audit partner performance: a network perspective
TRANSCRIPT
Audit Partner Performance: A Network Perspective
Joanne Horton University of Exeter Business School
İrem Tuna* London Business School
Anthony Wood University of Exeter Business School
First draft: 9 June 2014 This draft: 9 October 2014
*İrem Tuna gratefully acknowledges the financial support of the European Research Council (Grant ERC-2010-263525). We thank Emmanuel De George, Stuart Turley, and the seminar participants at the 5th EIASM Workshop on Audit Quality and at Queen Mary University of London for helpful comments.
1
Audit Partner Performance: A Network Perspective
ABSTRACT
We provide a partner level analysis on the association between the social capital of an audit partner and their audit quality and fees. We use social capital theory and techniques developed in social network analysis to measure the audit partner’s level of connectedness and investigate whether these connections provide information advantages and enhance their social influence. We argue that both of these potential network benefits support the audit partner attributes, namely personal capabilities and level of independence, necessary to provide high quality audits. Using a sample of French listed firms, we construct a network consisting of 4,159 board directors and 734 audit partners, mapping the connections between audit partners and directors, between directors and between audit partners. We find that better-connected audit partners provide higher quality audits and earn higher audit fees. The level of an audit partner’s social capital decreases if they have longer tenure and greater economic bonding with their clients.
2
INTRODUCTION
Auditors play an essential role in the effective functioning of corporate governance and
are among the most important gatekeepers in the capital markets (Coffee, 2005). The
audit process provides the market with ‘independent’ verification of the financial
statements, signaling that the management information is reliable, and thereby
reducing the cost of the information exchange between managers and investors
(Dopuch and Simunic, 1980, 1982). However, it is the management that controls the
process of hiring and firing auditors and there are quasi-rents associated with such
auditing contracts (DeAngelo 1981a, 1981b). Consequently auditors have incentives to
yield to pressure from management, implying that the quality of the information
contained in audited financial statement depends upon the ability of the audit partner
to resist such pressures (Becker et al. 1998). Audit quality is therefore determined by
the probability that the auditor will uncover a breach in the client’s accounting system
and will report the breach (DeAngelo 1981b). The ability to uncover the breach will
depend on the audit partners’ capabilities while the reporting of the breach will depend
on the auditor’s level of independence from the client (DeAngelo 1981a). If the auditor
does not remain independent, the auditor will be less likely to report the irregularities
and audit quality will be impaired. Prior studies have identified various factors that
explain differences in the level of audit quality provided by auditors (Larcker and
Richardson, 2004, DeFond and Zhang 2014), and the effect of their implicit or explicit
incentives (Ruddock, Taylor, and Taylor 2006).
In this paper we investigate whether an audit partner’s network position is a
factor in determining audit quality and audit fees. Why would we expect the network
position of an audit partner to affect their audit quality and audit fees? Social capital
theory would suggest that centrally located individuals within a network have greater
access to a broader source of information, resources, people and instrumentalities, such
that their network position represents a structural source of influence or power. These
network benefits are ideally suited to enhance and reinforce partner attributes
necessary to produce high quality audits – auditor capabilities and independence. This
in turn should be rewarded with fee premiums based on an assessment of the quality of
the partner’s audits. For instance a better-connected audit partner will have greater
opportunities to be privy to more and better quality information, which ultimately
provides them with opportunities to enhance their personal capabilities and thereby
3
increase the likelihood of uncovering a breach in the clients’ accounting systems. At the
basic level the more connections (or better connections) an audit partner has, the more
peers with whom to consult, the wider the range of expertise to which she will have
access. For example, a better-connected audit partner will have greater opportunities to
achieve a comprehensive ‘business scan’ of the latest auditing and business practices
and of the overall business environment in which their client is situated (Useem, 1984).
Moreover, a better-connected audit partner will also have greater influence (or
power) over her connections potentially enhancing her level of independence – the
second partner attribute necessary for high quality audit. Many sociologists would
agree that an individual’s influence over others is a fundamental property of social
structures. Social influence in this context refers to the capacity of an audit partner to
influence others and the ability to resist others’ influences. Both of these elements
should increase the audit partner’s level of independence. For instance, the more
influence an audit partner gains from their connections the more likely they are to be
independent and the less likely they are to yield to management pressure.
Although economic agents may gain information advantages as well as enhanced
influence though their network connections and network position (Coleman 1988; Burt
1992; Bonacich 1972; Cohen et al. 2011; Horton and Serafeim 2012), it is not clear that
such advantages would exist among audit partners leading to better performance. First,
studies have shown that audit partners who audit more firms and cover more industries
provide relatively lower audit quality (Goodwin, 2011; Gul et al. 2013; Ittonen et al.,
2010). This result is counter-intuitive from a network perspective, because if
connections do provide information advantages and enhanced influence to an audit
partner then audit partners who audit relatively more clients over more industries are
more likely to be better-connected. Second, studies have also found that ties between
audit partners and directors do not always results in higher audit quality – indeed the
opposite is true – alumni ties for instance have been associated with poor quality audit
(Lennox, 2005; Menon and Williams, 2004). Third, information flows from one audit
partner to another competing audit partner only if complementarities exist in the
information structure (Stein 2008). Whether such complementaries, which allow
information exchange through networks, exist is an empirical question.
To investigate the possible impact of an audit partner’s network on audit quality
and fees, we construct a network of French listed firms, their directors (executive and
4
non-executive), and their external audit partners. There are two key advantages of
focusing on the French audit market. First, France has a long history of requiring the
engagement partner’s signature on the audit report thereby enabling us to identify each
individual audit partner over a longer period of time compared to the UK who only
recently required such a signature. Second, France has one of the lowest levels of audit
firm concentration due to its joint audit requirement – i.e. that two independent audit
firms must audit every listed firm1. Since each audit partner has the potential to be
connected to a greater number of firms and to other audit partners outside of their
immediate audit firm, the French audit market provides a richer and denser network
over what could be observed in alternate audit markets.
The network we construct comprises of 4,159 board directors from 326 client
firms and 734 audit partners from 192 audit firms. We measure the quality of the audit
partner’s network position and use this measure to proxy for the likelihood that the
audit partner is able to obtain valuable information and generate greater social
influence. An audit partner’s position within the network will determine the quality,
type, timing and amount of the information she receives, as well as the amount of
influence she could potentially exert over others, or inversely prevent her from being
influenced by others, within the network. The proxy does not rely on the presumption
that an audit partner will have explicit one-to-one conversations with all her
connections. The proxy merely captures the increased likelihood that an audit partner
with a relatively better network position will have more opportunities to be privy to
more or better information and to be able to create higher level of social influence. 2
We find that the more central an audit partner is within the network the better
the quality of the audit. We find that both abnormal working capital accruals, and the
likelihood of firms reporting a small earnings increase, are negatively related to quality
of the audit partner’s network position. An alternative explanation for this
performance-enhancing role of audit partner’s network position could simply be audit
partner experience. It is plausible that the better-networked audit partner is also the
most senior and more experienced partner. To rule out the possibility that our network
1Mandatory joint audit has been required in France since 1966 (Huber, 2011). 2 We acknowledge that this network is incomplete. Both directors and audit partners have many other possible avenues through which they can obtain information or influence, for example golf club membership, political affiliations, professional bodies etc. Nevertheless, these social or ‘grey ties’ will add noise to our network estimates potentially biasing downwards the effects we can identify.
5
measure merely proxies for experience, our models explicitly control for a variety of
dimensions of audit partner experience, both time variant and time invariant. For
example, in addition to controlling for the generic ability of the audit partner using
partner fixed effects we also include the partner’s industry expertise, tenure with the
client, the number of clients the partner audits during the year and the audit firm the
partner works for. These controls are important since the concept of social capital and
human capital have clear parallels, although they are not identical (Field, 2003)3. Under
all specifications the audit partner’s network measure continues to be negative and
significantly associated with all the audit quality measures.
Our findings also hold when we limit our sample to those clients whose audit
team comprised of one Big4 audit firm and one non-Big4 audit firm, thereby reducing
the possibility of self-selection (Ferguson et al 2003; Francis et al 2005) and other
potential confounding effects (Chan and Wu, 2011). The results continue to hold.
Since an audit partner’s position in the network can explain variation in audit
quality, we also investigate whether it has implications for the level of audit fees
charged to the client. We find evidence of a significantly positive association between
the audit fees and the network position of the audit partner. This result is consistent
with the findings of Zerni (2012) and Taylor (2011), that different audit partners – even
from the same audit firm – can earn different levels of audit fees.
Finally we provide evidence on the determinants of our audit partner’s network
position. We find that an audit partner’s network position improves if the partner
works for an audit firm that is an industry specialist, is employed by a Big4 audit firm,
audits large clients and numerous clients, and declines if the audit partner has a long
tenure with a client, is associated with high levels of non-audit fees and has higher
levels of economic dependence on the client.
We make numerous contributions to the literature. First, our paper contributes
to the growing literature on social network effects in economics (Hong, Kubik and Stein
2005; Hochberg, Ljungqvist and Lu 2007; Cohen et al. 2011; Cohen et al. 2008; Stein
2008). Given the importance of the auditors within the capital markets, understanding
the factors that affect their performance and implicit incentives is important. Our
analysis provides a deeper understanding of the possible drivers of cross-sectional 3 As Gould (2002) notes; “even if the network position is initially caused by exogenous personal qualities, central actors likely receive a disproportionate share of new nominations, not because of personal qualities, but by virtue of having already received many nominations”.
6
variation in performance of audit partners, and to the best of our knowledge is the first
to show that audit partners’ position in their professional network is a significant
determinant of audit quality and audit fees.
Second, we answer the call for more research at audit partner level (Reynolds
and Francis 2000; DeFond and Francis 2005; Chen et al. 2009; Chen et al. 2010) in order
to gain a deeper understanding of audit partner behavior and its implications for audit
quality, providing further evidence that audit quality is not uniform across audit
partners (O’Keefe and Westort, 1992; Carcello et al. 1996; Carcello et al 2000; Carey and
Simnett 2006; Gul et al. 2013). Third, we also add to the audit fee literature by
providing further evidence that audit partners earn individual audit fees that are not
fully explained by the audit firm at which they work (Taylor 2011; Burrows and Black
1998). Fourth, we provide evidence regarding the determinants of an audit partner’s
network position and contribute to the audit independence literature. Assuming our
proxy captures, inter alia, the level of audit partner independence, we find, consistent
with the prior literature, that certain partner attributes, such as long tenure and high
levels of economic dependency reduce the advantageous network position of the audit
partner and thereby reduce the opportunities to be more independent.
Last but not least, our findings have potential implications for audit partners,
audit firms, as well as regulators. Our evidence suggests that enhancing one’s network
position should be an important strategic consideration for an incumbent audit partner
and their audit firm. Moreover, the policymakers should also be interested in our
findings as they highlight that an environment in which audit partners have
opportunities to enhance their network position could contribute to increased audit
quality. We are however very cautious to imply that these opportunities only exist in a
joint-audit setting, such as the one investigated here. Our findings do however suggest
that policymakers should examine audit quality and audit market structure
enhancement together as a complex whole.
The paper is organized as follows. In Section I, we review the related literature
and form the hypotheses of our study. Section II describes the sample. Section III
constructs the necessary measures and presents the research design. Section IV
discusses the results on the association between an audit partner’s network position,
therefore social influence, and audit quality and audit fees. Section V presents the
7
results on the determinants of the audit partner’s social influence. We conclude in
section VI.
I. Literature Review and Hypothesis Development
Audit Quality
DeAngelo (1981) defines audit quality as ‘the market assessed joint probability that a
given auditor will both discover a breach in a client’s accounting system, and report the
breach’. This definition is often interpreted to break down audit quality into two
components: (1) the likelihood that an auditor discovers existing misstatements, and
(2) appropriately acts on the discovery. The first component is related to an auditor’s
capabilities and level of effort while the latter relates to an auditor’s objectivity,
professional skepticism, and independence.4 These two components also suggest that
different aspects of the audit can influence overall audit quality.
Prior literature confirms the there is a direct link between the quality of the
audit and individual audit partner capabilities such as knowledge and expertise.
Domain-specific knowledge (e.g. knowledge accumulated through client, task and
industry experience) is associated with higher quality auditor judgment (Knechel et al.
2013; Bonner and Lewis 1990). A number of experimental studies have found industry-
experienced auditors are better able to detect errors among clients within their
industry specialization than those outside their specialization (Owhoso et al. 2002;
Bedard and Biggs 1991; Wright and Wright 1997). Hamilton and Wright (1982) find a
negative association between years of experience and consensus on internal control
evaluation. While Ashton (1991) finds that months of general audit experience are not
related to how accurately auditors judge the frequency of specific accounting errors.
Moreover, differences in audit partner behaviors and attitudes (Ponemon and Gabhart
1990; Trompeter 1994; Tan 1999; Emby et al 2002; Ayers and Kaplan 2003) have also
been found to influence the behavior and attitude of their audit staff (DeZoort and Lord
1994; Tan et al 1997; Kaplan and Lord 2001; Lord and DeZoort 2011; Wilks 2002).
Recent empirical studies also document a positive association between audit
quality and individual audit partner specialism (Chin and Chi, 2009; Chi and Chin 2011; 4 Francis (2011) notes that audit quality is a complex concept and cannot be reduced to a simple definition. He argues that there are gradations of audit quality across a continuum from low- to high- quality audits and that quality is affected by a number of elements. However, even under this more rigorous approach to defining and understanding audit quality, audit partner capabilities and the need for independence are still identified as important elements.
8
Kallunki et al. 2009; Duh et al. 2009; Ittonen et al., 2010). Extending previous studies
Ittonen et al. (2010) find that clients of audit partners with both higher public
specialization and longer auditing experience report lower discretionary accruals.
While Chi and Chin (2011) find differential discretionary accruals is driven by a
combination of firm and partner industry expertise, although the differential likelihood
of a modified audit opinion is primarily attributable to the signing audit partner’s
industry specialism.
The link between independence and audit quality has also been empirically
tested but given audit partner independence ‘in fact’ is generally unobservable, studies
tend to investigate the relationship between audit quality and certain audit partner
features that could potentially reduce an audit partner’s level of independence5 (either
in fact or in appearance). For instance alumni affiliations (Menon and Williams 2004;
Lennox 2005; Lennox and Park 2007; JeanJean et al. 2013), long tenure with client firm
(Carey and Simnett 2006; Bamber and Iyer 2007; Chen et al 2008; Chi et al. 2009; Gul et
al. 2013), and economic dependency/bonding (Chen et al. 2010; Chi et al. 2012) in some
instances have been found to reduce audit quality, although there is no evidence to
suggest client importance impacts audit partner independence.
Social Capital Theory
Given audit partners need to have both the personal capabilities and be
independent, the auditing industry is an ideal testing ground to investigate whether an
audit partner’s network position enhances these two antecedents by increasing the
level of information available to them and by providing them with the opportunity to
increase their social influence. Both of these network advantages potentially could
increase the quality of their audits and determine their audit fee premiums. Social
capital is based on the notion that the actions of individuals are facilitated by their
membership in social networks. Therefore social capital can explain the differential
success of individuals and firms in their competitive environments (Alder and Kwon,
2002). As Burt (2000) notes social capital is a metaphor for advantage: people who do
better in business are better-connected in one way or another. For example, research
shows that social ties help individuals (firms) gain access to information about job
opportunities (potential clients) (Burt, 1992; Meyerson, 1994; Fernandez et al., 2000), 5 Bartlett (1993) notes that independence is a matter of degree, rather than an absolute concept.
9
new innovations (Burt, 1987) and provide a finer grained information set (Uzzi, 1997).
Prior research documents that agents at the core of the social network have more
valuable information (Freeman 1980; Cohen et al. 2011). Many network studies
repeatedly demonstrate a link between social capital and performance in a variety of
settings. For example, Mizruchi and Stearns (2001) show that loan officers in a large
commercial bank with more social capital are more likely to close a successful deal.
Hochberg, Ljungqvist and Lu (2007) find that better-network venture capitalists have
better performance. Horton et al. (2009) find director’s connectedness is positively
associated with compensation and with their firm’s future performance, while Horton
and Serafeim (2012) find better-connected analysts make more accurate, timely and
bold forecasts.
In addition to providing informational advantages, being better-connected has
also been found to be associated with an individual’s level of power/influence. Power is
defined here as the ability to influence others or not be influenced by others.
Power/influence is inherently a structural phenomenon where one actor’s influence
over another needs to be considered within a wider network of relationships (Pfeffer,
1981; McClurg and Young 2011), and that it is distinct from the influence gained from
individual attributes (Brass and Krackhardt 2012). As Brass and Krackhardt (2012)
state “the structure of social networks strongly affects the extent to which personal
attributes, cognition and behavior result in power and influence within an
organization”. Network analysis emphasizes that power and influence are inherently
relational, and individual does not have power in abstract; they have power because
they can dominate others (Hammerman and Riddle, 2005).
Prior research demonstrates that being better-connected not only increases
access to different resources but also increases legitimacy and social influence and
power (Brass, 1984, 1985, 1992; Mizruchi and Potts 1998; Brass and Burkhardt 1992,
1993; Burkhardt and Brass 1990; Krackardt, 1990; Friedkin 1991; Sparrowe and Liden
1997; Mehra, Kilduff and Brass 2001; Ibarra and Andrews 1993; Setton and Mossholder
2002; Reagans and Zuckermann, 2001; Grosser, Kidwell-Lopez and Labianca 2010).
Consistent with the power-dependency theory, networks create tangible imbalances by
regulating the availability of alternative exchange partners (Cook et al 1983; Emerson
1972; Molm 1990). As Galaskiewicz (1979) noted, it is not the level of resources per se
10
that determines influence but rather the ‘set of resources that actors [could] mobilize
through their existing set of social relationships’ (p.151).
The study of inter-corporate power relations, via interlocking directorships, was
one of the earliest streams of research to employ social network techniques (Bearden et
al 1975). Recent studies continue to find that the better-connected the board member
the greater their influence on the decision-making process of the board (Stevenson and
Radin 2008). Specifically Stevenson and Radin (2008) find that social capital of board
members in the form of ties to others is a much stronger factor in gaining influence on
the board as compared to the human capital attributes of board members such as
management experience or committee membership. El-Khatib et al. (2012) find that a
CEOs personal network provides them with enough power to pursue any acquisition,
regardless of the impact on shareholder wealth. Renneboog and Zhao (2011) examine
the relation between directors’ networks, CEO compensation and pay-for-performance
sensitivities. They find that a CEO’ s social influence and power is positively associated
to their compensation and negatively related to their pay-for-performance, suggesting
potential rent extraction by these powerful CEOs.
Connections to others may also create the perception of social influence and
power. For instance, from a prism perspective, these central actors are more likely to be
viewed by others as more powerful due to their connections to other prominent
individuals (Kilduff and Krackhardt 1994)6. Whether or not the perception is accurate,
these actors are more likely to receive deferential treatment based on that perception.
For example, as reported by Kilduff and Krackhardt (1994) when approached for a loan
the wealthy Baron de Rothschild replied, ‘I won’t give you a loan myself but I will walk
arm-in-arm with you across the floor of the stock exchange and you will soon have
willing lenders to spare’ (Cialdini 1989, p.45).
Based on the social network research noted above, if audit partners are able to
obtain an informational advantage as well as some level of social influence through their
network position, this could enhance the quality of their audits. Better-connected audit
partners will have access to a wider range of expertise, contacts, and pools of social
capital, thereby providing them with greater opportunities to achieve a comprehensive
‘business scan’ of the latest auditing and business practices and of the overall business
6 Deephouse and Suchman (2008) describe this form of power as an individual having greater visibility and legitimacy.
11
environment in which their client is situated, all of which would provide the audit
partner with greater capabilities to perform higher quality audits. Moreover, given the
ability to influence others or not be influenced by others is a major condition for
independence, an audit partner with relatively more (and more powerful) connections
is more likely to resist management pressure. In a recent paper, Francis and Yu (2009)
elude to the network benefits of audit partners when they suggest that their findings -
larger offices of Big 4 auditors have higher audit quality - are consistent with the
argument that auditors working in large offices have more peers with whom to consult,
which improves audit quality.
One could also equate the idea of being better-connected to the concept of
‘reputational capital’ introduced by DeAngelo (1981a). DeAngelo (1981a) argues that
audit partners with a higher ‘reputational capital’ have greater incentives to be
independent since they earn client-specific rents, which they stand to lose in the event
of an audit failure. One could use this same argument for audit partners who have
greater levels of social capital since in the event of an audit failure they would
potentially risk the loss of these connections and any benefits (potential or actual)
associated with them, which in turn also means that these audit partners have greater
incentives to be independent. Therefore our first hypothesis is:
Hypothesis 1: An audit partner with more social capital will provide a higher
quality audit.
Past research has also examined audit partner fees (Zerni, 2012; Taylor, 2011). Taylor
(2011) finds that individual audit partners earn individual audit fee premiums (or
discounts) that are not explainable by the audit firms of which they are members. Zerni
(2012) finds that engagement partners who are both industry and public firm
specialists earn higher fees. Therefore if audit partners gain information and influence
benefits from their network position and consequently are able to provide better
quality audits, then the audit services market should reward individual audit partners
with fee premiums based on an assessment of the quality of their audits (Taylor 2011).
Similar to the demand for partners with industry specialism, if demand exists for audit
partners with higher social capital, then that demand gives the audit partner greater
12
‘power’ in pricing – relative to audit partners with less social capital. Therefore our
second hypothesis is:
Hypothesis 2: An audit partner with more social capital will earn higher audit fees.
II. Sample
Network sample and construction
To construct the network, we obtained an initial sample from the Boardex
database for the years 2004 to 2010. In total we obtained data for 326 unique listed
French firms (1,859 firm-year observations), with a total 19,957 directorships and
4,159 unique directors (See Table 1, Panel A). Following a review of the firms yearly
audit reports we were able to identify 734 unique audit partners working for 192
unique audit firms. Of the firms within the Boardex database we were unable to identify
the audit partners for 61 of the firm-year observations.7
For each year we construct the network by connecting audit partners to audit
partners, audit partners to directors, and directors to directors. A director is directly
linked to another director if they sit on the same board, and is directly linked to an audit
partner if the audit partner audits their company. An audit partner is also linked to
another audit partner if they work for the same audit firm or if they audit the same
client during the year (see Appendix). We assume that it is more probable that audit
partners and directors interact with each other under these conditions. For example,
audit partner A is more likely to communicate with audit partner B if they are part of
the same audit team compared to audit partner C that neither audits the same clients
nor works in the same auditing firm as audit partner A. Similarly, if audit partner A
audits company E then she is more likely to be associated with director D who sits on
the board of company E than director F who sits on the board of company G, which
audit partner A does not audit. One could argue that audit partners should only be
connected to directors if the director is a member of the audit committee. However, one
would have to assume a) that the audit partner would never have the opportunity to
communicate with a director who is not on the audit committee, and b) that there is no
potential benefit to an audit partner of being seen to be associated with a particular
7 Despite not being able to identify the audit partners for some firm-year observations, we continue to include the directors in the network, so the network is as complete as possible.
13
director. We believe both of these assumptions are unlikely to hold.8 First, certainly in
France there is a strong ‘collegiality’ principle in the sense that board sub-committees
have only a consultative role, they are not a substitute for board responsibilities –
rendering the board of directors collectively responsible for financial reporting quality –
thereby providing an incentive for directors to communicate with the audit partner(s)
(Bouton Report, 2002, p.6). Second, audit partners are required to attend board
meetings when the board approves the accounts; otherwise audit partners attendance
is optional. Third, Kilduff and Krackhardt (1994) highlight that even being perceived to
be connected to an individual benefits a person’s influence, reputation, status, etc.,
which is consistent with the old proverb ‘you are known by the company you keep’.
Analysis Sample
To test Hypothesis 1 and 2, we collected all the relevant client-specific
information from Datastream (except the number of subsidiaries which we hand
collected). Audit partner data such as tenure, audit office, audit fees and non-audit fees
was also hand collected directly from the annual reports. 9
In practice, joint-auditorships in France are often characterized by a lead auditor
who imposes its quality standards and a ‘second’ auditor of a lower caliber (Piot and
Janin 2007; Andrè et al. 2011). Following an investigation of the audit fees paid to each
independent auditor, and consistent with Le Maux (2004), we find that the distribution
of fees (both audit and non-audit) between the joint-auditors – and probably the
budgeted audit hours – is far from being equal, with the lead auditor charging on
average nearly 60% more audit fees than the non-lead auditor. We therefore contend
that the quality of the external audit can be proxied by the characteristics of the audit
partners from the dominant (lead) audit firm. Therefore, although both lead and non-
lead audit partners are included when we construct the network to capture the network
position of audit partners, we exclude the non-lead auditor partners from the analysis
(Piot and Janin 2007). This results in a total 2,096 observations remaining for analysis.
8 To note, if in constructing our network we have not identified the ‘right’ connections then this will only bias against us finding any association between our network measure and audit quality and fees 9 The disclosure of the audit fees for French listed companies became compulsory in 2003. This disclosure requires companies to distinguish (a) between fees paid for statutory audits and those paid for other services (non-audit fees), and (b) between the specific fees paid to each of the two independent auditors.
14
The observations from the financial industry are also excluded (590) as their total asset
base and financial information is not comparable to those of the other companies.
To obtain the data for the audit quality analysis, we delete 67 observations with
missing data. We also exclude the first year the firms introduced IFRS due to
comparability issues (given the use of lagged accounting data) resulting in the exclusion
of a further 188 observations. These procedures result in a total of 1,251 observations
for the audit quality analysis. Table 1 Panel A summarizes the sample selection process.
Shown in Table 1 Panel C is the available data for the fee analysis. Again we were
unable to identify data for some of the control variables but we did not need to exclude
the year IFRS was implemented so the final audit fee sample comprises of 1,585
observations.
[Insert Table 1]
III. Measurement of Variables and Research Design
Network measure
‘It does not only matter that actors have a lot of contacts, but they have highly
connected, prominent contacts.’ (Hannerman & Riddle, 2005)
On the basis of its theoretical relevance we use the eigenvector centrality
measure (EVC) to proxy for an audit partner’s social capital. Friedkin (1991) describes
this index as ‘total effects centrality’ in the context of network influence models and for
many researchers centrality is EVC (Faust, 1997). The EVC measure has been widely
used in a variety of settings to capture the opportunities available to individuals in
obtaining information and for creating social influence and power (Borgatti, 2006;
Mizruchi and Bunting, 1981; Frenandez and McAdam 1988; Friedkin, 1991; Scott and
Davis, 2007), from interlocking corporate boards of directors (Mintz and Scwartz,
1981a, b; Mizruchi and Bunting, 1981; Mizruchi 1982; Roy 1983; Rosenthal et al., 1985;
Liu, 2009; Renneboog and Zhao, 2011; El-Khatib et al. 2012; Foge et al. 2013), to
venture capitalists (Hochberg et al., 2007), and inter-bank networks (Bech et al. 2010;
Cohen-Cole et al. 2010; Akram and Christophersen 2010).
Eigenvector centrality is defined as the principle eigenvector of the adjacency
matrix defining the network (See Appendix). 10 It is a recursive measure of degree (the
10 There are several variations of this simple eigenvector centrality, for example, PageRank the network centrality originally used in Google search engines.
15
number of direct connections one has), whereby the actor’s centrality is defined as the
sum of her ties to others weighted by their respective centralities. So while degree
centrality simply counts the number of direct connections an individual has, EVC takes
into account their ‘quality’. Thus connections to neighbors that are in turn well
connected themselves are rated higher than connections to neighbors that are weakly
connected i.e. this measure does not assume all ties are equal. For example, suppose
audit partner A and audit partner B each has the same number of direct connections.
However, audit partner A’s connections happen to be fairly isolated individuals (e.g.
their client firms have very few board-interlocks, and their joint audit partners are from
a small one person audit firm, who does not audit many clients). In contrast audit
partner B connections each have lots of connections, who also have lots of connections,
and so on (e.g. their client firms have many board-interlocks, and their joint audit
partners are from large audit firms and audit many clients). Under this scenario despite
having the same number of direct connections audit partner B is clearly better-
connected e.g. she has the ‘right connections’. Audit partner B’s ties to individuals with
many connections creates greater opportunities for indirect flows of information and
social influence whether this influence is actual or perceived (Kilduff and Krachardt,
1994).
As indicated in the prior literature this measure has several advantages over
other centrality measures such as closeness and betweenness (Bonacich 2007) when
trying to capture both informational benefits and the opportunity to create social
influence and power.11 Unlike these other centrality measures, which weights every
connection equally, EVC is based on the idea that a connection to other powerful actors
contributes more to a given actor’s power and informational opportunities than equal
connections to less powerful actors. Moreover the EVC measure assumes that
information, influence etc. is able to move via unrestricted walks rather than being
constrained by triads, paths or geodesics – as is the case with closeness for example.
The measure does not in any way assume that things flow by transferring or by
replicating to one’s connections at a time. Rather it is consistent with a mechanism in
11 Although betweenness centrality is argued to be associated with power and influence, this is more related to the control of information flowing through the connections rather than the power and influence that these connections generate.
16
which each actor can affect all its neighbors simultaneously (Borgatti, 2005). Hence the
EVC measure is ideally suited for influence type processes.12
Empirical Models: Audit Quality
Higher-quality audits should mitigate errors or earnings management in financial
statements, which is often introduced through the use or abuse of the accruals process
(Richardson et al. 2006; Ettredge et al. 2010). Therefore to test Hypothesis 1 we follow
prior literature and capture audit quality using two proxies a) abnormal working capital
accruals (Ashbaugh et al. 2003; Carey and Simnett 2006; Frankel et al. 2002; Carcello
and Li 2013) and, b) the likelihood of firms reporting a small earnings increase (Frankel
et al. 2002; Ashbaugh et al. 2003; Carey and Simnett 2006; Carcello and Li 2013). Ideally
we would have liked to use other proxies such as going concern opinions and earnings
restatements (Carey and Simnett 2006; Geiger and Raghunanden 2002; Myers et al.
2003) but the number of such cases is very small for our sample.13
Abnormal working capital accruals
We measure abnormal working capital accruals (AWCA) utilizing the DeFond and Park
(2001) model, which has also been used in subsequent audit partner research (Carey
and Simnett 2006). AWCA is the difference between reported working capital and an
expected level of working capital needed to support current sales level, where a historic
relation of working capital to sales captures expected working capital. We focus on
working capital accruals for two reasons. First, DeFond and Park (2001) find this
measure to be a more powerful test in comparison to a test using total accruals. Second,
it avoids the estimation problems, associated with the more sophisticated models
(Jones 1991) especially for studies with European data (Francis et al. 2009; Cameran et
al. 2012; Maijoor and Vanstraelen 2006; Andrè et al. 2011). 14
12 Costenbader and Valente (2003) evaluated the stability of centrality measures when networks are sampled in the face of inaccurate or incomplete network data. It turns out that the most robust centrality measure under these circumstances is the EVC measure. Consequently, not being able to identify a number of actors, as is the case for our sample, should not cause significant problems. 13 An investigation of audit reports and financial statements revealed that less than 0.5% of our sample firms had either a going concern opinion or a restatement during 2004 - 2010. 14 A cross-sectional Jones (1991) model is not used to estimate abnormal accruals because it requires more data than the approach in DeFond and Park (2001) and would cause a large reduction in the number of observations. In our French sample, if two-digit SIC codes are used to define industry categories, as is typically the case, then only a small fraction of our industries would have the rule-of-thumb of 10 minimum observations required to estimate the Jones model. In addition, a recent study suggests that the Jones model does not perform well on non-U.S. data, and this may in part be caused by small industry samples in international settings (Peek et al. 2013).
17
If an audit partner’s social capital – their eigenvector centrality score (EVC) - is
associated with the quality of their audit, then we expect lower absolute AWCA for audit
partners with higher EVC scores. We therefore estimate the following model:
|𝐴𝑊𝐶𝐴| = ∝ +𝛽1𝐸𝑉𝐶𝑝𝑡 + �𝛾𝑖 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒+ �𝜌𝑖 𝐴𝑢𝑑𝑖𝑡 𝐹𝑖𝑟𝑚 𝐹.𝐸.
+�𝜑𝑖𝐴𝑢𝑑𝑖𝑡 𝑂𝑓𝑓𝑖𝑐𝑒 𝐹.𝐸. + �𝜇𝑖 𝐴𝑢𝑑𝑖𝑡 𝑃𝑎𝑟𝑡𝑛𝑒𝑟 𝐹.𝐸. +
�𝛿𝑖 𝑌𝑒𝑎𝑟 𝐹.𝐸. +�𝛾𝑖 𝐶𝑙𝑖𝑒𝑛𝑡 𝐹𝑖𝑟𝑚 𝐹.𝐸. + 𝜖
(1)
where AWCA is the DeFond and Park (2001) measure of abnormal working capital
accruals: AWCAt = WCt –[(WCt-1/St-1) * St], with WC = non-cash working capital in the
current year which calculated as (current assets – cash and short-term investments) –
(current liabilities – short-term debt) and S = sales.
Control Variables
Audit Partner Specific
We include a number of audit partner controls that have been found in the prior
literature to be associated with audit quality: whether the audit partner is an industry
specialist (AP_INDEXP), the number of clients the partner audits during the year (BUSY),
and the partner’s firm-specific experience (TENURE). Studies document a positive
association between audit quality and individual audit partner industry specialization
(Kallunki et al. 2009; Duh et al. 2009; Ittonen et al., 2010); a negative association with
audit quality and the number of clients the audit partner audits (Goodwin 2011 and Gul
et al. 2013), whilst mix results have been found in relation to auditor tenure and audit
quality (Chi and Huang 2005; Chen et al. 2008; Carey and Simnett 2006; Manry et al.
2008). We also control for the audit firm and audit office the partner works for, and
whether the audit firm is an industry expert (AF_INDEXP).
In addition, an audit partner’s time-invariant generic ability, which cannot be
directly observed and measured, could also be an important variable. In particular,
generic ability could be positively correlated with an audit partner’s network position
e.g. high ability audit partners are just better-connected. Therefore, in addition to the
audit firm and audit office fixed effects we also include audit partner fixed effects. If the
audit partner’s EVC score captures the partner’s generic ability rather than her social
18
capital, then the EVC coefficient should be insignificant given partner fixed-effects are
also included.
Firm Specific
Following prior research (Becker et al. 1998; Myers et al. 2003; Johnson et al.
2002; Frankel et al. 2002; Ashbaugh et al. 2003; DeFond et al. 2002; Andrè et al. 2011)
we control for firm size (SIZE), profitability (ROA, and LOSS), probability of bankruptcy
(PBANK) based on Zmijewski (1984), leverage (LEV), book-to-market ratio (BM), last
year’s total current accruals (LAGCAR), current assets (CA), cash flow from operations
(CFFO), sales growth (SALESG), age of the company (AGE), busy audit period
(DECEMBER), proportion of non-audit fees to audit fees (NAAF) and how many audit
partners signed the audit report (NUM_PARTNERS). In addition, given the French joint
audit regime, we also control for the possible combinations of audit firms: whether the
client firm is jointly audited by two Big 4 audit firms (TWO_BIG4) or whether the client
firm is audited by one Big4 audit firm and one non-Big4 audit firm (ONE_BIG4). Notes to
Table 2 contain each variable’s definition.
The resulting model may also be biased due to the omission of unobservable
client characteristics, such as the extent of the client’s preference for a particular audit
firm, combination of audit firms, or a specific partner, and these client characteristics
may also be correlated with audit quality and with the audit partner’s social capital
(Minutti-Meza 2013). One approach to this potential problem is to use the method of
instrumental variables (IVs) and two-stage least squares (2SLS) as in Zerni (2012).
Unfortunately, we are unable to find a suitable IV. We therefore include client fixed
effects in our regressions. Assuming any client-related omitted variables are time
invariant and captured by the client-specific fixed-effects, this approach mitigates this
endogeneity concern.
Small Earnings Increase Analysis
Our second proxy of audit quality measures the extent of earnings management by
investigating the likelihood of reporting a small earnings increase (Carey & Simnett
2006; Carecello and Li 2013). Prior studies document that firms avoid reporting losses,
tend to report small increases in earnings more frequently than small decreases, and
manage earnings towards a target (Bannister and Newman 1996; Degeorge et. al 1999;
Matsumoto 2002; Abarbanell and Lehavy 2003; Hutton 2005). We therefore estimate
19
the following logistic regression model to test the association between the audit
partner’s social capital and a small earnings increase:
𝐼𝑁𝐶𝑅𝐸𝐴𝑆𝐸 = ∝ +𝛽1𝐸𝑉𝐶𝑝𝑡 + �𝛾𝑖 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒+ �𝜌𝑖 𝐴𝑢𝑑𝑖𝑡 𝐹𝑖𝑟𝑚 𝐹.𝐸.
+�𝜑𝑖𝐴𝑢𝑑𝑖𝑡 𝑂𝑓𝑓𝑖𝑐𝑒 𝐹.𝐸. + �𝜇𝑖 𝐴𝑢𝑑𝑖𝑡 𝑃𝑎𝑟𝑡𝑛𝑒𝑟 𝐹.𝐸. +
�𝛿𝑖 𝑌𝑒𝑎𝑟 𝐹.𝐸. +�𝛾𝑖 𝐶𝑙𝑖𝑒𝑛𝑡 𝐹𝑖𝑟𝑚 𝐹.𝐸. + 𝜖
(2)
INCREASE is an indicator variable coded one if the difference between a firm’s income
before extraordinary items in years t and t-1 (scaled by the market value at the end of
the year t-1) falls in the interval [0.00, 0.02] and zero otherwise. If an audit partner’s
social capital is associated with the quality of their audit then we expect EVC to be
negatively associated with INCREASE. Control variables are the same as in Eq (1) and
are defined in the notes to Table 2.
Empirical Model: Audit Fee To test our second hypothesis we examine whether audit fees are associated with the
social capital of the audit partners. Following a stream of audit fee studies (Ferguson et
al. 2003; Francis et al. 2005; Hay et al. 2006) we control for various determinants of
audit fees in the following regression model:
𝐿𝑁𝐴𝐹𝐸𝐸 = ∝ +𝛽1𝐸𝑉𝐶𝑝𝑡 + �𝛾𝑖 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒 + �𝜌𝑖 𝐴𝑢𝑑𝑖𝑡 𝐹𝑖𝑟𝑚 𝐹.𝐸.
+�𝜑𝑖𝐴𝑢𝑑𝑖𝑡 𝑂𝑓𝑓𝑖𝑐𝑒 𝐹.𝐸. + �𝜇𝑖 𝐴𝑢𝑑𝑖𝑡 𝑃𝑎𝑟𝑡𝑛𝑒𝑟 𝐹.𝐸. +
�𝛿𝑖 𝑌𝑒𝑎𝑟 𝐹.𝐸. +�𝛾𝑖 𝐶𝑙𝑖𝑒𝑛𝑡 𝐹𝑖𝑟𝑚 𝐹.𝐸. + 𝜖
(3) We expect EVC to be positively associated with LNAFEE if demand exists for audit
partners with social capital. Control variables are the same as in Eq. (1) and Eq (2)
except we control for the number of subsidiaries the client firm has (SUBSID) but not for
last year’s total current accruals (LCACCR). Moreover, since we do not need to exclude
the first year the firms implemented IFRS, we include an additional control variable
20
(IFRS) to capture this period. We adjust for firm and year clustering effects in each of
the regression models.
IV. Descriptive Statistics and Results
Audit Quality Analysis
Table 2 Panel A provides descriptive statistics for the audit quality sample. The average
absolute AWCA is 0.078 with an approximate standard deviation of 0.115 and 39.7% of
the sample firms report a small earnings increase (INCREASE). The average eigenvector
centrality score (EVC) for audit partners is 0.267 and 9% of the audit partners are
identified as industry experts (AP_INDEXP). Audit partners are on average BUSY with
2.2 clients per year and have an average TENURE of approximately 4 years. The average
log of firm size is 7.631 and approximately 15% of the firms are loss making. The
average return on assets (ROA) is 0.031 with an average leverage (LEV) of 0.623 and a
book-to-market (BM) of 0.794. 48% of firms are audited by one Big4 and one non-Big4
auditor (ONE_BIG4), while 30% of firms are audited by two Big4 auditors (TWO_BIG4).
Proportion of non-audit fees to audit fees on average is approximately 6%. This is
consistent with the long standing practice in France that auditors are not allowed to
provide both auditing and consulting services to a company. This principle of separation
has been much stricter in France than in English-speaking countries (Andre 2011).
[Insert Table 2]
To obtain a better feel for our network measure, Table 3 presents the
correlations between the EVC and other audit partner and firm characteristics. Above
the diagonal we present correlations for the audit quality sample and below the
diagonal for the audit fee sample. Ex ante we expect the EVC score of an audit partner to
have a positive association with their level of busyness (BUSY) since covering more
firms and possibly more industries will provide, relatively, more opportunities for an
audit partner to gain more social capital. In addition, working for a Big 4 audit firm,
auditing larger clients, and being part of a two Big4 audit team should also increase
their social capital. Table 3 confirms these predictions. Interestingly the audit firm’s
industry expertise (AF_INDEXP) is more highly correlated with an audit partner’s social
capital than the size of the client firm or being part of a two Big4 audit team, while audit
partner tenure (TENURE) is negatively associated with their level of social capital.
[Insert Table 3]
21
Abnormal working capital accruals
Table 4 Panel A presents the estimates of Eq. (1). We find that audit partners
with higher social capital are associated with significantly lower levels of absolute
working capital accruals, indicated by the negative coefficient on EVC significant at the
one percent level. Column (1) only includes industry and year fixed effects, while
column (2) includes all fixed effects. In column (3) we substitute the audit partner’s
industry expertise (AP_INDEXP) with an alternative measure (AP_INDEXP_2) similar to
the one used by Zerni (2012), whereby an audit partner is identified as an industry
expert if they a) audit at least one of the top two largest firms (based on total assets)
within the industry in the fiscal year, and b) audit at least two additional clients
belonging to that industry.15 The results are not sensitive to this new measure. Under all
specifications, the coefficients for the client-specific control variables all have signs
consistent with expectations, except for LEV, CFFO and Ln(AGE), which are insignificant.
ROA and LCACCR are negative and significant at the five percent level, while SALESG and
PBANK are positively and significant at the one and ten percent level respectively.
Coefficients on both TWO_BIG4 and ONE_BIG4 are positive (consistent with Andrè et al.
2011) and significant at the ten percent level.
[Insert Table 4 Panel A]
In untabulated analysis, we split the accrual sample into two groups, firms with
positive abnormal working capital accruals and firms with negative abnormal working
capital accruals, to examine signed accruals. We find that audit partners with high levels
of social capital are associated with higher accruals quality by significantly reducing
both positive and negative accruals.
To assess the sensitivity of our findings to possible adverse effects of self-
selection bias (Ferguson et al. 2003; Francis et al. 2005) we re-run the model on a
subsample – those firms whose audit team comprised of only one Big4 audit firm. Table
4 Panel B presents these results. Under both specifications, EVC is negative and
significant at either the five or ten percent level.
[Insert Table 4 Panel B]
15 We use a less strict identification procedure than that of Zerni (2012) since imposing Zerni’s (2012) procedure yields an extremely small number of audit partners identified as industry specialists. Zerni (2012) identifies an audit partner as an industrial specialist if the individual auditor is (i) ranked among the top two auditors in industry k in fiscal year t based on the amount of audited total assets and (ii) audited at least five clients belonging to industry k in fiscal year t.
22
Overall these results provide support for hypothesis 1. Audit partners with
higher EVC scores, therefore social capital, are associated with higher levels of audit
quality.
Small earnings increase
Table 5 Panels A and B present our logistic regression of the likelihood of firms
reporting a small earnings increase. Consistent with hypothesis 1, EVC is negative and
significant at the one percent level under both model specifications for the full sample -
and negative and significant at five percent level under both specifications for the
ONE_BIG4 sample. For instance, Panel A column (1) reports the coefficient on EVC is
negative and significant (-0.065 with a p-value of <0.001) for our full sample, indicating
that firms whose audit partners have higher EVC scores are less likely to report small
earning increases. This finding is robust to substituting our alternative audit partner
industry specialism measure (AP_INDEXP_2).
In Panel B where we focus on the ONE_BIG4 sample, we report similar results. In
column (1), we find the coefficient on EVC is negative and significant (-0.113 with a p-
value of <0.05). As for our control variables, for the full sample, the results suggest that
larger firms and firms that are older are more likely to report small earnings increases.
On the other hand, loss firms, firms with high levels of current assets to total assets and
firms audited by two Big 4 audit firms are less likely to report a small earnings increase.
In addition for the ONE_BIG4 sample, firms audited by an audit firm with industry
expertise are more likely to report a small earnings increase, while firms with a greater
chance of bankruptcy and higher levels of cash flow from operations and higher levels
of non-audit services are less likely to report a small earnings increase.
[Insert Table 5]
Thus these overall findings tabulated in Table 4 and Table 5 provide support for
our hypothesis (1) that audit partners with higher EVC scores provide higher quality
audits, consistent with both their ability and influence due to social capital obtained via
their network connections benefiting the audit quality.
Audit Fee Analysis
Table 2 Panel B provides descriptive statistics for the audit fee sample. The average
audit fee is €3,019,000 with a standard deviation of €4,383,000. The EVC score and
control variables are not significantly different from the audit quality sample; the EVC
23
score average is 0.24 and has a standard deviation of 0.04. The additional control
variable – the number of subsidiaries (SUBSID) – has an average of 25 with a standard
deviation of approximately 39.
[Insert Table 6 Panel A]
Table 6 Panel A presents the estimates of Eq. (2) using the full sample. We find
that audit partners’ EVC score is positively associated with audit fees and this
association is significant at the ten percent level. Column (1) only includes industry and
year fixed effects, while column (2) includes all but industry fixed effects. In column (3)
we substitute the audit partner’s industry expertise (AP_INDEXP) with an alternative
measure (AP_INDEXP_2) similar to the one used by Zerni (2012). The results are not
sensitive to these new specifications. Under all specifications the coefficients for the
client-specific control variables all have signs consistent with expectations, except for
LOSS, LEV, CA and SALESG, which are insignificant in most specifications. SIZE and
LNSUBSID are positive and significant at the one percent and PBANK is positive and
significant at the five percent level while ROA and BM are negative and significant at the
ten and five percent level respectively. Coefficients on both TWO_BIG4 and ONE_BIG4
are positive and significant at the one percent level consistent with the findings of the
prior literature.
Again to assess the sensitivity of our findings to possible adverse effects of self-
selection bias (Ferguson et al. 2003; Francis et al. 2005) we re-run the model on a
subsample of firms whose audit team consists of only one Big4 audit firm. Table 6 Panel
B presents these results. Under both specifications, EVC is positive and significant at
either the five or 10 percent level. Unlike the results presented in panel A, the
coefficient on the alternative measure of audit partner industry expertise, AP_INDEXP_2,
is positive and significant at the one percent level in column (2). This is consistent with
the findings of Goodwin and Wu (2013) that auditor industry expertise fee premium is
mainly a partner level phenomenon. In unreported tests, we also assess the sensitivity
of the results to an alternative measure of audit firm industry expertise constructed in
the same way as AP_INDEXP_2. The results are robust: the coefficients on audit partner
social capital (EVC) and expertise (AP_INDEXP_2) continue to be positive and significant,
whilst that on the alternative measure of audit firm industry expertise is positive but
not significant.
[Insert Table 6 Panel B]
24
Overall these results provide support for hypothesis 2. Audit partners with
higher levels of social capital earn higher levels of audit fees.
V. Determinants of an audit partner’s social influence.
The results so far suggest that better-networked audit partners provide better
quality audits and command higher audit fees. In Table 7, we present our analysis of the
determinants of audit partners’ EVC score, i.e. social capital. Consistent with the
univariate results in Table 3, the results show that an audit partner’s EVC score is
higher if she audits many clients, works for a Big4 audit firm, is part of a two Big4 audit
team, audits larger clients, works in larger audit-office, and works for an audit firm who
is identified as an industry specialist. An audit partner’s EVC score is lower as the length
of the client relationship increases (TENURE) and if the client is economically significant
for the audit partner (ECON_SIG) with respect to audit fees paid. The regression model
in Table 7 has high explanatory power with an adjusted R-squared of approximately 71
percent.
[Insert Table 7]
One could claim that these results, with respect to an audit partners’ busyness,
tenure, and economic significance of the client are consistent with the prior audit
independence literature. If, as we argue, an audit partners EVC score captures inter alia
the partner’s level of social influence and thereby their level of independence, our
findings also reveal a negative association with auditor independence and the length of
the partner’s client relationship as well as the economic significance of their client.
(Carey and Simnett 2006; Bamber and Iyer 2007; Chi et al. 2009; Ye et al. 2011) The
results are also consistent with the DeAngelo’s (1981a) argument that being busy
auditing many clients, and thereby creating ‘reputational capital’ increases an auditor’s
independence.
Is this a joint-audit phenomenon?
To investigate whether these results hold in a non-joint audit regime we
constructed the network of German audit partners from 2006 to 2010. We chose
Germany since it is a country with both a single audit regime and a requirement for
audit partner signatures. We hand collected from the annual reports the audit partner
data such as, partner name, tenure, audit office, audit fees and non-audit fees. The
25
network comprises of 4,316 directors, 239 client firms, with 870 audit partners from 35
audit firms.
We are unable to replicate our earlier results using the sample of German firms.
One possible explanation is that under a joint-audit regime there are greater
opportunities for an audit partner to create social capital, which they are able to
leverage and provide higher quality audits. In the French setting, there is lower audit
firm concentration, greater chance to audit many clients, along with the opportunity to
work with audit partners from competing audit firms, compared to the German setting.
Further investigation of our German sample reveals that unlike French audit partners,
we observe German audit partners on average conducting only one audit per year. The
mean for BUSY is 1.4 and the median is 1, compared to France where the mean of BUSY
was 2.2 and the median is 2. Consequently, in any given year the German audit partner
is on average only connected to one client (via their directors) and connected to audit
partners from their own audit firm. This results in a network with very low density
providing very little opportunity for the audit partners to create informational
advantages and social influence beyond their audit firm and their one client. Prior
research finds that as density increases (and the number of ties between network
members grows) communication across the network becomes more efficient and power
imbalances are able to flourish (Rowley, 1997).
This would be the first paper to provide some empirical support to suggest that a
joint-audit regime provides opportunities to increase audit quality (Holm and
Thinggaard 2011; Lesage et al 2012; Lesage and Ratzinger-Sakel 2012). To-date prior
research has been unable to find any association between joint-audits and audit quality,
except in the case of voluntary joint-audits (Zerni et al. 2012). However, at this stage we
are very cautious to suggest that the lack of any significant association between a
German audit partner’s social capital and audit quality indicates that audit partners can
only create and leverage their social capital in a joint-audit regime. It might be the case
that we did not construct the ‘right’ network, and that German audit partners are part of
a denser network that we are unable to observe, e.g. connections through consulting
service or connections from being members of advisory boards, etc., through which they
are able to create social capital.
26
VI. Conclusion
This paper is the first to connect an audit partner’s structural position in a network to
her performance. We find evidence in support of the hypothesis that audit partners
appear to gain benefits through their position in the network, which enhance their audit
quality and enable them to command an audit fee premium. We also document that
audit partners’ social capital is higher if they audit more clients, and is lower if the
partners have longer tenure with their clients and the more economically significant the
clients are to the audit partner.
This research suggests that regulators should consider providing an
environment in which audit partners have opportunities to enhance their networks, and
in turn to improve audit quality. This however may or may not manifest itself in a joint
audit system. Our findings do suggest that policymakers should examine audit quality
and audit market structure enhancement together as a complex whole.
27
Appendix Construction of the Network
The diagram below shows a network of 10 people (actors). Each line represents a
connection between the individuals. In the context of this study we can consider, for
example, actors 1 and 2 as directors of two different firms, firm A and firm B. These
directors do not sit on any other boards and therefore there is no possibility that they
are connected to each other. Actor 3 is an audit partner who works for audit firm C and
audits firms A and B. Therefore she is connected both to Director 1 and Director 2. Actor
6 is a director in firm F that audit partner 3 also audits, hence audit partner 3 is also
connected to Director 6. Director 6 does not sit on the same board with either Director 1
or Director 2 and therefore is not connected to either of them. Actors 4, 5 and 7 are
audit partners working for different audit firms and jointly audit with actor 3. Therefore
actor 3 is also connected to: actor 4 since they jointly audit firm A, actor 5 since the
jointly audit firm B, and actor 7 since they jointly audit firm F. Actors 4, 5 and 7 do not
audit the same firms and therefore are not directly connected to each other. Actor 8 is a
director in Firm H and also sits on the same board of directors as Director 6 in Firm F.
Therefore, Director 6 and Director 8 are also connected to each other through sitting on
the same board Firm Z. Actors 9 and 10 are audit partners working for different audit
firms who jointly audit firm H. Thereby they are both connected to Director 8 but not
Director 6. They are also directly connected to each other since they audit the same
firm, Firm H.
Director 1 Auditor 5 Auditor 7 Auditor 9 (Firm A) (Audit Firm D) (Audit Firm G) (Audit Firm I) Auditor 3, Director 6, Director 8, (Audit Firm C) (Firm F) (Firm H) Auditor 10 (Audit Firm J) Director 2 (Firm B) Auditor 4 (Audit Firm E)
28
Eigenvector centrality measure
The eigenvector centrality (EVC) is a relative score recursively defined as a function of
the number of strengths of connections to an actor’s neighbors and as well as those
neighbors’ centralities. Let x(i) be the EVC score of node vi. Then,
𝑥(𝑖) =1λ
� 𝑥(𝑗)𝑗∈Γ(υ𝑖)
=1λ�𝐴𝑖,𝑗𝑥(𝑗)𝑁
𝑗=1
Here λ is a constant. Equation (1) can be rewritten in vector form, where x ={x(1), x(2),
x(3),…..,x(N)}′ is the vector of EVC scores of all nodes.
𝐱 =1λ𝐀𝐱
λx = Ax
This is the well-known eigenvector equation where this centrality takes its name from λ
is an eigenvalue and x is the corresponding eigenvector of matrix A. Obviously several
eigenvalue/eigenvector pairs exist for an adjacency matrix A. The EVC of nodes are
defined on the basis of the Perron eigenvalue λA (the Perron eigenvalue is the largest of
all eigenvalues of A and is also called the principal eigenvalue). If λ is any other
eigenvalue of A then λA> |λ|. The eigenvector x = {x(1), x(2),….,x(N)}′ corresponding to
the Perron eigenvalue or principal eigenvector. Thus the EVC of a node vi is the
corresponding element x(i) of the Perron eigenvector x. Note that when the adjacency
matrix A is symmetric all elements of the principal eigenvector x are positive.
Computing a node’s EVC it takes into consideration its neighbor’s EVC scores.
Because of its recursive definition, EVC is suited to measure nodes’ power to influence
other nodes in the network both directly and indirectly through its neighbors.
Connections to neighbors that are in turn well connected themselves are rated higher
than connections to neighbors that are weakly connected.
29
REFERENCES
Abarbanell, J., and Lehavy, R. 2003. Biased forecasts or biased earnings? The role of reported earnings in explaining apparent bias and over-/underreaction in analysts’ earnings forecasts. Journal of Accounting and Economics, 36 (1–3): 105–146. Adler, P.S., and Kwon S. 2002. Social capital: Prospects for a new concept, Academy of Management Review, 27: 17-40. Akram, Q.F., and Christophersen, C. Interbank overnight interest rates – gains from systemic importance. Norges Bank working paper. Available at SSRN: http://ssrn.com/abstract=1636763. Andrè, P., Broye, G., Pong, C.K.M., and Schatt, A. 2011. Audit fees, big four premium and institutional settings: The devil is in the details!. Available at SSRN: http://ssrn.com/abstract=1554842. Ashbaugh, H., LaFond, R., and Mayhew, B. 2003. Do nonaudit services compromise auditor independence? Further evidence. The Accounting Review 78 (3): 611-639. Ashton, A. H. 1991. Experience and error frequency knowledge as potential determinants of audit expertise. The Accounting Review, 218-239. Bamber, E. M., and Iyer, V.M. 2007. Auditors’ identification with their clients and its effect on auditors’ objectivity. Auditing: A Journal of Practice and Theory. Vol 26(2): 1-24. Bannister, J.W., and H.A. Newman. 1996. Accrual usage to manage earnings toward financial analysts’ forecasts. Review of Quantitative Finance and Accounting, 7 (3): 259–278. Bartlett, R.W. 1993 A Scale of Perceived Independence: New Evidence on an Old Concept, Accounting, Auditing & Accountability Journal, 6(2): 52-67. Bearden, J., Atwood, W., Freitag, P. Hendricks, C., Mintz, B., and Schwartz, M. 1975 The nature and extent of bank centrality in corporate networks. Paper presented at the Annual Meeting of the American Sociological Association, San Francisco, August. Beasley, M. S., J. V. Carcello, and D. Hermanson. 2000. Should you offer your job to an external auditor? The Journal of Corporate Accounting and Finance 11(4): 35-42. Bech, M., Chapman, J., and Garratt, R., 2010. Which bank is the ‘central’ bank? Journal of Monetary Economics, 57(3): 352-363. Becker,C . L., DeFond, M.L., Jiambalvo, J., and Subramanyam, K.R. 1998. The effect of audit quality on earnings management. Contemporary Accounting Research 15 (Spring): 1-24. Bedard, J. C., Biggs, S. 1991. The effect of domain-specific experience on evaluation of management representations in analytical procedures. Auditing: A Journal of Practice & Theory 10 (Supplement): 77-90. Bonacich, P. 1972. Factoring and weighing approaches to status scores and clique identification. Journal of Mathematical Sociology 2: 113–120. Bonacich, P. 1987. Power and centrality: a family of measures. American Journal of Sociology 92: 1170-1182. Bonacich, P. 2007. Some unique properties of eigenvector centrality. Social Networks 29(4): 555-564 Bonner, S.E., and Lewis, B.L. 1990. Determinants of auditor expertise. Journal of Accounting Research, 28 (Supplement): 1-20. Borgatti, S.P., 2005. Centrality and network flow. Social Networks 27 (1), 55–71
30
Borgatti, S.P. 2006. Identifying sets of key players in a social network. Computional Math Organization Theory 12:21-34. Bouton, D. 2002. Promoting Better Corporate Governance in Listed Companies, Chairman of the AFEP/AGREF-MEDEF Working Group. Brass, D. J. 1984. Being in the right place: A structural analysis of individual influence in an organization. Administrative Science Quarterly 29: 518-539. Brass, D. J. 1985. Men’s and women’s networks: A study of interaction patterns and influence in an organization. Academy of Management Journal 28: 327-343. Brass, D. J. 1992. Power in organizations: A social network perspective. In Moore, G., and J. A Whitt (Eds). Research in Politics and Society, Greenwich, CT: JAI Press: 295-323. Brass, D. J., and Burkhardt. M. E. 1992. Centrality and power in organizations. In N. Nohria and R. Eccles (Eds.), Networks and organizations: Structure, form, and action (pp. 191-215). Boston: Harvard Business School Press. Brass, D. J., and Burkhardt, M.E. 1993. Potential power and power use: An investigation of structure and behavior. Academy of Management Journal 36(3): 441-470 Brass, D.J., and Krackhardt, D. 2012. Power, politics, and social networks in organizations. In G.R. Ferris & D. C. Treadway (Eds.), Politics in organizations: Theory and research considerations, 355-375. New York: Routlege. (5) Burkhardt, M. E., and. Brass, D.J. 1990. Changing patterns or patterns of change: The effects of a change in technology on social network structure and power. Administrative Science Quarterly 35: 104-127. Burt, R. 1987. Social contagion and innovation: Cohesion versus structural equivalence, American Journal of Sociology, 92: 1287-1335. Burt, R.S. 1992. Structural Holes: The Social Structure of Competition, Cambridge MA: Harvard University Press. Burt, R. 2000. Structural holes versus network closure as social capital. (Pre-print for a chapter in Social Capital: Theory and Research. Edited by Lin, N., Cook, S. and Burt, R. 2001. Cameran, M., Francis, J., Marra, A., and Pettinicchio, A. 2012. The cost of mandatory auditor rotation: evidence from the Italian experience, Working paper Carey, P., and Simnett, R. 2006. Audit partner tenure and audit quality, The Accounting Review, 81: 653- 676. Carcello, J. V., Hermanson, D.R., and Huss, H.F. 1996. Inappropriate audit partner behaviour: Views of partners and senior managers. Behavioral Research in Accounting (Supplement): 245-268. Carcello, J. V., Hermanson, D.R., and Huss, H.F. 2000. Going-concern opinions: The effects of partner compensation plans and client size. Auditing: A Journal of Practice & Theory (Spring): 67-77. Carcello, J.V., and Li, C. 2013 Costs and benefits of requiring an engagement partner signature: Recent experience in the United Kingdom. The Accounting Review, 88(5): 1511-1546. Chan, K.H., and Wu, D. 2011. Aggregate Quasi Rents and Auditor Independence: Evidence from Audit Firm Mergers in China. Contemporary Accounting Research 28(1): 175–213. Chen, C.Y., Lin, C.J., and Lin, Y.C. 2008. Audit partner tenure, audit firm tenure, and discretionary accruals: Does long auditor tenure impair earnings quality? Contemporary Accounting Research, 25:
31
415–445. Chen, C.J.P., Su, X., and Wu, X. 2009. Forced audit firm change, continued partner-client relationship, and financial reporting quality, Auditing: A Journal of Practice & Theory, 28(2): 227-246. Chen, C.J.P., Su, X., and Wu, X. 2010. Client importance, institutional improvements, and audit quality in China, an office and individual auditor level analysis, The Accounting Review, 85(1): 127-158. Chi, H and Chin, C. 2011. Firm versus partner measures of auditor industry expertise and effects on auditor auality, AUDITING: A Journal of Practice & Theory: 30(2): 201-229. Chi, W., and H. Huang. 2005. Discretionary accruals, audit-firm tenure and audit-partner tenure: Empirical evidence from Taiwan. Journal of Contemporary Accounting and Economics 1 (June): 65-92. Chi, W., Huang. H., Liao, T., and Xie, H. 2009. Mandatory audit partner rotation, audit quality, and market perception: Evidence from Taiwan. Contemporary Accounting Research 26(2): 359–91. Chi, W., E.B. Douthett Jr., L. L. Lisic, 2012. Client importance and audit partner independence. Journal of Accounting and Public Policy 31: 320-336. Chi, W., Douthett Jr, E.B., and Lisic, L. L. 2012. Client importance and audit partner independence. Journal of Accounting and Public Policy 31: 320-336. Cialdini, R. B. 1989. Indirect tactics of impression management: Beyond basking. In R. A. Giacalone & P. Rosenfield (Eds.), Impression management in the organization (pp. 45-56). Hillsdale, NJ: Erlbaum. Coleman, J. S. 1988. Social Capital in the Creation of Human Capital, American Journal of Sociology 94, Supplement: Organizations and Institutions: Sociological and Economic Approaches to the Analysis of Social Structure: 95-120. Coffee, J.C. 2005. Gatekeepers: The Professions and Corporate Governance. Clarendon Lectures in Management Studies, pp.298, ISBN: 978-0-19-928809-0. Cohen, L., Frazzini, A., and Malloy,C.J. 2011. Sell Side School Ties. The Journal of Finance, 65(August): 1409–1437. Cohen, L., Frazzini, A., and Malloy,C.J. 2008. The small world of investing: Board connections and mutual fund returns, Journal of Political Economy, 116: 951- 981. Cohen-Cole, E., Patacchini, E., and Zenou, Y. 2010. Systemic risk and network formation in the interbank market. Maryland working paper, available at http://www.rhsmith.umd.edu/files/Documents/Centers/CFP/SystemicRisk.pdf Cook, K.S., Emerson, R.M., Gilmore, M.R., and Yamagishi, T. 1983. The distribution of power in exchange networks: Theory and experimental results. American Journal of Sociology, 89: 275-305. Costenbader, E., and Valente, T.W. 2003. The stability of centrality measures when networks are sampled. Social Networks 25: 283-307. DeAngelo, L.E. 1981a. Auditor Size and Audit Quality, Journal of Accounting and Economics, 3 (August): 183-199. DeAngelo, L.E. 1981b. Auditor Independence, “Low Balling”, and Disclosure Regulation, Journal of Accounting and Economics, 3 (August): 113-127. DeFond, M., Raghunandan, K., and Subramanyam, K. R. 2002. Do non-audit service fees impair auditor independence? Evidence from going concern opinion decisions, Journal of Accounting Research (June): 1247–1274.
32
DeFond, M., Francis, J. 2005. Auditing research after Sarbanes-Oxley. Auditing: A Journal of Practice and Theory 25: 5-30. DeFond, M.L., and Park, C.W. 2001. The reversal of abnormal accruals and the market valuation of earnings surprises. The Accounting Review 76 (July): 375-404. DeFond, M., and J. Zhang, 2014. A Review of the archival auditing research. Available at SSRN: http://ssrn.com/abstract=2411228 or http://dx.doi.org/10.2139/ssrn.2411228 Degeorge, F., Patel, J., and Zeckhauser, R. 1999. Earnings management to exceed thresholds. The Journal of Business, 72 (1): 1–34. Dopuch, N., and Simunic, D. A. 1982. Competition in auditing: An assessment. In. Fourth Symposium on Auditing Research, University of Illinois. Dopuch, N., and Simunic, D. A. 1980. The nature and competition of the auditing profession: A descriptive and normative view’, in J.W. Buckley and J.F. Weston (eds) Regulation and the Accounting Profession, Lifetime Learning Publications. Duh, R.-R., Hou, C.-P., and Huang, D.-F. 2009. Audit partner’s general experience, industry specialization, and audit quality. Working paper, National Taiwan University, China University of Technology, and National Dong Hwa University. El-Khatib, R., Fogel, K., and Jandik, T. CEO network centrality and merger performance. Available at SSRN: http://ssrn.co,/abstract=2024484 Emerson, R.M. 1972. Exchange Theory, Part II: Exchange Relations and Networks. (: 58-87) in Sociological Theories in Progress, edited by J. Berger, M. Zelditch, and B. Anderson. Vol.2. Boston: Houghton Mifflin. Ettredge, M., Scholz, S., Smith, K., and Sun, L. 2010. How do misstatements begin? Evidence of biased financial reporting preceding misstatements. Journal of Business Finance and Accounting 37: 332–355. Fargher, N.L., Mayorga, D., and Trotman, K.T. 2005. A field-based analysis of audit work paper review. Auditing: A Journal of Practice and Theory 24(2): 85–110. Faust, K. 1997. Centrality in affiliation networks, Social Networks, 19:157-191. Ferguson, A., Francis, J., and Stokes, D. 2003. The effects of firm-wide and office-level industry expertise on audit pricing. The Accounting Review 78(2): 429-448. Fernández, R.M., Castilla, E.J., and Moore, P. 2000. Social Capital at Work: Networks and Employment at a Phone Center, American Journal of Sociology, 105(5): 1288-1356. Field, J. 2003. Social Capital, Routledge, London Fogel, K., Ma, L., and Morck, R. 2013. Powerfully independent directors, Illinois working paper, available at https://business.illinois.edu/finance/papers/2013/Morck.pdf Francis, J.R., and Yu, M.D. 2009. The effect of Big 4 office size on audit quality, The Accounting Review, 84 (5): 1521-1552. Francis, J. 2011. A framework for understanding and researching audit quality. Auditing: A Journal of Practice and Theory, 30(2): 125-152. Francis, J.R., Reichelt, K., and Wang, D. 2005. The pricing of national and city-Specific reputations for industry expertise in the U.S. audit market. The Accounting Review: 80(1): 113-136. Frankel, R.M., Johnson, M.F., and Nelson, K.K. 2002. The relationship between audit fees for nonaudit services and earnings quality. The Accounting Review 77 (Supplement): 71-105.
33
Freeman, L.C. 1977. Set of Measures of Centrality Based on Betweenness. Sociometry 40:35-41. Freeman, L.C. 1979. Centrality in Social Networks Conceptual Clarification. Social Networks 1:215-239. Freeman, L.C. 1980. The gatekeeper, pair-dependency and structural centrality, Quality and Quantity, 14: 585-592 Frenandez, R.M., and McAdam, D. 1988. Social networks and social movement: Multiorganizational fields and recruitment to Mississippi freedom summer. Sociological Forum, 3:441-60. Friedkin, N.E. 1981. The Development of Structure in Random Networks: Analysis of the Effects of Increasing Network Density on Five Measures of Structure. Social Networks 3: 41-52. Friedkin, N.E.1991. Theoretical Foundations for Centrality Measures. American Journal of Sociology 96: 1478-1504. Freriksson, A., and Kankaanpӓӓ, J. 2012. Audit committee independence, network centrality and earnings quality. Available at SSRN: http://ssrn.com/abstract=2171573. Galaskiewicz, J. 1979. Exchange networks and community politics. Beverly Hills, CA: Sage. Geiger, M. A., Raghunandan, K., and Rama, D.V. 2006. Auditor decision-making in different litigation environments: The Private Securities Litigation Reform Act, audit reports and audit firm size. Journal of Accounting and Public Policy 25 (3): 332-353. Goodwin, J. 2011. Audit partner busyness and audit quality. Available at SSRN: http://ssrn.com/abstract=1799681. Goodwin, J., and Wu, D. 2013. Is the effect of industry expertise on audit pricing an office-level or a partner level phenomenon? Review of Accounting Studies (forthcoming). Gould, S.J. 2002. The structure of evolutionary thought. Harvard University Press, Cambridge Grosser, T J., Lopez-Kidwell, V., and Labianca, G. 2010. A social network analysis of gossip in organizational life. Group and Organization Management, 35: 177-212. Gul, F.A., Wu, D., and Yang, Z., 2013. Do individual auditors affect audit quality? Evidence from archival data. The Accounting Review, 88(6): 1993-2023. Hay, D.C., Knechel, W.R., and Wong, N. 2006. Audit fees: A meta-analysis of the effect of supply and demand attributes. Contemporary Accounting Research 23(1): 141-191. Hamilton, R. E., and Wright, W.F 1982. Internal control judgments and effects of experience: Replications and extensions. Journal of Accounting Research Autumn(II): 756-65. Hamilton, J., Ruddock, C., Stokes, D., and Taylor, S. 2005. Audit partner rotation, earnings quality and earnings conservatism. Working paper, University of Technology, Sydney and University of New South Wales. [Online] Available: http://ssrn.com/abstract=740846. Hanneman, R.A., and Riddle, M., 2005. Introduction to social network methods. Riverside, CA: University of California, Riverside (published in digital form at http://faculty.ucr.edu/~hanneman/) Hochberg, Y.V., Ljungqvist, A. and Lu, Y. 2007. Whom you know matters: Venture capital networks and investment performance, Journal of Finance, 62: 251-296. Holm, C. and Thinggaard, F. 2011. Joint audits – benefit or burden? Working paper. Available from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1702867, accessed 31/08/2012.
34
Hong, H., Kubik, J.D., and Solomon, A. 2000, Security analysts’ career concerns and herding of earnings forecasts, Rand Journal of Economics, 31: 121-144. Horton, J, Millo, Y., and Serafeim, G. 2009. Paid for connections or too connected to be good? Social networks and executive and non-executive directors’ compensation, Journal of Business Finance and Accounting, 39(3/4):?? Horton, J., and Serafeim, G. 2012. Security Analyst Networks, Performance and Career Outcomes, Harvard Working paper. Hutton, A. P. 2005. Determinants of managerial earnings guidance prior to regulation fair disclosure and bias in analysts' earnings forecasts. Contemporary Accounting Research, 22(4): 867–914. Ibarra, H., and Andrews, S.B. 1993. Power, social influence and sense-making: Effects of network centrality and proximity on employee perceptions. Administrative Science Quarterly, 38: 277-303. Ittonen, K., Johnstone, K., and Myllymäki, E. R. 2010. Characteristics of individual audit partners and their client portfolios: Effects on discretionary accruals and audit fees. Working paper, University of Vaasa, and University of Wisconsin School of Business. Jackson, M. 2008. Social and Economic Networks. NY: Princeton University Press JeanJean, T., Marmousez, S., and Sirois, L-P. 2013. Social networks, auditor independence and earnings quality. Working Paper. Jones, J. 1991. Earnings management during import relief investigations. Journal of Accounting Research 29 (Autumn): 193-228. Kallunki, J.-P., Nilsson, H., Sahlström, P., and Zerni, M. 2009. Do the leading audit partner characteristics affect audit quality? Working paper, University of Oulu and Umeå School of Business. Kilduff, M., and Krackhardt, D. 1994. Bringing the individual back in: A structural analysis of the internal market for reputation in organizations. Academy of Management Journal, 37: 87-108. Knechel, R., Vanstraelen, A., Zerni, M., 2013. Does the identity of engagement partners matter? An analysis of the persistence and economic consequences of low audit partner quality. Working paper. Knoke, D., and Burt, R.S. 1983. Prominence. In R.S. Burt & M.J. Miner (Eds.), Applied network analysis: A methodological introduction (pp. 195-222). Beverly Hills, CA: Sage. Krackhardt, D. 1990. Assessing the political landscape: Structure, cognition, and power in organizations. Administrative Science Quarterly, 35: 342-369. Larcker, D. , and S. A. Richardson, 2004. Fees paid to audit firms, accrual choices, and corporate governance. Journal of Accounting Research, 42: 625-658. Le Maux, J. 2004. Le co-commissariat aux comptes a la francaise. Revue Francaise de la Comptabilite (May): 34–38. Lesage, C. and Ratzinger-Sakel, N. 2012. Are the findings of international studies reliable? The joint audit example, http://papers.ssrn.com/abstract=2176729, accessed 12/11/2012. Lesage, C., Ratzinger-Sakel, N. and Kettunen, J. 2012. Struggle over joint audit: On behalf of public interest?, Working paper. Liu, T. 2009. An empirical study about the effects of interlocking directorates strategy on the firm's output in the dynamic environment. jcai :818-820, International Joint Conference on Artificial Intelligence
35
Lennox, C. L. 2005. Audit quality and executive officers' affiliations with CPA firms. Journal of Accounting and Economics, 39 (2): 201-231. Maijoor, S. and Vanstraelen, A. (2006) Earnings management within Europe: the effects of member state audit environment, audit firm quality and international capital markets. Accounting and Business Research 36(1): 33–52. Manry, D.L., Mock, T.J., Turner, J.L. 2008. Does increased audit partner tenure reduce audit quality, Journal of Accounting, Auditing and Finance, 23(4): 553-572 Matsumoto, D. A. 2002. Management’s incentives to avoid negative earnings surprises. The Accounting Review, 77(3): 483–514 McClurg, S., and Young, J.K. 2011. A Relational Political Science, guest editors and authors of the introduction to a symposium on the study of social networks in Political Science, PS: Political Science and Politics 44(1): 39-43. Mehra, A., Kilduff, M., and Brass, D.J. 2001. The social networks of high and low self-monitors: Implications for workplace performance. Administrative Science Quarterly, 46: 121-146. Menon, K., and Williams, D.D. 2004. Former audit partners and abnormal accruals. The Accounting Review. 79 (October): 1095-1118. Meyerson, E.M. 1994, The impact of human capital, social capital and compensation, ACTA sociologica. 37(4).: ?? Minutti-Meza, M. 2013, Does auditor industry specialization improve audit quality. Journal of Accounting Research, 51(4): 779-817. Mintz, B.A., and Schwartz, M. 1981a. Interlocking directorates and interest group formation. American Sociological Review, 46: 851-868. Mintz, B.A., and Schwartz, M. 1981b. The structure of intercorporate unity in American business. Social Problems 28: 87-103. Mizruchi, M.S. and Bunting, D. 1981. Influence in corporate networks: An examination of four measures. Administrative Science Quarterly, 26(3): 475-489. Mizruchi, M.S.. 1982 The American Corporate Network, 1904-1974, Beverly Hills: Sage Publications, 216 pp. Mizruchi, M. S., and Potts, B.B. 1998. Centrality and power revisited: actor success in group decision making. Social Networks 20: 353–387. Mizruchi, M. S., and Stearns, L.B. 2001, Getting deals done: the use of social networks in bank decision making, American Sociological Review 66: 647-671. Molm, L.D. 1990. Structure, Action and Outcomes: The Dynamics of Power in Exchange Relations. American Sociological Review 55:427-47. Myers, J.N, Myers, L.A., and Palmrose, Z-V. 2003. Mandatory Auditor Rotation: Evidence from Restatements, Working paper, available from http://aaahq.org/audit/midyear/04midyear/papers/myers.pdf Owhoso, V.E., Messier Jr, W.F., and Lynch, J.G. 2002. Error Detection by Industry-Specialized Teams during Sequential Audit Review. Journal of Accounting Research 40(3): 883-900. Parlin, J. C., and Bartlett, R.W. 1994. Prior employment effects and independence in fact. Business and Professional Ethics Journal 13 (1 and 2): 185-202.
36
Peek, E., Meuwissen, R.H.G., Moers, F., and Vanstraelen, A. 2013. Comparing abnormal accruals estimates across samples: An international test. European Accounting Review, 22(3): 533-572. Pfeffer, J. 1981. Power in Organizations, Marshfield, Mass.: Pitman Publishing Inc. Piot, C., and Janin, R. 2007. External auditors, audit committees and earnings management in France. European Accounting Review 16 (2): 429-454. Proctor, C.H., and Loomis, C.P. 1951. Analysis of sociometric data, in P.W. Holland, and S. Leinhardt, ed.: Research Methods in Social Relations 561–586 (Dryden Press, New York) Reagans, R. E., and Zuckerman, E.W. 2001. Networks, diversity and performance: The social capital of R&D teams. Organization Science, 12: 502-518. Renneboog, L., and Zhao, Y. 2011. Us Knows Us in the UK: On Director Networks and CEO Compensation. Working Paper, Tilburg. Reynolds, J. K., and Francis, J.R. 2000. Does size matter? The influence of large clients on office-level auditor reporting decisions. Journal of Accounting and Economics 30 (December): 375- 400. Richardson, S., Sloan, R., Soliman, M., and Tuna, I. 2006. The implications of firm growth and accounting distortions for accruals and profitability. The Accounting Review, 81 (3): 713-743. Rosenthal, N., Fingrutd, M., Ethier, M., Karant, R., and McDonald, D. 1985, Social movements and network analysis: A case study of nineteenth-century women's reform in New York state, American Journal of Sociology 90: 1022-1054. Rowley, T.J. 1997. Moving beyond dyadic ties: A network theory of stakeholder influences. Academy of Management Review, 22: 887-910. Roy, W.G. 1983. Interlocking directorates and the corporate revolution. Social Science History, 7, 143-164. Ruddock, C., Taylor, S. B., and Taylor, S. L. 2006. Nonaudit services and earnings conservatism: Is auditor independence impaired?, Contemporary Accounting Research,23, 701-746. Sabiduss, G. 1966. Centrality Index of a Graph. Psychometrika 31:581-581 Scott, W. R., and Davis, G. F. 2007. Organizations and organizing: Rational, natural, and open system perspectives. New Jersey: Pearson Prentice Hall. Settoon, R.P., and Mossholder, K.W. 2002. Relationship quality and relationship context as antecedents of person- and task-focused interpersonal citizenship behavior. Journal of Applied Psychology, 87: 255-267. Sparrowe, R.T., and Liden, R.C. 1997. Process and structure in leader-member exchange. Academy of Management Review, 22: 522-552 Stein, J.C. 2008. Conversations among competitors, American Economic Review, 98(5): 2151-2162. Stevenson, W.B, and Radin, R.F. 2008. Social capital and social influence on the board of directors. Journal of Management Studies, 46(1): 16-44. Taylor, S.D. 2011 Does audit fee homogeneity exist? Premiums and discounts attributable to individual partners. AUDITING: A Journal of Practice & Theory, 30(4): 249-272. Useem, M. 1984. The Inner Circle: Large Corporations and the Rise of Business Political Activity, New York: Oxford University Press.
37
Uzzi, B. 1997. Social structure and competition in inter-firm networks: The paradox of embeddedness, Administrative Science Quarterly, 42: 35-67. Ye, P. Carson, E., and Simnett, R. 2011. Threats to auditor independence: The impact of relationship and economic bonds', Auditing: A Journal of Practice and Theory, 30(1); 121-148. Watts, D. J., and Strogatz, S.H. 1998. Collective dynamics of 'small-world' networks. Nature 393 (6684): 440–442 Wright, A., and Wright, S. 1997. An examination of factors affecting the decision to waive audit adjustments. Journal of Accounting, Auditing and Finance 12 (Winter): 15–36. Zerni, M. 2012. Audit partner specialization and audit fees: Some evidence from Sweden. Contemporary Accounting Research, 29(1): 312–340 Zerni, M., Haapamäki, E., Järvinen, T. and Niemi, L. 2012, Do joint audits improve audit quality? Evidence from voluntary joint audits. European Accounting Review, 21(4):731-765. Zmijewski, M. E. 1984. Methodological issues related to the estimation of financial distress prediction models. Journal of Accounting Research (Supplement): 59–82.
38
Table 1 Sample Selection
Panel A: Network Sample
Year # of Directors # of Companies with
Directors # of Audit Partners # of Audit Firms
2004 2375 207 395 109 2005 2757 253 460 115 2006 2807 261 500 113 2007 3020 284 556 118 2008 3049 293 572 112 2009 3012 288 590 117 2010 2937 273 577 110 Total 19,957 1,859 3,650 794
Unique 4,159 326 734 192 Panel B: Sample for audit quality analysis
Number of Observations
Audit partner observations sample from 2004-2010 3,650 Delete:
Financial firms (590) Missing data to construct control variables ( 67) Partner’s from non-lead audit firms (1,554)
1,439 Delete: First year implemented IFRS (188)
Sample for audit quality analysis 1,251
Panel C: Sample for audit fee analysis
Number of Observations
Audit partner observations sample from 2004-2010 3,650 Delete:
Financial firms (590) Missing data to construct control variables ( 59) Partner’s from non-lead audit firms (1,554)
Sample for audit fee analysis 1,485
39
Table 2 Descriptive Statistics Panel A: Accounting Quality Analysis (n=1,251) Variable Mean Std. Deviation Quartile 1 Quartile 3 ABS_AWCA 0.078 0.115 0.020 0.089 INCREASE 0.397 0.489 0.000 1.000 EVC 0.267 0.044 0.001 0.016 AP_INDEXP 0.094 0.450 0.000 0.000 BUSY 2.209 1.511 1.000 3.000 TENURE 4.080 2.854 2.000 6.000 AF_INDEXP 0.282 0.450 0.000 1.000 SIZE 12338.070 27885.330 424.452 10734.000 Ln(SIZE) 7.631 2.029 6.051 9.281 ROA 0.031 0.082 0.015 0.059 LOSS 0.146 0.354 0.000 0.000 PBANK 0.114 0.169 0.009 0.160 LEV 0.623 0.265 0.504 0.737 BM 0.794 0.817 0.355 0.932 LCACCR -0.003 0.133 -0.001 -0.049 CA 0.481 0.187 0.348 0.607 CFFO 0.077 0.074 0.047 0.108 SALESG 0.084 0.321 0.011 0.127 AGE 30.909 19.558 15.000 50.000 Ln(AGE) 3.209 0.703 2.708 3.912 DECEMBER 0.846 0.361 1.000 1.000 NAAF 0.058 0.114 0.000 0.069 NUM_PART 1.296 0.457 1.000 2.000 TWO_BIG4 0.297 0.457 0.000 1.000 ONE_BIG4 0.478 0.500 0.000 1.000 Panel B: Audit Fee Analysis (n=1,485) Variable Mean Std. Deviation Quartile 1 Quartile 3 AUDIT_FEE (th.) 3019.220 4383.590 323.000 4104.610 Ln(AUDIT_FEE) 13.889 1.549 12.685 15.228 EVC 0.241 0.043 0.001 0.013 AP_INDEXP 0.096 0.294 0.000 0.000 BUSY 2.247 1.615 1.000 3.000 TENURE 4.063 2.874 2.000 6.000 AF_INDEXP 0.285 0.452 0.000 1.000 SUBSID 24.796 38.799 6.000 23.000 Ln(SUBSID) 2.620 1.091 1.946 3.178 SIZE (mil.) 11777.170 26978.970 388.500 9426.000 Ln(SIZE) 7.545 2.051 5.962 9.151 ROA 0.032 0.083 0.016 0.060 LOSS 0.141 0.349 0.000 0.000 PBANK 0.114 0.168 0.009 0.156 LEV 0.622 0.252 0.504 0.735 BM 0.850 0.889 0.369 1.010 CA 0.487 0.192 0.348 0.616 CFFO 0.076 0.079 0.047 0.107 SALESG 0.099 0.333 0.008 0.140 AGE 30.119 19.488 14.000 49.000 Ln(AGE) 3.167 0.739 2.639 3.892 DECEMBER 0.844 0.363 1.000 1.000 NAAF 0.059 0.113 0.000 0.074 NUM_PART 1.295 0.456 1.000 2.000 TWO_BIG4 0.286 0.452 0.000 1.000 ONE_BIG4 0.483 0.499 0.000 1.000
40
Variables Definition ABS_AWCA = Absolute abnormal working capital accruals divided by average total assets at the
end of year t. Abnormal working capital accruals is measured as follows: Non-cash working capitalt – ((Non-cash working capitalt-1/Salest-1) * Salest), where: Non-cash working capital = (Current assets - Cash and short term investments) – (Current liabilities - Short-term debt).
INCREASE = an indicator variable coded one if the difference between a firm’s income before extraordinary items in years t and t-1 (scaled by the market value at the end of the year t-1) falls in the interval [0.00, 0.02] and zero otherwise.
AUDIT_FEE = The audit fees (in thousands of euros) in time t. Ln(A_FEE) = The natural logarithm of audit fee in time t. EVC = The audit partner’s eigenvector centrality score at time t. AP_INDEXP = An indicator variable with the value of one if the audit partner is identified as an
industry expert for industry i in time t and zero otherwise. In line with prior studies (Balsam et al. 2003; Chin and Chi 2009) we measure market share using the number of clients audited by an audit partner within an industry. We then rank audit partners in each industry by their market share and identify an audit partner as an industry expert if they are the largest supplier in the industry. Chin and Chi (2009) argues that such a base avoids a bias toward large clients that is implied by using sales or firm size as the base.
BUSY = The number of audits signed off by the audit partner in year t. TENURE = The number of consecutive years that the incumbent audit partner has audited the
client by year t. AF_INDEXP = An indicator variable with the value of one if the audit firm is identified as an
industry expert for industry i in time t, ad zero otherwise. In line with prior studies (Balsam et al. 2003; Chin and Chi 2009) we measure market share using the number of clients audited by an audit firm within an industry. We then rank audit firms in each industry by their market share and identify an audit firm as an industry expert if they are the largest supplier in the industry.
SIZE = Total assets (in millions of euros) at the end of year t. Ln(SIZE) = The natural logarithm of total assets at the end of year t. ROA = Earnings before extraordinary items in year t divided by total assets at the end of
year t. LOSS = An indicator variable with a value of one if net income in year t is less than zero,
and zero otherwise. PBANK = Probability of bankruptcy as measured by adjusted Zmijewski (1984) score. LEV = Total liabilities divided by total assets. BM = The book value at the end of year t divided by market value at the end of year t. LCACCR = The prior year’s total current accruals (net income before extraordinary items +
depreciation and amortization – operating cash flows) divided by lagged total assets.
SUBSID = The number of subsidiaries in time t. Ln(SUBSID) = The natural logarithm of the number of subsidiaries in time t. CA = The current assets at the end of year t divided by total assets at the end of year t. CFFO = The cash flow from operations divided by total assets at the end of year t. SALESG = The percentage increase in sales over year t. AGE = The number of years since the firm was listed. Ln(AGE) = The natural logarithm of AGE DECEMBER = An indicator variable with a value of one if the fiscal year end is December 31st,
and zero otherwise. NAAF = The level of non-audit fees to audit fees in year t. NUM_PART = The number of audit partners who signed the audit report on behalf of their audit
firm in year t. TWO_BIG4 = An indicator variable with a value of one if the firm was audited by two Big 4 audit
firms in year t, and zero otherwise. ONE_BIG4 = An indicator variable with a value of one if the firm was audited by one Big 4 audit
firm and one non-Big 4 audit firm in year t, and zero otherwise.
41
Table 3 Correlation coefficients for audit quality sample (above diagonal) and fee sample (below diagonal).
Audit Quality Sample (n=1,251) Fe
e Sa
mpl
e (n
=1,4
85)
Variables
EV
C AP
_IN
DEXP
BU
SY
TE
NUR
E
AF
_IN
DEXP
LN
(SIZ
E)
RO
A LO
SS
PB
ANK
LE
V BM
LC
ACCR
CA
CFFO
EVC 0.068* 0.064* -0.079* 0.352* 0.108* -0.031 0.029 -0.016 0.009 -0.042 -0.011 -0.068* -0.014 AP_INDEXP 0.079* 0.403 0.085* 0.132* 0.138* 0.055 -0.025 -0.004 0.008 0.004 -0.023 -0.041 0.060 BUSY 0.093* 0.421* 0.128* 0.092* 0.083* 0.049 -0.026 0.027 0.050 0.027 -0.017 -0.082* 0.042 TENURE -0.072* 0.078* 0.119* -0.023 0.002 0.025 -0.008 0.013 -0.007 -0.025 0.012 0.028 0.025 AF_INDEXP 0.305* 0.135* 0.071* -0.031 -0.003 -0.028 0.037 -0.057 -0.083* 0.010 -0.028 -0.114* -0.003 Ln(SIZE) 0.111* 0.167* 0.094* 0.019 -0.008 0.064* -0.154* 0.117* 0.185* -0.245* -0.035 -0.451* 0.038 ROA -0.061 0.045* 0.059 0.020 -0.022 0.068* -0.596* -0.267* 0.031 0.278* 0.046 0.054 0.601* LOSS 0.046 -0.007 -0.025 -0.008 0.031 -0.154* -0.596* 0.262* 0.029 -0.082* -0.063 0.007 -0.310* PBANK -0.000 -0.012 0.014 0.009 -0.043 0.120* -0.283* 0.270* 0.738* -0.169* -0.122* 0.075* -0.130* LEV -0.011 0.002 0.041 0.001 -0.082* 0.202* 0.026* 0.029 0.729* -0.118* -0.047 0.050 0.068* BM -0.042 -0.002 0.041 -0.056 -0.013 -0.291* 0.225* -0.072* -0.173* -0.143* 0.055 0.203* 0.233* LCACCR -0.011 -0.012 -0.008 0.016 -0.018 -0.032 0.046 -0.053 -0.103* -0.481 0.076* -0.021 0.033 CA -0.068* -0.020 -0.070 0.019 -0.975 -0.045 0.035 0.169 0.081* 0.039 0.228* -0.016 -0.068* CFFO -0.041 0.057 0.058 0.018 -0.002 0.062 0.649* -0.322* -0.159* 0.064 0.159* 0.020 -0.102* SALESG 0.035 -0.016 0.014 -0.006 -0.007 -0.087* 0.060 -0.045 -0.070* -0.065 0.201* 0.001 0.059 -0.011 Ln(AGE) -0.006 0.024 0.018 0.082* 0.045 0.303* 0.032 -0.099* -0.089* -0.044 -0.181* -0.032 -0.153* 0.014 DECEMBER -0.027 0.020 -0.003 0.016 -0.083* 0.122* 0.103* -0.108* -0.059 -0.008 0.076* -0.027 -0.032 0.090 NAAF -0.016 -0.023 0.039 0.006 0.034 0.159* 0.046 -0.074* -0.057 -0.021 -0.058 0.001 -0.003 0.010 NUM_PART 0.013 0.096* 0.061* -0.040 -0.068* 0.449* 0.042 -0.051 0.058 0.113* -0.108* -0.037 -0.096* -0.012 TWO_BIG 0.073* 0.073* 0.051 -0.050 -0.046 0.361* 0.054 -0.026 -0.020 0.050 -0.005 -0.003 -0.101* 0.071* ONE_BIG 0.156* 0.020 0.072 -0.035 0.307* -0.117* -0.077* 0.072* 0.077* -0.013 0.007 0.014 0.007 -0.026 Ln(SUBSID) -0.049 -0.046 0.019 0.003 -0.048 0.145* 0.038 0.007 0.004 0.022 0.015 -0.047 -0.127* 0.106*
CONT’D
42
Audit Quality Sample (n=1,251)
Fee
Sam
ple
(n=1
,485
)
Variables
SA
LESG
Ln
(AGE
)
DE
CEM
BER
N
AAF
N
UM_P
ART
TW
O_BI
G
ON
E_BI
G
EVC 0.065 0.004 -0.038 -0.015 -0.006 0.069* 0.165* AP_INDEXP -0.001 0.034 0.039 -0.017 0.085 0.083* 0.009 BUSY 0.004 0.038 -0.003 0.038 0.033 0.051 0.067 TENURE 0.011 0.053 0.010 0.001 -0.045 -0.044 -0.027 AF_INDEXP -0.005 0.051 -0.072 0.036 -0.064 -0.035 0.289* Ln(SIZE) -0.061 0.313* 0.135* 0.156* 0.442* 0.358* -0.116* ROA 0.077* 0.040 0.113* 0.039 0.025 0.060 -0.076* LOSS -0.061 -0.109* -0.118* -0.067 -0.030 -0.037 0.070 PBANK -0.062 -0.087* -0.045 -0.046 0.072 -0.003 0.062 LEV -0.052 -0.046 0.004 -0.014 0.107* 0.064 -0.027 BM 0.165* -0.145* 0.069 -0.047 -0.122* 0.009 0.013 LCACCR -0.006 -0.025 -0.033 0.003 -0.038 0.005 0.006 CA 0.060 -0.152* -0.022 0.015 -0.105* -0.093* -0.003 CFFO -0.015 0.007 0.096* -0.001 -0.035 0.069 -0.020 SALESG -0.082 0.042 -0.046 -0.018 -0.039 0.042 Ln(AGE) -0.107* 0.019 0.163* 0.151* 0.147* -0.108* DECEMBER 0.041 0.027 -0.034 0.112* 0.094* -0.118* NAAF -0.054 0.155* -0.046 0.159* 0.063 0.055 NUM_PART -0.027 0.150* 0.123* 0.135* 0.199* -0.052 TWO_BIG -0.039 0.155* 0.095* 0.058 0.208* -0.623* ONE_BIG 0.042 -0.111* -0.113* 0.052 -0.055 -0.612* Ln(SUBSID) -0.040 -0.007 -0.032 0.040 0.038 0.068* -0.023
Note: The table presents Pearson correlation coefficients for both audit quality and fee samples. Variables are defined in Table 2. * denotes two-tailed significance at the 1% level.
43
Table 4: Regression results for audit quality: Abnormal working capital accruals
Panel A: Full Sample
(1) (2) (3) +/- Coeff. t-stat Coeff. t-stat. Coeff. t-stat. Intercept 0.169 3.93*** -0.076 -0.26 -0.075 -0.26 Log EVC - -0.003 -2.28*** -0.007 -2.54*** -0.007 -2.71*** AP_INDEXP ? 0.009 0.94 0.033 1.44 AP_INDEXP_2 ? 0.020 1.30 BUSY - 0.001 0.54 0.003 0.64 0.005 1.21 TENURE - -0.001 -0.13 0.003 1.35 0.003 1.49 AF_INDEXP ? -0.010 -1.42 -0.010 -0.76 -0.012 -0.96 Ln(SIZE) - -0.006 -3.02*** -0.007 -0.28 -0.006 -0.22 ROA - -0.183 -1.31* -0.345 -1.99** -0.336 -1.94** LOSS + 0.022 1.48* 0.010 0.53 0.011 0.58 PBANK + 0.119 2.06** 0.178 1.78* 0.182 1.77* LEV + -0.022 -0.97 -0.086 -0.95 -0.093 -1.02 BM - -0.010 -2.29*** -0.004 -0.46 -0.002 -0.25 LCACCR - -0.155 -4.27*** -0.134 -2.36** -0.135 -2.35** CA + -0.034 -1.66* 0.074 0.85 0.075 0.85 CFFO - 0.012 0.17 0.075 0.85 0.068 0.78 SALESG + 0.089 6.19*** 0.087 4.63*** 0.087 4.57*** Ln(AGE) - 0.001 0.12 0.020 0.30 0.026 0.39 DECEMBER + 0.003 0.36 0.007 0.19 0.004 0.11 NAAF - -0.012 -0.53 -0.009 -0.17 -0.009 -0.16 NUM_PART ? 0.004 0.55 0.006 0.45 0.008 0.67 TWO_BIG4 ? 0.011 1.22 0.387 1.85* 0.372 1.77* ONE_BIG4 ? 0.180 2.28** 0.369 1.74* 0.353 1.66* Year F.E. Yes Yes Yes Industry F.E. Yes No No Client Firm F.E No Yes Yes Audit Office F.E. No Yes Yes Audit Firm F.E. No Yes Yes Audit Partner F.E No Yes Yes N 1,246 1,246 1,246 Adjusted R2 25.06% 38.45% 38.05%
44
Panel B: One_Big4 Sample Note: The dependent variable is absolute abnormal working capital accruals divided by total assets (ABS_AWCA). AP_INDEXP_2 is an indicator variable with the value of one if the audit partner is an industry expert and zero otherwise. An audit partner is identified as an industry expert if they a) audit at least one of the top two largest firms (based on total assets) within the industry in the fiscal year, and b) audit at least two additional clients belonging to that industry. The t-statistics are adjusted for firm and year clustering effects ***, **, *: Significant at 1%, 5%, and 10% level. P-values are one-tailed for signed expectations and two-tailed for unsigned expectations. All other variables are defined in Table 2. After calculating DFBETAS, we exclude any outliers.
(1) (2) +/- Coeff. t-stat. Coeff. t-stat. Intercept 0.877 0.68 0.830 0.65 Log EVC - -0.009 -1.83* -0.010 -2.11** AP_INDEXP ? 0.047 1.17 AP_INDEXP_2 ? 0.013 0.43 BUSY - 0.012 1.31 0.016 1.85* TENURE - 0.010 1.49 0.010 1.46 AF_INDEXP ? -0.045 -1.84* -0.045 -1.89* Ln(SIZE) - -0.093 -0.87 -0.088 -0.80 ROA - -0.141 -0.34 -0.130 -0.31 LOSS + 0.025 0.43 0.025 0.42 PBANK + 0.116 0.75 0.121 0.77 LEV + 0.008 0.05 0.006 0.04 BM - -0.022 -0.70 -0.017 -0.54 LCACCR - -0.136 -1.23 -0.141 -1.25 CA + 0.056 0.32 0.058 0.33 CFFO - 0.032 0.20 0.023 0.14 SALESG + 0.100 3.99*** 0.100 3.91*** Ln(AGE) - -0.072 -0.56 -0.062 -0.51 DECEMBER + -0.011 -0.21 -0.010 -0.21 NAAF - -0.157 -0.71 -0.152 -0.67 NUM_PART ? -0.003 -0.12 0.001 0.05 Year F.E. Yes Yes Client Firm F.E Yes Yes Audit Office F.E. Yes Yes Audit Firm F.E. Yes Yes Audit Partner F.E Yes Yes N 595 595 Adjusted R2 23.89% 23.38%
45
Table 5: Regression results for audit quality: Likelihood of a firm reporting a small
earnings increase. Panel A: Full sample.
(1) (2) +/- Coeff. Chi-Sqr. Coeff. Chi-Sqr. Intercept -1.487 6.31*** -1.434 6.04** Log EVC - -0.065 6.78*** -0.064 6.46*** AP_INDEXP ? -0.124 0.42 AP_INDEXP_2 ? -0.353 0.77 BUSY - 0.033 0.69 0.028 0.59 TENURE - 0.010 0.28 0.008 0.20 AF_INDEXP ? 0.118 0.68 0.112 0.64 Ln(SIZE) - 0.104 6.48*** 0.105 6.64*** ROA - 1.730 0.98 1.679 0.6793 LOSS + -2.323 44.68*** -2.330 44.70*** PBANK + -0.273 0.20 -0.289 0.23 LEV + -0.037 0.01 -0.038 0.01 BM - -0.138 2.14 -0.133 2.01 LCACCR - -0.242 1.17 -0.245 1.20 CA + -0.700 3.36* -0.715 3.56* CFFO - -1.304 0.96 -1.313 0.98 SALESG + 0.017 0.01 0.036 0.03 Ln(AGE) - 0.189 5.00** 0.186 4.74** DECEMBER + -0.087 0.26 -0.097 0.31 NAAF - -0.789 1.92 -0.763 1.83 NUM_PART ? 0.220 2.45 0.217 2.41 TWO_BIG4 ? -0.596 10.46*** -0.602 10.75*** ONE_BIG4 ? -0.092 0.32 -0.100 0.34 Year F.E. Yes Yes Client Firm F.E Yes Yes Audit Office F.E. Yes Yes Audit Firm F.E. Yes Yes Audit Partner F.E Yes Yes N 1,242 1,242 Pseudo R2 20.76% 20.80%
46
Panel B: One_Big4 Sample. (1) (2) +/- Coeff. Chi-Sqr. Coeff. Chi-Sqr. Intercept -2.710 8.04*** -2.673 8.15*** Log EVC - -0.113 3.94** -0.114 4.01** AP_INDEXP ? -0.080 0.09 AP_INDEXP_2 ? 0.060 0.01 BUSY - 0.026 0.25 0.019 0.15 TENURE - 0.054 2.79* 0.054 2.86* AF_INDEXP ? 0.509 6.26** 0.502 6.26** Ln(SIZE) - 0.091 2.43 0.089 2.33 ROA - 2.719 1.77 2.710 1.75 LOSS + -2.229 27.52*** -2.230 27.55*** PBANK + -2.256 6.09** -2.254 6.10** LEV + 1.784 3.29* 1.786 3.32* BM - 0.069 0.20 0.070 0.20 LCACCR - -0.361 2.52 -0.360 2.52 CA + -1.014 2.56 -1.021 2.61 CFFO - -2.789 2.78* -2.806 2.81* SALESG + -0.263 0.63 -0.274 0.66 Ln(AGE) - 0.255 3.77* 0.254 3.70** DECEMBER + -0.056 0.05 -0.056 0.05 NAAF - -2.648 7.54*** -2.620 7.59*** NUM_PART ? -0.049 0.05 -0.049 0.05 Year F.E. Yes Yes Client Firm F.E Yes Yes Audit Office F.E. Yes Yes Audit Firm F.E. Yes Yes Audit Partner F.E
Yes Yes
N 595 595 Pseudo R2 24.39% 24.38% Note: The dependent variable is INCREASE. AP_INDEXP_2 is an indicator variable with the value of one if the audit partner is an industry expert and zero otherwise. An audit partner is identified as an industry expert if they a) audit at least one of the top two largest firms (based on total assets) within the industry in the fiscal year, and b) audit at least two additional clients belonging to that industry. The t-statistics are adjusted for firm and year clustering effects ***, **, *: Significant at 1%, 5%, and 10% level. P-values are one-tailed for signed expectations and two-tailed for unsigned expectations. All other variables are defined in Table 2. After calculating DFBETAS, we exclude any outliers.
47
Table 6: Regression results for audit fees Panel A: Full sample. (1) (2) (3) +/- Coeff. t-stat. Coeff. t-stat. Coeff. t-stat. Intercept 7.118 39.33*** 10.415 14.28*** 10.438 14.28*** Log EVC + 0.012 1.83** 0.011 1.85** 0.011 1.72* * AP_INDEXP + -0.043 -0.85 0.035 0.85 AP_INDEXP_2 + 0.097 0.85 BUSY ? 0.0264 2.79*** -0.012 -1.02 -0.011 -0.93 TENURE ? 0.0063 1.16 -0.001 -0.04 -0.001 -0.04 AF_INDEXP + 0.0273 0.71 0.012 0.25 0.012 0.25 Ln(SIZE) + 0.6522 48.25*** 0.434 6.47*** 0.434 6.46*** ROA - -0.5363 -2.08** -0.353 -1.67* -0.348 -1.67* LOSS + -0.0022 -0.04 -0.024 -0.59 -0.023 -0.57 PBANK + 0.7711 5.72*** 0.316 2.14** 0.317 2.15** LEV + -0.1709 -2.39*** -0.079 -0.80 -0.075 -0.76 BM - -0.0637 -2.95*** -0.050 -2.43** -0.050 -2.42** CA + 0.0271 2.45*** -0.093 -0.40 -0.088 -0.38 CFFO + -0.0586 -0.23 0.090 0.47 0.092 0.49 SALESG + -0.0446 -0.87 -0.026 -0.78 -0.027 -0.82 Ln(AGE) - 0.0008 0.03 -0.123 -1.02 -0.132 -1.09 Ln(SUBSID) + 0.1000 5.44*** 0.145 2.92*** 0.148 2.99*** DECEMBER + -0.1217 -2.07*** 0.024 0.35 0.019 0.28 NAAF - 0.2195 1.62 -0.222 -1.54 -0.224 -1.55 NUM_PART + 0.2659 6.46*** 0.022 0.52 0.024 0.56 TWO_BIG4 + 0.0930 1.36 2.250 4.66*** 2.270 4.71*** ONE_BIG4 + 0.2886 5.08*** 2.513 5.12*** 2.535 5.16*** IFRS + 0.0191 0.21 0.005 0.13 0.002 0.04 Year F.E. Yes Yes Yes Industry F.E. Yes No No Client Firm F.E No Yes Yes Audit Office F.E. No Yes Yes Audit Firm F.E. No Yes Yes Audit Partner F.E No Yes Yes N 1,478 1,478 1,478 Adjusted R2 88.01% 97.63% 97.63%
48
Panel B: One_Big4 Sample.
(1) (2) +/- Coeff. t-stat. Coeff. t-stat. Intercept 11.249 10.67*** 11.714 11.18*** Log EVC + 0.015 1.96** 0.013 1.76* AP_INDEXP + 0.028 0.48 AP_INDEXP_2 + 0.267 2.58*** BUSY ? 0.002 0.12 -0.001 -0.05 TENURE ? -0.010 -1.06 -0.011 -1.19 AF_INDEXP + -0.006 -0.10 -0.004 -0.08 Ln(SIZE) + 0.504 6.01*** 0.508 6.11*** ROA - -0.450 -1.81* -0.488 -1.98** LOSS + -0.053 -0.90 -0.049 -0.83 PBANK + 0.220 1.24 0.227 1.29 LEV + 0.375 1.25 0.348 1.17 BM - 0.026 0.64 0.026 0.66 CA + -0.313 -1.14 -0.288 -1.08 CFFO + 0.398 1.62 0.429 1.87* SALESG + -0.026 -0.71 -0.032 -0.89 Ln(AGE) - -0.231 -1.27 -0.345 -1.91* Ln(SUBSID) + 0.147 2.22** 0.159 2.45** DECEMBER + 0.100 1.25 0.094 1.17 NAAF - -0.174 -0.91 -0.179 -0.95 NUM_PART + 0.001 0.01 0.010 0.15 IFRS + 0.001 0.02 -0.011 -0.24 Year F.E. Yes Yes Client Firm F.E Yes Yes Audit Office F.E. Yes Yes Audit Firm F.E. Yes Yes Audit Partner F.E Yes Yes N 705 705 Adjusted R2 94.86% 97.55%
Note: The dependent variable is Ln(A_FEE).. AP_INDEXP_2 is an indicator variable with the value of one if the audit partner is an industry expert and zero otherwise. An audit partner is identified as an industry expert if they a) audit at least one of the top two largest firms (based on total assets) within the industry in the fiscal year, and b) audit at least two additional clients belonging to that industry. The t-statistics are adjusted for firm and year clustering effects ***, **, *: Significant at 1%, 5%, and 10% level. P-values are one-tailed for signed expectations and two-tailed for unsigned expectations. All other variables are defined in Table 2. After calculating DFBETAS, we exclude any outliers.
49
Table 7: Determinants of an audit partner’s social capital: EVC score
+/- Coeff. t-stat. Intercept -9.388 -12.96*** AP_INDEXP + -0.049 -0.23 BUSY + 0.252 2.28** TENURE - -0.068 -3.01*** AF_INDEXP + 1.069 7.39*** NAAF - -0.409 -0.96 ECON_SIG - -0.658 -2.20** TWO_BIG4 + 3.485 15.39*** ONE_BIG4 + 3.176 15.26*** Ln(SIZE) + 0.262 5.56*** ROA ? -1.269 -1.55 PBANK ? 0.281 0.57 LEV ? 0.027 0.09 BM ? -0.087 -1.05 Ln(AGE) ? -0.007 -0.06 NUM_PART ? -0.097 -0.61 Paris Office + 0.404 1.80* Year F.E. Yes Industry F.E Yes N 1,493 Adjusted R2 70.54%
Note: The dependent variable is the natural logarithm of EVC. ECON_SIG is the level of audit fees from the client firm divided by total audit fees the partner receives from all other audits conducted in the fiscal year t. The t-statistics are adjusted for audit partner and year clustering effects ***, **, *: Significant at 1%, 5%, and 10% level. P-values are one-tailed for signed expectations and two-tailed for unsigned expectations. All other variables are defined in Table 2. After calculating DFBETAS, we exclude any outliers.