The effect of auditor gender and the moderating effect of audit committee characteristics on the readability of KAMs

(1)

The effect of auditor gender and the

moderating effect of audit committee

characteristics on the readability of KAMs

Master Thesis, MSc Accountancy

Mark Jan IJkema (S2948486)

University of Groningen, Faculty of Economics and Business

Van Tuyll van Serooskerkenweg 16-2

1076 JL, Amsterdam

E-mail:

m.j.ijkema@student.rug.nl

Tel.: 06 – 37 56 48 33

Supervisor: Prof. Dr. D.A. de Waard

Second assessor: Dr. D.B. Veltrop

16

th

of January 2020

Wordcount: 10.339

ABSTRACT: With the introduction of the extended audit report, steps were taken to make the audit process more transparent and give users of financial information more firm-specific information. In this article, a quantitative research is performed on the relation between auditor gender and the moderating effect of audit committee characteristics on the readability of Key Audit Matters (KAMs) on listed companies in the UK, The Netherlands, and Germany. Key Audit Matters are the firm-specific risks that are included in the extended audit report. The sample used consists of 605 firm years from 2016 to 2017. Results show that female audit partners have a positive effect on the readability of KAMs. Further tests show that meeting frequency of audit committees have a negative effect on KAMs readability. The test did not provide evidence on the expected moderating effect of the audit committee characteristics. KEYWORDS: Key Audit Matters, Readability, External Auditors, Audit Committee

(2)

ACKNOWLEDGEMENTS:

By writing this acknowledgment an intensive period of writing on my thesis come to an end. Deloitte gave me the opportunity to do my internship at their office in Amsterdam. I have been working on my thesis with pleasure at this location. I will thank my supervisor Prof. Dr. de Waard for the interesting discussions on the theme related to the thesis and the feedback and support I received and I also thank my colleagues of Deloitte for the advice and co-student Gineke Spoelstra for the cooperation to collect data used in the research.

(3)

-1. INTRODUCTION & LITERATURE CONTRIBUTION

ue to several audit scandals in the last two decades (i.e. Enron, Imtech and WorldCom), the audit profession suffered reputational damage and lost the confidence of the society. The audit opinion was questioned after these major scandals. According to Humphrey, Loft and Woods (2009) the pass/fail model, wherein the auditor gives his/her view on the financial statements, was no longer sufficient for a proper audit process. The audit report presented almost no firm-specific information that is useful for investors’ decision making. In reaction to the investors’ demand and to restore the image of the external auditor, the Financial Reporting Council (FRC) in the UK issued new standards for the audit report in 2013. By introducing requirements to the audit report, where auditors provide an explanation on the 1) scope of the audit 2) firm-specific risks 3) the application of materiality and 4) auditors’ responsibility, FRC tries to rebuild the trust of the society and improve the quality and transparency of the audit report. Reliability and transparency were the main objectives of the 2013 issued standards for the new form audit report (FRC, 2016). By adding Key Audit Matters (KAMs) to the audit report as it exists now, FRC responds to the investors' demand for more company-specific information that enhance transparency on the audit process. Following the example set by the FRC in the UK, regulators in several countries around the world introduced new and updated standards with the aim to generate more transparency. The Netherlands was first in following the example of the FRC. The Nederlandse Beroepsorganisatie van Accountants (NBA) was accountable for the introduction of these requirements in 2014 (NBA, 2014). The NBA is an organisation established by law that is responsible for the continuous improvement of the audit profession in the Netherlands. These changes and the introduction of the KAMs have led to the ‘Extended Auditor’s Report’, from now on ‘Extended Audit Report’. External auditors are responsible for specifying KAMs to the annual report of companies. Since this is the case, the characteristics of the auditor do have an influence on how KAMs are reported. By disclosing the KAMs in an effective way, the (extended) audit report should enhance effective communication between the auditor and the users of the audit report. Due to better communication of the audit process to the intended users, the expectation gap reduces (IFAC, ISA 701). The expectation gap is the difference between what the users of the audit report expect from the auditor and what auditors actually do (Chye Koh and Woo, 1998). Providing this extra information to the users of financial statements, also the decision usefulness of the audit opinion should increase for professional- and even nonprofessional investors (Low and Siesfeld, 1998). A precondition for increased decision usefulness is that the communication of financial information is well performed. To assess the level of communication by financial and non-financial statements the measure ‘readability’ can be used as a variable (Smith and Smith, 1971).

The communication of financial statements to users of the financial information is not only affected by corporate governance characteristics of the external auditor but also characteristics of the board and the audit committee (AC) (Velte, 2017; Landry, Bernardi and Bosco, 2014; Dechow et al.,

(5)

1996; Dezoort and Salterio, 2001; Abbott et al., 2001). Since the late 90s, corporate governance is more and more a tool to improve the quality of financial statements disclosures and audit reports (Cohen, Krishnamoorthy and Wright, 2002). In the UK the use of corporate governance instruments all started with the issuance of the Report of the Committee on the Financial Aspects of Corporate Governance in December 1992. One recommendation of the report was that the board of UK listed firms should consist of at least 3 external directors, but these recommendations were all non-compulsory (Committee on the financial aspects of corporate governance & Cadbury, 1992). In the years following a lot of countries, including the Netherlands, introduced a corporate governance code that assists in long-term value creation. The revised Dutch Corporate Governance code of 2008 also included guidance on board composition which would support the creation of corporate value (De Nederlandse Corporate Governance Code: Beginselen van deugdelijk ondernemingsbestuur en best practice bepalingen, 2008). Prior research shows that the gender diversity of board or subcommittee members have an influence on the financial and non-financial performance of a company (Landry, Bernardi and Bosco, 2014), but also the size of the board or AC plays a role in financial reporting quality (Bedard & Gendron, 2010). Another characteristic in the board and AC that affects financial reporting is the number of meetings (Abbott et al., 2001; Raghunandan, 1996; Vafeas, 2005).

Besides corporate governance, ACs are a recent topic in research as well. ACs are a mechanism to develop the quality of the mandatory audit process (Ghafran and O’Sullivan, 2012). In 2003, the Sarbanes-Oxley Act (SOX) was established. This legislation was created to protect shareholders and users of the financial statements from financial reporting errors. The objective of the introduction of the SOX was to monitor the financial reporting of financial professionals. And by doing this, rebuild and enhance the society’s and investors’ trust level by adding independence and expertise requirements to the ACs (U.S. Congres, 2002). The same initiatives were imposed in many other countries around the world. Still, changes in regulations of ACs are common and regulations are continuously improving. During the past decade, the focus in the literature of the ACs moved to how AC characteristics impact elements of the financial reporting process (Ghafran and O’Sullivan, 2012).

The AC is the subcommittee to be in closest contact with the external auditor. The auditor is the watchdog of the society, however even the auditor needs help in performing a proper audit (Levitt, 2000). Besides, the AC takes care of the independence of the external auditor. The independence of the auditor is a recurrent ongoing hot topic (Carcello and Neal, 2003). Since the auditor is responsible for setting up the KAMs of a company, characteristics of the auditor do play a role in the quality of these KAMs. In this research, we will focus on the effect of the auditor gender, because the increasing number of women in the auditing field demands for research and the consequences of the effect of gender of the auditor on audit disclosures (Khalifa, 2013).

The AC also requires a high-quality audit, because the external auditor assists the AC in its monitoring duties (Velte, 2017). For that reason, the AC characteristics and the characteristics of the external auditor should have an impact on the interaction between the AC and the external auditor (Velte,

(6)

2017; He, Pittman, Rui and Wu, 2017). According to Johnstone, Sutton and Warfield (2001) corporate governance mechanisms provide some assurance that audit independence risk is being managed. They state that: ‘’Appropriately functioning boards and ACs should provide a neutral, well-informed buffer between auditors and management (Johnstone et al., 2001).’’

As mentioned before, the aim is that KAMs should add more content and value for the user to the audit report. This should lead to better-informed investment decisions based on the more firm-specific information and the transparency on the audit process that is given. In this paper, the effect of auditor gender and several characteristics of the audit committee on the readability of KAMs will be assessed. If KAMs are readable, users of the information are better able to understand the KAMs and thus make better-informed decisions. Therefore, the research question is as follows:

“To what extent does the gender of the external auditor affect the readability of Key Audit Matters in the UK, the Netherlands and Germany? And what is the moderating effect of Audit Committee

Characteristics?”

This research responds to the demand for more research on the effect of auditor gender and AC characteristics on auditor reports (Velte, 2017; Khalifa, 2013). Besides that, the study will contribute to the literature since there is a research gap regarding the effect of corporate governance characteristics on the readability of KAM disclosures. The study will provide new insights for regulators and practitioners.

The literature and research available on extended audit reports are limited because extended audit reports only exist in the UK since 2013 and in other countries since 2014. Although, some research is done on the value relevance of the extended audit report. Porumb et al. (2019) examined the effect of KAMs on debt contracting terms. They found that the introduction of the KAMs are associated with more favourable loan contracting terms. Another finding was that extra information available for lenders decreases adverse selection problems (Porumb et al., 2019). Christensen, Glover and Wolfe (2013) researched the effect of KAMs on investment decisions of nonprofessional investors. Their outcomes provided evidence that nonprofessional investors are more likely to stop investing due to the disclosure of KAMs. According to this study, KAM disclosures are decision-relevant for users. However, research of Lennox, Schmidt and Thompson (2015) also focused on the responses of individual investors. The results of this study show that investors make use of other information sources to base decisions on.

In 2017, Velte (2017) was the first in researching the effect of a corporate governance characteristic on the readability of KAM disclosures. His research was limited to only one variable. Velte examined the effect of gender diversity by assessing the percentage of women in the AC on KAM readability. The focus of the research was only on UK Listed companies. Velte (2017) stated in his paper that more research is needed on the effect of AC characteristics on the readability of KAMs and he also

(7)

mentioned that more countries should be included in future research to make comparisons or finding some differences across countries. Probably, different countries will show different results.

To the best of my knowledge, this is the first research that is assessing the effect of auditor gender on the readability of the KAMs. Furthermore, the moderating effect of AC characteristics is included in this paper. Namely, in this research the effect of gender diversity in the AC, meeting frequency of the AC and the size of the AC are included as well. As mentioned above, the research done on KAMs is relatively scarce, let alone the readability of KAMs. However, there is prior research that took into account auditor gender. In this paper is stated that auditors' gender is influencing the audit process from two perspectives. On the one hand is stated that women are more risk-averse than men (Dwyer, Gilkenson and List, 2002; Olsen and Cox, 2001), but on the other hand, O’Donnell and Johnson (2001) state that women are more complete and more specific in information processing and reporting. The gender differences are possibly a factor that influence the readability of KAMs.

Expected is that this research will contribute to the existing literature if can be concluded that auditor gender influences the readability of the KAMs and if characteristics of the AC have a moderating effect on this relation. A significant effect of these relations can be used by regulators to enhance regulations on corporate governance of external auditors and characteristics of ACs. If these variables affect the readability, the results can be used enhancing requirements to improve the value of KAMs disclosures for the users of audit reports in the future, but also information value for investors will increase what helps to reduce the information gap.

(8)

2. THEORETICAL BACKGROUND & HYPOTHESES

In this section, the focus is on the underlying theories (2.1), background information (2.2) and the development of the hypothesis (2.3). First of all, the agency, stakeholder, legitimacy and Limpergs’ inspired confidence theory will be discussed. Secondly, the audit report and the extended audit report will be mentioned. Lastly, the hypotheses are formulated to answer the research question.

2.1 Agency -, stakeholder -, the legitimacy - and Limpergs’ Inspired confidence theory

In most studies the disclosure of financial and non-financial information is explained by the

principal-agent theory (agency theory) and the stakeholder theory. The agency theory describes the problem that

exists between the principal and the agent due to the separation of ownership and control (Adams, 1994). Consequently, information asymmetry exists, because the agent will behave in his own interest. In this study the principal-agent problem can be applied to the stakeholders (principals) and the client of an audit firm (agents). Finally, the external auditor should publish KAMs which contain firm-specific information for the stakeholders. The stakeholders (principals) asked for more information next to the financial statements and audit opinion. The KAMs are a manner to provide extra firm-specific information to users. By disclosing KAMs the information gap between the firm (agent), and stakeholders (principal) will decrease. The information gap is the lack of information that one party has compared to the other party that works on a common problem (Doughty and Pica, 1986). Finally, the external auditor is responsible for the publication of the KAMs, but the AC is given the opportunity to discuss the results of the audit and propose adjustments in the report.

As mentioned, this problem can also be seen from the stakeholder theory. In the case of the stakeholder theory the firm should create value for its stakeholders by aligning the firms’ goals with the stakeholders’ interests. Stakeholders are all persons, groups or organisations that affect or are affected by the firm (Freeman and McVea, 2001). Morals and values in an organisation do play a key role in this theory. To maintain and create relationships firms should focus on issues like corporate social responsibility and social contracts, and align them with the interest of the stakeholders. Where the agency theory is focussing on the shareholders only, the stakeholder theory is considering other groups and individuals who are affecting the company or are affected by the company. Stakeholders want strict audits of the financial reports and a modified opinion by the auditor when the financial statements contain material misstatements. In this situation the auditor does act as a ‘gatekeeper’ for stakeholders. The audit committee also benefits from a proper external audit, because the external auditor is assisting the audit committee and vice versa (Velte, 2017).

The theories mentioned above are from the perspective of the stakeholders and shareholders. However, this research can also be seen from the perspective of the company because the KAMs also depends on the work done by the internal audit function and also the AC is able to have an indirect effect on the reporting of the KAMs. The legitimacy theory emphasizes that organizations need to behave

(9)

within the boundaries society expects from the company (Deegan, 2019). These boundaries depend on time and place, so the firms are dependent on the perceptions of a society at a specific time and at a specific place. Otherwise, firms will not be in line with the stakeholders' expectations and possibly will fear reputational damage (Comyns, 2016). For firms it is thus extremely important to follow the demands and requirements of the society to be recognized as a legitimate firm. In this case, firms behave legitimately by disclosing by the society requested additional information to the audit report.

Another theory that is emphasizing the supply and demand for auditing is Limpergs’ theory of

inspired confidence. The demand for audit-related services is a consequence of the interest of third

parties in a firm. Third parties expecting reliable reporting of financial statements by the management of the firm. An auditor should supervise and enforce the reliability of the financial statements. By giving assurance on an assurance engagement, the auditor always seeks to meet the expectations of users of financial statements in a critical and independent way, because the auditor has to give assurance on the requirements of the society. Changing expectations of society will thus affect the work of an auditor (Limperg, 1932).

Viewed from the different theories and perspectives the main goal of the extended audit report is to decline the information gap between the users of financial reports and the company and enhance the transparency of the audit process. In the next section, the methodology is presented about how auditor gender and several audit committee characteristics may have an effect on the readability of KAMs.

(10)

2.2 The (Extended) audit report and audit committees

2.2.1 (Extended) audit report by the external auditor

The audit report has come in for much criticism, especially with respect to the limited information included in the report and the standard layout. After several collapses (i.e. Enron, Imtech and WorldCom), the opinion of the public was that auditors were unable to report the effects that entail significant risks. In the crisis, firms that received an unqualified audit opinion collapsed only after a few months. Auditors failed to mention the risks of this crisis and the significant risks the crisis had on these firms. The limited risk reporting in the crisis started a movement in regulations and standard-setting.

Since 2013, countries started to adopt new standards on improving communication with users of audit reports, because the informational value of the audit report was questioned. With the introduction of new standards, the extended audit report was established. This new standard required auditors to provide more firm-specific information as mentioned in the introduction of the paper. The information should give users of the financial statements a better understanding of what firm-specific risks play an important role (FRC, 2016). The auditor must provide an explanation on the following topics:

1) scope of the audit

2) the application of materiality 3) identify firm-specific risks (KAMs) 4) Auditor’s responsibility

The part in the extended audit report that is the most important for users of financial statements is the KAMs. The KAMs give useful insights for stakeholders that make investment decisions based on the audit report. Therefore, this research is focused on the readability of the KAMs.

2.2.2 Audit committees

During the last two decades, ACs received increased importance in corporate governance requirements and regulations. The importance of the monitoring role of the AC on the financial reporting process is addressed by the regulations of the Sarbanes-Oxley act (2002). From on 2002, this regulation is applicable for ACs, ACs are since that moment responsible for the appointment, compensation and oversight of the external auditor. Not only the regulations or the requirements emphasize the importance of the AC but more and more literature also focuses on the role of the AC as an important corporate governance factor to enhance the audit quality (Ittonen, Miettinen and Vähämaa, 2010; He et al., 2017).

(11)

Nowadays, according to the SOX 301, the AC is seen as one of the most important subcommittees, because the AC is the monitoring mechanism for shareholders’ interest, negotiating about the engagement terms/fee and hire and fire auditors. The AC is also overseeing the financial reporting done by the internal auditors of the company (DeZoort, 1997). Besides, the AC is an intermediary between the board and the external auditor to enhance communication between both parties. The AC and the external auditor have to meet at least once a year. During these meetings the engagement is discussed and issues in accordance with the audit process will be analyzed (FRC, 2012). The meetings should improve the quality of the audit. The main function of the AC is to increase the quality of the audit by guaranteeing high-quality audit information, safeguarding for complying with regulations, monitoring internal auditors, the appointment of the external auditor and ensuring the independence of the external auditor (SOX 301).

Based on the indirect and direct effect of the AC on the financial information disclosure, AC can influence the operations, strategy and performance of a firm (Karim, Robin and Suh, 2016). Concluding, the cooperation between the AC and the external auditors is that the AC and, with specific reference to AC characteristics, may have an impact on what is reported in the annual financial statements and the audit report.

2.3 Hypotheses development

2.3.1 Dependent variable ‘Readability of KAMs’

The dependent variable in this research is the ‘Readability of KAMs’. The goal of readability is to enhance effective communication of decision or valuation-relevant information (Loughran and McDonald, 2014). In the case that KAMs are readable individual investors with less financial knowledge and expertise are able to make better informed decisions. The definition of readability according to Loughran and McDonald (2014) is as follows:

“Readability is a measure of the complexity of the text in a report based on applying textual analysis algorithms and metrics.”

Although readability measures are widely used in literature there are some critics on readability formulas and the use of these formulas. Readability formulas are developed in the context of primary school education, but it is not validated that these formulas are useful for assessing the readability for high educated readers (Feenstra, 2012). Besides, Jones and Shoemaker (1994) also mention that readability formulas consider characteristics of the texts and not considering the individual characteristics of the reader like the level of education, expertise and experience.

Another critic is that most of the readability formulas focus only on word length and sentence length, readability should include more than these two measures (Jones, 1997). For that reason, in this paper two readability measures will be applied. The Fog index determines the readability on the basis of the

(12)

average number of words per sentence and percentage of complex words, whilst the Flesch reading ease formula looks at the number of syllables per word.

A. Flesch reading ease formula

The first formula that will be used to measure the readability of KAMs is the Flesch reading ease formula. This formula is used in the study of Velte as well (2017) and is considered as the most accurate readability formula. By using the Flesch reading ease formula one can determine how easy or how difficult the context of an English text is (Li, 2008). The score is between 0 and 100, where 0 is very difficult and 100 equals easy to read. The application of the formula is simple and reads as follows:

206.835 - (1,015 𝑥 (

𝑡𝑜𝑡𝑎𝑙 𝑤𝑜𝑟𝑑𝑠

𝑡𝑜𝑡𝑎𝑙 𝑠𝑒𝑛𝑡𝑒𝑛𝑐𝑒𝑠

)– (84,6 𝑥 ((

𝑡𝑜𝑡𝑎𝑙 𝑠𝑦𝑙𝑙𝑎𝑏𝑙𝑒𝑠 𝑡𝑜𝑡𝑎𝑙 𝑤𝑜𝑟𝑑𝑠

))

Table 2. Flesch reading ease score difficulty level

B. Fog Index

The second measure that will be used is the Fog index. The Fog index is the most used measure of readability in the accounting literature (Loughran and McDonald, 2014). The Fog index consists of two components. The first component is the ‘average number of words per sentence’ and the second component is ‘complex words’. Both components are criticized, although ‘average number of words per sentence’ correlates with several measures of readability, Loughran and McDonald (2014) state that the length of sentences in financial reports are less precious than in non-financial disclosure. The ‘complex words’ component is in accounting literature not that meaningful as well. Complex words are measured by the number of syllables, a word existing of more syllables decreases readability and is therefore more difficult. Words with more than two syllables are indicated as difficult. In financial disclosures, there are a lot of words that exist of two or more syllables but are not difficult to investors with financial expertise. Nevertheless, as mentioned before the Fog Index is the most used measure of readability in accounting research. Where a high score indicates a text is difficult to read and a low score indicates a text is more readable. The outcome estimates and represents the number of years of education that is

Flesch reading ease Score Difficulty level of readability

0 – 30 1. Very difficult 30 – 50 2. Difficult 50 – 60 3. Fairly difficult 60 – 70 4. Plain English 70 – 80 5. Fairly easy 80 – 90 6. Easy 90 – 100 7. Very easy

(13)

needed to understand the text after reading it once (Boritz, Hayes and Timosheko, 2016). The formula is clarified as follows:

𝐹𝑜𝑔 𝑖𝑛𝑑𝑒𝑥 = 0.4(𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑤𝑜𝑟𝑑𝑠 𝑝𝑒𝑟 𝑠𝑒𝑛𝑡𝑒𝑛𝑐𝑒

+ 𝑝𝑒𝑟𝑐𝑒𝑛𝑡 𝑜𝑓 𝑐𝑜𝑚𝑝𝑙𝑒𝑥 𝑤𝑜𝑟𝑑𝑠)

Table 1. Fog index difficulty level

2.3.2 Independent variables

In this research, the chosen variables are auditor gender and several audit committee characteristics such as the presence of female in the AC, AC size and the number of meetings of the AC in a year. The reason for applying these variables is the expected relation between de auditor and the readability of KAMs as the auditor is responsible for publishing the KAMs. It is also expected that the AC has a moderating effect on this relationship. In addition, will be assessed whether the AC characteristics affect the readability of KAMs on itself. Since, the AC is a corporate governance mechanism to protect shareholders’ interests (DeZoort, 1997). Another duty of an AC is to emphasize and consider the interest of shareholders. Furthermore, AC responsibilities are the appointment of the external auditor and ensure the transparency of the firms’ financial and non-financial information disclosure to their shareholders (Chiang and He, 2010).

Auditor Gender

The main relationship assessed in the paper is the effect of auditor gender on KAMs. Same as the increase of women on corporate boards, there was also an increase in the number of women in the audit profession in the late 90s. Therefore, researchers believe that it was really important to consider differences in gender in accounting research. The auditor's gender can illustrate the corporate governance effect of the external auditor on the readability of KAMs. Namely, the auditor is finally liable for the preparation of the KAMs (IFAC, ISA 701).

The selectivity hypothesis explains the differences in the level of processing information by men and women (O'Donnell and Johnson, 2001). Based on prior research women tend to base their judgments on an elaborating set of information while men base their decisions on only a limited and distinctive set of information. Because of this specific set of information that men make use of, the information is fast

Fog Index

Difficulty level of readability

0-5 1. Very easy 5-10 2. Easy

10-15 3. Plain English 15-20 4. Difficult 20-25 5. Very difficult

(14)

and easy to process. Therefore, men are faster in processing information, but also less extensive in the provided information (Meyers-Levy, 1998). Another difference between men and women is the level of risk aversion. Several psychology researchers show that men are more willing to accept actions that have negative consequences (Dwyer et al., 2002; Olsen and Cox, 2001). This implies that women are more conservative than men. Next to that, women scoring better on communicative abilities, which leads to a competitive advantage over men in functions where communication is an important part of the job (Schubert, 2006). Ittonen, Vähämaa and Vähämaa (2013) found that female audit partners do engage to a lesser extent in abnormal accruals, this confirms the conservative behaviour of women.

Due to the conservative behaviour and risk aversion of women and the more extensive provision of information, the description of the KAMs by a woman audit partner will be more elaborated and described in a way easier to understand than the descriptions written by men. Besides, the advantage women have compared to men in communication abilities, I will expect that the KAMs processed by women are easier to read.

H1: The readability of KAMs will increase if the audit engagement partner is a woman.

Gender diversity on Audit Committee

In the late 90s, there was an increasing move of women that practice a senior management function or in other corporate positions (Elsas and Graves, 1997). Till that moment there was only attention for diversity in terms of tenure, education, background and age and insufficient attention for gender diversity in literature. However, due to the movement of women to important management positions a debate in the literature about gender diversity arose. The mentioned failures in the introduction were another reason for increased attention to corporate governance reforms (Campbell and Minguez-Vera, 2007).

That women should be represented on boards can be explained in two different ways. On the one hand, there is an ethical reason, on the other hand, the economic one. Ethically it should be immoral to exclude women from boards based on gender. From the economically perspective boards should be composed in a way firm performance will increase. The presence of women on boards enhances firm performance (Campbell and Minguez-Vera 2007; Landry, Bernadi & Bosco, 2014). Various studies show that gender differences in boards lead to more effective communication capabilities since women are more effective in listening (Wood, Polek, and Aiken, 1985). Women also have increased incentives to ask critical questions and have different leadership styles than men (Konrad, Kramer, and Erkut, 2008). These gender diversity characteristics make that women behave better in group activities and more effective in solving problems. A broader view leads to a better understanding of the complexities of businesses and thus higher quality decision-making processes (Dallas, 2002; Campbell and Minguez, 2007). Another positive effect of diversity is that creativity and innovation will increase (Campbell and Minguez-Vera, 2007). The abovementioned arguments show several reasons to include women in the

(15)

boards and committees of organisations. Including women in the boards lead to firm value and enhanced financial performance (Campbell and Minguez-Vera 2007; Landry, Bernadi & Bosco, 2014).

The composition of the board of directors and subcommittees is indisputably the most important corporate governance determinant. The AC come to be perceived as the subcommittee that is most important of all board subcommittees (Klein, 1998 & Xie, Davidson and Dadalt, 2003). In earlier stage independence, expertise and accounting education were seen as critical factors that influence the ACs effectiveness, but gender diversity received increasing attention as an AC effectiveness increasing factor (Klein, 2002 & Abbott, Parker and Peters, 2004). Already some research is done on what effect women in audit committees have on financial and non-financial performance (Cucari, DeFalco, and Orlando, 2017; Thiruvadi and Huang, 2011; Law Chapple, Kent and Routledge, 2012). The representativeness of women in the AC (ACGEN) affecting auditor’s assessment of audit risk due to role the AC is playing in the improved communication with external and internal auditors, but also the enhanced monitoring role of the AC on the board of directors due to female representation plays a great role in the effectiveness of reporting processes (Ittonen, Miettinen and Vähämaa, 2010). Huse and Solberg (2006) state in their research that women come to meetings more prepared than men, which has a positive impact on committee effectiveness.

Velte (2017) did research on the effect of the percentage of women in the audit committee on KAM disclosures. Velte (2017) was first in assessing this relationship, but his study focussed on UK companies only. Expected was that the representativeness of women in the audit committee lead to better communication and monitoring capabilities. Therefore, cooperation with the external auditor enhanced and readability of the KAM disclosures increase. Therefore, the second hypothesis to be examined in this paper is as follows:

H2a: Female representativeness in the AC will enhance the readability of the KAMs.

Meeting frequency of Audit Committee

Likewise gender diversity, research is also focussing on meeting frequency (ACFREQ) since the early 90s. However, the corporate governance codes do not stipulate how often ACs should meet. For this reason, there is variance in audit committee meetings between firms (Sharma, Naiker and Lee, 2009). Because the number of meetings a year are not regulated, it is interesting to assess the effect of more frequent meetings. Several researchers already did some research on the effect of meeting frequency on various aspects of firms. Vafeas (1999), Lipton and Lorsch (1992) and Conger, Finegold and Lawler (1998) state that ACs that meet more often and at least quarterly are more active and able to serve their monitoring role to the board. In more active boards the directors do perform better, which leads to higher stock returns. Hoque, Islam and Azam (2013) also find that firm performance is increasing due to a higher number of meetings of the AC. They find that these results are stronger for large Australian firms. The existence of an AC as a subcommittee is negatively related to the perceived audit risk and a higher

(16)

number of AC meetings has a positive influence on this relationship (Steward and Munro, 2007). Prior research on meeting frequency of the AC state that ACs that are more active and thus do meet more frequently are less likely to be sanctioned by the Securities and Exchange Commission (SEC) (Abbott and Parker, 2000). Mc Mullen and Raghunandan (1996), Abbott et al. (2004) and Vafeas (2005) found that firms that experiencing reporting problems do have less frequent AC meetings. Krishnan and Visvanathan (2007) state that audit committees that meet more frequently were more likely to detect control weaknesses. Lastly, AC that meet more frequently have a significant negative effect on discretionary accruals (Xie, Wallace, Davidson and Dadalt, 2002). Due to the positive effect of more frequent AC meetings in prior research, in this research is expected that more frequent meetings will lead to more developed and detailed KAMs, consequently, the readability of KAMs will decrease. Therefore, the hypothesis is as follows:

H2b: More frequent meetings of the AC will lead to decreased readability of KAMs disclosure.

Size of the Audit Committee

The third AC characteristic that is included in the research is AC size (ACSIZE). Already some research is done on the effect of AC size. According to the resource dependence theory more members shall ensure that the available knowledge will increase in the committee. Audit committees that exist of more members are expected to include more experience, skills and expertise (Bedard & Gendron, 2010). On the other hand, if committees are too large the communication in the committees will decline and overall performance will as well. A reason for this effect could be that a small board is better able to make useful decisions and cope with obtained information (Chan and Li, 2008). Based on prior research which shows varied outcomes, I expect that the Size of the Audit Committee influences the readability of KAMs, but no direction is given to the relation. Therefore, the hypothesis is as follows:

H2c: There is a relationship between the audit committee size and the readability of KAMs.

At least we expect a moderating effect of AC characteristics on the relationship between auditor gender and the readability of KAMs due to the possible impact AC can have on the audit process and the reporting of the external auditor. As mentioned before, the AC is the subcommittee that is in closest contact with the external auditor and takes care of the independence of the external auditor (Levitt, 2000). Further, AC has to oversee and monitor the audit process and is able to fire the external auditor. After all, the AC has the opportunity to discuss with the auditor the work the auditor performed and thus is able to affect what is reported in the KAMs (He et al., 2017). Due to the role of the AC and the characteristics of both the AC and the external auditor can affect the communication and activities of the external auditor I expect that the AC characteristics have an effect on the relationship between auditor gender and the readability of the KAMs.

(17)

H3: Audit committee characteristics have a moderating effect on the relationship between auditor gender and readability of KAMs.

[Appendix 1 & 2]

2.3.3 Control variables:

As set out above, in this research the effect of auditor gender, Audit Committee characteristics and the moderating effect of AC on the relationship between auditor gender and KAMs readability is assessed. However, there are other variables that will affect these relations. Therefore, this research includes several control variables to control for the effect of these variables on the dependent variable. The control variables used in the study are control variables that are used more often in other corporate governance studies. In the research we control for Industry effects (INDj), Firm size (Sizej) and Legal

systems (LGS) (Velte, 2017; Buallay, Hamdan and Zureigat, 2017; Johnson and Greening, 1999).

Industry effects

The first variable that will be controlled for is industry effects (INDj). Listed companies around the

world differentiate among industries. For example, financial service firms are more complicated and carry greater risks than other firm types. This manifests itself in more risks and more complex risks reported in the KAMs. To classify the industries of the firms the Standard industrial classification (SIC) code of the companies is used.

Firm Size

Secondly, the control variable Firm Size (SIZEj) is included in the formula. According to Li (2008) Firm

size is positively related to the complexity of auditor reports. Due to a better control framework, Larger firms tend to publish more complex reports. There is a possibility that the size of the firm affects the risks mentioned in the extended audit report. As a consequence, the KAMs will be more and more specified and detailed for large companies. To measure firm size the logarithm of the total assets will be used like many papers come before us. Because in the Netherlands and Germany totals assets are reported in euros and in the UK the total assets are reported in pounds all currencies are converted to pounds based on the exchange rate on 31-12-2016 for 2016 and 31-12-2017 for 2017.

Legal system

The last control variable added to the model is legal systems (LGS). Because three different countries are included in the sample it is important to control for country differences. Legal systems affect the disclosures of firms and therefore also the work done by the auditor in timing and level. Ball, Kothari

(18)

and Robin (2000) state that disclosure standards in common-law countries reduce agency costs due to more transparency. In the sample, The Netherlands and Germany cope with civil law and in the UK common law is applied.

3. RESEARCH METHODOLOGY

This chapter describes how the research is established. First, a description of the sample will be given (3.1). After that, the statistical model is outlined (3.2). This part also contains a table and short description of all the variables included in de statistical model. In section 3.3 information will be given on the tests that are used to test the hypotheses in the research.

3.1 Sample description

The research method that is applied to answer the research question is quantitative research. Part of the data used for this study is already collected and comes from publicly available annual reports and audit reports. Our supervisor Dr. de Waard collected all the annual reports and audit reports of the firms that are included in the sample together with his team. Dr. de Waard and his team reported all the hand-collected data in a spreadsheet. This data contains information about KAMs, Total Assets and information about the ACs. The variables FOG-Index, Flesch reading ease score and auditor gender were hand collected by myself. The FOG-index and Flesch reading ease scores were determined by means of the website www.readabilityformulas.com. This website offers software to measure the readability of texts since 2003, first by selling software programs and now by a freely available website. For auditor gender, I observed the auditor reports which were already available. To determine the gender of the audit partner LinkedIn is used. If no LinkedIn profile was available to verify the gender of an auditor, company websites of the Big-4 firms are used to identify the gender. The SIC codes of the firms are derived from Orbis research database.

Velte (2017) mentioned in his research that besides the UK other countries should be included in future research. Therefore, the sample that will be subject to this research consists of Dutch companies which are all listed companies on the AEX and AMX, listed companies of Germany DAX 30 Stock Market and FTSE 250 listed companies of UK. Only listed companies are included in the sample, because non-listed companies are not obliged to report KAMs. By including 3 countries it is tried to chart differences in the readability of KAMs across these countries. In total 354 companies are included in the research. The data of the companies derives from the years 2016 and 2017. The sample includes 605 observations. The selection of firms is based on the at the beginning of 2015 market capitalization of these countries.

So, firms that were listed in the UK, Germany or The Netherlands on the first of January 2015 are represented in the sample that is used for this research. Using recent years, the research will be generalizable and valid for the present. Due to missing data, the final example consists of 605 firm-year

(19)

observations. The reasons for missing data vary from no reported KAMs, Dutch formulated KAMs and the absence of other important variables in the annual or audit reports.

Table 3. Sample

Number of organizations 354

Years included in sample 2

Number of observations 708

Observations with missing data in 2016 60

Observations with missing data in 2017 43

Final sample 605

In this paper is opted for countries that are on the G20. The UK is mentioned individually, and The Netherlands and Germany are mentioned as part of Europe on the G20. The G20 is a group of countries that cover together approximately 90% of the global gross national product, so it can be concluded that these countries are well developed and big economies (Oecd.org, n.d.). The UK companies differentiate from the West-European companies because they apply the one-tier board system. In the one-tier board systems the emphasis is on the shareholder, whilst the focus of two-tier boards is more towards the stakeholders of the company (Jungmann, 2006). Despite from that, not one of the two systems proved to be better what is understandable, because both systems exist along for more than 100 years. The Netherlands differs from the UK and Germany because the Netherlands score low on masculinity on the Hofstede’s scale. People in masculine countries tend to be more competitive and ambitious than feminine countries. Another characteristic of masculine countries is that there are few women in management positions (Hofstede, 2011). The differences between these countries could possibly expose different outcomes in the research.

(20)

3.2 Statistical model

Table 4. Definition of variables

Description Type of

variable

KAM_REA The readability score of the KAMs. Interval

AUGEN Auditor gender (0 = Man, 1 = Woman). Nominal

ACGEN Representativeness of female in AC (0 = No, 1 = Yes). Nominal

ACFREQ Meeting frequency of the audit committee. Interval

ACSIZE The number of members in the AC. Interval

ACCOM A computed combined variable of ACGEN, ACFREQ and ACSIZE. Interval INDj Kind of industry a firm is operating in measured by SIC. Firms

operating in an high-risk industry are given 1 (SIC codes 2833-2836, 3570-3577, 3600-3674. 5200-5961 or 7370-7374), and 0 otherwise.

Dummy

SIZEj Firm size measured by the natural logarithm of total assets. Interval

LGS Code law (0), Civil law (1). Dummy

In the table above the variables included in the statistical model are defined. This oversight can be used to understand the quantitative research done. The hypotheses of this quantitative research are tested by the following statistical models. The first equation will assess the effect of the auditor gender and the AC characteristics individual direct effect on KAMs readability including the control variables INDj,

SIZEj and LGS. In model 2, the combined computed variable of the AC characteristics (ACCOM) is

included as moderator including the control variables INDj, SIZEj and LGS.

(1) 𝐾𝐴𝑀𝑅𝐸𝐴 = β0 AUGEN + β1 ACGEN + β2 ACFREQ + β3 AC SIZE + β4 IN𝐷𝑗+ 𝛽5 𝑆𝐼𝑍𝐸𝑗+

𝛽6 𝐿𝐺𝑆 + 𝜀𝑖

(2) 𝐾𝐴𝑀𝑅𝐸𝐴 = β0 AUGEN + (β0 AUGEN) ∗ (β1 𝐴𝐶𝐶𝑂𝑀) + β5 IN𝐷𝑗+ 𝛽5 𝑆𝐼𝑍𝐸𝑗+ 𝛽6 𝐿𝐺𝑆 + 𝜀𝑖

The models will give insight into the effect of the several independent variables (auditor gender and AC characteristics) on the dependent variable readability of the KAMs. The standard error is included in the formula to prevent for coincidences and increase reliability.

(21)

3.3. Description of the tests

First of all, descriptive statistics are given on the variables in the research. Some additional information is given on the distribution of the sample which is important for the research. Also, the readability across countries is reviewed. Subsequently, in the correlation matrix is verified how the variables relate to each other.

For assessing the hypotheses in the paper two different tests are conducted. Hypotheses 1 and 2a are assessed by the Mann-Whitney U test, because the independent variables in these models are a nominal variable. Namely, the independent variable in hypothesis 1 is gender (0 = Man, 1 = Woman). The independent variable of hypothesis 2a is conducted as follows: 0 = no female representativeness in AC, 1 = female representativeness in AC. The relations of hypotheses 2b, 2c and 3 are tested with linear regression. All dependent variables in these hypotheses 2b, 2c and 3 are interval variables with a fixed zero point.

(22)

4. RESULTS

4.1 Descriptive statistics

Table 5 gives a representation of the descriptive statistics of all variables included in the research. The mean of the Flesch index is 25,42 (Std. Dev. = 6,26) which indicates that KAMs are very difficult to read. The maximum score in the sample is 50,1 this score corresponds to the score ‘fairly difficult’. The mean of the FOG index is 19,54 (Std. Dev.= 1,85) which indicates that KAMs are difficult to read. Remarkable is that the mean of the FOG index compared to the Flesch index is not categorized into the most difficult range. However, the scores align the expectations that auditor reports and components of auditor reports such as KAMs are difficult to read.

Female representatives in the AC range from 0%-75% with a mean of 30,47% (Std. Dev. = 19,72%). As presented in table 6 in the total sample about 11,4% of the observations a female auditor was responsible for the reporting of the KAMs. Besides, in 80,7% of the cases female are represented in the AC.

Further, the average AC consists of 4 people. Logically, the minimum number of persons in an AC is 1, the maximum number of people in the AC in the sample consists of 8. On average an AC meets 4,86 times a year. Only 2 ACs in the total sample do never meet, the maximum number of meetings is 14. As mentioned before, there are no requirements about how often AC should meet.

From the total assets, the natural log is token to control for the skewness of the extreme amounts. In the sample 66 firms operate in a high-risk industry. To control for the effect on the KAMs high-risk firms are given the dummy variable 1.

Table 5. Descriptive statistics

Dependent variables N Mean Min. Max.

Std. Deviation Flesch Index 605 25,42 8,7 50,1 6,26 FOG Index 605 19,54 14,2 25,0 1,85 Independent variables AUGEN 605 0,11 0 1 0,297 % male AC 605 69,53 25 100 19,74 % Female AC 605 30,47 0 75 19,74 FEMALE IN AC 605 0,81 0 1 0,40 Meetings AC 605 4,86 0 14 1,76 AC Size ACCOM 605 605 4,05 13,13 1 1,67 8 28 1,16 6,65 Control variables INDj 605 0,11 0 1 0,31

Log total assets 605 22,48 14,13 28,86 1,97

(23)

In table 6 some extra information is given about females that act as an audit partner and the number of ACs that contain at least one woman. Besides, also the number of firms in the sample that operate in high-risk industries is given.

4.1.1 Country differences

To determine if there are differences in readability between countries, the countries in the sample are divided (UK, The Netherlands and Germany). In table 7 the descriptives of the readability scores are presented. No significant differences are revealed in the variances of the readability scores across countries in the ANOVA test on the Flesch reading ease (Appendix 3). However, results show a significant effect in the variances among the FOG-index ( p < 0,1). Appendix 4 also declares that only 1% of the variance in means is a consequence of the countries.

Another note is that the variance between the minimum and maximum scores for the UK is larger than the differences between the minimum and maximum for The Netherlands and Germany. A reason for this could be a difference in sample size. The sample of the UK (N = 465) is more than 4 times bigger than The Netherlands (N = 101) and almost 12 times bigger than the German (N = 39) sample.

Table 7. Country differences

Germany N Mean Minimum Maximum Std. Deviation

Flesch Index 39 24,49 8,90 35,50 5,82 FOG Index 39 19,60 16,90 24,50 1,61 The Netherlands Flesch Index 101 24,92 10,50 40,90 6,95 FOG Index 101 19,14 14,30 23,50 1,95 UK Flesch Index 465 25,61 8,70 50,10 6,14 FOG Index 465 19,62 14,20 25,00 1,84

Table 6. Additional descriptives

Frequency

N Absolute Relative

Female audit partner 605 69 11,4%

Female representatives AC 605 488 80,7%

(24)

4.1.2 Correlation analysis

Subsequently, table 8 represents the correlations between the dependent, independent and

control variables in the research. The fixed assets are not included. The first and even the

strongest correlation presented in table 8 is the one between the FOG and Flesch index

(B = -0.876, p < 0.01). This negative relation is logical and explainable. FOG index will increase

if the readability decreases. However, the Flesch index decreases when readability increases.

Prior research show this correlation as well (Xu et al., 2019). Surprisingly, no correlation is found

between the readability scores and auditor gender. Also, no correlation is found between

women's representativeness in the AC and the readability scores.

AC size and AC meetings do present a correlation with the readability scores. If the size of the AC increases KAMS will become increasingly difficult (Flesch: B = -0.088, p < 0,05; FOG: B = 0.083, p < 0,05). Number of AC meetings reveal the same relation with the Flesch and FOG index ( B = -0.240, p < 0.01; B = 0.181, p < 0.01). These correlations are consistent with hypothesis 2b and 2c. Notable is that legal systems correlate to the FOG index (B = 0.084, p < 0.05) but no correlation is found according to the Flesch index. Further, the log of the total assets correlates with the number of meetings of the AC and the Size of the AC. What is in accordance with prior research (Menon and Williams, 1994; Raghundanan and Rama, 2007; Mendez and Garcia, 2007). Mendez and Garcia (2007) state that due to the fact that large companies tend to be more complex AC activities increase.

(25)

T ab le 8. P ea rso n c or re la ti o n s F le sc h Inde x F O G I nde x A U G E N % F em al e A C F E M A L E I N AC M ee ti ng s A C A C s iz e IN D j L og t ot al as se ts L G S 1. F le sc h Inde x 1 .00 2. F O G Inde x -. 876 ** * 1 .00 3. A U G E N .052 -. 045 1 .00 4. % F em al e AC .026 -. 014 .025 1 .00 5. F E M A L E IN A C .005 -. 013 -. 018 -. 757 ** * 1 .00 6. 6. M ee ti ng s AC -. 240 ** * .181 * ** -. 001 .038 -. 102 * 1. 00 7. A C s iz e -. 088 ** .083 ** .053 .044 -. 258 ** * .086 ** 1 .00 8. IN D j .024 .023 .008 .012 .003 .028 .027 1 .00 9. L og t ot al as se ts -. 290 ** * .182 * ** .001 -. 082 ** .012 .483 * ** .299 * ** -. 046 1 .00 10. L G S .050 .084 ** -. 025 .250 * ** -. 228 ** * -. 059 .100 * -. 072* -. 135 ** * 1 .00

(26)

4.2 Statistical analyses

4.2.1 Mann-Whitney U test (H1 and H2a)

In hypothesis 1 is stated that female auditors increase the readability of KAMs based on prior literature that stated that women are more risk-averse than men (Schubert, 2006) and that women have an advantage in communication abilities compared to men (Dwyer et al. 2002; Olsen and Cox, 2001). To test hypothesis 1, the Mann-Whitney U test is used. In this test, the difference in (man/woman) in a group on a dependent variable (Readability of KAMs) can be assessed.

First, some descriptives on the variables are given. The sample consists of 605 auditors, whereof 536 are men audit partners and 69 are women. The mean for men audit partners is 25,23, for women the mean is 26,29. Based on a 10% significance level, the Mann-Whitney U test in table 10 shows that KAMs of women's audit partners are easier to read (p < 0,1). However, the Z-score declares that the variability in the mean ranks is accounted for by only 0,5% on the independent variable.

ղ2 = 𝑍

2

𝑛 − 1=

−1,6842

605 − 1 = 0,00495

Further, the histograms in appendix 5 show a normal distribution which shows that the mean of the readability in the women sample is a bit more to the right. Although the KAMs of women are easier to read, both means fall into the category ‘very difficult to read’.

Table 9. Descriptive statistics auditor gender

Flesch Men N Valid 536 Missing 0 Mean 25,2996 Std. Error of Mean 0,2659 Median 25,7000 Std. Deviation 6,1623 Skewness 0,085 Std. Error of Skewness 0,105 Women N Valid 69 Missing 0 Mean 26,3275 Std. Error of Mean 0,8397 Median 27,2000 Std. Deviation 6,9752 Skewness -0,518 Std. Error of Skewness 0,289

(27)

Subsequently, the Mann-Whitney U test is also performed on the FOG-index. The mean for men audit partners is 19,58, for women the mean is 19,32. The FOG-index is contrary to the Flesch reading ease, because score 0 means easy to read and a score of 25 means very difficult to read. Remarkable is that both means are marked as ‘Difficult’ instead of the ‘very difficult’ classification under the Flesch reading ease formula. The FOG-index also shows that the KAMs of women's audit partners are easier to read, but the differences for the index are very small as well. However, the results of the FOG-index are not significant compared to the results of the Flesch reading ease (p > 0,1). Further, the histograms in appendix 6 show the normal distribution for women is more to the left.

Hypothesis 1 thus can be accepted for the Flesch reading ease formula (p < 0,1). However, Hypothesis 1 cannot be accepted for the FOG-index (p > 0,1).

Table 10. Test Statistics of Mann-Whitney U auditor gender Flesch reading ease

Mann-Whitney U 16190,000

Wilcoxon W 160106,000

Z -1,684

Asymp. Sig. (2-tailed) 0,092*

a. Grouping Variable: AUGEN

***, **, * coefficients are respectively statistically significant at 1, 5 and 10 percent level

Table 11. Descriptive statistics auditor gender

FOG Men N Valid 536 Missing 0 Mean 19,5644 Std. Error of Mean 0,07782 Median 19,5000 Std. Deviation 1,8034 Skewness 0,028 Std. Error of Skewness 0,105 Women N Valid 69 Missing 0 Mean 19,3145 Std. Error of Mean 0,2617 Median 19,1000 Std. Deviation 2,1742 Skewness 0,377 Std. Error of Skewness 0,289

(28)

To assess the effect of female representativeness in AC on KAMs readability the Mann-Whitney U test is here applied as well. Female representativeness (0: No, 1: Yes) on KAMs readability’s Flesch and FOG (scores) is the same equation done for hypothesis 1. In 488 ACs of the total sample of 605, females are represented. The mean of the Flesch reading ease for ACs where no females are represented is 25,4197. For ACs where females are represented the mean is 25,3293. So, the descriptives show almost no differences between ACs were females are not represented and represented.

In table 14 is stated that the effect of female representativeness of Flesch reading ease is not significant (p > 0,1), based on this information we can reject H2a. The Mann-Whitney U test is also conducted for the FOG-index. No differences in outcomes for no female and female representativeness on ACs are found (Mean without female: 19,4933 – Mean with female: 19,5677). The test results show no significance (p > 0,1) for the effect of female representativeness on the FOG-index.

Table 12. Test statistics Mann-Whitney U auditor gender

FOG

Z -1,309

Asymp. Sig. (2-tailed) 0,190

a. Grouping Variable: AUGEN

Table 13. Descriptive statistics Flesch No female in AC N Valid 117 Missing 0 Mean 25,4197 Std. Error of Mean 0,61176 Median 26,4000 Std. Deviation 6,61724 Skewness -0,233 Std. Error of Skewness 0,224 Female in AC N Valid 488 Missing 0 Mean 25,3293 Std. Error of Mean 0,28421 Median 25,7000 Std. Deviation 6,27838 Skewness -0,073 Std. Error of Skewness 0,111

(29)

Table 14. Test statistics Mann-Whitney U Female representativeness

Flesch

Z -0,391

a. Grouping Variable: FEMALE IN AC

Table 15. Descriptive statistics on female representativeness in AC

FOG No female in AC N Valid 117 Missing 0 Mean 19,4933 Std. Error of Mean 0,18490 Median 19,3000 Std. Deviation 1,99996 Skewness 0,207 Std. Error of Skewness 0,224 Female in AC N Valid 488 Missing 0 Mean 19,5677 Std. Error of Mean 0,08438 Median 19,5000 Std. Deviation 1,86391 Skewness 0,181 Std. Error of Skewness 0,111

Table 16. Test statistics Mann-Whitney U Female representativeness

FOG

Z -0,380

(30)

4.2.2 Linear regression (H2b and H2c)

To test hypotheses H2b and H2c linear regression is applied. Hypothesis 2b assessed the relation between meeting frequency and KAMs readability on the Flesch reading ease score. To eliminate the effect of the variables INDj, SIZEj and LGS these variables are included as control variables in the model. As stated in table 8 (Pearson correlations), SIZEj correlates on an 1% significance level with the Flesch reading ease and FOG-index. This relation emerges again in table 17. Hypothesis 2b is significant on a 1% level, the meeting frequency of AC thus have an effect on the readability of KAMs according to the Flesch reading ease formula. This effect is negative according to table 17.

Table. 18 Model Summary linear regression (Flesch)

Change Statistics Model R R Square Adjusted R Square Std. Error of the Estimate R Square Change F Change Sig. F Change 1. ,291a 0,085 0,080 6,08043 0,085 18,544 0,000 2. ,313b 0,098 0,092 6,04182 0,013 8,706 0,003

a. Predictors: (Constant), Law, INDDUMMY, Log total assets

b. Predictors: (Constant), Law, INDDUMMY, Log total assets, Meetings AC

Table 17. Results linear regression analyses (Flesch)

Model 1 Model 2 (H2b) Model 3 (H2c) Constant 46.126 _43.770 46.117 Independent variable Meeting frequency _-.472*** AC Size -.017 Control variables INDj .257 .391 .260 SIZEj .931*** _.725*** .928*** LGS .593 .590 .165 F-value 18.544 16.263 13.886 R-squared 0.085 0.098 0.085 Adj R-squared 0.080 0.092 0.079

(31)

In table 18 is the effect of meeting frequency of the AC declared. Model 1 presents the effect of the control variables. The control variables declare 8,5% (p < 0,1) of the effect on the readability according to the Flesch reading ease. Meeting frequency declares only 1,3% (p < 0,1) of the effect on KAMs readability. An increase in the number of AC members does lead to a lower Flesch reading ease score. This means readability will decrease.

To assess the effect of Meeting frequency and AC Size on the FOG-index, the linear variable is conducted on the FOG-index measure as well. The linear regression on the FOG-index presents the opposite result of the regression of the Flesch. This is in line with the expectations because the FOG and Flesch have contrary scores. What indicates that the frequency of the AC has a decreasing effect on the readability according to the FOG-index (B = .137, p < 0,01) and also a decreasing effect on the readability according to the Flesch reading ease (B = -.472, p < 0,01).

Table 19. Results linear regression analyses (FOG)

Model 1 Model 2 (H2b) Model 3 (H2c)

Constant 14,802 15.442 14.814 Independent variable Meeting frequency .128*** AC Size .021 Control variables INDj .251 .241 .247 SIZEj .193*** .137*** .189*** LGS .508*** .502*** .499*** F-value 9.928 9.251 7.459 R-squared .047 .058 0.085 Adj R-squared .042 .052 0.079

(32)

Table 20 reveals that the meeting frequency of the AC declares only 1,1 % of the model (p < 0,01).

Table. 20 Model Summary linear regression (FOG)

Change Statistics Model R R Square Adjusted R Square Std. Error of the Estimate R Square Change F Change Sig. F Change 1 ,217a 0,047 0,042 1,84895 0,047 9,928 0,000 2 ,241b 0,058 0,052 1,83990 0,011 6,925 0,009

a. Predictors: (Constant), Law, INDDUMMY, Log total assets

b. Predictors: (Constant), Law, INDDUMMY, Log total assets, Meetings AC

4.2.3 Linear regression on hypothesis 3

To test the moderating effect of AC characteristics on the relation between auditor gender and KAMs several adjustments have to be made to the data. First, a new variable is computed for the AC characteristics. The variables Female in AC%, AC meeting frequency, and AC Size are combined to the variable AC characteristics. Subsequently, the independent variable auditor gender is centralized. The moderating variable is calculated as the product of the independent variable and the new variable AC characteristics. All the variables (independent, dependent, moderator and control variables) are included in the linear regression.

As mentioned in table 21, no significant effect of the moderator variable is detected in the linear regression. Based on this information, hypothesis 3 can be rejected. AC characteristics do not have an effect on the relation between auditor gender and the readability of the KAMs.

Table 21. Results linear regression analyses (Flesch)

Model 1 Model 4 (H3) Constant 46.126 46.064 Independent variable Moderator -.004 Auditor gender 1.132 AC Characteristics -.011 Control variables INDj .257 .257 SIZEj -.931*** -.930*** LGS .158 .220 F-value 18.544 9.585 R-squared 0.085 0.088 Adj R-squared 0.80 0.079

(33)

In table 22 the same test as in table 21 is conducted, here the effect of the moderator is tested on the FOG-index. In this test also no significant outcomes of the moderator variable are found. The significant effect of the control variables of SIZEj and LGS corresponds with the outcomes in table 19.

Table 22. Results linear regression analyses (FOG)

Model 1 Model 4 (H3) Constant 14.802 14.866 Independent variable Moderator -,002 Auditor gender -.213 AC Characteristics -.004 Control variables INDj .251 .255 SIZEj .193*** .193*** LGS .508*** .520*** F-value 9.928 5.156 R-squared 0.085 0.088 Adj R-squared 0.80 0.079

The effect of auditor gender and the moderating effect of audit committee characteristics on the readability of KAMs