THE HOMOGENIZIATION OF THE LONG-FORM AUDITOR’S REPORT: EVIDENCE FROM EUROPE

(1)

THE HOMOGENIZIATION OF THE LONG-FORM

AUDITOR’S REPORT: EVIDENCE FROM EUROPE

Thesis MSc Accountancy and MSc Controlling

University of Groningen

Gineke Spoelstra

S2936127

Oosterstraat 12-1

9711 NT Groningen

06-20488691

m.g.spoelstra@student.rug.nl

16

th

_{of January 2020}

Prof. Dr. D.A. de Waard

KEYWORDS: LONG-FORM AUDITOR’S REPORT, KEY AUDIT MATTERS,

LEGITIMACY THEORY, INSTITUTIONAL ISOMORPHISM,

(2)

THE HOMOGENIZIATION OF THE LONG-FORM

AUDITOR’S REPORT: EVIDENCE FROM EUROPE

ABSTRACT: Regulators are requiring the issuing of a long-form auditor’s report in order to provide more

information about the performed audit. The long-form auditor’s report requires auditors to report on, among other aspects, Key Audit Matters (KAMs). This study attempts to investigate whether the long-form auditor’s report is growing towards a standardized report in which other jurisdictions learn from lessons of the early adopters and whether the process of isomorphism influences the issuing of long-form auditor’s reports for financial institutions in Europe. The sample for this study consists of Dutch, UK, German and Italian listed financial institutions. The data is collected from 472 long-form audit opinions which have been issued from 2013 up to and including 2018. Overall, the results of this thesis indicate that the long-form auditor’s report is partly evolving towards a homogenous report for listed companies in the financial sector in Europe. Firstly, auditors are influenced by coercive isomorphism, the legal system, which leads to more diverse audit opinions. Secondly, auditors are influenced by mimetic isomorphism, the financial year in which the report is issued, which leads to less diverse audit opinions when time progresses. Thirdly, auditors are influenced by normative isomorphism, the issuing audit firm, which leads to less diverse audit opinions issued by auditors of the same audit firm. However, late adopting countries undergo their own process in implementing the long-form audit report instead of learn from lessons from early adopting countries. Results of this study should be of interest for investors, auditors and regulators. The positive influence of coercive isomorphism on the long-form auditor’s report and the implementing of their own process by late adopting countries confirms that new regulation has the intended effect of providing more useful information to investors. On the other side, the results of mimetic and normative isomorphism confirm the concern that issuing of the long-form auditor’s report becomes a standard ticking the box procedure.

Keywords: long-form auditor’s report, key audit matters, legitimacy theory, institutional isomorphism,

(3)

TABLE OF CONTENT

1. Introduction ... 4

2. Theoretical framework ... 8

2.1 Legitimacy theory ... 8

2.2 Institutional theory ... 9

2.3 Institutional isomorphism……….. 10

2.3.1 Coercive isomorphism... 11

2.3.2 Mimetic isomorphism ... 12

2.3.3 Normative isomorphism ... 13

3. Research methodology ... 15

3.1 Sample and data collection ... 15

3.2 Variables ... 16

3.2.1 Dependent variables ... 16

3.2.2 Independent variables ... 17

3.2.3 Control variables ... 18

3.3 Statistical model ... 19

4. Results……… ... 21

4.1 Dataset adjustments ... 21

4.2 Descriptive statistics ... 21

4.3 Normality test... 23

4.4 Correlation analysis ... 24

4.5 Independent samples t-test ... 27

4.6 Regression analyses ... 29

4.6.1 Control variables ... 29

4.6.2 Legal system ... 29

4.6.3 Financial year ... 30

4.6.4 Issuing big 4 audit firm ... 31

4.6.5 Complete model ... 32

5. Conclusion & Discussion ... 34

5.1 Conclusion and discussion ... 34

5.2 Limitations and recommendations for future research ... 36

5.3 Overall conclusion ... 37

6. References ... 38

(4)

I. INTRODUCTION

ecently Accountant.nl published an article in which the relevance of the long-form auditor’s report is discussed (Accountant.nl, 2019). Around 1900 the method of reporting audit findings was also a topic for much discussion both among leading practitioners and the business press. However a signature of a reputable accountant was sufficient to attest to the reliability of the accounts (Chandler and Edwards, 1996). The deepening financial crisis in 2008 triggered an increase in criticism about the role and value of external audits (Sikka, 2009). During that period a large number of financial enterprises either collapsed or have sought state support within a short period after receiving an unqualified audit opinion. Academic research in the last decade has repeatedly emphasized that there may be a need for change, motivating audit report reforms (e.g. Gray et al. 2011; Gold, Gronewold and Pott, 2012; Carcello, 2012). In 2012, the chairman of the IAASB stressed that ‘users of audited financial statements are calling for more pertinent information for their decision making in today’s global business environment with increasingly complex financial reporting requirements’ (IAASB, 2012). The Financial Reporting Council (FRC) responded by the introduction of new requirements for the auditor’s report in the UK. These requirements, resulting in the so-called long-form

auditor’s report, went in effect for periods commencing on or after October 2012. In addition to the

(un)qualified opinion, the auditor’s report must describe the risks of material misstatements that had the greatest effect on the audit, the application of materiality and the scope of the audit (FRC, 2015). The risks of material misstatement are typically reported as Key Audit Matters (KAMs). KAMs are intended to provide transparency as to how certain issues affected the auditor’s approach and to highlight for users the matters that require the most attention (IAASB, 2016). In 2014, shortly after the FRC, the Royal Netherlands Institute of Chartered Accountants (NBA) issued similar requirements (NBA, 2014). In January 2015, after the implementation of requirements in the UK and the Netherlands, the International Auditing and Assurance Standards Board (IAASB) released a revised standard (ISA 700 – revised) which is compulsory for listed entities and comes into effect for auditors of financial statements for periods ending on or after 15 December 2016 (IFAC, 2017).

The auditor’s report has increasingly been a subject of research. This research frequently focuses on the audit expectation gap between how the auditor perceives his role and what the public expects from the auditor (e.g. Best et al., 2001; Lin and Chen, 2004; Porter, 1993). More recent studies investigate whether a long-form-auditor’s report results in a smaller audit expectation gap. Results show that a more expanded auditor’s report has little impact on shareholders perceptions (Chong and Pflugrath, 2008) or found no significant effect of the long-form auditor’s report on the expectation gap (Boolaky and Quick, 2016; Gold et al. 2012). Results of prior studies on the informative and communicative value of the long-form auditor’s report and the KAMs are mixed. Lennox, Schmidt and Thompson (2017) suggest that the new disclosures lack of incremental information content because users were

(5)

already informed about the majority of the risks before the risks were disclosed in the auditor’s report. The archival study of Gutierrez et al. (2018) regarding investor’s reaction on the additional information in the auditor’s report aligns with the findings of Lennox et al. (2017) concluding that KAM disclosure does not influence investor behavior. Köhler et al. (2016a) found that investment professionals’ assessments of the economic situation of a company are influenced by variations in the KAM disclosures. Professional investors perceive the disclosure of KAMs as positive while nonprofessional investors perceive it as non-communicative as they may have difficulties to process the new information. Research of Christensen, Glover and Wolfe (2014) shows that nonprofessional investors

do change their decisions when the audit report contains KAMs. Additionally, Sirois, Bedard and Bera

(2018) found, using an eye-tracking study, that KAMs have an attention-directing effect. Participants pay relatively more attention to KAM-related information in the financial statements. However, when exposed to an auditor’s report with several KAMs, participants devote less attention to the remaining of the financial statements. Other recent findings demonstrate that KAMs sections may serve as a beneficial mechanism for reducing managerial earnings management activities only when the informational precision in risk disclosure is high (Klueber, Gold and Pott, 2018). Research of Porumb et al. (2018) shows that lenders perceive borrowers with more KAMs to be riskier, which translates into less favorable loan contracting terms. Therefore, according to the research of Porumb et al. (2018) the expanded audit report disclosures contain relevant information for lenders. In sum, due to the mixed results in prior studies on the value of KAMs it can be concluded that reporting KAMs does not necessarily add more informative and communicative value to the auditor’s report.

Before the implementation of ISA 700 the UK’s FRC and the Dutch NBA introduced similar requirements in 2013 and 2014 respectively. These requirements were seen as very much in the vanguard of the new reporting model, providing early adopter lessons for other jurisdictions (Smith, 2018). The discussion paper of Köhler, Quick and Willekens, (2016b) shows the need for studies on the risk that the long-form auditor’s report becomes a standard ticking the box procedure and thereby falls back towards the boilerplate (pass/fail) model. Regarding a redesigned auditor’s report, financial statement users suggested that new material added to the auditor’s report (long-form auditor’s report) does not matter if that added material is boilerplate in nature (Gray et. al., 2011). Moreover, feedback on the long-form auditor’s report has indicated that homogeneity in auditor reports is not helpful (ICAEW, 2017). According the IAEW (2017) every company and every audit is different and so each audit report should also be different. The reports should be insightful, dynamic, relevant and a useful source of information. Therefore the long-form auditor’s report need to avoid boilerplate content as much as possible.

(6)

To what extent are the KAM disclosures in the long-form auditor’s report evolving towards a homogenous report for listed companies in the financial sector in Europe?

The focus of this study is on long-form auditor’s reports in the financial sector. The long-form auditor’s report is a result of the financial crisis which has been caused by companies in the financial sector (Sikka, 2009). Moreover, financial institutions are in a highly regulated and supervised industry that influences the number and type of disclosed KAMs (Pinto and Morais, 2019). More specially, their findings show that more complex business segments and more precise accounting standards lead to the disclosure of a higher number of KAMs. According to Dunn and Mayhew (2004) in regulated industries the high level of required disclosure and monitoring can lead auditors to have lower incentives to disclose KAMs. Results of Pinto and Morais (2019) and Dunn and Mayhew (2004) are not consistent, however they show that the kind of industry influences the number and type of disclosed KAMs. This research paper contributes to the existing literature regarding the long-form auditor’s report. The long-form auditor’s report is mandatory for six years in the UK, five years in the Netherlands and two years in other European countries. Most studies on long-form auditor’s report are focused on Anglo- Saxon countries such as the US, UK and Australia (Gutierrez et al., 2018; Lennox, et al., 2017; Gay, Schelluch and Reid, 1997) and Asian countries such as Singapore and China (Best, Bucky and Tan, 2001; Lin and Chen, 2004). Little research is done on the differences across multiple jurisdictions (Lennox, et al. 2017, Gold and Heilmann, 2018). Therefore this thesis focuses on the homogenization between different countries. Moreover, the majority of prior research studied the long-form auditor’s report in a qualitative way or studies a one-year period (Gold et al., 2012; Köhler., 2016a; Best et al., 2001), this thesis investigates how the long-form auditor’s report is evolving during the years and if this process in leaded by the early adopters UK and NL. Aforementioned studies provide insights on the effects of the long-form auditor’s report. However there is a need for more research (Köhler, Quick and Willekens, 2016b).

Legitimacy theory (Dowling and Pfeffer, 1975) and the theory of institutional isomorphism (DiMaggio and Powell, 1983) are theories that might explain how the reporting on KAMs evolves over time. The mechanism of isomorphism explains how auditors increase their legitimacy by adapting a long-form auditor’s report towards a template which is perceived as legitimate within the auditing working field (Lee and Pennings, 2002; DiMaggio and Powell, 1983). This study tries to contribute to existing literature by analyzing how long-form auditor’s reports evolve and how this process is influenced by coercive, mimetic and normative isomorphic pressures. Additionally, this thesis should be relevant for financial statement users, auditors and regulators. An understanding of how the long-form auditor’s reports evolve can give insight in the informativeness of the report for investors, since they expressed the need for more company-specific information in the audit report (Lennox et al., 2017). Moreover,

(7)

describe KAMs in a client-specific manner since existing literature provided that long-form auditor’s reports should be a useful source of information. (i.e. ICAEW, 2017; Lennox et al., 2017; Köhler et al., 2016b; Gray et. al., 2011). Lastly, this thesis should be of interest for regulators since it gives these groups better insight in the effectiveness of the regulations. The expanding of the audit report is a recent development for audit firms. When research shows that added material is boilerplate in nature, changes in regulations could be needed.

The remainder of this thesis is organized as follows. Section 2 presents the theoretical background and develops hypotheses based on the literature discussed. Section 3 describes the research methodology for testing the hypotheses, while section 4 presents the results. The final section concludes with a discussion of the major findings and the study’s limitations and ideas for future research.

(8)

2. THEORETICAL FRAMEWORK

This section describes the theoretical framework that forms the basis for this study. The first paragraph discusses the legitimacy theory. The second paragraph describes the institutional theory. The third paragraph describes the three mechanisms through which institutional isomorphism occurs. Building upon the theory, five hypotheses will be developed.

2.1 Legitimacy theory

Legitimacy theory might explain how the reporting on KAMs will evolve over time. Legitimacy theory states that organizations should continually seek to ensure their operations fall within the bounds of respective communities and to be perceived as ‘legitimate’ by various stakeholder groups (Deegan and Samkin, 2009). Expectations of society at large have to be fulfilled by the organization (Dowling and Pfeffer, 1975), not merely the owners’ or investors’ requirements as in for example the agency theory (Jensen and Meckling, 1976). Deegan (2006) argues there is a ‘social contract’ between the society and the organization with respect to the state of organizational legitimacy. The reputation of being legitimate is crucial for the survival of organizations (An, Davey and Eggleton, 2011). This creates a challenge for organizations to stay legitimate with respect to changing requirements of society (Deegan, 2006). In a working field legitimacy can be maintained by adapting operations towards a template that is perceived as legitimate (DiMaggio and Powell, 1983). The introduction of the long-form auditor’s report can be seen as a template which can enhance the legitimacy of the auditor. In this template auditors disclose KAMs that are perceived as legitimate by investors. However, these are not necessary company- specific. If this is the case, the purposes of increasing the usefulness of the audit opinion for investors and providing company-specific information, has not been fulfilled (Lennox et al., 2017). Misalignment of expectations about and within the auditing system, threatens the legitimacy (Power, 2003).

In the auditors working field, more transparency leads to more trust and therefore enhanced legitimacy (Heald, 2017). The main goal of the long-form auditor’s report is providing greater transparency on how the audit is conducted and what risks of material misstatements were identified (FRC, 2015). The long-form auditor’s report is therefore expected to enhance legitimacy. However, following the legitimacy theory there could be expected that it is not needed to require a mandatory long-form auditor’s report. Especially during this century, where the legitimacy of auditing is under fire as perhaps never before (Power, 2000), you would expect auditors to enrich their reporting to increase their legitimacy when society is calling for more transparent information about the auditor’s process. Because the effort of the profession alone to respond to the various corporate failures are inadequate, independent regulators put monitors in place to prevent following failures (Maroun and Solomon, 2014). The SOX, with mandatory audit quality provisions, serves as example.

(9)

Studies about legitimacy are divided in strategic and institutional legitimacy (Suchman, 1995). Strategic legitimacy assumes that changes are induced by an managerial perspective. This study is about institutional legitimacy since the long- form auditor’s report is a result of sector-wide changes (FRC). Institutional legitimacy is built on institutional theory which is discussed in the next paragraph.

2.2 Institutional theory

Institutional theory has succeeded in becoming the dominant theory in studying macro-organizational phenomena (Suddaby, 2010). An institution is defined as a cognitive, normative or regulatory structure or activity that provides stability and meaning for social behavior (Suchman, 1995). According to (Selznick, 1996) institutional theory traces the emergence of distinctive forms, processes, strategies, outlooks and competences as they emerge from patterns of organizational interaction and adaption. Institutionalism can be divided in ‘old’ institutionalism and ‘new’ institutionalism. Old institutionalism is focused on the individual organization, whereas new institutionalism is primarily related to organizations-in sectors (DiMaggio and Powell, 1991).To date, three institutional approaches have probably had most influence on accounting research: old institutional economics (OIE), new institutional economics (NIE) and new institutional sociology (NIS) (Moll, Burns and Major, 2006). Drawing from OIE, researchers would generally consider why and how particular behaviors or structures emerge, sustain or change over time rather than merely what structures exist at a given point in time (Moll et al., 2006). The main focus of NIE research is on making optimal choices by organizational actors with a view to maximizing the economic outcomes of their activities (Spicer, 1988) and explaining the existence of some institutions and the non-existence of other (Moll et al., 2006). On the other hand, NIS adopts a broader approach for focusing on macro (external) and micro (internal) organizational contexts (DiMaggio and Powell, 1991; Greenwood and Hinings, 1996). Research in the new institutional sociology has focused on the process of institutional isomorphism (Beckert, 2010).

Institutional theory is not usually regarded as a theory of organizational change, but as an explanation of similarity (isomorphism) and stability of organizational arrangements in a given population or field of organizations (Greenwood and Hinings, 1996). Institutional theory examines organizational forms and explains the reasons for having homogeneous characteristics in organizations within a same organizational field. By organizational field, DiMaggio and Powell (1983) mean those organizations that, in the aggregate, constitute a recognized area of institutional life: key suppliers, resource and product consumers, regulatory agencies, and other organizations that produce similar services or products. In the competitive nature of the auditing working field organizations that do not adapt towards the legitimate template are winnowed out. This process will lead to the homogenization of the working field (Lee and Pennings, 2002; Dimaggio and Powell, 1983). The concept that best captures the process of homogenization of an organizational field is called institutional isomorphism. This concept is described

(10)

2.3 Institutional isomorphism

As mentioned before, institutional isomorphism is the concept that best captures the process of homogenization (DiMaggio and Powell, 1983). Institutional isomorphism can be described as the constraining process that forces one unit in a population to resemble other units that face the same set of environmental conditions (Hawley, 1968). Similarity can make it easier for organizations to transact with other organizations, to attract career-minded staff, and to be acknowledged as legitimate and reputable (DiMaggio and Powell, 1983). Following the theory of isomorphism, three isomorphic processes are described: coercive, normative and mimetic (DiMaggio and Powell, 1983). All three isomorphic processes lead organizations to adopt similar structures and management practices in their field, irrespective of their actual usefulness or efficiency (Carpenter and Feroz, 2001). The expectation in this thesis is that long-form auditor’s reports are evolving into a homogeneous and standardized report, due to the processes of institutional isomorphism. This expectation challenges the expectation one could state following the legitimacy theory, that the auditor would be motivated to tailor KAM disclosures towards the desire of investors to gain legitimacy. However, over decades research showed that organizational isomorphism increases organizational legitimacy (Meyer and Rowan, 1977; DiMaggio and Powell, 1983; Deephouse, 1996; Maingot; 2006). Organizations that conform to other organizations are recognized by regulators and the general public as being more legitimate than those that deviate from normal behavior (Deephouse, 1996). As mentioned in the introduction section, the requirements in the UK and the Netherlands were seen as very much in the vanguard of the new reporting model, providing early adopter lessons for other jurisdictions. It is expected that, due to isomorphic forces, the long-form auditor’s report is growing towards a standardized report in which other jurisdictions learn from lessons from the early adopters. This thesis focuses on the diversity of audit opinions. Therefore, it is expected that long-form audit opinions issued in the years after the mandating of the long-form auditor’s report by the FRC and the NBA are more diverse than long-form audit opinions issued in the years after the implementation of the requirements by the IAASB for listed companies in other European countries. The first hypothesis expects that auditors from ‘late’ adopting countries will be reporting increasingly on the same KAMs, which leads to less diverse long-form audit reports.

Hypothesis 1: Long-form audit opinions issued by late adopters are less diverse than long-form audit opinions issued by early adopters

In the ‘ iron cage’ paper of DiMaggio and Powell (1983) introduced a framework on institutional isomorphism, which is created to obtain more understanding of the institutional theory. As mentioned above, three isomorphic processes are described: scoercive, mimetic and normative. The three processes can overlap and intermingle, but they tend to derive from different conditions (Frumkin and Galakiewics, 2004). Bearing in mind that the three types of isomorphic forces generally operate simultaneously, we discuss each

(11)

2.3.1 Coercive isomorphism

Coercive isomorphism results from both formal and informal pressures exerted on organizations by other organizations upon which they are dependent and by cultural expectations in the society within which organizations function (DiMaggio and Powell, 1983). In case of the homogenization of the long- form auditor’s report, formal pressures can derive from the fact that the long-form auditor’s report has been mandated (IFAC, 2017). On the other side, informal pressures can stem from investors that call for a more informative audit report (Lennox, 2017). Related to this mechanism of isomorphism, several studies highlighted the importance of the legal environment in the disclosure and auditing of information (Zhou, Simmett and Green, 2013; Francis et al., 2011; Martinez-Ferrero and Garcia- Sanchez, 2017). These studies compare common versus code law countries and the strength of the enforcement mechanism, which La Porta et al. (1997) used as a proxy for the strength of the legal system. Investors and creditors often enjoy greater protection of their interests in countries with a common law system in which a company is considered as an instrument to create and maximize shareholder value. This can be clarified by the fact that countries based on the common law system are more shareholder-oriented, whereas countries based on the code law system more stakeholder-oriented (La Porta et al. 1998). The framework provided by La Porta et al. (1997, 1998) is used by Francis, Khaurana and Pereira (2003) to argue that increased agency conflicts, associated with more widespread ownership and greater external financing, result in a higher demand for transparent disclosure. Stakeholders would have direct access to information in code law countries where ownership is more concentrated, and consequently the demand for disclosure should be lower (Ball, Kothari and Robbin, 2000). In contrast, stakeholders in common law countries demand for more disclosure to reduce information asymmetry problems. In countries with a strong protection of investors (common law countries), higher requirements for financial disclosure should also imply a stronger demand for a credible control of accounting data (Broyce and Weil, 2008; DeFond, 1992). Choi and Wong (2002) investigated the relationship between national legal environments and auditor choice. They found that the demand for a high quality auditor is likely to be higher in common law countries. These results are consistent with Francis et al. (2003), who have shown that Big Five auditors, which are seen as high quality auditors, have larger market shares in common law countries than in code law countries.

As stated, in common law countries there is a higher demand for disclosure and a stronger demand for a credible control of accounting data. From this reasoning, it is expected that auditors are more pressured in common law countries. The general assumption behind coercive isomorphism is that formal and informal pressures lead to homogenization of the long-form auditor’s report. However, pressures are exerted to create more diverse long-form auditor’s reports. Therefore, the following hypothesis state

(12)

that the long-form auditor’s report is more diverse in common law countries because pressures to engage in coercive isomorphism are stronger in these countries.

Hypothesis 2a: Long-form audit opinions issued by auditors in common law countries are subject to stronger pressures of coercive isomorphism, resulting in more diverse audit opinions

In contrast, other authors found, using the same framework provided by La Porta et al. (1997, 1998), that companies adopt corporate governance practices as a mechanism for adapting to poor legal environments (Durney and Kim, 2005). External audits play a more important role in mitigating agency problems in weak legal environment countries than in strong legal environment countries (Choi and Wong, 2004). Audit facilities can serve as an alternative device to improve the information environment in code law countries (Choi et al., 2005). Moreover, Broye and Weill (2008) stated that poor protection rights may trigger off a heavy demand for credible auditors as a substitute for corporate governance mechanisms to limit managerial opportunism. Zhou et al. (2013) support this substitute effect, showing that there is a greater demand to enhance the credibility and transparency of information issued in countries where the legal enforcement system is weaker. This means that the demand for a credible control of accounting data and transparency is higher in code law countries. Therefore, the alternative hypothesis expects that the long-form auditor’s report is more diverse in code law countries because pressures to engage in coercive isomorphism are stronger in these countries.

Hypothesis 2b: Long-form audit opinions issued by auditors in code law countries are subject to stronger pressures of coercive isomorphism, resulting in more diverse audit opinions 2.3.2 Mimetic isomorphism

Among the three processes of institutional isomorphism one process, mimetic, has received disproportionate attention. (Mizruchi and Fein, 1999). The discussion of mimetic isomorphism is consistent with the dominantly held view among leading organizational researches. According to DiMaggio and Powell (1983) mimetic isomorphism is a response to uncertainty. In situations in which a clear course of action is unavailable, organizational leaders may decide that the best response is to mimic a peer that they perceive to be successful. They defined this imitating response to uncertainty as ‘modeling’. The introduction of the long-form auditor’s report is a change in the environment of audit firms which leads to uncertainty. Mimetic isomorphism can, alongside to uncertainty, be driven by what March (1981) called ‘obligatory action’. According to his model, once enough social actors do things a certain way, that particular course of action becomes institutionalized, and thereafter, other social actors will undertake that course of action without thinking. Haveman (1993) assumed that organizations imitate organizations within their population (industry) as the actions of these organizations tend to me more salient than the actions of organizations in other populations (industries). Organizations in their

(13)

population, which are imitated, are similar in terms of structure, strategy, resources and constraints (Haveman, 1993).

Research, in the context of corporate environmental reporting, showed that the mimetic process in enhanced in highly concentrated industries and weakened when a firm is subject to public media exposure (Aerts, Cormier and Magnan, 2006). Wang, Yuan and Wu (2017) argued that an audit firm will imitate peers in the same identity group more than peers in different identity groups when determining its audit quality. Moreover, aforementioned study of Havenman (1993), showed that organizations attend to the actions of successful organizations and will imitate their behavior. Slack and Hinings (1994) examined the process of isomorphic change by looking at change holistically. Using a population of sport organizations, they found that over time there is an increase in the level of homogeneity of these organizations. The fact that organizations become less diverse when time progresses can be linked to the long-form audit opinion. At the moment of introduction of the long-for auditors report in 2013 by the FRC, auditors could not engage in mimetic isomorphism because examples of long-form audit opinions were not available. However, as time progresses more long-form audit opinion are issued which makes is possible for auditors to engage in mimetic isomorphism. Castrogiovanni (1996) supports this by stating that mature industries provide the opportunity for mimetic isomorphism because new entrants have an opportunity to observe and learn from the established competitors. Using a similar argument, mimetic pressures are said to be more salient in mature institutional fields (Honig and Karlsson, 2004). Because auditors engage more in mimetic isomorphism when time progresses, the third hypothesis expects that more recent long-form audit opinions are less diverse.

Hypothesis 3: More recent long-form audit opinions are subject to stronger pressures of mimetic isomorphism, resulting in less diverse audit opinions

2.3.3 Normative isomorphism

Normative isomorphism is the third mechanism trough which institutional isomorphism occurs. In the original paper of Dimaggio and Powell (1983) the discussion of normative isomorphism receives the greatest attention (Mizruchi and Fein, 1999). DiMaggio and Powell (1983) viewed normative isomorphism as a result of professionalization. Formal education and the cognitive base produced by universities on the one hand and growth and elaboration of professional networks on the other hand are two important aspects of professionalization (DiMaggio and Powell, 1983). This could lead to a taken- for-granted model when issuing the long-form auditor’s report.

Big 4 audit firms have a professional network which can lead to such a model. Big 4 audit firms describe themselves as ‘‘one firm worldwide’’, providing a consistently high-quality service (Arnold, Bernardi and Neidemeyer, 2009). Suddaby et al. (2007) posit the growth of transnational service firms, which

(14)

constitute a new organizational form that is in sharp contrast to the regional and national companies that have dominated for so long. Big 4 audit firms are organized in highly integrated networks cross- nationally (Lenz and James, 2007). The global central governing bodies of the Big 4 audit firms seek to organize and integrate the member firms by imposing common standards and practices on them to operate as a ‘‘seamless web’’, even though the firms remain independently owned and managed on a national basis (Malhorta, Morris and Hinings, 2006). The member firms are required to comply with these standards and practices to retain their membership in the network (Zimmermann and Volckmer, 2012). Consequently, the reporting practices is likely to be coordinated by the particular central governing body of the Big 4 audit firm nowadays (Girdhar and Jeppersen, 2018). The main limitation for cross-border cooperation between the firms remains national regulation. According to (Zimmermann and Volckmer, 2012) some field already have been harmonized at European level regarding the harmonization of the accounting relevant regulation. Between European countries and the US, the differences are much larger (Fosbre, Kraft and Fosbre, 2009). However, this thesis focuses on European countries. The expectation is that the long-form auditor’s report of the same audit firm is less diverse than the long-form auditor’s report of other audit firms due to the fact that member firms are required to comply with the standards and practices to retain their membership in the specific Big 4 network. Because auditors of the same audit firm engage in more normative isomorphism, the last hypothesis expects that long-form audit opinions issued by auditors of the same audit firm are less diverse.

Hypothesis 4: Long-form audit opinions issued by auditors of the same audit firm are subject to stronger pressures of normative isomorphism, resulting in less diverse audit opinions

(15)

3. RESEARCH METHODOLOGY

This chapter described the methodology used to perform this research. In the first paragraph the sample and data collection will be explained. The second paragraph describes the dependent, independent and control variables used in this study and how these variables have been measured. The third paragraph gives an oversight of the variables and sets out the statistical model.

3.1 Sample and data collection

This study uses data of long-form auditor’s reports for British, Dutch, German and Italian stock-listed firms. Data for British stock-listed firms is collected for financial years 2013 up and including 2018. For Dutch stock-listed firms the data is collected for financial years 2014 up and including 2018. However, 23 Dutch firms voluntarily adopted the long-form auditor’s report in 2013. This data is included in the sample. Data for German and Italian stock-listed firms is collected for financial years 2016, 2017 and 2018. The data collection for the KAMs is part of a larger research and has therefore been hand collected by multiple students.

The UK sample in the database consists of the FTSE100 firms and the first 150 of the FTSE250 firms, which are in total 250 firms. The Dutch sample consists of 73 companies that are listed on the AEX, AMX and AScX. The German and Italian sample consists respectively of 30 companies listed on the DAX and 40 companies listed on the FTSE MIB. This results in a total sample size of 384 companies with an extended audit opinion, 274 over six years (2013-2017), 50 over five years (2014-2018) and 60 for three years (2016-2018). Only companies active in the financial sector are selected in the sample since the focus of this study is on long-form auditor’s reports in the financial sector

.

The Standard Industrial Classification (SIC) code in the Orbis database in used to distinguish companies active in the financial sector. Companies with a SIC code between 6000 and 6700 are selected, which includes finance, insurance and real estate (FIRE) companies. Companies with a long-form auditor’s report that does not contain KAMs and companies which received a long-form auditor’s report form a non-Big 4 auditor are excluded. After excluding, the entire sample consists 472 firm-year observations, which is presented in table 1. The companies included in the sample are presented in appendix 1.

(16)

Table 1: Sample distribution

Initial number of cases 2074

Number of cases in non-financial sector 1489

Number of cases in financial sector 585

Cases with missing data on KAMs 95

Cases with non Big4 auditor or two auditors 18

Final sample 472

3.2 Variables

This section describes and shows how to measure respectively the dependent, independent and control variables used in this study.

3.2.1 Dependent variable

The goal of this study is to determine whether the long-form auditor’s report is growing towards a standardized report in which other jurisdictions learn from lessons of the early adopting countries and whether the process of isomorphism influences the issuing of long-form auditor’s reports. The homogenization of the long-form auditor’s report the mechanisms of isomorphism could influence the diversity of long- form audit opinions. Therefore, the dependent variable of this study is the diversity of the long-form audit opinion (DIV). The diversity of the long-form auditor’s opinion is measured by the variety of the KAMs disclosed. Variety is diversity within an unit based on the fact that they differ from each other on categorical characteristics (Harrison and Klein, 2007). Variety can be used to measure diversity for a qualitative variable (Argesti and Argesti, 1978), which is in this case diversity of the audit opinion. To measure variety, the KAMs are coded into nineteen different categories. The categorization of the KAMs is presented in appendix 2. Harrison and Klein (2007) showed two operationalizations of variety, namely Blau’s index of heterogeneity and Teachman’s entropy index. Variety is commonly measured by both indexes, which are linearly correlated (McDonald and Dimmick, 2003). The only advantage one of these two operationalizations of variety has over the other is that Blau’s index occupies a somewhat tidier range of 0 to 1 (Harrison and Klein). Therefore, Blau’s index of heterogeneity is used to measure the diversity of long-form audit opinions. The formula to calculated the Blau index is shown below:

𝐵𝐵𝐵𝐵𝐵𝐵𝑢𝑢′_{𝑠𝑠 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑜𝑜𝑜𝑜 ℎ𝑖𝑖𝑒𝑒𝑖𝑖𝑒𝑒𝑜𝑜𝑒𝑒𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑒𝑒𝑒𝑒: 1 − ��𝑝𝑝} 𝑘𝑘2�

(17)

In the Blau’s index of heterogeneity p corresponds to the proportion of group members in kth category and the variety can take k possible categories. For a long form audit opinion which has four KAMs: one focuses on taxation, two on revenues and one on non-current assets, the Blau index can be calculated as follows: (1-(1/4)² + (2/4)² + (1/4)²) = 0.56. The index reflects the chance that two randomly selected key audit matters belong to different categories. In this case the chance is 56%. If Blau’s index of heterogeneity equals zero (it’s minimum value), all reported key audit matters are in the same category and there is no variety. In contrast, the higher the value, the more dispersed key audit matters over the categories. The maximum value is achieved when key audit matters are spread equally over all possible categories (Harrison and Klein, 2007). The maximum is limited by the number of categories, k, which are in these case nineteen. However, not every long-form audit opinion has the same amount of KAMs. Audit opinions with more KAMs, a higher k, have a higher maximum of diversity. This issue can be mitigated by standardizing Blau’s index by dividing it by its maximum, yielding the index of qualitative variation (IQV). The IQV controls for the number of categories in an audit opinion (Argesti and Argesti, 1978). The following formula is used to calculate the IQV:

𝐼𝐼𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑜𝑜𝑜𝑜 𝑞𝑞𝑢𝑢𝐵𝐵𝐵𝐵𝑖𝑖𝑒𝑒𝐵𝐵𝑒𝑒𝑖𝑖𝑞𝑞𝑖𝑖 𝑞𝑞𝐵𝐵𝑒𝑒𝑖𝑖𝐵𝐵𝑒𝑒𝑖𝑖𝑜𝑜𝑖𝑖 (𝐼𝐼𝐼𝐼𝐼𝐼): 1 − ��𝑝𝑝𝑘𝑘2� /(1 −1_𝑘𝑘)

The theoretical maximum of the example is computed as follows: 1- (1/4) = 0.75. Therefore the IQV is 0.56/0.75 = 0.75. Researchers have stated that comparisons between variables with a different number of categories still makes sense, as long as larger number of categories contributes to greater diversity (Argesti and Argesti, 1978; Solanas et al., 2012). An audit opinion with more KAMs is thus more diverse. In contrast, the IQV gives good insights in the diversity of the audit opinion since it measures only the variety of the audit opinion in terms of categories of KAMs disclosed. Therefore, the analyses are performed both using the Blau’s index of heterogeneity and the IQV.

Summarized, the diversity of long-form audit opinions is operationalized in two ways. The first way is by the unstandardized diversity of the long-form audit opinion, Blau’s index of heterogeneity (DIV) and the second by the standardized diversity of the long-form audit opinion, the IQV (DIV_STD).

3.2.2 Independent variables

To test the hypotheses, this study used four different independent variables. These are the adopting period (ADOP), the legal system (common or code law) in which the audit opinion is formed (LAW) the financial year on which the audit option is formed (FY) and which Big 4 audit firm issued the long- form audit option (BIG4).

The first variable is the adopting period (ADOP). The long form audit report is issued more early in the UK and the Netherlands. Germany and Italy are late adopters. In this thesis we focus om the big economies. The G20 consists of four European countries, respectively France, Germany, Italy and the

(18)

UK. French auditors had to disclose justifications of assessments (JOAs) in expanded audit reports, since 2003. These disclosures include a summary of auditor’s assessments and performed procedures and are therefore comparable to KAM disclosures (Bedard, Gonthier-Besacier and Schatt, 2019). Because of the deviating maturation of the long-form auditor’s report in France, this country is not taken into account.

The second independent variable is the legal system (LAW) in which the long-form audit opinion is issued. Following Martinez-Ferrero and Garcia-Sanchez (2017), legal system is used as determinant of coercive isomorphism. As mentioned in the theoretical section, national legal systems are classified in code and common law systems (La-Porta et al., 1997). Consistent with Francis and Wang (2008) and Abughazaleh, Connell and Princen (2015) a dummy variable is created. The variable is coded as zero for the common law country (UK) and one for the code law countries (the Netherlands, Germany and Italy).

The third independent variable is the year (FY) on which the long-form auditor’s report is formed. This variable is measured by the end of the financial year of the audited entity. Similarly, this is the case for entities with a split financial year. This means that when a long-form audit opinion is issued for an entity with a split financial year of 2016-2017, it is measured as an long-form audit opinion that is issued in 2017.

The fourth and last independent variable is the audit firm (BIG4) that issued the long-form auditor’s report. In the sample selection, audit opinions issued by non-Big 4 companies are excluded. For the Big 4 audit firms a dummy variable is created, with PwC as point of reference since it is the audit company that issued the most reports for companies in the financial sector. The coding of the dummy variable is presented in Appendix 3.

3.2.3 Control variables

Following prior literature (Pinto and Morais, 2019; Velte, 2018; Laitinen and Laitinen, 1998; Johl, Jubb and Houghton, 2007; Molyneux and Thornton, 1992; Lennox et al., 2017) the control variables used in this study are firm size (SIZE), the profitability of the audited entity (PROF) and the firm complexity (CMPLX).

Pinto and Morais (2019) and Velte (2018) found a significant positive relation between firm size (SIZE) and the number of KAMs. According to Pinto and Morais (2019), large firms are more complex, requiring more work from the auditors, and posing greater risk to the auditor’s liability. Moreover, Velte (2018) stated that firm size is linked to greater audit resources. The amount of KAMs disclosed can influence the variety of KAMs and thus the size of the audited entity is controlled for. The size of the

(19)

Total assets in euros and dollars are converted to pound to be able to compare on this variable properly. The conversion rate of 31 December of the different years are used. This conversion rates are presented in Appendix 3.

The profitability of the audited entity (PROF), which is the second control variable, influences the amount of KAMs disclosed. Firms with lower profitability present a higher risk of failure, consequently auditors may need to extend the scope of their work and disclose more KAMs (Pinto and Morais, 2019). Also, firms with lower profitability tend to use more creative accounting in their preparation of financial statements, which increases the probability of a qualified opinions and the disclosure of more KAMs (Laitinen and Laitinen, 1998; Johl, Jubb and Houghton, 2007). Thus, the more profitability the audited entity, the less KAMs are disclosed by the audit firm. This in turn influences the diversity of the long- form audit opinion. The profitability (PROF) is measured by the Return on Assets (ROA), calculated by dividing the net income by total assets. ROA is a widely used measure for the profitability (Molyneux and Thornton, 1992). The ROA ratios for the different companies have been collected from the Orbis database or are calculated by hand when not available in the database. When calculated by hand, information is obtained from the annual reports.

The last control variable used in this study is firm complexity (CMPLX). Research showed that the disclosure of KAMs could be affected by the firm complexity (Lennox et al., 2017; Pinto and Morais, 2019). In more complex firms, there are more areas of risk that lead to an increase in the number of disclosed KAMs (Pinto and Morais, 2019). Results of Lennox et al. (2017) support this argument by showing that a greater complexity relates to a higher number of KAMs. The amount of KAMs disclosed can influence the variety of KAMs and thus the complexity is controlled for. Following Lennox et al. (2017) the natural logarithm of the number of subsidiaries will proxy for firm complexity (CMPLX). The number of subsidiaries have been collected from the Orbis database. Due to time barriers the number of subsidiaries for each firm is assumed to be stable over years.

3.3 Statistical model

This study in conducted in a quantitative way by performing statistical tests. An independent samples t-test was conducted to compare the diversity of long-form audit opinions in early adopting countries (UK and NL) and late adopting countries (Germany and Italy). To perform the analysis the long-form audit opinions of the late adopters in 2016 and 2017 are combined in one sample and compared with the long form audit opinions of early adopters in 2013 and 2014. Thereafter, the sample of late adopters is compared with the long-form audit opinions of early adopters in 2015 and 2016. If the long-form audit opinions of early adopters differ significantly from the early adopters sample in 2013/2014 and are significantly similar to the long-form audit opinions of early adopters in 2015/2016, we can conclude that the other jurisdictions learn from lessons from early adopters. The reason for combining two years in one sample is because the data shows that in the first adopting year not all companies follow the

(20)

regulation.

A multiple regression analysis was created to examine the association between the processes of isomorphism and the diversity of KAMs in long-form audit opinions. The following multiple regression analyses are used:

𝐷𝐷𝐼𝐼𝐼𝐼 = 𝛽𝛽

0

+ + 𝛽𝛽

1

𝐿𝐿𝐿𝐿𝐿𝐿 + 𝛽𝛽

2

𝐹𝐹𝐹𝐹 + 𝛽𝛽

3

𝐵𝐵𝐼𝐼𝐵𝐵4 + 𝛽𝛽

4

𝑆𝑆𝐼𝐼𝑆𝑆𝑆𝑆 + 𝛽𝛽

5

𝑃𝑃𝑃𝑃𝑃𝑃𝐹𝐹 + 𝛽𝛽

6

𝐶𝐶𝐶𝐶𝑃𝑃𝐿𝐿𝐶𝐶 + 𝜀𝜀ᵢ

𝐷𝐷𝐼𝐼𝐼𝐼_𝑆𝑆𝑆𝑆𝐷𝐷 = 𝛽𝛽

0

+ + 𝛽𝛽

1

𝐿𝐿𝐿𝐿𝐿𝐿 + 𝛽𝛽

2

𝐹𝐹𝐹𝐹 + 𝛽𝛽

3

𝐵𝐵𝐼𝐼𝐵𝐵4 + 𝛽𝛽

4

𝑆𝑆𝐼𝐼𝑆𝑆𝑆𝑆 + 𝛽𝛽

5

𝑃𝑃𝑃𝑃𝑃𝑃𝐹𝐹 + 𝛽𝛽

6

𝐶𝐶𝐶𝐶𝑃𝑃𝐿𝐿𝐶𝐶 + 𝜀𝜀ᵢ

In this formula, 𝛽𝛽0 is constant, 𝛽𝛽𝑖𝑖 represent the independent and control variables used in this thesis and

εᵢ is the distortion term. Table 2 summarizes the variables, measurement of variables and type of variables used in the regression models.

Table 2: Description of variables

Variable Measure Type of

variable Dependent

DIV Blau Index of heterogeneity Interval

DIV_STD Index of Quality Variation (IQV) Interval

Independent

LAW The country in which the long form audit opinion is issued. Common law (UK) = 0 and code law (NL, Germany and Italy) = 1

Dummy FY The financial year on which the long form audit opinion is issued

(2013-2018)

Interval BIG4 The Big 4 audit firm that issued the long form audit opinion (PwC,

KPMG, EY or Deloitte)

Dummy

Control

SIZE The natural logarithm of the total assets Interval

PROF Return on assets obtained from the Orbis database or calculated by hand with information from the annual report

Interval CMPLX The natural logarithm of the number of subsidiaries obtained from

(21)

4. RESULTS

In this chapter, the results of the conducted research is presented. The first paragraph describes the transformation of outliers in the dataset. an overview is provided of the analyses and the corresponding results that were derived.

4.1 Dataset adjustments

Outliers skew the normal distribution of data which can lead to distorted results (Field, 2013). Therefore, variables were subject to the process of winsorizing. First, outliers are identified by creating boundaries that are calculated by the mean of a variable plus/minus the standard deviation times three. Thereafter, all data that falls outside these boundaries is revalued by the mean plus/minus the standard deviation times three. This resulted in the downward revaluation of 4 values to 44,20171 in the control variable ROA and downward revaluation of 12 values to 2576 in the control variable CMPLX.

4.2 Descriptive statistics

Table 3 gives a representation of the descriptive statistics of the variables used in this study. The table shows the sample, mean, standard deviation, minimum and maximum of the dependent and control variables for 472 long for audit opinions. The descriptive statistics for the independent variables are shown in a separate table, because the standard deviation, minimum and maximum do not provide useful information about these variables. Table 4 gives the descriptive statistics for testing the first hypothesis. The mean of the Blau index (DIV), which is the unstandardized diversity of the long form audit opinion, is 0,5493 (St. Dev. = 0,2307). This indicates that the chance that two randomly selected KAMs from the entire sample are different is 54,93%. The Index of Quality Variation (DIV_STD), which is de standardized diversity, shows a mean of 0,8096 (St. Dev. = 0,3190). The chance that two randomly selected KAMs are different when measuring the long form audit opinion diversity in a standardized way is higher, namely 80,96%. The Blau index of diversity ranges from 0 to 0,84. However, the Index of Quality Variation ranges from 0 to 1. This is because the maximum of the Blau index of diversity is limited by the number of categories (Harrison and Klein, 2007). The natural logarithm of the size of all audited entities in the sample is on average 21,4327 (St. Dev. = 5,2284). The profitability of the audited entity shows a mean of 4,6203 (St. Dev. = 7,2310). On average the natural logarithm of the number of subsidiaries is 4,2265 (St. Dev.=1,9987) For the control variable profitability it appears that the standard deviation is higher than the mean. This indicates that the profitability of the firms is quite spread out.

(22)

Table 3: Descriptive statistics for dependent and control variables

Dependent variables N Mean St. Dev. Min Max

DIV 472 0,5493 0,2307 0 0,84 DIV_STD 472 0,8096 0,3190 0 1 Control variables SIZE 472 21,4327 5,2284 6,4312 29,1188 PROF 472 4,6203 7,2310 -12,33 44,20 CMPLX 472 4,2265 1,9987 0 7,8540

In table 4 the descriptive statistics between the early and late adopters are presented. In 2013 and 2014 the early adopting countries issued 124 long-form audit opinions. This long-form audit opinions score on average 0,5728 (St. Dev. =0,2233) on the Blau index and 0,8127 (St. Dev. =0,3006) on the IQV. In 2015 and 2016 the early adopting countries issued some more long-from audit opinions, namely 166. These 166 opinions show an average of 0,5256 (St. Dev. =0,2428) on the Blau index and 0,7998 (St. Dev.=0,3446) on the IQV. However, the most diverse opinions are issued for the late adopting countries in 2016 and 2017. The 24 long-form audit opinions issued in these countries score on average 0,6667 (St. Dev. =0,1183) on the Blau index and 0,9158 (St. Dev. =0,1253) on the IQV.

Table 4: Descriptive statistics for differences between early and late adopters

Variables N Mean (Blau index) St. Dev. (Blau index) Mean (IQV) St. Dev (IQV)

Early adopters (2013- 2014) 124 0,5728 0,2233 0,8127 0,3006 Early adopters (2015- 2016) 166 0,5356 0,2428 0,7998 0,3446 Late adopters (2016- 2017) 24 0,6667 0,1183 0,9158 0,1253

Table 5 presents the descriptive statistics for the common and code law countries, table 6 shows the descriptive statistics of each year included in this study and table 7 the descriptives of the Big 4 audit firms. 356 long-form audit opinions are issued in the common law country (UK) and 116 in the code law countries (NL, Germany and Italy). The reason for this is because the FTSE100 and the first 150 companies of the FTSE250 are taken into account for the UK. For the other countries smaller indexes are used. Respectively the unstandardized and the standardized diversity shows a higher mean in code law compared to common law countries. The most (94) long-form audit opinions are issued in 2017 and 2018 and the least (45) in 2013. For the years 2014, 2015 and 2016 respectively 76, 81 and 82 long-form audit opinions are issued. The mean for both the Blau index and the IQV fluctuates, but decreases over time. The descriptive statistics for the independent variable audit firms shows that PwC issued the most

(23)

long-form audit opinions with 170. KPMG issued 126 long form audit-opinions and both EY and Deloitte issued 88 long-form audit opinions. Both the standardized and unstandardized indexes show the highest mean for audit opinions issued by PwC and the lowest mean for audit opinions issued by KPMG.

Table 5: Descriptive statistics per law origin

Law origin Frequency Percentage Cumulative Mean (DIV) (DIV_STD) Mean

Common 356 75,42 75,42 0,5270 0,7878

Code 116 24,58 100 0,6176 0,8764

Total 472 100 - - -

Table 6: Descriptive statistics per year

Year Frequency Percentage Cumulative Mean (DIV) (DIV_STD) Mean

2013 45 9,53 9,53 0,6006 0,8342 2014 76 16,10 25,63 0,5569 0,7926 2015 81 17,16 42,79 0,5407 0,8182 2016 82 17,37 60,16 0,5567 0,8140 2017 94 19,92 80,08 0,5573 0,8226 2018 94 19,92 100 0,5098 0,7848 Total 472 100 - - -

Table 7: Descriptive statistics per audit firm

Audit firm Frequency Percentage Cumulative Mean (DIV) (DIV_STD) Mean

PwC 170 36,03 36,03 0,5897 0,8613

KPMG 126 26,69 62,72 0,4954 0,7259

EY 88 18,64 81,36 0,5694 0,8536

Deloitte 88 18,64 100 0,5272 0,7855

Total 472 100 - - -

As stated in paragraph 3.2.2 a dummy variable for the legal system and the issuing audit firm is created. When using a dummy variable a rule of thumb is to work with a reference group which is left out of the rest of the analyses. In this study PwC issued to most extended audit opinions and it used as the point of reference and left out in the rest of the analyses.

4.3 Normality test

To perform statistical analyses, it is assumed that the dependent variable is normally distributed. In order to statistically test if the diversity of the long-form audit opinion is normally distributed, a Shapiro-Wilk test is conducted. The Shapiro-Wilk test is appropriate for small sample sizes (<50), but can also handle samples sizes as large as 2000 and therefore appropriate for the relatively small late adopters (2016-2017) group.

(24)

When normality is tested using a Shapiro-Wilk test, a non-significant p-value indicates that we may assume that the data is normally distributed. As presented in table 8, the p-value of the dependent variable for the total sample is for both the Blau index and the IQV 0,000, indicating that the dependent variable is not normally distributed. Additional, in the subsample the p-value of both early adopters samples is 0,000. This indicates that these are also not normally distributed. However, the Central Limit Theorem states that the distribution of sample means approximates a normal distribution as the sample size gets larger. Sample size equal or greater than 30 are considered sufficient for the CLT to hold. The late adopters subsample is less than 30. Nevertheless, the p-value of the Blau index and IQV for this subsample is greater than 0,05 an is therefore not significant at a 0.05 level. Therefore, we can assume a normal distribution of the dependent variable for all samples used in this study.

4.4 Correlation analysis

The correlation analysis checks for multicollinearity between the different variables. If multicollinearity exists it is hard to obtain unique estimates of the regression coefficients. The concept of multicollinearity is assumed when the correlation between two different variables is lower than -0.8 or higher than 0.8 (Field, 2013). From the results of the Pearson correlation analysis in table 9 it can be concluded that multicollinearity exists between the Blau index and the IQV (0.886). However, this is not a problem for the regression analyses since both variables are placed in different regression models. Moreover, this relation is logical and explainable. The unstandardized measure of diversity will increase when the standardized measure of diversity increases. Based on the correlation matrix two significant relations exist between the dependent and the independent variables. Firstly, a negative relation between the diversity of the long-form audit opinion and the legal system of the country (-0.169 and -0.120). This indicates that code law countries issue more diverse long-form audit opinions compared with common law countries. Secondly, a negative relationship between the diversity of the long-form audit opinion and the audit firm KPMG (-0.136 and -0.156). This indicates that KPMG issues less diverse long-form audit opinions in comparison PwC. No significant correlations are found between EY and Deloitte and PwC as reference group.

Table 8: Shapiro-Wilk test of normality for total sample and subsample

Statistic df Sig. Blau index 0,778 472 0,000 Early adopters (2013-2014) 0,761 124 0,000 Early adopters (2015-2016) 0,771 166 0,000 Late adopters (2016-2017) 0,923 24 0,067 IQV 0,614 472 0,000 Early adopters (2013-2014) 0,637 124 0,000 Early adopters (2015-2016) 0,588 166 0,000 Late adopters (2016-2017) 0,856 24 0,059

(25)

Table 9: Correlation matrix Variables 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 1. DIV 1,00 2. DIV_STD 0,886** 1,00 3. LAW -0,169** -0,120** 1,00 4. FY -0,078 -0,018 -0,203** 1,00 5. KPMG -0,136** -0,156** -0,056 0,038 1,00 6. EY 0,039 0,063 -0,043 0,020 -0,289** 1,00 7. Deloitte -0,046 -0,036 0,058 -0,003 -0,289** -0,229** 1,00 8. SIZE 0,194** 0,074 -0,293** 0,226** 0,046 0,068 -0,077 1,00 9. PROF -0,110* -0,030 0,256** -0,107* -0,013 -0,054 -0,035 -0,265** 1,00 10. CMPLX 0,155** 0,023 -0,033 0,138** 0,018 -0,025 -0,120** 0,696** -0,065 1,00

(26)

In addition, the variance inflation factor (VIF) of each variable are calculated and presented in table 10. A VIF higher than 5 indicates issues of multicollinearity (Stine, 1995). The low values of VIF indicate that there are no issues of multicollinearity since the values are less than 5.

Table 10: Variation inflation factor (VIF) – Blau index

Variables Tolerance VIF

LAW 0,814 1,228 FY 0,926 1,080 KPMG 0,771 1,296 EY 0,785 1,274 Deloitte 0,784 1,275 SIZE 0,405 2,471 PROF 0,875 1,143 CMPLX 0,459 2,178

a. Dependent variable: Blau index/IQV

4.5 Independent samples t-test

An independent samples t-test was conducted to test the first hypothesis that states that long-form audit opinions of late adopters are less diverse than long-form audit opinions of early adopters. Table 11 up to and including 14 present the results of the conducted independent samples t-tests using respectively the Blau index and the IQV.

Table 11: Results independent samples t-test early adopters (2013-2014) – late adopters (2016-

2017) Blau index

Levene’s test for equality of variances T-test

F-statistic Sig. T-statistic Df Sig.

6,093 0,015 -2,989 60,301 0,004

2017) Blau index

7,796 0,006 -4,275 56,609 0,000

2017) IQV

(27)

2017) IQV

10,554 0,001 -3,134 86,388 0,002

Firstly, the independent samples t-test was conducted using the Blau index. The assumption of equal variances between early adopters (2013-2014) and late adopters (2016-2017) was tested and rejected via Levene’s F test, F(60,3) = 6,093, p. = 0,015. The independent samples t-test was associated with significant effect at the 0.05 level (t (60,3) = -2,989, p < 0,05, Cohen’s D = 0,53). Thus, long-form audit opinions issued by early adopters (2013-2014) were associated with a statistically significant different Blau index than late adopters (2016-2017). The assumption of equal variances between early adopters (2015-2016) late dopters (2016-2017) was also rejected via Levene’s F test, F(56,6) = 7,796, p. = 0,006. The independent samples t-test was associated with a significant effect at the 0.05 level (t (56,6) = -4,275, p<0,01, Cohen’s D = 0,69). Thus, long-form audit opinions issued by early adopters (2015-2016) were also associated with a statistically significant different Blau index than late adopters. Secondly, the independent samples t-test was conducted using the IQV. The assumption of equal variances between early adopters (2013-2014) and late adopters (2016-2017) was tested and rejected via Levene’s F test, F(83,4) = 6,422, p. = 0,012. The independent samples t-test was associated with significant effect at the 0.05 level (t (83,4) = -2,771, p < 0,05, Cohen’s D = 0,45). Thus, long-form audit opinions issued by early adopters (2013-2014) were associated with a statistically significant different IQV than late adopters (2016-2017). The assumption of equal variances between early adopters (2015-2016) late adopters (2016-2017) was also rejected via Levene’s F test, F(86,4) = 10,554, p. = 0,001. The independent samples t-test was associated with a significant effect at the 0.05 level (t (86,4 = -3,134, p<0,05, Cohen’s D = 0,45). Thus, long-form audit opinions issued by early adopters (2015-2016) were also associated with a statistically significant different IQV than late adopters.

That means that long form audit opinions issued by auditors in early adopting countries are associated with a statistically significant different level of diversity than audit opinions issued by auditors in late adopting countries. The hypothesis assumes that long-form audit opinions issued by late adopting countries are less diverse than long-form audit opinions issued by early adopters. However, the results show that long-form audit opinions issued by auditors in late adopting countries are significantly more diverse than audit opinions issued by auditors in early adopting countries, both compared to the early adopters in early years (2013-2014) and to later years (2015-2016). Therefore we reject hypothesis 1 as we found evidence that contradicted our hypothesis.

(28)

4.6 Regression analysis

Since the dependent variable is measured in two different ways, each regression is performed twice with two different dependent variables. The unstandardized diversity (Blau index) of the long-form audit opinion is the dependent variable in the models labeled ‘A’. In the models labeled ‘B’, the standardized diversity (IQV) of the long-form audit opinion is the dependent variable.

4.6.1 Control variables

Before the hypothesis were tested we ran a linear regression to test the effect of our control variables. The results are presented in table 15.

Table 15: Results regression analysis model 1

Variables Model 1A Model 1B

Constant 0,403*** 0,703***

SIZE 0,006** 0,007

PROF -0,002 0,928

CMPLX 0,006 0,411

Adjusted R-Square 0,037 0,001

*** Significant on a 1% level, **Significant on a 5% level, * Significant on a 10% level

The adjusted R-square of model 1A is 0,037, which indicates that the variation in the diversity of the long-form audit opinion is explained for 3,7% by the control variables size, profitability and complexity. Model 1A indicates a significant positive relation (β=0,006, p<0,05) between the size of the audited entity and the unstandardized diversity of the audit opinion. Thus, this model indicates that when the audited firm is larger, the diversity of the long-form audit opinion increases. In model 1B the adjusted R-square is 0,001. This means that in model 1B the control variables explain the variation in the standardized diversity of the long-form audit opinions for 0,1%. Model 1B indicates no significant relationships between the control variables and the standardized diversity of the long-form audit opinion.

4.6.2 Legal systems

The second model is used to test hypotheses 2a and 2b. Hypothesis 2a proposes that long-form audit opinions issued by auditors in common law countries are subject to stronger pressures of coercive isomorphism, resulting in more diverse audit opinions. Hypothesis 2b is the contradicting hypothesis which proposes that long-form audit reports issued by auditors in code law countries are subject to stronger pressures of coercive isomorphism, resulting in more diverse audit opinions. Both hypothesis assume a positive relationship between the legal system and the diversity of the extended audit opinion. Table 16 presents that results of the performed regression to test hypotheses 2a and 2b.

(29)

Constant 0,421*** 0,722*** LAW 0,070*** 0,078** SIZE 0,003 0,004 PROF -0,002 0,001 CMPLX 0,011 -0,004 Adjusted R-Square 0,049 0,008

*** Significant on a 1% level, **Significant on a 5% level, * Significant on a 10% level

When including the independent variable LAW in the regression model, the explanatory power increases in both models. The explanatory power increases from 3,7% in model 1A to 4,9% in model 2A and from 0,1% in model 1B to 0,8% in model 2B. Model 2A shows a significant positive relationship (β= 0,070, p<0,01) between LAW and the unstandardized diversity of the long-form audit opinion. This means that long-form audit opinions issued by code law countries are significantly more diverse than long-form audit opinions issued by code law countries. Model 2B also indicates a significant positive relationship (β= 0,078, p<0,05) between LAW and the standardized diversity of the long-form audit opinion. However this relationship is less significant. Hypothesis 2a expected a positive relationship between long-form audit opinions issued by auditors in common law countries and the diversity of the long-form audit opinion and is therefore rejected. However, hypothesis 2b expected a positive relationship between long-form audit opinions issued by auditors in code law countries and the diversity of the long-form audit opinion and is therefore accepted.

4.6.3 Financial year

The third model tests the third hypothesis which states that more recent long-form audit opinions are subject to stronger pressures of mimetic isomorphism, resulting in less diverse audit opinions. A negative relationship is expected between the year on which the long-form audit opinion is issued and the diversity of the long-form audit opinion. Table 17 shows the results of the performed regression analysis to test hypothesis 3.

Constant 37,834*** 15,799 FY -0,019*** -0,007 SIZE 0,007** 0,007* PROF -0,002 0,000 CMPLX 0,410 -0,009 Adjusted R-Square 0,051 0,000