The effect of audit tenure on audit quality

Amsterdam Business School

The effect of audit tenure on audit quality

Final Version Name: Telleman, P.J. Student number: 10291660 Supervisor S.W.Bissessur Date: 18 / 06 / 2015 Amount of words: 15.627

MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam

Abstract

This study examines the effects of auditor tenure on the audit quality. Prior research showcases the ongoing discussing between opponents and proponent of longer auditor tenure, as longer tenure might influence the independence of the auditor towards its client (Chen & Lin, 2010; Davis, Soo, & Trompeter, 2000; Sinason, Jones, & Shelton, 2001) and a shorter tenure might result in unnecessary cost and inefficiency (Geiger & Raghunandan, 2002; U.S. Governement Accountabilty Office (GAO), 2003). The hypothesis for this study states that a decrease in auditor’s tenure has no effect on the discretionary accruals, used as proxy for audit quality. This study uses the discretional part of the accruals, as provided by Jones (1991), to determine the association between auditor’s tenure and the discretional accruals. This study provides insignificant results showing that auditor’s tenure has no association with discretional accruals reported by listed North American companies from 2003 till 2013. This study supplement prior research and gives opportunities for further studies to use different measurement methods or a more complete dataset.

Statement of Originality

This document is written by student Paulus Johannes Telleman, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Table of contents

1 Introduction ... 4

2 Literature review ... 5

2.1 Audit Quality ... 5

2.1.1 What determines audit quality? ... 5

2.1.2 How to measure audit quality? ... 7

2.2 Audit Tenure ... 9

2.2.1 Negative effects of a longer audit tenure ... 9

2.2.2 Positive effect of a longer audit tenure... 10

2.3 Audit Tenure in relation to Audit Quality ... 11

3 Methodology ... 13

3.1 Hypothesis ... 13

3.2 Prior literature on discretionary accrual models ... 13

3.3 Empirical model ... 16 3.4 Data ... 18 4 Results ... 21 4.1 Sample characteristics ... 21 4.2 Descriptive statistics ... 23 4.3 Empirical results ... 25 4.3.1 Jones-model (1991) ... 25

4.3.2 Association between auditors tenure and the discretional accrual ... 29

5 Conclusion ... 31

1 Introduction

The recent financial crisis from 2007 raised serious doubts about auditor independence - the cornerstone of the audit profession and synonym for audit quality (Karim & Sunder, 2011; Herz, Iannaconi, & Ryan, 2001). It was not that the auditors where involved in a lot of the primary causes of the crisis. Auditing is not meant to stop companies from making the wrong business moves. According to IASB (2010), it’s the auditor’s task to safeguard, that the information provided within the financial reporting gives an accurate and fair view of the company’s financials.

In response to the financial crisis, the European commissioner Barnier (European Commision, 2010) presented his proposals on statutory audits, in which the most important proposal is considered to be a mandatory auditor’s rotation for audit firms. Barnier believes that the rotation will prevent conflicts of interest and will safeguard the independence of the auditor. This proposal suggests that Barnier believes that longer auditor tenure will result in lower audit performance by the auditor. It is expected, that an auditor will be less tempted to demand necessary adjustments to the companies financials, when the auditor has been with the client for a longer period of time (Sinason, Jones, & Shelton, 2001). The auditor is assumed to have built a good partnership with the client’s management and will try to secure future audit fees. This study will verify whether this assumption is accurate.

This study examines the effects of auditor tenure on the audit quality of listed North American companies from 2003 till 2013. As proxy for audit quality, this study uses the discretional part of the accruals as provided by Jones (1991). The hypothesis for this study states that a decrease in auditor tenure has no effect on the discretionary accruals. The regression analyses show that auditor’s tenure has no association with reported discretional accruals. The results however were insignificant, therefore the hypothesis for this study is confirmed. This study supplements prior research and gives opportunities for further studies to use different measurement methods or a more complete dataset.

This paper first discusses what other authors have written on auditor tenure and audit quality. Does increased tenure at the client influence the performance of the auditor and does the relationship between the auditor and the client change over the course of time? What is audit quality and how is audit quality measured? What variables increase the quality of an audit? These are some of questions the literature review is trying to answer. In the third paragraph, the methodology of this paper will be further discussed. It contains a review on models used in prior literature, to determine the model that is most appropriate for this study. The two different regressions that will be used for this study are explained, as well as the data on which the model will be used. An overview of the composition of the data and the regression outcomes is presented in the fourth paragraph. The fifth and last paragraph contains the conclusion, the contribution of this study to prior literature and suggestions for further research.

2 Literature review

The separation of ownership and control in public companies creates a conflict between management and the stakeholders. Due to conflicts of interest and asymmetric information, financial reporting’s are prepared by management (Chi, Huang, Liao, & Xie, 2009) to overcome the information gap between the two. According to IASB (2010) it’s the auditor’s task to safeguard the accurateness and fair view of the information provided within the companies financials. The financial reporting helps the user of the report to predict future benefits (Finger, 1994). The extent in which the auditors succeed in safeguarding the accuracy and fairness of the financial reporting depends on the quality of the audit. Audit quality is defined as the probability that the auditor will both discover and report a breach in the client´s accounting system (DeAngelo', 1981; Deis & Giroux, 1992). Audits are of high quality when the engagement team personnel makes good decisions regarding the specific tests that are used and the evidence extracted from these tests (Francis J. R., 2011). As such, audit quality contributes to the credibility of financial disclosure (Mansi, Maxwell, & Miller, 2003) and determines the reliability of the financial reporting (Behn, Choi, & Kang, 2008).

2.1.1 What determines audit quality?

There are multiple variables that influences audit quality (Francis J. R., 2011). Some of these are easily observable and measurable, like audit-team factors such as previous experience, team size and team composition. Others are more difficult to measure, such as the professional skepticism and conservatism of the auditor (Knapp, 1991) or the auditor characteristics and educational background (Gul, Wu, & Yang, 2013). Each individual auditor has an economically and statistically influence on the audit quality, even when audit firms try to maintain consistency by quality-control mechanism, like standardized audit approach (Jeppesn, 2007).

There haven´t been many studies investigating which variables has the biggest influence on audit quality. When searching for previous literature on the variables of audit quality, it isn´t surprising that most studies revolve around measurable variables. The authors of these papers perform quantitative empirical research, using various observable outcomes to proxy for audit quality, like the audit opinion given and the change of auditors (Beattie, Fearnley, & Hines, 2012). The most investigated variables on audit quality seem to be focused on three variables: audit firm size, industry specialization and auditor tenure. The effects of audit firm size and industry specialization will be discussed below. The effects of auditor tenure in relation to the audit quality will be discussed in paragraph 2.3. There are other variables that can influence the audit quality, like the shareholders involvement, economic risk, audit committee activities, risk of regulatory action, audit firm ethics, auditor partner rotation, audit inspections, and regulation (Beattie, Fearnley, & Hines, 2012; Chi, Lisic, Long, & Wang, 2013; Dao, Raghunandan, & Rama, 2012; Gavious, 2007; Gong, Li, & Xie, 2009). Since the amount of research on these variables is limited this thesis will only discuss the three variables mentioned earlier.

Prior research shows multiple reasons why audit firm size can be seen as a proxy for audit quality (Griffin, Lont, & Sun, 2010; Francis J. R., 2004). DeAngelo (1981) argues that a larger audit firm has a greater reputation to lose if firms misreport. These incentives drive large audit firms to provide high quality audits (Ruiz-Barbadillo, Gómez-Aguilar, & Carrera, 2009). The auditor firm size also influences the extent in which the auditor is economical dependent on its clients (Dart, 2011). Larger audit firms have a bigger client portfolio, making the risk of losing one individual client less important than it would be for a smaller audit firm. Larger audit firms therefore have lower economic risk and economic dependence factors, which enhances the audit quality (Lim & Tan, 2010). Lastly, larger audit firms got more in-house experience dealing with public companies than smaller audit firms. This gives auditors a larger peer group to consult if problems arises (Danos, Eichenseher, & Holt, 1989).

When assessing the audit firm size, the most commonly used distinction is whether the audit firm belongs to one of the Big Four audit firms or not (DeAngelo', 1981; Francis & Yu, 2009). When comparing the Big Four audit firms to smaller audit firms, prior research show that Big Four auditors are (a) capable of delivering higher audit quality (Francis & Yu, 2009; Choi, Kim, Kim, & Zang, 2010) (b) are more likely to issue an going-concern opinion (Francis & Krishnan, 1999) and (c) are more capable to resist clients of using aggressive earnings management strategies than smaller audit firms (Geiger & Rama, 2006; Becker, Defond, Jiambalvo, & Subramanyam, 2010). Big Four audit firms are associated with lower absolute values of cost of capital (Khurana & Raman, 2004), cost of debt (Pittman & Fortin, 2004) and discretionary accruals, which is an indicator for poor audit quality as discusses in the next paragraph (Becker., DeFond, Jiambalvo, & Subramanyam, 1998; Francis & Krishnan, 1999). Even in industries that naturally have higher amounts of discretionary accruals, such as firms with high investments opportunities, bigger audit firms can curb the discretionary accruals better than smaller firm can (Lai, 2009). It is also found that Big Four firms, when controlled for clientele size, are sued less frequently than non-Big four firms (Feroz, Park, & Pastena, 1991), because they have a reputation to protect and face greater litigations costs than smaller audit firm (DeAngelo', 1981; Weber, Willenborg, & Zhang, 2008). Banks share this perception by offering a lower interest rate on private loans, when a firm is audited by a Big Four audit firm compared to a non-Big four audit firm (Chu, Mathieu, & Mbagwu, 2013). Lastly, the fact that Big Four audit firms sell their services for a higher fee than smaller audit firms, can be considered an proxy for high quality services, because auditors of the Big Four cannot sell low-quality audits at a higher quality price (Chang, Gygax, Oon, & Zhang, 2008). Big Four audit firms doesn’t always outperform smaller practices in terms of audit quality, but past research suggests that on average this may be the case (Boone, Khurana, & Raman, 2010; Francis J. R., 2004).

The auditor industry specialization is also an indicator of audit quality (Reichelt & Wang, 2010). When it comes to the auditors branch, industry specialization relates to the auditor having a relative significant portion of the industry market share within their client portfolio (Solomon, Shields, & Whittington, 1999) or the level of non-audit services acquired within the industry market (Lim & Tan, 2008). This gives the non-auditor more opportunities

errors and identify extraordinary transactions and risks when auditing the client (Bedard & Biggs, 1991; Krishnan, 2003). Prior research shows that an audit industry-specialist provides higher quality financial statements (Balsam, Krishnan, & Yang, 2003; 2010) and has greater incentive to protect their high quality industry-reputation (Lim & Tan, 2008). Industry specialist have stricter quality standards on their staff allowing them perform more effective procedures to measure a client’s risk of business failure (Owhoso, Messier, & Lynch Jr., 2002). Prior research also shows that industry specialist charge a higher audit fee, which allows auditors to perform higher quality audits, as discussed earlier (DeFond, Francis, & Wong, 2000). The use of industry specialist is negatively associated with restatements (Romanus, Maher, & Fleming, 2008), accrual-based earnings management (Chi, Lisic, & Pevzner, 2011) and financial fraud (Carcello & Albert, 2004) and positively associated with disclosure quality (Dunn & Mayhew, 2004). This is consistent with prior research on audit firm size, namely that audit firms with a larger industry specific portfolio have more in-house experience and resources to provide higher audit quality. Further research shows that auditor industry specialization leads to higher forecast accuracy and less forecast dispersion (Behn, Choi, & Kang, 2008), resulting from an increase in quality of financial information used to predict future earnings. Firms that choose industry-specialist auditors and meet or beat analysts’ forecasts are less likely to use discretionary current accruals to meet or beat analysts’ forecasts ex post (Burnett, Cripe, Martin, & McAllister, 2012; Reichelt & Wang, 2009).

2.1.2 How to measure audit quality?

Auditing reduces information asymmetries that exist between managers and firm stakeholders by verifying the validity of financial statements. Whether auditing has been performed accordingly, depends on the quality of the auditor. High quality auditors are capable to detect misstatements in the financial statements and object to these misstatements towards management. (Becker, Defond, Jiambalvo, & Subramanyam, 2010). A misstatement in financial reporting isn’t uncommon, because managers have incentives to adjust earnings to maximize firm and managers wealth. Meeting or beating earnings benchmarks is not random, suggesting that managers bias financial statements to meet earnings targets (Brugstahler & Dichev, 1997), also known as earnings management. Earnings management is intended to obtain private gains, by either using accruals or real activities, thereby lowering the reporting quality (Chi, Lisic, & Pevzner, 2011). Management is limited in managing earnings when there is a greater transparency, provided by the work performed by the external auditor (Hunton, Libby, & Mazza, 2006; Prawitt, Smith, & Wood, 2009).

Audit quality is difficult to measure, because the only thing that is clearly observable is the audit report itself. Other variables, like the audit strategy used or the professional skepticism of the auditor, aren’t visible or reported to the public. When investigating prior studies, there are three measurements given to identify the quality of the audit: the accurateness of the audit opinion given, accurateness of issuing a going-concern opinion and the amount of accruals reported in the financial statements. These measurements are discussed below. It’s the auditor’s task, to write an auditor opinion report regarding the client’s reported financial statements. In short, the auditor can either issue a clean auditor report (with no issues found) or report about any bias found during the audit. The accurateness of the audit opinion can only be expressed as either an ‘audit failure’ or a ‘no

audit failure’. An audit failure occurs when the auditor is a) not independent or b) if an auditor incorrectly issues a clean report (Francis J. R., 2011). Audit failures, regarding the financial statements, are identified when there is a successful civil litigation against the auditors. Although this is easily observable, prior research has shown that the number of legal actions against auditors is quiet small (Feroz, Park, & Pastena, 1991; Dechow P. , Ge, Larson, & Sloan, 2011). Using audit failures regarding audit opinions as measurement for audit quality provides insufficient data to be used in data analyses.

Another way to measure audit quality is by analyzing client business failures in relationship with issuing a going-concern opinion. Not only does the auditor need to notify possible bias within the financial statement, but they also need to alert stakeholders if they believe the contingency of the firm is at risk, by issuing a going-concern opinion. Just like the audit opinion, the accurateness of the going-concern opinion can be expressed as either an ‘audit failure’ or a ‘no audit failure’. When it comes to audit failure regarding going-concern opinions, prior research finds that auditors are conservative and issue about seven times more going-concern opinions as there are client failures. When there actually is a client failure, the auditor usually gets it wrong and fails to issue a going-concern report in the year prior (Lennox, 1999). However, the actual amounts of this type of error is relatively very low as is represents less than 1 percent of the total audit engagements. Therefor using audit failures regarding going-concern opinions as measurement for audit quality provides insufficient data to be used in data analyses.

Most prior studies use the amount of accruals in the financial statement to measure audit quality, because the financial statement should be viewed as a joint outcome from the management and the auditor. The earnings quality as captured by abnormal accruals is affected by the audit firm and thus reflects audit quality (Chi, Huang, Liao, & Xie, 2009). Large accruals can signify low quality and persistence of earnings, resulting from managers trying to adjust earnings. There are multiple factors that can constrain managements limit to manage earnings, like the firm’s internal governance (Dechow, Sloan, & Sweeney, 1995), previous accounting decisions made by the firm (Sweeney, 1994) and the cost of getting imposed, should the earnings management attempt be revealed. The one factor that’s interesting for this study is the external auditor, who’s responsible for evaluation the quality of the accruals and to request a necessary adjustment if bias has been found. Firms that go through high quality audits are less likely to use accrual based- earnings management to meet or beat analysts’ forecasts (Burnett, Cripe, Martin, & McAllister, 2012; Balsam, Krishnan, & Yang, 2003). Therefor if the company financials contains low quality accruals, than the auditors has failed to identify possible bias resulting from low audit quality.

When looking at accruals, it is important to distinguish discretionary accruals and non- discretionary accruals. Discretionary accruals are, in contrast with non-discretionary accruals, a non-obligatory expense (such as an anticipated bonus for management) that is yet to be realized but is recorded in the account books. Discretionary accruals can be viewed as evidence of auditor quality based on three assumptions. The first assumption is that an independent auditor requires the client to file objective or unbiased financial statements. Second, discretionary

allowed by the auditor. Third, just meeting or beating earnings benchmark is not random, suggesting that managers bias financial statements to meet earnings targets (Brugstahler & Dichev, 1997). An objective auditor will not allow clients to bias earnings to achieve these benchmarks, even if his economic dependence on the client increases.

2.2 Audit Tenure

The auditor’s tenure has always been called in question, especially after European commissioner Barnier’s (European Commision, 2010) proposal for mandatory audit rotation. The length of the auditor tenure influences the way the audit is performed. In the first couple of years, the auditor has to use more resources to get to know the auditor’s organisation. But what happens when an audit firms stays for a longer period of time? Will the audit keep improving itself in quality or will indolence kicks in? The opinions on the subject are divided, because there are reasonable arguments to both support or oppose a longer auditor’s tenure. These arguments are discussed below.

2.2.1 Negative effects of a longer audit tenure

The opponents of longer auditor tenure argue that an audit rotation leads to a fresh look at the company’s financials, to the benefit of investors and regulators (Seidman, 2001). Over time auditors are likely to follow the audit procedures from the previous audit, which can result in auditors failing to incorporate new evidence or changes in the client´s situation (Brody & Moscove, 1998). The longer the auditor remains at the client, the less sharp the auditor will be when performing the audit, leading to undetected errors in the financial reporting (Copley & Doucet, 1993).

When auditing the same client for multiple years, the auditor might become convinced of the client’s financials and personal integrity, that evidence of errors in the financial statements is unconsciously overlooked or downgraded (Bates, Ingram, & Philip, 1982; Kim & Yi, 2009). It is even possible that auditors start to identify himself as someone working for the client, which will cloud the auditor’s professional skepticism (Brody & Moscove, 1998). It is likely that this loss of independence will result in auditors compromising on their client’s accounting and reporting choices, by being less tempted to demand necessary adjustments to the companies financials (Chen & Lin, 2010; Davis, Soo, & Trompeter, 2000; Sinason, Jones, & Shelton, 2001). Financial statements are perceived to be more uncertain, which lead to an increase in socials cost. Investments becomes riskier, for which investors will demand higher risk premiums, thus the cost of capital will increases as well (Botosan & Plumlee, 2002).

Another argument is that auditors are likely to settle disputes in the auditor´s favor in fear of losing the client. An auditor will be less tempted to demand necessary adjustments to the company’s financials, when the auditor tries to secure future audit fees (Sinason, Jones, & Shelton, 2001). Auditors practice low-balling, by offering a discounted price for audit services in order to build a relationship with the client, that could become profitable later by increases in fees or offering additional services (DeAngelo', 1981; Davis & Trompeter, 1993). Low-balling increases auditors willingness to comply with the client’s desires (Lee & Gu, 1998). A mandatory rotation

of audit firms, as suggested by European commissioner Barnier’s (European Commision, 2010), limits the future profits of retaining a client (Quick, 2012). The probability that clients will switch auditors increases after a critical auditors report. This decreases the likelihood that the auditor will file such a report (Moore, Tetlock, Tanlu, & Bazerman, 2006).

2.2.2 Positive effect of a longer audit tenure

The proponents of longer auditor tenure argue that longer auditor’s tenure gives the auditor more knowledge on the client´s business, operations and systems. Knowledge is essential for the auditor to detect material misstatements. For the most part, the knowledge necessary is client-specific (Joihnson, Khurana, & Reynolds, 2002), which means that there is a learning curve for new auditors to acquire this knowledge (Knapp, 1991). Longer auditor tenure can enable auditors to place greater constraints on extreme management decisions in the financial reporting (Myers , Myjers, & Omer, 2003). Auditor expertise can remedy the reduced vigilance through overfamiliarity caused by longer auditor tenure (Solomon, Shields, & Whittington, 1999; Dart, 2011).

The cost of changing auditor results in loss of auditor knowledge that is not carried forward to the new auditor, even with the mandatory transfer of documents from the old auditor to the new (U.S. Governement Accountabilty Office (GAO), 2003). Because of this, prior research shows that there are more failures in the earlier years of the audit client relationship, than in the later; earnings management is at its highest in the period after an audit rotation (Geiger & Raghunandan, 2002). Clients will need to devote more resources to help new auditors understand their operations and systems. Knowledge of clients operations is critical to identify sources of audit risk and perform more efficient audits. The possible benefits of shorter audit tenure, like being more objective, maybe don’t weight out the higher first-year audit costs, thus leading to an overall increase in audit costs.

Lastly normal audit and client staff rotation provides the same fresh look, that shorter auditor tenure provides (Bates, Ingram, & Philip, 1982). Furthermore, firm’s internal quality controls procedures should insure the objectivity and compliance when auditing long time clients. Firms failing to maintain independence for an engagement may suffer a direct loss from litigation, by those to whom the auditor is responsible or indirect losses from loss of reputation (Simon & Francis, 1988). The tenure of the audit therefore isn´t relevant for maintaining objectivity throughout the years and for increasing total firm audit fees. Auditors' reputations for performing quality audits is positively associated with their ability to earn higher fees and attract clients (Wilson Jr. & Grimlund, 1990), as higher fees increases the pressure to perform on the auditor (Dao, Raghunandan, & Rama, 2012).

To summarize, the importance of the auditor tenure is dependable on two types of cost. A longer tenure might influence the independence of the auditor towards its client and a shorter tenure might result in unnecessary cost and inefficiency. This study will try to determine the impact of auditor tenure on the audit quality and thus whether the opponents or proponents of longer auditor tenure are right.

2.3 Audit Tenure in relation to Audit Quality

The fraude and bankruptcies during the recession period in 2008 has put the relationship between auditors and clients under the microscope. It’s the society’s perspective, that it’s the auditor’s task to identify fraud and to also alert the market of the possibility of bankruptcy, by issuing a going-concern opinion. The recent crisis showed that in some cases, auditors failed to report these issues. Audit firms experienced higher risk, because they suffered from slower receivables collection and clients where trying to reduce the audit fees, thus giving the auditors less resources to increase audit effort (Ettredge, Fuerherm, & Li, 2014). Regulators believe that it was the length of the auditor tenure that threatened the audit quality during the recession (Johnson, Khurana, & Reynolds, 2002). This is contrary to the investor’s perspective. They believe that longer auditor tenure improves audit quality, since the reported earnings have a bigger influence on the investor’s decision making as the auditor tenure increases (Ghosh & Moon, 2004; Mansi, Maxwell, & Miller, 2003). Investors believe that the auditor’s economic dependence impairs the auditor’s independence more than the auditor tenure (Dart, 2011). The auditors agree with the investor and claim that there is a positive relation between tenure and auditor quality, and that mandatory rotation would most likely impair audit quality (AICPA, 1992). These contradictions may be the reason why the auditor tenure is being heavily debated.

When looking at previous studies, multiple researchers found that a shorter tenure has a negative impact on the audit quality. Geiger & Raghunandan (2002) found that there is more audit reporting failures in the early years of the audit-client relationship than in the later. Johnson, Khurana, & Reynolds (2002) found evidence that a short tenure of two to three years is associated with a lower reporting quality compared to the reporting quality of four to eight years tenure. There was no evidence found that a tenure of eight years or more reduces the reporting quality, compared to a four to eight year tenure. Manry, Mock, & Turner (2008) chose to look at the audit partner tenure (and therefor audit tenure) and actually did found evidence that increased tenure provides better audit quality. The same goes for Lim & Tan (2010), who found that firms with extended auditor tenure have higher audit quality, measured by looking at accrual quality. Also when looking at earnings management as proxy for audit quality, Davis, Soo, & Trompeter (2010) and Gul, Fung, & Jaggi (2009) found a positive relationship between auditor tenure and audit quality. Jackson, Moldrich, & Roebuck (2008) found that the auditor tenure neither increases nor decreases the audit quality.

Assuming that auditor conservatism leads to higher audit quality, Cameran, Prencipe, & Trombetta (2012) found that the auditor conservatism increases in the last three-years of an engagement period, suggesting that a mandatory auditor rotation does not improve audit quality. Jenkins & Velury (2008) found a positive association between the conservatism in reported earnings and the length of the auditor–client relationship. Differently, Kramer et al (2011) show that conservatism in reported earnings decreases as the tenure of the audit firm increases.

Ruiz-Barbadillo, Gómez-Aguilar, & Carrera (2009) studied the mandatory rotation of audit firms in Spain in 1991-1994 and found that there was no evidence that the mandatory rotation is associated with a higher ratio of issuing going-concern opinions. Summer (1998) even found that rotation could impair the auditor independence

rather than enhance it. Nagy (2005) examined the effect of the mandatory auditor change for former Arthur Anderson clients, after the fall of Arthur Anderson. Although the author wasn’t able to form a conclusion on bigger companies, research showed that the audit quality increased for smaller companies forced to switch from Author Anderson.

Davis, Soo, & Trompeter (2000) performed research between the length of the auditor tenure and earnings management. They find a negative relation between tenure and analyst forecast errors. These findings suggest that when the auditor tenure increases, the ability to meet earnings forecast will increase as well. Chi, Lisic, & Pevzner (2011) found that longer auditor tenure is associated with higher levels of real earnings management among firms with incentives to manage earnings. The auditor designation by authorities in Korea, gave Kim & Yi (2009) the possibility to research whether mandatory audit designation influences the extent of earnings management. They found designated auditors are likely to be less lenient towards managerial discretion and that mandatory auditor replacement enhances audit quality. This gives evidence that setting limits on the auditor tenure increases the financial reporting quality.

So what is the optimal auditor tenure? Given what has been discussed above, the optimal tenure should be the tenure in which the costs of independence, resulting from a longer tenure, and the cost of inefficiency, resulting from a shorter tenure, are at a minimum. Regulators in Europe seem to think four years, as that is what they included in their Green Paper as the right tenure (European Commision, 2010), which can be described as relative short audit tenure. Regulators may prefer a short audit tenure, not only because of the seemingly decrease in objectivity, but also because it provides regulators with more flexibility, in terms of future policy adjustments (Eleftheriou, Komarev, Klumpes, & Farouk, 2013).

3 Methodology

My study will investigate the relationship between auditor’s tenure and audit quality, by using an empirical model on collected data from COMPUSTAT. I will determine the audit quality for listed North American companies from 2003 till 2013 and thereby supplementing prior research, by investigating a similar research question in a more recent period time. The auditor has been called into question, as to their role within the economic crisis of 2008 and the rollout of new regulations regarding the auditor tenure. This study will determine the association between the auditor’s tenure and the change in audit quality, through regression analysis. The methodology will be explained in detail below.

3.1 Hypothesis

Prior research shows that there is a relationship between the auditor tenure and the audit quality. The opponents of longer auditor tenure state that a longer tenure influences the independence of the auditor towards its client, as the auditor is more focused on securing future audit fees. The proponents of a longer tenure claim that a shorter tenure might result in unnecessary cost and inefficiency, as new auditors have a learning curve to acquire client specific knowledge. This study aims to contribute towards this discussion by calculating the discretionary part of the accruals, as a proxy for audit quality, and determine its coloration with audit tenure. For my study, I will research the following hypothesis:

H0: Decrease in auditor’s tenure has no effect on the discretionary accruals

3.2 Prior literature on discretionary accrual models

For the hypothesis I consider the role of the auditor tenure on the audit quality provided by the auditor. During the discussion of prior literature on audit quality in paragraph 2.2, I’ve found three different ways to measure audit quality: the accurateness of the audit opinion given, accurateness of issuing a going-concern opinion and the amount of accruals reported in the financial statements. Prior research found that audit failures regarding audit opinions and going-concern opinions as measurement for audit quality, provides insufficient data to be used in data analyses. Therefor I will be using discretionary accruals as my proxy for audit quality.

Accruals are used to mitigate transitory variation in cash flow. Cash flow can cause variation from normal operating activities or from manipulative variation in the working capital (Dechow, 1994). Accruals that are used to manage earnings are categorized as discretionary accruals. Therefor the discretionary part of the accruals is what affects the earnings quality and therefor also the audit quality. Prior studies have used different ways to measure the discretionary accruals, which I´ll discuss in the paragraph.

When examining prior literature, most use a model based on the cross-sectional Jones (1991) accruals estimation model. The Jones-model is used in a wide range of studies (Dechow, Sloan, & Sweeney, 1995; Kothari, Leone, & Wasley, 2005; Myers , Myjers, & Omer, 2003), to measure the non-discretionary accruals as a function of change in revenue and level of property, plant and equipment. The model is based upon the change in working

capital. Working capital depends on changes in revenue and non-working capital accounts, which are reflected in property, plant and equipment. These variables are used to control for changes in accruals that are due to changes in the firm’s economic situation. The portion of accruals unexplained by normal operating activities is assumed to be due to earnings management and thus distinguished as discretionary accruals.

Dechow, Sloan, & Sweeney (1995) evaluated alternative accruals based models for measuring earnings management. For one of these models, they modified the Jones-model by adjusting the change in revenues for the change in receivables. The original Jones-model assumed that adjustments made by management are not exercised on the revenues in the estimation period or event period. Dechow, Sloan, & Sweeney believed this would result in bias, assuming that it is easier to manage earnings by exercising discretion over the recognition of credit sales compared to the recognition of cash sales. By reducing the change in revenue with the change in receivables, there should no longer be bias in samples where earnings management was performed on revenues. After testing multiple models, they found that the modified Jones-model provided the most powerful tests of earnings management, thus confirming their assumptions. McNichols (2000) discussed the trade-offs of three research designs commonly used to measure earnings mangement, including modified Jones-model. The findings show that the model doesn’t take long-term earnings growth into consideration, wich can be misspecified and result into misleading result. This is not controlled by the current year change of sales and therefor estimed discreationary accruals are significanly associated with analyst projection of long-term earnings growth.

Larcker & Richardson (2004) expanded the modified Jones-model by adding additional proxies that shown to be correlated with measures of unexpected accruals. They included changes in the book-to-market ratio, to control for expected operating growth measured as the ratio of the book value of common equity. It’s expected to see large accruals for growing firms, as an increase in inventory wouldn’t be opportunistic managerial behavior, but due to the firm’s growth operations. They also included operating cash flows, to control for current operating performance, because the measured discretionary accruals at firms that show extreme levels of performance are likely to be misspecified.

Performance adjusted discretionary accruals is another improvement on the Jones-model, recommended by Kothari, Leone & Wasley (2005). The authors chose to develop alternative measures of discretionary accruals, based on (stratified-) random samples. They began with the simple model used by Dechow, Sloan & Sweeny (1995), using cash flow and accruals to adjust the model for the effect of performance on accruals. They choose to incorporate return on assets as performance measure, because a) earnings deflated by assets equals return on assets and b) prior research, analyzing abnormal stock returns and operating performance, finds matching on return on assets results in better specified and give more powerful tests compare to other variables. The results suggest that the performance-matched discretionary accruals measure is a viable alternative to the existing discretionary models, although there where instances where this measure was misspecified.

Jones-performance matching systematically reduces the power of the tests, but the mechanism by which this reduction of power occurs is unknown. This is important, because the users are likely to attribute any loss of power to change and continue to adopt the model. Keung & Shih finds that the loss of power occurs, because the control firms used for performance matching are likely to have more abnormal accruals unrelated to the event of interest than the treatment firm with upwards earnings management. Second, the extent in which the performance matching reduces the power of the tests has been underestimated. Implementing the control firms, that Kothari, Leone & Wasley use, requires knowledge of the treatment firms’ original return on assets, instead of the reported return on assets. This results in a higher loss of power, than suggested by Kothari, Leone & Wasley. Keung & Shih results are consistent whith the two issues discussed. More research is neccassary to develop better solutions.

The models discussed above all measure the discreationary accruals as a sum of aggregated accruals. It’s expected that manager manipulate various components of accruals simultaneously in a consistent matter, for example by positive manipulation in accrual asset accounts or negative manipulation in accrual liability accounts. If firms are actively manipulating more than one accrual, than measuring aggregated accruals may understate the earnings management behaviour. Ibrahim (2009) shows that it is usefull to separate this sum of aggregated accruals into accounts receivable, inventory, accounts payable, other working capital and depreciation. By applying two types of measures, Ibrahim shows that capturing the consistency between the discreationary components of accruals could be usefull, but further research is required to determine how they could be used in better detecting earnings management.

Dechow & Dichev (2002) worked on developing a new measure for the quality of accruals. Because accruals are bases on assumptions and estimates of futures cash flows, they decrease in quality as the estimation error of the accruals increases. Therefor the quality of accruals can be measured as the residuals from changes in working capital of past, present and future operating cash flows. This reveals that accrual quality is systematically related to observable and recurring firm characteristic. Their research on the interrelations between accrual quality, level of accruals and earnings persistence showed that high level of accruals signifies low-quality accruals. This is because accruals are used to solve timing and mismatching issues on cash flows, so large accruals results in higher incurring estimation errors. McNichols (2002) later added the growth in revenue, in an attempt to reflect performance, and property, plant and equipment to expand the model to a broader measure of accruals, that includes deprecation. This increased the explanatory power, reducing measurement error. Francis, LaFond, Olsson, & Schipper (2005) modified the model even further by by decomposing the standard deviation of the residual into firm-level measures of innate estimation errors and discretionary estimation errors. This allowed them to make statements about intentional and unintentional estimation errors by management, which Dechow & Dichev did not. The results show, that investors are able to price accrual quality, because low accrual quality is associated with larger cost of debt and equity.

McNichols & Stubben (2008) and Stubben (2010) use a model measuring discretionary revenues. The model tries to explain the change in accounts receivable, by looking at the annual sales revenue. Any residual is the

discretionary revenue, results in lower reporting quality. Both authors used this model over the discretionary accruals model, because both researches where fixed on investments and investments are closely related to accruals. Still, this model could be used for this paper, because Stubben shows that measures of discretionary revenue show less measurement error and bias than discretionary accruals. Burgstahler, Hail, & Leuz (2006) try to measure the absolute volume of accruals to cash flow. The authors attempted to capture the extent to which accruals are used to reduce the variability of earnings, ass intended. Using accruals for other purposes will decrease the reporting quality. They used a ratio of the standard deviation of operating income dividend, by the standard deviation of cash flows from operations.

Dechow, Hutton, Kim, & Sloan (2012) found that existing techniques for measuring discretionary accruals lack power, because of priors concerning the reversal of accruals. These priors exist because accrual accounting that is being used to manage earnings in one period must reverse in another period. If researchers correctly identifies in which period the earnings management originates and reverses, incorporating the reversal essentially doubles the amount of variation in discretionary accruals that is attributed to earning management. Dechow, Hutton, Kim, & Sloan found that incorporating the reversals increases the test power of earnings management by 40%. This study will use is the cross-sectional variation of the Jones (1991) accruals estimation model. The model is incorporating multiple variables, that are straightforward and easy to acquire, therefor it’s to be expected that a large data sample can be selected. Investing more time and cost on a more complex model is inefficient. Still this is considered to be a limitation on this study.

3.3 Empirical model

The framework for this research can be divided into two regressions. The first regression is the Jones-model (1991), that calculates the discretional accruals (2). The outcome of the discretional accruals calculated, is used in the second regression (3), which determines the correlation between auditors tenure and the discretional accruals. This will be explained in more detail in this paragraph.

The empirical model used by Jones is based on the expectations model used by DeAngelo (1981). DeAngelo measured the abnormal accruals by subtracting the total amount of accruals with the normal accruals. In other words, DeAngelo separated the total amount of accruals in discretionary accruals and non-discretionary accruals: (1) ∆𝑇𝑇𝑇𝑇𝑡𝑡= (𝑇𝑇𝑇𝑇𝑡𝑡− 𝑇𝑇𝑇𝑇𝑡𝑡−1) = (𝐷𝐷𝑇𝑇𝑡𝑡− 𝐷𝐷𝑇𝑇𝑡𝑡−1) − (𝑁𝑁𝑇𝑇𝑡𝑡− 𝑁𝑁𝑇𝑇𝑡𝑡−1)

Where:

- (𝑇𝑇𝑇𝑇𝑡𝑡− 𝑇𝑇𝑇𝑇𝑡𝑡−1) = Change in total accruals

- (𝐷𝐷𝑇𝑇𝑡𝑡− 𝐷𝐷𝑇𝑇𝑡𝑡−1) = Change in discretionary accruals

- (𝑁𝑁𝑇𝑇𝑡𝑡− 𝑁𝑁𝑇𝑇𝑡𝑡−1) = Change in non-discretionary accruals

the abnormal accruals was negative during the event period. DeAngelo found a significant negative change in accruals in year 0, suggesting that managers are making income-decreasing accruals. The results should be taken with a grain of salt, because the changes in earnings, cash flows and revenues were negative as well. Several working capital accounts depend heavenly on the economic circumstances of the firm, including accruals. The industry in which the firm operates influence the way earnings and accruals comes to exist.

Jones (1991) created an expectation model that tries to control for the effects of economic circumstances, unlike the model used by DeAngelo (1981). This will serve as my first regression for this study:

(2) 𝑇𝑇𝑇𝑇𝑡𝑡 𝑇𝑇𝑡𝑡−1 = 𝛼𝛼 + 𝛽𝛽1∙ � 1 𝑇𝑇𝑖𝑖𝑡𝑡−1� + 𝛽𝛽2∙ � ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖𝑡𝑡 𝑇𝑇𝑖𝑖𝑡𝑡−1 � + 𝛽𝛽3∙ � 𝑃𝑃𝑃𝑃𝑅𝑅𝑖𝑖𝑡𝑡 𝑇𝑇𝑖𝑖𝑡𝑡−1� + 𝜀𝜀𝑖𝑖𝑡𝑡 𝑇𝑇𝑇𝑇 Where:

- 𝑇𝑇𝑇𝑇𝑖𝑖,𝑡𝑡 = Total accruals for year t by industry

- 𝑇𝑇𝑖𝑖,𝑡𝑡−1 = Total assets for year t-1 by industry

- ∆𝑅𝑅𝑅𝑅𝑅𝑅𝑖𝑖,𝑡𝑡 = Change in revenues for year t by industry

- 𝑃𝑃𝑃𝑃𝑅𝑅𝑖𝑖,𝑡𝑡 = Property, plant and equipment for year t by industry

- 𝜀𝜀𝑖𝑖,𝑡𝑡𝑇𝑇𝑇𝑇 = Error term for year t by industry

Total accruals include changes in working accounts, such as accounts payable, accounts receivable and inventory. Because these accounts are all dependable on revenue, Jones (1991) includes changes in revenue into the expectation model to control for changes in working capital. Revenue can be affected by earnings management, but earnings itself is not completely exogenous and can therefore still be used as control. The Jones-model also includes changes in property, plant and equipment to control for changes in non-working capital. Jones did include the change in this account, because tot depreciation expense is included in the total accruals measure. All variables in the accruals model are scaled by total assets at t-1, to reduce the heteroscedasticity disturbance; when there are sub-populations that have different variability from others. The residual from regression represents the discretionary part of the accruals. In other words, the part of the total accruals that is not explained by total assets, change in revenue and property, plant and equipment is considered to be the doing of management managing earnings.

For this study, I will use the Jones-model (1991) to determine the total discretionary accruals for the dataset described in paragraph 3.4. The discretionary accruals will serve as a proxy for audit quality, which I´ll will measure for each firm for each year between 2003 and 2013. This gives me the data I need to perform analyses on the second regression:

(3) 𝜀𝜀𝑡𝑡𝑇𝑇𝑇𝑇 = 𝛼𝛼 + 𝛽𝛽1∙ 𝑇𝑇𝑇𝑇𝑡𝑡+ 𝛽𝛽2∙ 𝐹𝐹𝑇𝑇𝑡𝑡+ 𝛽𝛽3∙ �_𝑇𝑇𝑇𝑇𝑇𝑇_𝑡𝑡−1𝑡𝑡� + 𝛽𝛽4∙ �𝐶𝐶𝐶𝐶𝐶𝐶_{𝑇𝑇𝑡𝑡−1}𝑡𝑡� + 𝜈𝜈𝑡𝑡

Where:

- 𝜀𝜀𝑡𝑡𝑇𝑇𝑇𝑇 = Error term for year t; taken from the first regression (2)

- 𝑇𝑇𝑇𝑇𝑡𝑡 = Audit tenure at year t

- 𝐹𝐹𝑇𝑇𝑡𝑡 = Firm age at year t

- 𝑇𝑇𝑇𝑇𝑡𝑡 = Total assets at year t by industry

- 𝐶𝐶𝐹𝐹𝐶𝐶𝑡𝑡 = Cash flow from operations at year t by industry

- 𝜈𝜈𝑡𝑡 = Error term for year t by industry

As mentioned before, the residual part of the first regression (1) contains the portion of the total accruals that is not explained by total assets, change in revenue and property, plant and equipment. In the second regression (2), I’ll try to determine the extent in which the auditors’ tenure determines the residual from the first regression. If the residual from the second regression is at a minimum, than the variable chosen, including auditor tenure, would determine the discretionary accruals. With this regression I’ll test whether the auditor tenure, as variable in the second regression, has a significant impact on the residual from the first regression, namely the discretionary part of the accruals. This will allow me to answer my research question, as prior literature has proven that the discretionary part of accruals can serve as a proxy for audit quality.

In order to control for factors affecting discretionary accruals, that are unrelated to auditor tenure, I include control variables. I include firm age to control for the difference in discretionary accruals of firms with different life cycles and for the firm age itself (Myers , Myjers, & Omer, 2003; Gul, Fung, & Jaggi, 2009). The firm age is measured by each successive year of data available in the dataset. It is therefore assumed that a company exists for as many years as there is data in the dataset Total assets are included because large firms tend to record larger, more stable accruals (Dechow & Dichev, 2002). To control for the negative association between accruals and cash flows, I also include cash flow from operations (Dechow, 1994; Larcker & Richardson, 2004). Just like the first regression, all variables are scaled by total assets at t-1, to reduce the heteroscedasticity disturbance. 3.4 Data

For this research, a sample is constructed from COMPUSTAT. The sample used for this study contains the most recent data for the United States, covering data from companies listed from 2003 till 2013. Consistent with prior research (Ghosh & Moon, 2004); a range of ten years is selected to provide some variations for auditor tenure.

As stated before, after the presentation of European commissioner Barnier (European Commision, 2010) proposals on mandatory auditor’s rotation, the auditor tenure has become a hot topic within the European audit market. From that perspective, it would be interesting to select a data sample from European listed companies to verify whether Barnier assumptions are correct. After extracting test samples from COMPUSTAT for European companies, we’ve found that excluding insufficient data leaves with a rather small data sample to draw good conclusions. The biggest financial data from COMPUSTAT contains data from the North American market. Test samples show that the North American market contains more sufficient data for this research, than

There are many differences between the European and North American market and the debate on mandatory auditor rotation is not actually in North America. Nevertheless, to draw solid conclusions a solid data set is required. Therefor this research will use the North American dataset. The mean objective of this paper is to determine the impact of determining the extent on which the auditor tenure influences the reporting quality. The variables that determine reporting quality and the influence that the auditor has on this, is the same for Europe and North America. Therefor conclusions drawn from the North American dataset can be used to draw conclusions for the European market, making this research relevant for the European auditor profession.

We´ve extracted data from COMPUSTAT for North American listed firms, requesting the following variables required for first and second regression:

- Total assets

- Total property, plant and equipment - Total revenue

- Total cash flow from operating activities - Auditor

- Industry

The data from the above variables is used to calculate the missing data, namely change in revenue, firm age and auditor tenure. To accurately determine the firm and audit tenure, I’ll extract an additional dataset from COMPUSTAT containing all know data regarding known auditors for each firm from 1982 onwards. The auditor tenure will be calculated by determining how many years the same auditor has performed an audit for the same firm. When a firm chooses to go with a different auditor, the auditor tenure will be reset to one year. To determine the firm age, I presume that a company exists for as many years as there is data in COMPUSTAT. Measuring firm and audit tenure this way isn´t fully accurate, but it will suffice for the purpose of this study and should not be considered to be a limitation on this study.

The total database acquired contained 134.864 firm-years divided over 18.548 North American firms. Consistent with prior research, I exclude firm-year observations that do not have sufficient data to compute total accruals or the variables needed to perform the Jones model and the correlation between auditors tenure (Kothari, Leone, & Wasley, 2005; Dechow & Dichev, 2002). This means that if a single data within one firm year is missing, that the firm is filtered out of the dataset. It is there for possible for a firm to exist in one year, but not in the next. The audit and firm tenure are determined separately and are therefore unaffected. To control for outliers, that might influence the data results, I calculated the standard scores (z-values) for all variables and excluded all values above 2.5 from the dataset. A standard score value above 2.5 is considered to be divergent towards the remainder of the database. Lastly, the firms categorized as financial and public administration firms are excluded from the dataset, as these firms financials are affected by regulation. The discretionary accruals for these types of firms are not purely based on economic circumstances and are therefore potentially biased. After filtering out these firms, insufficient data and outliers, the total data consist of 45,412 firm-years divided over

8,595 North American firms. The sample characteristics in the next paragraph will provide a descriptive statistic on this dataset.

The variables calculated within the Jones-model are dependable on the industry characteristics, in which the North American firms operate. The industries are divided in 8 different categorizations according to the North American Industry Classification Code applied by COMPUSTAT. This ranges from manufacturing companies till educational services. The categorization ‘Other industries’ contains all other industries that couldn’t be placed within the other industries, like certain holding and consultancy companies. Some firms may have large residuals because of variation induced by industry classification rather than earnings management or errors. For example, the model may have a poorer fit in growth industries, caused by timing and mismatching issues on cash flows, so large accruals results in higher incurring estimation errors. It is also expected, that the residuals will vary across periods of time. It is expected that the economic circumstances in 2003 will deviate a from the circumstances in 2013, after the economic crisis around 2008. To take the industry classification and economic circumstances into account within the Jones-model, I will run the regressions for each industry classification in the dataset and for each year from 2003 till 2013. A possible downside to running the regression per industry-year is the amount of separate regressions that will be performed. With 10 year in data and 8 different industries, there will be a total of 80 regressions divided over the total dataset, thus dividing the total dataset into smaller datasets. This will decrease the power of the results.

I will not control for industry classification within the second regression. The correlation between auditor’s tenure and the discretional accrual goes about the extent in which the auditor tenure influences the audit quality. As audits remained mandatory, it is not necessary to control for the industry classification in which the data is collected. Auditors have to deliver audit quality no matter in which industry the audit is performed. To maintain the assumption of independence I will perform the regression for each year separately, as the same firm provides multiple data throughout 2003 till 2013.

The dataset used within this study is extracted from COMPUSTAT into an Excel spreadsheet. All variables from both models are then calculated within the spreadsheet, as well as the standard score value to exclude al outliers. All invalid data within the spreadsheet has been labeled and has been excluded into the regression analyses. The spreadsheet was then imported within the statistical program IBM SPSS Statistics in which the regressions where executed.

4 Results

4.1 Sample characteristics

Table 1 provides some descriptive information about the sample distribution of year and length of audit tenure data set used. Looking at the distribution of the amount of firms within one year in table 1.a, it seems that the data is distributed evenly, with no year exceeding 11% of the total dataset, although the number of observations declines throughout 2003 till 2013. The reasons for the diminishing number of firms are the outliers and invalid line items excluded from the dataset. The original dataset was evenly distributed around 9% for each year. A possible explanation could be that the data from the most recent year has not yet been collected. The difference between 2003 and 2013 is not significant. This study takes the effects of specific industries and fiscal years into account, when analyzing the dataset. Therefor the distribution between the different industries within the dataset should not influence the results.

TABLE 1.A DISTRIBUTION BY FISCAL YEAR

No. of observations % of firms

2003 4.981 11% 2004 4.749 10% 2005 4.546 10% 2006 4.460 10% 2007 4.253 9% 2008 4.056 9% 2009 3.953 9% 2010 3.813 8% 2011 3.663 8% 2012 3.661 8% 2013 3.277 7% 45.412 100% Notes:

No. of observations : The factual distribution of the dataset over the fiscal years 2003-2013 % of firms : The relative distribution of the dataset over the fiscal years 2003-2013

Table 1.b shows the distribution over the different types of industry, within the dataset. Almost half of the population consists of manufacturing companies and a quarter consists of information and services companies. Manufacturing companies are expected to use accruals to report for products still under production or for larger orders that will result in revenue in the upcoming book year. For information and services companies accruals can be used to report upcoming revenues from contracts to be signed. There are four industries which only occupy a small part of the dataset, namely agriculture, education, entertainment and other services. In the original dataset, the finance and manufacturing companies occupied about a 36% of the total dataset. The

reason why the information and services companies deliver invalid data is unclear. As explained before, this study takes the effects of specific industries and fiscal years into account, therefor the distribution between the different industries within the dataset should not influence the results.

TABLE 1.B DISTRIBUTION BY INDUSTRY

No. of observations % of firms

1. Agriculture, Forestry, Fishing and Hunting 258 1%

2. Mining, Utilities, Construction 5.623 12%

3. Manufacturing 20.478 45%

4. Wholesale Trade, Retail Trade, Transportation and Warehousing 5.593 12%

5. Information, Services and Management 10.404 23%

6. Educational Services, Health Care and Social Assistance 1.295 3%

7. Arts, Entertainment and Recreation, Accommodation and Food Services

1.528 3%

8. Other Services 233 1%

45.412 100%

Notes:

No. of observations : The factual distribution of the dataset over the industries 1-9 % of firms : The relative distribution of the dataset over the industries 1-9

Looking at table 1.c, it shows that 50% of the firms will keep the current auditor for a period between 1-6 years. It further shows that the number of firms for the audit tenure category decreases when the audit tenure increases. This is consistent with the fact that auditor’s tenure will expire sometime, whether it’s after a small or longer period of time, and will not influence the results.

TABLE 1.C DISTRIBUTION BY AUDIT TENURE

No. of observations % of firms

1-3 years 13,258 29% 4-6 years 10.588 23% 7-9 year 7.374 16% 10-12 year 4,850 11% 13-15 year 3.071 7% 16-18 year 1.979 4% 19 years or more 4.292 9% 45.412 Notes:

No. of observations : The factual distribution of the dataset by audit tenure % of firms : The relative distribution of the dataset by audit tenure

4.2 Descriptive statistics

Table 2 provides descriptive statistics for the data sample. Table 2.a shows the mean and standard deviation analyses per year. Looking at the means throughout 2003-2013, there are a couple figures that stand out. The assets and net cash flow in 2003 are higher from the other years. In 2004 all variables seem to differ from the other years, except net cash flows. This is noticeable, because there was no major shift in economic circumstances in 2004. The revenue, property, plant and equipment and change in 2010 seem to differ from the other years, for which the reason is also unclear.

TABLE 2.A MEAN (STANDARD DEVIATION) BY YEAR

Accruals Assets Revenue PPE ΔAssets NFC

2003 -0,25 0,23 0,97 0,85 2,61 -0,53 (6,35) (7,25) (38,80) (5,43) (75,39) (35,19) 2004 -0,68 0,26 2,82 1,18 3,21 -0,11 (21,61) (7,53) (116,74) (16,35) (59,23) (10,81) 2005 -0,14 0,09 0,28 0,71 1,40 -0,01 (1,32) (0,72) (4,81) (1,29) (3,96) (1,09) 2006 -0,35 0,13 0,49 0,74 1,85 -0,26 (12,93) (2,79) (11,56) (2,54) (24,46) (17,65) 2007 -0,57 0,07 0,28 0,71 1,44 -0,03 (22,96) (0,66) (6,72) (2,11) (4,24) (1,37) 2008 -0,18 0,04 0,12 0,63 1,10 0,02 (1,90) (0,29) (0,58) (1,13) (1,31) (0,49) 2009 -0,12 0,08 -0,02 0,69 1,16 0,02 (4,32) (1,52) (2,90) (2,16) (3,03) (0,88) 2010 -0,32 0,06 4,62 1,38 4,46 0,01 (9,27) (0,64) (261,02) (29,88) (159,39) (3,99) 2011 -0,20 0,08 0,56 1,27 2,09 0,02 (2,34) (0,86) (21,67) (29,61) (46,53) (2,73) 2012 -0,32 0,07 0,44 0,76 1,60 0,02 (7,58) (0,84) (10,80) (2,38) (14,63) (1,55) 2013 -0,21 0,05 0,13 0,71 1,37 -0,01 (1,81) (0,36) (2,00) (1,37) (6,13) (0,84) Notes:

Accruals : Mean of the scaled accruals by year = (earnings – net cash flow) / assetst-1

Assets : Mean of the scaled assets by year = 1 / assetst-1

Revenue : Mean of the scaled revenue by year = earnings / assetst-1

PPE : Mean of the scaled property, plant and equipment by year = property, plant and equipment / assetst-1 ΔAssets : Mean of the scaled change in assets by year = assetst / assetst-1

Table 2.b show the mean and standard deviation analyses per industry. Looking at the means throughout the industries, there aren’t many figures that stand out. With the exception of the variable property, plant and equipment, the variables do not differ from each other. Looking at property, plant and equipment, industry 2 and 7 differ from the other industries. Both industries require investments in fixed assets to generate revenue. The divergence is therefor expected.

TABLE 2.B MEAN (STANDARD DEVIATION) BY INDUSTRY

Accruals Assets Revenue PPE ΔAssets NFC

1. Agriculture, Forestry, Fishing and Hunting -0,06 0,04 0,11 0,72 1,29 0,03 (0,46) (0,22) (0,60) (0,41) (2,19) (0,31) 2. Mining, Utilities, Construction -0,13 0,06 0,28 1,52 1,78 0,11

(3,76) (1,40) (6,71) (9,31) (13,00) (0,99)

3. Manufacturing -0,29 0,13 0,49 0,74 1,89 -0,20

(11,05) (5,02) (26,66) (14,23) (47,69) (18,09) 4. Wholesale Trade, Retail Trade, Transportation

and Warehousing

-0,39 0,05 1,70 0,76 1,73 0,09 (19,92) (1,26) (97,26) (3,18) (29,51) (2,18) 5. Information, Services and Management -0,44 0,15 2,16 0,79 2,68 -0,16

(11,59) (2,22) (160,12) (16,16) (94,80) (11,89) 6. Educational Services, Health Care and Social

Assistance

-0,18 0,06 0,30 0,58 1,28 0,04 (1,85) (0,35) (1,82) (0,48) (1,14) (0,91) 7. Arts, Entertainment and Recreation,

Accommodation and Food Services

-0,28 0,03 0,61 1,54 2,94 0,19 (7,22) (0,20) (20,51) (21,85) (70,83) (3,81)

8. Other Services -0,20 0,04 0,08 0,57 1,10 0,04

(1,04) (0,15) (0,49) (0,54) (0,78) (0,34)

Notes:

Accruals : Mean of the scaled accruals by industry = (earnings – net cash flow) / assetst-1

Assets : Mean of the scaled assets by industry = 1 / assetst-1

Revenue : Mean of the scaled revenue by industry = earnings / assetst-1

PPE : Mean of the scaled property, plant and equipment by industry = property, plant and equipment / assetst-1

ΔAssets : Mean of the scaled change in assets by industry = assetst / assetst-1

4.3 Empirical results

As discussed earlier, this study will be divided into two regressions. The first regression is the Jones-model (1991) that calculates the discretional accruals. The outcome of the discretional accruals calculated, is used within the second regression, which is used to determine the correlation between auditors tenure and the discretional accruals. In this paragraph I will discuss the analyses performed and the results for both regressions. 4.3.1 Jones-model (1991)

Jones (1991) model tries to control the effects of economic circumstances on the total accruals reported. It is expected that there will be differences between the results for specific industries, as certain industries will need larger accruals to present a fair view of the financial reporting. In order to control for the influence of the industry -specific effects, the regression is run for each industry separately. Furthermore, the data of one firm is divided in 10 separate data within the dataset, consistent with data from 2003 till 2013. It is expected that the 10 years of data for one firm will have some sort of correlation. To ensure the assumption of independence, the regression is also run for each year separately.

The result from the Jones model is presented in the table 3.a. Looking at the results, it shows that overall the variables from Jones model explain a good part of the total accruals, with an average coefficient of determination of 52%. The Jones model exists from three in dependable variables, namely assets scaled, revenue scaled and property, plant and equipment scaled. Not all variables have a significant beta on the total accruals. The assets have a significant beta on accruals in all industries. Assets relates to properties which can result in revenue in a later stage. The fact that almost all industries show a significant correlation between accruals and assets was therefore expected. Looking at revenues, a couple of industries show a significant beta, like industry 2, 3, 4 and 5. These industries all rely on revenue to persist and aren’t funded by governement for example. Analyzing the beta between industries within each year, doesn’t show differences that stand out, which is expected as revenues should have the same effect on accruals, no matter in what industry the firm operates. The property, plant and equipment only have a partly significant influence on the accruals. Only industry 2 and 7 show a significant result, with just a minor negative beta. This is consistent with both industries requiring large investments to create revenue, thus needing accruals to fix mismatching issues. Analyzing the results next to the descriptive statistics shows that the industries that occupy the biggest portion of the dataset also show the most significant results for each of the variables. Taking into account that a bigger dataset (N) give the most powerful results, it can be stated that the variables assets, revenue and property, plant and equipment have a significant influence on the accruals.

When analyzing the differences in the beta between the industries within each year, there are a couple of betas that differ from the other industries. Looking at the assets, the betas of industries within one year that stand out are industry 2 in 2011, industry 3 in 2004 and industry 8 in 2006 and 2010. When analyzing the betas of revenue within one industry throughout 2003 till 2013, it’s the betas of 2006 in industry 5 and industry 8, 2007 in industry 4 and 2009 in industry 2 that differ. Lastly when comparing the property, plant and equipment betas of

one industry throughout 2003 till 2013, the betas that stands out are industry 4 in 2007, industry 5 in 2012 and industry 7 in 2010. Why these figures differ from the other industries is unclear, as only a small portion is explained by analyzes in the descriptive, like the differences of the variables means in 2010. The differences in the means in 2004 isn´t shown in the results of 2004 as might have been expected.