Joint audit in the French setting: How to team up best?

(1)

Joint audit in the French setting:

How to team up best?

J. Hoekstra (Jorrit) S2334526

j.hoekstra.23@student.rug.nl

MSc Accountancy & Controlling | Track Accountancy Faculty of Economics and Business

University of Groningen Supervisor: dr. V.A. Porumb Second assessor: Prof. Dr. D. de Waard

14 juni 2017 Aantal woorden: 10.411

(2)

Abstract

This paper investigates if audit partners’ tenure, age, gender, and Big 4 affiliation have an impact on audit quality in the French joint audit setting. The French audit market is one of the few audit markets where audit firms are required to enclose the names of individual engagement and review partners. My results are robust to using six alternative audit quality measures. I draw on a sample of non-financial listed companies and find that the tenure, age, gender and Big 4 affiliation are significantly associated with audit quality. Specifically, being audited by two Big 4 firms, having both audit partners of the same gender and having a bigger difference in age and tenure increases audit quality. In contrast, when audited by at least one non-Big 4 firm and when the audit team consists of both male and female partners, the audit quality decreases, decreases audit quality. Taken together, the findings of this paper add to the growing literature on joint audit by documenting that the characteristics of individual auditors impact the quality of audit.

(3)

“You have brains in your head. You have feet in your shoes.

You can steer yourself any direction you choose. You're on your own.

And you know what you know.

And YOU are the one who'll decide where to go...” Dr. Seuss

(4)

1 Introduction

According to the October 2010 Green Paper, the European Commission (EC) suggested that joint audits could increase the audit market in Europe. Some parties do not share this opinion, since the positive effects of joint audit do not outweigh the downside of joint audit, namely the audit costs. In this paper I investigate how, in a joint audit setting, the audit quality can be optimized by considering the composition of the audit team. Specifically, I assess whether the age, gender, tenure and audit firm of audit partners have a significant effect on the audit quality in the French joint audit setting.

An increasing amount of literature calls for audit quality to be measured at individual partner level. The recent paper of Knechel, Vanstraelen and Zerni (2015) documents that aggressive or conservative reporting is not randomly allocated over the engagements, but that it depends on the individual auditor. Additionally, Gul, Wu and Yang (2013) draw on a sample of Chinese firms to show that auditor characteristics can be used to explain inter-firm variations in audit quality. This finding is corroborated by Cameran, Campa and Francis (2016) as they find an association between the characteristics of individual auditors and audit outcome in the UK market.1 While this research stream has been developing over the last decade, no paper to date has analyzed how the characteristics of audit partners impact audit quality in a joint-audit setting.

Recent literature shows the existence of individual auditor effects. Most papers investigate the auditor effects in a single audit setup. In this paper I examine the auditor effects in a joint audit setting. By doing this, I will expand the literature by combining two fields of research, namely: the joint audit and individual auditor effects. The aim of the project is to determine how to maximize the quality delivered by audit firms in a joint audit setting, by assessing audit partners’ performance based on their age, gender, tenure and audit firm affiliation. The understanding of these effects can lead to a lower cost of debt, greater credit access and less likeliness to post collateral (Aobdia, Lin and Petacchi, 2015).

I focus on the French audit market, as it has specific features. It has relative strict litigation, which makes it relatively easier to judge whether an auditor has fully fulfilled its role,

1_{This evidence is also found by Aobdia, Lin and Petacchi (2015), in the form of an influence of the individual}

auditor on the Earnings Response Coefficient (ERC). However, in the article of Francis, Pinnuck and Watanabe (2014) evidence is found consistent with audit style increasing the comparability of reported earnings within a Big 4 auditor’s clientele. Which means that the firm style has more influence over audit quality, than the auditor style. In the article of Francis et al. (2014), there is a focus on the combination of auditors to be chosen by the organization, Big-4 or non-Big-4. They find that companies using one Big 4 auditor paired with a non-Big 4 auditor have smaller income-increasing abnormal accruals than those using no Big 4 auditors and that such an effect is even stronger for companies that use two Big 4 auditors (Deng, Lu, Simunic and Ye, 2012).

(6)

and it requires both the enclosure of the audit partners responsible for the joint audit and the joint audit itself. These three features are of great importance to this paper, as it presents the opportunity to collect personal information about the audit partners, and test this information for significant effects on audit quality. France is one of the only countries in Europe where joint audit is a requirement by law. I draw a sample on 142 non-financial listed firms from France, for the years 2003-2015. Because most of the data I use to test the hypotheses was not readily available from databases, I hand-collected some of the variables. I combine the hand-collected data with data from Datastream, where I obtain the financial data of all companies, and Asset4, where I get data on audit fees and restatements. This results in a unique dataset that allows to test my hypotheses.2

Given the framework from the agency theory and the stakeholder theory, four main hypotheses are generated and tested. For the first hypothesis, I expect to find a significant positive relation, since an older person has a higher chance of being more experienced relative a young person. Therefore, the higher the age of the auditors, the higher the audit quality. I find conclusive results to support Hypothesis 1, since that audit quality is positively determined by the age of the audit partners. One of six audit quality measures shows evidence to support the direct relation between the audit partner age and audit quality. However, four of six audit quality measures show a significant positive relation for the relation between the difference in age of the audit partners and the audit quality. When the difference in audit partner age increases, so does the audit quality.

My expectation for the second hypothesis, is a significant positive relationship for two female audit partners. I find evidence to conclude that audit quality is significantly influenced by gender of the audit partners. The outcome showed multiple significant relations, showing that two male or two female audit partners increase the audit quality, while one male and one female audit partner decrease audit quality. One of the main findings in this paper is, however, that when two male auditors, both working for a Big 4 audit firm, audit a company, the positive relation to audit quality is reversed, and becomes highly significantly negative, thus decreasing the audit quality.

Further, I expect to find a significant positive relation between the tenure of the audit partners and the audit quality. The higher the tenure of both audit partners, the more experienced they are. More experience will increase their skill in recognizing material misstatements and thereby increase the audit quality. The results show there is no significant relation between the

(7)

tenure of the auditors and the audit quality. I do, however, find a highly significant positive relation for the difference in tenure of the audit partners and the audit quality. This implies that the higher the difference in tenure between the audit partners, the higher the audit quality.

For the last hypothesis, I expect to find that when a firm is audited by one or two Big 4 firms, the audit quality is higher than the audit quality when audited by no Big 4 firms. The results are mixed. Two of our six audit quality measures show significant results. Being audited by two Big 4 firms significantly influences audit quality, however, due to mixed results, I cannot conclude on the nature of this relationship.

This paper contributes to the literature by combining two fields of research. Both the influence of the characteristics of an individual auditor on the audit quality (Gul et al., 2013; Aobdia et al., 2015; Cameran et al., 2016), as the influence of a joint audit on the audit quality (Andre et al., 2014) have been investigated in prior papers. These subjects were, however, never combined. This will be the first paper to combine both subjects into one research. Therefore, by conducting this research, I fill an important gap in the literature. The results of this paper show, that when auditors are matched on their characteristics and firm they work for, it is possible to increase audit quality, given the significant relations found in this paper. Additionally, it does not only show that it is possible, the results show how to gain this audit quality increase.

The rest of the thesis is structured as follows. In Section 2, I develop the theoretical framework, including the institutional background and the literature review. In section 3, I develop my hypotheses. In section 4, I show the data sources and research method. Section 5 contains all results from testing my hypotheses and section 6 contains the conclusion of this paper.

2 Theoretical framework

2.1 Institutional background

2.1.1 What is a joint audit?

In previous literature, joint audit is described as an audit in which two or more auditors from separate audit firm are appointed to audit financial statements of an audit client; develop the audit plan jointly; performing the audit work jointly; making periodic cross reviews and mutual quality controls; issuing and signing a single audit report; bearing joint liability in case of audit failure (Zerni, Elina, Jarvinen and Niemi 2012, Alanezi, Alfrashaid and Alolushi 2012,

(8)

Baldauf and Steckel 2012, Ratzinger-Sakel, Audousset-Coulier, Ketunnen and Lesage 2013, Paugam and Casta 2012).

A sharp distinction should be made between joint audit and dual audit. The dual audit setting differs from the joint audit at a few critical points. In a dual audit setting, the audit plan is developed separately, the audit work is performed separately, there are no mutual quality controls and periodic cross reviews and it is possible that two or more audit reports are issued. A last form of audit in which two or more separate auditors are involved is the double audit. In a double audit setting, a single auditor is required to perform the audit work twice (El Assy, 2015). In this paper I focus on the joint audit setting as described above.

How did the joint audit emerge in France?

De Beelde, Gonthier-Besacier and Mikol (2009) state that the traditional audit function in France originated from around the 1850-1875. The French auditors, in French called “Commissaires des comptes”, were viewed as an extension of the government, due to their role as a safe keeper of litigation the French state has created (Gonthier-Besacier and Schatt, 2007). Where does the litigation regarding the audit come from?

Although there were functions which had a lot in common with the nowadays audit function, the real function of auditors was created with the law of May 23, 1863. This law introduced a statutory audit for a limited number of companies. Article 15 of this new law stated that a company had to elect one or more persons (commissaires) to check the balance sheet and the accounts presented by the directors. These persons were called, commissaires des comptes (Houpin and Bosvieux, 1935). The auditing system was heavily criticized for its lack of independence and competency regulations, since the commissaires des comptes mostly were employees of the company (Mikol, 1993).

An innovative decree on August 8 1935 made it impossible for directors, employees or relatives to serve as auditors. There were other changes as well: The provision of non-audit services to audit clients was not allowed, auditors could not receive any fees apart from an audit fee, auditors could only become director but only after a five year cooling-down period and the auditors had a mandate for three years, which could not be ended before the end of the term (Praquin, 2012). To protect the public savings, the 1935 decree forced listed companies to acquire at least one auditor from a list maintained by the Courts of Appeal. The commissaires on this list were tested for their technical competency. This was the start of the joint audit in France (De Beelde et al., 2009). This is regarded as the start of the joint audit in France

(9)

In 1966 the auditing profession was comprehensively organized with a new decree. The decree focused on defining the objectives and requirements with respect to examinations, independence and the professional body of the auditor and has has set an obligation to appoint two auditors for any listed company, or companies with a share capital above a certain threshold. (Audousset-Coullier, 2008). In 1984 the scope of the joint audit changed, joint audit was now made mandatory for all the companies preparing financial statements.

2.1.2 What does the audit market in France and in the EU look like?

The French audit market in the year 2016 is largely serviced by the Big 4 audit firms and two other large international firms (Granth Thornton and Mazars). Compared to the European Union average, the Big 4 firms have a smaller market share in France, while the market share of the relative small ‘other auditor’ firms, is relatively big compared to the European Union average. Another difference between the French audit market and the other European countries’ audit market is the number of firms that audit 2% or more of the total market. This is due to the fact that there a lot more small firms in France. A last remarking difference between the market share distribution of France and the United Kingdom on one side, and Spain, Italy and Germany on the other side, is that the top 4 audit firms in market share do not consist of only Big 4 firms. In France, Mazars owns a larger market share than PwC, while in the United Kingdom, Grant Thornton owns a larger market share than EY. For the direct comparison of the French and EU market share distribution, see Appendix 1 and 2.

Previous literature joint audit

André et al. (2014) investigate whether joint audits are associated with higher audit fees. The research is conducted by comparing the audit fees from French listed companies with British and Italian companies from 2007 till 2011. They find that the audit fees in France are significantly higher than the fees in Britain and Italy. They, however, do not find statistically significant differences in the size of accruals. Therefore they conclude that the higher audit fee is not related to a higher audit quality. Similarly, Deng, Lu, Simunic and Ye (2012) test whether “two heads are better than one”. In their paper they argue that, while it is possible that the audit quality increases due to the joint audit, it can also create problems like free-riding and opinion-shopping. Free-riding is taking advantage of the audit-work of other firms, while holding back on their own investment in the audit. Opinion shopping is trying to get the opinion from the

(10)

auditor that is the most positive for the company being audited. Deng et al. (2012) describe opinion shopping as: “having more draws in a lottery”.

El Assy (2015) further investigates the effect of joint audit on audit quality. The research is conducted by using 32 listed Egyptian companies, for the period of 2009 through 2013. The results show that companies audited by joint auditors are more conservative than the companies audited by only one auditor. Francis, Richard and Vanstraelen (2009) conduct their research in France, for listed non-financial companies. They find that firms using a Big 4 auditor, have smaller income-increasing accruals than firms who don’t use a Big 4 auditor.

In contrast to my paper, Francis et al (2009) only focus on the impact of joint audit, where this paper focuses on the impact of partner characteristics in the joint audit setting.

Which type of rule of law is prevalent in France?

The legal systems in the world can be distinguished into two groups: the civil law and the common law. The French legal system is a good example of a civil law system. However, the civil law system has differences in itself, these systems differ in Germany, France and Scandinavia. The biggest influence of the law was made by Napoleon in 1801. Napoleon wanted legislation that empowered the state and minimized the independence of the French judges, by making the state the sole interpreter and source of the law (Levine 2005).

Schlesinger, Baade, Damaska and Herzog (1988) wrote: “The Napoleonic code strove both to eliminate jurisprudence, the law created by judges by interpreting statues and adjudicating disputes, and to impose strict procedural formalism on court processes to eradicate juridical discretion. The main reason for Napoleon to unify and strengthen the law was the enormous corruption by the French judges”. Currently, France still is a civil law country. Civil law is codified, which means that that the legal system is comprehensively and continuously updated, so that all matters that can be brought before a court are specified in the law. The emphasis for the judge lies on establishing the facts and applying the law on these facts (Levine, 2005).

The French setting therefore means a high degree of legislation, which also means that the requirements for an audit are clear. This clarity in requirements for an audit, makes it easier to judge whether an audit partner has (or has not) fulfilled his by law described role.

(11)

2.2 Literature review

2.2.1 Agency theory

From an economic point of view, the separation between ownership and management explains the existence of the accountant. Researchers define an organization as a collection of contracts, resulting from a legal form with legal personality (Jensen, and Meckling, 1976; Fama, 1980; Fama and Jensen, 1983). With ownership, they mean the owners. Those can be shareholders or investors (principals). The interests of the principles can differ or conflict with the interests of managers (agents). If this is the case, an agent's actions may be suboptimal to the principal’s interests. This is enhanced by the occurrence of information asymmetry. An agent has more information regarding the company than the principal. The principal is dependent on the amount and reliability of information that the organization publishes. An auditor reduces the costs regarding the agency problems (Watts and Zimmerman, 1983). A principal puts faith in the auditor’s work. Limperg, an auditor himself, saw the role of the auditor as: “Being the trustee of the society”. The auditor’s existence is thus derived from the established trust of the society in the audit profession (Dassen, 1989). This research is to be placed within the framework of the agency theory.

2.2.2 Stakeholder theory

For listed companies in France, the financial statements and corresponding documents about the audit and partner disclosure are publicly available. The user of an annual report can be an owner, as in the agency theory, but it can also be a supplier, customer, employee or any other stakeholder. Unlike the agency theory, the stakeholder theory focusses on a much broader ‘audience’ than the agency theory, which only focuses on the owners. The maximalization of the shareholder value does not by definition have the highest priority for the management of an enterprise. A stakeholder is defined as anyone, each group or individual, who can influence or is influenced by the actions of an enterprise (Deegan, and Blomquist, 2006; Freeman et al., 2010).

2.2.3 Audit quality

In this paper I test for audit quality, but what exactly is audit quality? “Audit quality is determined by an auditor’s ability to discover breaches of accounting standards and the auditor’s incentives to report such breaches. Audit quality is a product of auditor competence

(12)

and independence” (Gul et al., 2013). DeFond and Zhang (2014), however define audit quality as: “Greater assurance of higher reporting quality”.

This means that, when testing for audit quality, all auditors have to be checked for competence and independence. Since this is a lot of work, and therefore barely possible, research is conducted to find another way of measuring the competence and independence of an auditor. These are so-called proxies, a variable that is not directly relevant itself, but that serves in place of an unobservable or immeasurable variable (Wickens, M.R., 1972).

There are multiple ways of measuring audit quality. In the research of Deangelo (1981), auditfirm size is used as a proxy for audit quality, since for bigger firms, clients are relatively smaller and bigger firms have a greater reputation risk. In the article of Francis and Yu (2009) it is added that, bigger firms achieve higher quality, given their higher ‘in-house’ knowledge. They also note that due to the fact that auditors in bigger firms have larger engagement hours, therefore they are more experienced in finding material misstatements. Thus, auditors in bigger offices are more likely to detect and report material misstatements.

Another proxy that is widely used, also in this paper, for measuring audit quality, are accruals. Although earnings management is not always the general accepted earnings principle, firms that manage earnings are widely viewed as having lower quality earnings (Francis and Yu, 2009).

Rajgopal, Srinivasan and Zheng (2015) find that in much of the papers investigating audit quality, 5 different proxies are used: Big N auditor, discretionary accruals, going concern opinions, accrual quality and meet or beat a quality earnings target. They also acknowledge that the use of these proxies is mostly motivated by a cost-benefit consideration.

According to DeFond and Zhang (2014), proxies can be categorized into input-based proxies and output-based proxies. Input-based proxies refer to auditor-specific characteristics and auditor fees. The most popular example of auditor-specific characteristics, is whether the company is audited by a Big N auditor or not. The audit fees are used to measure how much effort an auditor puts in the client’s audit. In the article of Frankel, Johnson and Nelson (2002), the ratio between the audit fees and non-audit fees are used as proxy for independence. Lawrence, Minutti-Meza and Zhang (2011) investigate whether the difference in Big 4 and non-Big 4 audit quality is a reflection of the clients characteristics. They do not find conclusive evidence for this.

Output-based measures are material misstatements, going concern opinions and financial reporting characteristics, but also the earnings response coefficient, stock price changes and cost of debt measures (DeFond and Zhang, 2014).

(13)

2.2.4 Individual auditor features

Recent literature shows a significant relationship between the identity of individual audit partners and the Earnings Response Coefficient (ERC) (Aobdia et al., 2015), which proves that audit partner features can be influential and thus is a factor to account for in research. Nelson and Tan (2005) note that: “Auditors need to perform a variety of tasks to form an overall assurance or attestation opinion. To do so, various personal attributes of the auditor (skills and personality) influence the outcome”. Therefore, it does seem likely that the characteristics of individual audit partners influence the outcome of the audit (Gul et al., 2013).

Cameran et al. (2016) find a relation between the characteristics of an auditor and the audit outcome for the UK market, where they only focused on Big 4 firms. The recent literature shows a trend in the level at which the audit quality-analysis is conducted, going from firm-level in the earlier stages to individual auditor firm-level now (Cameran et al., 2016). This conclusion was also made by Church, Davis and McCraken (2008), where they spoke out for more research on the existence of a systematic relationship between the individual characteristics and audit quality. This relationship was also investigated in the article of Cameran et al (2016), where they split the influence on the audit quality into three layers, individual audit characteristics, audit office and audit firm. They conclude that the influence of partner fixed effects explain more of the variation than the combined fixed effects of audit firms and offices. In this paper I investigate the auditor effects, for age, gender, tenure and Big 4 affiliation on audit quality in a joint audit setting.

3 Hypothesis development

Audit quality is important for audit firms, but for the company being audited as well. According to Skinner and Srinivasan (2012) and Swanquist and Whited (2015), the reputation loss resulting from low quality audits is associated with severe capital market consequences. Therefore it is important to investigate what exactly determines audit quality in the French joint audit setting.

Age

The age of an auditor may influence the quality of the auditor. Literature shows that when people age, they become more risk-averse (Hang and Wanna, 2004). Another factor to take into account is that older people tend to have more experience, for which I can imagine the

(14)

audit quality to increase. Given that in a joint audit setting the audit team can result in different combinations of individual auditors of different ages, this represents a salient setting to test empirically the impact of age on audit quality. I expect to find a positive relation between the age of the audit partners and the audit quality. Moreover, the affiliation of an auditor to a Big 4 firms is likely to moderate this relationship. Since previous literature documents that Big 4 auditors are of better quality relative to non-Big4 auditors (André et al., 2016; Francis and Yu, 2009), I expect that the impact of age on audit quality will not be significant when the auditors works in a Big 4 audit firm. Therefore, I formulate the following hypotheses:

Hypothesis 1: Audit quality is determined by the age of the audit partners in a joint audit setting. Hypothesis 1A: Audit quality is determined by the average age of the audit partners in a joint

audit setting.

Hypothesis 1B: Audit quality is determined by the difference in age of the audit partners in a

joint audit setting.

Hypothesis 1C: Whether a company is audited by two Big 4 firms or not, influences the relation

of the average audit partner age on the audit quality in a joint audit setting.

Hypothesis 1D: Whether a company is audited by two Big 4 firms or not, influences the relation

of the difference in audit partner age on the audit quality in a joint audit setting.

Gender

Given previous literature, women in general are more risk-averse (Jianakoplos and Bernasek, 1998; Schubert, Brown Gysler and Brachinger. 1998), this also applies to financial risks (Maxfield, Shapiro, Gupta and Hass, 2010). I expect to find a significant relation between the gender of the audit partners and the audit quality. Specifically, I expect to find that two female auditors pertain a higher audit quality, due to the fact that they are less likely to accept risks regarding the audit. Moreover, the affiliation of an auditor to a Big 4 firms is likely to moderate this relationship. Since previous literature documents that Big 4 auditors are of better quality relative to non-Big4 auditors (André et al., 2016; Francis and Yu, 2009), I expect that the impact of gender on audit quality will not be significant when the auditors works in a Big 4 audit firm. Therefore, I formulate the following hypotheses:

Hypothesis 2: Audit quality is determined by the gender of the audit partners in a joint audit

(15)

Hypothesis 2A: Whether both auditors work at Big 4 firms, influences the relation of the gender

of the audit partners on the audit quality in a joint audit setting.

Hypothesis 2B: A company audited by two female auditors obtain a higher level of audit quality

in a joint audit setting.

Tenure

The third hypothesis focuses on the tenure of the audit partners and audit quality. Literature shows there is correlation between the tenure of an auditor and his/her audit quality, as shown in the articles of Ghosh and Moon (2005), Knechel and Vanstraelen (2007) and Francis (2004). In general, the articles show that an auditor with low tenure show lower audit quality. However, it is possible that the auditor experiences loss of independence due to the relationship the auditor built with the client. To test whether the above conclusion is also applicable to the French joint audit setting, I formulate the following hypotheses:

Hypothesis 3: Audit quality is determined by the tenure of the audit partners in a joint audit

setting.

Hypothesis 3A: Audit quality is determined by the average tenure of the audit partners in a

joint audit setting.

Hypothesis 3B: Audit quality is determined by the difference in tenure of the audit partners in

a joint audit setting.

Hypothesis 3C: Whether a firm is audited by two Big 4 firms, influences the relation of the

tenure of the audit partners on the audit quality in a joint audit setting.

Big 4 / non-Big 4

The auditing literature concludes that in general the audit quality increases when a company is audited by a Big 4 firm (El Assy, 2015; André et al., 2016; Francis and Yu, 2009). According to the article of Cameran, Campa and Francis (2016), Big 4 firms have higher reputational incentives to maintain strong quality control systems. These systems can also be in place for the non-Big 4 audit firms, they may, however, not have these same incentives. In comparison to Big 4 firms they do not have the same risks regarding their reputation. This effect should be visible in a joint audit setting as well. Therefore I expect to find higher audit quality for two Big 4 auditors. Accordingly, I present the following hypotheses:

(16)

Hypothesis 4: Audit quality is determined by whether a company is audited by two Big 4 firms,

one Big 4 firm and one non-Big 4 firm, or two non-Big 4 firms in joint audit setting.

Hypothesis 4A: Audit quality is determined by whether a company is audited by at least one

(17)

4 Data sources and research design

4.1 Data sources

I conduct this study using data from French listed non-financial firms for a number of reasons. First, France is one of the only countries that has mandatory joint audits for listed firms. This creates the opportunity to research the association between audit quality and individual auditor characteristics in a joint audit setting. Second, the French listed companies are required by law to file their audited financial statements. Third, French law requires the auditor to sign the audited statements.

I start the data selection considering all French listed non-financial firms with financial reporting data available in WorldScope for the 2003-2015 period. I exclude the French financial listed firms, since the law requires other forms of financial reporting for these firms. This results in 8,203 firm year observations. For these firms I download the financial reporting data from DataStream and Auditanalytics. Not all firms had data readily available from DataStream and Auditanalytics, so I lose those firms. This results in 631 firms, with 8,203 firm-year observations. Each observation is one year of one firm. Each firm-year observation in our sample is an audit of a firm in a year and has been conducted by at least two auditors. This means that each observation has two signatures.

For the audit partners I hand collect the name, audit firm, age, gender and tenure of the auditor.3_{This data is mostly publicly available from annual reports of these firms, the audit firm}

websites, or social media sites as LinkedIn or Facebook. Subsequently, I compute audit quality variables with the financial data obtained from DataStream. I also add the Restatement and Audit fee variables from Asset 4. In line with the article of Francis et al. (2004), I winsorize the continuous variables at 1 and 99 percent to eliminate the effect of potential outliers.

Not all the variables were available for all 631 companies, therefore I delete the firm-year observations when they did not have two auditors. This would leave 1.003 firm-firm-year observations. After this, I delete the firm-year observations for which there was no audit quality data available. This results in a total of 782 firm year-observations. The dataset is unbalanced, since not all the potential observations are in the dataset. Since N x T is 8,203 and my dataset consists of 782 observations. The outcome of the data selection and data collection process is a final sample 831 observations containing 63 firms, 40 territorial offices and 400 individual signing partners. Please see Table 2 for the descriptive statistics of the final sample.

3_{This data is mostly publicly available from annual reports of these firms, the audit firm websites, or social media}

(18)

4.2 Empirical models

For measuring the drivers of audit quality, I distinguish multiple levels of influence: Audit fees, audit firms, audit offices and individual audit partners. This paper is focused on the influence of audit partner characteristics. I regress each of our audit quality measures on partner characteristics. I do this after controlling for time-varying audit quality determinants. In order to determine if a set of fixed effects has a significant impact on the determined proxies, I look at the significance of the coefficient of the relation between the variable and the audit quality measure.

For measuring the drivers of audit quality, I use multiple proxies, which I divide into two categories as done by DeFond and Zhang (2014), input- and output-based proxies. I mostly use output-based proxies. The proxies I use are defined below. I use these measures to further run OLS models with individual effects for audit partners, audit offices and audit firms. By describing the fit of the models and explanatory power of the groups of variables, I distinguish the drivers of audit quality.

The first proxy of audit quality is financial restatement. This proxy shows whether a firm needed to make a restatement in a certain year. The proxy is defined as a dummy variable, so the values can be 0 or 1.

RESTATi,t (1)

The second proxy of audit quality I use are the total accruals as defined by Healy (1985). Healy defines accruals as the difference between the reported financial results and the operational cash flow.

AC = DEP + ∆ AR + ∆ INV - ∆ AP,

Where:

AC = Total accruals DEP = Depreciation

∆ AR = Value of accounts receivable – value of accounts receivable last year ∆ INV = Value of inventory – value of inventory last year

(19)

The calculated total accruals are then scaled by lagged total assets:

AbsTotalAccrualsi,t = ACi,t / TAi,t-1, (2)

Where:

TAi,t-1 = Lagged total assets, the value of last year’s total assets.

The next model I use is the Modified Jones Model (1991). This model estimates the abnormal accruals attempts to control for the effects of changes in a firms economic circumstances on non-discretionary accruals. The model was introduced by Jones (1991) and enriched by Dechow, Sloan and Sweeney (1995). The estimation of abnormal or discretionary accruals by regressing working capital accruals on selected determinants is conducted as follows:

WCACCRi,t = β0,t + β1,t*(∆SALESi,t - ∆RECi,t) + ԑi,t ,

Where:

WCACCR is computed as: (∆CAi,t - ∆CASHi,t ) – (∆CLi,t - ∆CPLTDi,t - ∆ITi,t);

SALESi,t represents the sales of firm i in period t;

RECi,t represents accounts receivables of firm i at time t;

CAi,t stands for current assets of firm i at period t;

CASHi,t stands for current assets of firm i at time t;

CPLTDi,t stands for the level of long term debt of firm i at period t,

ITi,t stands for the income tax payable of firm i in period t.

All variables have been scaled with lagged total assets (TAi,t-1).

With the residual from the WCACCR, The abnormal working capital accruals are calculated:

AWCAi,t = WCACCRi,t – (β0,t + β1,t* (∆SALESi,t - ∆RECi,t)) (3)

(20)

In addition to model 3, I also use a long-term version of the Modified Jones model, which includes Gross Property Plant and Equipment (GPPE). This additional explanatory factor for normal accruals is obtained by the following model:

TACCRi,t = β0,t + β1,t * (∆SALESi,t - ∆RECi,t) + β2,t * GPPEi,t + ԑi,t

TACCRi,t is computed as the difference between income before extraordinary items

(IBEIi,t) and cashflow from operations (CFOi,t), again scaled by lagging total assets. GPPEi,t is

the gross property, plant and equipment firm i in period t.

Again, using the residual from estimating TACCR, model 4 can be calculated:

ABNTACCRi,t = TACCRi,t - (β0,t + β1,t * (∆SALESi,t - ∆RECi,t) + β2,t * GPPEi,t). (4)

In this model (4), β0,t, β1,t and β2,t are regression coefficients that have to be estimated.

In this paper I also refer to this measure as MD2.

The next model I use, is the modified Dechow and Dichev model, by Dechow and Dichev (2002). This model adjusts the Jones model to include past, present and future cash flows in the regression of total accruals on its normal determinants.

The unadjusted total accruals are computed as follows: WCACCRi,t = β0,t + β1,t* CFOi,t-1 + β2,t* CFOi,t + β3,t* CFOi,t+1 + ԑi,t

Where CFOi,t+π is cash flow from operations of firm i in period π = -1, 0, 1.

ABNWCACCRi,t = WCACCRi,t – ( β0,t + β1,t* CFOi,t-1 + β2,t* CFOi,t + β3,t* CFOi,t+1 ) (5)

This measure is also referred to as DD1.

The sixth audit quality measure I use in this paper, is the long-term Dechow and Dichev (2002) model. This model includes GPPE, the gross property plant and equipment.

The long-form Dechow and Dichev (2002) model can be specified as:

TACCRi,t = β0,t + β1,t* CFOi,t-1 + β2,t* CFOi,t + β3,t* CFOi,t+1 + β4,t * GPPEi,t + ԑi,t

(21)

ABNACCRi,t = TACCRi,t - ( β0,t + β1,t* CFOi,t-1 + β2,t* CFOi,t + β3,t* CFOi,t+1 + β4,t * PEi,t +

ԑi,t (6)

Managers can have different objectives, in one organization the objective is to pay as less tax as possible, so the objective is to decrease the earnings (income decreasing accruals), while in other organizations it is important to show as much result as possible (income increasing accruals), I consider the absolute value of the abnormal accruals measures. For all models I include a set of control variables for other factors that can affect audit quality, as done in other researches. The following control variables are added to our model: Size of the firm (Cameran et al., 2016 and Johnson, Khurana and Reynolds, 2002); Company leverage (Cameran et al., 2016 and Dechow, Sloan and Sweeney, 1995); Return on assets (Cameran et al., 2016); The presence of loss (Gul et al., 2013 and Cameran et al., 2016); lastly I added FeeRatio and SalesGrowth. Hoitash, Markelevich and Barragato (2007) find a significant positive relationship between audit quality and audit fees. Therefore in this paper I control for the size of the audit fee. These variables are defined in Table 1.

Before testing the data for my hypotheses, I run the Pearson Correlation Coefficient test. This test determines whether the dataset shows multicollinearity, which means that two variables can act too much alike. The Pearson Correlation Coefficient test prevents a research from showing the same measures twice.

To test the hypothesis, I use the Ordinary Least Squared method. The panel regression models I use for testing my hypotheses, are shown below. The variables mentioned are defined in Table. The models are used to test the hypotheses.

I test my first main hypothesis, whether audit quality is determined by the age of the audit partners in a joint audit setting, by using the following panel regression models:

1A:

AQi,t = α0 + ∑ α1,y * AudAgeAveragei,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * AudAgeAveragei,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t + α5 *

SGi,t + ∑αy,6 * DPOLi,t + ∑ α7,y * DYt + ∑ α8,y * DIndustryt + ԑi,t

1B:

AQi,t = α0 + ∑ α1,y * AudAgeDifferencei,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * AudAgeDifferencei,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t + α5 *

(22)

1C:

AQi,t = α0 + ∑ α1,y * DB4B4*AudAgeAveragei,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * DB4B4*AudAgeAveragei,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t +

α5 * SGi,t + ∑αy,6 * DPOLi,t + ∑ α7,y * DYt + ∑ α8,y * DIndustryt + α9 * FeeRatioi,t + ԑi,t

1D:

AQi,t = α0 + ∑ α1,y * DB4B4*AudAgeDifferencei,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * DB4B4*AudAgeDifferencei,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t

+ α5 * SGi,t + ∑αy,6 * DPOLi,t + ∑ α7,y * DYt + ∑ α8,y * DIndustryt + α9 * FeeRatioi,t + ԑi,t

The second main hypothesis, whether audit quality is determined by gender of the audit partners in a joint audit setting, is tested using the following panel regression models:

2A:

AQi,t = α0 + ∑ α1,y * DB4B4*DGenMaleMalei,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * DB4B4*DGenMaleMale,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t +

α5 * SGi,t + ∑αy,6 * DPOLi,t + ∑ α7,y * DYt + ∑ α8,y * DIndustryt + ԑi,t

2B:

AQi,t = α0 + ∑ α1,y * DGenFemaleFemalei,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * DB4B4*DGenMaleMale,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t +

α5 * SGi,t + ∑αy,6 * DPOLi,t + ∑ α7,y * DYt + ∑ α8,y * DIndustryt + ԑi,t

The third main hypothesis, whether audit quality is determined by the tenure of the audit partners in a joint audit setting, is tested by using the following panel regression models: 3A:

AQi,t = α0 + ∑ α1,y * AudTenAveragei,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * AudTenAveragei,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t + α5 *

SGi,t + ∑αy,6 * DPOLi,t + ∑ α7,y * DYt + ∑ α8,y * DIndustryt + ԑi,t

3B:

AQi,t = α0 + ∑ α1,y * AudTenDifferencei,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * AudTenDifferencei,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t + α5 *

(23)

3C:

AQi,t = α0 + ∑ α1,y * DB4B4*AudTenAveragei,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * DB4B4*AudTenAveragei,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t +

α5 * SGi,t + ∑αy,6 * DPOLi,t + ∑ α7,y * DYt + ∑ α8,y * DIndustryt + α9 * FeeRatioi,t + ԑi,t

3D:

AQi,t = α0 + ∑ α1,y * DB4B4*AudTenDifferencei,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * DB4B4*AudTenDifference,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t

+ α5 * SGi,t + ∑αy,6 * DPOLi,t + ∑ α7,y * DYt + ∑ α8,y * DIndustryt + α9 * FeeRatioi,t + ԑi,t

The fourth main hypothesis, whether audit quality is determined being audited by zero, one or two Big 4 firms in a joint audit setting, is tested using the following panel regression models:

4A:

AQi,t = α0 + ∑ α1,y * DB4B4i,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * DB4B4i,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t + α5 * SGi,t + ∑αy,6

* DPOLi,t + ∑ α7,y * DYt + ∑ α8,y * DIndustryt + α9 * FeeRatioi,t + ԑi,t

4B:

AQi,t = α0 + ∑ α1,y * 1or2Big4i,t + Control variables + ԑi,t

AQi,t = α0 + ∑ α1,y * 1or2Big4i,t + α2 * LogSizei,t + α3 * LEVi,t + α4 * ROAi,t + α5 * SGi,t +

∑αy,6 * DPOLi,t + ∑ α7,y * DYt + ∑ α8,y * DIndustryt + ԑi,t

(24)

5 Results

5.1 Descriptive statistics

In Table 2 I present descriptive statistics of audit firms, audit offices and audit partners. This table gives an interesting insight in the French audit market. This table translates the distribution of the number of observations, i.e. client-years. In total 10% of the audit firms in my sample audit more than 100 client-years, while 38 audit firms only audit 1-5 client-years. For the individual partners, these amount are lower. While 369 auditors obtain 1-5 client-years in this sample, there are no auditors with over 20 client-years. Table 6 presents the Pearson Correlation coefficient. Table 3 presents the descriptive statistics of the variables used in this paper. The Table shows the dependent, independent and control variables.

5.2 Individual characteristics

In Table 4 and 5 I present the results of regressing our 6 audit quality measures on their normal determinants. Due to a lack of observations for the FeeRatio variable, I regress the six audit quality measures with (Table 4) and without FeeRatio (Table 5).

5.3 Age

In Table 7 I present the results of regressing the average age of the audit partners with all six audit quality measures. Column 3 shows that there is a significant positive relationship (0.004) between the average auditor partner age and the audit quality for model MD1. The positive relationship shows that when the average age increases, so does the audit quality measure, so the audit quality decreases.

Table 8 shows multiple significant relations between the difference in auditor age and audit quality. Columns 3, 4, 5 and 6 show a significant negative relationship, which means that when the difference in auditor age increases, the audit quality increases as well.

Table 9 shows the outcome for regressing the average auditor age on the the audit quality, with the Big 4 dummy as a moderating variable. By lack of results for all models except MD2, I only show the outcome for MD2. The Table shows that if the Big 4 dummy and the average auditor age are tested individually, the both show a significant negative relationship to the audit quality, so a positive relationship with the audit quality. However, if the Big 4 dummy acts as a moderator variable, the relation becomes significantly positive (0.004).

(25)

Table 10 presents the outcome of the regression of the difference in age of the audit partners and the audit quality measure DD1, with the Big 4 dummy as a moderating variable. Due to a lack of significant results, only the outcome for audit quality measure DD1 is shown. The outcome in Table 10 shows that when the Big 4 dummy (column 1) is regressed on audit quality, it shows a non-significant positive relation (0.012) with the audit quality measure. The outcome of regressing the difference in audit partner age and the audit quality measure is significantly negative (-0.002). When the Big 4 dummy as a moderating variable is added to the column 2, it results in column 3. Column 3 shows three significant relations with the audit quality measure. The Big 4 dummy strengthens the relation of the difference in audit partner age on the audit quality measure, which means that coëfficient of the relation becomes higher, from -0.002 to -0.006, while maintaining its significance.

5.4 Gender

Table 11 shows the result of regressing all gender dummies (male-male, male-female, female-female) on the audit quality measures. The Restat variable showed significant results for all three gender dummies. Column 1 of Table 11 shows a significant negative (-0.042) relation between the male-male gender dummy and the audit quality measure. Column 2 shows a significant positive relation between the male-female gender dummy and the audit quality measure. This would mean that when both genders are represented in an audit team, the audit quality decreases. Lastly, column 3 shows a highly significant negative (-0.134) relation between the female-female gender dummy and the audit quality measure. Also, there is a difference in the coefficient for columns 1 and 3, -0.042 and -0.134, respectively. Table 12 shows the outcome of regressing the male-male gender dummy on audit quality, with a moderating big 4 dummy. Column 1 shows that when the Big 4 dummy is tested individually, it shows a significant negative (-0.012) influence on the audit quality measure. Column 2 shows a non-significant positive relationship between the male-male gender dummy and audit quality. When these are combined to column 3 and the Big 4 dummy is used as a moderating variable, a highly significant positive (0.005) relation is shown. This means that the male-male gender dummy increases audit quality, same as with two Big 4 firms who also increase audit quality. But when two male partners working for a Big 4 firm audit a company, the audit quality decreases.

(26)

5.5 Tenure

Table 13 shows the outcome of regressing the average tenure of the audit partners on the audit quality measures. This results in no significant relations, while the coefficients vary from -0.001 to 0.006.

Table 14, column 3, 5 and 6 show significant and a highly significant negative (-0.006, -0.006 and -0.007) relation between the difference in audit partner tenure. This would imply that when the difference between the audit partners’ tenure increases, the audit quality increases as well.

Table 15 and 16 show the outcome of using a Big 4 dummy as a moderating variable. Both tables show only non-significant results.

5.6 Big 4

In Table 17 I present the outcome of regressing all Big 4 dummies (Big 4-Big 4, Big 4 – Non-Big 4, Non-Big 4 – Non-Big 4). I find significant results for 2 audit quality measures. In column 1, 2 and 3 I find the outcome for regressing on the absolute total accruals audit quality measure. These columns show a significant negative relation (-0.020) for two Big 4 audit firms, a significant positive relation (0.026) for one Big 4 and one non-Big 4 audit firm and a highly significant negative relation for two non-Big 4 audit firms. This implies that being audited by one Big 4 and one non-Big 4 firm decreases the audit quality.

Columns 4 and 5 show a significant relation for the Big 4 dummy and the DD1 audit quality measurement model. Column 4 shows a significant positive relation (0.076) for two Big 4 audit firms, while column 5 shows a significant negative relation for the Big 4 dummy with one Big 4 and one non-Big 4 firm. It seems that both audit quality measures give different outcomes for the hypothesis that was set-up. I therefore confirm that audit quality is determined by the audit firms that conduct the audit. I cannot, however, conclude on the nature of the relationship. I recommend future research to find out this relationship.

Table 18 shows the outcome of regressing the dummy for at least one Big 4 firm on the audit quality measures. This results in one highly significant positive relation for the absolute total accruals (column 2).

5.7 Robustness

In this paper, robustness is tested by adding the control variable FeeRatio. Our datasample consists of 781 observations. Not all variables are respresented 781 times in the datasample. For the variable FeeRatio, there are 256 observations. This means that when a

(27)

model is tested, including FeeRatio as a control variable, and still shows a significant result with FeeRatio included, the relation can be regarded as robust. Also the fact that in this paper, six audit quality measures are used help us to establish robust results. Normally, for example in the paper of Cameran et al. (2016), in the case of audit quality, another measure is computed to measure audit quality, only with other variables used from another paper. In this paper, I use six audit quality measures, to ensure that when there is significant evidence for multiple models, I can assume that these results are robust.

Table 8 shows robust results, since four of six audit quality measures show significant results. Table 9 shows significant results for a model in which FeeRatio is included. Table 14 shows robust results, since three of six audit quality measures show significant results. Table 15 shows robust due to the control variable FeeRatio, but it also shows significant results for multiple models, as well. Table 18 shows robust results, since FeeRatio is included in the model.

6 Conclusion

Several previous studies have evaluated the individual auditors effects on audit quality, or whether joint audit increases audit quality. This paper combines both subjects into one, by testing auditor characteristics on audit quality in a joint audit setting.

Based on the results, I conclude that there is evidence to support all main hypotheses. The paper shows evidence for each main hypothesis due to the significant results in at least one of the sub hypotheses. Additionally, I found that when a company is audited by two male auditors or by two Big 4 audit firms, the audit quality increases. However, when the company is audited by two male auditors from two Big 4 firms, this relation reverses. Two male Big 4 auditors, decrease audit quality for a company.

I also find evidence for the influence of two Big 4 auditors, on the relation between the average age of the auditors and the audit quality. Individually, both two big 4 auditors, and an increasing average age of auditors increase the audit quality. However, if combined, this relation reverses to a significant negative impact on audit quality. Evidence for these relations are unprecedented, and therefore fill an important gap in the literature.

Nonetheless, not all results are in line with expectation. Table 18 shows evidence that having at least one Big 4 auditor, decreases audit quality. However, evidence from André et al. (2016) shows that having at least one Big 4 auditor increases audit quality. This is explained by the fact that when tested individually, having two Big 4 auditors and having one Big 4 auditor

(28)

show different (significant) results. Combining both into one measure, can be compared to taking a weighted average of the coefficients of the individual measures.

This paper contributes in a practical way as well, since it shows practitioners who to choose for their audits, how to match or on which audit to focus. In this paper the term practitioners should be specified: Practitioners are all people who practice a profession as auditor, or a profession that is somehow related to auditor. Therefore practitioners are: the companies that are subject of the audit, audit partners from audit firms, but regulators as well. Given the results from this paper, companies are better suited to make a choice for an audit firm: two Big 4, a combination of Big 4 and non-Big 4, or two non-Big 4 firms. Given the choice the company makes, the audit firms can match their partners to achieve statistically the highest possibility for a high-quality audit. Lastly, the regulators can inform themselves with the results from this paper, with which they can focus on joint audits where statistical probability for a high-quality audit is the lowest.

With regard to future research, I suggest taking a look at the relation between the differences of the age and tenure of both auditors. In this paper evidence is found (for both relations for multiple models) for a positive relation on audit quality in a joint audit setting. However, there is no scientific explanation found for this relation. I also recommend to increase the size of the data sample, especially for the FeeRatio and the six audit quality measures. Since the availability for big companies is higher than for small companies, it might be the case that the big companies are overrepresented in the sample. Since big companies, most of the time, go for big audit firms, it is harder to find representative results for relations involving two non-Big 4 firms. Another potential future research subject are the difference audit quality measures. It is hard to determine when a proxy for audit quality is sufficient, or in case of mixed results (as in Table 17), which results to follow. Therefore, it is important to extent the literature regarding the use of audit quality measures.

(29)

7 References

Alanezi, F., Alfaraih, M., Alrashaid, E., and Albolushi, S., 2012. Dual/joint auditors and the level of compliance with international financial reporting standards (IFRS-required disclosure): The case of financial institutions in Kuwait. Journal of Economic and Administrative Sciences, 28 (2): 109-129.

André, P., Broye, G., Pong, C. and Schatt, A., 2014. Are Joint Audits Associated with Higher Audit Fees? European Accounting Review, 25 (2): 245-274.

Audousset-Coulier, S., 2008, determinants of the voluntary disclosure of audit fees and audit pricing in a joint audit setting. PhD dissertation, HEC Paris.

Aobdia, D., Lin, C.J. and Petacchi, R., 2015. Capital market consequences of audit partner quality. The Accounting Review, 90 (6): 2143-2176.

Baldauf, J., and Steckel, R., 2012. Joint Audit and Accuracy of the Auditor's Report: An Empirical Study. International Journal of Economic Sciences and Applied Research, 5 (2): 7-42.

Bartov, E., Gul, F. A., and Tsui, J. S., 2000. Discretionary-accruals models and audit qualifications. Journal of Accounting and Economics, 30 (3): 421-452.

Beelde, de I., Gonthier-Besacier, N., and Mikol, A., 2009. Internationalizing the French auditing profession. The Accounting Historians Journal, 36 (1): 29-59.

Bertrand, M., and A. Schoar. 2003. Managing with style: The effect of managers on firm policies. Quarterly Journal of Economics, 118 (4): 1169–1208.

Brooks, C., 2002. Introductory econometics for finance, Cambridge University Press. Cameran, M., Campa, D. and Francis, J.R., 2016. Partner Identity and audit outcomes:

Evidence from the UK Engagement Partner Disclosure Rule, working paper. Church, B., Davis S. and McCraken S., 2008. The auditor’s reporting model: A literature

overview and research synthesis. Accounting Horizons, 22 (1): 69–90.

Dassen, R.J.M., 1989. De Leer van het Gewekte Vertrouwen: Agency avant la lettre? Maandblad voor Accountancy en Bedrijfseconomie, 63 (9): 341 – 352.

DeAngelo, L., 1981. Auditor size and audit quality. Journal of Accounting and Economics, 3 (3): 183-199.

Dechow, P.M., Sloan, R. and Sweeney, A., 1995. Detecting earnings management. The Accounting Review, 70 (2): 193-225.

(30)

Deegan, C., Blomquist, C., 2006. Stakeholders influence on corporate reporting: An exploration of the interaction between WWF-Australia and the Australian minerals industry. Accounting, Organizations and Society, 31 (5): 343 – 372.

DeFond, M. and Zhang, J., 2014. A review of archival auditing research. Journal of Accounting and Economics, 58 (2): 275-326.

Deng, M., Lu, T., Simunic, D.A. and Le, M., 2012. Do Joint Audits Improve or Impair Audit Quality? Journal of Accounting Research, 52 (5): 1029-1060.

El Assy, M.G., 2015. The Effect of Joint Audit on Audit Quality: Empirical Evident from Companies Listed on the Egyptian Stock Exchange. International Journal of Accounting and Financial Reporting, 5 (2): 2162-3082.

Eshleman, J. D., and Guo, P., 2014. Do Big 4 auditors provide higher audit quality after controlling for the endogenous choice of auditor?Auditing: A Journal of Practice and Theory,33 (4): 197-219.

European Commission, 2010. Green paper – audit policy: Lessons from the crisis, Brussels. Retrieved from

http://ec.europa.eu/internal_market/consultations/docs/2010/audit/green_paper_audit_en.pdf Fama, E.F.,1980. Agency Problems and the Theory of the Firm. Journal of Political

Economics, 88 (2): 288 – 307.

Fama, E.F., Jensen, M.C., 1983. Agency Problems and Residual Claims. The Journal of Law and Economics, 26 (2): 327 – 349.

Francis, J. R., 2004. What do we know about audit quality? The British Accounting Review, 36 (4): 345-368.

Francis, J.R. and Wang, D., 2008. The joint effect of investor protection and Big 4 audits on earnings quality around the world. Contemporary Accounting Research, 25 (1): 157-191.

Francis, J.R., Richard, C. and Verstraelen, A., 2009. Assessing France’s Joint Audit Requirement: Are two heads better than one? Auditing: A journal of practice and Theory, 28 (2): 35-63.

Francis, J.R. and Yu, M.D., 2009. Big 4 Office Size and Audit Quality. The Accounting Review, 84 (5): 1521-1552.

Francis, J. R., 2011. A framework for understanding and researching audit quality. Auditing: A Journal of Practice and Theory, 30 (2): 125–152.

Francis, J. R., Pinnuck, M. L., and Watanabe, O., 2013. Auditor style and financial statement comparability. The Accounting Review, 89 (2): 605-633.

(31)

Frankel, R. M., Johnson, M. F., and Nelson, K. K., 2002. The relation between auditors' fees for nonaudit services and earnings management. The Accounting Review, 77 (1): 71-105.

Freeman, R. E., Harrison, J., Wicks, A., Parmar, B., Colle, S. de, 2010. Stakeholder theory: The state of the art. Cambridge: Cambridge University Press.

Ghosh, A. and Moon, D., 2005. Auditor tenure and perceptions of audit quality. The Accounting Review, 80 (2): 585-612.

Gonthier-Besacier, N., and Schatt, A., 2007. Determinants of audit fees for French quoted firms. Managerial Auditing Journal, 22 (2): 139-160.

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.

Hoitash, R., Markelevich, A. and Barragato, C.A., 2007. Auditor fees and audit quality. Managerial Auditing Journal, 22 (2): 761-786.

Houpin, C. and Bosvieux, H., 1935. Traite general theorique et pratique des societes civiles et commerciales et des associations.

Jensen, M.C., Meckling, W.H., 1976. Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure. Journal of Financial Economics, 3 (4): 305 – 360.

Jianakoplos, N.A. and Bernasek, A., 1998. Are women more risk averse? Economic inquiry, 36 (4), 620-630.

Johnson, V.E., Khurana, I.K. and Reynolds, J.K., 2002. Audit-firm tenure and the quality of financial reports. Contemporary Accounting Research, 19 (4): 637-660.

Knechel, W. R., and Vanstraelen, A., 2007. The relationship between auditor tenure and audit quality implied by going concern opinions. AUDITING: A journal of practice & theory, 26 (1): 113-131.

Knechel, R.W., Vanstraelen, A. and M. Zerni, 2015. Does the identity of engagement partners matter? An analysis of audit partner reporting decisions. Contemporary Accounting research, 32 (4): 1443-1478.

Lawrence, A., Minutti-Meza, M., and Zhang, P., 2011. Can Big 4 versus non-Big 4 differences in audit-quality proxies be attributed to client characteristics? The Accounting

Review, 86 (1): 259-286.

Levine, R., 2005. Law, endowments and property rights.The Journal of Economic Perspectives,19 (3): 61-88.

Maxfield, S., Shapiro, M., Gupta, V., and Hass, S., 2010. Gender and risk: women, risk taking and risk aversion.Gender in Management: An International Journal,25 (7), 586-604.

(32)

Milok, A., 1993. The evolution of Auditing and the Independent Auditor in France. European Accounting Review, 2 (1): 1-16.

Paugam, L., and Casta, J. (2012). Joint Audit, Game Theory, and Impairment-Testing Disclosures. Working paper available at: i .

Praquin, N., 2012. Commercial legislation and the emergence of corporate auditing in France, 1856–1935. Accounting History Review, 22 (2): 161-189.

Ratzinger-Sakel, N., Audousset-Coulier, S., Kettunen, J., and Lesage, C., 2013. Joint Audit: Issues and Challenges for Researchers and Policy-Makers. Accounting in Europe, 10 (2): 175-199.

Rajgopal, S., Srinivasan, S., and Zheng, X., 2015. Measuring audit quality. Working paper. Schubert, R., Brown, M., Gysler, M., and Brachinger, H. W., 1999. Financial

decision-making: are women really more risk-averse?The American Economic Review,89 (2), 381-385.

Skinner, D.J. and Srinivasan, S., 2012. Audit quality and auditor reputation: Evidence from Japan. The Accounting Review. 87 (5): 1737-1765.

Swanquist, Q.T. and Whited, R.L., 2015. Do clients avoid ‘contaminated’ offices? The economic consequences of low-quality audits. The Accounting Review, 90 (6): 2537-2570.

Watts, R.L., Zimmerman, J.L., 1983. Agency problems, auditing, and the theory of the firm: some evidence. The Journal of Law and Economics, 26 (3): 613 – 633.

Wickens, M.R., 1972. A note on the use of proxy variables. Econometrica: Journal of the Econometric Society, 759-761.

Zerni, M., Elina, H., Jarvinen, T., and Niemi, L., 2012. Do joint audits improve audit quality? Evidence from voluntary joint audits. European Accounting Review, 21 (4): 731-765.

(33)

APPENDIX 1

Audit market share distribution for the 5 biggest EU countries and the EU average

Country Big 4 market share Market share firms >2% Number of firms >2% Percentage of other auditors France 46% 60% 6 40% Germany 51% 66% 8 34% Italy 80% 95% 8 5% Spain 81% 91% 7 9% United Kingdom 59% 85% 8 15% European Union 61% 77% 8 23% Source: http://www.auditanalytics.com/blog/eu-auditor-market-share-2016/

(34)

APPENDIX 2

(35)

Table 1

Variable definitions

Variable Explanation

Dependent variables

Restat Dummy variable that takes the value of 1 if there was a restatement.

AbsTotAccruals The absolute total accruals, scaled by lagged total assets. Computed by Healy’s model

MD1 The Modified Jones model

MD2 The long-term Modified Jones model

DD1 The Dechow and Dichev base model

DD2 The long-term Dechow and Dichev modified model

Independent variables

DB4B4 Dummy variable that takes the value of 1 if both auditors work for a Big 4 audit firm

DB4NB4 Dummy variable that takes the value of 1 if one of two auditors work for a Big 4 audit firm

DNB4NB4 Dummy variable that takes the value of 1 if both auditors do not work for a Big 4 audit firm

DGenMaleMale Dummy variable for gender that takes the value of 1 if both auditors are male DGenMaleFemale Dummy variable for gender that takes the value of 1 if one of two auditors is

male

DGenFemaleFemale Dummy variable for gender that takes the value of 1 if none of both auditors are male

AudAgeDifference Variable for the agegap between both auditors AudAgeAverage Average age for both auditors

AudTenDifference The difference in tenure for both auditors AudTenAverage The average tenure for both auditors

Control variables

DY Dummy for controlling the year effects

DIndustry Dummy for the industry in which a company operates

DpoL Dummy variable for the presence of losses, variable takes the value of one when there is a loss in that year

Leverage Total debt divided by total assets for that year FeeRatio The audit fee divided by the total assets for that year

ROA Net income divided by total assets

Size Log size of total assets

SalesGrowth The growth in sales compared to last year (salest – salest-1) / salest-1

(36)

Table 2

Descriptive statistics of audit firm, offices, and audit partners Panel A: Number of auditors grouped by client-year observations

1-5 client-years 38 (47,5%) 16 (36,4%) 369 (80,9%) 5-10 client-years 24 (30%) 13 (29,6%) 84 (18,5%) 11-15 client-years 7 (8,8%) 2 (4,5%) 2 (0,4%) 16-20 client-years 3 (3,8%) 4 (9,1%) 1 (0,2%) 21-100 client-years 0 (0%) 6 (13,6%) 0 (0%) 101+ client-years 8 (10%) 3 (6,8%) 0 (0%) Total 80 (100%) 44 (100%) 456 (100%)

This table represents the composition of our data on audit firms, audit offices, and individual audit partners grouped by client and client-year observations.

(37)

Table 3

Descriptive statistics, in order of dependent, independent and control variables

Variable Name Obsverations Mean Standard

deviation Min (% 0) Max (% 1) Restatement 273 0.04395 0.20537 ( 95.7% ) ( 4.3% )

Abs Total Accruals 629 0.07996 0.10257 0.00122 0.52444

MD1 401 0.08968 0.14380 0.00163 0.75869 MD2 619 0.05206 0.04734 0.00152 0.19894 DD1 647 0.05833 0.06390 0.00141 0.29395 DD2 596 0.05214 0.05660 0.00093 0.25355 DB4B4 781 0.22279 0.41638 (77.8% ) ( 22.2% ) DB4NB4 781 0.59539 0.49113 ( 40.5% ) ( 59.5% ) DNB4NB4 781 0.18181 0.38594 ( 88.9% ) ( 18.1% ) DGenMaleMale 781 0.82714 0.37836 ( 17.3% ) ( 82.7% ) DGenMaleFemale 781 0.16133 0.36807 ( 83.9% ) ( 16.1% ) DGenFemaleFemale 781 0.01152 0.10679 ( 98.9% ) ( 1.1% ) AudAgeDifference 424 7.74528 5.87708 0 24 AudAgeAverage 424 43.2818 6.36917 35.5 58.5 AudTenDifference 349 18.4541 6.44545 1 35.5 AudTenAverage 349 7.98567 5.51459 5 23 Presence of losses 781 0.24967 0.43310 Leverage 779 0.23886 0.17050 0.00143 0.85902 Feeratio 256 0.00065 0.00041 0.00007 0.00190 Return on Assets 779 0.01416 008466 -0.30544 0.19508 Sales Growth 628 0.05986 0.21555 -0.49922 0.86792 Size 779 13.63419 2.30134 8.18868 19.3527

This table reports descriptive statistics of the variables used in our empiricial models. Please see Appendix 1 for variable definitions.

Joint audit in the French setting: How to team up best?