Do Joint Audits Improve Audit Quality? Evidence from Voluntary Joint Audits
ABSTRACT This study examines whether the decision to voluntarily (i.e. without a statutory obligation) employ two audit firms to conduct a joint audit is related to audit quality. We use separate samples and empirical designs for public and privately held companies in Sweden, where a sufficient number of companies have a joint audit on a voluntary basis. Our empirical findings suggest that companies opting to employ joint audits have a higher degree of earnings conservatism, lower abnormal accruals, better credit ratings and lower perceived risk of becoming insolvent within the next year than other firms. These findings are robust to the use of a propensity score matching technique to control for the differences in client characteristics between firms that employ joint audits and those that use single Big 4 auditors (i.e. auditor self-selection). We also find evidence that the choice of a joint audit is associated with substantial increases in the fees paid by the client firm, suggesting a higher perceived level of quality. Collectively, our analyses support the view that voluntary joint audits are positively associated with audit quality in a relatively low litigious setting both for public and private firms.
- SourceAvailable from: wlkc.gdqy.edu.cn
Article: Auditor size and audit quality[show abstract] [hide abstract]
ABSTRACT: Regulators and small audit firms allege that audit firm size does not affect audit quality and therefore should be irrelevant in the selection of an auditor. Contrary to this view, the current paper argues that audit quality is not independent of audit firm size, even when auditors initially possesses identical technological capabilities. In particular, when incumbent auditors earn client-specific quasi-rents, auditors with a greater number of clients have ‘more to lose’ by failing to report a discovered breach in a particular client's records. This collateral aspect increases the audit quality supplied by larger audit firms. The implications for some recent recommendations of the AICPA Special Committee on Small and Medium Sized Firms are developed.Journal of Accounting and Economics. 01/1981;
- [show abstract] [hide abstract]
ABSTRACT: In a competitive market, audit prices can vary if the clients believe that the quality of audits varies. Previous research links auditor independence, a key element of audit quality, to auditor size and consequently suggests a positive association between audit quality and auditor size. Moreover, by using the dichotomy approach (Big Five/non-Big Five), numerous studies in many countries have found that the largest audit firms with international reputations earn fee premiums due to their perceived higher quality. Little is known, however, about pricing differences arising from product differentiation among small audit firms. This study examines the relation between auditor size and audit prices by using the data on hourly billing rates and the auditor characteristics from 103 small Finnish audit firms. This study documents, after controlling for the auditor's technical capability, the positive association between auditor size and audit pricing. The results suggest that both size and technical capability have a positive impact on auditor remuneration, implying that product differentiation also takes place among these small audit firms.European Accounting Review 02/2004; 13(3):541-560. · 1.15 Impact Factor
- Econometrica. 04/2001; 48:817-830.
Do Joint Audits Improve Audit
Quality? Evidence from Voluntary
MIKKO ZERNI∗, ELINA HAAPAMA¨KI∗, TUUKKA JA¨RVINEN∗
and LASSE NIEMI∗∗
∗Department of Accounting and Finance, University of Vaasa, PO Box 700, FI-65101, Finland;
∗∗Department of Accounting, Aalto University School of Economics, Finland
statutory obligation) employ two audit firms to conduct a joint audit is related to audit
quality. We use separate samples and empirical designs for public and privately held
companies in Sweden, where a sufficient number of companies have a joint audit on a
voluntary basis. Our empirical findings suggest that companies opting to employ joint
audits have a higher degree of earnings conservatism, lower abnormal accruals, better
credit ratings and lower perceived risk of becoming insolvent within the next year than
other firms. These findings are robust to the use of a propensity score matching technique
to control for the differences in client characteristics between firms that employ joint
audits and those that use single Big 4 auditors (i.e. auditor self-selection). We also find
evidence that the choice of a joint audit is associated with substantial increases in the fees
paid by the client firm, suggesting a higher perceived level of quality. Collectively, our
analyses support the view that voluntary joint audits are positively associated with audit
quality in a relatively low litigious setting both for public and private firms.
This study examines whether the decision to voluntarily (i.e. without a
In recent years, there has been increased concern regarding auditor independence,
a necessity for audit quality. Calls for more regulation and governance to improve
auditor independence have been made, with the ultimate goal of restoring trust in
the quality of financial statement audits (Eilifsen and Willekens, 2008). A recent
European Accounting Review
iFirst Article, 1–35, 2012
iFirst Article, 1–35, 2012
Correspondence Address: Mikko Zerni, Department of Accounting and Finance, University of Vaasa,
PO Box 700, FI-65101, Finland. Email: ZERMP@uwasa.fi
Paper accepted by Laurence van Lent.
European Accounting Review
0963-8180 Print/1468-4497 Online/12/000001–35 # 2012 European Accounting Association
Published by Routledge Journals, Taylor & Francis Ltd on behalf of the EAA.
example is the Green Paper ‘Audit Policy: Lessons from the Crisis’ of the Euro-
pean Commission, which is aimed at stimulating discussion on how to improve
audit regulation to increase audit quality (and audit market competition).1The
Green Paper proposes several regulatory actions as possible remedies for the
alleged lack of market trust in auditor independence, such as joint audits,
auditor rotation, audit committees, and restrictions on the services that auditors
are allowed to provide to their clients. The focus of this study is on joint audits.
In the Green Paper, in some letters commenting on it, and in the financial press,
the idea of adopting joint audits has been raised as a potential way to enhance
audit quality and to stimulate audit market competition (see, for instance,
Andre ´ et al., 2009, pp. 5-6; European Commission, 2010, pp. 15–16; Financial
Times, 2007; Herbinet, 2007a; Kauppalehti, 2011a, 2011b; Mazars, 2010). In a
joint audit, two different audit firms jointly form an opinion of a client’s financial
statements. Both audit firms are also jointly liable for the issued audit opinion.
Proponents of joint audits argue that joint audits have the potential to safeguard
auditor independence. For example, the audit firm Mazars conjectures that ‘[joint
audits are] an advanced form of governance of an audit enhancing in particular
independence and the auditors’ ability to stand their own ground in the event
of a difference of view with the audited entity’ (Mazars, 2010, Response to Euro-
pean Commission’s Green Paper, ‘Audit Policy: Lessons from the Crisis’, p. 31).
Opponents of mandating joint audits argue that joint audits increase the cost of
auditing with little effect on audit quality. The issue of adopting joint audits is
highly controversial. For instance, in their response letters to the Green Paper,
each of the Big 4 audit firms opposed mandatory joint audits while smaller
audit firms were generally in favour of such a requirement.
This study contributes to the audit literature by investigating the timely topic of
joint audits, addressing in particular the alleged benefits and costs associated with
joint audits.2More specifically, we examine whether joint audits are associated
with audit quality and/or audit fees in a voluntary setting, which is fundamentally
different from previous studies related to a mandatory setting (e.g. Andre ´ et al.,
2009; Francis et al., 2009; Gonthier-Besacier and Schatt, 2007; Piot, 2007;
Thinggaard and Kiertzner 2008). Our study relates to Sweden, where a sufficient
number of companies have a joint audit on a voluntary basis. The use of a volun-
tary setting is beneficial because it allows for the determination of the potential
effects of joint audits per se on audit quality and fees. Our study also contributes
to the recent stream of audit studies examining the effects of country-level insti-
tutional factors on auditing (e.g. Choi et al., 2008, 2009; Francis and Wang,
2008), as we study the potential effect of a voluntary joint audit on audit
quality in a relatively low litigious setting (Choi et al., 2008; Wingate, 1997).
The results from the empirical tests for public and privately held companies
separately suggest that companies employing voluntary joint audits have a
higher degree of earnings conservatism, lower abnormal accruals, better credit
ratings and lower risk forecasts of becoming insolvent within the next year
than other firms.3For private firms, with a sufficiently large initial sample
2 M. Zerni et al.
(over 65,000 unique firms), we are able to address auditor self-selection by using
a propensity score matching technique and find that our results are robust in this
respect. We attribute the observed better credit ratings and more favourable risk
forecasts among privately held companies to a higher (perceived) assurance and/
or insurance value of joint audits. Analyses of audit fees among the sample of
publicly listed firms show that joint audits are associated with considerably
higher fees. We interpret this fee premium for joint audits as an indication of
additional value for clients arising from the voluntary decision to hire two audi-
tors instead of one. Collectively, the empirical findings of our study support the
view that joint audits are positively associated with audit quality.
The structure of the paper is as follows. The next section reviews the relevant
literature and develops the hypotheses. Section 3 describes the data and the
matching procedure used for the sample of privately held firms. Section 4
describes the tests of earnings conservatism and reports the results. Sections 5
and 6 describe the tests used for abnormal accruals and credit risk, respectively.
Section 7 reports our results for audit fee tests, and Section 8 concludes the study.
2.Literature Review and Hypotheses Development
2.1. Joint Audits and Actual Audit Quality
Francis et al. (2009) examine auditor-pair choices and their effects on audit
quality in France, where joint audits are mandatory, and report evidence consist-
ent with an agency-driven demand to employ higher quality auditor pairs.4They
also find that client firms employing higher quality auditor pairs have smaller
abnormal income-increasing accruals than the firms that do not use Big 4 auditors
(i.e. those that use two non-Big 4 auditors) and that this effect is strongest when
client firms use two Big 4 auditors.
At least three arguments suggest that joint audits would contribute positively to
audit quality, thereby giving credibility to financial statements. First, by appoint-
ing two different audit firms, the client firm allows audit firms to rotate, safe-
guarding auditor independence, but also retains the remaining auditor’s
knowledge and understanding of the client’s business operations, thereby avoid-
ing the potential downside of auditor rotation of a discontinuity in competence
(Carcello and Nagy, 2004). Second, the threat to auditor independence due to
economic bonding is likely to be a less significant issue with the joint audit
approach than it is with the single auditor approach. This is simply because, in
joint audits, the audit fees and lucrative consulting fees are distributed between
two different audit firms (i.e. there are lower fees at stake).5Consequently, the
two different audit firms may take a stronger stand against pressure from the man-
agers and/or controlling owners and report their opinions on the clients’ accounts
more independently (Mazars, 2010; Zerni et al., 2010). Finally, it is ex ante less
likely that both Big 4 firms (or one Big 4 firm and one non-Big 4 firm) would sim-
ultaneously acquiesce to client pressure and not report the discovered breach(s)
Do Joint Audits Improve Audit Quality?3
than that a single (Big 4) audit firm would do so. In essence, non-reporting of the
discovered breach(s) and willing to sign-off on financial statements that signifi-
cantly depart from GAAP would require three-party collusion.6In a game-theor-
etic sense (Kargupta et al., 2007), the expected penalties of being caught for
substandard reporting are more likely to exceed its expected benefits for at
least one of the auditors in a joint audit setting compared with auditors in
single auditor engagements, increasing the likelihood of truthful reporting.
Opponents of mandatory joint audits present two main arguments that joint
audits do not increase audit quality. First, joint audits may suffer from a potential
‘free-rider problem’. This problem may occur if one of the auditors attempts to
‘shirk’ and rely on the other auditor’s effort during the audit. Second, it may
be difficult for two competitive audit firms to cooperate closely while conducting
the audit, resulting in insufficient information exchange. Competition between
auditors aiming to acquire a larger share of the business in the upcoming years
may hinder cooperation and even compromise audit quality because of insuffi-
cient information exchange. Moreover, accounting standards containing con-
siderable discretion can make cooperation difficult and lead to conflicts
(Neveling 2007). Thinggaard and Kiertzner (2008) report that after the abolition
of the mandatory two-auditor system in Denmark in 2004, 15 of the 63 (23.8%)
companies investigated retained two auditors in the next year.
In sum, the existing theories and models on audit production (e.g. Simunic,
1980) and quality (DeAngelo, 1981a, 1981b) examine single audit firm settings
and therefore are not applicable to settings in which two separate firms share
the work and legal liability of an audit. In addition, the few existing empirical
studies on joint audits relate to a mandatory setting. Thus, whether joint audits
improve earnings quality as compared with audits conducted by single audit
firms is an empirical question. Given the absence of convincing theories and
the above-discussed opposing views on the impacts of joint audits on audit
quality, we state our first hypothesis in null form:
H1: Joint audits are not significantly associated with audit quality.
2.2. Joint Audits and Perceived Audit Quality
According to financial theory, a higher credibility of financial information
reduces the cost of capital by reducing investors’ information risk (Botosan,
1997; Coles and Lowenstein, 1988; Jensen and Meckling, 1976; Lambert
et al., 2007). Datta et al. (1999) report evidence that firms lower their interest
rates by developing their reputations. The appointment of a high-quality, high-
reputation auditor(s) may thus help more efficiently to resolve contracting pro-
blems by reducing information risk about borrowers (Jensen and Meckling,
1976; Watts and Zimmerman, 1986). The increased credibility of the firms’ finan-
cial reporting would justify lenders and credit rating agencies considering auditor
choice when pricing debt contracts (Pittman and Fortin, 2004). Given that the
4 M. Zerni et al.
appointment of two separate audit firms instead of one signals ‘good news’ to the
market (Teoh and Wong, 1993; Titman and Trueman, 1986) about the client
firm’s financial reporting quality, we would expect joint audits to be associated
with lower perceived credit risk.
There are at least two complementary explanations as to why joint audits may
be perceived as of higher quality/value compared with single auditor audits.
First, as noted before, lenders and other users of financial statements consider
the probability that both auditors simultaneously acquiesce to client pressure to
be lower than the probability that either of them do so alone (i.e. higher assurance
value). The second explanation, the ‘insurance hypothesis’ (e.g. Wallace 1980),
predicts that audits are expected to add value by providing a type of implicit
insurance to investors. In the case of audit failure, investors can sue auditors to
recover their losses if an investment or credit loss results from misstated financial
statements. The auditor is deemed to be a ‘deep pocket’ because the audit firm
carries malpractice insurance and is often the only solvent defendant in a
lawsuit.7Moreover, the larger the audit firm, the larger the insurance value or
‘bond of wealth’ from which to recover losses and, hence, the higher the value
of an audit (Dye, 1993). In the case of joint audits, the two audit firms together
by definition have deeper pockets (i.e. higher insurance value) than either of
them do alone.8The deeper pockets of the two audit firms should be reflected
in a lower credit risk for investors providing funding for the audit client.
Consistent with the above viewpoints, Zerni et al. (2010) document both stat-
istically and economically significant equity discounts due to the entrenchment
problem9for the clients of non-Big 4 auditors (i.e. the largest discount) and
Big 4 auditors (i.e. the second largest discount) but do not find a statistically sig-
nificant discount for the firms that employ joint audits. These findings are consist-
ent with the above explanations of the higher assurance and/or insurance value of
joint audits compared with single auditor audits. Our next two hypotheses add to
Zerni et al.’s (2010) findings by examining whether joint audits are associated
with a lower level of perceived credit risk and thus facilitate access to debt
capital. Formally, these hypotheses can be stated as follows:
H2a: Joint audits are associated with better credit ratings.
H2b: Joint audits are associated with more favourable risk forecasts of
3.1. Public Company Sample
For listed Swedish companies, we obtain data from the Worldscope database and
audit fees from the annual reports available on the companies’ homepages. This
results in an initial sample of 1667 firm-year observations. Joint audits are
Do Joint Audits Improve Audit Quality?5
signed by two engagement partners representing different audit firms. We elimin-
ate 91 observations by excluding the firms in the finance sector (SIC codes 6000-
6500)because oftheirdistinct characteristics andbecause thesefirmsare required
to employ joint audits by law. The number of (joint audit) observations varies
between different tests due to use of different subsamples, and limitations
arising from calculations of dependent and explanatory variables.
A professional association for authorised public accountants, approved public
accountants, and other highly qualified professionals in the accountancy sector in
Sweden (FAR) recommends the design in Figure 1 for the audit-fee-related foot-
note in the annual report.
The design recommended in the footnote reflects the Swedish tradition of fre-
quently employing joint audits. Among all Swedish non-financial publicly listed
firms the proportion of voluntary joint audits has been roughly 10%. For joint
audits, we aggregate the audit fees into single amounts because both audit
firms are jointly liable for the issued audit opinion.
3.2. Propensity Score Matched Private Company Sample
For privately held clients, we obtain most of the data from UC AB, a leading
Swedish business and credit information agency owned by the major Swedish
Figure 1. Recommended scheme for reporting audit and non-audit fees in footnotes for
6 M. Zerni et al.
banks.10In addition, we use data on firms where individual auditors are employed
to identify joint audits among remaining firms. These data are retrieved from
Revisorsna ¨mnden (The Supervisory Board of Public Accountants), a governmen-
tal authority under the Ministry of Justice that handles all matters related to char-
tered accountants. Because only a few client firms hired joint audit pairs
composed of two non-Big 4 companies, we excluded these observations to sim-
plify our empirical models, reducing the number of interaction variables needed.
After excluding the finance sector and the observations with missing values
needed for our matching model, we identified 973 joint audit observations
from 191 different firms. For each of the above 973 treatment observations we
attempt to find as closely matched pair as possible through a propensity score
matching technique. Despite the number of joint audit observations varying
across different matched pair tests, the proportion of joint audits in each test is
always exactly 50% (please, refer also Note 15).
3.2.1. Matching procedure and propensity score matching model
A sufficiently large initial sample size of privately held firms (over 65,000 unique
the differences in client characteristics between the treatment group (i.e. the firms
ence et al., 2011; Lennox et al., 2012; Rosenbaum and Rubin, 1983). Propensity
going treatment, which in our case is the probability of employing a joint audit.
Before estimating the matching model, we exclude all of the single audits con-
ment and control groups. We then match, without replacement, each Big 4 auditor
client with a joint audit client that had the closest predicted value (according to the
We use the following logit model to estimate the probability of employing joint
audits on the basis of prior auditor selection studies (e.g. DeFond, 1992; Francis
and Wilson, 1988; Johnson and Lys, 1990; Lennox, 2005) (firm and time sub-
PROB(JAUDIT) = a + b1SIZE + b2DTA + b3CONTROL
+ b4GROUP + b5LOGAGE + b6CASH + b7LOSS
+ b8CA CL + b9ROA + fixed effects
where the dependent variable JAUDIT is an indicator variable for the joint audit
decision. Explanatory variables are defined as follows: SIZE is the natural logar-
ithm of total assets; DTA is the ratio of debt to total assets; CONTROL is an
Do Joint Audits Improve Audit Quality?7
indicator variable for the presence of a controlling shareholder with at least a 25%
stake of voting power; GROUP is an indicator variable for group affiliation;
LOGAGE is the natural logarithm of the client age in years; CASH is the ratio
of cash and cash equivalents to total assets; LOSS is an indicator variable for
accounting losses; CA_CL is the ratio of current assets to current liabilities;
and ROA is the return on assets. We also add indicator variables for the economic
sectors and different years.14
Using the 1% upper bound for the difference between propensity scores, we
were able to find a match for 597 joint audit observations from 135 different
firms between 2001 and 2007 resulting in a matched sample of 1194 firm-
years. However, because of missing values for variables used in our empirical
models, the final sample sizes in our tests are reduced to between 1160 and
848 firm-year observations.15
3.2.2. Results of estimating the propensity score matching model
Table 1 presents the results of the estimation model (1). As can be seen from
column (1) of Table 1, all of our chosen explanatory variables are significant pre-
dictors of the joint audit decision. The estimated magnitudes of the coefficients
indicate that client size, company age, group affiliation and the presence of a con-
trolling shareholder appear to be the most focal positive determinants of joint
audit selection. The return on assets, the leverage and the ratio of current
assets to total assets are all estimated to be significantly negative, but the magni-
tudes of their coefficients are considerably smaller compared with the above-
mentioned four variables. Finally, we find that accounting losses and liquidity
have positive impacts on the likelihood of employing a joint audit.
The estimated results reported in column (2) of Table 1 indicate that the match-
ing procedure balances the differences regarding all dimensions used in the pro-
pensity matching approach. The likelihood ratio test cannot reject the global
null hypothesis that all of the coefficients are zero (p-value ¼ 0.935). Hence,
given ‘perfect’ matching, the remaining differences between the two groups of
firms should reflect only the treatment effect, and a simple univariate t-test of
the differences in means should be sufficient to estimate the treatment effects
(Dehejia and Wahba, 2002; Heckman et al., 1997; Zhao, 2004). However, we
in client characteristics between the treatment and control groups.
4.Multivariate Analysis: Earnings Conservatism
4.1. Earnings Conservatism Model Specification for Publicly Listed
Our first test uses the earnings conservatism framework of Basu (1997) to deter-
mine whether there are differences in timely loss recognition between the clients
of dual and single auditors. In this framework, positive annual stock returns are
8M. Zerni et al.
used to indicate ‘good news’ whereas negative annual stock returns are assumed
to reflect ‘bad news’ in the current fiscal year. The premise of earnings conserva-
tism is that losses are recognised more quickly than gains, which are usually
deferred until they are realised. Accordingly, earnings conservatism exists if
the contemporaneous accounting earnings recognise bad news more quickly
than good news. Conservative earnings reduce the level of information asymme-
try and make earnings more useful for contracting purposes (e.g. LaFond and
Watts, 2008; Watts, 2003). By estimating the following model that builds on
Basu (1997), we test whether a firm’s choice of auditor affects the degree of con-
EARN = b0+ b1R + b2DR + b3DR × R + b4JOINT + b5JOINT
× R + b6JOINT × DR + b7JOINT × DR × R + b8BIG + b9BIG
× R + b10BIG × DR + b11BIG × DR × R + b12LMV + b13P/B
+ b14DTA + fixed effects + 1
Table 1. Logit model for the joint audit decision used in the matching process.
Matching Process After Matching
Variable Coef.Prob. Coef.Prob.
Annual fixed effects?
Economic sector fixed effects?
Likelihood ratio, x2
–2 Log Likelihood
N (# joint audits)
SIZE is the natural logarithm of total assets in thousands of Swedish kronor; DTA is the ratio of debt to
total assets; CONTROL is a dummy variable with a value of one if the firm has a controlling
shareholder, otherwise zero; GROUP is a dummy variable with a value of one if the firm is affiliated
with a group of companies, otherwise zero; LOGAGE is the natural logarithm of firm age in years;
CASH is the ratio of cash and cash equivalents to total assets; LOSS is a dummy variable with a value
of one if earnings are negative, otherwise zero; CA_CL is the ratio of current assets to current
liabilities; ROA is the interest-adjusted return on opening total assets.
Do Joint Audits Improve Audit Quality?9
where R is the buy-and-hold annual stock return (including dividends) and DR is
a dummy variable with a value of one if the return is negative, otherwise zero. We
add controls for firm size (LMV), leverage (DTA), and growth (P/B). However,
we omit the interactions with LMV, P/B and DTA to avoid multicollinearity pro-
blems. We include the year and industry-fixed effects and estimate the parameters
by adjusting the standard errors with respect to heteroscedasticity and within-firm
clustering (Rogers, 1993).
The standard Basu (1997) model includes the first four parameters in equation
(2). In the Basu framework, a positive sign on the coefficient on the interaction
variable DR∗R (b3) implies that accounting earnings recognise bad news
(timely loss recognition) more quickly than good news.16The primary coeffi-
cients of interest are the interaction variables JOINT∗DR∗R (b7) and BIG∗
employ a joint audit or a single Big 4 audit firm on earnings conservatism relative
to clients of single non-Big 4 audit firms, respectively.
∗R (b11), which measure the incremental effects of the decisions to
4.2. Specification of an Earnings Conservatism Model for Privately Held
Because private companies do not have stock returns, we rely on Ball and Shiva-
kumar’s (2005) metrics to examine the timeliness of loss recognition among pri-
vately held clients. Ball and Shivakumar’s (2005) approach builds on Dechow
et al.’s (1998) work and exploits the likelihood that timely loss recognition
occurs through accounting accruals. In particular, this approach hypothesises that
economic losses are more likely to be recognised on a timely basis in the form
of unrealised (i.e. non-cash) accrued charges against income whereas economic
gainsaremorelikelytoberecognisedwhen theyarerealised (BallandShivakumar
2005). Following Ball and Shivakumar (2005), we estimate the following piece-
wise-linear model to examine whether a privately held client firm’s choice of
auditor affects the timely loss recognition through accounting accruals:
ACCR = b0+ b1DCFO + b2CFO + b3DCFO × CFO + b4JOINT
+ b5JOINT × CFO + b6JOINT × DCFO + b7JOINT × DCFO
× CFO + fixed effects + 1
where cash flow from operations (CFO) is measured as the earnings before excep-
tional and extraordinary items less accruals. We define accruals (ACCR) as the
change in current assets (less the change in cash and cash equivalents) from the
prior year minus the change in current liabilities (less the change in short-term
debt and the current portion of long-term debt) minus depreciation. We standardise
both the accruals and the cash from the operations by the beginning-of-period total
assets. DCFO is a dummy variable taking the value 1 if the cash flow (CFO) is
negative and zero otherwise. Similar to Dechow et al. (1998) and Ball and
10M. Zerni et al.
Shivakumar (2005), we predict a negative coefficient for cash flows (b2). Losses
are hypothesised to be recognised faster via accruals than gains; hence, we
predict a positive incremental coefficient (b3) for negative cash flows. As in
Basu’s (1997) model, the primary coefficient of interest is the coefficient on the
interaction variable JOINT∗DCFO∗CFO (b7), which measures the incremental
effect of the choice of a joint audit on earnings conservatism relative to the single
Big 4 auditors.
4.3. Results for the Earnings Conservatism Tests
Table 2 reports the descriptive statistics for the variables used in our analyses of
earnings conservatism. Panel A reports the descriptive statistics for the sample of
public firms whereas panel B reports the descriptive statistics for the privately
held companies. Recall that no joint audits were conducted by two non-Big 4
firms in our analyses.17
Table 2. Descriptive statistics for the earnings conservatism tests.
Panel A: Public firm sample (N ¼ 1,257)
VariableMean Std.Min 25 %Median75 % Max
Panel B: Private firm sample (N ¼ 1,160)
extraordinary items scaled by the stock price at the beginning of the period; R is the annual raw stock
zero; JOINT is a dummy variable with a value of one if the client firm employs a joint audit, otherwise
zero; BIG is a dummy variable with a value of one if the client firm employs a single Big 4 audit firm,
otherwise zero; LMV is the natural logarithm of the market value of equity in thousands of Swedish
accounting accruals divided by the opening total assets. Accruals are defined as the change in current
assets (less the change in cash and cash equivalents) from the prior year minus the change in current
a dummy variable with a value of one if cash flow (CFO) is negative, otherwise zero.
Do Joint Audits Improve Audit Quality?11
Table 3. Regression results for earnings conservatism tests.
Panel A: Public firm sample (N ¼ 1,257). Results from the model based on Basu (1997).
Variable Exp. SignCoef. Prob. Coef.Prob. Coef.Prob.
Annual fixed effects?
Industry fixed effects?
-2 Res Log Likelihood
M. Zerni et al.
Panel B: Private firm sample (N ¼ 1,160). Results from the model based on Ball and Shivakumar (2005).
VariableExp. Sign Coef. Prob.Coef.Prob.
Annual fixed effects?
Industry fixed effects?
-2 Res Log Likelihood
The dependent variable EARN is defined as the earnings per share before extraordinary items scaled by the stock price at the beginning of the period. R is the annual
raw stock return; DR is a dummy variable with a value of one if the annual raw stock return is negative, otherwise zero; JOINT is a dummy variable with a value of
one if the client firm employs a joint audit, otherwise zero; and BIG is a dummy variable with a value of one if the client firm employs a single Big 4 audit firm,
otherwise zero. LMVisthe natural logarithmof the market value of equityin thousandsof Swedish kronor; P/B isthe price-to-bookratio; andDTAis theratio ofdebt
to total assets. The dependent variable ACCR is the amount of accounting accruals divided by the opening total assets. Accruals are defined as the change in current
assets (less the change in cash and cash equivalents) from the prior year, minus the change in current liabilities (less the change in short-term debt and current portion
of long-term debt), minus depreciation. CFO is measured as the earnings before exceptional and extraordinary items less accruals; DCFO is a dummy variable with a
value of one if cash flow (CFO) is negative, otherwise zero. Statistical significances are calculated by adjusting the standard errors for heteroskedasticity (White
1980) and firm-level clustering (Petersen 2009). Asterisks∗∗∗,∗∗, and∗denote two-tailed statistical significance at the 1%, 5%, and 10% levels, respectively.
Do Joint Audits Improve Audit Quality?
Our public firm sample includes a total of 129 joint audit observations from 28
unique firms (not tabulated). Panel A reveals that Big 4 firms dominate the
Swedish public company audit market, with a combined market share of single
auditor clients of 90%. The mean (median) annual raw stock return is 14.1%
(10.4%); furthermore, in approximately 41% of firm-years, the stock return is
negative, allowing for an examination of the asymmetric recognition of economic
gains and losses.
For privately held companies, the mean (median) amount of accruals is –0.048
(–0.020), with a range from –0.718 to 0.706. Furthermore, the mean (median)
operating cash flow is 0.091 (0.061), and the cash flow is negative in approxi-
mately 27% of observations.18We conclude that there is a sufficient number of
observations with ‘bad news’ to capture the expected timely recognition of
losses in both of the earnings conservatism models.
Panel A of Table 3 reports the results for the Basu (1997) sample of public
companies whereas panel B reports the results for Ball and Shivakumar’s
(2005) measure of earnings conservatism.
As column 1 of Table 3 indicates, the standard Basu (1997) model appears to
work relatively well with the Swedish data. The estimated coefficient of the inter-
action DR∗R is positive and highly significant, which suggests more timely loss
Column 2 reports our extension of the Basu model, which is defined in equation
(2). In column 3, we allow a firm-specific random intercept to control for the het-
erogeneity in firm characteristics that may affect the risk-return relationship. In
columns (2) and (3), the three-way interaction term ‘R∗DR∗JOINT’ (‘R∗DR∗BIG’)
tests the incremental earnings conservatism of the joint auditor (single Big 4
auditor) clients relative to the non-Big 4 auditors’ clients. The coefficient of the
interaction term ‘R∗DR∗JOINT’ is positive and significant at the 5% level. More-
over, the coefficient of ‘R∗DR∗BIG’ is positive and significant, although the mag-
nitude is lower. Collectively, these findings suggest that the clients of dual auditors
havethemostconservativeearningswhileclientsof singlenon-Big4 auditorshave
the least conservative earnings. In terms of earnings conservatism, the clients of
single Big 4 auditors appear to fall between these two classes.19
Panel B of Table 3 presents the results from the regression identified in
equation (3). As can be seen from column 1 of panel B, the estimated coefficient
on the interaction DCFO∗CFO is positive and significant at the 1% level. Thus,
the negative relationship between accruals and cash flow is less pronounced
during losses than during gains. In other words, the empirical evidence supports
the view that relative to gains, economic losses are recognised more quickly via
Column 2 reports our extension of the model, which allows the asymmetric
recognition of unrealised gains and losses to vary between clients of single Big
4 firms and clients that employ joint audits. We find that the test variable JOIN-
T∗DCFO∗CFO is estimated to be significantly positive, which suggests that, on
average, the clients of dual auditors have more conservative earnings than their
14M. Zerni et al.
propensity-score-matched counterparts. Overall, the results of the accruals-based
test, like those based on the Basu (1997) model, suggest that the client firms that
employ joint audits make more conservative reports than the clients of single
auditors. Joint liability essentially means that both auditors bear the (\,potentially)
incremental audit risk arising from the likelihood that the other auditor fails to
adequately perform its share of the audit work. One explanation for the increased
conservatism may thus relate to each auditor’s responses to information asymme-
try regarding the other auditor’s actions. In other words, each auditor responds to
the potential ‘free-rider problem’ by enforcing more conservative accounting
5.Multivariate Analysis: Abnormal Accruals
5.1. Abnormal Accruals Test Specifications
Our second set of tests employs abnormal working capital accruals as a proxy for
audit quality. For both the public and private samples, we define working capital
accruals (WA) as the change in the current assets (less the change in cash and cash
equivalents) from the prior year minus the change in current liabilities (less the
change in short-term debt and the current portion of long-term debt). We then
specify the abnormal working capital accruals (ABWA) as the actual accruals
minus the ‘expected’ accruals. To project the expected level of accruals, we
turn to an expectation model similar to that used by DeFond and Park (2001)
and Francis et al. (2009). In this approach, the level of expected accruals is
based on each firm’s prior-year linear relationship between sales and working
capital accruals.20We calculate the expected accruals in the following manner:
Expectedaccruals = SALESt× (WAt−1/SALESt−1)
We eliminated observations with extreme abnormal accruals values. In par-
ticular, observations were deleted if the absolute value of the abnormal
working capital accruals scaled by the lagged total assets was above 0.99
(0.71), which is approximately equivalent to deleting the top and bottom 2.5%
of the distribution for public (private) clients. Our multivariate models then
regress (absolute) abnormal accruals on our research variables and control vari-
ables based on prior studies in the field (e.g. Teoh et al., 1998; Becker et al.,
1998). Again, for reasons of data availability, we specify slightly different
models for public and private firms. For public firms, we estimate the following
|ABWA| = a + b1SIZE + b2P/B + b3SALESG + b4OCF + b5LOSS
+ b6DTA + b7LAGWA + b8OVAR + b9JOINT + b10BIG
+ fixed effects + 1
Do Joint Audits Improve Audit Quality?15