E V I DE N C E F RO M T H E DU T C H A ND B E L GI A N M A R K E T S F O R P U B LI C I NT ER ES T C O M PA N Y AU D IT S
Abstract
This paper studies the relationship between audit firm tenure and audit quality among Dutch and Belgian public interest companies in a non-mandatory audit firm rotation environment, using total accruals and abnormal working capital accruals as a proxy for earnings management. The study is motivated by calls from the European Commission for limits on the maximum length of auditor tenure, arguing that lengthy auditor-client relationships pose a threat to the auditor’s independence and, hence, to the audit quality provided. The results, after controlling for the effects of client firm size, leverage, cash flow from operations, growth, and auditor type (BIG 4 or non-BIG 4), do not suggest that lower audit quality occurs with extended audit firm tenure. In contrast, there is some evidence that suggests that audit quality may in fact improve slightly with extended audit firm tenure, although the level of audit quality is only marginally affected by the duration of the audit firm-client relationship.
The results therefore do not support the contention that mandatory auditor rotation will improve audit quality, although future research on the relationship between audit firm tenure and audit quality in a mandatory rotation environment could provide different insights.
Author: Daan Vermei
Student Number: S2209314
Address: Resedastraat 1-6 9713TN Groningen
Phone: +31 614428897
E-mail: daan@vermei.nl
Institute: Rijksuniversiteit Groningen
Faculty: Economics and Business
Program: MSc. Accountancy
Course: Masterthesis Accountancy EBM869A20
Field of research: The Accountancy Profession Supervisor: Prof. Dr. R.L. ter Hoeven
Date June 2015
Table of contents
1. Introduction ... 2
1.1 Audit regulation ... 2
1.2 Scientific Contribution ... 3
2. Theoretical Framework... 5
2.1 Background: the need for high-quality auditing ... 5
2.1.1 Auditing as a monitoring device in agency conflicts ... 5
2.1.2 The confidence-inspiring role of auditing ... 6
2.2 Audit quality ... 6
2.2.1 Auditor ability ... 7
2.2.2 Auditor independence... 7
2.3 Audit firm tenure ... 7
2.3.1 Background: the relationship between auditor tenure and audit quality ... 8
2.4 Hypothesis development ... 8
3. Research Design ... 10
3.1 Sample ... 10
3.2 Measuring earnings management... 10
3.2.1 Total accruals model ... 11
3.2.2 Defond & Park model ... 12
3.3 Explanatory variable ... 12
3.4 Control variables ... 13
4. Results ... 14
4.1 Univariate results ... 14
4.2 Multivariate results ... 15
4.2.1 Nonparametric test ... 16
4.2.2 Linear regression ... 16
4.3 Sensitivity analysis ... 17
4.3.1 Sensitivity analysis: nonparametric test ... 17
4.3.2 Sensitivity analysis: linear regression ... 18
4.4 Additional analyses ... 19
5. Conclusion ... 20
6. References ... 23
1. Introduction
In its recent reports, the Netherlands Authority for the Financial Markets (AFM) has identified substantial problems and risks regarding the quality of auditors’ reports on the financial statements of public interest companies (AFM 2013 & AFM 2014). At the same time, as a result of recent financial scandals and audit failures, it became clear that the gap between what society expects of auditors and what is actually supplied by auditors is widening, resulting in a loss of confidence inspired by the accountant and public debate about the role and functioning of the accounting profession, a phenomenon that was already identified more than 80 years ago by Limperg (1932).
1.1 Audit regulation
On a European level, there has recently been criticism of the auditing profession as well, in the form of the Barnier commission’s Green Paper (European Commission, 2010), discussing the role and scope of the audit in the context of the financial market regulatory reforms after the 2008 financial crisis. One of the key focal points is auditor independence. It is argued that the threat of familiarity between the auditor and the client may not be overcome by key audit partner rotation alone, but that mandatory audit firm rotation should be considered to this end.
The Commission acknowledges the potential loss of client-specific knowledge as a result of audit firm rotation, which is to be taken into consideration. Rotation rules should include the mandatory rotation of key audit partners as well as audit firms, to prevent partners from changing audit firms to take audit clients along with them, which would then offer limited improvements to independence. The recommendations regarding audit firm and audit partner rotation from the Green Paper were adopted in regulation by the European Commission and will come into effect in 2016. The new European regulation requires its members’ public interest companies to switch between audit firms after a maximum period of ten years, with possible exceptions in case of publicly tendering the offer or in case of joint audits.
Additionally, key audit partners will not be allowed to carry out the statutory audit of a single company for more than seven consecutive years, after which a ‘cooling-off period’ of at least four years will apply (Regulation (EU) No 537/2014).
In the U.S., the Sarbanes-Oxley act (2002) mandates key audit partner rotation after five consecutive years for public interest companies, but poses no restrictions on the duration of the relation between audit firm and auditee, as the U.S. The General Accounting office (GAO) concludes that “mandatory audit firm rotation may not be the most efficient way to strengthen auditor independence’’ (GAO, 2003). In The Netherlands, new regulations will come into force in 2016, stretching the maximum period that an audit firm can be the auditor of a public interest company to ten years consecutively instead of the eight year period that was proposed before, bringing Dutch regulation in line with European regulation. This regulation will be combined with a mandatory rotation of the key audit partner after five consecutive years and a
‘cooling-off period’ of four years (De Vries et al., 2014). In Belgium, the new European
regulations will have to be incorporated in the law as well, and effective as of 2016. Until this
is the case, norms issued by the Belgian regulator of auditors (Instituut van de
Bedrijfsrevisoren, IBR) require audit partner rotation after six consecutive years of auditing a
client, while audit firm rotation is not mandatory until the incorporation of the European
regulations. Additionally, a cooling-off period of at least two years applies to the audit
partner-audit client relationship (IBR, 2007). The IBR has issued a memorandum to the Belgian Government (IBR, 2014) advising it to choose for an audit firm rotation period of no less than ten years, applicable to public interest companies only.
It is apparent that various regulatory institutions have implemented restrictions on the duration of auditor-client relations, in an effort to improve audit quality. This quality improvement is argued to result from improved audit partner and audit firm independence, lower risks of auditor complacency, and enhanced skepticism by introducing a fresh view of a new accountant (Shockley, 1981). An opposite argument, however, takes into consideration the possibility that the costs associated with mandatory auditor rotation may actually exceed benefits. For example, auditors are still acquiring crucial knowledge about the company they audit in the early years of the client-auditor relationships, possibly lowering audit quality (Carcello & Nagy, 2004a).
The argument for audit partner rotation over audit firm rotation is the retention of the audit firm’s institutional knowledge of the auditee, while periodically offering a fresh perspective of the audit client by assigning a new audit partner, simultaneously improving independence (PWC, 2012). However, consecutive audit partners are allowed to be from the same audit firm and other members of the audit team are potentially retained beyond the audit partner switch. In the case of audit firm rotation, the whole audit team, including the audit partner, would be rotated, potentially offering a stronger improvement to auditor independence. Audit firm rotation, compared to audit partner rotation, would also suggest a bigger loss of institutional knowledge, however. I therefore expect the suggested detrimental effect of losing institutional knowledge about the auditee and the favorable effect of improved independence on audit quality to be stronger for audit firm rotation than for audit partner rotation. As both audit firm and audit partner rotation will be mandatory as of 2016 in the European Union, I will focus on the type of auditor rotation that I expect to have the strongest effect on audit quality, that is, audit firm rotation.
Focusing on the auditing market for Dutch and Belgian public interest companies, I ask: what are the effects of audit firm tenure and audit firm rotation on audit quality in the Dutch and Belgian markets for public interest company audits?
1.2 Scientific Contribution
Previous studies generally consider audit firm tenure of up to three years to be short tenure, and eight years or more to be long tenure. Some studies classify more than three years as
‘longer’ tenure, while others make a distinction between the medium term (generally between four and six to eight years) and the long term (generally between seven and nine years or more) (Johnson et al., 2002; Geiger & Raghunandan, 2002; Carcello & Nagy, 2004a; Gul et al., 2007; Knechel & Vanstraelen, 2007; Boone et al., 2008). The results of those studies provide mixed results. Some actually find that longer audit partner and firm tenure actually increase earnings quality (Myers et al., 2003; Chen et al., 2008) and that long audit firm tenure causes investors and information intermediaries to perceive earnings quality to be higher (Ghosh & Moon, 2005). Others do not find a strong relationship between long audit firm tenure and audit quality (Geiger & Raghunandan, 2002; Johnson et al., 2002; Carcello &
Nagy, 2004a; Knechel & Vanstraelen, 2007). There is, however, evidence of lower audit
quality with more lengthy audit firm tenure as well (Kealey et al., 2007; Boone et al., 2008).
Regarding short audit firm tenures, the results of previous studies clearly point in the direction of lower audit quality when auditor tenure is short (Geiger & Raghunandan, 2002; Johnson et al., 2002; Carcello & Nagy, 2004; Davis et al. 2009).
Some studies provide interesting implications apart from the main relationship between audit firm tenure and audit quality. Davis et al. (2009) compare the pre-SOX period to the post-SOX period and find that, consistent with the evidence above, audit quality was lower in both short and long tenure periods (compared to the medium term) in the pre-SOX period, but that this relationship disappeared after SOX (2002), which could have been caused by increased scrutiny and the threat of legal sanctions introduced by SOX. Lim & Tan (2010) examine the effect of fee dependence and industry specialization on the relationship between audit firm tenure and audit quality. They suggest that longer tenure increases industry specialization and consequently audit quality, and that strong fee dependence weakens this relationship.
At this point, it is worth noting that the previous research has been conducted in order to investigate if mandatory audit firm rotation could enhance audit quality through increased auditor independence. As little data is available to test the effects of mandatory audit firm rotation on audit quality, most studies have been conducted in a non-mandatory audit firm rotation environment. I intend to contribute to the scientific field by extending the literature on the relationship between audit firm tenure and audit quality, by investigating this relationship in the Netherlands and Belgium. Most previous studies were conducted in a pre- SOX, US setting (Johnson et al. 2002; Geiger & Raghunandan, 2002; Carcello & Nagy, 2004a; Davis et al. 2009), though not exclusively. Studies by Chen et al. (2008) and Knechel
& Vanstraelen (2007), for example, examine the relationship between audit firm and partner tenure and audit quality in Taiwan and Belgium respectively. The implementation of SOX (2002) and stricter oversight over the accounting profession following various accounting failures (e.g. Enron, 2001; Parmalat, 2003; Lehman Brothers, 2008) could well have changed the factors affecting audit quality. Others focus solely on big audit firms and therefore cannot examine the effects of audit firm size on the level of independence and audit quality (Boone et al., 2008; Johnson et al., 2002).
New regulation, effective as of 2016 in The Netherlands and Belgium, will limit the amount of years that an audit firm can consecutively audit the same company to ten years, in line with European Union regulation. Considering the mixed results of previous literature, this raises the question of whether this limit will actually improve audit quality and, if yes, whether the limit of ten years would be optimal. Is longer audit firm tenure associated with lower audit quality through impaired independence, and if yes, at what length of tenure does this occur? Or do audit firms become more knowledgeable about their audit clients as their tenure extends, making them more able to deliver greater audit quality with limited detrimental effects as a result of supposedly lower independence.
The remainder of the paper is organized as follows. Section two discusses the theoretical framework that underpins the paper and states the hypotheses to be investigated.
Section three elaborates on research design to allow for testing of the hypotheses. Section four
presents the results, and section five concludes the paper.
2. Theoretical Framework
This section discusses the theories that underlie and give direction to the research question posed in the previous section. It consecutively discusses the need for high-quality audits, the constituents of audit quality, and auditor tenure. The section concludes by stating the hypotheses that will be tested.
2.1 Background: the need for high-quality auditing
In the current research setting, the external audit is mandatory, but the firm has the discretion to choose its own auditor at any time, thereby potentially influencing audit quality and thus the potential of insiders of the firm to manage earnings. The agency theory (Jensen &
Meckling, 1976) and the theory of inspired confidence (Limperg, 1932) are two theories explaining the need for high-quality auditing that I consider relevant to the posed research question in the chosen research setting.
2.1.1 Auditing as a monitoring device in agency conflicts
Jensen & Meckling (1976) identify conflicts of interest between utility-maximizing agents and principals. These conflicts can arise in all situations requiring cooperative efforts in which the involved parties have incentives to maximize their own utility. Relevant examples of such agency problems are those between managers and shareholders (Jensen & Meckling ,1976), between controlling shareholders and minority shareholders (La porta et al., 2000), and between shareholders and debtholders (Leland, 1998). In the first agency problem, as some decision-making authority is delegated to the agent by the principals, the agent will have the opportunity to act in its own best interest at the cost of the interests of the principal (Jensen &
Meckling, 1976). In the second agency problem, controlling shareholders essentially are insiders, like managers, and have decision-making power enabling them to control and use assets for purposes that are not in the interest of outside investors (La Porta et al., 2000).
Finally, in the third agency problem, equity holders can increase risk after a debt is in place, diverting value form debtholders to themselves (Leland, 1998).
One way in which the principal can limit selfish behavior by the agent, is to monitor the actions of the agent to reduce its opportunity to act at the cost of the principal. By reducing information asymmetries between agents and principals, auditing is such a monitoring device that limits the incentives of the agents to act at the expense of the principal. For the first two agency conflicts, this mainly pertains to misappropriation of assets by the agent (controlling shareholder or management) at the expense of outside shareholders. Considering the third agency problem, auditing can reduce the agent’s incentives to act at the expense of debtholders by making sure that covenants related to the debt contract with the firm are not violated (Jensen & Meckling, 1976).
Agency costs consist of the monitoring costs, bonding costs, and the residual loss.
Monitoring costs are the costs that the principal incurs to limit the agent’s actions that harm the interests of the principal, for example, by mandating an audit of the financial statements.
Bonding costs are costs incurred by the agent to guarantee the principal that he will not act at
the expense of the principal. In the present environment, the annual audit is mandatory for
public interest entities, both in The Netherlands and in Belgium. This begs the question
whether principals and agents can actually exert influence over the amount of monitoring and
bonding through external auditing. Since external audits are conducted by different auditors, the mandatory external audits could still inhibit differences in the effectiveness of auditing as a monitoring device to mitigate opportunistic behavior of insiders. High quality auditors are more likely to detect these opportunistic kinds of behavior by management. Therefore, high quality auditing deters earnings management, because detection and revelation of misreporting damages management’s reputation and reduces firm value (Becker et al. 1998).
2.1.2 The confidence-inspiring role of auditing
Limperg (1932) investigates the economic function of the auditing accountant and argues that its function lies in the verification of the information presented to the community by the parties with interests in the firm. He argues that this function of confidential agent of the community at large is the most important function of the accountant, more important than his function as confidential agent of his principals (i.e. the shareholders). As financing of production has become a societal issue, the community is no longer satisfied with the accountability provided by the supervisory directors, who lack the amount of time and expertise to effectively and efficiently carry out the audit of increasingly complex organizations. An internal audit function could solve these problems of time and expertise, but is only economically efficient for very large companies. Additionally, the use of the public accountant is preferable to such an internal function because the variety of organizations that the accountant audits strengthens his competence, and because the independence of the external audit office is beneficial to the technical efficiency of the audit (Limperg, 1932).
The function of the confidential agent therefore arises because the community has too little confidence in the communication of, and opinion on, firm information provided by officials other than the confidential agent. The gap between the amount of confidence placed in the accounting function by the community and the actual fulfillment by the accountant of its function arises from either excessive expectations by the community or from the accountant falling short of fulfilling his function (Limperg, 1932). This function stems from the need for expert and independent auditors dedicating their time to auditing the financial statements of the firm, in order provide some level of assurance that the presentation by the firm of its position and performance is accurate. The extent to which the auditor issues the appropriate opinion on the financial statements, and therefore fulfills his function, is the audit quality delivered by the auditor.
2.2 Audit quality
The theories of agency costs and inspired confidence support the notion that audit quality is essential to the value of the auditing function as a monitoring device on behalf of stakeholders, shareholders, and the community at large. Audit quality is a function of (1) the ability of the auditor to detect material misstatements in the client’s financial statements, and (2) the probability that the auditor will correct or report such an error (DeAngelo, 1981a;
DeAngelo, 1981b; Watts & Zimmerman, 1981). This definition of audit quality mainly relates
to technical audit quality, following the agency theory in which audit quality is a determinant
of the effectiveness of auditing as a monitoring device. Limperg (1932) mainly refers to
perceived audit quality, where society is the percipient of audit quality. We could argue,
however, that both factors are interrelated, as impaired technical audit quality does affect
perceived audit quality when news of accounting failures reaches society. Therefore, this paper will focus on audit quality in the technical sense.
2.2.1 Auditor ability
Given a misstatement in the financial reports, the probability that an audit firm will discover that breach in the client’s accounting system depends on the audit firm’s ability to make the discovery (DeAngelo, 1981b). A recent report by the International Auditing and Assurance Standards Board (IAASB) views appropriate audit quality to be achieved when the users of the financial statements can rely on the auditor’s opinion on the statements, based on sufficient appropriate audit evidence. The engagement team collecting such evidence should:
exhibit appropriate values, ethics, and attitudes; be sufficiently knowledgeable and experienced; have sufficient time allocated to perform the audit work; apply a rigorous audit process and quality control procedures; provide valuable and timely reports; and interact appropriately with a variety of different stakeholders. Of these requirements, factors relating to the auditor’s ability to discover a material misstatement arguably are knowledgeability, experience, allocated time, and the audit process and quality control (IAASB, 2013).
2.2.2 Auditor independence
Auditor independence essentially is the probability that an auditor will report a misstatement in the client’s accounts, given that such a misstatement exists and is discovered by the auditor (DeAngelo, 1981b). DeAngelo (1981a) argues that the incentives of the auditor to tell the truth, when doing so disfavors his client, determine the probability that the auditor reports a discovered breach. The revenues earned by the auditor and exceeding avoidable costs are termed quasi-rents by DeAngelo (1981a). Quasi-rents arise in multi-period settings because auditors invest in the client relationship in a certain period in the expectation that future revenues exceed future costs. Essentially, when no client-specific quasi-rents are earned from a client, an auditor is indifferent to whether to continue or terminate the client relationship.
This indifference causes the auditor to lack an incentive to conceal a discovered breach.
However, due to start-up costs in audit technology and transaction costs of switching auditors, incumbent auditors can raise future fees above the avoidable costs of producing audits, causing auditors to earn quasi rents on their clients (DeAngelo, 1981a). Because of the significant costs to the audit firm of switching between clients, audit firms may be more compliant with management’s demands or choose not to report an error in the financial statements. Thus, the higher the amount of quasi-rent streams earned by the audit firm, the stronger the threat to the audit firm’s independence.
2.3 Audit firm tenure
The purpose of this paper is to study the effects of audit firm tenure and rotation on the level of audit quality supplied by the audit firm. Extended audit firm tenure causes the audit firm to gain a better understanding of the client through client-specific expertise. On the other hand, longer tenure can cause overfamiliarity with the client, potentially reducing vigilance.
Additionally, high client-specific quasi rent streams can cause the audit firm to accept more of
the client’s demands, lowering independence (Lim & Tan, 2010). The differing levels of
client-specific knowledge at different levels of tenure affect the ability part of the audit quality
function. Overfamiliarity and the amount of quasi-rent streams in turn affect auditor independence in the audit quality function. This section subsequently discusses the previous research on the relationship between auditor tenure and audit quality, and develops the hypotheses that I aim to test in this study.
2.3.1 Background: the relationship between auditor tenure and audit quality
While regulators and oversight institutions often use the intuitively appealing argument that longer auditor tenure affects auditor independence negatively due to impaired vigilance and overfamiliarity (SOX, 2002; European Commission, 2010; PCAOB, 2006), much studies find that longer audit firm tenure does not in fact have a negative impact on audit quality in non- mandatory auditing environments (Johnson et al. 2002; Myers et al. 2003; Carcello & Nagy, 2004a, Knechel & Vanstraelen, 2007; Chen et al. 2008). Some studies even find improved audit quality with longer audit firm tenure, both in the technical sense (Myers et al. 2003) and as perceived by investors and information intermediaries (Ghosh & Moon, 2005). There are some studies that do find that audit quality is impaired after longer periods of audit firm tenure. Boone et al. (2008) suggest that audit quality is impaired after approximately 13 years of consecutive audit firm tenure, and Davis et al. (2009) find impaired audit quality after the audit firm-client relation extends beyond 15 years. Additionally, Kealey et al. (2007) find that audit quality as perceived by successor audit firms was lower for clients that had had longer tenure with their previous audit firm. While these studies prove to be inconclusive as to whether longer audit firm tenure impairs audit quality through reduced independence, the results on short audit firm-client relationships generally suggest that audit quality is impaired in short-tenure situations (Geiger & Raghunandan, 2002; Johnson et al. 2002; Carcello &
Nagy, 2004a; Davis et al. 2009)
The different expected and found effects of audit firm tenure on audit quality in the short-term of the audit firm-client relationship and the long-term of the audit firm-client relationship implies that a non-linear, downward-opening parabolic relationship between audit firm tenure and audit quality may exist. Since European regulation requires member states to impose the new auditor rotation rules as of 2016, the question arises which maximum-tenure period would be most appropriate. The European commission gives its member states some discretion over the imposed maximum period of the audit firm-client relationship, but it cannot be longer than ten years. This period may be too short overall, or shorter periods may be more appropriate across different countries.
2.4 Hypothesis development
The start-up costs associated with initial audit engagements are significant, as the new audit
firm has to become familiar with the client’s organization. The lack of knowledge of the
client’s business, operations, systems, and accounting policies could therefore lower audit
quality in the early years of the audit firm-client engagement because the audit firm’s ability
to detect material misstatements is lower (Carcello & Nagy, 2004a). As the audit firm
becomes more and more familiar with the client firm, its ability to detect financial reporting
misstatements increases. Longer audit firm tenure is therefore expected to have a positive
effect on audit quality through the first part of the audit quality function, i.e. the auditor’s
ability to detect material misstatements increases with extended audit firm tenure.
H1: A greater length of audit firm tenure will cause higher audit quality
On the other hand, longer audit firm tenure may also affect the second part of the audit quality function, as it may reduce the probability that the audit firm will correct or report material misstatements, if found (i.e. the audit firm becomes less independent). The level of auditor independence effectively is the probability that the auditor will report a breach in the client’s accounting system if it discovers such a breach (DeAngelo, 1981a; DeAngelo 1981b).
Shockley (1981) states that audit firms may become less independent as audit firm tenure increases, because they become complacent, apply less rigorous audit procedures, lack innovative audit initiatives, and develop confidence in their client as the amount of previous years without material misstatements grows. DeAngelo (1981a) argues that auditors may view their relationship with a client as lasting in perpetuity and that audit firms therefore view their income from the audit fees received from that client as perpetual rents. As the auditor expects to keep earning these rents, it may become less independent. Additionally, a new audit firm may introduce a new, skeptical view of the organization and thereby improve audit quality (Carcello & Nagy, 2004a). While several studies did not find support for the hypothesis that long audit firm tenure impairs audit quality, most of them were conducted in a pre-SOX, US setting (Johnson et al. 2002; Geiger & Raghunandan, 2002; Carcello & Nagy, 2004a; Davis et al. 2009). The implementation of SOX (2002) and stricter oversight over the accounting profession following various accounting failures (e.g. Enron, 2001; Parmalat, 2003; and Lehman Brothers, 2008) could well have changed the factors affecting audit quality.
Additionally, the audit markets I investigate are non-US, and could therefore produce different results compared to previous studies undertaken in a US setting. Furthermore, Boone et al.
(2008) and Davis et al. (2009) do find that perceived audit quality decreases after a certain length of audit firm-client tenure, 13 and 15 years respectively.
H2: Long audit firm tenure will cause lower audit quality compared to short and medium audit firm tenure
As there clearly is rather strong evidence for the short-term learning effect of audit firm tenure
on audit quality (Geiger & Raghunandan, 2002; Johnson et al., 2002; Carcello & Nagy,
2004a; Davis et al. 2009), and to a lesser extent for the long-term impaired-independence
effect of audit firm tenure on audit quality (Boone et al. 2008; Davis et al., 2009), I expect the
relationship between audit firm tenure and audit quality to be positive (hypothesis 1) in the
short run, but potentially reversed in the long run (hypothesis 2). This perspective of audit
firm tenure being beneficial to audit quality in the short run, but detrimental to audit quality
after a certain period of time opens up the possibility for a nonlinear relationship between
audit firm tenure and audit quality, in which there is an optimum level of audit firm tenure
where the sum of the knowledge-acquisition and impaired-independence effects causes audit
quality to be the highest. Boone et al. (2008) do find such a parabolic relationship between
audit firm tenure and perceived audit quality. Therefore, accepting both hypothesis 1 and 2
would imply a nonlinear relationship between audit firm tenure and audit quality.
3. Research Design
3.1 Sample
The sample consists of firms listed on either the Euronext Amsterdam or the Euronext Brussels stock exchange. As of March 2015, Orbis identifies 138 firms mainly listed on Euronext Amsterdam and 161 firms mainly listed on Euronext Brussels. The sample period stretches the firm years 2004-2013, with a limited amount of sample firm years for 2004, because this data had to be handpicked to enlarge the amount of firm years with short tenure.
The population consists of 299 listed firms before elimination of sample elements. The first elimination of sample elements excludes companies with operating revenues less than one million USD as of 2013, companies with no recent financial data, and Public authorities, states, and governments (ORBIS, Bureau van Dijk). These elimination processes result in an intermediary sample of 90 firms listed on Euronext Amsterdam and 128 firms listed on Euronext Brussels. Previous research (Zerni et al., 2012) further shows that joint audits have effects on audit quality. To prevent noise in the sample, I thus eliminate 7 firms that had joint auditors, instead of a single auditor, during the sample period. Firms with US SIC codes 6000 through 6999 are excluded from the sample because their unique working capital structures render it problematic to measure earnings management using accruals (Becker et al., 1998;
Dechow et al., 2003; Maijoor & Vanstraelen, 2006). This additional elimination of 39 financing firms results in an intermediary sample of 164 firms. Eliminations resulting from a lack of data for firm years during the sample period amounted to 68, leaving a final sample of 96 firms.
3.2 Measuring earnings management
The amount of reported earnings consists of the sum of net operating cash flows and accruals.
The role of accruals in this process is to adjust the recognition of cash flows over time to better represent firm performance (Dechow & Dichev, 2002). The actual amount of cash flows is independent from financial reporting choices and is therefore hard to manipulate by management. The accrual portion of earnings is partly dependent on financial reporting choices and can therefore be manipulated by management, given the allowed discretion over financial reporting (Hooghiemstra et al., 2008). In practice, it is hard to measure the proportion of accruals that is economically justified and the amount that is not, since only management itself ultimately knows which part of reported accruals is economically justified and which part it has manipulated to manage reported earnings in its desired direction.
The scientific literature uses a variety of measures to measure the discretionary part of total accruals. Early models, like the Healy model (Healy, 1985) and the DeAngelo model (DeAngelo, 1986) use total accruals as a proxy for expected nondiscretionary accruals.
Dechow et al. (1995) argue that these models fail to model the impact of economic
circumstances on the amount of accruals, resulting in the omission of relevant variables and,
consequently, inflated standard errors. A widely used discretionary accruals model is the Jones
model (1991), which takes into consideration the effect of changes in a firm’s economic
circumstances on nondiscretionary accruals by incorporating total assets, property, plant and
equipment, and changes in revenues as variables. The 1991 Jones model assumes, however,
that revenues are nondiscretionary. In reality, we might assume that earnings can be managed
through discretionary revenues (e.g. an increase in receivables). Treating this discretionary component of accruals as nondiscretionary will bias the estimate of earnings management downwards. Dechow et al. (1995) propose a modification of the 1991 Jones model to adjust for the tendency to treat discretionary accruals as nondiscretionary when discretion is exercised over revenues. The adjustment entails incorporating the change in receivables compared to the previous year in the model by assuming that all changes in credit sales in the event period result from earnings management. The reasoning behind this modification is that it is easier to manage earnings by exercising discretion over the recognition of credit sales than over the recognition of cash sales.
The models proposed by Jones (1991) and Dechow et al. (1995) use regression to estimate the amount of accruals that is discretionary by partitioning the sample into firm years in which earnings management is hypothesized (the event period) and firm years in which no earnings management is hypothesized (the estimation period). The estimation period is then used to obtain firm-specific parameters which are used to predict the amount of nondiscretionary accruals in the estimation period. To measure earnings management in the event period, the predicted level of nondiscretionary accruals is then subtracted from the total amount of accruals (Dechow et al., 1995). To estimate the firm-specific parameters, the estimation period should be separated from the event period and arguably hypothesized to be characterized by no earnings management. Since I expect increased earnings management when audit firm tenure is both short and long, and because the exact length of the period that would define tenure as either long or short is unclear, the identification of the estimation period and its estimated parameters would at least be very complex, if not invalid. Matching firms based on industry (Kothari et al., 2005) or using firms in the same industry to estimate the firm-specific parameters (Dechow & Sloan, 1991) is impossible given the relatively small sample size.
3.2.1 Total accruals model
Dechow & Schrand (2004) argue that focusing on total accruals as a simple and easy way to identify accruals is appropriate in many circumstances. This argument is supported by Dechow et al. (2003), who show that the correlation between estimated discretionary accruals from accrual models like the modified Jones model (Dechow et al., 1995) and total accruals is more than 80%. Therefore, total accruals (TA) are used as a measure for earnings manipulation. Consistent with, among others, Healy (1985), Jones (1991) and Dechow et al.
(1995), TA as a proportion of total assets are calculated by dividing the change in noncash working capital less depreciation and amortization by lagged total assets.
𝑻𝑨
𝒕𝑨
𝒕−𝟏= (∆𝑪𝑨
𝒕− ∆𝑪𝑳
𝒕− ∆𝑪𝒂𝒔𝒉
𝒕+ ∆𝑺𝑻𝑫
𝒕− 𝑫𝒆𝒑
𝒕)/(𝑨
𝒕−𝟏) ( 1 )
Where:
TAt = total accruals in year t
∆CAt = change in current assets in year t compared to year t-1
∆CLt = change in current liabilities in year t compared to year t-1
∆Casht = change in cash and cash equivalents in year t compared to year t-1
∆STDt = change in debt included in current liabilities in year t compared to year t-1
Dept = depreciation and amortization expense in year t At-1 = total assets in year t-1
The change (∆) is computed between time t and t-1.
3.2.2 Defond & Park model
Since there is a variety of measures available that can be used as a proxy for earnings management, I include the model designed by Defond & Park (2001) to do a sensitivity analysis. The Defond & Park (2001) model uses the difference between realized working capital and the expected level of working capital, based on the historical relation between working capital and sales levels, as a proxy for abnormal accruals. The reasoning behind this model is that a certain amount of working capital is needed to support a certain level of revenues (Hooghiemstra et al., 2008). While Defond & Park (2001) calculate abnormal working capital accruals quarterly, I follow the approach consistent with Hooghiemstra et al.
(2008) by computing abnormal working capital accruals yearly and scaling them by revenue in the same year.
𝑨𝑾𝑪𝑨
𝒕/𝑺
𝒕= (𝑾𝑪
𝒕− (𝑾𝑪
𝒕−𝟏⁄ 𝑺
𝒕−𝟏) × 𝑺
𝒕)/𝑺
𝒕( 2 )
Where:
AWCAt = Abnormal working capital accruals in year t
WCt = Net working capital in year t ((current assets – cash and cash equivalents) – (current liabilities – short term debt))
WCt-1 = Net working capital in year t-1 St = Revenue in year t
St-1 = Revenue in year t-1
3.3 Explanatory variable
The explanatory variable, audit firm tenure, is measured by the number of consecutive years that the audit firm has audited a client. Auditor data for both the researched sample years (2004-2013) and the years before 2004 is found on company.info. For most companies, company.info provides annual reports and information on the audit firm that has audited these firms back to the early 1990s. Therefore, the auditor-client relationship can have lasted 10 years before the sample period, allowing long audit firm tenures to be researched. This does cause a limitation to the audit firm tenure data, as the length of tenure may actually have been longer when the audit firm-client relationship extends back in time beyond the years recorded in the company.info database. To determine the length of tenure, audit firm switches had to be identified. Auditor data was handpicked from annual reports, as Orbis only provides this information for the most recent year of their database. During the sample period, an audit firm switch was identified for 39 firms, while no audit firm switch was identified for 125 firms.
In some cases, the audit partner or firm joined or merged with another audit firm. In that
case, when the signing audit partner remains the same between the years of the audit firm
switch, I consider the audit firm-client relationship to be extended and sum both numbers of
years to arrive at audit firm tenure in years. This approach seems appropriate, as I expect both
the auditor learning and impaired independence arguments to apply, especially when the
previous auditor has merged with another firm and the same audit team audits a specific client.
3.4 Control variables
Defond & Jiambalvo (1993) argue that firms that risk violating debt covenants are more inclined to manage earnings so as to avoid violating these covenants, or obtain a better bargaining position in the event of renegotiation. Therefore, I include leverage (LEVt) as a control variable, as a larger amount of debt compared to the total assets of the firm would give managers incentives to manage earnings to avoid violating debt covenants. Following a similar argumentation, firms with higher cash flow from operations relative to assets (CFOt) are less likely to manage earnings, as they are less likely to be in financial distress or at risk of violating debt covenants (Gul et al., 2007)
Following DeAngelo (1981b), audit firm size is expected to positively affect audit quality, because larger audit firms have less incentives to behave opportunistically, as a discovery of such opportunistic behavior would negatively affect a greater amount of quasi-rent streams.
Additionally, a larger audit firm with more clients is less dependent on revenues generated from a specific client, and thus has less incentives to accept earnings management by that firm. I use the dummy variable Big N/Non-Big N (BIG4t) to proxy for audit firm size, which is tested and found to be positively influencing audit quality (Becker et al., 1998; Francis &
Krishnan, 1999; Carcello & Nagy, 2004b).
As the client firm is larger, its financial reporting system is expected to be more sophisticated, because smaller firms often have less resources available and more often experience rapid growth, causing the accounting systems to be unable to keep up with the growing firm (Johnson et al. 2002; Gul et al. 2007). Additionally, the likelihood of litigation is higher for larger clients, providing an incentive for auditors to be more independent (Lys &
Watts, 1994). Therefore, client firm size (CFSt), measured by the log of total assets, is expected to negatively influence earnings management.
Previous research (Menon & Williams, 2004; Burgstahler et al., 2006; Chen et al., 2008)
indicates that sales growth (GROWTHt), defined as the annual percentage of change in
revenue, might influence the amount of accruals. The expected relationship of sales growth
with the amount of discretionary accruals is positive, as investments in working capital could
be a result of expanding operations.
4. Results
4.1 Univariate results
The univariate results present descriptive statistics of the variables used in this paper and correlations between these variables. These results include both the main dependent variable, absolute total accruals, and the dependent variable used for the sensitivity analysis, absolute abnormal working capital accruals.
Table 1 presents mean and standard deviation (SD) values for both the dependent and independent variables, and the control variables that are part of the research model. Outliers and their potential effect on the results are accounted for by winsorizing the data at 2.5% at both tails. The values are split between the short tenure group (audit firm tenure of up to three years), the group with tenure of medium length (four to seven years of audit firm tenure), and the group with long tenure (audit firm tenure of eight or more years). The amounts of firm years in each group (N) are 112, 159, and 524 respectively. Additionally, the final two columns represent the means and standard deviations for the sample as a whole.
Table 1
Descriptive statistics
Short tenure Medium tenure Long tenure Total
Mean SD Mean SD Mean SD Mean SD
ABS (TA) 0.092 0.071 0.079 0.062 0.075 0.059 0.078 0.062 ABS (AWCA) 0.282 0.814 0.082 0.180 0.092 0.383 0.117 0.448
AFT 2.040 0.827 5.510 1.141 13.870 3.910 10.530 5.746
CFS 5.408 0.709 5.452 0.756 5.820 0.859 5.688 0.839
LEV 0.537 0.203 0.563 0.168 0.589 0.150 0.576 0.163
CFO 0.071 0.114 0.087 0.105 0.084 0.083 0.083 0.092
BIG4 0.740 0.440 0.600 0.492 0.840 0.364 0.780 0.415
GROWTH 0.106 0.322 0.069 0.264 0.051 0.213 0.062 0.242
N 112 159 524 795
Note: the descriptive statistics are calculated based on a sample of 96 firms and over the period 2004- 2013. Short tenure = 1-3 years of tenure, medium tenure = 4-7 years of tenure, long tenure = 8 or more years of tenure
Variable definitions:
ABS (TA): client firm absolute value of total accruals
ABS (AWCA): client firm absolute value of abnormal working capital accruals AFT: audit firm tenure
CFS: client firm size LEV: client firm leverage
CFO: client firm cash flows from operations
BIG4: takes value 1 if a company is audited by a Big 4 audit firm in year, if not, it takes value 0 GROWTH: client firm growth
The descriptive statistics in table 1 suggest that the amount of absolute total accruals and
absolute working capital accruals are higher for the short tenure group compared to the
medium tenure group. Absolute total accruals are again higher for the medium tenure group
compared to the long tenure group, though slightly. Absolute abnormal working capital
accruals are slightly higher for the long tenure group compared to the medium tenure group.
With regard to accruals, the differences between the short tenure group and medium tenure group appear quite sizeable, while the difference between the medium and long tenure groups seems unclear. These results provide a first indication that the relationship between audit firm tenure and absolute accruals does not appear nonlinear.
Audit firm tenure for the total sample was 10.5 years on average, with a minimum value of 1 year and a maximum value of 23 years, although some tenures, especially those that date back to the early 1990s according to company.info, may have been longer in practice.
With regard to the control variables, client firms in the longer audit firm tenure groups are somewhat larger, more levered, and experience lower growth on average. The long audit firm tenure group is most often audited by a Big 4 firm (84% of the cases), followed by the short tenure group (74% of the cases), and the medium tenure group (60% of the cases). Cash flow from operations is lowest for the short tenure group and higher but equal among the medium and long tenure groups. The correlation matrix (table 2) shows that there are no two variables correlated so strongly that multicollinearity poses a threat to the validity of the results.
Table 2 Correlation matrix
1. 2. 3. 4. 5. 6. 7. 8.
1. ABS (TA) 1
2. ABS (AWCA) 0.236** 1
3. AFT -0.109** -0.106** 1
4. CFS -0.169** -0.129** 0.395** 1
5. LEV 0.015 -0.049 0.128** 0.279** 1
6. CFO 0.069 -0.334** 0.042 0.182** -0.088* 1
7. GROWTH 0.016 0.155** -0.101** -0.034 -0.023 0.099** 1
8. BIG4 -0.027 0.004 0.282** 0.425** 0.098** 0.111** -0.046 1
*, **: significant at the 5% and 1% levels respectively. Pearson 2-tailed test.
Variable definitions:
ABS (TA): client firm absolute value of total accruals AFT: audit firm tenure
CFS: client firm size LEV: client firm leverage
CFO: client firm cash flows from operations
BIG4: takes value 1 if a company is audited by a Big 4 audit firm in year, if not, it takes value 0 GROWTH: client firm growth
4.2 Multivariate results
The main model predicts the independent variable, the amount of absolute total accruals. The
Kolgomorov-Smirnov and Shapiro-Wilk tests of normality were significant (p < 0.01) for the
dependent variable, implying that the variable is not normally distributed, rendering some
parametric tests inappropriate. Therefore, section 4.2.1 analyzes if there are differences in the
amount of absolute total accruals between groups consisting of short tenure firm years (1-3
years), medium tenure firm years (4-7 years), and long tenure firm years (8 or more years)
using a nonparametric test. Section 4.2.2 analyzes the potential linear relationship between the independent variable audit firm tenure and absolute total accruals.
4.2.1 Nonparametric test
A nonparametric test, the Kruskal-Wallis test, is used to test for differences in absolute total accruals between the different groups of tenure (Table 3). The null-hypothesis that the distribution of absolute total accruals is the same across the different group of tenure is retained (p < 0.093), suggesting that no significant relationship between audit firm tenure group and absolute total accruals exist. Therefore, hypothesis 2 is rejected at this stage, while section 4.2.2 tests for hypothesis 1 using linear regression.
H2: Long audit firm tenure will cause lower audit quality compared to short and medium audit firm tenure (reject)
Table 3
Kruskal-Wallis test for ABS (TA)
Short tenure Medium tenure Long tenure
Mean Rank 439.87 401.31 388.05
N 112 159 524
Short -/- Medium Medium -/- Long Short -/- Long
Mean rank difference 38.56 13.26 51.82
Note: short tenure = 1-3 years of tenure, medium tenure = 4-7 years of tenure, long tenure = 8 or more years of tenure. Overall significance is 0.093.
ABS (TA): absolute value of total accruals
4.2.2 Linear regression
Tabel 4 depicts the results of the linear regression model:
𝐴𝐵𝑆(𝑇𝐴) = 𝛽
0+ 𝛽
1𝐶𝐹𝑆 + 𝛽
2𝐿𝐸𝑉 + 𝛽
3𝐶𝐹𝑂 + 𝛽
4𝐺𝑅𝑂𝑊𝑇𝐻 + 𝛽
5𝐵𝐼𝐺4 + 𝛽
6𝐴𝐹𝑇 + 𝜀
Table 4
Model summary (dependent variable: ABS(TA))
Model 1 Model 2
Intercept 0.143** (0.015) 0.143** (0.015)
Control variables
CFS -0.016** (0.003) -0.016** (0.003)
LEV 0.033* (0.014) 0.033* (0.014)
CFO 0.075** (0.024) 0.075** (0.024)
GROWTH -0.001 (0.009) -0.000 (0.009)
BIG4 0.008 (0.006) 0.009 (0.006)
Main effects
AFT -0.001 (0.000)
R-squared 0.048 0.050
R-squared change 0.003
N 795 795
F-value 7.887** 6.939**
*, **: significant at the 5% and 1% levels respectively
Variable definitions
ABS (TA): client firm absolute value of total accruals AFT: audit firm tenure
CFS: client firm size LEV: client firm leverage
CFO: client firm cash flows from operations
BIG4: takes value 1 if a company is audited by a Big 4 audit firm in year, if not, it takes value 0 GROWTH: client firm growth
The results from model 2 in table 4 show that audit firm tenure is not significantly associated with the amount of absolute total accruals (p = 0.14). With regard to the control variables, growth (p = 0.93) and Big four audit firm (p = 0.12) are not significantly related to absolute total accruals. Client firm size is negatively related to the amount of absolute total accruals (p
< 0.01), in line with Johnson et al. (2002) and Geiger & Raghunandan (2002). In line with Carey & Simnett (2006), leverage is positively related to the amount of absolute total accruals (p < 0.05). The amount of cash flows from operations has a significant positive relation with the amount of absolute total accruals (p < 0.01), which is a surprising result because the need for earnings management is argued to be lower when cash flows from operations are high and the chance of financial distress subsequently low (Johnson et al., 2002; Gul et al., 2007).
Model 1, consisting of the control variables only, has an R-squared of 0.048, which means that only 4.8% of the variance in absolute total accruals is explained by the variables in that model. Model 2, including the independent variable audit firm tenure, adds only 0.3% of explanatory power, bringing the R-squared value to 0.050. The F-values for model 1 and model 2 are 7.887 (p < 0.01) and 6.939 (p < 0.01) respectively, implying that the regression model is a good fit of the data. The inability to find a significant relationship between audit firm tenure and the amount of absolute total accruals leads to the rejection of hypothesis 1.
H1: A greater length of audit firm tenure will cause higher audit quality (reject) 4.3 Sensitivity analysis
To test whether the results are sensitive to the method chosen to proxy for audit quality, a sensitivity analysis is employed using absolute abnormal working capital accruals as a proxy for audit quality. Section 4.3.1 contains the non-parametric Kruskal-Wallis test to test for differences between the groups of different audit firm tenure. Section 4.3.2 provides results on the linear regression model.
4.3.1 Sensitivity analysis: nonparametric test
The Kruskal-Wallis test is ran to test for differences in absolute abnormal working capital
accruals in a nonparametric setting (Table 5). The null-hypothesis that the distribution of
absolute abnormal working capital accruals is the same across different groups of tenure is
rejected in this instance (p < 0.01). An analysis of the pairwise comparisons shows that
absolute abnormal working capital accruals are lower for the long tenure group compared to
both the medium and short tenure groups (p < 0.01). Therefore, hypothesis 2 is not only
rejected, but evidence is found that the opposite of hypothesis 2 is true: long audit firm tenure
causes higher audit quality compared to short and medium firm tenure
H2: Long audit firm tenure will cause lower audit quality compared to short and medium audit firm tenure (reject)
Table 5.
Kruskal-Wallis test for ABS (AWCA)
Short tenure Medium tenure Long tenure
Mean Rank 447.54 436.94 375.60
N 112 159 524
Short -/- Medium Medium -/- Long Short -/- Long
Mean rank difference 10.60 61.34** 71.94**
**: significant at the 1% level
Note: short tenure = 1-3 years of tenure, medium tenure = 4-7 years of tenure, long tenure = 8 or more years of tenure. Overall significance is 0.001.
ABS (AWCA): client firm absolute value of abnormal working capital accruals
4.3.2 Sensitivity analysis: linear regression
Table 6 depicts the results of the linear regression model:
𝐴𝐵𝑆(𝐴𝑊𝐶𝐴) = 𝛽
0+ 𝛽
1𝐶𝐹𝑆 + 𝛽
2𝐿𝐸𝑉 + 𝛽
3𝐶𝐹𝑂 + 𝛽
4𝐺𝑅𝑂𝑊𝑇𝐻 + 𝛽
5𝐵𝐼𝐺4 + 𝛽
6𝐴𝐹𝑇 + 𝜀
Table 6.
Model summary (dependent variable: ABS(AWCA))
Model 1 Model 2
Intercept 0.499** (0.015) 0.476** (0.015)
Control variables
CFS -0.042* (0.020) -0.030 (0.021)
LEV -0.172 (0.094) -0.168 (0.094)
CFO -1.718** (0.164) -1.726** (0.163)
GROWTH 0.351** (0.061) 0.341** (0.061)
BIG4 0.100* (0.039) 0.110** (0.039)
Main effects
AFT -0.006* (0.003)
R-square 0.161 0.166
R-square change 0.004*
N 795 795
F-value 30.385** 26.070**
*, **: significant at the 5% and 1% levels respectively Variable definitions
ABS (AWCA): client firm absolute value of abnormal working capital accruals AFT: audit firm tenure
CFS: client firm size LEV: client firm leverage
CFO: client firm cash flows from operations
BIG4: takes value 1 if a company is audited by a Big 4 audit firm in year, if not, it takes value 0 GROWTH: client firm growth
