The influence of CEO Narcissism on Perceived Audit- and Fraud Risk:
The moderating role of Corporate Governance Strength.
(Evidence from the Netherlands)
University of Groningen
Faculty of Economics & Business
Department of Accountancy & Controlling
MSc Thesis Accountancy
January, 2018
Name: BSc. Jeroen Omri Nathan Zeijl
Address: Oosterweg 62a
Postal code: 9724 CK Groningen
Phone number: + 31 6 46 59 38 68
Email: j.o.n.zeijl@student.rug.nl
Student number: s2367866
First Assessor: Prof. dr. J.A. (Jim) Emanuels
Co-assessor: EMA. M.L. (My Lan) Lu RA. MSc
Second Assessor: Dr. Y. (Yasemin) Karaibrahimoglu
External Supervisor E. (Elmer) Zoomer MSc
Word count: 12.611
ABSTRCT
This paper examines what impact CEO Narcissism has on the perceived audit- and fraud risk of auditors and how this relationship is moderated by Corporate Governance Strength (CGS). Auditors’ risk perceptions are seen as a highly important and essential for a thorough risk assessment and a good audit quality. This study tries to gain insight in the way personality characteristics influence audit risk perceptions. This may help auditors in meeting regulatory expectations and requirements concerning the incorporation of management’s integrity and fraud attitudes when making risk assessments. The research is done by means of an experiment in the form of a 2 x 2 factorial design – between subjects. A total of 125 auditors from Deloitte, KPMG, PWC and Flynth participated in the experiment via Qualtrics. The results show that observable indicators of CEO narcissism are positive and significant related to perceived audit- and fraud risk. This means that companies with more narcissistic CEOs are seen as a higher risk for the audit. Also, for auditor experience, a positive significant relation is found with perceived audit- and fraud risk. This suggests that the more experienced an auditor is, the more efficient risk assessments can be made, based on the perceptions concerning the risk of personality characteristics. Due to a response rate below 80%, a non-response bias could affect the reliability of the results. Because the perception of auditors is a subjective measure, the results should be interpreted with caution.
Table of Content
ABSTRCT ... 2
1. Introduction ... 5
2. Theoretical Background ... 7
2.1 Narcissism and CEOs... 8
2.2 CEO Narcissism, fraud and audit risk ... 9
2.3 Audit- and fraud risk perception ... 10
2.4 Corporate Governance Strength ... 11
2.5 Co-variables ... 12
Auditor Experience. ... 12
Auditor Gender. ... 12
Auditor Narcissism. ... 13
2.6 Upper Echelon Theory ... 13
2.7 Conceptual Model ... 14
3. Methodology ... 14
3.1 Experimental Design ... 14
3.1.1 Participants and Data Collection ... 15
3.2 Dependent variable ... 15
3.3 Independent variables ... 16
3.3.1 CEO Narcissism (CEON) ... 16
3.3.2 Corporate Governance Strength (CGSTR) ... 17
AC Experience ... 17
AC diligence. ... 17
AC-CEO social ties ... 17
3.4 Co-variables ... 18
3.4.1 Auditor Experience (AUDEXPI). ... 18
3.4.2 Auditor Gender (AUDGEN). ... 19
3.4.3 Auditor Narcissism (AUDNAR). ... 19
3.5 Control variables ... 19
3.6 Statistic model ... 20
3.7 Validity, reliability and common method bias ... 21
4. Results ... 22
4.1 Descriptive statistics ... 23
4.2.1 Non-response ... 24
4.2 Missing values ... 24
4.3 Manipulation check tests ... 25
4.4 Cronbach’s Alpha ... 26
4.5 Multicollinearity ... 27
4.7 Regression analysis ... 29 Model 1 ... 29 Model 2 ... 29 Model 3 ... 30 Model 4 ... 30 Model 5 ... 30 Model 6 ... 30 Model 7 ... 30 Model 8 ... 30 4.8 Additional analysis ... 34
Adjusted risk-taking propensity ... 34
5. Discussion and Conclusion ... 38
5.1 Summary ... 38
5.2 Findings and Conclusion ... 38
5.3 Discussion ... 40
5.4 Implications and Recommendations for future research ... 40
6. Reference list ... 42
7. Appendices ... 48
“
V
isibly irritated stood ‘Renske Leijten’ behind the rostrum in the House of representatives. ‘When is the limit reached?’ asked the SP-politician at a high tone, after she had listed a series of fraud scandals involving accountants in recent years. From Housing Corporation Vestia to meat processor Weyl. “Ok, one more than: cremation Group ‘De Facultatieve’. ‘When I think about accountants, I think of scandals’ she said (Knoop, 2017).1. Introduction
Since multiple fraud scandals, litigations and the financial crisis in 2008 have violated the integrity, image and expertise of accountants, the damaged confidence in the accountancy profession is an issue (Van Der Ven, 2015; Knoop, 2017). Ever since, audit firms tried to improve the quality as pressure increases, but a desired level of quality is yet to be reached (Van Der Heijden, 2017). In December 2012, the PCAOB (2012) reported instances in which auditors failed to apply appropriate professional skepticism in their dealings with client management. In 2015 the AFM concluded that 45% of the examined audits from 2014 did not meet the desired quality (Van Der Ven, 2015; Knoop, 2017). Recently the position of the accountancy profession became even more under pressure as it was criticized by de AFM again, about 60% of the examined audits from 2016 were of insufficient quality (Van Der Heijden, 2017; Knoop 2017).
To obtain reasonable assurance whether the financial statements are free of material misstatements due to error or fraud, an audit must be planned and performed (PCAOB, 2012). However, inappropriate accounting may be caused by auditors who occasionally develop improper levels of trust in management. Incentives or pressures (e.g. realizing high client satisfaction ratings, avoiding important management conflicts and building or maintaining long-term audit engagements), which can be created by circumstances inherent to the audit environment, can influence auditors’ correct use of professional skepticism and allow unconscious prejudice to prevail. To realize the desired audit quality, skepticism and a realistic audit risk perception are necessary (PCAOB, 2012).
As stated by Fukukawa & Mock (2011), “One of the key elements of audit quality is the individual risk assessments auditors make” (p.75). The auditors’ objective is to reduce the overall audit risk in the financial statements to an appropriately low level (PCAOB, 2012). Furthermore, because “the risk of fraud is an integral component of overall audit risk” (Allen et al., 2006, p.159), a careful consideration of the audit risk, including the fraud risk judgments, is crucial for the overall audit quality and the audit process (Boyle, Carpenter & Hermanson, 2012).
The judgement of entity-level controls, which includes control environment and tone at the top, is also fundamental for the assessment of audit risk (Lauck, Rakestraw & Stein, 2016). The aim of the tone at the top, settled by the top management, is to pass on high ethical standards and integrity in the performance of all activities through the organization (Hunton, Hoitash, & Thibodeau, 2011). The importance of the “tone at the top” and “integrity” were already discussed in the COSO report on
fraudulent financial reporting of 1999 (Rijsenbilt & Commandeur, 2013). A firm’s tone-at-the-top or ethical environment is strongly influenced by CEOs and therefore demands critical attention from auditors when the client control environments and fraud risks are evaluated (Judd, Olsen & Stekelburg, 2016; Rijsenbilt & Commandeur, 2013; Hunton et al., 2011). As a matter of fact, dysfunctional “tone at the top” has been linked to executives and those in charge with governance, who possess indicators of narcissism (Johnson et al., 2013; Amernic and Craig, 2010).
Apostolou, Hassel, Weber & Sumners (2001) found that the “management characteristics” and the “influence over control environment” categories, assessed as the most important red flags of Statements of Auditing Standards (SAS) No. 82, are related to the possibility for management fraud. Furthermore, personality characteristics seem to be a very dominant fraud-risk component too (Cohen, Ding, Lesage & Stolowy, 2012). In addition, previous research shows that personality characteristics of narcissistic CEOs are related to the risk of fraud, a fraud risk attitude and fraudulent or unethical behavior (Chatterjee & Hambrick 2007; Rijsenbilt & Commandeur, 2013; Johnson, Kuhn, Apostolou, & Hassel, 2013; Judd et al., 2016). This is in accordance with literature on the upper echelon theory, which classified narcissism as a crucial personality element of CEOs, which can affect strategic and organizational decisions (Zhu & Chen, 2015; Chatterjee & Hambrick, 2007; Judd et al., 2016). Nevertheless, the judgement of fraud risk is one of the hardest challenges auditors face (Boyle et al., 2012).
Prior research examined personality characteristics, such as narcissism, in relation to fraud attitude risk and audit risk (Johnson et al., 2013; Lauck et al., 2016). However, research about auditors’ perceptions regarding these risks is still limited. Judd et al., (2016) examined the relation between CEO narcissism and external audit fees (as a proxy for perceived audit risk). They stated that clients with relatively more narcissistic CEOs and charged with a risk premium and/or auditors have to perform more work. In contrast, Johnstone (2000) found by experiment that “partners did not use more proactive risk-adaption strategies such as adjusting the audit fee, making plans about necessary audit evidence, making plans about personal assignment, and/or adjusting the amount of data collected during the client-acceptance process to make higher-risk clients more acceptable” (p. 3). This means there are contrary results in the current literature.
Corporate governance, which is the responsibility of the board of directors (Cadbury, 1992), “is about ensuring that a company is run properly” (Collier, 1997, p.70). To ensure integrity of the financial reporting process, oversight and monitoring activities are undertaken by the board and the Audit Committee (hereafter: AC) (Cohen & Hanno, 2000). According to Abbot, Park & Parker (2000), high AC effectiveness is assumed to increase the quality of financial reporting. They found that organizations with an AC, which consists of independent directors who meet at least twice a year, are less likely to be sanctioned for fraudulent or misleading reporting. Further, Beasley, Carcello, Hermanson & Lapides (2000) found that the CG mechanisms of fraud and no-fraud firms are very different, in specific for AC-independence, AC-diligence and the existence of an AC.
Research of Rijsenbilt (2011) confirmed that the actions of top executives, and CEOs in particular, define the ethical climate of a firm. In addition, as previous studies suggest, narcissism is related to fraudulent behavior (Chatterjee & Hambrick 2007; Rijsenbilt & Commandeur, 2013; Johnson, Kuhn, Apostolou, & Hassel, 2013; Judd et al., 2016). Furthermore, as Cohen and Hanno (2000) mentioned, previous auditing research suggests that studying the impact of various factors on audit judgments helps improving risk assessment models and the development of decision aids. So, since many of world’s present CEOs have narcissistic personality (De Hoogh, Den Hartog & Nevicka, 2015) it is interesting and relevant to study to what extent their personality characteristics have an impact on audit risk assessments, from an audit perspective. This study will be executed in the form of an experiment. The fundamental question of this experiment is: ‘To what extent does CEO narcissism
influence the audit- and fraud risk perception of auditors in the Netherlands, and how is corporate governance strength moderating this relationship?’
This research question intervenes on current events regarding the accountancy profession, such as pressure due to fraud scandals and insufficient quality. This study will broaden and refine the research of Johnson et al., (2013) by specifically focusing on the impact of narcissistic CEOs instead of client managers on perceived audit risk. Also, as suggested by Johnson et al., (2013), the contribution of auditor experience on auditors’ risk assessments will be examined in this study. The purpose of this paper is to give an understanding of the extent that auditors incorporate CEO personality characteristics in their audit risk judgement. This may assist future audit processes and help developing client acceptance guidelines (Cohen and Hanno, 2000). Also, in response to Boyle, DeZoort & Hermanson (2015), this study will contribute to current literature by extending linkages of fraud risk assessments. This may contribute to the development of improved policies and guidelines about audit risk judgements. Consequently, practical insights for auditors can be generated and contributions can be made in realizing a more desirable audit quality. Most research around this topic uses data from the U.S. and consists of archival studies. This study will enhance current literature by taking an experiment-based approach, based on data from the Netherlands.
The structure of the remnant of this study will be as follows. First, the theoretical link between CEO narcissism and fraud, the moderating role of corporate governance strength and auditor characteristics, including the development of hypothesis, will be elaborated. This is followed up by a description of the methodology and the development of the experimental material. Then, the statistical model and experimental results will be presented and interpreted. Finally, the paper will conclude with a discussion of the results, the main limitations and the recommendations for future research.
2. Theoretical Background
This chapter will provide further information concerning narcissistic personality characteristics of CEOs and will illustrate its influence on perceived audit- and fraud risk. Also, an explanation on the proposed
moderating role of corporate governance strength will be provided. Furthermore, the reasoning for the development of hypotheses regarding auditor gender, auditor experience and auditor narcissism is given.
2.1 Narcissism and CEOs
“The term narcissism comes from the Greek myth of Narcissus, the story of a young man who fell in love with his own image” (O’Reilly et al., 2013, p.2). In 1914, Freud described this abnormal form of self-love on the basis of multiple specifications such as “self-admiration and a tendency to see others as an extension of one’s self”’ (Rijsenbilt & Commandeur, 2013, p414). Until research in the 80’s, narcissism was mainly defined in categories. Individuals were narcissistic (abnormal) or were not narcissistic (normal) (Rijsenbilt & Commandeur, 2013). Nowadays, narcissism1 is a personality
spectrum on which people can vary from a low to high degree (American Psychiatric Association, 2013; Rijsenbilt & Commandeur, 2013). In psychological literature, researchers distinguish between ‘clinical narcissism’ (narcissistic personality disorder (NPD)) and ‘trait narcissism’ or ‘subclinical narcissism’ (Narcissism) (Johnson et al., 2013). Narcissistic Personality Disorder (NPD) can be defined as a mental disorder whereby individuals exhibit low levels of empathy and high levels of grandiosity, have a strong desire for admiration and feel superior to others (American Psychiatric Association, 2013)2. Narcissism,
in terms of a subclinical variant of NPD, emerged from research of Raskin & Hall (1979) who were the first to construct a Narcissistic Personality Inventory (NPI) (Spain, Harms & LeBreton, 2014). Within the personality ‘trait narcissism’, the individual’s behavior is driven by the interaction of his personal motivations, relationships with others and self-regulatory strategies (Johnson et al., 2013). Therefore, narcissism contains aspects from narcissistic personality disorder, such as dominance, superiority, grandiosity and entitlement (Spain et al., 2014).
The personality characteristics of ‘trait narcissism’ are normally distributed in the adult population and considered as stable (Johnson et al., 2013). However, this is certainly not the case for Chief Executive Officers (hereafter: CEO), since many CEOs have a narcissistic personality (De Hoogh et al., 2015; Nevicka et al., Ten Velden, De Hoogh, & Van Vianen, 2011). According to research of
1 In the remainder of the paper the term ‘‘narcissism’’ refers to trait narcissism. 2 NPD starts at early adulthood and occurs in variety of context
, as indicated by five (or more) of the following aspect
1) Has a grandiose sense of self-importance (e.g., exaggerates achievements and talents, expects to be recognized as superior without commensurate achievements).
2) Is preoccupied with fantasies of unlimited successes, power, brilliance, beauty, or ideal love.
3) Believes that he or she is “special” and unique and can only be understood by, or should associate with, other special or high-status people (or institutions).
4) Requires excessive admiration.
5) Has a sense of entitlement (i.e., unreasonable expectations of especially favorable treatment or automatic compliance with his or her expectations).
6) Is interpersonally exploitative (i.e., takes advantage of others to achieve his or her own ends). 7) Lacks empathy: is unwilling to recognize or identify with the feelings and needs of others. 8) Is often envious of others or believes that others are envious of him or her.
Rosenthal & Pittinsky (2006), “It is clear that a significant number of world leaders have grandiose belief systems and leadership styles… whose aspirations, judgements, and decisions both good and bad, are driven by unyielding arrogance and self-absorption” (O’Reilly et al., 2014, p.3). There is strong evidence that more narcissistic people are more likely to become leaders, because of their dominant and grandiose personality (Brunell et al., 2008). Narcissists have a strong need for recognition, affirmation and reward (Ham, Lang, Seybert & Wang, 2017), and they are willing to take more risks in order to realize their desires (Olsen & Stekelburg, 2016). Because leadership roles offer the perfect opportunity to demonstrate their superiority to others (Nevicka et al., 2011), it is no surprise that many CEOs have a narcissistic personality (De Hoogh et al, 2015; Nevicka et al., 2011).
2.2 CEO Narcissism, fraud and audit risk
Narcissistic CEOs, who are often described as dominant, arrogant and selfish, have low levels of empathy for situations and feelings of others and a low sense of guilt (Schwartz, 2013; Nevicka et al., 2011). Narcissists have a strong desire for affirmation and recognition of their inflated sense of self-importance, self-esteem and self-praise (Judd et al., 2016; Olsen et al., 2014; Spain et al., 2014; Schwartz, 2013). In order to achieve those desires, they use their power and dominance (Rijsenbilt & Commandeur, 2013) to exploit others without shame or remorse (Schwarz, 2013; Ham et al., 2017). Further, they dominate decision making and fail to take criticism (Ham et al, 2017). This results in narcissistic CEOs who initiate challenging, risky or bold actions to obtain continuous external self-affirmation, praise and admiration (Chatterjee & Hambrick, 2007; Rijsenbilt & Commandeur, 2013; Zhu & Chen, 2014; Judd et al., 2016). However, this mask of grandiosity, high-confidence, and a never-ending desire for affirmation just camouflages a fragile self-esteem, vulnerable to the slightest criticism (Schwartz, 2013; Olsen et al., 2013).
Chief executive officers (CEOs) with a higher narcissistic personality turn out to manage their firms very different compared to less narcissistic CEOs (Zhu & Chen, 2014). The more narcissistic a CEO, the more a CEO tends to “play loose with the company’s reported financial position in order to shun remediation strategies and to live in a fantasy world of delusion about the company’s financial strength” Amernic & Craig 2010, p. 80). Narcissistic CEOs feel superior to others, believe they are above the law (Olsen & Stekelburg, 2016) and think the rules do not apply to them (Ham et al., 2017). Because of their constant fixation with themselves and their lack of moral sensibility (Olsen & Stekelburg, 2016), narcissists are more likely to violate rules and ethical norms for their own benefit (Ham et al, 2017). Rijsenbilt (2011) concluded that when top executives lack those ethical norms and values, the propensity for fraud increases. Prior research underlines this as, compared to other CEOs, narcissistic CEOs are more likely to cheat or behave unethical (Judd et al., 2016; Olsen & Stekelburg, 2016; O’Reilly et al., 2014).
The belief that behavioral- and personality characteristics of executives can relate to fraud is not unique (Rijsenbilt & Commandeur, 2013). Because narcissistic CEOs have a need for affirmation and
self-praise they are more likely to take bold and risky actions with extreme outcomes (Spain et al., 2014; Judd et al., 2016). In fact, previous research has already shown that personality traits play a key role in the likelihood of financial reporting fraud (Johnson et al., 2013; Rijsenbilt & Commandeur, 2013). Previous research shows that more narcissistic CEOs are more likely to engage in unethical or fraudulent behavior (Judd et al., 2016; Rijsenbilt & Commandeur, 2013; Ham et al., 2017; Olsen & Stekelburg, 2016). In line with this behavior, narcissistic CEOs are more likely to violate integrity standards, cheat or be unfaithful (O’Reilly et al., 2014). Narcissistic CEOs also engage more often in white collar crimes, academic fraud and lie about own achievements (Ham et al., 2017).
2.3 Audit- and fraud risk perception
Auditors have the purpose to analyze and evaluate the financial statements in such way, that the audit risk is reduced to an appropriate low level (PCAOB, 2010). Their responsibility is to identify and respond to high fraud risk situations (PCAOB, 2012). To increase the likelihood of fraud detection, an auditor should make explicit fraud risk assessments (Loebbecke, Eining & Willingham, 1989; Knapp & Knapp, 2001). Fraud is an intentional act designed to deceive or mislead another party (Knapp & Knapp 2001, p26.). As mentioned by Knapp & Knapp (2001), explicit fraud risk assessments are included as a required audit procedure in SAS No. 82, Consideration of Fraud in a Financial Statement Audit. Beside regulation claims, auditors nowadays also have a social claim due to multiple scandals. As stated by Loebbecke, Eining & Willingham (1989) “Auditors must ‘battle’ to maintain a sense of skepticism in light of economic pressure and natural tendencies to the contrary” (p25). When making a fraud risk assessment, auditors should also take client management’s attitude towards fraud into account (Johnson et al., 2013; PCAOB 2010), since personality characteristics seem to be a very dominant fraud-risk component (Cohen et al., 2012). Johnson et al., (2013) found that when a manager is characterized as narcissistic, the fraud risk assessment is significantly higher.
Apostolou et al., (2001) found that auditors classify red flags related to management characteristics and influence over control twice as important to the fraud risk assessment than red flags related to operating, financial stability and industry conditions. Hence, auditors increase their audit scope when their clients are perceived as high fraud risk and lacking integrity, so related evidence for their audit opinion can be added (Judd et al., 2016). High levels of CEO narcissism are related to more risky and unethical behavior (Judd et al., 2016; Chatterjee & Hambrick, 2007) and to low personal integrity (Johnson et al., 2013; O’Reilly, 2014). When considering the auditor’s fraud risk assessment, the integrity of management is a key element of the fraud attitude (Johnson et al., 2013). Previous research has already shown that personality traits play a key role in the likelihood of financial reporting fraud (Johnson et al., 2013; Rijsenbilt & Commandeur, 2013). In addition, higher assessments of executive personality characteristics, in particular narcissism, lead to increased auditor fraud risk perceptions (Johnson et al., 2013). Further, Judd et al., (2016) concluded that “Compared to other firms, firms with narcissistic CEOs are more likely to engage in risky business practices and to be perceived
as greater fraud risks, both of which increase audit risk” (p.1). Suggesting that the risky, unethical and non-integer behavior of a narcissistic CEO will pose a greater risk to the firm’s auditor, the following hypothesis is assumed.
H1: Auditors’ audit- and fraud risk perception will be higher for CEOs that show behavior and attitudes
consistent with high levels of narcissism.
2.4 Corporate Governance Strength
Corporate Governance Strength. “Corporate governance (hereafter: CG) is the system by which companies are directed and controlled” (Cadbury, 1992, p.14). This includes, for instance, the board, external auditors and ACs. In order to asses and improve the financial reporting quality, auditors and the AC have to cooperate (Cohen et al., 2002). On the one hand, “external auditors need effectively performing AC members in response to communicated areas of concerns, such as potential fraud and internal control weaknesses” (DeZoort, 1998, p.18). Contrarily, in order to protect shareholders, auditors have to provide information and assistance to the AC (DeZoort, 1998). According to Cohen, Krishnamoorthy & Wright (2002), such CG factors are considered as important by auditors, especially in the client acceptance phase and in an international context. Though, as mentioned by Gul et al., (2008) “the firm’s board is perhaps the most important single control mechanism that seeks to control management opportunistic behavior” (p.8). Furthermore, ACs play a critical role when it comes to ensuring the integrity of financial reports (in US companies) (Cadbury, 1992). In addition, to work effectively, all members of an AC need to be financially literate (DeZoort and Salterio, 2001). It is assumed that understanding the CG type in place helps auditors by assessing various client risks and therefore a more effective and efficient audit can be planned (Cohen et al., 2002).
Anytime the governance structure is weak, auditors should evaluate the resulting impact on the audit. Therefore, it is important for auditors to recognize that weak governance mechanisms are associated with financial fraud across a number of time periods and industries (Beasley et al., 2000). Abbot et al., (2003) suggest that improved oversight of the management-auditor relationship can be realized with greater independence and expertise of ACs. Also, knowledge differences between the AC members lead to systematic judgement differences of those members, concerning accounting disputes between auditors and management (DeZoort and Salterio, 2001). Previous literature suggests that “strong CG (including an independent AC) has the potential to increase audit effectiveness and efficiency by reducing (1) the auditor’s perceptions of client business risk; (2) the auditor’s control risk judgements for specific audit assertions; and (3) the amount of planned substantive testing” (DeZoort and Salterio, 2001, p.32). Thus, because the current or potential client risk may be affected by the CG strength, it is assumable that such factors will affect auditors’ risk judgments and clients’ acceptance decisions (Cohen et al., 2002). To be more specific, when executive narcissism is combined with
less-effective CG, organizational misconduct can increase (Benimahd et al., 2013). Hence, the following hypothesis are formulated.
H2a: Corporate Governance Strength will negatively affect the audit- and fraud risk perception of an
auditor.
H2ba: The relationship between CEO narcissism and the audit- and fraud risk perception of an auditor
will be negatively moderated by Corporate Governance Strength.
2.5 Co-variables
Auditor Experience. According to Knapp & Knapp (2001), the existing knowledge of an auditor contributes to the interpretation and understanding of information. They suggest that the knowledge changes when individuals get more experienced, which may influence the analytical performance. Shelton, Whittington & Landsittel (2001) argue that an individual’s experience level is important for the risk assessment process. “For analytical procedures to be beneficial in enabling an auditor to effectively asses the risk of financial statement fraud, an auditor must possess the relevant knowledge required to uncover the fraud if it is present and the ability to “trigger”, or activate, that knowledge when needed” (Knapp & Knapp, 2001, p.28). Experience is an opportunity to gain knowledge (Hammersley, 2011), this due to the fact that auditors are required to attend in certain training opportunities during their career. So, the more experience (in years) auditors have, the more expertise they gain through training activities. According to Abdolmohammadi & Wright (1987) there is evidence that the effects of experience on decision-making are significant, especially in highly technical fields, where task complexity is explicitly considered, such as auditing. Hammersley (2011) also found a positive correlation between participation in an audit where fraud was discovered and the auditor’s total audit experience. Based on the assumption that the audit risk assessment may be affected by the auditors’ experience, the following hypothesis is suggested.
H3a: Auditor experience will positively affect the audit- and fraud risk perception of an auditor.
Auditor Gender
. Previous research shows that women are more conservative and risk averse than men (Cumming et al., 2015; Gul et al., 2008). Consequently, women are less inclined to take extreme risks and try to prevent losses (Ittonen & Peni, 2012). In addition, women are also more ethically sensitive when it comes to dilemma situations (Cumming et al., 2015; Owhoso, 2002). According to Breesch & Branson (2009), there are differences between men and women in the way information is processed. They state that women tend to use a holistic view, and process all information, where men process information more selectively. Also, in contrast to men, women appear to be more likely to recognizedisturbing ethical situations and presume them to be unacceptable (Owhoso, 2002). Siegrist, Gutscher & Earle (2005) found that gender was significantly associated with the judgment of risk. They concluded that females perceive higher risk than males. Hence, the following is suggested.
H3b: When auditor gender is female the audit- and fraud risk perception of an auditor will be higher.
Auditor Narcissism.
In the accounting profession, which is seen as a decision-making discipline, all (senior) managers base their decisions on accountants and auditor reports (Banimahd, Dilami & Javanmard, 2013). However, there are differences in how individuals process information and therefore how they perceive risks. For example, processing information can be influenced by auditor experience (Knapp & Knapp, 2001; Shelton et al., 2001) or auditor gender (Breesch and Branson, 2009). However, high levels of narcissism also have been linked to ‘distorted judgments’ of one’s capabilities (Brown, Akers & Giacomino, 2013), and risky decision making (Chatterjee & Hambrick, 2007; Rijsenbilt & Commandeur, 2013). According to Banimahd et al., (2013) the judgment and decision-making of those auditors may be influenced by their narcissistic personality motivations. This can cause overstated and biased reporting of firm performance. The behavior of narcissistic accounting professionals “will impact how they interact with each other and client personnel, as well as how they process, evaluate and summarize information and form conclusions” (Akers et al., 2014, p.171). Highly narcissistic auditors may reflect this on their auditing judgment about management and hence cause overstated and biased reporting of firm performance (Banimahd et al., 2013). This prediction is formulated as follows.H3c: Auditor narcissism will negatively affect the audit- and fraud risk perception of an auditor.
2.6 Upper Echelon Theory
According to the upper echelon theory, firm-related decisions are affected by managers experiences, values and their personalities (Hambrick, 2007). Because all individuals are different, managers operating in similar situations will act differently (Ham et al., 2017). Therefore, the upper echelons theory indicates that all actions and organizational choices of a firm’s top management are a reflection of their characteristics (Judd et al., 2016).
2.7 Conceptual Model
H2b- H2a - H1 + H3a+ H3b + H3c-3. Methodology
In this chapter the method for the executed research will be elaborated. First of all, a description of the experimental design, participants and data collection is given. Then, a specific introduction of the dependent variable, namely the audit- and fraud risk perception of an auditor, will be given. This is followed up by an explanation of the independent experimental variables and the co-variables. Finally, the validation for this research is given.
3.1 Experimental Design
The first step of our research was to collect data about the audit risk perception of auditors3. Therefore,
data was gathered by means of a 2 x 2 factorial design – between subjects. The experiment included manipulations for both CEO narcissism and CG strength at two levels (high or low). So, four different groups to be measured. To measure all elements, existing instruments from prior research of Johnson et al., (2013) have been adjusted and specifically developed for the Dutch situation. We mainly focused on the original audit risk judgement casus of Johnson et al., (2013) as it appeared to be the best case, after having considered alternatives. Thereby, the Johnson et al., (2013) case was adjusted to fit the Dutch experimental setting and to be able to measure two other experimental variables, namely CEO
3 In this paper the term auditor includes all respondents of the experiment that work in an audit firm. That means that both CPA and non-CPA that work in an audit firm are included.
Corporate Governance
Strength
Auditor Narcissism
Perceived Audit Risk
CEO Narcissism
Auditor Experience
narcissism and CG strength. Furthermore, the variable of motivation from Johnson et al., (2013) was excluded.
To study the experimental treatments of CEO narcissism and CGS on perceived audit risk, the following 2 (high/low narcissism) x 2 (high/ow CGS) between subjects factorial design was created.
Experimental Cases
CEO Narcissism
High
(Indicators of highly CEO Narcissism)
Low
(Indicators of low CEO narcissism) Corporate Governance Strength High (Indicators of strong CGS)
Case 1: Six AC meetings a year, no social (friendship) ties
and high CEO narcissism.
Case 2: Six AC meetings a year, no social (friendship) ties and Low CEO narcissism.
Low (Indicators of weak CGS)
Case 3: Two AC meetings a year, strong social ties a high
CEO narcissism.
Case 3: Two AC meetings a year, strong social ties a low
CEO narcissism.
Table 1: Experimental design and treatments
3.1.1 Participants and Data Collection
For the experiment a total of 190 Dutch auditors from KPMG, Deloitte, EY, PWC and Flynth in the Netherlands were selected. The participants were randomly and double blindly assigned to one of the four possible experimental conditions. All participants completed the experimental case individually. The data for the experiment were gathered with online surveys in Qualtrics. All participants were provided with background information about a construction audit client, its employees and its audit committee. This included the manipulations of CEO narcissism and CG strength. After reading the case, the participants were asked to fill in six questionnaires in order to measure their perceptions. This concerned perceptions about the level of audit- and fraud risk, CEO narcissism, CGS, the respondents’ risk-taking propensity, the respondents’ personal narcissism (based on the valid NPI-16 questionnaire of Ames, Rose & Anderson (2006) and a personality questionnaire.
3.2 Dependent variable
Perceived Audit Risk (PARTOT) - To measure an individual’s perception of the audit risk, questions were used based on the study of Johnson et al., (2013). The dependent measure to obtain the overall audit risk perception (PAR) was formulated as follows: My assessment of the risk of a material
misstatement in the 2017 financial statements of Groenendijk (before the start of the audit) is… Because “the risk of fraud is an integral component of overall audit risk” (Allen et al., 2006, p.159), also the perception of the fraud risk (PFR) was explicitly measured. This was formulated as follows: My perception of the fraud risk within Groenendijk is…
In line with research of Johnson et al., (2013) “The responses to the audit and fraud risk perception questions were measured on a seven-point scale, with endpoint labels of 1 = ‘very low’ and 7 = ‘very high’. The scale midpoint of 4 was labeled ‘moderate’” (p.210). Audit risk is "the risk that the auditor may unknowingly fail to appropriately modify his opinion on financial statements that are materially misstated" (e.g., as proxied by inherent risk and control risk) (AICPA, 2006, AU Section 312.02).
3.3 Independent variables
3.3.1 CEO Narcissism (CEON)
In business research, the Narcissistic Personality Inventory (NPI) of Raskin and Hall (1979) is a commonly used measure for narcissism (Olsen et al., 2014; Johnson et al., 2013). To determine the experimental condition for CEO narcissism in this study, we focus on the NPI items that “best matched up with the narcissism constructs of grandiosity and entitlement/exploitativeness” (p.210), in line with the study of Johnson et al., (2013). Therefore, two different proxies were included in the case scenarios. To begin with, the treatment of the high CEO narcissism. This case condition contained multiple person references (“I,” “me”, “my”) of the CEO during his dialogue with the auditor. The use of first-person singular reference(s), which reflects self-absorption (Chatterjee & Hambrick, 2007), has been used as a measure of narcissism (Johnson et al., 2013; Rijsenbilt, 2011). This is in line with prior research which has shown that first-person references are associated with high narcissism (Raskin & Shaw, 1988; Chatterjee & Hambrick, 2007; Rijsenbilt, 20110). Furthermore, the case situation contained a photograph of the board from the annual report in which the CEO was positioned prominently. A CEO photograph has been considered as a standard feature of an annual report (Chatterjee & Hambrick, 2007). However, not all companies include a CEO photograph and when one is included, the form and prominence varies (Olsen et al., 2014). In the low CEO narcissism case situation no first-person references were included and the CEO was not prominently positioned on the photo in the annual report.
3.3.1.1 Narcissism Manipulation Check
In order to validate the effectiveness of the manipulated experimental variable, respondents had to answer some statements in a short questionnaire after having read the case. For narcissism, participants gave their perception about the CEOs personality traits. In line with Johnson et al., (2013) “Participants gave their response to each statement on a seven-point scale where 1 = “strongly agree,” 4 = “neutral,” and 7 = “strongly disagree”. Example statements included descriptions of the manager as “demands
respect,” “has a high opinion of himself,” and “is even-tempered” (reverse-scored). Statements unrelated to narcissism were also included to disguise the nature of the manipulation check task” (p.210).
3.3.2 Corporate Governance Strength (CGSTR)
The experimental case condition for CGSTR is determined on the base of three different aspects. This condition consists of proxies for AC experience, AC diligence and AC-CEO social ties.
AC Experience
. The experience and expertise of an AC member is seen as an important aspect for the effectiveness of the AC (DeZoort, 1998). For instance, ACs with a financial expert enable AC members to perform their monitoring role better (Cassel, Giroux, Myers & Omer, 2012). DeZoort (1998) support this with evidence that experience in a general domain or specific task significantly affects the judgment of AC members’ internal control assessments. Also, Johnstone, Li and Rupley (2011) find that improvements in AC competence (financial expertise) are positive related with internal control material weakness remediation. According to Cassel et al., (2012) ACs with a financial expert are more capable to constrain earnings management and the companies’ financial statements are less likely to be restated. Additional results of DeZoort (1998) show that “experienced AC members made more consistent judgments, had higher self-insight, higher consensus, and higher technical content levels for additional items offered than did the members without experience”(p.18). Also, previous research found a link between AC expertise and the quality of financial reporting (Cassel et al., 2012).AC diligence.
Research of Cassel et al., (2012) suggests that auditors from Big N (Big 5 accounting) companies take clients’ CG mechanism into account when making client portfolio decisions. Cassel et al., (2012) found a positive association between AC diligence and financial reporting quality and Thiruvadi (2012) stated that increased AC diligence established good CG practices. In addition, diligence of the AC also leads to fewer restatements and good financial reporting quality. In addition, previous literature shows that companies in which fraud occurred had significantly fewer AC meetings (generally one per year), except in financial industry (Beasley et al., 2000). Since AC meeting frequency is an important component of its effectiveness, it has often been used as a proxy to measure AC diligence (Abbot et al., 2003). Therefore, AC meeting frequency was used to measure AC diligence.AC-CEO social ties
. Within current literature there is no unambiguous definition of independence. For instance, Rijsenbilt (2011) distinguishes between executive and non-executive (independent) directors. Wu and Li (2015) differentiate between inside and outside directors. They state, “A director is defined to be an “outsider” if he or she is not a shareholder, an employee, or a person who is associated with the firm’s management. The general belief among governance experts is that board independence is required for effective governance (Uzun, Szewczyk & Varma, 2004). Wu and Li (2014) found that CG and the quality of board monitoring can improve due to increased board independence. In addition, previous research found evidence that firms where financial statement fraud occurs, have less independent directors compared to non-fraudulent firms (Beasley, 1996; Wu and Li, 2014)Therefore, it is desirable to have a skeptical and independent board (Rijsenbilt, 2011). However, “If seemingly independent audit committees are socially tied to the CEO, then they cannot be considered truly independent” (Bruynseels & Cardinaels, 2014, p.116). In addition, Bruynseels & Cardinaels (2014) found that the quality of the audit committee’s oversight may be reduced and that CEOs tend to engage more in earnings management when there are social “friendship” ties4 between CEOs and the audit
committee. When those social ties are present, auditors are less likely to report internal control weaknesses or to issue going-concern opinions. In line with Bruynseels & Cardinaels (2014), we distinguish entirely independent audit committees from audit committees whose members have social ties with the CEO. These social friendship ties are defined non-familial, informal ties between individuals that arise from non-professional activities like golf clubs, charity organizations, country clubs or other social networks (Bruynseels & Cardinaels, 2014; Krishnan, Raman, Yang & Yu, 2011).
3.3.2.1 Corporate Governance Strength Manipulation
For the high CG strength situation, the AC consists of members with strong financial- and banking backgrounds, a lot of auditing experience and explicit experience in the construction industry. They meet six times a year and there are no social ties between the CEO and the AC. For the low CG strength situation, the AC consists of members with little or no audit-, financial- or banking experience. Though, there is an experienced executive from the construction branch in the AC. Furthermore, the AC meets twice a year and the extend of social ties is high, as the CEO has social friendship ties with two members of the AC.
3.4 Co-variables
The following co-variables are independent variables which are expected to influence the audit risk perception of an auditor but are not part of the experimental 2x2 design. For this study the effects of the co-variables auditor experience, auditor gender and auditor narcissism on auditors are included.
3.4.1 Auditor Experience (AUDEXPI). As stated by Bonner & Lewis (1990), “Most studies of expertise have divided subjects into groups of experts and novices on the basis of years of experience or tenure-based titles” (p.2). Information concerning auditor experience (general fraud and industry) will be gathered by means of a personality background questionnaire, based on prior research of Hammersley (2011). He states “General experience is typically measured as longevity as an auditor” (Hammersley, 2011, p.106). Further, also based on Hammersley (2011), industry experience is measured as the amount of time in a task over a career. “Similarly, fraud experience can be measured as the direct experience on audits on which financial reporting fraud was discovered, or as the amount of indirect experience received from training on fraud topics” (Hammersley, 2011, p.106).
4 Social ties which established through non-professional activities are labelled as ‘‘friendship’’ ties because these ties are originate from the CEO’s personal network.
To measure the total auditor experience an experience INDEX will be build. This will include the total years of audit experience (AUDYEARS), the achievement of a CPA title (CPA), the experience with fraud (FRAUDEX), the experience within the construction sector (INDUSTEX) and finally the functional level of the auditor (AUDRANK) will be taken into account. In general, the more experienced in terms of AUDYEARS the greater the chance an auditors becomes CPA. Since they have more years of experience, CPA auditors have gained more knowledge by training than auditors without CPA title. Hence, in this study the assumption was made that CPA auditors and auditors with relatively more years’ experience (AUDYEARS) have gained more indirect experience in the form of training, for example concerning fraud risk factors and risk assessments. Both FRAUDEX, INDUSTEX and RA/CPA will be measured as dummy variable in which 0 means absence or inexperienced and 1 means present or experienced. The minimum index score is 3 and the maximum score is 45. See below for an overview of the experience INDEX.
Table 2: Description of the weighting of the index variables
3.4.2 Auditor Gender (AUDGEN). The gender of every individual respondent will be measured as a dummy variable by means of a personality background questionnaire. The gender parameter for male is 0 and for female it is 1.
3.4.3 Auditor Narcissism (AUDNAR). Literature shows that narcissism appears to influence the way information is processed and how decisions and judgements are made (Akers et al., 2014; Brown et al., 2013; Banimahd et al., 2013). To explicitly measure personal differences of narcissism, a derivative of the Narcissistic Personality Inventory (NPI) of Raskin and Hall (1979) is used, which is a commonly used measure for narcissism (Olsen et al., 2014; Johnson et al., 2013,). This study is based on the valid NPI-16 questionnaire of Ames, Rose & Anderson (2005). In line with Johnson et al., (2013) this questionnaire contains fundamental aspects of narcissism, such as grandiosity and entitlement/exploitativeness. Within this questionnaire, every narcissism answer will account for one point, with a minimum score of 0 (not narcissistic) and a maximum score of 16 (highly narcissistic).
3.5 Control variables
In order to improve the accuracy of the results, this research will control for alternative variables that influence the audit risk perception of an auditor. Based on prior research, the following control variables were selected: auditor age (AUDAGE) and likability of the CEO (CEOLIKE). AUDAGE is measured
EXPERIENCE INDEX Score
Auditor Experience in Years # of years
CPA title Yes = 1, No = 0
Experience with Fraud Yes = 1, No = 0
Experience with construction industry clients Yes = 1, No = 0 Functional Rank (Sr-staff, Manager, Sr Manager, Partners/Director) 1-4
as continuous variable. Siegrist, Gutscher & Earle (2005) found that age was significantly associated with the judgment of risk. They concluded that older people perceived higher risks than younger people. The measure for CEO likability is a statement which controls for the influenced reaction of auditors on their perception of audit risk. The statement for CEO likability: “Michel van de Beek is likable” is similar to the research of Johnson et al., (2013). This statement was measured with a seven-point Likert Scale whereby “1 = totally disagree” and “7 = totally agree”. When an individual dislikes the behavior and actions (personality) of the CEO, this can affect the auditor’s audit risk perception. CEO likability is included as control variable because it significantly correlated in the narcissism manipulation check test.
3.6 Statistic model
With the data that was gathered through a 2 x 2 factorial design between subjects, a quantitative research was conducted by means of a multiple regression analysis. The choice for a multiple regression analysis has derived from the following logical reasoning. Both independent variables and the dependent variable consist of ordinal data. Therefore, an ANCOVA or ordered logistic regression seems obvious. However, perceived audit risk can be viewed as continuous variable since the data is gathered through a points Likert Scale. This is also the case for CEON and CGSTR, as they are measured with a seven-point Likert Scale to. Since both dependent and independent variables are assumed to have a continuous underlying concept, it is possible to carry out a multiple regression. This analysis is preferred above an ANOVA, this will give a better insight in the effect size and also the direction of the relation will come clear. The formula for this model is as follows.
PARTOTi = β0 + β1 * CEOLIKE + β2 * AUDAGE + CEON * β3 * CGSTR + β4 (CEON * CGSTR) + β5 * AUDEXPI + β6 * AUDNAR + β * AUDGEN + εi
β0 = constant coefficient βi = direction coefficient εі = error-term
Description variables
PARTOT Total perceived audit risk
PAR Perceived Audit Risk
PFR Perceived Fraud Risk
CGSTR Corporate Governance Strength
CEON CEO Narcissism
AUDEXPI (INDEX) Auditor Experience Index
AUDRANK Auditor Rank
CPA Possession of CPA title
AUDYEARS Audit experience in years
FRAUDEX Experience with prior fraud
INDUSTEX Experience with industry
AUDGEN Auditor Gender
AUDNAR Auditor Narcissism
AUDAGE Auditor Age
CEOLIKE CEO likability
RISKTAKE Risk-taking propensity score
adjPAR Adjusted Perceived Audit Risk
adjPFR Adjusted Perceived Fraud Risk
Table 3: Description of all research variables.
3.7 Validity, reliability and common method bias
During the development process of the experiment, the case materials were reviewed repeatedly and the experimental protocol was approved, both in order to increase the internal and external validity5. Hence,
we were able to build a realistic and understandable fundament for the case. Accordingly, the quality as a whole was increased and guaranteed. To filter and exclude for the remainder of irregularities and ambiguity the case was pilot tested with seven CPA auditors6. After the pilot the volunteers were briefly
interviewed and the purpose of the study (including treatments) was discussed. Their findings were verified through questionnaires and discussed by telephone, in order to check that the manipulations came out sufficiently7. Finally, to verify that the experiment was ethical and complied with all ethical
standards, the case was approved with a statement of no objection by the ethical committee of the University of Groningen (Appendix 1).
Prior to the experiment, all the participants were personally approached to create a high commitment from the participants. Two days prior to the launch of the experiment, all participants were reminded to achieve a high response rate. To enhance the reliability of the data, all participants had to fill in the survey within one week. Furthermore, the survey was intentionally set out on a Thursday, since Fridays are the most quite working days for auditors. To further increase the response also a reminder was sent out directly after the weekend. Also, a number of precautionary measures was added
5 Support was provided by Prof. Philip Wallage (VU), Prof. Arnie Wright (Northeastern University), Herman van Brenk (Phd VU), Dr. Yasemin Karaibrahimoglu, Prof. dr. Jim Emanuels and My Lan Lu (All RUG). 6 The development and pretest procedures led to minor modifications to the instrument.
7 A skype call with Prof. dr. Jim Emanuels, Dr. Yasemin Karaibrahimoglu, MSc. My Lan Lu, Prof. dr. Phillip van Wallage and a MSc student
to increase the overall validity of the experiment and to reduce for possibility of “common method bias”8. As stated by Podsakoff, Mackenzie, Lee & Podsakoff (2003), “Most researchers agree that
common method variance is a potential problem in behavioral research” (p.879). They argue, “systematic measurement error is a particularly serious problem because it provides an alternative explanation for the observed relationships between measures of different constructs that is independent of the one hypothesized” (p.879). Therefore, all respondents received the same information before the start of the case, including the following which is in line with Podsakoff et al., (2003).
The experimental cases were thoroughly tested to exclude for irregularities and ambiguity. Participants are explicitly told the experiment is anonymous, answers are processed with care
and personal information cannot be derived from provided answers. Respondents are told that gathered information is only used for statistical analysis. Hence, sincerity and honesty of respondents is stimulated, and so does the quality of their answers.
To reduce pressure on respondents and to decrease the possibility for socially desirable answers, the respondents were told, there is no ‘right’ or ‘wrong’ answer.
To limit the knowledge of the respondents as much as possible, the research title is presented vague and respondents received no information about the interest of the researchers. Therefore, the title was ‘Auditor decision making’ instead of ‘Auditors risk perception’.
A variety of answer options is given (seven-point Likert Scale, pre-defined answer options and open questions.
By launching the experiment on Qualtrics (online platform) we control for highly social desirable answers and lower accuracy as with face to face interviews.
Respondents were personally and individually approached for the study. This reduced the possibility of consultation between respondents and prevented for awareness within the department.
4. Results
In this chapter the results concerning all formulated hypotheses will be presented and explained. First of all, insight in the dataset will be provided through the descriptive statistics. This is followed by an explanation on the manipulation check test and the Cronbach’s alpha. Hereafter, the multicollinearity- and correlation analysis are discussed. Finally, the result of the multiple regression and the additional analysis will be discussed with regard to the postulated hypothesis.
8A common method bias can occur when variations in responses are caused by the instrument rather than the actual perceptions of the respondents which the instrument tries to uncover.
4.1 Descriptive statistics
Eventually, 125 of the 190 approached respondents participated in the experiment. This lead to a response rate of 65,8%. All responses were gathered within one week. The sample of 125 respondents consisted out of 98 were males and 27 females. Table 3 gives further insight in the equal distribution per case for the auditors’ gender, function and age.
Furthermore, table 4 presents a description of the sample based on the auditors’ functional rank. This includes the gender, age, years’ experience and possession of a CPA title. Accordingly, the average age was 34 years with a minimum of 23 and a maximum of 59. The average work experience was 11,3 years, with a minimum of 2 years and a maximum of 40 years. In total, 57 respondents (45,6%) were a CPA and 68 (54,4%) were not a CPA. The sample consisted of 64 (51,2%) Senior-staff auditors, 22 (17,6%) Managers, 9 (7,2%) Senior Managers and 30 (24%) Partners/Directors9. Remarkable here is
the fact that one senior staff member had 36 years of experience was 57 years old. Also, the only senior manager without CPA title was 59 years old. This strongly increases the average age and average experience per functional level. In addition, it is notable that the NPI-16 score increased almost parallel to the functional levels. In other words, the higher functional rank, the more narcissistic an auditor is. Finally, when considering the normality of the dependent variable, a normal distribution occurred. This is indicated by both kurtosis and skewness scores, which were looking good with values of respectively (ϓ = 0,105) and (𝑆 = −0,589).
Table 4: Experimental cell sizes and distribution
Sample description
Male Female Total Av.
Age Min. Age Max. Age CPA % CPA Mean Exp. Min. Exp. Max. Exp. Av. NPI-16 Senior-Staff 42 22 64 28.6 23 57 6 9.4% 6 2 36 3.91 Manager 21 1 22 31.4 27 40 15 68.2% 9.1 6 18 5.45 Senior Manager 9 0 9 39.0 31 59 8 88.9% 16.9 8 40 5.56 Partner/Director 26 4 30 45.8 37 58 28 93.3% 22.7 15 37 4.90 Total Sample 98 27 125 34.0 23 59 57 45.6% 11.3 2 40 4.54 Table 5: Total sample description of respondents by function and total.
9 1. Senior-staff (Sr. associate/assistant manager/junior manager), 2. Manager, 3. Senior Manager 4. Partner/Director Although differences in functionality names exist between firms, the functional allocation reconciles with comparable responsibilities at all participating firms.
Treatment distribution Narcissism CGSTR Respondents Male Female Sr.
Staff Manager Sr. Manager Partner/ Director Av. Age
Case 1 High High 31 26 5 15 6 4 6 34.1
Case 2 Low High 31 24 7 19 3 1 8 32.4
Case 3 High Low 32 26 6 15 5 3 9 34.6
Case 4 Low Low 31 22 9 15 8 1 7 34.9
4.2.1 Non-response
The most relevant deviation in the non-response sample, in comparison to the response sample, is the higher percentage of functional levels that are above Senior-Staff. Since the response rate is below 80%, this might form a treat for the validity of this research. Hence, this can cause a lower reliability and limit the interpretability of the results (Smit, 2015). The potential impact and implications will be discussed in the discussion chapter.
Non-response sample description
Male Female Total CPA % CPA
Senior-Staff 21 7 28 4 11,1%
Manager 11 3 14 11 78,6%
Senior Manager 7 2 9 8 88,9%
Partner/Director 14 0 14 14 100,0%
Total Sample 53 12 65 37 56,9%
Table 6: Total non-response sample description by function and total
Furthermore, the following table presents the descriptive statistics of all variables that were included in the correlation, multiple regression and additional analysis.
Descriptive statistics
Variable Mean Median SD Minimum Maximum
PAR 5.61 6 0.79 3 7 PFR 4.94 5 1.05 2 7 CGSTR 3.73 4 1.38 1 6,25 CEON 4.72 4.5 0.73 3,50 6,50 AUDGEN 0.22 0 0.41 0 1 Z-AUDEXP .00 -.37 3.55 -4.73 7.67 AUDYEARS 11.33 8 8.98 2 40 CPA 0.46 0 0.50 0 1 FRAUDEX 0.36 0 0.48 0 1 INDUSTEX 0.59 1 0.49 0 1 AUDRANK 2.04 1 1.25 1 4 AUDNAR 4.54 4 2.63 0 13 AUDAGE 33.99 30 8.98 23 59 CEOLIKE 3.66 4 1.19 1 6 adjPAR 5.61 5.6 .014 2.67 11.96 adjPFR 4.93 5 .014 1.67 10.25
Table 7: Descriptive statistics of dependent-, independent-, co- and control variables. This includes continuous
and dichotomous variables. Also adjPAR and adjPFR is included for the additional analysis.
4.2 Missing values
In general, there are a few options for dealing with missing values. This can be done by excluding the concerned respondent, excluding respondents for the particular item or by a replacing the value with the
mean. For this study the missing values10 for age were replaced by a prediction. The predictions are
based on the factors that appear within the peer group. This includes average age, average years of experience and the functional level. To make an accurate prediction for the auditors’ years of experience, the age, functional level and the achievement of a CPA title are included11.
4.3 Manipulation check tests
To check if the manipulation for CEO narcissism and CGS were perceived as intended, an independent sample t-test was done. In total, six of the nine questions are related to the CEO narcissism manipulation and the other three are control items. For five out of the six manipulation statements a significant difference was indicated between the high and low narcissism condition. Hence, it is assumed that the manipulation for CEO narcissism was effective and successful. In addition, the Cohen’s d12 is calculated
to measure the effect size for each item. The Cohen’s d shows moderate size effects for all the manipulation statements, except for the integrity statement. Hence, all six items were used for the narcissism manipulation (Table 8).
Table 8: Manipulation check test CEO narcissism.
*, **, ***, **** represent significance level of 0.10, 0.05, 0.01 and < 0.001, respectively (one-tailed).
10 The missing value of age is based on the average age of peers in the same functional level. For example; a partner with 15 years of experience is estimated to be 39 because of the following. Partners with 15 to 17 years of experience are 39 year old on average. In contrast, the average age of a partner with 22 years of experience is 46. Also, the average experience of respondents between 37 and 42 years of age, is 16.7 years. Furthermore, a partner with 15 year experience is assumed to be one of the youngest partners when compared with its peers. 11 Experience is predicted on average age, the functional level of peers and achievement of a CPA title is included. For instance, auditors without CPA cannot become partner.
12 Cohen’s d is defined as the difference between two means divided by a standard deviation.
Mean (SD)
High CEO Narcissism
Low CEO Narcissism
Manipulation Check statement: Michel van Beek… (n=63) (n=62) t
Effect size (d)
1. is competent. (Control) 4.90 5.05 -.660 (0.255) 0.00
(1.34) (1.08)
2. demands respect. (Entitlement) 5.38 4.9 2.243 (0.014)** 0.41
(1.17) (1.21)
3. is a team builder. (Grandiosity: reverse-scored) 3.17 3.85 -2.674 (0.004)*** 0.58 (1.36) (1.48)
4. has been successful in his career. (Control) 4.98 5.02 -.164 (.435) 0.00 (1.16) (1.02)
5. has a high opinion of himself. (Grandiosity) 5.57 4.79 4.040 (0.000)**** 0.73 (0.98) (1.18)
6. is even-tempered. (Grandiosity; reverse-scored) 3.35 3.73 -1.902 (0.029)** 0.35 (1.17) (1.04)
7. is assertive. (Entitlement) 5.22 4.84 1.920 (0.028)** 0.34
(1.10) (1.13)
8. is likable. (Control) 3.41 3.9 -2.355 (0.010)*** 0.00
(1.27) (1.05)
9. has a high integrity. (reverse-scored) 3.98 3.98 0.002 (0.499) 0.00 (0.89) (0.71)
For the manipulation of CGSTR four statements were included. The results of the manipulation check show that all four statements were perceived by respondents as they were intended to be. Also, the size effects indicate a large effect for the manipulation. Accordingly, the manipulation is considered as effective on the base of the independent sample t-test. Hence, all four items were used in the construct for CGSTR (Table 9).
Table 9: Manipulation check test CGSTR.
**** represent a two-tailed significance level of < 0.001.
The following table shows what impact the CEO narcissism manipulations has on how it is perceived as an audit- and fraud risk increasing factor. The effects size is labelled as low, medium or high for respectively 0.2, 0.5 and 0.8 (Cohen, 1988). Indicating a small to medium effect for both audit- and fraud risk.
Mean (SD)
High Narcissism Low Narcissism
Dependent measure (n=63) (n=62) t Effect size (d)
5.78 5.44 2.458 (0.008)*** 0.29
1. Overall Audit Risk (.63) (1.58)
5.13 4.74 2.071 (0.020)** 0.38
2. Overall Fraud Risk (1.01) (1.07)
Table 10: Perceived narcissism effects on audit risk perception. **, *** represent significance levels of 0.05 and < 0.01 (one-tailed)
4.4 Cronbach’s Alpha
In this study the degree of CEO narcissism, CGSTR and perceived audit- and fraud risk are formed by multiple factors. For instance, the experimental condition for CEO narcissism is based on how the CEO is positioned on a picture of the board and on his language style. To measure the audit- and fraud risk perception with regard to CEO narcissism, the respondents were asked nine different questions. All questions were measured on a seven-point Likert Scale13. Therein, 1 means “very low” or “totally
13 Likert Scale whereby “1 = very low or totally disagree” and “7 = very high or totally agree”
Mean (SD)
High CGS Low CGS
Manipulation Check statement: The Audit
Committee… (n=63) (n=62) t
Effect size (d)
1. is independent. 4.63 2.87 6.399 (0.000)**** 1.03
(1.86) (1.58)
2. has many financial- and audit knowledge. 4.92 2.98 7.026 (0.000)**** 1.27 (1.50) (1.58)
3. spends a lot of time on monitoring tasks. 4.15 2.83 6.298 (0.000)**** 1.14 (1.02) (1.30)
4. its overall quality is high. 4.47 3.02 6.390 (0.000)**** 1.15
