Word count: [9.277] Date: 16 June 2017

(1)

MSc Accountancy & Controlling

Abstract

Audits conducted by audit firms are one of the main instruments by which financial markets attain assurance over financial information. The focus of this study is to determine to what extent the audit partner individually and the audit firm as an institution are responsible for the assurance provided by the audit firm. Using a sample consisting of 857 German companies and 5.931 firm-year observations this paper studies the effect a rotation of audit firm and audit partner has on audit quality. The findings of this study suggest that the audit firm (and the institutional knowledge it has) is the main determinant of audit quality. More controversial, they also suggest that a rotation of audit partner is sufficient to preserve auditor independence and thus to maintain audit quality over time.

Keywords: audit firm rotation, audit partner rotation, audit quality

Personalia: Tom Pat (s2800772) Institute: University of Groningen Supervisor: dr. C.A. Huijgen

Co-assessor: N. Hussain, PhD

Word count: [9.277]

(2)

1. Introduction

This chapter gives a detailed overview of the subject that is at study. Background information is given explaining the context of the problem, why it is relevant and how this study will contribute to existing research. It concludes with the main research question and a description of the structure of the paper.

1.1 Problem statement

There is a need within society for expert and independent examination and attestation of the financial information that is reported by companies (Francis et al., 2003). Audits conducted by audit firms are one of the main instruments by which financial markets attain this assurance. Auditors thus possess a crucial economic role within society as a confidential agent of the community at large (Limperg, 1985).

Public confidence in the auditors expertise and independence is conditional for the existence of the function. As Limperg (1985, p. 16) states: “The confidence in the effectiveness of the audit and in the opinion of the accountant [read: auditor] thus forms the raison d'etre of his function.” This confidence is to be preserved by the auditor through, in adherence with society‟s expectations, the provision of reasonable assurance over financial information. Reasonable assurance is however only attainable if the auditor exhibits certain levels of audit quality in conducting audits.

After a series of accounting scandals at the beginning of the twenty-first century many regulatory bodies were suspect that audit quality provided by audit firms was not up to society‟s needs. As such, regulations were introduced with the intention to increase audit quality while also preserving it over time (SOx, 2002; Code-Tabaksblat, 2003).

Regulatory bodies‟ main criticism was that the independence of auditors was impaired; contending that the threat of familiarity between auditor and client may damage independence and thus audit quality over a longer period of time. As a response, most regulators adopted restrictions on the duration of auditor-client relationships to increase audit quality (European Commission, 2010).

In practice regulatory bodies have either chosen for a mandatory rotation of audit partners within the audit firm (United States) or more rigorously for a mandatory rotation of audit firms (European Union). This study will focus on this regulatory gap and study the difference in impact a rotation of audit partner or audit firm has on audit quality. The main goal is to gain insight in to what extent the audit partner individually and the audit firm as an institution are responsible for audit quality.

(4)

Page | 4

1.2 Relevance

There has been a vast amount of studies on the effect of audit firm rotation on audit quality (Chan et al., 2014; Edwards, 2014; Corbella et al., 2015; Kim et al., 2015; Cameran et al. 2016; Reid & Carcello, 2017). The empirical findings of these studies are however all but conclusive. Chan et al. (2014) find that audit firm rotations could significantly disrupt normal operations. Corbella et al. (2015), Kim et al. (2015) and Cameran et al. (2016) found evidence suggesting a long term improvement in audit quality. Edwards (2014) denotes both the possible costs and increments of audit firm rotation. Reid & Carcello (2017) however find that investors generally react negatively to the possibility of mandatory audit firm rotation. This last finding is important as it is worth noting that while all mentioned research referred to mandatory audit firm rotation, most of these studies were conducted in non-mandatory environments.

A small amount of studies has done research on the effect of audit partner rotation (Chi et al., 2009; Gul et al., 2013; Lennox et al., 2014; Arthur et al., 2017). These studies are however mostly based in a Chinese setting because of their disclosure requirements regarding the identity of audit partners. This area is thus still relatively unexplored because of scarce data availability. Litt et al. (2014, p. 59) states: “research on the effect of audit partner rotation… ….is virtually non-existent, largely due to the absence of publically available information on audit partners.” It is thus scientifically relevant to study the singular impact of both audit firm and audit partner rotation on audit quality. This study additionally aims to expand existing research by also addressing the differences in the effect both types of rotation have on audit quality.

There is also societal relevance to be found in researching the effect that audit firm and audit partner rotation have on audit quality. Regulatory bodies remain divided as to which form of auditor rotation should be mandated to instil and maintain high levels of audit quality in financial markets. As this study researches both types of rotation, it could provide guidance as to which may be best to accomplish this.

1.3 Contribution

This study will contribute to existing literature twofold. First, the main purpose of this study is to assess both audit firm and audit partner rotation and the affect either has on audit quality. These relations and the differences between them are interesting to study because they give insight in the importance of the audit partner individually as opposed to the audit firm as an institution and which of both is the main determinant of the audit quality provided by the firm. The findings of this study will also be interesting for regulatory bodies in guiding them as to which of both restrictions on the audit-client relationship is preferable in terms of being the most effective method of institutionalising high levels of audit quality. As far as I am aware this study is the first to research the effect of both audit firm and audit partner rotations.

(5)

Page | 5

Second, this study will provide additional scientific evidence on the effect of audit partner rotation by using a large sample in a Western institutional setting. This study is able to obtain a large sample because Germany already has additional disclosure requirements providing the identity of external auditors starting from 1998 (HGB, 2016, § 322). The German setting is interesting because previous research was mostly conducted in a non-Western Chinese setting (Chi et al., 2009; Gul et al., 2013; Lennox et al., 2014). As Hooghiemstra et al. (2015, p. 360) states: “the nature and characteristics of agency conflicts may be culturally determined.” The cultural dimensions individualism and uncertainty avoidance by Hofstede (2001) are according to Hooghiemstra et al. the most explanatory in terms of exacerbating agency conflicts. China differs vastly from Germany on both these dimensions. China scores respectively 20 and 30 on individualism and uncertainty avoidance as opposed to Germany scoring respectively 67 and 65. The auditor has an essential role in solving existing agency conflicts and as such the effect may be vastly different in the German setting (with more agency conflicts) as opposed to the Chinese setting (with less agency conflicts).

1.4 Research question

The research question of this study is: ‘To what extent are the audit firm as an institution and audit partner individually responsible for provided audit quality?’

The focus of this study is to determine to what extent the audit partner individually and the audit firm as an institution are responsible for audit quality. The root source of audit quality is best studied indirectly by studying disruptions of the audit-client relationship on both levels. The effect a change of audit firm or partner has on audit quality will give insight in to what extent both were responsible for the provided audit quality before given change. Thus, to be able to answer the research question, this study will focus on the effect a rotation of audit firm or audit partner will have on audit quality.

1.5 Structure of the paper

The structure of the remainder of the paper is as follows. In chapter two the theoretical framework will be elaborated and explained including the hypotheses of the study. Chapter three describes the methodological framework that is used to conduct the statistical analyses. The empirical findings of the study are reported in chapter four and chapter five concludes with a summary and discussion of the main findings of the study.

(6)

Page | 6

2. Theoretical framework

In this chapter the theoretical framework that underlies the study is elaborated. It will start with an agency perspective of auditing and consecutively review the theory underlying audit quality and the impact of auditor rotation. The chapter concludes with the hypotheses of the study.

2.1 Agency perspective of auditing

As noted in the introduction there is a need within society for expert and independent examination and attestation of the financial information that is reported by companies (Francis et al., 2003). The need for attested financial information can be attributed to the (agency) conflicts that arise from the separation of ownership and control within a company. (Armstrong et al., 2010).

Smith (1776) noted already in the 1700‟s that separation of ownership and control within a company might cause the manager to make decisions contrary to the interests of the owners. Agency theory attempts to explain the relationship between principal (c.q. owners) and agent (c.q. management) and makes propositions on how to resolve the (agency) conflicts which might arise from it. Scott (2015, p. 358) defines agency theory as: “theory that studies the design of contracts to motivate a rational agent to act on behalf of a principal when the agent‟s interests would otherwise conflict with those principal.”

Agency conflicts play a dominant role within companies because of the existence of information asymmetry between management and contracting parties like equity holders and debt holders (Ndofor et al., 2015). The company‟s management typically possesses better company-specific information than outside contracting parties, which they may not share because it may be detrimental to the personal interest of management (Verrecchia, 2001). Thus, to be able to monitor management effectively, outside contracting parties needed a device to ensure that the assertions by management do indeed faithfully represent the company‟s performance.

Auditing has taken onto itself this credibility-adding role. Watts & Zimmerman (1986) argue that audited financial information is able to reduce existing information asymmetries between management and outside contractors. An independent audit gives outside contractors the reasonable assurance they desire that information stated by management is fair and complete (Eilifsen, 2010). Healy & Palepu (2003) in conclusion suggest that auditing has become a critical component of well-functioning capital markets.

(7)

Page | 7 2.2 Audit quality

In the framework for audit quality released by the International Auditing and Assurance Standards Board (IAASB, 2013) appropriate audit quality is established when users of financial statements can rely upon the auditor‟s opinion, knowing that it was based on sufficient appropriate evidence.

DeAngelo (1981, p. 186) states that audit quality is “the market-assessed joint probability that a given auditor will both (a) discover a breach in the client‟s accounting system, and (b) report the breach.” In this, in academic research widely used definition, audit quality is thus a combination of the competence and independence of the auditor.

Auditor competence is defined as a state of expertise that is sufficient to achieve explicit audit objectives (Lee & Stone, 1995) Flint (1988) states that, to conclude an audit, the auditor must have sufficient knowledge, information, qualification and experience. It is also linked to the level of effort devoted by the auditor according to Knechel et al. (2013, p. 388) as: “the discovery of a misstatement requires that appropriate resources be effectively utilized in the audit process.”

Auditor independence is the probability that an auditor will report a misstatement when discovered (DeAngelo, 1981). Watts & Zimmerman (1981) conclude that it is unlikely that auditors are perfectly independent from their clients. Audit firms have a future economic interest in the client and switching costs are significant due to necessary investments in the early years of an audit. The higher the economic interest and switching costs, the lower will be the probability that the auditor will report a breach (Watts & Zimmerman, 1981). Independence is observed as crucial and more close to the essence of auditing by both Schandl (1978) and Boritz (1992) and could thus be deemed more important than competence.

2.3 Audit quality in relation to financial reporting quality

DeAngelo‟s (1981) definition portrays auditing primarily as a two-phased process and implicitly states that the auditor only has a responsibility with regard to compliance with accounting standards. While her definition still largely conforms, the profession has not remained idle over the years. A recurring debate regarding the responsibilities of auditors (Chandler & Edwards, 1996), was reignited after accounting scandals at the beginning of the twenty-first age.

Defond & Zhang (2014, p. 280) argue that a high quality audit should not confine itself to whether: “the client‟s accounting choices are in technical compliance with GAAP, but also how faithfully the financial statements reflect the company‟s underlying economics.” This is consistent with auditing standard No. 14,

(8)

Page | 8

promulgated by the Public Company Accounting Oversight Board in 2010, which requires auditors to evaluate the qualitative aspects of a company‟s accounting practices (PCAOB, 2010). Jurisprudence also confirms the wider scope of responsibilities as the Supreme Court held auditors legally liable stating: “mere compliance with professional accounting standards was not a sufficient defense [sic], and the appropriate test was whether the financial statements fairly represented …‟s financial position” (Ball, 2009, p. 303).

The above statement extends the auditor‟s responsibility to financial reporting quality and implicitly states that audit quality is an important determinant of financial reporting quality. The achievable level of financial reporting quality is however constrained by the quality of the pre-audited statements that are provided to the auditor (Defond & Zhang, 2014). Defond & Zhang further state that the quality of pre-audited statements are dependent on the quality of the company‟s financial reporting system and the innate characteristics of its business.

To conclude, while the auditor has a responsibility with regard to financial reporting quality, financial reporting quality does not necessarily equal audit quality. Instead, financial reporting quality is a function of audit quality, quality of the financial reporting system and a company‟s innate characteristics.

2.4 Audit quality in relation to auditor rotation

One theme that serves as a continuous catalyst for controversy within the profession is that of (impaired) auditor independence and the undermining effect is has on audit quality (Williams & Wilder, 2016). It is first suggested by Mautz & Sharaf (1961) that the objectivity of the auditor may diminish with passage of time. Hoyle (1978) states that the audit programme might become a mere routine in lengthy auditor-client relationships. A rotation of auditor has since then been argued to increase overall audit quality due to an increase in auditor independence, decrease in auditor complacency and supporting professional scepticism (Shockley, 1981).

The downward effect of rotation has however also been emphasized. Rotation would also lead to a loss in auditor competence (Carcello & Nagy, 2004). During the start of the auditor-client relationship the auditor still needs to acquire crucial knowledge about the company, which would thus impair audit quality in the first years of engagement. Complementary to this, Myers et al. (2005) find that financial reporting problems are also more likely to occur during these years.

(9)

Page | 9

While the effect of long client-tenure on audit quality is not yet conclusive, there is considerable evidence that audit quality is impaired in the first years of auditing a client (Johnson et al., 2002; Davis et al., 2009). Thus, in the context of this study, I conclude that a rotation of auditor is expected to have at least a short term negative effect on audit quality.

2.5 The audit firm and audit partner

Thus far I consistently exert the term auditor without making a further distinction towards either the audit firm or audit partner. The reason for this is that both firm and partner carry similar (if not the same) responsibilities with respect to the profession and society. However, the question remains as to whom of both is most responsible for the level of audit quality that is provided. Lorsch & Tierney (2002, p. 14) argue that professional service firms are: “knowledge engines for business.” As such, an audit firm builds institutional knowledge about the client over the years that it is providing audits for that client. DeAngelo (1981) supports this view by arguing that audit quality also depends on audit firm size as bigger firms are economically less dependent on each individual client. Francis & Yu (2009) find support for the importance of the audit firm by finding that bigger offices provide higher quality audits.

The audit partner however has a decisive role in the service the audit firm provides. This person is responsible for the process of the audit and the eventual opinion given. The existing research (Gul et al., 2013; Lit et al., 2014) supports the importance of the audit partner by providing evidence of lower reporting quality after an audit partner rotation. However, to which extent the audit partner contributes to the level of audit quality provided by the audit firm is not yet clear.

Research has shown that in the first years of auditing a client audit quality will be impaired (Johnson et al., 2002; Carcello & Nagy, 2004; Davis et al., 2009), thus both audit firm and audit partner rotations are expected to have a negative effect on audit quality shortly after the event. Additionally, after an audit firm rotation there will be no retention of client knowledge. Ceteris paribus, an audit firm rotation is expected to lead to a stronger impairment of audit quality then an audit partner rotation and thus I hypothesize:

Hypothesis 1: the rotation of an audit firm has a short-term negative effect on audit quality. Hypothesis 2: the rotation of an audit partner has a short-term negative effect on audit quality.

Hypothesis 3: the rotation of an audit firm has a more dominant short-term negative effect on audit quality than the rotation of a partner individually.

(10)

Page | 10

3. Methodological framework

This chapter will elaborate on the methodologies used for this study and account for the established validity and reliability. It will first elaborate on the sample used for the analyses and second thoroughly discuss and attest the research design of the study.

3.1 Sample selection

A dataset of German companies listed on the Frankfurt Stock Exchange (Frankfurter Wertpapierbörse) during the years of 1998 – 2010 has been made available for this study. This study was able to obtain a large dataset because Germany already has additional disclosure requirements providing the identity of external auditors starting from 1998 (HGB, 2016, § 322). All data within the dataset was manually collected from public financial statements.

The German government introduced mandatory audit partner rotation in 1998. Rotation of the audit partner is mandatory after seven consecutive years of auditing a client followed by a three-year cooling off period (HGB, 2016, § 319a). The time-series within the dataset is before the introduction of the German Audit Reform Act (Abschlussprüfungsreformgesetz) in 2014, which requires mandatory audit firm rotation, thus audit firm rotation is still only voluntarily. Audit client-tenures within the sample are consequently expected to be overall quite long.

Germany has a highly concentrated market of auditors. In a time-series analysis of public firms from 1997 to 2007 the big four has a clear dominance with a cumulative market share of 93% (Velte & Stiglbauer, 2012). Among the big four PWC and KPMG lead the market having an average market share of 52% based on audit assignments, while Deloitte and Ernst & Young have much less and the largest non-big four is even classified as rather insignificant (Quick & Sattler, 2011).

The German setting is characterised as high in terms of agency conflicts. According to Hofstede (2001) Germany has both high individualism and uncertainty avoidance. These cultural dimensions are the most explanatory in terms of exacerbating agency conflicts (Hooghiemstra et al, 2015). Additionally, Germany is a country with low litigation risk (Defond & Zhang, 2014). This further exacerbates agency conflicts as this lowers incentives for auditors to maintain independence (Hope & Langli, 2010). Altogether, these conditions put more emphasis on the importance of the auditor within society and thus may lead to more clear inferences of changes in audit quality.

(11)

Page | 11

The dataset contains financial information from the period of 1998 to 2010. It does not contain companies which are in regulated and financial industries. Credentials of audit partners became public onward from 1998, thus these observations start in 1999. They end after 2009. The initial sample consists of 857 companies and 5.931 firm-year observations. As this study focuses on the effect of a rotation of auditors there will be some loss of data within the dataset until the occurrence of the first rotation. To measure the impact on audit quality in a valid manner the research design focuses on a three year window after a rotation of audit firm or partner. Additionally, because client-tenure is not known at the start of the sample I exclude the first three firm-year observations unless a rotation takes place during this period. This is done to eliminate potential bias from rotations that took place just before the sample period.

With regard to the sample of audit firm rotations, I exclude a firm-year observation if an audit partner rotation takes place within the assigned audit firm. Not doing this would lead to interferences from audit partner rotations in the audit firm rotation sample and consequently create measurement noise. In line with this, I also exclude all audit partner rotations that were due to a rotation of audit firms from the audit partner rotations sample. All institutional knowledge of the audit firm is lost after these rotations making these audit partner rotation observations essentially audit firm rotations. Including them would wrongly attribute the measured effect to audit partner rotations and possibly skew results.

Lastly, in conformance with both Bertrand & Schaor (2003) and Gul et al. (2013) an audit partner must meet certain conditions to become a part of the sample. First, he/she needs to have audited a client for at least three consecutive years and second there are at least three years in which someone else has. The first criteria is imposed to give the audit partner the chance to „imprint his mark‟ on the audit and the second to maintain a balanced sample of pre- and post-rotation observations.

Considering all the above, the research period will start in 2000 and end after 2009. The final audit firm rotation sample consists of 2586 firm-year observations and 492 audit firm rotations and the final audit partner rotation sample 2234 firm-year observations and 272 audit partner rotations.

3.2 Research design

3.2.1 Independent variables: audit firm and audit partner rotation

The audit firm as well as the partner that conducted the audit of the financial statements are disclosed by public companies in Germany. As the dataset is based on these disclosures it contains the name of both the audit firm and partner. All companies within the dataset are identified based on unique Worldscope codes to be able to also oversee mergers or name changes if they occur during the research period. All audit

(12)

Page | 12

firms and partners within the dataset possess a unique numeric code to ensure that there are no misconstructs within the dataset due to a similarity in names.

I measure a rotation of an audit firm or audit partner with an indicator variable that activates in the year after a rotation took place. To measure the effect on audit quality the research design focuses on a three year post-rotation window. In accordance, the indicator variable deactivates again once this post-rotation window closes. As client-tenure at the beginning of the sample is not known, the first three firm-year observations of each company are excluded from the sample (unless a rotation takes place) to avoid potential bias due to rotations that took place just before the research period. Both variables were constructed based on the available data within the dataset. Rotations of audit firms and audit partners were identified in a structured manner using formulae which indicate when a change occurred. The activation (and deactivation) of both indicator variables were constructed accordingly.

3.2.2 Dependent variable: audit quality

It is difficult to determine audit quality in a direct manner because the amount of assurance provided by auditors is not observable (Defond & Zhang, 2014; Knechel et al., 2013). Researchers thus have had to make use of proxies to measure audit quality. Behn et al. (2008) state that it is common to consider the overall quality of financial reporting. The argument for using financial reporting quality as a proxy for audit quality is that high quality auditing constrains opportunistic earnings manipulation. This method has a significant limitation in that financial reporting quality is a function of audit quality, quality of the financial reporting system and the company‟s innate characteristics (Defond & Zhang, 2014). As such, it does not directly infer audit quality and requires several control variables to be able to disentangle these other elements.

There are however also arguments in favour of using financial reporting quality. First, in the previous chapter I established that the auditor also has a responsibility with regard to financial reporting quality. The given assurance hence should be as such that the financial statements faithfully represent the company‟s financial position. Second, it goes beyond compliance and is also able to measure possible earnings manipulation that is technically not in violation with GAAP. The third argument relates specifically to the use of discretionary accruals as is the case in this study. These are due to their continuous nature able to capture variations in audit quality within every firm-year observation in the sample and not restricted by observations that only indicate the lower bound of the audit quality continuum (i.e. restatements or going concern opinions).

(13)

Page | 13

Financial reporting quality is most commonly measured by referring to the discretionary accruals of a company (Knechel et al., 2013). Discretionary accruals are the accrual portion of earnings that can be manipulated by management through financial reporting choices (Hooghiemstra et al., 2008). The main objective of auditors should be to ensure that the annual report faithfully represents underlying reality. Discretionary accruals represent a form of earnings manipulation by management which would do this objective harm. Francis & Michas (2013) find a basis for this view by finding a relation between higher levels of discretionary accruals and past restatements of annual reports. Discretionary accruals are thus a reasonable proxy to measure financial reporting quality.

To estimate the level of discretionary accruals the model proposed by Jones (1991) will be used as modified by Dechow et al. (1995). More specifically this method estimates the component of total accruals that is non-discretionary and consequently measures the discretionary accruals indirectly as a residual. This model is generally accepted as most accurate when determining the discretionary accruals of financial statements (Scott, 2015). The mathematical equation of this model is as follows:

( ) [ ( ) ] ( ) (1)

In equation (1), TAijt are the total accruals. The component of the total accruals that is non-discretionary is estimated based on variables that indicate non-discretionary items (cash revenue growth and property plant and equipment) along with industry parameters (α0, α1, α2, α3). Lastly, the component of accruals that

is discretionary is measured by the residual error term εijt. The subscript after each variable denotes the measurement for company i in industry j in year t or t – 1. All variables within equation (1) are scaled by the company‟s total assets at prior year-end (Aijt – 1) to eliminate firm size effects. The other terms within equation (1) are as follows: ∆REVijt is the revenues in year t minus revenues in year t – 1; ∆RECijt is net receivables in year t minus net receivables in year t – 1; PPEijt is gross property plant and equipment at the end of year t.

Prior studies (Johnson et al., 2002; Myers et al., 2003) found that the industry in which the company operates influences the level of accruals. Thus, the estimation of the parameters α0, α1, α2, α3 is conducted

in a cross-sectional manner. Industry classification is based on SIC-codes, as modified by Barth et al. (1998). The parameters are calculated once for each industry based on the entire research period. This decision was made because separate estimations for each individual year would lower the amount of observations for each parameter estimation significantly and thus arguably do more harm than good.

(14)

Page | 14

After the industry parameters are estimated for each industry the non-discretionary component of accruals ( ) can be calculated for each firm-year observation within the dataset. This is done by completing the mathematical equation for non-discretionary accruals once more with the corresponding industry-based parameters. The industry parameters are an indication of the amount of accruals that is systematic within the industry the company operates.

Lastly, discretionary accruals are determined with the following mathematical equation:

(2)

In equation (2), is the estimation of discretionary accruals which is calculated by subtracting the total accruals with the estimated non-discretionary accruals . This is the final outcome of the Modified Jones Model, as it was termed by Dechow et al. (1995).

3.2.3 Control variables

As financial reporting quality is a construct of multiple components (of which one is audit quality) it is critical to control for the other elements to be able to disentangle the effect of these confounding variables on financial reporting quality. Simultaneously, this isolates the effect on audit quality and lowers measurement error.

The control variables in this study conform to the model suggested by Defond & Zhang (2014) for audit quality research based on discretionary accruals. The variables in this model are likely to be correlated with the constricting elements of financial reporting quality.

Watts & Zimmerman (1986) argue that larger companies should exhibit lower levels of accruals because of more extensive public scrutiny. This statement is also corroborated by prior research (Johnson et al., 2002; Myers et al., 2003; Dechow et al., 2012). Thus I include SIZE. Following Johnson et al. (2002) it is expected that more leveraged companies have greater incentives to manipulate earnings to avoid breaching debt covenants. As such I include LEVERAGE. Following this line of reasoning; companies experiencing losses (Carey & Simnet, 2006) and those with a higher probability of bankruptcy (Carcello et al., 1995) also have increased incentives to manipulate earnings. Thus I include LOSS and BANKRUPTCY. BANKRUPTCY is based on the model proposed by Zmijewski (1984).

(15)

Page | 15

Dechow (1994) finds that accruals and cash flows are negatively correlated on average, thus I also control for CASHROA. Additionally, Dechow et al. (1995) argue that discretionary accrual models fail to completely extract non-discretionary accruals that are correlated with firm performance. Thus I add PERFORMANCE to control for this limitation. Myers et al. (2003) find that accruals are positively related with growth-opportunities, thus I add GROWTH. Francis et al. (1999) and Gul et al. (2013) both find that larger audit firms are more conservative, therefore less accepting of discretionary accruals and thus I include BIG4.

Bartov et al. (2001) argues that a high book-to-market ratio attracts investors (as it indicates asset undervaluation) and thus also increases the litigation risk of the auditor. The auditor will then be less inclined to allow earnings manipulation, thus I control for BOOKTOMARKET. Following Gul et al. (2009) I include YEAR to control for year-fixed macro-economic effects. The definition and measurement of all control variables used for this study can be found below.

Definition control

variables Measurement

SIZE Natural logarithm of the book value of total assets at year-end.

LEVERAGE Total liabilities divided by total assets at year-end.

LOSS Indicator variable that indicates if a company reported a loss in prior year.

BANKRUPTCY Probability of bankruptcy as measured by Zmijewski scores.

CASHROA Cash flows from operations divided by the total assets at year-end.

PERFORMANCE Earnings before tax divided by the total assets at year-end.

GROWTH The growth rate of revenues based on prior years’ revenues.

BIG4 Indicator variable that indicates if a company is audited by a big four audit firm. BOOKTOMARKET The book-value of equity divided by the market-value of equity

YEAR Indicator variable that indicates the year over which the financial statements

report.

3.2.4 Regression model

To empirically test the hypotheses, I propose a multivariate analysis with as dependent variable the absolute value of discretionary accruals as determined by the Modified Jones Model. This choice is made on the premise that both upward and downward earnings manipulation lowers financial reporting quality (Myers et al., 2003). The explanatory side of the equation includes the treatment variables and necessary control variables and a residual term εit. As an overview, the framework of the regression model is as follows:

( ) ∑

(16)

Page | 16

In order to test the relationship between audit firm and audit partner rotation and audit quality it is necessary to first test the intervening influence of control variables on discretionary accruals, as they can then be held constant when I test the hypotheses. Accordingly, the first regression model consists solely of control variables:

Where in equation (4), is the absolute value of discretionary accruals and all other definitions have remained as before. After this, the linear regression models in equation (5) and (6) are used to estimate the singular association between audit firm and audit partner rotations and audit quality. To first test the hypotheses I thus use:

Inhere, FirmRot and PartnerRot are the indicator variables used to measure the effect of respectively an audit firm or audit partner rotation. Lastly, the multivariate regression models in equation (7) and (8) contain the complete model which is used to analyse the effect on audit quality in a more substantial manner. The sample of audit firm and audit partner rotations are not entirely identical, due to differences in requirements and the elimination of interfering observations. As such, I cannot use both variables in the same regression model and am obligated to measure the impact of both types of rotations in separate regressions: (4) (5) (6) (7) (8)

(17)

Page | 17

4. Empirical findings

This chapter will present and discuss the empirical findings of the study. Statistical analyses were conducted to ascertain whether there is evidence in support of the established hypotheses. The chapter concludes by discussing the robustness of the findings with additional sensitivity analyses.

4.1 Descriptive statistics

The descriptive analysis is conducted to get a better understanding of the dataset used for this study. Table 1 and 2 present measures of central tendency and spread for all variables included in the complete regression models. Univariate outliers were dealt with by winsorizing all continuous variables at three standard deviations. The missing observations with regard to the independent variables are due to sample-requirements, as with the control variables they are due to divisions by zero or missing audit firm data.

The distribution of audit firms within the sample corroborates prior research in that the German market is highly concentrated. The sample consists of 245 individual audit firms. The big four were responsible for 90% (11 billion) of the accumulative company-revenues within the sample as opposed to the 10% (1,2 billion) that all other 241 audit firms were responsible for. It is even more extreme as PWC and KPMG, when taken together, are solely responsible for 75% (9,1 billion) of the company-revenues within the sample. When looking at individual audit assignments the absolute difference is less significant as the big four have conducted 2.726 assignments and all other firms 2.287.

Table 1: Unsegregated descriptive statistics

n Mean _DeviationStd. Minimum Maximum

Valid Missing |DA| 5930 0 ,015 ,035 ,000 ,312 FirmRot 2586 3344 ,377 ,485 ,000 1 PartnerRot 2234 3696 ,359 ,480 ,000 1 SIZE 5930 0 5,048 2,125 -,740 11,458 LEVERAGE 5930 0 ,585 ,240 -,324 1,507 BANKRUPTCY 5928 2 -,962 1,719 -7,915 6,136 LOSS 5929 1 ,299 ,458 ,000 1 CASHROA 5930 0 ,060 ,167 -,535 ,648 PERFORMANCE 5930 0 ,000 ,169 -,776 ,755 GROWTH 5915 15 ,102 ,631 -1 10,520 BIG4 5013 917 ,544 ,498 ,000 1 BOOKTOMARKET 5928 2 ,726 ,703 -2,171 3,630

The unsegregated descriptive statistics reveal no peculiarities within the sample. As it should, the estimations of accruals with the modified jones model do not exceed one. All indicator variables do also

(18)

Page | 18

not exceed one. The variables SIZE, LEVERAGE and GROWTH are all adequate with no minimum below one. All other remaining control variables are also seemingly normally distributed based on mean, standard deviation and the minimum and maximum observation. To get a more thorough understanding of the sample the statistics are also presented segregated toward both audit firm and audit partner rotation.

Table 2: Segregated descriptive statistics

FirmRot PartnerRot

Pre-rotation Post-rotation Pre-rotation Post-rotation

_Mean _{Deviation Mean}Std. _DeviationStd. _Mean _{Deviation Mean}Std. _DeviationStd.

|DA| ,012 ,033 ,015 ,033 ,013 ,033 ,009 ,023 SIZE 5,493 2,206 4,572 1,909 5,272 2,148 5,864 2,178 LEVERAGE ,593 ,214 ,596 ,272 ,579 ,218 ,621 ,212 BANKRUPTCY -1,028 1,517 -,860 1,955 -1,107 1,527 -,864 1,509 LOSS ,220 ,415 ,386 ,488 ,242 ,429 ,237 ,425 CASHROA ,077 ,139 ,061 ,182 ,076 ,150 ,077 ,139 PERFORMANCE ,023 ,133 -,010 ,175 ,023 ,142 ,025 ,126 GROWTH ,047 ,376 ,099 ,824 ,061 ,454 ,035 ,219 BIG4 ,561 ,493 ,548 ,498 ,564 ,496 ,661 ,474 BOOKTOMARKET ,763 ,675 ,710 ,818 ,753 ,646 ,731 ,696 n 1610 976 1432 802

These statistics suggest that the average amount of accruals is higher after a rotation of audit firms. This is expected as in the first years of auditing a client crucial knowledge is still unknown. More noteworthy is that the average amount of accruals is lower in the sample with observations after a partner rotation took place. This could possibly be due to an increase in professional scepticism in the post rotation window along with a substantial preservation of auditor competence. This would however suggest that audit quality is not significantly impaired after an audit partner rotation and consequently would attribute the audit firm (and the institutional knowledge it has) as the main determinant of audit quality.

The control variables may also be partially responsible for the difference in accruals. The size of a client is expected to be negatively correlated with the amount of accruals and might as such intervene in the relationship. All control variables indicative of financial difficulties (LEVERAGE, BANKRUPTCY, LOSS) are on average slightly higher in the post rotation window but no substantial differences can be noted. What is remarkable is that the average amount of growth is almost double post audit firm rotation and the inverse is observed post partner rotation. This could also skew the descriptive results of accruals slightly. Big four audit firms are more dominant within the post audit partner rotation sample, possibly indicating that audit partner rotations occur more frequently within the big four as opposed to non-big four audit firms. Lastly, the book to market ratio of clients does not differ significantly among both samples.

(19)

Page | 19 Table 3: Pearson correlation matrix

|DA| Fir mR ot Pa rtn er R ot SIZE LE VER A G E B A N KR UP TCY LOS S CA SHR OA PER FOR M A N CE G R OW TH BIG₄ B OO K TO M A R KET |DA| 1 FirmRot ,044 ** ₁ PartnerRot -,059 ***_-,206*** ₁ SIZE -,250 ***_-,207***_,131*** ₁ LEVERAGE -,054 ***_,007 _,094***_,189*** ₁ BANKRUPTCY ,008 ,048 **_,076***_-,031**_,872*** ₁ LOSS ,126 ***_,181***_-,006_-,298***_,118***_,287*** ₁ CASHROA -,041 ***_-,049**_,003 _,155***_-,064***_-,353***_-,261*** ₁ PERFORMANCE -,101 ***_-,107***_,007 _,259***_-,179***_-,609***_-,423***_,619*** ₁ GROWTH ,171 ***_,045**_-,033_-,005_-,062***_-,094***_,001_-,009 _,092*** ₁ BIG4 -,045 ***_-,013 _,095***_,355***_,062***_,011_-,086***_,078***_,088***_-,030** ₁ BOOKTO MARKET -,048 *** -,035 ** -,016 -,006 -,284 *** -,212 *** ,021 ,001 -,024 * -,046 *** -,046 *** 1

***. Correlation is significant at the 0.01 level (2-tailed). **. Correlation is significant at the 0.05 level (2-tailed). *. Correlation is significant at the 0.1 level (2-tailed).

The Pearson correlation matrix indicates that both audit firm and audit partner rotation is significantly correlated with the amount of accruals presented in financial statements. It again suggests the aforementioned inverse relation between audit partner rotation and accruals. As expected most control variables are also correlated with the amount of accruals. While multicollinearity is not an issue among the independent variables, it might be present amongst multiple control variables. BANKRUPTCY is heavily correlated with both LEVERAGE (,872) and PERFORMANCE (-,609) and additionally PERFORMANCE is also extensively correlated with CASHROA (,619). This is not inexplicable as a high chance of bankruptcy would likely be accompanied with high amounts of leverage and negative performance. The other correlation is also explainable as one is based on cash flow from operations and the other on earnings, which are both performance indicators. This could pose a threat to the validity of the regression model so I conduct additional VIF-tests to determine if these variables are indeed mutually correlated. VIF-scores above 10 are not acceptable and require that changes be made to the model (Field, 2009). These tests lead to VIF-scores of: 26,0 for LEVERAGE, 39,5 for BANKRUPTCY, 10,6 for PERFORMANCE and 1,7 for CASHROA. I thus establish that multicollinearity is a threat in the current regression model. Elimination of both BANKRUPTCY and CASHROA from the regression has as a result that no VIF-score exceeds 2.0. As multicollinearity is then no longer a threat I proceed to the multivariate analysis without these control variables.

(20)

Page | 20 4.2 Multivariate analysis

The outcomes of the regression model are presented in table 4. The outcome of the first regression is as expected in that all control variables have a significant influence on discretionary accruals. The direction of almost all coefficients are also as anticipated. SIZE, PERFORMANCE and BOOKTOMARKET have an inverse relation (β = -,003; -,010; -,002, p < 0,01) and LOSS and GROWTH lead to higher amounts of discretionary accruals (β = ,004 and ,007, p < 0,01). Remarkable is that more leveraged firms display lower accruals (β = -,004, p < 0,1) and that big four audit firms tolerate higher levels of accruals (β = ,003, p < 0,01). The former could possibly be due to more extensive scrutiny from debt holders. Armstrong et al. (2010) argue that debt contracting may also lead to financial transparency as this would reduce the cost of monitoring for debt holders and hence also the cost of such financing. It is reasonable that companies that rely more heavily on debt contracting try to reduce the costs that come with such financing. The reason that big four audit firms tolerate higher levels of accruals might be due to the high concentration within the German audit market. Francis et al. (2013) find that earnings quality is lower in countries where market concentration among the big four is higher. The German market might be an illustration of this as the leading two of the big four are responsible for 75% of all company-revenues within the sample. Lastly, while not presented in the table, none of the year indicator variables turn out to be significant thus there are no significant year-fixed effects.

Table 4: Regression model

Dependent Variable: |DA|

Control variables Firm rotation Partner rotation Complete model _{firm rotation} Complete model _{partner rotation} Unstanda

rdized B t-value Unstandardized B t-value Unstandardized B t-value Unstandardized B t-value Unstandardized B t-value (Intercept) ,031 ***_-3,899 _,012***_18,738 _,013***_16,712 _,030***_10,580 _,028***_9,891 FirmRot ,003 ** _1,842 _-,001 _-1,161 PartnerRot -,004 ***_-2,802 _-,002 _-1,208 SIZE -,003 ***_-13,876 _-,003***_{-9,661 -,003}***_-8,061 LEVERAGE -,004 * _-,860 _-,004 _{-1,251 -,006}* _-1,886 LOSS ,004 ***_2,923 _,004***_2,723 _,003* _1,930 PERFORMANCE -,010 ***_3,503 _-,013***_{-2,742 -,012}**_-2,367 GROWTH ,007 ***_9,622 _,009***_8,625 _,011***_6,997 BIG4 ,003 ***_2,993 _,003**_2,225 _,005***_3,578 BOOKTOMARKET -,002 ***_-3,417 _{-,001 -1,631 -,001 -,875} R Squared ,088 ,002 ,004 ,093 ,073 F-value 28,443*** _5,124** _7,849*** _15,344*** _10,177***

(21)

Page | 21

The outcome of the univariate regressions of both independent variables do not differ from what the descriptive statistics already suggested. Audit firm rotation leads to a significant increase in discretionary accruals (β = ,003, p < 0,05) and audit partner rotation leads to a significant decrease in discretionary accruals (β = -,004, p < 0,01). This is remarkable as either rotation was expected to lead to higher levels of accruals within the post-rotation window. As elaborated in the descriptive statistics this suggests that an audit partner rotation does not lower competence and impair audit quality. Even more, it suggests an increase in audit quality arguably due to higher independence. This outcome is in support of hypothesis 1 and 3 and does not find any support for hypothesis 2 as no negative effect on discretionary accruals is found after an audit partner rotation.

Lastly, the final regressions show the outcome of the complete regression model. The control variables within this model all exhibit similar coefficients but slightly less significant. The coefficient of PERFORMANCE and GROWTH do both increase notably when compared to the regression of control variables. It is unfortunate that both independent variables are not significant within the complete regression models. Rather remarkable is that the direction of audit firm rotation has become inverse suggesting lower levels of accruals (β = -,001). This outcome is peculiar but not very reliable as it is not significant (p = ,224). For now, it is safe to conclude that the whole model is a good fit of data as F-tests for all regressions are significant. The R-squared corroborates this with an explanatory value of 88% for the control variables, 93% for the complete audit firm model and 73% for the complete audit partner model.

4.3 Sensitivity analysis

To test whether the outcomes of the multivariate analysis are robust I conduct a sensitivity analysis. I test the robustness of discretionary accruals by using a method other than the Modified Jones Model to estimate discretionary accruals. The model I use is the model designed by Defond & Park (2001) which focuses on the reversal of discretionary working capital accruals. This model is a good robustness check as it is not reliant on estimations of industry parameters and, as Defond & Jiambalvo (1994) argue, it focuses on working capital accruals which are more susceptible to earnings manipulation. The mathematical equation of this model is as follows:

* ( ) + ₍₉₎

(22)

Page | 22

In which, is the amount of discretionary working capital accruals, is the net working capital in year t or t – 1 and lastly are the revenues in year t or t – 1. Net working capital is calculated as the difference between current assets minus cash and current liabilities minus short term debt. All variables are scaled by the company‟s total assets at prior year-end to eliminate firm size effects. Outliers were handled in the same way as in the main regression by winsorization at three standard deviations.

Table 5: Robustness regression model

Dependent Variable: DEFOND&PARK

Control variables Firm rotation Partner rotation Complete model _{firm rotation} Complete model _{partner rotation} Unstanda

rdized B t-value Unstandardized B t-value Unstandardized B t-value Unstandardized B t-value Unstandardized B t-value (Intercept) ,132 ***_12,880 _,084***_23,706 _,091***_30,200 _,119***_9,054 _,113***_9,865 FirmRot ,040 ***_6,972 _,024***_4,106 PartnerRot -,016 ***_-3,137 _-,011**_-2,040 SIZE -,011 ***_-10,033 _-,009***_{-6,065 -,007}***_-5,926 LEVERAGE ,043 ***_4,486 _,034***_2,624 _,028**_2,222 LOSS ,013 ** _2,509 _,022***_3,085 _,017***_2,658 PERFORMANCE -,049 ***_-3,447 _-,052**_-2,489 _,008 _,389 GROWTH ,022 ***_6,443 _,016***_3,136 _,030***_4,678 BIG4 -,005 -1,037 ,000 -,050 ,003 ,641 BOOKTOMARKET -,008 ***_-2,688 _,002 _,446 _{,000 -,126} R Squared ,061 ,019 ,004 ,067 ,044 F-value 19,035*** _48,609*** _9,842*** _10,725*** _5,980***

Regressions included indicator variables to determine year-fixed effects.

The regression model based on the estimation of discretionary accruals by Defond & Park (2001) corroborates the results shown in the previous paragraph. Almost all control variables are significantly related with discretionary accruals and coefficients for every variable are also as anticipated. That means that, unlike with the Modified Jones Model, leverage increases discretionary accruals and big four audit firms decreases them. Also, none of the year indicator variables turn out to be significant. When inspecting all coefficients, the Defond & Park model arguably leads to more clear inferences of changes in audit quality. The only peculiarity is that PERFORMANCE, BIG4 and BOOKTOMARKET invert in the complete audit partner model but these coefficients are far from significant (p = ,698; ,522; ,900) and thus not reliable.

(23)

Page | 23

The outcome of the univariate regression model confirms, at the highest significance level, the relationship between audit firm rotation and discretionary accruals (β = ,040, p < 0,01) and audit partner rotation and discretionary accruals (β = -,016, p < 0,01). These coefficients also hold in the multivariate regressions with the complete audit firm model (β = ,024, p < 0,01) and audit partner model (β = -,011, p < 0,05), thus reinforcing the outcome of the main regression comprehensively.

The outcomes of the robustness regression model corroborates the findings of the main regressions. As such I find enough support to accept hypothesis 1 and 3 and argue that audit firm rotation has a (more dominant) negative effect on audit quality than audit partner rotation. I do not find any support for hypothesis 2 and thus reject this hypothesis and argue that audit partner rotation has no short term negative effect on audit quality.

Lastly, I conduct some additional analyses (which are not presented) to determine the influence of other potential confounding factors and further explore the outcomes of this study. First, I repeated the regression with a post-rotation window of two years instead of three as did Carey and Simnet (2006). This does not lead to any significant differences with regard to the coefficients in the complete main and robustness regression models for audit firm and audit partner rotations. Second, I control for the collapse of Arthur Anderson in 2002. The rotations due to this event are not representative for all the other observations in the sample which were all voluntarily of nature. As this thus could have led to noise within the regression model I test the influence of this event. The event itself, based on a three year window post-collapse, however turns out to have had no significant effect on the level of accruals of companies which were audited by Arthur Anderson before the collapse.

Overall, none of the additional analyses lead to outcomes that would cast doubt on the findings of this study. Altogether, there is no direct reason to assume that the findings of the main and sensitivity analyses are not a valid inference of the effect an audit firm and audit partner rotation has on audit quality.

(24)

Page | 24

5. Discussion

This chapter summarises the empirical findings of the study and discusses how these findings should be interpreted according to the researcher and concludes with some limitations of this study and interesting avenues for future research.

5.1 Findings

Audits conducted by audit firms are one of the main instruments by which financial markets attain assurance over financial information. The focus of this study has been to determine to what extent the audit partner individually and the audit firm as an institution are responsible for the assurance provided by the audit firm. More specifically the goal was to gain insight in to what extent aforementioned actors are responsible for audit quality.

Previous research has shown that in the first years of auditing a client audit quality will be impaired. Thus both audit firm and audit partner rotations were expected to have a negative effect on audit quality directly after the event. After an audit firm rotation there is no retention of client knowledge, thus these were expected to lead to a stronger impairment of audit quality.

The findings of this study corroborate to a large extent these expectations on which the hypotheses were also based. I find that discretionary accruals are higher after a rotation of audit firms and that they thus have a short term negative impact on audit quality. More noteworthy is that I find that discretionary accruals are lower after an audit partner rotation takes place. This is remarkable as both types of rotations were expected to lead to higher levels of accruals due to a decrease in client knowledge. As this is not the case, this outcome suggests that after an audit partner rotation audit quality is not impaired. Even more, it suggests an increase in audit quality arguably due to an increase in auditor scepticism.

Both the main regression model and the sensitivity analysis confirm these findings. As such I find enough support to accept hypothesis 1 and hypothesis 3 and argue that audit firm rotation has a (more dominant) short term negative effect on audit quality then audit partner rotation. I do not find any support for hypothesis 2 and thus reject this hypothesis and argue that audit partner rotation has no short term negative effect on audit quality.

(25)

Page | 25 5.2 Implications

The findings of this study suggest that the audit firm (and the institutional knowledge it has) is the main determinant of audit quality. More controversial, it also suggests that a rotation of audit partner is sufficient to preserve auditor independence and thus to maintain audit quality over time. While an audit firm rotation leads to a short term impairment of audit quality, an audit partner rotation does the inverse and increases audit quality.

This finding attests the response of regulators who have chosen to only implement restrictions on audit partner client-tenures. While academic accounting research is mostly ignored by policy makers (Francis, 2004), this study might be a positive exemption and also guide other regulators to follow in the aforementioned approach. As explained, the German setting is characterised as high in terms of agency conflicts and as such is ideal for observing the effectiveness of the auditor within society.

As this study is one of the first to be conducted in a Western institutional setting the findings are still preliminary of nature. Future research could contribute by corroborating or expanding this study‟s findings. More scientific evidence regarding the effect of an audit partner rotation is needed to get a more comprehensive understanding of the impact it has on audit quality.

5.3 Limitations

This study is subject to various limitations. The first limitation is that audit firm rotations within the research period are all voluntarily of nature and as such they might not be representative of audit firm rotations which are mandated by regulators. Second, this study makes use of multiple methods to estimate discretionary accruals and uses this to proxy for audit quality. While these methods have their benefits they also remain constrained in that they do not directly infer audit quality. As such I was required to use several control variables to be able to disentangle audit quality from these other elements. Third, although I was in possession of one sample with both types of rotations, I was not able to validly combine the effect of audit firm and audit partner rotations in one regression model. This due to the requirements which were implemented to keep both variables from interfering among each other. Lastly, this study only gives insight in the short term effect of auditor rotation and does not measure the potential diminishing of auditor independence with passage of time. This would have been interesting as it gives insight in the effect of tenure on auditor-client relationships and at which moment in time a rotation is most preferable.

(26)

Page | 26

References

Armstrong, C.S., Guay, W.R. & Weber, J.P. (2010). The role of information and financial reporting in corporate governance and debt contracting. Journal Of Accounting And Economics, 50(2): 179-234.

Arthur N., Ho S. & Endrawes M. (2017). Impact of Partner Change on Audit Quality: An Analysis of Partner and Firm Specialisation Effects. Australian Accounting Review (2017): 1835-2561.

Ball, R. (2009). Market and Political/Regulatory Perspectives on the Recent Accounting Scandals. Journal Of Accounting Research, 47(2): 277-323.

Barth, M.E., Beaver, W.H. & Landsman, W.R. (1998). Relative valuation roles of equity book value and net income as a function of financial health. Journal of Accounting and Economics, 25: 1-34.

Bartov, E., Gul, F.A. & Tsui, J.S.L. (2001). Discretionary accruals models and audit qualifications. Journal of Accounting and Economics 30: 421-452.

Behn, B., Choi, J. & Kang., T. (2008). Audit quality and properties of analyst earnings forecasts. The Accounting Review 83 (3): 327–349.

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

Boritz, J. E. (1992). Expert and Novice Problem-Solving Behavior in Audit Planning: Discussant. Auditing, 11: 21-32.

Cameran, M., Prencipe, A. & Trombetta, M. (2016). Mandatory Audit Firm Rotation and Audit Quality. European Accounting Review, 25(1): 35-58.

Carcello, J.V., Hermanson, R.H. & Huss, H.F. (1995). Temporal changes in bankruptcy-related reporting. Auditing: A Journal of Practice and Theory, 14 (2): 133-143.

(27)

Page | 27

Carcello, J.V. & Nagy, A.L. (2004). Audit firm tenure and fraudulent financial reporting. Auditing: A Journal of Practice & Theory, 23(2): 55-69.

Carey, P. & Simnett, R. (2006). Audit Partner Tenure and Audit Quality. The Accounting Review, 81(3): 653-676.

Chan K.C., Farrell B. & Healy P. (2014). Audit firm rotation - Concerns and considerations. Journal Of Applied Business Research, 30(1): 227-232.

Chandler, R. & Edwards, J. R. (1996). Recurring issues in auditing: back to the future? Accounting, Auditing And Accountability Journal, 9(2): 4-29.

Chi W., Huang H., Liao Y. & Xie H. (2009). Mandatory audit partner rotation, audit quality, and market perception: Evidence from Taiwan. Contemporary Accounting Research, 26(2): 359-391.

Corbella, S., Florio, C., Gotti, G. & Mastrolia, S.A. (2015). Audit firm rotation, audit fees and audit quality: The experience of Italian public companies. Journal Of International Accounting, Auditing And Taxation, 25: 46-66.

Davis, L.R., Soo, B.S. & Trompeter, G.M. (2009). Auditor tenure and the ability to meet or beat earnings forecasts. Contemporary Accounting Research, 26(2): 517-548.

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

Dechow, P.M. (1994). Accounting earnings and cash flows as measures of firm performance: the role of accounting accruals. Journal of Accounting and Economics, 18 (1): 3-42.

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

Dechow, P.M., Hutton, A.P., Kim, J.H. & Sloan, R.G. (2012). Detecting Earnings Management: A New Approach. Journal of Accounting Research, 50(2): 275-334.