Me, myself and I: the effect of CEO narcissism on audit pricing and the moderating role of gender diversity

(1)

Me, myself and I: the effect of CEO

narcissism on audit pricing and the

moderating role of gender diversity

Combined master thesis University of Groningen Faculty of Economics and Business

MSc Accountancy and Controlling

Student name: S.A. ten Boom Student number: 3271021

Supervisor: dr. Y. Karaibrahimoglu Assessor: prof. dr. J.A. Emanuels

Co-assessor: G.C. Helminck, RA MSc EMA Date: June, 25, 2018

(2)

2 Abstract

Prior studies have focused on outcomes related to the CEO’s organization, this research investigates how external auditors respond to organizations with a narcissistic CEO. Specifically, this research examines whether CEO narcissism is positively related to audit pricing in the two-tier boards’ structure. Moreover, the underlying mechanism of this relationship is studied, including the indirect effect of CEO narcissism on audit pricing, through audit risk and the moderation of the indirect effect by gender diversity in the board of directors. This research draws on quantitative research using panel data with 251 firm year observations of Dutch listed organizations over the period 2013-2016. The findings indicate a positive relationship between CEO narcissism and audit pricing. However, audit risk does not significantly mediate this relationship. The results suggest that the indirect effect of CEO narcissism on audit pricing, through audit risk is significantly moderated at moderate and high levels of gender diversity in the board of directors. However, at low levels of gender diversity in the board of directors there is no significant indirect effect and the variable gender diversity in the board of directors is not significantly related to audit risk. For the audit literature, this research provides evidence that also the CEO’s personality aspect, the narcissistic trait, affects the audit pricing. The outcomes of this research should be taken into account with the

mandatory trainings of Accounting Associations for external auditors. These trainings should specifically highlight the behavioural aspects, like CEO narcissism, and their influence on the audit pricing.

Keywords: audit pricing, audit risk, CEO narcissism, gender diversity, board of directors, two-tier board’s structure

Acknowledgement: I would like to thank my supervisors dr. Y. Karaibrahimoglu, prof. dr. J.A. Emanuels, and G.C. Helminck RA MSc EMA for all your support regarding the development of the thesis.

(3)

3 Table of contents

1. Introduction ... 4

2. Theoretical background... 8

2.1 Theory ... 8

2.2 Literature review and hypothesis development ... 10

3. Methodology ... 15

3.1 Measures ... 15

3.2 Data collection and analysis procedure ... 17

3.3 Reliability and validity ... 18

3.4 Specifications of the tests ... 19

4. Results ... 22

4.1 Descriptive statistics and correlations ... 23

4.2 Results from the analyses ... 24

4.3 Robustness tests ... 28

5 Conclusions and discussion ... 32

5.1 Conclusions ... 32

5.2 Implications and limitations ... 33

5.3 Future research ... 35

References ... 36

(4)

4 1. Introduction

Nowadays, many headlines and other items read with the titles, “beware of generation

narcissism” (Derksen, 2016; van der Veen, 2017), “the narcissism epidemic – living in the age of entitlement” (Twenge & Campbell, 2010) or “Me! Me! Me! Are we living through a

narcissism epidemic?” (Williams, 2016). A commonly wide used definition of narcissism is “a pervasive pattern of grandiosity where self-admiration and self-importance with no

empathy defines an individual” (American Psychiatric Association, Task Force on DSM- IV., 2000, p. 717). There is also narcissism in business, an example of this is the Volkswagen scandal. Volkswagen made a commitment to “partnership-based conduct”, “sustainability”, and “social responsibility” (The Volkswagen Group, 2015). However, the diesel emissions scandal seriously conflicted with these values. The failure of tone at the top to execute what they preach leads to cynical employees, harmful corporate climates, and cultures that enable narcissistic behaviours (Crête, 2016). Another example is Uber, the founder, Travis Kalanick, stepped down as chief executive officer (CEO) of Uber in 2017. A female engineer blamed the organization of “ignoring her complaints of pervasive harassment, promoting less qualified men over her, retaliating against her for raising concerns, and paying her a lower salary than men doing the same work” (Brown, 2017). This is exemplary for organizations run by narcissistic leaders. The workplace culture reflects the values and behaviour of the narcissistic leaders (Brown, 2017).

The CEO plays a central role within the top executive team and even exercises greater and greater influence over their organization’s performance and actions according to evidence of existing empirical research (Quigley & Hambrick, 2015). CEOs can sometimes have

dominating and out of balance influences. The extent of the CEO influence within the top executive team is determined by personal characteristics of CEOs (Chatterjee & Hambrick, 2007, 2011; Hiller & Hambrick, 2005). Drawing on research indicating that narcissism is a personality dimension, rather than a pathological disorder (Judge et al., 2006; O’Reilly et al., 2014). Narcissism is an essential requirement for effective leadership due to productive influences, but the other side of the coin is that narcissism forms a potential threat (Bushman & Baumeister, 1998; Rhodewalt & Morf, 1998).

Prior studies have focused on outcomes related to the CEO’s organization. These studies show that narcissistic CEOs are more likely to engage in risky business practices and demonstrate a lack of respect for compliance with regulations and rules (Amernic & Craig,

(5)

5 2010; Johnson et al., 2013). In other words, narcissistic CEOs seem to exhibit greater control and inherent risk. Nevertheless, how does an essential counterparty, the organization’s external auditor, responds to an organization having a more narcissistic CEO? It can be argued that this characteristic poses greater audit risk to an organization’s auditor. Various studies theorize that the audit risk affects the audit price and that higher risk is associated with higher audit pricing (Hay et al. 2006; Venkataraman et al., 2008). The audit risk model

(Srivastava & Shafer, 1992) contains the equation of: audit risk = inherent risk x control risk x detection risk. This model can be used to determine the audit risk and, in this way, explains the audit pricing. External auditors assess organizations with poor internal governance as having higher levels of control and inherent risk. In this way, auditors need to perform more audit work and time, bear greater audit risk, and price higher audit accordingly (Carcello et al., 2002). When an external auditor is faced with the heightened risk of CEO narcissism, it is expected that external auditors modify their fee arrangement to maintain an acceptable level of audit risk.

A study of Judd et al. (2017) has examined the impact of CEO narcissism on audit pricing worldwide. The board of directors operates in a variety of systems to monitor management because organizations operate in different business contexts. Two categories of board

structures are prevailing: the one-tier board and the two-tier board. One-tier systems are more presumably to be found in an Anglo-Saxon context, while two-tier systems are regular in continental Europe (Adams & Fereira, 2007). A single board system is the one-tier board, consists of both non-executive directors and executive directors (Solomon, 2013). The two-tier boards’ structure is characterized by a formal separation between non-executive directors and executive directors, that both acts separately (Bezemer et al., 2014). This might result in less aggressive performance targets and more monitoring the narcissistic CEO (Carrasco, 2005). The question then arises; do narcissistic CEOs still have an impact on the audit pricing even with a high monitoring quality? It can be argued that there a research gap is between the worldwide results of Judd et al. (2017) and the two-tier boards’ structure. For this reason, a sample of listed firms will be used from the Netherlands, where people are used to a two-tier boards’ structure.

The recent public debate about improving the gender balance and increasing female board participation motivates this research to include gender diversity in the board of directors as moderator on the relationship between CEO narcissism and audit pricing (Dekker, 2018). The

(6)

6 board of directors has an important role in monitoring and to deter CEOs from opportunistic behaviour (Campbell & Minguez-Vera, 2007; Rose, 2007), and thus excessive level of audit risk. The literature shows that gender diversity affects the quality of decisions and the monitoring role (Adams & Ferreira, 2007; Lakhal et al., 2015). Female board members are related to behaviour that is more ethical and socially desirable actions compared to male board members (Barua et al., 2010; Peni & Vähämaa, 2010). For this reason, gender-diverse boards presumably have a much higher level of monitoring a narcissistic CEO. However, female board members are still belonging to a minority in the board of directors and is continued as a worldwide concern (Arun et al., 2015). For example, the percentage of female non-executive directors has increased in the Netherlands (Egon Zehnder, 2016; Hendrikse & Pouwels 2016). Nevertheless, females are still underrepresented in the board of directors (European commission, 2016) and there is a long way to go in moving from an average of 10.2% on 31 May 2016 to the target of 30% (Seierstad, Gabaldon, & Mensi-Klarbach, 2017).

The effect of gender diversity in the board of directors has received little attention from finance and accounting researchers. This is remarkable, considering that gender diversity can be an important determinant of the level of audit risk. Particularly, gender diversity advances a heterogeneous and broader way of thinking with respect to the decision-making process within the board of directors. Therefore, the purpose of this study is to discover to what extend CEO narcissism has an impact on audit pricing in the two-tier boards’ structure. Moreover, the underlying mechanism of this relationship will be studied, including the indirect effect of CEO narcissism on audit pricing, through audit risk and the moderation of the indirect effect by gender diversity in the board of directors. This leads to the following research question of this study:

“To what extent is CEO narcissism related to audit pricing in the two-tier boards’ structure and to what extent is this relationship mediated by audit risk and does gender diversity in the board of directors moderates this relationship through the mediator audit risk?”

This research will make several contributions. Firstly, this research will give a better understanding of the relationship between CEO narcissism and audit pricing in a European context. There is not much research available about an essential counterparty, the

(7)

7 Therefore, this research will contribute to the executive personality characteristics in an audit setting. A study of Judd et al. (2017) is closely related to the research field of this study. However, the composite measure of this research of CEO narcissism did not take into account the CEO’s use of first-person singular pronouns in interviews (Chatterjee & Hambrick, 2007). Moreover, the research of Judd et al. (2017) did not control for one-tier or two-tier board system. Therefore, this research will complement the worldwide results of Judd et al. (2017) by using a sample of listed firms in the Netherlands, where people are used to supervise executive directors in the two-tier board system. Secondly, the proposed research is expected to deliver a valuable contribution to the existing empirical research and literature.

Specifically, this research will contribute to the audit literature by examining determinants of external audit fees (Hay et al., 2006). Thirdly, this research will fill up the gap about a

possible influence of gender diversity in the board of directors on the relationship between CEO narcissism and audit pricing. Recent research (Kalelkar & Khan, 2016) examines the effect of CEO functional background on the level of audit pricing. The influence of gender diversity in the board of directors through audit risk on the relationship between CEO narcissism and audit pricing has not been examined. Finally, practical and societal contributions will be made because regulators and policymakers may be interested in the outcomes of this research. An understanding of the possible positive influence of gender diversity in the board of directors may be helpful by implementing or adjusting quotas and laws concerning the presence of female non-executive directors. Furthermore, the outcomes of this study may influence external auditor’s training by adapting the formulated objectives and the content to behavioural aspects, like CEO narcissism, and their influence on the audit risk (Simon, 2012). The objective of the mandatory training is to educate external auditors in such a way that they can better identify risks and act on it (NBA, 2017). However, both the formulated objectives and the content of the training do not specifically highlight the behavioural aspects, like CEO narcissism, and their influence on the audit risk and audit pricing.

This study is organized as follows: first, the theoretical background is presented on the subject before further elaboration. Next, the methodology is described in chapter three. Subsequently, the results of the research will be discussed in chapter four. Finally, the consequences of the results will be discussed and formulated in conclusions and suggestions for further research.

(8)

8 2. Theoretical background

In this study, the influence of CEO narcissism on audit pricing in the two-tier boards’ structure will be examined and to what extent this relation is mediated by audit risk and the moderating effect of board gender diversity through the mediator audit risk on this

relationship. This chapter will discuss the relevant literature about CEO narcissism, audit pricing, audit risk, and gender diversity in the board of directors. The hypotheses are developed based on the theories. At the end of this chapter, the conceptual model will visualize the expectations with regard to this study.

2.1 Theory

This research is rooted in the upper echelon theory, agency theory, and resource dependence theory. The first section introduces and explains the aforementioned theories.

2.1.1 Upper echelon theory

The upper echelon theory can be seen as the traditional theory that is used for research on CEO narcissism and their influences on organizational level outcomes. Hambrick and Mason published this management theory in 1984. The influence that characteristics of executive personalities can have on organizational level outcomes is grounded in the upper echelon theory and the proposition is that “cognitions, perceptions, and values of executives manifest themselves in the decision-making process”. Consequently, the (strategic) choices in

organizations are reflections of values and cognitive bases of executives, and in this way the organization is a reflection of it’s executive(s) (Hambrick & Mason, 1984). For the reason that CEOs operating in comparable situations act variously (Ham et al., 2017). Therefore, the upper echelons theory argues that all organizational choices and actions of organization’s CEOs are a reflection of their characteristics (Judd et al., 2017)

The CEO’s personality is reflected by “the tone at the top” and influences the external auditors’ determination of the organization’s control and inherent risks, as a result of its prevalent impact on the organization’s financial reporting and auditors’ ability to rely on the organization’s internal controls (Cohen, Krishnamoorthy, & Wright, 2002; Hammersley, 2011; COSO, 2013; Schmidt, 2014; Patelli & Pedrini, 2015).

(9)

9 2.1.2 Agency theory

Agency theory explains the conflict between the manager (agent) and shareholders

(principals). This theory is based on two assumptions of goal incongruence and asymmetric information between the agent and the principals (Jensen & Meckling, 1976). It is assumed that the agent act in self-interest contrary to the principals’ interest. This assumed self-interest of an agent leads to information asymmetry. The principals are not completely aware of what the agent has done. A narcissistic CEO acts in self-interest, and possibly misaligns with the interest of the principals. CEO narcissism can thus increase agency problems.

The external audit can be explained as a function, to monitor the behaviour of the agent. Research argues that the external audit function has a corporate governance role in reducing agency costs by mitigating agency conflicts among stakeholders (Jensen & Meckling, 1976; Watts & Zimmerman, 1983; Fan & Wong, 2005). For this reason, the agency theory is applicable for this research. The external auditor will reduce the audit risk to an acceptable level by conducting an initial risk assessment, and perform procedures based on that

assessment. The external auditor has the obligation to act in the public interest. The ultimate objective of the external auditor is to issue an opinion on the fairness and give reasonable assurance about the organization’s financial statements (ISA 200).

A form of internal governance is the board of directors. The board of directors wants to monitor the actions of the agent in order to help mitigate the agency conflicts. The benefits of gender diversity in the board of directors are supported by the agency theory. Gender diversity in the board of directors allows for greater monitoring actions that align agents’ interests with those of the principals to improve organizational outcomes (Adams & Ferreira, 2009). The gender diversity in the board of directors increases board independence according to Carter et al. (2003). Carter et al. (2003) assume that a more diverse board can be regarded as a better controlling mechanism, which helps to make decisions that are effective and thus lower audit risk (Dalton et al., 1999; Kiel & Nicholson, 2003).

(10)

10 2.1.3 Resource dependence theory

The resource dependence theory provides a theoretical foundation for the role of the board of directors as a resource to an organization (Johnson et al., 1996; Hillman et al., 2000). A board of directors exists as a provider of resources to executive directors so that non-executive directors help them to achieve organizational goals (Hillman et al., 2000; Hillman & Daziel, 2003). In other words, the board of directors is considered as a link between the organization and the external resources that an organization needs from the external environment for superior performance. The basis for theoretical arguments for diversity in the board of directors originates from the resource dependence theory. Diversity contributes to the

potential to improve the information provided by the board of directors because of the unique information held by diverse non-executive directors. Hillman et al. (2000) argue that female non-executive directors bring different benefits and resources. Therefore, gender diversity may bring diverse perspectives and non-traditional approaches to problems.

2.2 Literature review and hypothesis development

In the next part, the main concepts of the theoretical background will be defined and the underlying arguments for the hypotheses will be introduced.

2.2.1 CEO narcissism

In the 19th century, the term narcissism is originated and used by Havelock Ellis (Ellis, 1898) to explain a clinical condition of perverse self-love. This term influences the work of Freud (1914) and he identified several specifications of narcissism. These specifications include a tendency to see others as an extension of one’s self and self-admiration. The history of the word narcissism comes from the unfeeling, proud, and beautiful Greek Narcissus who declined the love of a nymph and has been punished with self-love (Apeldoorn & Beijer, 1996). Narcissism reflects the personality characteristic of self-love that includes a set of characteristics like self-confidence, egoism, vanity, hubris, dominance, lack of empathy, and self-esteem. Narcissists usually lack self-confidence and self-esteem and try to compensate these deficiencies by constantly looking for affirmation and presenting themselves as being more important than others. In consideration of protecting themselves from being criticized by others (Freud, 1914). The outcome of Ham et al. (2017) their study was that narcissism is related to unethical behaviour. The tendency of a narcissistic CEO is to set their personal and

(11)

11 organizational goals impossibly high so that affirmation can be achieved. In order to obtain these goals, they are willing to behave unethically (Duchon & Drake, 2009).

The commonly wide used definition of Diagnostic and Statistical Manual of Mental Disorders (DSM) will be used to define narcissism, “a pervasive pattern of grandiosity where self-admiration and self-importance with no empathy defines an individual” (American

Psychiatric Association, Task Force on DSM- IV., 2000, p. 717). The Narcissistic Personality Inventory (NPI) is frequently used to measure narcissism (Raskin & Terry, 1988). However, it is difficult to obtain access to CEOs willing to complete self-reported psychological measures. Although, inconspicuous measures have been shown to be credible alternatives to

self-reported measures (Webb & Weick, 1979). For this reason, CEO narcissism will be measured by following a measure employed in prior accounting literature (Olsen et al., 2014; Olsen & Stekelberg, 2016).

2.2.2 Audit pricing

Both the demand side (audit risk) and supply side (audit market concentration) determine audit pricing. Prior research indicates audit risk as the main contributor of audit pricing (Elliott et al., 2008; Hogan & Wilkins, 2008; Morgan & Stocken, 1998; Zhu & Sun, 2012). During the audit planning process, external auditors must make risk assessments of their clients (Auditing standard Nos. 8, 12, and 13). The audit risk model (Srivastava & Shafer, 1992) supports these assessments and contains the equation of: audit risk = inherent risk x control risk x detection risk. Whereby, International Standard on Auditing (ISA) 200 formulates audit risk as “the risk that an auditor expresses an inappropriate opinion on the financial statements” and detection risk is “the risk that the auditors fail to detect a material misstatement in the financial statements” (ISA 200). Particularly, external auditors assess an organization’s control and inherent risks to determine the level of detection risk needed to accomplish an acceptable level of audit risk (Hogan & Wilkins, 2008). The external auditor reduces detection risk in case of a high level of control and inherent risk. Consequently, the external auditor will accomplish additional substantive testing, that brings about an increase in audit pricing. It can be argued that, in comparison with other organizations, those with

narcissistic CEOs pose greater audit risk that could in turn affect audit pricing. The following hypothesis can be stated, taking the above-mentioned arguments in account.

(12)

12 2.2.2.1 Inherent risk

Inherent risk is “the risk of a material misstatement in the financial statements arising due omission or error as a result of factors other than the failure of controls” (ISA 200). Narcissistic CEOs are expected to pose greater inherent risks as a result of heightened financial reporting and business risk. The common deﬁnition of business risk is “the probability that an external auditor will suffer a loss (e.g. litigation, sanctions imposed by regulatory agencies, impaired reputation, etc.) because of a client relationship” (ISA 200). According to Chatterjee and Hambrick (2007) are narcissistic CEOs presumably more than other CEOs be engaged in bold strategic actions, such as frequently acquisitions. These actions contribute to variability of performance (Chatterjee & Hambrick, 2007). Furthermore, they are presumably more engaged in operational activities to improve short-term earnings, regardless of the possible long-term negative consequences to the organization (Olsen et al., 2014). According to Olsen and Stekelberg (2016) are narcissistic CEOs more likely to occupy with questionable business practices, such as participation in corporate tax shelters and

fraudulent activities (Rijsenbilt & Commandeur, 2013). In other words, narcissistic CEOs take on greater business risks and make operational decisions to enhance perceptions of themselves.

2.2.2.2 Control risk

Control risk is “the risk of a material misstatement in the financial statements arising due to absence or failure in the operation of relevant controls of the organization” (ISA 200). Narcissistic CEOs usually exhibit desirable leadership characteristics, such as confidence, charisma, and drive. However, they also exhibit conceit and exploitativeness, more disruptive characteristics (Resick et al., 2009). Narcissistic CEOs are characterized by a feeling that they are above the law. For this reason, they demonstrate a lack of respect for compliance with regulations and rules (Twenge & Campbell, 2003). Besides, narcissistic CEOs use their organizations as vehicles to achieve grandiose and self-serving aims while their organization exhibit poor organizational decision making (Conger, 2002; Amernic & Craig, 2010; Craig & Amernic, 2011). Moreover, narcissistic CEOs are presumably less sensitive to penalties and more motivated by rewards (Foster et al., 2009). These characteristics are difficult considering that CEOs are legally mandated to present their organization’s financial statements fairly, per Section 302 of the Sarbanes-Oxley Act of 2002. Organizations with narcissistic CEOs

(13)

13 2.2.2.3 Audit risk

The audit risk can explain the relationship between audit pricing and CEO narcissism by using the audit risk model (Srivastava & Shafer, 1992). Narcissistic CEOs seem to exhibit greater control and inherent risks that may have an impact on the professional services of the external auditor. When an external auditor is faced with the heightened risk of CEO

narcissism, it is expected that auditors modify the extent, nature, and timing of planned substantive procedures to maintain an acceptable level of audit risk. For this reason, it is expected that external auditors modify their audit pricing to maintain an acceptable level of audit risk (Hay et al., 2006; Hogan & Wilkins, 2008). The following hypothesis can be stated, taking the above-mentioned arguments in account.

Hypothesis 2: The relationship between CEO narcissism and audit pricing is positively mediated by audit risk.

2.2.3 Gender diversity in the board of directors

Fama and Jensen (1983) describe the function of the board of directors from the agency theory perspective, it is a structure that monitors the executive directors of the organization, in such a way that the executive directors make decisions according to the preferences of the principals. In other words, an internal device to monitor the executive directors. Non-executive directors are better at monitoring the organization because they have no financial interest other than improving their own image on the job market (Boo & Sharma, 2008). In contrast, executive directors are remunerated based on their performance and therefore they have incentives to manipulate earnings and participation in fraudulent activities (Olsen et al., 2014).

In the Netherlands, non-executive directors are used to supervise executive directors in the two-tier board system (Adams & Ferreira, 2007). Since January 1st 2012, it is allowed to implement a one-tier board structure, but most organizations continue to employ the two-tier approach.

According to Lakhal et al. (2015) is the decision-making process by executive directors influenced by the structure of the board of directors. The extent of female representation on the board of directors is a form of structuring. The resource dependence theory argues that it is expected that diversity in the board of directors may affect the effectiveness of the board (Adams & Ferreira, 2007; Coffey & Wang, 1998; Lakhal et al., 2015). For the reason that

(14)

14 male and female board members have a different way of working and critical thinking in the organization. Prior studies have shown that a more diverse board of directors will decrease conflicts of interest between agents and principals because it will exclude that an individual or group dominate the decision-making process (Lakhal et al., 2015). However, Mosakowski (2000) suggests that diversity in the board of directors will create more critical questions and opinions, which may lead to less effective decision-making and more conflicts. Nevertheless, Lakhal et al. (2015) suggest that board of directors with female non-executive directors is related to greater monitoring and supervision actions that will reduce agency costs and

aligning agent’s interest with those of the principals. In other words, it is expected that gender diversity in the board of directors contribute to lower audit risk, so lower level of audit

pricing. Specifically, the board of directors will weaken the positive relationship between CEO narcissism and audit pricing through the mediator audit risk. The following hypotheses can be stated, taking the above-mentioned arguments in account.

Hypothesis 3: The indirect effect of CEO narcissism on audit pricing, through audit risk is negatively moderated by gender diversity in the board of directors.

Hypothesis 4: Gender diversity is negatively related to audit risk.

The conceptual model visualizes the relationships between the dependent variable (audit pricing) and the independent variable (CEO narcissism), and how they relate to each other. A mediator variable (audit risk) is included, to explain the relationship between audit pricing and CEO narcissism. Figure 1 shows that a moderated mediation effect is expected between CEO narcissism and audit pricing.

CEO narcissism

Gender diversity

Audit risk Audit pricing

H1 (+)

H2 (+) H2 (+)

H3 (-)

H4 (-)

(15)

15 3. Methodology

In the previous chapter, the literature is reviewed concerning CEO narcissism, audit pricing, audit risk, and gender diversity in the board of directors. This chapter will explain how the data of this study will be obtained. The measures and data collection procedure will be elucidated in sections 3.1 and 3.2. Furthermore, section 3.3 discusses the reliability and validity of this study. This chapter concludes with the specification of the tests in section 3.4.

3.1 Measures

As stated before, this study focuses on the research question: “To what extent is CEO narcissism related to audit pricing in the two-tier boards’ structure and to what extent is this relationship mediated by audit risk and does gender diversity in the board of directors moderates this relationship through the mediator audit risk?”. This section describes the dependent, independent, mediator, moderator, and control variables that will be used in this research and how these variables will be measured.

3.1.1 Dependent variable: external audit fees

The dependent variable of this study is external audit fees, which will be used as a

measurement of audit pricing. This variable can be described as the audit fees charged by the auditor to the client and will be measured as the total fees for audit of financial statements in the previous year. The external audit fees variable is stated as LNAUDIT and is the natural log of audit fees.

3.1.2 Independent variable: CEO narcissism

In this study, the independent variable is CEO narcissism. This variable will be measured by following a measure employed in prior accounting literature (Olsen et al., 2014; Olsen & Stekelberg, 2016) and based on two factors:

(1) The prominence of the CEO’s photograph in the organization’s annual report

Chatterjee and Hambrick (2007) mention that a CEO has high control and strong opinions about how they are portrayed in the organization’s annual report. Moreover, the CEO’s photograph can be used to showcase himself or herself, capture the CEO’s self-admiration and vanity. For this reason, it is expected that a CEO high in narcissism has a high visibility within the annual report.

(16)

16 Following the method used by Olsen et al. (2014), the photograph prominence is coded from one to five:

• 1 point, no CEO photograph

• 2 points, a photograph of CEO pictured with other executives • 3 points, CEO pictured alone and occupying less than half the page • 4 points, CEO pictured alone and occupying more than half the page

• 5 points, CEO pictured alone and occupying the full page in the annual report (2) The CEO’s use of first-person singular pronouns in interviews

According to Chatterjee and Hambrick (2007), first-person singular pronouns (I and my) reflect self-absorption and exploitativeness by catching how the narcissistic CEO takes credit for what might not be entirely a personal achievement. For this reason, it is an indicator of narcissism and will be measured by the number of first-person singular pronouns used by a CEO, divided by the sum of pronouns plus all first-person plural pronouns (we and our). Digital transcripts of interviews of CEOs are used to isolate direct quotes. This approach is supported by the work of Raskin and Shaw (1988) in which they show that the first-person singular pronouns that are used by a CEO are significantly correlated with the NPI scores.

Chatterjee and Hambrick (2007) include three additional factors in their CEO narcissism measure. One of these factors is the CEO’s prominence in the organization’s press releases. However, data limitations restrict access to the content of press release data in a readable format. For this reason, the necessary press release information cannot be gathered. Besides, the CEO’s relative cash and non-cash compensations are missing due to data restrictions. It can be argued that the two-factors included capturing traits showed by narcissists such as vanity, exploitativeness, and self-admiration. Following prior literature, the two factors of the narcissism measure over the CEO are averaged (NARCISSISM).

3.1.3 Mediator: key audit matters

The mediator of this research is audit risk. The proxy that will be used is the number of key audit matters in the organization’s annual report. These key audit matters are informative because they are the specific issues of an organization which the external auditor regards important when auditing the financial statements. The mediator variable is stated as AUDR and is the natural log of number of key audit matters in the annual report.

(17)

17 3.1.4 Moderator: gender diversity

The moderator within this research is gender diversity in the board of directors. Following prior research (Lakhal et al., 2015; Rose, 2007; Srinidhi et al., 2011), the number of females on the board to the total number of members on the board of directors is used as a measure of gender diversity (WOM).

3.1.5 Control variables

This research will include several control variables in the used econometric tests. These variables might have an important effect on audit pricing and will add quality of the results in this research. The following control variables as determinants of external audit fees can be distinguished: profitability, cash flow from operations, leverage, size, foreign, ratio of receivables and inventory to total assets, and growth (Kealey et al., 2007; Venkataraman, et al., 2008; Hogan & Wilkins, 2008). Hogan and Wilkins (2008) argue that organization’s business risk will increase audit pricing. To control the organization’s business risk, the return on assets measured as the amount of net income, divided by total assets (PROFITABILITY) is included. Moreover, the cash flow from operations (CFO) and leverage (LEVERAGE) are included. To capture client size and audit complexity, the natural logarithm of total assets (SIZE), the foreign operations (FOREIGN), and the ratio of receivables and inventory to total assets (INVREC) are included. In this research also, a control for organization growth is included, the market-to-book ratio (GROWTH).

3.2 Data collection and analysis procedure

This study seeks to find evidence to the relation between CEO narcissism and audit pricing in the two-tier boards’ structure, and the mediating role of audit risk and the moderating role of gender diversity in the board of directors on this relationship through the mediator audit risk. As mentioned in the introduction, organizations from the Netherlands are subject to this research.

Panel data, generally referred to as longitudinal data, is used to conduct this research. In this way, two dimensions, a time series and a cross-sectional dimension, are involved in the observations and a larger number of data points will be obtained which decreases collinearity between explanatory variables and increases the degrees of freedom (Hsiao, 2007). STATA statistical software is used for the data analysis due to its suitability to panel data regression (Bai et al., 2004; Black et al., 2006; Xia & Zhu, 2009; Shan & McIver, 2011).

(18)

18 The analysis of the effect of CEO’s narcissism on audit fees is started with a sample of 131 Dutch organizations, which are listed on the Amsterdam Small Cap Index (AScX),

Amsterdam Midkap Index (AMX), Amsterdam Exchange Index (AEX), and EuroNext (ENX) from 2011 through 2016. Thomson Reuters Datastream and Asset4 are used to obtain the data for the analysis. The organization specific information is retrieved from the database of Thomson Reuters Datastream. The database Asset4 is used to collect data about gender diversity in the board of directors. Next, the CEO narcissism score is gathered by hand-collected data from the publicly available annual reports in 2011 and 2012. When the CEO is the same one for the rest of the period 2011-2016, the narcissism score is averaged over the years 2011 and 2012. However, when there is a CEO change in 2011, the narcissism score is computed over the years 2012 and 2013. Finally, the data of the mediator is gathered by hand-collected data from the publicly available annual reports from 2013 through 2016. The reason for this timeframe is that the disclosure of key audit matters became mandatory for listed organizations since October 2012. Taken together, this research uses a timeframe from 2013 through 2016. The final sample size is described in table 1.

Table 1 sample distribution

Initial number of organizations 131

Organizations with missing data 47

Number of organizations after elimination 84

Number of years used in this research 4

Sample size 336

Years with missing data 85

Final sample size 251

3.3 Reliability and validity

The data is collected by using several databases. The reliability of this data is secured by random checks. A small sample of organizations from the dataset is selected, and the variables are checked manually by collecting the corresponding annual reports. Several students, who have been provided instructions prior to the start of the collection process, provide the hand-collected data. The hand-collected data is gathered from the publicly available annual reports. Fogarty (2006) recommends collecting at least one manually

(19)

19 the data is publicly available and various researchers commonly accepted that the information included in annual reports is of high validity (Amernic et al., 2010; Fogarty, 2006).

Before the analysis of the data, the data is checked to assure that the general assumptions will be met, which are required to conduct statistical tests. Firstly, the variables are subject to the process of winsorizing. In this way, the outliers of the dataset will be discovered. The lower and upper bounds of the variables are computed using the mean minus/plus three times the standard deviation. The outliers correspond with the lower and upper bound values. By the process of winsorizing the results of analyses will be more reliable and a distorted view will be prevented. Furthermore, the natural logarithm is calculated for several variables, which is important to continuous variables. These continuous variables are tested whether the data is normally distributed. In this way, distortions will be prevented in the results of the analyses. A characteristic for the accounting research is the high R² due to control variables that have an effect on the audit process (e.g. DeFond & Francis, 2005). For this reason, the increment in R² will be calculated to discover the increase in explanatory power of the (independent)

variable(s) of interest (VI). In line with Collins et al. (1997) the incremental R² and scaled incremental R² are calculated as:

Δ R² VI = R²full - R² without VI

% R²VI = (R²full - R²without VI) / R² without VI 3.4 Specifications of the tests

In this paragraph, the specifications of the tests performed will be presented and elaborated. The hypotheses will be separately discussed below.

This research is conducted in a quantitative manner by performing statistical tests. Following Baron and Kenny (1986) four steps are taken into account to test the first and second

hypotheses. The first step of Baron and Kenny (1986) is testing the first hypothesis of this research. The aim is to find whether CEO narcissism is positively related to audit pricing. In other words, the direct effect of CEO narcissism on audit pricing. The following regression equation is estimated:

Regression equation 1 (step 1)

(20)

20 In the second hypothesis of this research, the aim is to find whether the relation between CEO narcissism and audit pricing is positively mediated by audit risk. In other words, an indirect effect of CEO narcissism on audit pricing that passes through audit risk (Shrout & Bolger, 2002). The remaining steps of the mediation analysis are: CEO narcissism on audit risk (step 2), audit risk predicting audit pricing (step 3), and CEO narcissism and audit risk predicting audit pricing (step 4). In order to test the second hypothesis, the following equations are estimated of the remaining steps:

AUDR = 𝛽0 + 𝛽1NARCISSISM + yCONTROLS + FIXED EFFECTS +

𝜀

𝑖

LNAUDIT = 𝛽0 + 𝛽1AUDR + yCONTROLS + FIXED EFFECTS +

𝜀

𝑖

LNAUDIT = 𝛽0 + 𝛽1NARCISSISM + 𝛽2AUDR + yCONTROLS + FIXED EFFECTS +

𝜀

𝑖

In the third hypothesis of this research, the aim is to find whether the indirect effect of CEO narcissism on audit pricing, through audit risk is negatively moderated by gender diversity in the board of directors. In order to estimate the moderating effect, the observed variable regressions approach of Hayes (2013) is one of the most popular analytical procedures (Cheung & Lau, 2015). This approach consists of the product of an independent variable and a moderator in the regression equations to measure the moderating effect. In line with Hayes (2013), the following equations are required to estimate the moderated mediation:

AUDR = a0 + a1NARCISSISM + a2WOM + a3(NARCISSISM x WOM) + yCONTROLS +

(21)

21 Regression equation 6 (step 2)

LNAUDIT = 𝛽0 + 𝛽1AUDR+ 𝛽2NARCISSISM + 𝛽3WOM + 𝛽4 (NARCISSISM x WOM) +

yCONTROLS + FIXED EFFECTS + 𝜀𝑖

The conditional impact of the indirect effect is indicated by using the following equation:

Conditional indirect effect = 𝛽1 (a1 + a3w)

In order to test the third hypothesis in more detail, Hayes (2015) recommends bootstrap confidence intervals. For this reason, bias corrected 95% bootstrapping confidence intervals for the indirect effects is conducted at low, moderate, and high level of gender diversity in the board of directors. The following equations are estimated, whereby _b is the number of bootstrap samples:

Low gender diversity = (_b[AUDR: NARCISSISM] + (mean - standard deviation) *_b[AUDR: NARCISSISM x WOM])*_b[LNAUDIT: AUDR]

Moderate gender diversity = (_b[AUDR: NARCISSISM]+(mean)*_b[AUDR: NARCISSISM x WOM])*_b[LNAUDIT: AUDR]

High gender diversity = (_b[AUDR: NARCISSISM]+(mean + standard deviation)*_b[AUDR : NARCISSISM x WOM])*_b[LNAUDIT: AUDR]

In the fourth hypothesis of this research, the aim is to find evidence that gender diversity in the board of directors is negatively related to audit risk. In testing this hypothesis, a linear regression analysis is applied. The following regression equation is estimated:

Regression equation 7

AUDR = 𝛽0 + 𝛽1WOM + yCONTROLS + FIXED EFFECTS +

𝜀

𝑖

In the formulas the constant is β0, βi represents the variables used in this research, the fixed effects, and ɛi is the distortion term.Despite the extensive use of control variables in the regression equations, one potential issue with the analysis can be that industry or time specific fundamentals drive the findings. Therefore, industry and time fixed effects will be included to isolate the effects of the individual CEOs. The table below summarizes the variables that are used in the regressions.

(22)

22 Table 2 description of variables

Variables Measurement

Dependent variable

LNAUDIT The natural logarithm of external audit fees paid by a client organization.

Independent variable

NARCISSISM A composite measure of CEO narcissism based on: (1) Picture score

(2) First person singular pronouns

Mediator

AUDR The natural logarithm of number of key audit matters in the annual report.

Moderator

WOM The number of female non-executive directors divided by the total number of members on board of directors.

Control variables

PROFITABILITY The return on assets measured as the amount of net income, divided by total assets.

CFO Measured as the operating cash flow in the current year, divided by total assets in the preceding year.

LEVERAGE Measured as total liabilities divided by equity. SIZE Measured as a logarithm of total assets.

FOREIGN Measured as foreign sales divided by total sales. INVREC The ratio of receivables and inventory to total assets.

GROWTH Measured as the market value of equity as a percentage of book value of equity.

4. Results

In this chapter, an overview will be provided of the analyses and the relating results that were derived. The descriptive statistics and correlations of the dataset will be provided in section 4.1. Furthermore, section 4.2 discusses the outcomes of the tests.

(23)

23 4.1 Descriptive statistics and correlations

Table 3 provides the descriptive statistics of the variables included in the dataset used in this research. The descriptive statistics functions as a summary of the used data and provides the sample size (n), the mean, the standard deviation, the minimum, and the maximum.

Table 3 descriptive statistics

Variable n Mean Std. Dev. Min Max

Dependent variable LNAUDIT 251 6,830 1,786 3,556 10,221 Independent variable NARCISSISM 251 -0.152 0.452 -0.811 1.290 Mediator AUDR 251 1.160 0.409 0.000 1.946 Moderator WOM 251 0.177 0.148 0.000 0.500 Control variables PROFITABILITY 251 2.876 13.045 -48.210 44.300 CFO 251 12.592 58.801 -9.628 452.158 LEVERAGE 251 205.129 1860.586 -39.929 17088.000 SIZE 251 13,495 2,565 6.779 17,920 FOREIGN 251 53.511 37.456 0.000 100.000 INVREC 251 0.421 1.150 0.003 10.699 GROWTH 251 2.288 4.399 -20.370 22.670

Before the beginning of the analyses, the multicollinearity will be tested. The Pearson correlation analysis is used, to check the variables for multicollinearity. This analysis is important, to prevent distortions in the outcomes of the analyses. Appendix A depicts the correlations of the Pearson correlation analysis of the variables. The results of the Pearson correlation analyses show no multicollinearity between the variables because the coefficients do not exceed the value of 0.7 (Sinan & Alkan, 2015).

(24)

24 4.2 Results from the analyses

In this section, the results will be first presented and then discussed. By using (regression) equations, this research can make conclusions on whether to reject or accept the hypotheses. Firstly, the results of the first and second hypotheses will be analysed in paragraphs 4.2.1 and 4.2.2. Next, in paragraph 4.2.3 the results of the third hypothesis will be analysed, and the results of hypothesis 4 will be analysed in paragraph 4.2.4.

4.2.1 Results for CEO narcissism and audit pricing (H1)

In this paragraph, the first step of the mediation analysis of Baron and Kenny (1984) will be used to test the first hypothesis of this research – whether CEO narcissism is positively related to audit pricing. Table 4 provides the results of the performed regression to test the first hypothesis:

Table 4 results regression equations H1 and H2

Equation 1 Equation 2 Equation 3 Equation 4 Variables Step 1 LNAUDIT Step 2 AUDR Step 3 LNAUDIT Step 4 LNAUDIT Constant -1.776 0.558 -1.998 -1.613 NARCISSISM 0.227** 0.049 0.241*** PROFITABILITY -0.001 -0.005*** -0.001 -0.004 CFO 0.001 -0.001*** 0.001 0.000 LEVERAGE -0.000*** -0.000*** -0.000 -0.000* SIZE 0.602*** 0.045*** 0.630*** 0.615*** FOREIGN 0.001 0.001** 0.001 0.001 INVREC 0.316*** 0.001 0.301*** 0.316*** GROWTH -0.015** -0.013* -0.019* -0.018 AUDR -0.275 -0.291

Industry dummies Included Included Included Included Year dummies Included Included Included Included Clustered by: Audit company

name Audit company name Audit company name Audit company name R² 0.843 0.350 0.843 0.846 R² without VI 0.840 0.348 0.840 0.840 Change in R² 0.003 0.002 0.003 0.006 Incremental R² 0.357% 0.575% 0.357% 0.714%

Note: * Significant at the 10% level (= p < 0.1), ** Significant at the 5% level (= p < 0.05), *** Significant at the 1% level (= p < 0.01).

Table 4 shows that the increase in incremental R² is 0.003 (0.357%) for CEO narcissism. In the first regression equation of table 4 is CEO narcissism positively and significantly related to audit pricing (β=0.227, p=0.044). This implies that hypothesis 1, CEO narcissism is positively related to audit pricing, can be accepted.

(25)

25 4.2.2 Results for the mediating role of audit risk (H2)

In this paragraph, the same table of regression equations is used to test the second hypothesis of this research – whether the relationship between CEO narcissism and audit pricing is positively mediated by audit risk. Table 4 shows that audit risk does not positively mediate the relationship between CEO narcissism and audit pricing. The increase in incremental R² varies from 0.002 (0.575%) for CEO narcissism and audit risk to 0.006 (0.714%). In the first step, CEO narcissism is significantly and positively related to audit pricing (β=0.227;

p=0.044). In the second step, CEO narcissism is not significantly related to audit risk

(β=0.049; p=0.283). Moreover, in the third step is audit risk not significantly related to audit pricing (β=-0.275; p=0.354). According to Baron and Kenny (1986) is the mediation not possible or likely because two relationships are not significant. However, the final step is examined to see whether there is a significant effect of audit risk after controlling for CEO narcissism. When audit risk controlled, CEO narcissism is still significant (β=0.241; p=0.035), these findings do not support full mediation. Moreover, there is also no partial mediation because audit risk is not significant (β=-0.291; p=0.336). There is no ground for mediation and implies that hypothesis 2, the relationship between CEO narcissism and audit pricing is positively mediated by audit risk, needs to be rejected.

4.2.3 Results for the moderated mediation (H3)

In this paragraph, the third hypothesis of this research will be tested – whether the indirect effect of CEO narcissism on audit pricing, through audit risk is negatively moderated by gender diversity in the board of directors. First, the results of table 5 show that the interaction effect of CEO narcissism and gender diversity in the board of directors has no significant involvement on audit risk (β=0.771, p=0.148) in the fifth equation. This is the direct

moderating impact of gender diversity on the relation between CEO narcissism and audit risk. The incremental R² is 0.016 (4.598%) for the (independent) variables of interest (i.e.

NARCISSISM, WOM, WOM x NARCISSISM). Next, the results of the second step of table 5 show that the interaction of CEO narcissism and gender diversity in the board directors has a significant effect in audit pricing (β=1.840, p=0.018). The incremental R² is 0.009 (1.071%) for the (independent) variables of interest (i.e. NARCISSISM, WOM, WOM x

NARCISSISM, AUDR). Hayes (2013) provides the conditional indirect effect equation, which can be explained as the indirect effect of CEO narcissism on audit pricing through audit risk conditional on gender diversity in the board of directors. In this equation, the conditional indirect effect can be computed as 𝛽1 (a1 + a3w) = 0.040.

(26)

26 Table 5 results regression equations H3

Variables Equation 5 Step 1 AUDR Equation 6 Step 2 LNAUDIT Constant 0.471 -0.974 NARCISSISM -0.119 -0.150 WOM -0.001 0.306 NARCISSISM x WOM 0.771 1.840** PROFITABILITY -0.005*** -0.002 CFO -0.002*** 0.000 LEVERAGE -0.000*** -0.000* SIZE 0.048*** 0.615*** FOREIGN 0.001** 0.002 INVREC 0.008** 0.330*** GROWTH -0.011* -0.162 AUDR -0.335

Industry dummies Included Included

Year dummies Included Included

Clustered by: Audit company name Audit company name

R2 _0.364 _0.849

R² without VI 0.348 0.840

Change in R² 0.016 0.009

Incremental R² 4.598% 1.071%

In order to test this hypothesis in more detail, this research makes use of bias corrected 95% bootstrap confidence intervals for the indirect effects using 1,000 bootstrap samples. Table 6 shows that the conditional indirect effect increases as the value of the moderator, gender diversity in the board of directors, increases. The indirect effect is partially significant

(Edwards & Lambert, 2007) because the 95% confidence interval includes zero at low level of gender diversity, which indicates that no significant indirect effect has been found at low levels of gender diversity (β=0.104, 95% Conf. Interval: -0.108 to 0.316). However, the indirect effect is significant at moderate levels of gender diversity (β=0.117, 95% Conf. Interval: 0.020 to 0.323) and high levels of gender diversity (β=0.239, 95% Conf. Interval: 0.063 to 0.415). When zero is not present in the confidence intervals, the results depict evidence of the indirect effect. This implies that hypothesis 3, the indirect effect of CEO narcissism on audit pricing, through audit risk is negatively moderated by gender diversity in the board of directors, can be partially accepted. However, there is a positive indirect

moderating effect, instead of the hypothesized negative moderating effect (β=0.104; β=0.171; β=0.239).

(27)

27 Table 6 indirect effects of CEO narcissism through audit risk at values of gender diversity in the

board of directors LNAUDIT

Gender diversity β Confidence level of 95% for confidence intervals

Low gender diversity 0.104 -0.108 to 0.316

Moderate gender diversity 0.171 0.020 to 0.323

High gender diversity 0.239 0.063 to 0.415

4.2.4 Results for gender diversity and audit risk (H4)

In this paragraph, the fourth hypothesis of this research will be tested – whether gender diversity is negatively related to audit risk. Table 7 provides the results of the performed regression to the test the fourth hypothesis.

Table 7 results regression equation H4

Variables Equation 7 AUDR Constant 0.474 WOM -0.130 PROFITABILITY -0.004 CFO -0.001 LEVERAGE -0.000 SIZE 0.051 FOREIGN 0.001 INVREC -0.003 GROWTH -0.012

Industry dummies Included

Year dummies Included

Clustered by: Audit company name

R2 _0.350

R² without VI 0.348

Change in R² 0.002

Incremental R² 0.575

Table 7 depicts that the increase in incremental R² is 0.002 (0.575%) for the variable gender diversity in the board of directors. Moreover, table 7 shows that gender diversity in the board of directors is not significantly related to audit risk (β=-0.130, p=0.578). This implies that the fourth hypothesis, gender diversity is negatively related to audit pricing, needs to be rejected.

(28)

28 4.3 Robustness tests

This section will discuss the robustness tests and sensitivity to assure that the aforementioned results are not merely the effect of the choices that have been made in this research.

4.3.1 Propensity score matching

This paragraph will discuss the robustness test applying propensity score matching (PSM). PMS matches CEO-firm observations for high CEO narcissism and low CEO narcissism. This research uses a dummy variable that takes the value of 1 if the CEO’s narcissism measure is larger than the median, the value of 0 otherwise. Resulting in 130 of the 251 CEOs being categorized as high narcissistic CEOs. The next step is to match high narcissistic CEOs with low narcissistic CEOs based on organization-specific characteristics. The CEOs were matched using organization specific factors that might either limit or increase the degree of narcissism personality i.e. profitability, cash flow from operations, leverage, size, foreign, ratio of

receivables and inventory to total assets, and growth. These variables are as described in table 2. The results of the PMS test show that the degree of narcissism is robust (β=0.410, p=0.000) and the first regression equation does not suffer from endogeneity.

Table 8 Propensity Score Matching test

Variables LNAUDIT NARCISSISM 0.410*** PROFITABILITY -0.003 CFO 0.002* LEVERAGE -0.000 SIZE 0.598*** FOREIGN 0.001 INVREC 0.299*** GROWTH -0.006

Industry dummies Included

Year dummies Included

R2 _0.812

R² without VI 0.801

Change in R² 0.011

Incremental R² 1.373%

4.3.2 Robustness check CEO narcissism

In this paragraph, the robustness will be discussed by applying an extended measure for CEO narcissism. This variable will be measured by following a measure employed in prior

(29)

29 four factors: relative cash pay, relative non cash pay, picture score, and first person singular pronouns. In the main analyses, the composite measure of two factors (picture score and first person singular pronouns) was used as proxy. In order, to strengthen the results, the

robustness with the composite measure of four factors is checked. Table 9 reports the results of the robustness check and is consistent with the aforementioned analysis (β=1.184,

p=0.061). CEO narcissism is positively and significantly related to audit pricing and is consistent with the aforementioned first regression equation. Besides, audit risk does not significantly (p=0.772) mediate the relationship between CEO narcissism and audit pricing in the fourth regression equation. Table 10 indicate that the interaction effect of CEO narcissism and gender of diversity in the board of directors on audit risk has no significant involvement (β=-2.741, p=0.875). Moreover, the interaction effect of CEO narcissism and gender diversity in the board of directors on audit pricing has also no significant involvement (β=-15.283, p=0.627) and is not in line with the aforementioned regression equation 6. The conditional indirect effect equation is 𝛽1 (a1 + a3w) = -0.162. Table 11 shows that the indirect effect is not

significant because the 95% confidence intervals include zero (Edwards & Lambert, 2007). This implies that the aforementioned analysis of hypothesis 3 is not robust.

Table 9 robustness check CEO narcissism hypotheses 1 and 2

Variables Equation 1 LNAUDIT Equation 2 AUDR Equation 3 LNAUDIT Equation 4 LNAUDIT Constant -3.563 -9.305 -1.997 -5.378 NARCISSISM 1.184* -0.662 1.054 PROFITABILITY -0.146 -0.038 -0.001 -0.153 CFO -0.331 -0.001 0.001 -0.331 LEVERAGE -0.008 0.024 -0.000 -0.003 SIZE 0.523*** 0.824 0.630*** 0.684 FOREIGN 0.011* -0.008 0.001 0.010 INVREC 5.759 6.665 0.301*** 7.060 GROWTH 0.463 -0.388 -0.019* 0.388 AUDR -0.275 -0.195 Industry dummies

Included Included Included Included

Year dummies Included Included Included Included Clustered by: Audit company

name Audit company name Audit company name Audit company name R2 _0.989 _0.957 _0.843 _0.990 R² without VI 0.840 0.348 0.840 0.840 Change in R² 0.149 0.609 0.003 0.150 Incremental R² 17.738% 175% 0.357% 17.857%

(30)

30 Table 10 robustness check CEO narcissism hypotheses 3 and 4

Variables Equation 5 AUDR Equation 6 LNAUDIT Equation 7 AUDR Constant -6.600 -3.432 0.474 NARCISSISM 0.346 6.068 WOM 4.039 -1.553 -0.130 NARCISSISM x WOM -2.741 -15.283 PROFITABILITY -0.029 -0.033 -0.004 CFO -0.044 -0.042 -0.001 LEVERAGE 0.144 -0.126 -0.000 SIZE 0.466 0.670 0.051 FOREIGN -0.010 0.012 0.001 INVREC 9.442 12.927 -0.003 GROWTH -0.426 0.119 -0.012 AUDR 0.190

Industry dummies Included Included Included Year dummies Included Included Included

Clustered by: Audit company name Audit company name Audit company name

R2 _0.970 _0.992 _0.350

R² without VI 0.348 0.840 0.348 Change in R² 0.622 0.152 0.002 Incremental R² 178.736% 18.095% 0.575

Table 11 robustness check moderated mediation LNAUDIT

Gender diversity β Confidence level of 95% for confidence intervals

Low gender diversity 0.020 -1.683 to 1.723

Moderate gender diversity 0.012 -1.002 to 1.027

High gender diversity 0.005 -0.439 to 0.448

4.3.3 Robustness check abnormal number of key audit matters in the annual report In this paragraph, the robustness will be discussed by applying an extended measure for audit risk, measured as the number of abnormal key audit matters in the annual report. In the main analyses, the number of key audit matters in the annual report was used as proxy. In order, to strengthen the results, the robustness with abnormal number of key audit matters will be checked. Table 12 and 13 report the results of the robustness check. The relationship between CEO narcissism and audit pricing is not significantly and positively mediated by audit risk in the equations (β=-0.064, p=0.447; β=-0.236, p=0.223; β=-0.225, p=0.260). Table 13 indicate that the direct interaction effect of CEO narcissism and gender diversity in the board of directors on audit risk has no significant involvement (β=0.529, p=0.333).

(31)

31 The interaction effect of CEO narcissism and gender diversity in the board of directors on audit pricing is significant (β=1.736, p=0.006). These results are in line with the

aforementioned analyses. Moreover, the variable gender diversity in the board of directors is not significant (β=-0.134, p=0.589) and is in line with the aforementioned equation 7. This implies that the results from the robustness check regarding abnormal number of key audit matters are consistent with the aforementioned results and these results are robust.

Table 12 robustness check abnormal number of key audit matters hypothesis 2

Variables Equation 1 LNAUDIT Equation 2 AUDR Equation 3 LNAUDIT Equation 4 LNAUDIT Constant -1.776 0.812 -1.883 -1.562 NARCISSISM 0.227** -0.064 0.213* PROFITABILITY -0.001 -0.004** -0.001 -0.002 CFO 0.001 -0.002** 0.001 0.001 LEVERAGE -0.000*** -0.000** -0.000* -0.000* SIZE 0.602*** 0.026*** 0.620*** 0.606*** FOREIGN 0.001 0.001 0.001 0.001 INVREC 0.316*** 0.001 0.303*** 0.316*** GROWTH -0.015** -0.011** -0.019* -0.019* AUDR -0.236 -0.225 Industry dummies

Included Included Included Included

Year dummies Included Included Included Included Clustered by: Audit company

name Audit company name Audit company name Audit company name R2 _0.843 _0.203 _0.839 _0.841 R² without VI 0.840 0.201 0.840 0.840 Change in R² 0.003 0.002 -0.001 0.001 Incremental R² 0.357% 0.995% -0.119% 0.119%

(32)

32 Table 13 robustness check abnormal number of key audit matters hypotheses 3 and 4

Variables Equation 5 AUDR Equation 6 LNAUDIT Equation 7 AUDR Constant 0.607 -0.947 0.904 NARCISSISM -0.180* -0.158 WOM -0.065 0.277 -0.134 NARCISSISM x WOM 0.529 1.736*** PROFITABILITY -0.004** -0.002 -0.004 CFO -0.002*** 0.000 -0.002 LEVERAGE -0.000*** -0.000** -0.000 SIZE 0.029 0.604*** 0.024 FOREIGN 0.002** 0.002 0.002 INVREC 0.007 0.330*** 0.007 GROWTH -0.011 0.170* -0.011 AUDR -0.243

Industry dummies Included Included Included Year dummies Included Included Included

Clustered by: Audit company name Audit company name Audit company name

R2 _0.210 _0.845 _0.202

R² without VI 0.201 0.840 0.200 Change in R² 0.009 0.005 0.002 Incremental R² 4.478% 0.595% 1.000%

5 Conclusions and discussion

The theories, hypotheses, and results will be concluded in this chapter. In section 5.1 an overview will be given of the hypothesis is accepted or rejected and conclusions will be formed. Furthermore, the implications and limitations of this research will be elaborated in section 5.2. In section 5.3, some directions of future research will be provided.

5.1 Conclusions

The main goal of this study has been to answer the following research question: “To what

extent is CEO narcissism related to audit pricing in the two-tier boards’ structure and to what extent is this relationship mediated by audit risk and does gender diversity in the board of directors moderates this relationship through the mediator audit risk?”. This research

proposed that CEO narcissism affects the audit pricing. More specifically, this relationship could be explained by either the upper echelon theory, or the agency theory. Besides, it was proposed that audit risk positively mediates the relationship between CEO narcissism and audit pricing by the audit risk model to determine the audit risk and, in this way, explained the audit pricing. Furthermore, the literature has shown that gender diversity affects the quality of decisions and the monitoring role (Adams & Ferreira, 2009; Lakhal et al., 2015). Therefore,