Which factors do influence perceived audit risk?

(1)

Which factors do influence perceived audit

risk?

Master thesis University of Groningen Faculty of Economics and Business

MSc Accountancy

Abstract

This research examines the influence of CEO narcissism and corporate governance strength on perceived audit risk using an experimental design. Furthermore, I will explain the effects of risk-taking propensity and gender differences on the auditor’s risk assessment. A number of 125 auditors (CPAs/RAs) and other employees of audit firms had participated in an experiment in where CEO narcissism and corporate governance strength were manipulated. The results indicate that CEO narcissism and perceived audit risk are significant related.

January 20, 2018

Douwe Somsen S3239039

Supervisor: Emanuels, J.A. Second assessor: Karaibrahimoglu, Y.

(2)

1. Introduction

As a result of specific regulation updates (e.g. ISA 300, 315 and 330, PCAOB 2007 AS No. 5, and SAS Nos. 104-111), the importance of risk assessments in auditing is reemphasized (Fukukawa & Mock, 2011). In recent years, the audit is moved to the more risk-oriented practice (Fukukawa & Mock, 2011). Inaccurate conclusions may arise if the auditor does not assess the client risks in a properly way (Fukukawa & Mock, 2011). The most important input factor in the auditor’s risk assessment is the “tone at the top” at the client (Cohen, Krishnamoorthy & Wright, 2002; Hammersley, 2011; Patelli & Pedrini, 2015). Based on COSO (2013), “tone at the top” is a reflection of the attitudes, values, and individual characteristics of the top management. Because there is a relation between the CEO’s characteristics and organizational outcomes, it is interesting to study the specific impact of CEO narcissism (Chatterjee & Hambrick, 2007). 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 (Judd, Olsen & Stekelberg, 2017). In this study, I examine whether CEO narcissism affects auditor’s risk assessment. Much research related to narcissism is explored in the field of psychology, education, organizational studies, and strategy (Brown 1997; Campbell et al., 2000; Wallace & Baumeister 2002; Bergman et al., 2010; Westerman et al., 2011). These studies focus on the relation between CEO narcissism and the CEOs firm (e.g. the shareholders of that particular firm), based on the “rent extraction” theory or “efficient contracting” theory. From this point of view, this study contributes by studying the influence of CEO narcissism on the risk assessment of another third party, the external auditor based on the agency theory. There are two papers, Johnson et al. (2013) and Judd, Olsen & Stekelberg (2017), that are quite closely related to the research field of this research paper. These studies examine the relation between the auditors’ response to clients that have a more narcissistic CEO and the audit fees. These studies suggest that the audit fees increases when a firm has a more narcissistic CEO (Johnson et al., 2013; Judd, Olsen & Stekelberg, 2017). My research contributes to this by measuring the effect of CEO narcissism on the auditor’s risk assessment more directly using an experimental design.

There is a positive relation between an overconfident manager and significant misreporting (Schrand & Zechman, 2012). Continuing on this, strong corporate governance declines the likelihood of fraudulent financial reporting (Persons, 2005).Therefore, it is interesting to study the influence of corporate governance strength on the auditor’s risk assessment besides the effect of CEO narcissism. The task of the audit committee is to oversee the financial reporting process of the firm (Klein, 2002). In addition to the monitoring role of the audit committee, the external auditor has a monitoring role on behalf of the shareholders and that is an important aspect of corporate governance, because they verify the credibility of firm’s management accounting information (Lin & Liu, 2009). Because of that, it is

(4)

notable that the research that explores the relation between corporate governance strength and perceived audit risk is limited. Krishnamoorthy, Wright & Cohen (2002) argue that the audit committee should play a greater role than it currently does in ensuring the quality of the financial reporting process. Prior research that studied the relation between corporate governance strength and the auditor’s risk assessment, had focused on the effect of corporate governance strength on audit pricing (El’Fred & Divesh, 2008; Abdul Wahab, Zain & James, 2011; Bortolon, Sarlo Neto & Barreto Santos, 2013; Narcisa, 2016). My study contributes to this prior research by examining the specific relation between corporate governance strength and perceived audit risk. Contessotto & Moroney (2014) study the relation between audit committee effectiveness and audit risk. This study suggests that there is a negative relation between audit committee effectiveness and audit risk (Contessotto & Moroney, 2014). My study contributes to the study of Contessotto & Moroney (2014) by exploring the direct effect of the auditor’s perception of the influence of a strong/weak audit committee on the auditor’s risk assessment, because of the experimental design.

The results of my study do have implications for firm’s insiders (e.g. management and specific for the CEO) by creating awareness of the influence of CEO narcissism and the effect of corporate governance strength on the risk assessment of the auditor. Furthermore this study has implications for the firm’s outsiders (e.g. owners/shareholders of the firm and the external auditor) by explaining how perceived audit risk is affected by a narcissistic CEO and corporate governance strength.

The goal of this study is to explore the influence of CEO narcissism and corporate governance strength on perceived audit risk in a more directly way using an experimental design. Based on the agency theory, I examine to what extent CEO narcissism and corporate governance strength do have an influence on the auditor’s risk assessment. So, the research question of this research is:

To what extent do CEO narcissism and corporate governance strength influence the level of perceived audit risk?

The findings of this study were based on an experiment in which 125 Dutch auditors and other Dutch employees of audit firms responded to the questionnaires. The results of this experiment indicate that CEO narcissism is significantly positively related to perceived audit risk. The remaining variables that were studied are not significantly affecting the auditor’s risk assessment.

The remainder of the paper proceeds as follows. Section 2 provides the theoretical background and hypothesis development of this study. Section 3 provides the methodology and the results are described in section 4. The last section contains the conclusion and discussion.

(5)

2. Theoretical background and hypotheses development

2.1. Theoretical framework

From the agency theory, an agent could undertake actions that are not in the best interest of the principals (Eisenhardt, 1989). If this is the case, the agency problem arises, because two parties (e.g. management and owners) have different interests (Jensen & Meckling, 1976). The agency theory consists of two problems that could occur in an agency relationship: (1) the agency problem that arises when the goals of the principal and agent conflict and (2) it is difficult or expensive for the principal to verify what the agent is actually doing (Eisenhardt, 1989). The common aspect of research in agency theory is that there is a contract between the principal and the agent to monitor the agent (Eisenhardt, 1989).

A high level of narcissism could lead to dysfunctional behavior (Brown et al., 2009) or to commit acts of misconduct, like fraud, in order to achieve promotion, wealth or power (Williams, Nathanson & Paulhus, 2010). In this context, a more narcissistic CEO is more likely to make decisions with a high level of risk in it and to undertake fraudulent actions (Judd, Olsen & Stekelberg, 2017). The audit committee is responsible for monitoring the control environment and the associated “tone at the top” in the organization (Riner, 2017). In addition to that, organizations need contractual agreements to ensure that management acts in the best interest of the owner (Lin & Liu, 2009). Another actor then the audit committee, the external and independent auditor, needs to monitor these arrangements on behalf of the owners (Lin & Liu, 2009). The monitoring role of auditors includes reviewing the financial statements and look at the degree of aggressiveness and conservatism of the accounting policies and estimates (Rainsbury, Bradbury & Cahan, 2008). Management integrity significantly affects the risk assessment of the auditor (Iyer and Reckers, 2007). However, prior research did not examine the existence and significance of the response of auditors to individual characteristics of executives (Lauck, Rakestraw & Stein, 2017). Because the audit committee needs to monitor the narcissistic CEO because of agency problems and the external auditor needs to take these factors into consideration while assessing the audit risk, this research is based on the agency theory.

2.2. Perceived audit risk

The function of an external financial audit is to evaluate whether the information in the financial statements is free of material misstatements to provide assurance for financial reporting (Chow, 1982; Fan & Wong, 2005). The PCAOB defines audit risk as: “the risk that the auditor expresses an inappropriate audit opinion when the financial statements are materially misstated” (PCAOB 2010, paragraph 4). Audit risk consists of three aspects: inherent risk, control risk, and detection risk (Arens, Elder & Beasley, 2013). Perceived risk consists of two components: the perceived uncertainty of outcomes, and the perceived importance of negative consequences that are affiliated with the

(6)

outcomes of a choice (Bauer, 1960; Ross, 1975). Accordingly to these definitions, the definition of perceived audit risk is: the perceived uncertainty and importance of negative consequences if the auditor expresses an inappropriate audit opinion when the financial statements are materially misstated.

Fraud risk and audit risk are quite closely related. Fraud risk is the risk that a form of fraud occurs at the auditor’s client. Based on SAS. No. 82, there are two types of fraud that may result in financial statement misstatements: (1) fraudulent financial reporting and (2) misappropriation of assets. Auditor’s assessment of fraud risk involves: gather information to identify risks of material misstatement due to fraud, assess these risks based on the evaluation of the entity’s programs and controls, and respond to the results (SAS No. 99, 2002). Among researchers, there is the concern that auditors are not responding in an appropriately way to fraud cues (Asare & Wright, 2004; Hoffman & Zimbelman, 2009; Hammersley et al., 2011). An audit failure concerning fraud could occur, if the auditor fails to adjust low fraud risk assessments when the evidence indicates otherwise (Payne & Ramsay, 2005). After pre-engagement acceptance procedures, auditors are unlikely to change their beliefs that fraud risk is low (Wilks & Zimbelman, 2004). As stated before, audit risk is defined as: “the risk that the auditor expresses an inappropriate audit opinion when the financial statements are materially misstated” (PCAOB 2010, paragraph 4). Therefore, fraud risk and audit risk are quite closely related, because if the fraud risk is not detected in the risk assessment, then this directly affects the audit risk.

Behavioral studies examine the effect of executive characteristics on the risk assessment of the auditor (Beaulieu, 2001; Iyer & Reckers, 2007; Johnson et al., 2013). Lauck, Rakestraw & Stein (2017) argue that the evaluation of control environment and tone-at-the-top are important aspects of auditor’s risk assessment. A key element of the auditor’s risk assessment is the firm’s management integrity (Turner et al., 2003). Management integrity is the explanation for the risk structure of a client and this provides the base of the internal control environment (Kizirian, Mayhew & Sneathen, 2005). More specifically, management integrity is negatively related to auditor risk assessment (Kizirian, Mayhew & Sneathen, 2005). Because narcissistic characteristics of CEOs do have a strong influence on the ethical climate of their firms, CEO narcissism is an important aspect for auditors to take in consideration while they assess the audit risk (Judd, Olsen & Stekelberg, 2017).

2.3. CEO narcissism

Narcissism is originated by a Greek mythology in where Narcissus, a young man who fell in love with his own reflection, died as a result of his self-absorption (Ellis, 1898). Narcissistic persons are persons who admire themselves, who magnify themselves, and who see other persons as an extension of themselves (Freud, 1957). Positive characteristics such as authority, self-sufficiency, and superiority

(7)

predict effective leadership, promote organizational performance and attract loyal employees (Maccoby, 2000; Hogan & Kaiser, 2005). Positive characteristics relate to a moderate level of narcissism (Maccoby, 2003). The other side of narcissism is defined as a psychological construct that is defined as a sense of self-importance and uniqueness, entitlement, self-absorption, self-admiration, arrogance, exhibitionism, exploitativeness, and vanity (Emmons, 1987; American Psychiatric Association, 2000; Resick, Whitman, Weingarden & Hiller, 2009). These kinds of narcissism are elevated levels of narcissism (Maccoby, 2003). An example of elevated levels of narcissism is destructive narcissism as defined by Lubit (2002). These elevated levels of narcissism could lead to dysfunctional behavior (Brown et al., 2009) or to commit acts of misconduct, like fraud, in order to achieve promotion, wealth, or power (Williams, Nathanson & Paulhus, 2010). If a person has narcissistic characteristics, then this will have an impact on the interaction with people within and outside the firm (Akers, Giacomino & Weber, 2014). In addition to that, this has an impact on the way this person processes, evaluates and summarizes information and conclusions (Akers, Giacomino & Weber, 2014).

It is more likely that managers with greater narcissistic personality traits inflate reported earnings when there are positive social status implications such as praise, approval and confirmation (Hobson & Resutek, 2012). In addition to that, there is a positive relation between CEO narcissism and reported fraud (Rijsenbilt, 2011; Rijsenbilt & Commandeur, 2013).

There is little research that explores the direct relation between a narcissistic manager and the auditor’s risk assessment. Johnson et al. (2013) find that personal characteristics of executives affect the auditor’s fraud risk judgments. More specific, an increase in CEO narcissism lead to an higher assessment of fraud risk by the auditor (Johnson et al, 2013). Lauck, Rakestraw & Stein (2017) argue that the ability of the auditors to correct misstatements in the report or mitigate audit risks reduces as the perceived risk increases through the possible dysfunctional effects of the tone-at-the-top in an organization. Thereby, a narcissistic CEO is more likely to make risky decisions and to undertake fraudulent actions (Judd, Olsen & Stekelberg, 2017). This ends in the following hypothesis:

H1: CEO narcissism is positively related to perceived audit risk

2.4. Corporate governance strength

Corporate governance is another factor that influences the auditor’s risk assessment, so it is interesting to analyze the effect of corporate governance strength on perceived audit risk besides the influence of CEO narcissism (Cohen & Hanno, 2000). Berle & Means (1932) define corporate governance as many institutions that affect the process through which quasi-rents are distributed. Shleifer & Vishny (1997) define corporate governance as how the suppliers of finance get themselves a return on their

(8)

investment. More specific, corporate governance refers to how corporate objectives are established; the daily operations are managed; the interests of stakeholders are considered; they align behavior and activities that companies operate in a secure and correct manner; they protect the interest of their clients (Fernández-Gámez; García-Lagos & Sánchez-Serrano, 2016).

An important aspect of corporate governance in relation to the quality of accounting information is the composition of the board and especially the composition of the audit committee (Fama & Jensen, 1983). Auditors place high weight on the firm’s internal audit function to prevent and detect fraudulent reporting, because auditors’ familiarities with the internal audit role in reducing audit risk (Goodwin and Seow, 2002). The task of the audit committee is to oversee the financial reporting process of the firm (Klein, 2002). The audit committee plays a key role in improving the financial statement integrity (Blue Ribbon Committee, 1999). Contessotto & Moroney (2014) evolve an appropriate definition of an effective audit committee: “an effective audit committee is independent of management, ensures financial reporting integrity, appoints independent external auditors, monitors the audit process and oversees the risk management process” (p. 395). An effective audit committee declines the probability of fraudulent reporting (Beasley, 1996), fewer earnings restatements (Abbott & Parker, 2000), improved financial reporting integrity (Klein, 2002), reduced earnings management (Bédard et al., 2004), and the appointment of higher-quality auditors (Chen et al., 2005).

Audit committees enhance the effectiveness of internal and external auditors (Simnett et al., 1993). To enhance the effectiveness of the audit committee, meetings between the audit committee and external auditors without management can enhance objective exchange of information and reduce information asymmetry between the auditors and management (Raghunandan et al., 2001; Cohen et al., 2007). The audit committee reduces perceived audit risk and this decline in perceived audit risk is even larger when there is a higher frequency of meetings and a higher attendance at these meetings of audit partner attendance (Stewart & Munro, 2007). Strong internal monitoring as the basis of corporate governance reduces the risk assessment of the auditor (Boo & Sharma, 2008). Auditors make a more favorable audit risk assessment if a company has adopted strong corporate governance measures in a non-mandated environment (Sharma et al., 2008). An effective and independent audit committee can reduce aggressive financial reporting (Cohen et al., 2011). Thereby, the effective audit committee can support the auditor when discussing financial reporting issues with management (Cohen et al., 2011). DeZoort & Salterio (2001) and Cohen et al. (2004) state that strong corporate governance could reduce the perception of business and audit risk of the external auditor. This ends in the following hypothesis:

(9)

2.5. Auditor’s risk-taking propensity

Risk-taking propensity is a relevant individual difference variable to consider in audit contexts (Weber, 1978). Because of that, it is interesting to analyze the effect of auditor’s risk-taking propensity on perceived audit risk besides the effect of CEO narcissism and corporate governance strength. The definition of risk propensity is the current tendency of an individual to take or avoid risks (Sitkin & Pablo, 1992). Appelt et al. (2011) state that risk-taking propensity is the tendency to engage in behavior that bears the chance of losses as well as gains. In economic research, risk-taking propensity is often the estimation of the choice between monetary lotteries with varying probabilities of gains and/or losses (Holt & Laury, 2002). Risk attitude consists of two types: averse and risk-seeking (Kahneman & Tversky, 1982). In economic theories, the known probability distribution is an important characteristic of risk attitude (Mellers, Schwartz & Cooke, 1998). Cognitive biases are the reason for differences in the level of risk perception among the decision-making process of individuals (Barnes, 1984; Schwenk, 1984).

In audit context, Chung, Firth & Jeong-Bon (2003) argue that an external auditor has incentives to be more conservative. Thereby, risk-averse individuals who work together may suggest a more conservative work environment (Dong, 2017). A more conservative auditor tries to overrate the audit risk to mitigate the probability of an audit failure (Bigus, 2015). This ends in the following hypothesis:

H3: There is a negative relation between auditor’s risk-taking propensity and perceived audit risk

As stated before, the definition of perceived audit risk in this study is the perceived uncertainty and importance of negative consequences if the auditor expresses an inappropriate audit opinion when the financial statements are materially misstated. This is even more likely in doing the audit of a firm with a narcissistic CEO, because a narcissistic CEO is more likely to make risky decisions and to undertake fraudulent actions (Judd, Olsen & Stekelberg, 2017). As a result of this, the auditor will be more risk-averse in an audit with a narcissistic CEO, because inaccurate conclusions may arise if the client risks are not assessed properly by the auditor (Fukukawa & Mock, 2011).

Levin, Schneider & Gaeth (1998) describe three types of framing effects: (1) attribute framing effects, (2) goal framing effects, (3) risky choice framing effects. Risky choice framing effects arise when the propensity to take a risk depends on whether the potential outcomes are framed in a positive way or framed in a negative way (Levin, Schneider & Gaeth, 1998). Based on the increased risk caused by the narcissistic CEO, there is the negative framing effect of the auditors. The auditor’s risk-taking propensity and so the perceived audit risk will be affected if there is a narcissistic CEO

(10)

involved in the audit, because of the negative framing effect caused by the narcissistic CEO. Because of this negative framing effect, the auditor is more risk-averse, so the auditor’s risk-taking propensity decreases and that will result in an increase in perceived audit risk. This ends in the following

hypothesis:

H4: CEO narcissism does have a negative effect on the relation between risk-taking propensity and perceived audit risk

2.6. Auditor’s gender

There is a difference in the way in which men and women collect and process information (Meyers-Levy, 1989). Therefore, it is interesting to analyze whether there are differences in perceived audit risk between male and female auditors. Barke, Jenkins-Smith & Slovic (1997) find that men and women that have the same understanding and knowledge of a particular risk assessment still differ in their risk perceptions. In addition to that, women are more careful, well-reasoned and more conservative in comparison to men (Breesch & Branson, 2009). Chung & Monroe (2001) argue that the audit judgment accuracy of men is lower in complex decision tasks in comparison to the audit judgment accuracy in simple decision tasks. In comparison to men, women have the same audit judgment accuracy in complex and simple decision tasks (Chung & Monroe, 2001). The audit process consists of four stages: planning, risk assessment, collecting audit evidence, and evaluating the results and issuing the report (Ittonen & Peni, 2012). In this audit process, there arise differences between men and women in planning, group decision-making, risk tolerance and overconfidence (Ittonen & Peni, 2012). This may have a direct impact on the audit opinion, as these are a function of the audit risks (Hardies et al., 2015). In general, men assess risks at a lower level than woman assess risks (Brody, 1984). This finding is in line with Diberardinis et al. (1984), who conclude that women are more risk-averse than men. In addition to that, women are more conservative in comparison to men (Byrnes et al., 1999). More specific to the audit context, female auditors are more risk-averse than male auditors in their judgment (Gold, Hunton & Gomaa, 2009). Continuing on hypothesis 4 in where I hypothesized that there is a positive relation between auditor’s risk-taking propensity and perceived audit risk, this ends in the following hypothesis:

(11)

(12)

3. Methodology

3.1. Research model

An experiment was developed and executed for the purpose of this paper. The experiment consisted of two experimental variables and two conditions per variable (2x2 design). One of the experimental variables was CEO narcissism and the other experimental variable was corporate governance strength. In the experiment, the participant had to judge about the extent of perceived audit risk (on a 1-7 scale from very low to very high) based on the planning phase of an audit. The design of this experiment was based on the experimental design of Johnson et al. (2013). However, their experimental variables were CEO narcissism and motivation to conduct fraud. The possible conditions of our experiment were stated in table 1. The participants of the experiment were randomly assigned to one of the four possible experimental conditions. This condition mitigated the potential bias of the researchers that were involved in this research. Furthermore, the participants individually completed the experimental case. Because this condition was met, we mitigated the probability that people could have influenced each other while responding to the questionnaire. Because of that, participants responded in an personal and honest way about what they thought and by this method, we received reliable responses to draw up the conclusions of this study.

Table 1: Manipulation variables

CEO narcissism C G s treng th Low High Low

Low CEO narcissism / Low corporate governance strength

High CEO narcissism / Low corporate governance strength

H

igh

Low CEO narcissism / High corporate governance strength

High CEO narcissism / High corporate governance strength

The participants were personally approached through the researchers. The participants were included in the mailing list after they confirmed their participation. The questionnaire was sent by email on 16-12-2017 and the participants had the time till 23-16-12-2017 to fill in the questionnaire, which is quite a short response time. The reason why we have chosen for a short response time was that people had less time to talk to each other about the experiment. If people had time to talk about the cases and questionnaire, then people who filled in the questionnaire could found out that they have participated in an experiment and could have talked about it. This could have influenced the results, because the manipulation has been less effective in that case. As follow-up procedure, a reminder was sent by email at 20-12-2017 to the participants who did not respond to the questionnaires to increase the response rate. The questionnaires were executed in Qualtrics and the cases were randomly assigned to participants. Qualtrics is a well-known and often-used tool to distribute questionnaires. Customers like

(13)

Microsoft, Yahoo and Yamaha used Qualtrics. Furthermore, academic users like the top 100 business schools used Qualtrics.1 The participants received a link in their email through which they could participate. This was the only way that participants could fill in the questionnaires, so all responses were received via the same way. The case and the questionnaire were drawn up in Dutch, so only Dutch people participated in the experiment. The Ethics Committee of the Faculty of Economics and Business at the Rijksuniversiteit Groningen has reviewed the data collection and data storage procedure of this research project. The Ethics Committee has issued a statement of no objections and approved the procedures of data collection and storage. This statement was required to execute the experiment for our research. As a result of this statement, we can prove that there were no risks for participants, that anonymity was assured, and that participants were not explicitly misled and that they were informed about participation and had the choice to decline such participation. Before the experiment was carried out, the experiment had to be pilot tested. As Aiman-Smith & Markham (2004) stated, a pilot test of the survey is needed to check the reliability of the measures and to catch remaining glitches in the survey. The cases and questionnaires were pilot tested through six participants who work at the Rijksuniversiteit Groningen and have their base in the audit profession. We executed the pilot test to see whether the cases and questionnaires were clear and understandable for potential participants. Furthermore, we wanted to know what the pilot test participants thought about the manipulations that were used in the cases, because we needed to ensure that these would succeed in the experiment. They concluded that the experiment was, at overall, of a sufficient level. They came up with a few small improvement points. These points were discussed in a session with the researchers and the participants of the pilot test and this led to minor modifications in the case texts and questionnaires.

The questionnaire proceeded as follows. First of all, the participant read one of the four cases which were randomly and evenly distributed. After that, the participant responded to two questions about whether he/she thought there was a chance on a material misstatement and fraud risk based on the seven-point scale. In these questions, the participants estimated the perceived audit risk and the fraud risk. This was followed by the CEO narcissism manipulation check in the form of seven theses about the CEO, such as whether the CEO was competent and whether the CEO demanded respect, based on the seven-point scale. In the next section, there was a corporate governance strength manipulation check in the form of four theses about the audit committee. The participant had to state whether he/she thought the audit committee was independent and their financial/audit expertise for example, based on the seven-point scale. These manipulations checks allowed me to see whether the

(14)

manipulations in the cases have succeeded. The next question considered nine points from which the participant chose to take the risky choice in a lottery above the guaranteed amount of money. Through this question, we could determine the risk-taking propensity of the participant. This was followed by eight choices through which we determined the degree of narcissistic personal characteristics of the participant, based on the NPI-16. The last section of the questionnaire consisted of questions about personal characteristics of the participant. Questions that were asked in this section, were questions such as the age of the auditor and whether the auditor has experience with fraud cases in the audit. In this study, auditor’s gender, years of audit experience and years of audit experience in a specific industry were used from this personal characteristics questionnaire.

3.2. Participants

Johnson et al. (2013) used a sample of 101 participants in their experiment. Because my study is comparable to Johnson et al. (2013), a number of 100 participants would be sufficient. Johnson et al. (2013) used a sample of auditors who were working at the same audit firm. For my study, auditors (CPA/RA), people who work at an audit firm, people who are studying to become an auditor, and people who work at an audit firm as accountant, but are not studying to become an auditor (e.g. not studying to become CPA/RA) were approached. CPA is an abbreviation of certified public accountant. This is a statutory title of qualified accountants in the United States. This is also the title of qualified accountants in several countries in the English-speaking world. CPAs have a license to provide accounting services to the public. The CPA-title is similar to the title of RA (registeraccountant) in the Netherlands. RAs are signed up in the accountant register of the Nederlandse Beroepsorganisatie van Accountants (NBA). The difference between auditors (CPA/RA) and other people that work at an audit firm is that auditors have the responsibility to apply the laws and regulations that apply to financial statement audits. Furthermore, auditors need to achieve required and sufficient training in a year to keep up with the developments in the audit profession. In other words, the difference between auditors (CPA/RA) and other people who work at an audit firm lies in the compliance aspect. This could have affected their responses on the questionnaire, because they could look in a different way to risks because of responsibility. If a much bigger/smaller part of the participants is an auditor than the participants who were not an auditor, then this could have affected the outcomes of this study. Because of that, there was a question in the questionnaire that asked whether the participant was an auditor or not. The participants worked at ‘big four’ firms, but also at ‘non-big four’ firms. The ‘big four’ consists of Deloitte, EY, KPMG and PwC. The function level of the participants varied from senior staff to partner. All of the participants worked at an department of an audit organization which was located in the Netherlands.

(15)

3.3. Measurement of variables 3.3.1. Perceived audit risk

In the study of Johnson et al. (2013), the participant got a question, after he read the case, about what he thought the risk of fraud was and he had to choose an score based on a seven-point scale, with labels of 1 = “very low” and 7 = “very high”. The scale midpoint of 4 was labeled as “moderate”. By this method, Johnson et al. (2013) had determined the perceived fraud risk. The same method was applied in our experiment to determine perceived fraud risk. Furthermore, this method was applied in our experiment to determine perceived audit risk. In our experiment, the participant got a question, after he read the case, about what he thought the audit risk was and he had to choose an score based on a seven-point scale, with the same scale distribution as Johnson et al. (2013). The dependent variable in my study is perceived audit risk. Therefore, the analysis is executed with participants’ judgments about perceived audit risk and auditors’ judgments of perceived fraud risk were not further analyzed in this study.

3.3.2. CEO narcissism

Raskin & Hall (1979) have developed the narcissistic personality inventory (NPI) to make narcissism part of a personality dimension, rather than a clinical syndrome. This NPI is meant to measure narcissism. On some adjustments, the NPI is nowadays the measure of narcissism in extensive research (Chatterjee & Hambrick, 2007). The measure used in psychology research related to narcissism is also often based on the NPI. This is based on a forty-item forced-choice method. In accounting research this seems not to be realistic, because CEOs will probably not answer honest to sensitive questions (Cycyota & Harrison, 2006). As a result, many studies in this research field have based their measure of CEO narcissism more or less on the method that Chatterjee and Hambrick (2007) used to measure CEO narcissism (Rijsenbilt & Commandeur, 2013; Spangler, Gupta, Kim & Nazarian, 2012; Brennan & Conroy, 2013; Gerstner, König, Enders & Hambrick, 2013; Kim, 2013; Olsen, Dworkis & Young, 2014; Patel & Cooper, 2014; Zhu & Chen, 2015; Oesterle, Elosge C. & Elosge L., 2016).

The indicators through which CEO narcissism was manipulated in the experiment is based on the CEO narcissism measurement indicators of Chatterjee and Hambrick (2007): (1) prominence of the CEO’s photograph in the company’s annual report, (2) CEO’s prominence in the company’s press releases, (3) CEO’s use of first-person singular pronouns in interviews, and (4) CEO’s relative cash and non-cash compensation. Chatterjee and Hambrick (2011) stated that the CEO’s use of first-person singular pronouns in interviews to be unreliable after the introduction of Sarbanes-Oxley in 2002. Judd, Olsen & Stekelberg (2017) followed Olsen et al. (2014) who did not include the CEO’s prominence of the CEO’s photograph in the company’s annual report. As a result, Judd, Olsen &

(16)

Stekelberg (2017) introduced a measure of CEO narcissism that is based on three factors: (1) CEO’s relative cash pay, (2) CEO’s relative non-cash pay, and (3) size and prominence of the CEO’s photograph in the firm’s annual report. However, the correlation of the measure of CEO narcissism over time of Chatterjee & Hambrick (2007) was higher in comparison than the measure of CEO narcissism over time of Judd, Olsen & Stekelberg (2017).

The case in our experiment should not have been too complicated and the participant had to finish the case in half an hour. Therefore, we wanted to use the indicators that were most obvious in our opinion. Because of that, we focused on the following two specific indicators in the case through which CEO narcissism was manipulated, based on Chatterjee & Hambrick (2007): (1) use of language and (2) size of CEO’s picture.

(1) The first indicator was the use of language. In the ‘low narcissism’ case, it was clear that the CEO was equal to the other employees of the organization and in the ‘high narcissism’ case, it was clear that the CEO was too confident. The way in which this was executed was the usage of CEO’s first-person singular pronouns in the case, based on Chatterjee & Hambrick (2007).

(2) The second indicator was the size of the picture of a CEO in the case. In the ‘low CEO narcissism’ case, there was a photograph of the team in where the CEO was one of the team and stood in the background and in the ‘high CEO narcissism’ case, there was a photograph of the CEO in where he stood in the middle and the rest of team was blurred in the background.

Table 2: Aspects of CEO narcissism

Variable CEO narcissism

Aspect 1 Use of language

Aspect 2 Size of CEO’s picture

The participants responded to nine theses based on a seven-point scale, with labels of 1 = “totally disagree” and 7 = “totally agree”. These theses were about what they thought was appropriate to the CEO based on the case that was assigned to them. Six of the theses were used in the case as manipulation measures and three of the theses were included as controls. The measure of CEO narcissism was based on the average score of the following six manipulation theses: whether the CEO (1) demands respect, (2) was a teambuilder, (3) had a high self-esteem, (4) is well-balanced, (5) is assertive and (6) has a high integrity.

3.3.3. Corporate governance strength

The following indicators were used in the case of the experiment to manipulate the relative strength of corporate governance: (1) audit committee experience, (2) audit committee diligence, and (3) “socialities” between audit committee and CEO.

(17)

(1) The audit committee experience involved experience in the financial, audit or another specific sector. Bull & Sharp (1989) found that audit committee members need sufficient knowledge of accounting/auditing to reach an independent opinion on matters that are presented to them. This is important, because the matters that are reported to them are frequently of a technical nature (Braiotta, 1986; Green, 1994). In the ‘low narcissism’ case, members of the audit committee did have little specific financial/audit experience and in the ‘high narcissism’ case, members of the audit committee had much specific financial/audit experience.

(2) The audit committee diligence indicated the number of audit committee meetings. Stewart & Munro (2007) found that the frequency of audit meetings is significantly associated with a reduction of the perceived audit risk. In the ‘low narcissism’ case, the audit committee only had two meetings in a year and in the ‘high narcissism’ case, the audit committee had six meetings in a year.

(3) The “socialities” between the audit committee and the CEO indicated whether the members of the audit committee and the CEO have known each other before they have started in their role as a member of the audit committee or as the CEO. Cohen et al. (2010) found that members of the audit committee may appear to be independent when the focus is on economic ties, but they could be less independent if the CEO can nominate members with whom he/she has prior significant social or professional ties. In the ‘low narcissism’ case, members of the audit committee and the CEO did not knew each other from another association and in the ‘high narcissism’ case, members of the audit committee and the CEO knew each other from another association and from there, they got a good connection to each other.

Table 3: Aspects of corporate governance strength

Variable Corporate governance strength

Aspect 1 Audit committee experience

Aspect 2 Audit committee diligence

Aspect 3 “Socialities” audit committee - CEO

The participants responded to four theses based on a seven-point scale, with labels of 1 = “totally disagree” and 7 = “totally agree”. These theses were about what they thought was appropriate to the audit committee based on the case that was assigned to them. Four theses were used in the case as manipulation measures. The measure of corporate governance strength was based on the average score of the following four manipulation theses: whether the audit committee (1) was independent, (2) had sufficient financial and audit knowledge, (3) spent enough time on their audit committee tasks and (4) had a good overall quality.

(18)

3.3.4. Auditor’s risk-taking propensity

The measure of auditor’s risk-taking propensity was based on Young (1985). This measure involved a standard two-outcome lottery. The participant had to respond to a question in where was asked from which point he would choose the lottery above the guaranteed amount of money to receive. The questionnaire offered nine options in where the most risk-seeking choice was ‘€ 500 guaranteed or 10% chance on € 1000 and 90% chance on € 0’. The most risk-averse choice was ‘€ 500 guaranteed or 90% chance on € 1000 and 10% chance on € 0’. The other choices were between the most risk-seeking choice and the most risk-averse choice.

3.3.5. Auditor’s gender

In the last section of the questionnaire, the participants responded to questions about their personal characteristics that were based on the questionnaire that was used in Johnson et al. (2013). In one of the questions, the participant had to answer whether he/she is a man or a woman.

3.3.6. Control variables

In the question about the auditor’s personal characteristics, there were also questions included about the audit experience of the participant. Furthermore, the auditor was asked how many years he/she got audit experience and whether this experience was industry-specific, in this case, in the construction industry.

The more experienced decision-maker is more adept than the less experienced decision-maker at choosing the appropriate judgment variables and recognizing the relationships among them (Johnson, 1977; Johnson et al., 1984; Dreyfus & Dreyfus, 1986). Donnelly et al. (2011) found that the auditor’s experience level is negatively associated with dysfunctional audit behavior. In addition to that, affective information, which should be irrelevant to the risk assessment, had significantly influenced the risk assessment of inexperienced auditors, but did not influenced the risk assessment of the experienced auditors (Bhattacharjee & Moreno, 2002). Based on these prior studies, which showed that experience influenced auditor’s risk assessment, I controlled for the effect of auditor’s audit experience.

The role of task-specific knowledge was not always considered in experimental designs. As a result of this, there were mixed results in whether audit experience had an effect on audit experience (Bonner, 1990). Auditors’ decision-making effectiveness could have been improved by task specific-experience, which provides the opportunity for the development of enhanced knowledge structures (Biggs et al., 1987; Shelton, 1999). Hammersley (2011) found that industry experience is an example of task-specific knowledge. Furthermore, Moroney & Carey (2011) argued that industry experience affects auditor judgment on multiple aspects. Therefore, I controlled for the effect of auditor’s industry experience.

(19)

3.4. Method

Multiple regression analysis was used to test the hypotheses that were developed in section 2. This analysis was executed, because I wanted to model the relationship between perceived audit risk based on the value of several independent variables (e.g. predictor variables). Each value of an independent variable was associated with a value of perceived audit risk. Another feature of regression analysis in this study was that it predicted the value of perceived audit risk based on the value of the independent variables. More specifically, this analysis helped to understand how the value of perceived audit risk changed when the value of an independent variable was varied, while the other independent variables were held fixed. The overview of variables is stated in table 4. The following formula is elaborated for the purpose of the regression analysis:

PAR = β0 + β1 * CEONAR + β2 * CGS + β3 * RTP + β4 * (RTP * CEONAR) + β5 * EXP + β6 * INDEXP + εi

In this formula, β0 represents the constant coefficient, βi represents the various coefficients of this study and εi is the disturbance term.

Table 4: Overview of variables

Overview of variables Dependent variable

PAR Perceived audit risk

Independent variable

CEONAR CEO narcissism

CGS Corporate governance strength RTP Auditor's risk-taking propensity Control variables

EXP Years of audit experience

(20)

4. Results

4.1. Participants

The participants in the experiment were employees of different audit firms (big-four and non-big-four). Furthermore, the participants varied in age and in years of experiences. The number of participants was 125. As stated in the methodology section, based on Johnson et al. (2013) a number of 125 participants was a representative sample. In general, this means that this study included a good representation of the population, because our sample was larger than expected. A number of 190 people had confirmed their participation in the experiment, so the response rate was 65,8%. This means that 34,2% did not respond to the questionnaire, but almost 2/3 of the total participants did respond to the questionnaire. In general, this seems to be an acceptable response rate. As stated by Fincham (2008), the editor and associate editors of the Journal have been considered a response rate of 60% in survey research as an acceptable rate. Furthermore, Prairie Research Associates considered a response rate on mail surveys between 50% and 70% as an acceptable rate.2 The sample consisted of a distribution of auditors from several audit firms, differences in audit experience and CPAs/RAs (45,6%) as well as auditors who were not CPAs (54,4%), which made the sample even more representative.

The information of participants that did not respond to the questionnaires is stated in table 5. The distribution in auditor’s gender and function level was comparable between these groups. There was only a difference in the distribution whether the participant was a CPA/RA. However, this did not significantly influenced the results of this study. As a result of that, the non-response information was comparable to the information of the participants that responded to the questionnaires. A more detailed description of the demographic information of participants was stated in table 5. After that, information about the participants who did not respond to the questionnaire was stated in table 6.

4.2. Missing values

The missing values of this study involved 16 out of the 5.546 (0,3%) responses to the questionnaires. The values that were missing and used in this study were stated in table 5 by the use of footnotes. One method to handling missing data could have been mean substitution (Wyner, 2007). In this case, no data would be lost because of missing data, because the mean of all participants on that particular question is measured to create a new estimated value (Wyner, 2007). However, this method would have distorted the standard deviation, because the highest and lowest values were brought into the distribution then (Silva & Zárate, 2014). This could have affected the data if, for example, the participants with the highest age or most years of audit experience did not respond to these questions. Newman (2014) concluded that bias and inaccurate standard errors are the major problems in handling

(21)

missing values. These problems that were outlined by Newman (2014), were mitigated through the use of pairwise deletion in handling the missing values. Through the use of this method, more data was preserved than when the listwise deletion method was applied (Wyner, 2007). It would be inefficient if the remaining responses were thrown away of people that did not respond to a few questions (Wyner, 2007).

Participant demographic information

(n = 125)

Table 5: Participant demographic information

Mean age (in years) 343

Gender (in percentage male) 78,4%

Mean audit experience (in years) 11,34

Number of CPAs (percentage) 57 (45,6%)

Number of auditors who work at a ‘big four’ firm (percentage) 115 (92%)5

Distribution of participants by professional rank (percentage)

Senior staff 61 (48,8%)

Manager 22 (17,6%)

Senior manager 13 (10,4%)

Director 8 (6,4%)

Partner 21 (16,8%)

Auditors/accountants who have experience as auditor in 72 (57,6%)

the manufacturing industry (percentage)

Mean years of experience as auditor/accountant in 3,886

the manufacturing industry (in years)

Non-response information

(n = 65)

Table 6: Non-response information

Gender (in percentage male) 81,5%

Number of CPAs (percentage) 38 (58,5%)

Distribution of participants by professional rank (percentage)

Senior staff 27 (41,5%) Manager 15 (23,1%) Senior manager 9 (13.8%) Director 2 (3,1%) Partner 12 (18,5%)

3_{Two participants did not respond to the question of what age they were.}

4_{Two people did not respond to the question how many years they were working at the organization.}

5_{Because the four master students, who executed the experiment, all have worked at a ‘big four’ firm, this percentage was quite high.} However, all of the four master students have worked at a different ‘big four’ firm.

(22)

4.3. Correlation, manipulation check and reliability

The Pearson’s correlation matrix (table 7) was executed to determine whether there was correlation between the variables. More specific, this test showed if there was multicollinearity between the variables. If there would have been a high level of multicollinearity between variables, then the variables did influence each other. The result of multicollinearity would have been that it could have affected the outcomes of the regression analysis through which the hypotheses were tested. Variance inflation factor (VIF) is known as a more precise way to measure multicollinearity. If the VIF exceeded the value of 10, then there was a high level of multicollinearity (Blumberg et al., 2005). The VIF was stated ahead in this section below each regression model. The highest VIF of the regression models in this study was 1,637. This means that the multicollinearity was not an issue in this study.

Table 7: Pearson's correlation matrix

Sig. (2-tailed) 1 2 3 4 5 6 7

Dependent

1. Perceived audit risk -

Independent 2. CEO narcissism 0,324** - 3. CG strength -0,149 -0,224* - 4. Risk-taking propensity -0,021 -0,049 0,078 - 5. Gender -0,060 0,063 -0,104 -0,043 - Control 6. Audit experience 0,291** 0,130 -0,055 -0,044 -0,161 -

7. Audit experience in specific industry 0,137 -0,045 -0,013 0,001 -0,244* 0,229* -

** = significant at the 1% level * = significant at the 5% level

A correlation value between variables of more than 0,800 or less than -0,800 was considered as multicollinearity (Field, 2013). As stated in table 7, there was no value that exceeded the threshold, so there was not a high level of multicollinearity between the variables in this study. This means that from this viewpoint, the outcomes of the regression analysis were reliable and not affected by multicollinearity between variables.

After the conclusion that there was no effect of multicollinearity between variables, it was important to analyze whether the manipulations in the experiment have succeeded. This manipulation check was executed, because this gave an understanding about whether the participants of the high CEO narcissism case also have experienced this as high CEO narcissism and the same applies to the low CEO narcissism case and the high/low CG strength cases. If the manipulation of the experiment did not succeeded, then participants did respond to questions that they did not understood in the way it was meant to be. If this would have been the case, then the outcomes of the questionnaires could be

(23)

less reliable. The conclusions that were based on these outcomes could then be misguided. If the indicator in the manipulation check was significant at the 0,05 significance level, then the manipulation of that specific indicator had succeeded. The first of the six indicators were indicators through which CEO narcissism was measured. The last three indicators were controls in the model. After that, a reliability analysis was executed. The reliability analysis was used to determine whether the different indicators may be combined into one new variable. In general, the Cronbach’s Alpha’s score of approximately 0,700 or higher was considered as acceptable in most research studies.

The manipulation check of CEO narcissism is stated in table 8. This is followed by a reliability analysis for CEO narcissism. The manipulation check of corporate governance strength is stated in table 10. This is also followed by a reliability analysis for CG strength.

Manipulation check – CEO narcissism

Table 8: Manipulation check 1

Sig. (2-tailed) CEO narcissism High

(n = 63)

CEO narcissism Low (n = 62)

t (sign 2-tailed) The CEO:

Demands respect 5,38 (1,17) 4,90 (1,21) 2,240 (0,027)**

Has a high self-esteem 5,57 (0,98) 4,79 (1,18) 4,040 (0,000)***

Is a team-builder 3,17 (1,36) 3,85 (1,48) - 2,674 (0,009)**

Is well-balanced 3,35 (1,17) 3,73 (1,04) - 1,902 (0,059)*

Is assertive

Has a high integrity

5,22 (1,10) 3,98 (0,89) 4,84 (1,13) 3,98 (0,71) 1,920 (0,057)* 0,002 (0,999) Is competent (Control) Is successful (Control) Is sympathetic (Control) 4,90 (1,34) 4,98 (1,16) 3,41 (1,27) 5,05 (1,08) 5,02 (1,02) 3,90 (1,05) - 0,660 (0,511) -0,164 (0,870) - 2,355 (0,020)** * = significance level of <0,10 ** = significance level of <0,05 *** = significance level of <0,001

As stated in the table above, five out of the six CEO narcissism manipulations did succeeded, because they were significant at the 0,05 significance level. Only the integrity manipulation of CEO narcissism did not succeeded. However, because five out of the six manipulations did succeed, the manipulation of CEO narcissism was successful. There were also three controls in the manipulation of CEO

(24)

narcissism: whether the CEO was competent, successful, and sympathetic. One of the controls, the CEO was sympathetic, was significant. Because of this significance, the CEO sympathy score should be included as a control variable in this study, so this will not affect the relationships as stated in the hypotheses. The Pearson’s correlation test was executed to test whether there is multicollinearity between a sympathetic CEO and another variable. This would have been the case if the Pearson’s correlation score exceeded the 0,800 or -0,800. The highest value of Pearson’s correlation involved whether the CEO is sympathetic and CEO narcissism and is -0,597. However, this did not exceed the threshold value. Therefore, whether the CEO is sympathetic could be included as a control variable.

Reliability analysis – CEO narcissism

Table 9: Reliability analysis 1 Sig. (2-tailed)

Cronbach’s Alpha 0,694

Cronbach’s Alpha based on standardized items 0,700

N of items 6

The six items through which CEO narcissism was manipulated, were combined into one variable, based on the reliability analysis that is stated above. The Cronbach’s Alpha was 0,694, so this did not exceeded 0,700. However, this score was close to 0,700 and this means that this was an acceptable level to be reliable. This was plausible, because in the high narcissistic CEO case, the scores of the three narcissistic characteristics were at a higher level (respect, self-esteem and assertive) than the scores of the three less narcissistic characteristics (team-builder, well-balanced and high integrity).

Manipulation check – CG strength

Table 10: Manipulation check 2

Sig. (2-tailed) CG strength High

(n = 63)

CG strength Low (n = 62)

t (sign 2-tailed) The audit committee:

Is independent

Has sufficient audit/ financial knowledge

Spends enough time on their AC tasks

Has a good overall quality 4,63 (1,49) 4,92 (1,50) 4,15 (1,02) 4,47 (1,21) 2,87 (1,58) 2,98 (1,58) 2,83 (1,30) 3,02 (1,33) 6,399 (0,000)*** 7,026 (0,000)*** 6,298 (0,000)*** 6,390 (0,000)*** * = significance level of <0,10 ** = significance level of <0,05 *** = significance level of <0,001

(25)

As stated in the table above, all of the corporate governance strength manipulations did succeeded, because they were significant at the 0,05 significance level. All of the four corporate governance strength manipulations were further analyzed.

Reliability analysis – CG strength

Table 11: Reliability analysis 2 Sig. (2-tailed)

Cronbach’s Alpha 0,882

Cronbach’s Alpha based on standardized items 0,887

N of items 4

The four items through which corporate governance strength was manipulated, were combined into one variable, based on the reliability analysis that was stated above. The Cronbach’s Alpha was 0,882 and this exceeded 0,700. This means that this was an acceptable level to be reliable. This was plausible, because from table 10, it was clear that there were big differences in scores on all four aspects between the high corporate governance strength case and the low corporate governance strength case.

4.4. Testing hypotheses

The next step was the analysis whether perceived audit risk was influenced by the independent variables. A multiple linear regression was executed to predict perceived audit risk based on CEO narcissism and corporate governance strength. Preliminary analyses were performed to ensure that there was no violation of the assumption of normality, linearity and multicollinearity. All of the hypotheses were stated in one direction, so the two-tailed significance level is divided by two to measure the one-tailed significance level. However, the significance level of the control variables was two-tailed. In the regression models that were stated below, the control variables were tested first after which the main effects were stated. The model through which hypothesis 3 and 4 were tested involved a two-way interaction effect, so in that model the control variables, main effects and the two-way interaction effect were stated consecutively.

(26)

H1: CEO narcissism is positively related to perceived audit risk

Table 12: Analysis H1

Sig. (1-tailed) (1.) (2.)

Control variables B SE B SE

Audit experience 0,024** 0,008 0,021** 0,008

Audit experience in specific industry 0,019 0,020 0,022 0,020

CEO sympathy -0,143** 0,057 -0,038 0,070 Main effects CEO narcissism 0,287** 0,116 Adjusted R-square 0,113 0,150 Highest VIF 1,056 1,601 * = significance level of <0,10 ** = significance level of <0,05 *** = significance level of <0,001

The positive relation between CEO narcissism and perceived audit risk was significant at the 0,05 significance level, as hypothesized. This means that by every increase of one unit in CEO narcissism, the perceived audit risk increased by 0,287. In other words, the results of the experiment did support hypothesis 1.

H2: Corporate governance strength is negatively related to perceived audit risk

Sig. (1-tailed) (1.) (2.)

Audit experience 0,024** 0,008 0,023** 0,008

Audit experience in specific industry 0,019 0,020 0,018 0,020

CEO sympathy -0,143** 0,057 -0,129** 0,059 Main effects CG strength -0,052 0,051 Adjusted R-square 0,113 0,114 Highest VIF 1,056 1,058 * = significance level of <0,10 ** = significance level of <0,05 *** = significance level of <0,001

The negative relation between corporate governance strength and perceived audit risk was not found to be significant at the 0,05 significance level. In other words, the outcomes did not support hypothesis 2.

(27)

H3: There is a negative relation between auditor’s risk-taking propensity and perceived audit risk Table 14: Analysis H3

Sig. (1-tailed) (1.) (2.)

Audit experience 0,024** 0,008 0,024** 0,008 Audit experience in specific

industry 0,019 0,020 0,019 0,02 CEO sympathy -0,143** 0,057 -0,143** 0,058 Main effects Risk-taking propensity (RTP) -0,010 0,042 Adjusted R-square 0,113 0,136 Highest VIF 1,056 1,637 * = significance level of <0,10 ** = significance level of <0,05 *** = significance level of <0,001

The relation between risk-taking propensity and perceived audit risk was not found to be significant at the 0,05 significance level. In other words, the outcomes did not support hypothesis 3.

H4: CEO narcissism does have a negative effect on the relation between risk-taking propensity and perceived audit risk.

Sig. (1-tailed) (1.) (2.) (3.)

Control variables B SE B SE B SE

Audit experience 0,214 0,070 0,185** 0,070 0,184** 0,071 Audit experience in specific

industry 0,064 0,070 0,077 0,069 0,076 0,070

CEO sympathy -0,169 0,068 -0,045 0,084 -0,046 0,085

Main effects

Risk-taking propensity (RTP) -0,001 0,067 -0,001 0,068

CEO narcissism (CEOnar) 0,209** 0,085 0,208** 0,086

Two-way interaction RTP * CEOnar -0,005 0,068 Adjusted R-square 0,113 0,143 0,136 Highest VIF 1,056 1,584 1,637 * = significance level of <0,10 ** = significance level of <0,05 *** = significance level of <0,001

The independent variables were standardized and combined to create a two-way interaction variable for the purpose of testing hypothesis 4. Interaction effects create excessive amounts of

(28)

multicollinearity, so all of the independent and control variables were standardized to reduce multicollinearity between these variables and the associated problem of missing possible statistically significant results because of that. As a consequence of multicollinearity the regression coefficient could have not been true or the regression coefficient could not have been significant, because the standard errors of the beta coefficients were likely to be large (Grapentine, 1997). The (standardized) control variables were tested in model 1. The (standardized) main effects and control variables were tested in model 2. Model 3 involved the (standardized) two-way interaction effect, main effects and control variables. The two-way interaction effect was not found to be significant. In other words, the outcomes did not support hypothesis 4.

H5: Male auditor’s perceived audit risk is lower as compared to female auditor’s perceived audit risk Table 16: Analysis H5

Males’ perceived audit risk (mean) 5,63

Females’ perceived audit risk (mean) 5,52

t (Sig. 1-tailed) 0,255

* = significance level of <0,10 ** = significance level of <0,05 *** = significance level of <0,001

An independent samples t-test was executed to test the difference between males and females in perceived audit risk by comparing the means between these two independent groups. The assumption of normality was met and the Levene’s Test for equality of variances was executed to test the assumption of homogeneity of variance. Both of these assumptions were met, so the variances were assumed as equal. The 1-tailed significance was used, because the hypothesis was formulated in one direction. The results indicated that there was no significant difference between male and female auditors in perceive audit risk at the 0,05 significance level. In other words, the outcomes did not support hypothesis 5.

Which factors do influence perceived audit risk?

Which factors do influence perceived audit

risk?

Table of Contents

1.

Introduction

2.

Theoretical background and hypotheses development

3.

Methodology

4.

Results