unusual items (UNITEMS), the number of industries in
which the ?rm operates (NOIND), the existence of foreign
operations (FOREIGN), natural logarithm of the number of
subsidiaries (LNSUB), a ?scal year-end other than December
31 (FYE), long- and short-term investments in securities
scaled by sales (INVESTMENTS), acquisitions during the year
(ACQUISITIONS), negative earnings (LOSS), current ratio
(CURRATIO), Big 4 audit ?rm (BIG4), use of regular GAAP
(GAAP; as opposed to simpli?ed GAAP for smaller enter-
prises), changed to/from regular GAAP from/to simpli?ed
GAAP during the year (CHGAAP), auditor industry special-
ization (INDSPEC), change of auditor during the year (CHAU-
DITOR), an indicator for the last year of the incumbent
auditor (LASTYRAUDIT), audit report modi?cations
(MODOPINIONS), change in CEO during the year (CHCEO),
and whether the client resides in a municipality of?cially
classi?ed as a city or not (CITY). In addition, we include
?xed effects for the county in which the client ?rm is lo-
cated (COUNTY), two-digit industry (IND), and year (YR).
To test our six hypotheses of the demand for a Big 4
auditor, we employ the following logit regression model
with control variables (?rm subscript omitted):
BIG4
t
¼ b
0
þ b
1
OWNER CONC
t
þ b
2
OWNER SECOND
t
þ b
3
OWNER CEO
t
þ b
4
FAM CEO=OWNER
t
þ b
5
FAM BOARD=OWNER
t
þ b
6
FAM
B
OARD=CEO
t
þ b
7
LNTA
t
þ b
8
LEV
t
þ b
9
UNSECURED
t
þ b
10
LOSS
t
þ b
11
FOREIGN
t
þ b
12
EXANTEFIN
t
þ b
13
INCPIC
t
þ b
14
ROA
t
þ b
15
CITY
t
þRb
16;c
COUNTY
c
þRb
17;i
IND
i
þRb
18;y
YR
y
þg
t
ð2Þ
b
1
through b
6
test hypotheses H
1b
through H
6b
, while
the remaining variables serve as controls. BIG4 is an indica-
tor variable for whether the ?rm employs a Big 4 auditing
?rm or a forerunner. We expect that our agency variables
relate to the demand for a Big 4 auditor in the same way
that they relate to auditor effort. Speci?cally, our indicator
variable for a Big 4 auditor is expected to relate negatively
to OWNER_CONC, OWNER_SECOND, OWNER_CEO, and
FAM_BOARD/OWNER. For our two CEO family-related vari-
ables (FAM_CEO/OWNER and FAM_BOARD/CEO), we provide
expectations consistent with those for auditor effort. That
is, family relationships between the CEO and either the
major shareholder (FAM_CEO/OWNER) or board members
(FAM_BOARD/CEO) may decrease or increase the demand
for Big 4 auditors.
We base the control variables on prior research that has
examined the choice of Big 4 auditors (e.g., Khurana &
Raman, 2004; Knechel, Niemi, & Sundgren, 2008). In
particular, we control for ?rm size (LNTA), leverage (LEV),
unsecured debt (UNSECURED), loss ?rms (LOSS), the per-
centage of foreign subsidiaries (FOREIGN), operating cash
?ow less net investments in tangible and intangible ?xed
assets scaled by current assets (EXANTEFIN), increase in
paid-in capital (INCPIC), return on assets (ROA), and
whether the client ?rm is located in a city or not (CITY).
Finally, the logit regression models include county, indus-
try, and year ?xed effects. As an alternative speci?cation,
we additionally include controls for four alternative
ownership characteristics ( INSTITUTIONAL, STATE, INTER-
NATIONAL, and INDUSTRIAL).
Data on ownership and family relationships
We obtain our data from two sources. First, ?rm-spe-
ci?c information, such as the ?rm’s ?nancial information,
auditor, CEO, board members, and owners, comes from
Experian AS. Experian collects information from the
Brønnøysund Register Center (BRC), which is an adminis-
trative agency responsible for a number of national control
and registration schemes for business and industry. These
data are publicly available. Second, information on family
relationships comes from the National Register Of?ce
(NRO). In contrast to BRC data, NRO data are generally
not publicly available. Speci?cally, the social security num-
bers of any person having roles as owners, CEOs, or board
members are not available unless special permissions are
obtained. Family relationships among individuals are also
not publicly available. We gained access to these data
through a contract with the Centre for Corporate Gover-
nance Research (CCGR) at the Norwegian Business School.
CCGR obtained permissions to gather and merge data from
the two sources in accordance with con?dentiality and se-
crecy rules set forth by the Norwegian Data Inspectorate,
an independent administrative body that is set up to en-
sure that private and public storage of data are in accor-
dance with the Norwegian Personal Data Act.
In Norway, all limited liability ?rms, independent of size
and listing status, must send audited ?nancial statements
to the Center for Annual Accounts (CAA), which is part of
the BRC. According to the Accounting Act (paragraph 7–
26), limited liability ?rms that do not qualify as ‘‘small
enterprises’’ are required to disclose its 20 largest share-
holders and their shareholdings in the notes as long as indi-
vidual shareholdings exceed 1%. Firms that qualify as small
enterprises must disclose names and shareholdings of the
ten largest shareholders as long as the shareholder owns
5% or more. In addition, the Accounting Act requires disclo-
sure of names and shareholdings of all CEOs and members
of the Board and the Corporate Assembly. Experian collects
ownership information from the notes and to the extent
possible looks up the social security number of each owner.
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 507
The names of CEOs and board members are collected from
the Central Coordinating Register for Legal Entities (CCRLE),
which is also a part BRC. Data that identify CEOs and board
members (fromCCRLE) are merged with data on ownership
(from CAA) using the social security number for each indi-
vidual that occurs in the two databases.
16
The notes to the accounts specify direct ownership for
each ?rm. Utilizing the fact that direct ownership is avail-
able for all limited liability ?rms, data on ultimate owner-
ship for each ?rm has been calculated by tracing
ownership structures. Information on families’ ultimate
ownership and family in?uence on the CEO and boards
has been constructed by checking for relationships by
blood and/or marriage for any pair of persons that occurs
for owners, board members, CEOs, and Chairman using
data from NRO. Family relationships may cover up to four
generations. We assigned persons to the same family if
they (as of year-end 2007) are related through their par-
ents, grandparents, or by marriage. A single person without
relatives constitutes a family of size one.
Sample and results
Sample and descriptive statistics
Our sample selection procedure is explained in Table 1.
Our sample period spans 2000 through 2007. We include
only ?rms that are not publicly traded on a stock exchange.
We exclude ?rms with sales or total assets less than NOK
800,000 and ?rms with missing information concerning
audit fees or with fees below NOK 3000. We further elimi-
nate ?nancial ?rms, ?rms with missing information,
subsidiaries, and ?rm controlled by unknown owners or
the State.
17
Because the data provider did not provide a split
between audit and non-audit fees for years 2003 through
2005, our sample covers the 5-year period from 2000–2002
and 2006–2007. We have a ?nal sample of 185,109 ?rm-year
observations. Our tests are based on pooled samples, and we
use robust standard errors clustered at the ?rm level.
18
Panel A of Table 2 provides descriptive statistics. As can
be expected for a sample of private ?rms, ?rms are small
on average (mean SALES of NOK 16 million) but with a
large standard deviation. Audit fees (FEE) have a mean of
NOK 21,452, and 18.1% of ?rms employ Big 4 auditors. As
expected, ownership concentration (OWNER_CONC) is
quite high for private ?rms, with a mean Her?ndahl index
of 0.605. CEO ownership (OWNER_CEO) is also quite high
(mean of 49.2%). On average, the CEO comes from the larg-
est family (FAM_CEO/OWNER) in 70.6% of the cases, 71.2%
of board members come from the largest family
(FAM_BOARD/OWNER), and 15.1% of the board members
are family members of the CEO (FAM_BOARD/CEO). Panel
B presents Spearman correlation coef?cients among the
dependent variables, test variables, and primary control
variables.
19
As expected, most of the variables are signi?-
cantly correlated. However, unreported variance in?ation
factors and condition indices give no indication that multi-
collinearity is a serious issue in our multivariate analyses.
20
Important for our tests, and consistent with our unique data
on ownership and family relationships, we are interested in
testing the incremental effect of one agency characteristic
while holding the others constant.
Tests of hypotheses
Auditor effort tests
We report the results of tests of whether auditors’ fees
(i.e., auditors’ effort) vary with agency con?icts in private
?rms in Table 3. Column 1 reports our primary speci?ca-
tion with all test variables included simultaneously and
measured as continuous variables (except FAM_CEO/OWN-
ER which is an indicator variable). The regression has an
adjusted R
2
of 52.8% and we note that most of the control
variables are signi?cant in the predicted direction. For
example, audit fees are higher for ?rms that are larger,
have more ?nancial leverage, have unusual items, are more
diversi?ed (in terms of industries and subsidiaries), have
greater foreign operations, engage in acquisitions, and em-
ploy a Big 4 auditor.
21
The estimated coef?cient on OWNER_CONC is negative
and signi?cant, consistent with our prediction in H
1a
that
greater ownership concentration leads to easier monitor-
ing of managers, translating into less effort by external
auditors. Next, the coef?cient on OWNER_SECOND is nega-
tive and signi?cant, a ?nding which is in line with H
2a
and
recent research in ?nance suggesting that a strong second
shareholder can mitigate potential exploitation of minority
shareholders by the controlling shareholder. We further
?nd a negative and signi?cant coef?cient on OWNER_CEO,
consistent with H
3a
that CEO ownership aligns managers’
interests with those of the ?rms’ other stakeholders.
The next three variables, FAM_CEO/OWNER, FAM_
BOARD/OWNER, and FAM_BOARD/CEO, measure differences
in monitoring and independence. Whereas FAM_CEO/OWN-
ER captures the family relationship between the CEO and
major shareholders, the other two measure the degree of
16
For any pair of individuals, our data source has classi?ed the pair in
terms of a basic family relationship type. These types are parent,
grandparent, great grandparent, great-great grandparent, and marriage.
Using these basic types, we have generated 18 types of family relationships
between two individuals, including relationships like brother, sister, cousin,
uncle/aunt, and nephew/niece.
17
To mitigate the potential in?uence of outliers in such a large sample,
we remove observations with (absolute) studentized residuals exceeding
two. No inferences are affected by this choice. Similarly, no conclusions
change if we instead winsorize all continuous variables at the 1st and 99th
percentiles.
18
Results are consistent when employing the Fama–MacBeth approach.
19
Note that correlation coef?cients are provided for descriptive purposes
only and do not constitute tests of our hypotheses. Speci?cally, we are
testing our agency cost proxies against a proxy for abnormal audit fees;
hence it is essential to control for the normal factors that explain fees (i.e.,
risk and effort variables). It is also important to control for other agency-
related variables, as we do in our multiple regression tests.
20
Speci?cally, no variance in?ation factors are above 3.04 in any of the
speci?cations tabulated in Table 3 and Table 4.
21
The estimated coef?cient on INDSPEC is negative. The Pearson corre-
lation between audit fees and INDSPEC is signi?cantly positive (0.109) and
the negative regression coef?cient is likely caused by correlation with the
?rm size controls and BIG4 (i.e., INDSPEC is positive and signi?cant when
the model is estimated on a subset of ?rms with non-Big 4 auditors). In
addition, CHLEV has the opposite sign of expectations. No inferences are
affected if we exclude that (or any other) control variable from the test.
508 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
board independence based on family relationships. We
?nd a positive and signi?cant coef?cient for FAM_CEO/
OWNER. In other words, we document that when the CEO
is also a member of the largest family, auditors increase
their effort. Thus, our empirical ?ndings support the idea
that, because of the family relationship, these shareholders
no longer act as independent monitors in disciplining
CEOs’ decisions, thus increasing agency costs.
We further ?nd empirical support for H
5a
. Auditor effort
is decreasing in the proportion of board members related
to the largest family owner (FAM_BOARD/OWNER), consis-
tent with fewer agency con?icts between family owners
and the board. Finally, the empirical results indicate that
audit effort is increasing in the proportion of board mem-
bers related to the CEO (FAM_BOARD/CEO). This ?nding is
consistent with the loss of board independence when fam-
ily relationships exist between the CEO and board mem-
bers. Thus, to control for the increase in agency con?icts,
auditors exert more effort.
22
In the second column of Table 3, we report results of
regressions in which we replace continuous test variables
with indicator variables equal to one when equal to or
above the mean, zero otherwise. Results are similar for ?ve
the six tests. For FAM_CEO/OWNER (H
4a
), the coef?cients
changes from signi?cantly positive to only marginally sig-
ni?cant (t = 1.56).
In the next six columns, we provide results with one
test variable included at a time. We note two differences
from the full model results: OWNER_SECOND is no longer
signi?cant and FAM_CEO/OWNER is now signi?cantly neg-
ative (as opposed to signi?cantly positive). These differ-
ences highlight the importance of testing for multiple
agency settings simultaneously. As we discussed previ-
ously, prior studies have been limited to looking at one
or a few agency variables because of data limitations. We
are able to provide detailed tests of six agency settings.
Thus, we base our primary conclusions on the full model,
which we believe better controls for multiple agency
settings.
Demand for high-quality auditor tests
In Table 4, we report results of the analyses that mea-
sure the demand for higher-quality auditing by ?rms/own-
ers to reduce agency con?icts. As predicted in H
1b
, H
2b
, and
H
5b
, we observe negative and signi?cant coef?cients on the
three variables that capture non-CEO-related agency con-
?icts: OWNER_CONC, OWNER_SECOND, and FAM_BOARD/
OWNER. These results are consistent with those obtained
when testing audit fees (or auditor effort). In contrast,
the coef?cient for OWNER_CEO is negative, as predicted
in H
3b
, but not signi?cant.
For our two CEO family-related variables, FAM_CEO/
OWNER (H
4b
) and FAM_BOARD/CEO (H
6b
), we ?nd no sig-
ni?cant evidence that the demand for a Big 4 auditor is af-
fected when a family relationship exists between the CEO
and major shareholder or as the number of board mem-
bers related to the CEO increases. One interpretation of
these insigni?cant coef?cients is that, while hiring a high-
er-quality auditor may bring value to the ?rm as a whole
(in part by reducing agency con?icts), CEOs in these fam-
Table 1
Sample selection criteria.
2000 2001 2002 2003 2004 2005 2006 2007 Total
Private and public limited liability companies 136,140 138,745 141,146 141,991 144,426 157,710 180,709 192,011 1232,878
Exclusion criteria:
Public limited liability companies 582 571 529 492 463 459 471 384 3951
Operating revenue less than 800,000 NOK 55,180 56,092 57,081 56,142 55,151 64,325 81,038 87,257 512,266
Total assets less than 800,000 NOK 14,218 13,747 13,836 14,286 14,427 14,598 13,933 13,652 112,697
Missing information on audit fees, audit fees less than
3000 NOK, or unknown auditor
2465 2239 2034 71,071 74,385 78,328 13,904 2831 247,257
Missing information (e.g. industry af?liation, founding
date, prior year ?nancial statement) or ?rm using IFRS
6419 7400 6862 0 0 0 9252 11,690 41,623
Subsidiaries 11,088 12,131 12,199 0 0 0 16,274 25,242 76,934
Information on family ownership not available 7439 6134 7058 0 0 0 8421 8635 37,687
Aggregate ownership less than 50% 238 381 420 0 0 0 266 383 1688
Controlled by unspeci?ed owners or by the State 1298 724 678 0 0 0 781 693 4174
Observations with absolute value of studentized residuals
greater than two
1898 2008 2079 0 0 0 1694 1814 9493
Sample size 35,315 37,318 38,370 0 0 0 34,675 39,431 185,109
Note that the data provider does not provide a breakdown of audit remuneration into audit fees and non-audit fees for the years 2003, 2004, and 2005.
22
To assess the economic signi?cance of our results, we conduct two
tests. First, we consider the effect of a change from the ?rst to the third
quartile of the distribution of our test variables (holding other variables
constant). The effect for each translates into approximately NOK 1000 or
NOK 6000 in total. While this does not appear to be a large amount, note
again that the sample ?rms in our study are very small compared with
those examined in prior research. For example, using data from Statistics
Norway, the aggregate effect is equivalent to 18.8 man-hours based on
average salaries for full-time employed ‘‘senior of?cials and managers in
professional, scienti?c, and technical activities,’’ which includes auditors.
Compared with all full-time employed persons in Norway (in the profes-
sional, scienti?c, and technical activities), the aggregate effect is equivalent
to 34.9 (27.3) man-hours. In addition, we also compare these effects
directly with the average audit fees in our sample. The average effect for
our test variables is 4.7% and in aggregate the effect is 28.5% of average fees.
Second, we compare the effects of the test variables with those of the
control variables. In order to conduct such a comparison, we ?rst
standardize the estimated coef?cients so that they are all directly
comparable. Consistent with many prior studies, ?rm size is clearly the
most important determinant of fees as well as the demand for a Big 4
auditor. However, for the fee regressions, in untabulated analyses we
observe that OWNER_SECOND has an equal or greater effect than 12 of the
other control variables. The other test variables also fare quite well in such
a comparison. Overall we conclude that, although ?rm size is economically
more signi?cant than our test variables, the test variables have a non-trivial
effect on both audit effort and audit demand.
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 509
Table 2
Descriptive statistics.
Mean St. dev. 5% 25% Median 75% 95%
Panel A: Distribution of variables (N = 185,109)
Corporate governance
OWNER_CONC 0.605 0.296 0.194 0.355 0.500 1 1
OWNER_SECOND 0.219 0.184 0 0 0.226 0.35 0.5
OWNER_CEO 0.492 0.370 0 0.143 0.5 0.923 1
FAM_CEO/OWNER 0.706 0.455 0 0 1 1 1
FAM_BOARD/OWNER 0.712 0.341 0 0.4 1 1 1
FAM_BOARD/CEO 0.151 0.268 0 0 0 0.333 0.667
INSTITUTIONS 0.012 0.108 0 0 0 0 0
STATE 0.003 0.053 0 0 0 0 0
INTERNATIONAL 0.010 0.101 0 0 0 0 0
INDUSTRIAL 0.011 0.106 0 0 0 0 0
Financial variables
ACQUISITIONS 0.047 0.213 0 0 0 0 0
CHLEV À0.007 0.143 À0.221 À0.055 0 0.019 0.242
CURRATIO 1.679 1.727 0.401 0.968 1.246 1.772 4.067
DECPIC 0.012 0.111 0 0 0 0 0
EXANTEFIN 0.237 0.686 À0.871 À0.241 0.357 0.591 1.222
FEE 21.452 17.930 7 12 17 25 49
GROWTH 0.279 1.080 À0.307 À0.038 0.069 0.236 1.150
INCPIC 0.066 0.248 0 0 0 0 1
INTANG 0.020 0.066 0 0 0 0.006 0.109
INVESTMENTS 0.326 0.784 0.006 0.036 0.111 0.272 1.215
INVREC 0.181 0.170 0 0.061 0.147 0.249 0.487
LEV 0.267 0.286 0 0 0.180 0.457 0.832
LNFEE 2.868 0.602 1.946 2.485 2.833 3.219 3.892
LOSS 0.214 0.410 0 0 0 0 1
ROA 0.160 0.212 À0.131 0.042 0.124 0.256 0.561
SALES 16,242.9 182,098.9 1096.0 2523.0 5020.0 11,186.0 43,862.0
TA 13,860.4 133,659.0 935.0 1621.0 3139.0 7185.0 32,575.0
UNITEMS 0.036 0.186 0 0 0 0 0
UNSECURED À0.272 0.286 À0.884 À0.452 À0.153 À0.037 0.000
Other variables
AGE 12.814 12.100 2 5 10 16 33
EMPLOY 9.434 61.611 0 1.605 3.669 7.648 26.717
BIG4 0.181 0.385 0 0 0 0 1
CHAUDITOR 0.083 0.276 0 0 0 0 1
CHCEO 0.051 0.219 0 0 0 0 1
CHGAAP 0.006 0.077 0 0 0 0 0
CITY 0.558 0.497 0 0 1 1 1
FOREIGN 0.015 0.120 0 0 0 0 0
FYE 0.001 0.028 0 0 0 0 0
GAAP 0.036 0.187 0 0 0 0 0
INDSPEC 0.027 0.064 0 0.000 0.001 0.005 0.175
LASTYRAUDIT 0.084 0.277 0 0 0 0 1
MODOPINONS 0.287 0669 0 0 0 0 2
NOIND 1.514 0.843 1 1 1 2 3
SUB 0.267 1.507 0 0 0 0 1
Variables LNFEE 1 2 3 4 5 6 7 8 9 10 11 12 13
Panel B: Spearman correlation coef?cients (N = 185,109)
[1] BIG4 0.092
*
[2] OWNER_CONC À0.121
*
À0.050
*
[3] OWNER_SECOND 0.015
*
À0.030
*
À0.572
*
[4] OWNER_CEO À0.084
*
À0.052
*
0.589
*
À0.369
*
[5] FAM_CEO/OWNER À0.044
*
À0.042
*
0.286
*
À0.135
*
0.663
*
[6] FAM_BOARD/OWNER À0.171
*
À0.062
*
0.493
*
À0.227
*
0.376
*
0.416
*
[7] FAM_BOARD/CEO 0.029
*
0.002 À0.107
*
0.138
*
À0.031
*
0.342
*
0.273
*
[8] GROWTH À0.053
*
À0.006
*
À0.023
*
0.012
*
À0.015
*
À0.022
*
À0.039
*
À0.038
*
[9] LEV À0.001 0.015
*
À0.012
*
À0.002 À0.042
*
À0.012
*
À0.001 À0.040
*
À0.015
*
[10] LOSS 0.013
*
0.010
*
À0.011
*
À0.016
*
À0.022
*
À0.021
*
À0.041
*
À0.065
*
0.037
*
0.251
*
[11] ROA À0.101
*
À0.032
*
0.044
*
0.026
*
0.054
*
0.029
*
0.065
*
0.305
*
À0.107
*
À0.397
*
À0.273
*
[12] SALES 0.609
*
0.082
*
À0.130
*
0.035
*
À0.068
*
À0.038
*
À0.189
*
À0.129
*
0.171
*
À0.087
*
À0.063
*
0.017
*
[13] UNSECURED 0.062
*
À0.011
*
0.009
*
0.010
*
0.043
*
0.004 À0.045
*
À0.104
*
0.317
*
À0.524
*
À0.062
*
0.190
*
0.130
*
[14] EMPLOY 0.549
*
0.051
*
À0.149
*
0.051
*
À0.056
*
À0.030
*
À0.220
*
À0.173
*
0.184
*
À0.122
*
À0.050
*
À0.028
*
0.772
*
0.082
*
The table reports Spearman correlation coef?cients between dependent variables and tests and main control variables. The sample is described in Table 1
and variables are de?ned in the Appendix.
*
Signi?cant correlation coef?cient at p < 0.01.
510 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
ily-relationship settings may perceive that the additional
costs of hiring a Big 4 auditor is not necessary. Major
shareholders and board members could believe that inde-
pendent veri?cation of a CEO family member is not war-
ranted, allowing ?rm value to increase through reduced
audit fees. In addition, the CEO may have personal incen-
tives not to demand a Big 4 auditor to the extent that
these external monitors make it more dif?cult for the
CEO (or the CEO’s family) to extract resources from the
?rm. Thus, we attribute the weaker relation between
BIG4 and our CEO family-related agency variables to the
trade-off that exists when deciding the net bene?ts to hir-
ing a Big 4 auditor. These trade-offs are less apparent
when CEO family-related agency settings are not consid-
ered. We conclude that ?rms demand greater audit quality
in the presence of agency con?icts (that are not CEO fam-
ily-related).
23,24
We then provide results using indicator variables for
the test variables (measured as above) in Column 2. Results
are similar to those using continuous measures, except
that OWNER_CEO changes from insigni?cantly negative
(t = À1.26) to signi?cantly negative (t = À3.45), consistent
with H
3b
. The remaining columns tabulate results with test
variables included one at a time. Whereas FAM_CEO/OWN-
ER and FAM_BOARD/CEO were insigni?cant in the full mod-
el, they are now signi?cantly negative. These differences
from the full model again highlight an advantage of con-
trolling for multiple agency variables to provide more com-
plete inferences.
Additional analyses and sensitivity tests (untabulated)
Additional controls for ?rm size
In our tabulated analyses, we include two explicit con-
trols for client ?rm size (LNSALES and LNEMPLOY). We also
include additional variables that relate to size, including
BIG4, FOREIGN, LNSUB, ACQUISITIONS, and CITY. However,
given the potential importance of ?rm size as a control var-
iable, we conduct a number of additional tests. First, we re-
place LNEMPLOY (or LNSALES) with log of total assets in the
audit fee model. Second, to test for non-linear size effects,
we add square root terms of both LNSALES and LNEMPLOY
to the model. Third, we add the square of both the size
variables. Fourth, we include interaction terms between
the size controls and all variables included in the model.
Fifth, as a related test, we either break our size controls
into quintiles and introduce them separately as controls,
or we include these as both main and interacting effects
in the model. Finally, we exclude the smallest 10% of ?rms
(based on sales). In all these tests, results are consistent
with those tabulated.
Big 6 instead of Big 4
As the Big 4 audit ?rms have a relatively small market
share among private ?rms (18.1% in our sample), as an
additional test we replace Big 4 with Big 6. Speci?cally,
in addition to the international Big 4 ?rms we also include
two other large audit ?rms: BDO Noraudit and Inter Revis-
jon. Combined, the Big 6 ?rms have a market share of
29.9%. In fact, BDO Noraudit and Inter Revisjon are the
two largest audit ?rms in terms of number of private client
?rms in Norway, with Ernst and Young and PWC close
behind.
We ?rst repeat the auditor effort tests of Table 3 by
replacing BIG4 with BIG6 in the audit fee model. No infer-
ences are affected.
25
More importantly, we redo the auditor
demand test in Table 4 using Big 6 as the dependent vari-
able. Again, results are quite similar.
26
Non-linearity of CEO ownership
To explore the possibility that the effect of CEO owner-
ship is non-linear, we add the square root of CEO owner-
ship to the models. We ?rst note that adding this
variable does not affect the inferences for the other vari-
ables and further observe that the square root variable is
negative and marginally statistically signi?cant for the ef-
fort regression. Thus, there is some support of a non-linear
relation for CEO ownership.
Controls for CEO compensation
Executives can extract rents from the company by also
receiving excessive compensation. We test if our results
are robust to controlling for CEO compensation by includ-
ing either raw CEO compensation or CEO compensation
scaled by ?rm size. No inferences are affected in these
tests.
Time period effects
In the aftermath of well-known international account-
ing scandals and the introduction of SOX in the US, changes
were made to Norwegian accounting and auditing regula-
tions that were aimed at increasing accounting and audit
quality. The changes took effect between the two periods
covered in our sample. Even though we control for time-
period effects through year ?xed effects, we separately
examine the time periods 2000–2002 and 2006–2007 in
order to assess if the relations have changed. We ?nd al-
23
Landsman, Nelson, and Rountree (2009) model the ‘‘match’’ (or
‘‘misalignment’’) between the audit ?rm and its client and use this binary
variable in their empirical tests. Our regressions include proxies for all the
variables included in their model (i.e., LNSALES; LNEMPLOY; ACQUISITIONS;
CHLEV; INCPIC; DECPIC; GROWTH). No inferences are affected if we replace
the individual control variables with a binary variable computed following
Landsman et al.’s approach.
24
The change in the probability of a ?rm choosing a Big 4 auditor
increases by 55.5% (i.e., from 0.076 to 0.13) when all six test variables
change from the third to ?rst quartile (holding all other variables constant
at their mean values). The percentage changes in the probability of a ?rm
choosing a Big 4 auditor vary between 2.4% and 24.5% for the six variables.
We further observe that, using standardized coef?cients, OWNER_SECOND
has a greater effect than all non-size control variables. Again, the other test
variables continue do compare well.
25
As an alternative to controlling for Big 4 in the audit fee model, we run
the auditor effort tests separately for Big 4 and non-Big 4 client ?rms.
Although this is not one of our hypotheses, it is interesting to note that four
of the six estimated coef?cients are larger in magnitude for the Big 4 than
for the non-Big 4 subsample. This observation is consistent with Big 4 ?rms
being more sophisticated than other audit ?rms and thus responding more
strongly to variations in agency con?icts.
26
Speci?cally, in this speci?cation OWNER_CONC, OWNER_SECOND, and
FAM_BOARD/OWNER remain signi?cant in the hypothesized direction, while
OWNER_CEO is now signi?cantly negative (as predicted in H
3b
) and
FAM_CEO/OWNER also becomes signi?cantly negative (as discussed in H
4b
).
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 511
Table 3
OLS regression of audit fees on test variables and control variables.
Predicted sign Model
1 2 3 4 5 6 7 8
OWNER_CONC (H
1a
) À À0.032
***
À0.024
***
À0.039
***
(À4.16) (À5.22) (À8.37)
OWNER_SECOND (H
2a
) À À0.070
***
À0.032
***
0.004
(À7.15) (À7.56) (0.63)
OWNER_CEO (H
3a
) À À0.030
***
À0.023
***
À0.034
***
(À5.20) (À6.82) (À9.01)
FAM_CEO/OWNER (H
4a
) +/À 0.012
***
0.007
*
À0.009
***
(3.11) (1.79) (À3.16)
FAM_BOARD/OWNER (H
5a
) À À0.038
***
À0.021
***
À0.039
***
(À7.83) (À6.83) (À9.52)
FAM_BOARD/CEO (H
6a
) +/À 0.028
***
0.017
***
0.020
***
(4.91) (5.24) (3.98)
LNSALES + 0.234
***
0.235
***
0.236
***
0.237
***
0.236
***
0.236
***
0.235
***
0.237
***
(97.16) (97.36) (97.93) (98.34) (97.86) (98.14) (97.70) (98.38)
LNEMPLOY + 0.150
***
0.150
***
0.151
***
0.153
***
0.152
***
0.153
***
0.151
***
0.152
***
(53.56) (53.70) (54.12) (54.59) (54.47) (54.59) (53.81) (54.58)
LEV + 0.078
***
0.078
***
0.078
***
0.078
***
0.077
***
0.078
***
0.078
***
0.077
***
(13.67) (13.60) (13.63) (13.62) (13.56) (13.62) (13.62) (13.57)
CHLEV + À0.064
***
À0.064
***
À0.066
***
À0.066
***
À0.065
***
À0.066
***
À0.065
***
À0.066
***
(À8.66) (À8.66) (À8.88) (À8.90) (À8.83) (À8.90) (À8.81) (À8.87)
INVREC + 0.261
***
0.262
***
0.263
***
0.265
***
0.264
***
0.265
***
0.262
***
0.265
***
(32.36) (32.44) (32.61) (32.82) (32.74) (32.81) (32.44) (32.84)
GROWTH + À0.049
***
À0.049
***
À0.049
***
À0.049
***
À0.049
***
À0.049
***
À0.049
***
À0.049
***
(À43.35) (À43.34) (À43.22) (À43.27) (À43.34) (À43.31) (À43.39) (À43.22)
ROA À À0.153
***
À0.154
***
À0.160
***
À0.162
***
À0.159
***
À0.161
***
À0.157
***
À0.161
***
(À23.62) (À23.85) (À24.75) (À25.10) (À24.67) (À24.96) (À24.37) (À25.06)
INCPIC + 0.066
***
0.067
***
0.069
***
0.071
***
0.069
***
0.070
***
0.068
***
0.071
***
(16.01) (16.21) (16.69) (17.20) (16.83) (17.09) (16.53) (17.26)
DECPIC + 0.100
***
0.101
***
0.105
***
0.107
***
0.105
***
0.106
***
0.103
***
0.107
***
(10.67) (10.73) (11.14) (11.35) (11.18) (11.30) (10.95) (11.37)
INTANG + 0.104
***
0.106
***
0.110
***
0.114
***
0.112
***
0.113
***
0.109
***
0.114
***
(4.89) (4.97) (5.18) (5.35) (5.24) (5.32) (5.14) (5.37)
UNITEMS + 0.072
***
0.072
***
0.072
***
0.072
***
0.072
***
0.072
***
0.072
***
0.072
***
(12.22) (12.25) (12.22) (12.21) (12.24) (12.23) (12.23) (12.22)
NOIND + 0.032
***
0.032
***
0.032
***
0.032
***
0.032
***
0.032
***
0.033
***
0.032
***
(18.76) (18.66) (18.90) (18.75) (18.82) (18.80) (18.96) (18.65)
FOREIGN + 0.111
***
0.112
***
0.113
***
0.114
***
0.114
***
0.114
***
0.112
***
0.115
***
(8.43) (8.51) (8.57) (8.66) (8.62) (8.63) (8.48) (8.70)
LNSUB + 0.046
***
0.046
***
0.046
***
0.045
***
0.045
***
0.045
***
0.046
***
0.045
***
(9.20) (9.10) (9.18) (8.99) (8.99) (8.99) (9.20) (8.96)
FYE + 0.117
**
0.117
**
0.128
***
0.128
***
0.120
**
0.128
***
0.118
**
0.130
***
(2.43) (2.45) (2.66) (2.66) (2.51) (2.66) (2.47) (2.69)
INVESTMENTS + 0.066
***
0.066
***
0.067
***
0.067
***
0.067
***
0.067
***
0.067
***
0.067
***
(31.22) (31.17) (31.63) (31.45) (31.28) (31.45) (31.74) (31.31)
ACQUISITIONS + 0.065
***
0.066
***
0.066
***
0.066
***
0.066
***
0.066
***
0.066
***
0.067
***
(11.36) (11.43) (11.53) (11.57) (11.57) (11.57) (11.42) (11.58)
LOSS + 0.045
***
0.045
***
0.046
***
0.047
***
0.046
***
0.046
***
0.045
***
0.047
***
(16.30) (16.42) (16.86) (17.05) (16.85) (16.97) (16.52) (17.11)
CURRATIO + 0.001
*
0.001
*
0.002
**
0.002
**
0.002
**
0.002
**
0.002
**
0.002
**
(1.86) (1.92) (2.16) (2.31) (2.16) (2.28) (2.23) (2.25)
BIG4 + 0.116
***
0.116
***
0.118
***
0.118
***
0.117
***
0.118
***
0.117
***
0.118
***
(21.19) (21.21) (21.50) (21.61) (21.46) (21.57) (21.41) (21.61)
GAAP ? À0.074
***
À0.074
***
À0.074
***
À0.075
***
À0.076
***
À0.075
***
À0.073
***
À0.075
***
(À7.13) (À7.15) (À7.13) (À7.20) (À7.27) (À7.22) (À7.05) (À7.20)
CHGAAP + 0.083
***
0.083
***
0.085
***
0.086
***
0.085
***
0.086
***
0.084
***
0.086
***
(5.65) (5.66) (5.75) (5.82) (5.75) (5.80) (5.72) (5.82)
INDSPEC + À0.390
***
À0.390
***
À0.390
***
À0.394
***
À0.391
***
À0.394
***
À0.392
***
À0.393
***
(À11.67) (À11.64) (À11.64) (À11.70) (À11.64) (À11.69) (À11.67) (À11.67)
CHUDITOR + 0.006
*
0.006
*
0.006
*
0.006
*
0.006
*
0.006
*
0.006
*
0.006
*
(1.67) (1.71) (1.70) (1.67) (1.65) (1.68) (1.67) (1.67)
LASTYRAUDIT +/À 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001
(0.30) (0.31) (0.24) (0.23) (0.28) (0.24) (0.24) (0.23)
MODOPINIONS + 0.033
***
0.033
***
0.033
***
0.032
***
0.032
***
0.032
***
0.033
***
0.032
***
(18.69) (18.61) (18.28) (17.63) (18.02) (17.70) (18.19) (17.67)
CHCEO +/À 0.021
***
0.023
***
0.030
***
0.033
***
0.023
***
0.030
***
0.027
***
0.033
***
(4.61) (5.10) (6.53) (7.17) (4.99) (6.49) (6.03) (7.17)
512 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
most identical results across the two periods. The only dif-
ference is that OWNER_CONC is not signi?cant in the earlier
time period for the audit fee test (H
1a
). Overall, our robust-
ness tests corroborate the main ?ndings tabulated in the
paper.
Concluding remarks
This study provides empirical evidence of agency con-
?icts associated with ownership characteristics and family
relationships for private ?rms in the Norwegian market.
Using a detailed data set obtained by special permission
from the government, we are able to measure ultimate
ownership concentration, ultimate ownership of the sec-
ond largest shareholder and of the CEO, and family rela-
tionships among owners, board members, and CEOs for
all Norwegian private limited liability ?rms. These data al-
low us to simultaneously test multiple sources of agency
con?icts.
We ?rst test for the extent of agency con?icts using
audit fees. Controlling for a large number of client-?rm
and audit-?rm characteristics, we predict and ?nd that
ownership characteristics and family relationships explain
audit fees. Speci?cally, we ?nd that audit fees decrease
(i.e., audit effort decreases) as ownership concentration in-
creases and as the proportion of shares held by the second
largest shareholder increases. The ?rst result is consistent
with greater ownership concentration alleviating agency
costs between shareholders and the managers (i.e., vertical
agency costs). Larger shareholders can more ef?ciently
monitor managers. The second result is consistent with re-
duced agency costs between controlling shareholders and
minority shareholders (i.e., horizontal agency costs). A sec-
ond large shareholder serves as a monitor of a controlling
shareholder who potentially has the ability to extract pri-
vate bene?ts from minority shareholders.
Related to the CEO, we ?nd that as CEO ownership in-
creases, audit fees decrease. This is consistent with in-
creased ownership aligning the incentives of the CEO
with those of the ?rms’ other stakeholders. In contrast,
holding CEO ownership constant, we ?nd that when the
CEO is a member of the largest owning family, audit fees
increase (i.e., auditor effort increases). This ?nding sug-
gests that shareholders are less likely to act as independent
monitors of the CEO when a family relationship exists,
increasing the probability of misappropriation by the CEO
or extraction of private bene?ts by controlling owners.
For board independence, we ?nd that audit fees de-
crease as the proportion of board members from the larg-
est owning family increases. This ?nding re?ects the
likelihood that those inside the ?rm (board members) are
more likely to act as effective monitors on behalf of those
outside the ?rm (controlling owners) when a family rela-
tion exists. We also ?nd that boards seem to lose their
independence as the proportion of board members from
the CEO’s family increases. This result follows from the
natural expectation that family relationships with the
CEO impair the oversight responsibility of the board.
As our second test, we consider each of our agency set-
tings and the ?rm’s demand for a Big 4 auditor. Consistent
with results found for the auditor effort tests, the demand
for a Big 4 auditor decreases with ownership concentra-
tion, level of ownership by the second largest owner, and
family relationships between the board and the largest
owner. We ?nd no relation between demand for a Big 4
auditor and CEO ownership for our main tests but we do
in sensitivity tests. Finally, for our two CEO family-related
agency cost settings, we observe no impact on the demand
for a Big 4 auditor when the CEO is related to the major
shareholder or as the number of board members related
to the CEO increases. These insigni?cant relations likely re-
?ect a trade-off between the bene?ts of more credible
reporting from using a Big 4 auditor versus the potential
costs of increased fees associated with a Big 4 auditor
and the reduced ability of the CEO (or the CEO’s family)
to extract resources from the ?rm.
Table 3 (continued)
Predicted sign Model
1 2 3 4 5 6 7 8
CITY + 0.064
***
0.065
***
0.065
***
0.064
***
0.064
***
0.064
***
0.064
***
0.064
***
(18.73) (18.76) (18.77) (18.63) (18.66) (18.60) (18.51) (18.73)
Constant 0.376
***
0.349
***
0.317
***
0.282
***
0.311
***
0.292
***
0.327
***
0.278
***
(14.61) (13.86) (12.74) (11.53) (12.60) (11.86) (13.12) (11.38)
County ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Industry ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Year ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Observations 185,109 185,109 185,109 185,109 185,109 185,109 185,109 185,109
Adjusted R
2
0.529 0.529 0.528 0.528 0.528 0.528 0.528 0.528
This table presents OLSÀregression coef?cients (t-values in parentheses) for the impact of ownership characteristics and family relationships on audit fees.
The dependent variable is LNFEE, the natural log of audit fees in NOK 1000. The sample is described in Table 1 and variables are de?ned in the Appendix
except for the test variables in Model 2 which are measured as indicator variables de?ned as 1 if the test variable is greater than or equal to the mean and 0
otherwise. County, industry, and year ?xed effects are included but omitted from table to conserve space. All regressions employ robust standard errors
clustered at the ?rm level.
*
Statistical signi?cance at the 0.10 level (two-sided).
**
Statistical signi?cance at the 0.05 level (two-sided).
***
Statistical signi?cance at the 0.01 level (two-sided).
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 513
Our research makes a contribution by providing a de-
tailed investigation of the agency settings for private ?rms.
Several differences exist between private and public ?rms,
and results for one group are not necessarily expected to
be the same as those for the other. In addition, in aggregate
private ?rms provide the majority of economic activity
throughout the world. We obtain a unique database for
private ?rms in Norway to provide tests of the impact of
ownership characteristics and family relationships on
agency con?icts. This database entails combining owner-
ship characteristics and family relationships using social
security numbers, available only through special permis-
sion from the government.
There is ample opportunity for future research to further
explore the issues we examine in this study. Perhaps the
most obvious suggestion would be to test for economic
consequences of variations in the agency con?icts and/or
audit variables that we consider. Another possibility is to
make further use of the register data available through link-
ing databases using social security numbers. For example,
auditor independence issues could be examined by consid-
ering auditors’ personal income and wealth. Other research
could make use of interesting aspects of alternative institu-
tional environments, or use research methods other than
archival data to explore agency con?icts in private ?rms.
Acknowledgements
We appreciate valuable comments from Ahmed
Abdel-Meguid, Erlend Kvaal, Alastair Lawrence, Gilad
Livne, Salvatore Miglietta, Mark Peecher (Editor), Ann
Vanstraelen, Joost van Buuren, two anonymous reviewers,
Table 4
Logit regressions of choice of Big 4 auditor on ownership, CEO, and board characteristics, as well as control variables.
Predicted sign Model
1 2 3 4 5 6 7 8
OWNER_CONC (H
1b
) À À0.227
***
À0.096
***
À0.078
**
(À4.16) (À2.89) (À2.22)
OWNER_SECOND (H
2b
) À À0.753
***
À0.258
***
À0.357
***
(À10.06) (À8.70) (À6.58)
OWNER_CEO (H
3b
) À À0.052 À0.087
***
À0.106
***
(À1.26) (À3.45) (À3.81)
FAM_CEO/OWNER (H
4b
) +/À À0.027 À0.023 À0.109
***
(À0.93) (À0.88) (À5.27)
FAM_BOARD/OWNER (H
5b
) À À0.219
***
À0.139
***
À0.248
***
(À6.04) (À5.96) (À8.35)
FAM_BOARD/CEO (H
6b
) +/À 0.027 0.010 À0.119
***
(0.61) (0.39) (À3.06)
LNTA + 0.335
***
0.339
***
0.357
***
0.360
***
0.355
***
0.356
***
0.347
***
0.361
***
(34.58) (35.12) (37.49) (38.21) (37.18) (37.69) (36.53) (38.32)
LEV + À0.442
***
À0.450
***
À0.467
***
À0.465
***
À0.466
***
À0.468
***
À0.466
***
À0.473
***
(À9.92) (À10.11) (À10.49) (À10.47) (À10.48) (À10.54) (À10.46) (À10.63)
UNSECURED + 0.361
***
0.365
***
0.381
***
0.384
***
0.380
***
0.377
***
0.361
***
0.378
***
(7.84) (7.92) (8.29) (8.37) (8.28) (8.21) (7.82) (8.21)
LOSS ? 0.130
***
0.133
***
0.148
***
0.147
***
0.147
***
0.146
***
0.140
***
0.148
***
(7.24) (7.39) (8.27) (8.19) (8.17) (8.14) (7.76) (8.26)
FOREIGN + 0.319
***
0.321
***
0.340
***
0.338
***
0.340
***
0.338
***
0.327
***
0.337
***
(4.23) (4.26) (4.52) (4.48) (4.52) (4.49) (4.33) (4.47)
EXANTEFIN + 0.036
***
0.037
***
0.037
***
0.038
***
0.037
***
0.037
***
0.037
***
0.038
***
(3.37) (3.43) (3.48) (3.53) (3.44) (3.44) (3.47) (3.53)
INCPIC + À0.017 À0.011 0.012 0.016 0.011 0.009 À0.002 0.014
(À0.64) (À0.41) (0.44) (0.60) (0.41) (0.36) (À0.07) (0.54)
ROA À À0.374
***
À0.389
***
À0.443
***
À0.439
***
À0.438
***
À0.439
***
À0.414
***
À0.449
***
(À8.39) (À8.73) (À9.94) (À9.89) (À9.85) (À9.89) (À9.30) (À10.11)
CITY + 0.081
***
0.081
***
0.084
***
0.082
***
0.083
***
0.082
***
0.080
***
0.081
***
(3.21) (3.21) (3.31) (3.23) (3.28) (3.23) (3.15) (3.20)
Constant À4.300
***
À4.523
***
À4.945
***
À4.945
***
À4.917
***
À4.902
***
À4.731
***
À5.002
***
(À25.84) (À28.01) (À31.07) (À31.71) (À31.17) (À31.24) (À29.79) (À32.16)
Count ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Industry ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Year ?xed effects Yes Yes Yes Yes Yes Yes Yes Yes
Observations 185,109 185,109 185,109 185,109 185,109 185,109 185,109 185,109
Pseudo R
2
0.101 0.101 0.099 0.099 0.099 0.099 0.099 0.099
This table presents logistic regression coef?cients (z-values in parentheses) for the impact of ownership characteristics and family relationships on the
choice of Big 4 auditors. The sample is described in Table 1 and variables are de?ned in the Appendix except for the test variables in Model 2 which are
measured as indicator variables de?ned as 1 if the test variable is greater than or equal to mean and 0 otherwise. County, industry, and year ?xed effects are
included in the models but omitted from the table in order to conserve space. Robust standard errors clustered at the ?rm level are employed.
Ã
Statistical signi?cance at the 0.10 level (two-sided).
**
Statistical signi?cance at the 0.05 level (two-sided).
***
Statistical signi?cance at the 0.01 level (two-sided).
514 O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517
and participants at the 5th European Auditing Research
Network (EARNet) Symposium (Valencia), the European
Accounting Association Annual Meeting (Istanbul), the
16th Annual International Symposium on Audit Research
(Singapore), the American Accounting Association Annual
Meeting (San Francisco), the 3rd International EIASM
Workshop on Audit Quality (Bellagio), the 5th CCGR Work-
shop (Oslo), the 1st Nordic Conference on Financial
Accounting (Copenhagen), and the 6th European Auditing
Research Network (EARNet) Symposium (Bergen). We also
greatly appreciate a grant from the Centre for Corporate
Governance Research (BI Norwegian Business School),
and the support from its director, Professor Øyvind Bøhren.
Hope further acknowledges the support of the Deloitte
Professorship.
Appendix A. Appendix: De?nitions of variables
Test variables
OWNER_CONC Her?ndahl ownership
concentration ratio in year t based
on ultimate ownership
OWNER_SECOND Fraction of shares held by the
second largest owner in year t using
ultimate ownership
OWNER_CEO Fraction of shares held by the CEO in
year t, based on ultimate ownership
FAM_CEO/
OWNER
1 if the CEO is related to the largest
ultimate family owners through
blood or marriage
FAM_BOARD/
OWNER
Family in?uence on the board in
year t computed as the number of
board members from the largest
owning family divided by the total
number of board members
FAM_BOARD/
CEO
Family relationships between the
CEO and board members, de?ned as
the number of family members of
the CEO on the board divided by the
total number of board members
Other variables
ACQUISITIONS 1 if the ?rm has increased long-
term investments in other
companies from t À 1 to t, 0
otherwise
AGE Firm’s age in year t measured as
number of years since incorporation
BIG4 1 if the auditing ?rm is one of the
Big 4 auditing ?rms or their
forerunners in year t, 0 otherwise
CHAUDITOR 1 if the ?rm changed auditor in year
t, 0 otherwise
CHCEO 1 if the ?rm appointed a new CEO in
year t, 0 otherwise
CHGAAP 1 if the ?rm has changed from
simpli?ed GAAP to regular GAAP or
from regular GAAP to simpli?ed
GAAP during year t
CHLEV Change in interest bearing debt
from t À 1 to t = LEV
t
À LEV
tÀ1
CITY 1 if ?rm is located in a city, 0
otherwise
COUNTY
c
1 if ?rm is located in county c, 0
otherwise (c = 0, 1, . . ., 19)
CURRATIO Current ratio at the end of year t
DECPIC 1 if the ?rm decreased share capital
from t À 1 to t, 0 otherwise
EMPLOY Number of employees in year t is
estimated as sum wages in year t
divided by the average salary for full
time male and female workers in
year t if sum wages >0, 0 otherwise.
Data on average salary is taken from
Statistics Norway
EXANTEFIN Operating cash ?ow less net
investments in tangible and
intangible ?xed assets scaled by
current assets in year t. As in
Knechel et al. (2008), we use a
square root transformation of the
ratio multiplied by À1 if the
unadjusted ratio was negative.
Operating cash ?ow is estimated as
operating income + depreciation +
impairment À change in accounts
receivables À change in inventory +
change in accounts payable À
change in taxes payable. Net
investments in intangible and
tangible ?xed assets is estimated as
change in intangible and ?xed
assets during the year plus
depreciation and impairment since
cash ?ow statements are not
available
FEE Audit fee in NOK 1000, computed as
the total fee paid to the auditor for
auditing services in year t.
FOREIGN Percentage of foreign subsidiaries in
year t = the number of foreign
subsidiaries in year t
Ã
100/total
number of subsidiaries in year t
FYE 1 for ?scal years ending other than
December 31 in year t, 0 otherwise
GAAP 1 if ?rm uses regular GAAP in year t
and 0 otherwise
GROWTH Change in sales in year
t = (SALES
t
À SALES
tÀ1
)/SALES
tÀ1
INCPIC 1 if the ?rm increased share capital
from t À 1 to t, 0 otherwise
IND
i
1 if ?rm belongs to industry i, 0
otherwise. Two-digit industry codes
are used to classify ?rms into
industries. Firms in industries with
30 or less observations in a given
year are reclassi?ed to industry 99
(continued on next page)
O.-K. Hope et al. / Accounting, Organizations and Society 37 (2012) 500–517 515
INDUSTRIAL 1 if industrial investors own 20% or
more of the ?rm based on ultimate
ownership, 0 otherwise
INSTITUTIONAL 1 if institutional investors own 20%
or more of the ?rm based on
ultimate ownership, 0 otherwise
INTANG Intangible assets at the end of year t
scaled by SALES
t
INTERNATIONAL 1 if international investors own 20%
or more of the ?rm based on
ultimate ownership, 0 otherwise
INVESTMENTS (Long and short terminvestments in
securities + bank deposits + cash)
t
scaled by SALES
t
INVREC Inventory and accounts receivable
at the end of year t scaled by SALES
t
LASTYRAUDIT 1 in year t if the ?rm changed
auditor in year t + 1, 0 otherwise
LEV Long- and short-term interest
bearing debt/total assets, both at
the end of year t. Short-term
interest bearing debt = total short
term debt À accounts
payable À dividends À taxes
payable À VAT and social service
taxes À other short term debt
LNEMPLOY Natural logarithm of 1+ number of
employees = ln(1 + EMPLOY
t
)
LNFEE Natural logarithm of total fees paid
to the auditor for auditing services
in year t = ln(FEE
t
)
LNSALES Natural logarithm of total revenue
from operations = ln(SALES
t
)
LNSUB Natural logarithm of 1 + number of
subsidiaries in year t
LNTA Natural logarithm of total
assets = ln(TA
t
)
LOSS 1 if net income after taxes before
extraordinary item and taxes on
extraordinary item