Time flies when narcissistic CEOs have fun; the influence of narcissistic CEOs on audit report lag

(1)

Time flies when narcissistic CEOs have fun; the influence of narcissistic CEOs on

audit report lag

Abstract

This study examines a sample of companies from the United Kingdom to investigate whether narcissism under CEOs is associated with audit report lag. In addition, this study examines whether this relationship can be explained via increased audit risk. Furthermore, the moderating effects of employee/family shareholding are considered. This study shows that narcissistic CEOs increase the audit report lag, which can be partly explained via the mediator audit risk. Moreover, this study finds evidence of moderated mediation when employee/family shareholding is included in the analyses. Therefore, this study is a valuable addition to the current literature, by presenting evidence of the association between narcissism under CEOs and increased audit report lag.

Key words: Narcissism, Audit Report Lag, Audit Risk, CEO, Ownership Concentration JEL classifications: C33, M12, M42

Fenneke B. Korteweg S2569590

f.b.korteweg@student.rug.nl

MSc Accountancy & Controlling | Track Accountancy Faculty of Economics and Business

University of Groningen Supervisor: prof. Dr. J.A. Emanuels

Assessor: Dr. Y. Karaibrahimoglu

Second assessor: G.C. Helminck RA MSc EMA 25 June 2018

Number of words: 10.696

“The difference between God and Larry is that God does not believe he is Larry.” - (executive at Oracle about his boss Larry Ellison)

(2)

- 2 -

1. Introduction

A 132% increase in plastic surgery procedures1, vloggers jumping on moving trains for attention and ordinary people buying cars and houses they cannot afford. One might think these are just some current trends in society. However, these trends might be part of an overall rise of narcissism in society. Stinson et al. (2008) find a significant increase in the prevalence of narcissism under young adults and children. Culture’s focus on self-admiration and social media are believed to deluge this rise in narcissism (Bergman et al., 2011). Some researchers even believe this upsurge in narcissism is partly responsible for the severe financial crisis of 2008 (Twenge & Campbell, 2009).

This raises the question whether increased narcissistic tendencies in society would provoke repercussions. Therefore, the concept of narcissism has received increased attention in the academic world. Especially the influence of narcissistic CEOs has been a hot topic. According to O'Reilly et al. (2014), narcissistic CEOs tend be dominant, have an extreme amount of self-confidence, grandiosity and a lack of empathy. Since CEOs have a significant influence on the processes in their company, the companies with narcissistic CEOs act significantly disparate compared to companies without a narcissistic CEO. In line with this, Oesterle et al. (2016) find that narcissistic CEOs tend to intensify internationalization and perform more activities abroad. In addition, recent research of O'Reilly et al. (2017) shows that narcissistic CEOs increase the firm’s vulnerability to lawsuits, since they are more likely to act on the border of what is acceptable corporate behavior than other firms.

However, the consequences of narcissism under CEOs can reach far beyond business decisions. According to Maccoby (2000), narcissistic CEOs tend to have a negative influence on employees who do not agree with them. Narcissistic CEOs do not tolerate dissent of employees and want a group of yes-men. In addition, according to Chatterjee and Pollock (2017), narcissistic CEOs tend to have a significant impact on the board, since narcissistic CEOs chose high status board members to reflect positively upon themselves. Furthermore, the researchers find that

(4)

- 4 -

narcissistic CEOs also use manipulation techniques to influence board members, with less adequate monitoring of the board as their objective.

Given the impact a narcissistic CEO has on his company, employees and board, it is not surprising that a narcissistic CEO also has an effect on external actors, such as the auditor. Recent research of Judd et al. (2017) has shown that auditors recognize narcissistic CEOs. The effect is that, auditors appraise a higher audit risk and consequently these companies tend to pay a higher audit fee. However, the research of narcissistic CEOs in relation to the audit is new and quite limited up to this date.

This study elaborates on the relationship between the audit and the narcissistic CEO by researching whether companies with narcissistic CEOs, tend to have a higher audit report lag. Audit report lag (ARL) is defined according to Behn et al. (2006) and Blankley et al. (2015), who state that audit report lag represents the number of days elapsing between the end of the fiscal year and the completion of the audit. The reason that a relation is expected between a narcissistic CEO and audit report lag is that, narcissistic CEOs are believed to increase the audit risk, which increases the necessary activities and procedures of the auditor, which can increase the audit report lag. Although prior research has found many determinants of ARL, the presence of a narcissistic CEO has not been researched as being a possible determinant. This research therefore further develops the relationship between the narcissistic CEO and the audit and investigates another possible determinant of ARL.

The goal of this research is to determine whether the presence of a narcissistic CEO is a determinant of ARL. Furthermore, this study tries to determine what drives this relationship. This leads to the following research question:

Are narcissistic tendencies of the company’s CEO a predictor for higher audit report lag? For the possible drivers of this relationship this study examines the role of audit risk as a mediating variable. In addition, this study will examine whether employee/family shareholding is able to moderate the relationship between CEO narcissism, audit risk and audit report lag.

This study finds evidence that narcissistic CEOs increase the audit report lag. The possible explanation for this finding is the increased audit risk, which is found to partly mediate the relationship between narcissism of the CEO and the increased ARL. In addition, the effects of

(5)

- 5 -

employee/family ownership are found to moderate the relationship between narcissism, audit risk and ARL.

This study is relevant for researchers as well as for auditors and their clients. By researching the relationship between CEO narcissism and ARL, auditors can consider the possible effect of the personality of a CEO on their audit efficiency. By being aware of the possible determinants, the first step is taken towards preventing the negative effects a narcissistic CEO can have on ARL. In addition, ARL can have negative consequences for companies, such as higher perceived information asymmetry and negative stock market reactions (Krishnan & Yang, 2009). If the ARL is hampered, it will prevent the negative reactions from shareholders. Something that will greatly benefit auditors, audit companies and their clients. Furthermore, by investigating the moderating impacts that employee/family shareholding can have on this relationship, this study contributes to possible solutions that can combat the negative influences of having a narcissistic CEO. For academics, this study contributes to the literature on ARL determinants and is, to my knowledge, the first to examine the relationship between a narcissistic CEO and ARL. Therefore this study makes a valuable contribution to existing literature.

The remainder of this study is structured as follows, section two elaborates the relevant literature and develops the hypotheses, in section three the sample, data and methodology used will be described, section four will explain the results, and section five will conclude with and review some limitations of this study.

(6)

- 6 -

2. Theoretical Background and hypothesis development

This section provides the theoretical background on audit report lag and narcissistic CEOs. A literature review has been performed to formulate the hypotheses.

2.1 Audit Report Lag (ARL)

Audit report lag is defined according to Behn et al. (2006) and Blankley et al. (2015), who state that audit report lag (ARL) represents the number of days elapsing between the end of the fiscal year and the completion of the audit. Several studies on ARL have been conducted, because it is often viewed as the most important determinant of financial reporting timeliness (Abbott, Parker and Peters, 2012). Furthermore, Krishnan & Yang (2009) state that, ARL is relevant, because a higher ARL can have negative consequences such as a higher perceived information asymmetry and negative stock market reactions.

An explanation for the perceived information asymmetry can be found in the Agency theory. Over the years, corporations have grown considerably which eventually resulted in the separation of ownership and control. This separation is useful for companies, because they can attract more capital to invest in profitable corporate projects. However, this separation also has its disadvantages. The principals (investors) give their money to an agent (managers) to manage the projects. Since the principals are not able to be present constantly or to perfectly monitor the actions of the agent otherwise, they do not have full control over the actions of the agent. When the agent informs the principals about his actions, information asymmetry arises, since the agent has information about his actions that the principle is not able to obtain. This is where agency problems emerge. Due to the existence of the information asymmetry, the manager is able to untruthfully report his actions and assign his own interests above the interests of the shareholders (Jensen and Meckling, 1976).

According to Watts and Zimmerman (1983), an auditor of the information can reduce the information asymmetry by verifying the claim of the agent. Thus, the auditor is important for the principals to reduce the information asymmetry and to make sure the agent does not retrieve a private gain at the expense of the principals. Therefore, principals are dependent on the auditor to verify the claim of the agent. The ARL is for principals a measure of the effectiveness of the auditor and thereby the reliability of the claims made by the agent. Information that is delayed is

(7)

- 7 -

perceived to contain more information asymmetries than timely information, because according to Shu et al. (2015) the delay in information represents the hesitation of the auditor in preparing the audit report with the incomplete information of the client. In line with this study, Blankley et al. (2014) find that higher ARL is positively associated with financial restatements. Furthermore, the researchers find that the longer the financial report takes, the less likely that the audit report can truly reveal the financial condition of the client.

Shareholders therefore negatively perceive ARL, however, the reasons for the ARL can be numerous and do not always mean that the information contains more information asymmetries. According to Abernathy et al. (2017), the consensus on ARL is that audit-related factors and company-specific factors affect the ARL. In prior literature, many factors have been researched and examined. An overview of the most important factors can be found in the table below.

Table 1 – Determinants of ARL

Company-specific factors Effect on audit lag Confirmed by

Size - Carslaw and Kaplan (1991)

Ng and Tai (1994) Soltani (2002)

Bonsón‐Ponte et al. (2008)

Weak performance + Kross (1981)

Whittred and Zimmer (1984) Haw et al. (2013)

Complexity + McLelland and Giroux (2000)

Weaknesses in IC + Ettredge et al. (2006)

Impink et al. (2012) Khlif and Samaha (2014)

Audit-related factors

Unqualified audit opinion + McKeown (1991)

Soltani (2002) Internal audit assistance - Abbott et al. (2012)

(8)

- 8 -

Auditor Tenure - Dao and Palm (2014)

Lee et al. (2009) Lack of personnel resources + Behn et al. (2006)

As presented in the table, weak business performance, complex business structures, weaknesses in internal controls, an unqualified audit opinion and a lack of personnel resources in the audit firm are all positively associated with audit report lag. On the other hand, firms larger in size, internal audit assistance, size of the audit firm and auditor tenure are all negatively associated with audit report lag.

Even though a consensus on the main factors that determine ARL is present, these factors are not able to explain ARL fully. This gives rise to the question what other factors might determine ARL. This study will try to determine another important factor in explaining ARL, namely the presence of a narcissistic CEO.

2.2 Narcissism

According to the American Psychiatric Association (2013), narcissism is a personality disorder. This narcissistic personality disorder is characterized by significant impairments in personality and interpersonal functioning. These personality impairments include; excessive reference to others for self-definition and self-esteem regulation; exaggerated self-appraisal and setting goals based on gaining approval from others. People with a narcissistic personality disorder also have impairments in interpersonal functioning that include a lack of empathy and difficulty with intimacy. In addition, they often have feelings of grandiosity and seek attention excessively.

Narcissism, agency theory and information asymmetry.

The aforementioned agency problem is important in the explanation for the detrimental effects a narcissistic CEO can have on a company. As explained above, the agent and principal have conflicting interests and the agent can exploit this because of the information asymmetry in their relationship. In theory, agency problems would not arise if the interests of the principal and the agent would be perfectly aligned. However, in practice this is not a realistic scenario, the personal interests of the agent are practically always different from the interests of the principals. This has all sorts of negative consequences, e.g. problematic acquisitions, (Hayward and Hambrick, 1997), excessive risk-taking (Pathan, 2009) and lower financial leverage than optimal

(9)

- 9 -

(Jensen, 1986).Narcissistic CEOs are likely to fulfill a different role in the agency theory compared to regular CEOs. According to Ong et al. (2014), people with narcissistic tendencies are more likely to be seen as a good leader. They are charismatic (Khoo & Burch, 2008), appear socially skilled (Oltmanns et al., 2004) and they perform well in difficult tasks (Roberts et al., 2013). However, once they are installed as CEOs, narcissistic individuals (agents) are more likely to pursue their own interests above the interests of shareholders (principals). According to Campbell and Foster (2007), narcissistic individuals have fundamental narcissistic qualities, such as inflated self-views, entitlement, agentic concerns, lack of empathy. Due to these qualities, narcissistic individuals have certain interpersonal skills, such as confidence and charisma, but it also leads them to employ certain intrapsychic strategies. According to the researchers, these strategies include self-serving bias, fantasies of power, better-than-average effect and self-promotion. Because of these traits and strategies, narcissistic CEOs are even more likely to put their own interests above the interests of shareholders, which creates agency problems.

The consequences of a narcissistic CEO on the company are mainly caused by the desire of the narcissistic CEO to fulfill his private interests. In line with this, Olsen et al. (2013) argue that narcissistic CEOs have an increased interest in short-term gains, despite long-term harms. This study serves as a clear illustration showing that the narcissistic CEO assigns his own short-term interests above the long-term interests of the shareholders, since the long-term consequences do not affect the CEO personally. In conclusion, the expectation is that the narcissistic CEO is more likely to create agency problems, because of his strong desire to fulfill his needs and lack of empathy. The information asymmetry due to separation of ownership and control prevents the shareholders to hinder the narcissistic CEO to pursue his own desires.

Consequences of a narcissistic CEO. Narcissism was originally a concept used only in

psychology to describe personalities and consequently the behaviors of certain individuals. However, narcissism has been under the attention of business researchers, since first found that narcissistic CEO can have a significant impact on the decisions made and actions taken within a company. The impact of the personality of business leaders on a firm’s outcome are first described by Hambrick & Mason (1984) with the Upper Echelons theory. This theory explains that values of the decision maker determine the alternatives he considers and eventually these values determine the managerial perceptions. The managerial perceptions have, in turn, a direct effect on

(10)

- 10 -

the strategic choices made. Consequently, these strategic choices affect the organizational outcomes. Thus, despite the fact that other managers, the board and employees limit CEOs, they still have a significant impact on the decisions made in their company. Since the narcissistic CEOs have different personality traits than non-narcissistic CEOs, including a lack of empathy, the organizational outcomes of companies with a narcissistic CEO are likely to be different as well. As explained above, their interest and values are likely to be different. Therefore, narcissistic CEOs and their companies have been under increased attention under researchers.

Over the years, researchers have found several significant potential consequences for and differences in organizational outcomes of firms with a narcissistic CEO compared to those without a narcissistic CEO. These potential consequences of the presence of a narcissistic CEO have in general a negative effect on the organizational outcomes, but they can have positive effects on companies as well. Within the literature, no consensus about whether narcissistic CEOs positively or negatively affect companies is reached. As explained with the agency theory, the narcissistic CEOs are more likely to cause agency problems and the negative consequences thereof, because of their stronger tendency to assign their own interests above those of the shareholders. In the table below, a short overview is provided of the results found in recent years about narcissistic CEOs and organizational outcomes.

Table 2 – Organizational outcomes of companies with narcissistic CEOs Potential significant consequences of narcissistic

CEOs.

Found by:

Strategic dynamism and grandiosity leading to more volatile earnings

Chatterjee & Hambrick (2007)

Increased risky and aggressive organizational decision making

Foster & Brennan (2011)

Higher risk of fraud Rijsenbilt & Commandeur (2013)

Increased emphasis on short-terms gains, despite long-term potential harms to firm

Olsen et al. (2013)

Increased aggressive tax sheltering Olsen & Stekelberg (2015) Intensified internalization and more activities abroad Oesterle (2016)

(11)

- 11 - More likely to be involved in lawsuits, less likely to settle or take advice from experts

O’Reilly et al. (2017)

As presented in the table, the influences that a narcissistic CEO has on a company, which include increased risky and unethical behaviors, generally make the company riskier.

However, the impact of narcissistic CEOs is not limited to their own company and employees, but impacts relationships with external actors as well. Recent research of Judd et al. (2017) is one of the first studies that examines the relationship of a narcissistic CEO and the auditor. They are specifically interested if auditors recognize the potential increased risks of a narcissistic CEO. They find that auditors recognize an increased audit risk and therefore these companies tend to pay a higher audit fee. Given that Judd et al. (2017) finds that the narcissistic CEO has an impact on the auditor, and therefore on the audit, this research tries to expand on this finding by determining what might be the consequences besides the increased audit fee.

2.3 Hypotheses development

As narcissistic CEOs have an impact on various aspects of their company and the audit, it is likely that narcissistic CEOs also have an impact on the ARL. Narcissistic CEOs increase the audit risk of the company and therefore the auditor will take longer to complete the entire audit, which increases the audit report lag. This reason will be extended upon in the next section. However, other reasons which cause narcissistic CEOs companies’ to have longer audit report lag could be present. The narcissistic CEO might want to slow down the audit process, by intentionally delay delivering documents needed for the audit to prevent his actions from becoming public with the publication of the annual report. Therefore, the direct relation between narcissism and ARL is

examined first.

The first hypothesis is therefore as follows:

Hypothesis 1: Companies with a CEO with more narcissistic tendencies, have higher audit report lag.

Narcissistic CEO, Audit Risk and Audit Report Lag. This section will extend upon

why narcissistic CEOs increase the audit risk and therefore the ARL. The Association of International Certified Professional Accountants (1938) defines audit risk as ‘the risk that the

(12)

- 12 -

auditor may unknowingly fail to appropriately modify his opinion on financial statements that are materially misstated’ (p.1647). According to the Statements on Auditing Standards (107, the auditor determines the audit risk with the Audit Risk Model. The audit risk model consists of the following components:

Audit Risk = Inherent Risk x Control Risk x Detection Risk

First, the inherent risk is the risk of complete transactions of transaction classes. For instance, complex transactions are inherently more difficult to audit than simple calculations. Besides the risk of the transactions, the factors within the entity and the environment of the entity determine the inherent risk. Second, the control risk is the risk that a mistake in the financial statements will not be detected by the internal controls of the company. Therefore, if the auditor does not identify these possible deficiencies in the internal controls, they can significantly influence the reliability of the financial statements. Finally, the detection risk is the risk that the auditor does not detect and correct a mistake that materially could affect the financial statements. Due to the uniqueness of each firm, the audit risk differs by company. The audit risk should be held on an acceptable level to prevent an incorrect audit opinion. If the inherent risk and the control risk are significantly higher, the audit should perform more audit activities and substantive procedures to decrease the detection risk. Consequently, the auditor makes sure the audit risk does not reach undesirable levels. Hogan and Wilkins (2008) find evidence that the auditor indeed performs more audit activities if the internal controls are weaker and therefore the control risk of a company is higher.

There are specific characteristics of narcissistic CEOs and their companies that significantly affect the inherent and control risk. It is impossible to obtain certainty that this list holds for every company and narcissistic CEO, however, multiple reasons imply that narcissistic CEOs are likely to increase the inherent and control risk.

First, Buchholz et al. (2014) state that narcissistic CEOs are more likely to use earnings management as a form of self-enhancement. In line with this, Rijsenbilt and Commandeur (2013) argue that, narcissistic CEOs undertake challenging actions to obtain frequent praise and admiration. This makes these narcissistic CEOs engage in fraud sooner than less narcissistic CEOs, to make sure they receive this admiration even if the financial results originally provided no reason for admiration. The strong tendency of narcissistic CEOs to assign their own interests above those of the shareholders plays a role as well. The CEO fulfills his desire for admiration, but at the

(13)

- 13 -

expense of the shareholders. This tendency of manipulating the financial results increases the inherent risk of the company, because it inherently raises the risk of material mistakes in the financial statements.

Second, narcissistic CEOs are less likely to make adjustments after they get critique. According to Maccoby (2000), narcissistic CEOs do not tolerate critique from employees and rarely listen to it. Furthermore, according to Chatterjee and Pollock (2017), the board will also not be likely to give the necessary critique, since the CEO uses techniques to convince them of his own opinion. This increases the inherent risk, and thereby, ceteris paribus, the audit risk, since behaviors that are not in line with fair accounting will be less likely to be criticized by either employees or the board.

Third, the tone at the top is another important factor that auditors consider in the assessment of the audit risk. According to the COSO framework (2013), top management sets the tone at the top. By communicating the expectations of top management with regard to integrity and ethical values, they create the culture for the rest of the organization. Therefore, if a narcissistic CEO leads the organization, the CEO can construct a culture wherein his own tendencies to unethical values and behavior are dispersed throughout the entire organization. This distorted culture can increase the audit risk of the company, since the risk of material mistakes is inherently higher.

Lastly, the internal control deficiencies are likely to be greater in companies with narcissistic CEOs, which increases the control risk and thereby, ceteris paribus, the audit risk. According to Roberts (2001) and Campbell et al. (2011), narcissistic CEOs regard themselves as being above the law and they have little respect for laws and regulations. Therefore, narcissistic CEOs are more likely to override controls and tend to have weaker internal control systems. Judd et al. (2017) indeed find that narcissistic CEOs have a significant effect on the likelihood of internal control deficiencies within the company.

Following the audit risk model, if the inherent and control risk are increased, the auditor will respond by performing more audit activities and substantive procedures, to lower the detection risk. That the auditor indeed performs more activities has been confirmed by Judd et al. (2017), with evidence on audit fees. Whether increased risks and more substantive audit procedures lead to increased audit lag is debatable. According to O’Keefe et al. (1994), Caramanis and Lennox (2008) and Calderon et al. (2012), the risk of the company significantly affects the planned hours for the engagement. If during the planning phase, the increased risk is assessed, the audit firm will

(14)

- 14 -

increase the number of hours and the expertise of the personnel (Johnstone and Bedard, 2004). Therefore, the expectation is that the extra resources deployed by the audit firm prevent a part of the ARL.

However, it is likely that the increased risk of the narcissistic CEO does increase the ARL. Several experimental studies (Glover et al., 2000; Glover et al., 2003; Hammersly et al., 2011) have found that auditors have difficulties with properly modifying the audit in case of increased risks. Moreover, narcissism is not a factor that is considered explicitly in the risk assessment. According to Bierstaker et al. (2017), auditors might over rely on standard audit programs and decision support systems. The overreliance on standard checklists might cause the auditors to adjust the resources insufficiently. Lastly, Messier et al. (2008) found that partners from audit companies are overconfident in the abilities of their employees and therefore tend to assign insufficient resources on an audit engagement. These tendencies cause the ARL to increase, because audit companies are not fully prepared for the extra activities necessary with higher risk companies.

In addition, multiple researchers (Ettredge et al. 2006; Impink et al. 2012; Khlif and Samaha 2014) have indeed found that deficiencies in internal controls increase the audit activities that the auditor has to perform, which increases the ARL. Which shows again that auditors are not able to adjust the resources appropriately to prevent ARL. Other evidence comes from McLelland and Giroux (2000), who find that if the complexity of the firm increases, the auditor’s activities increase which eventually leads to a higher ARL. Meaning that even when risks are anticipated on in the planning phase of the audit, these risks still increased the ARL further on in the audit process. This evidence leads us to the conclusion that a narcissistic CEO increases the audit activities that should be performed, which increases the ARL. Therefore, the second hypothesis is as follows; Hypothesis 2: The relationship between narcissistic CEOs and audit report lag can be explained via audit risk as a partial mediator.

Narcissistic CEO, Audit report lag and Employee/Family shareholding. In the sections above

(15)

- 15 -

section, provides an explanation of the effect that employee/family shareholding could have on the relationship between narcissism and audit risk.

As explained in the literature section, narcissistic leaders tend to show behavior of which many believe they will make great leaders. The narcissistic leaders are charismatic, appear socially skilled and perform well in difficult tasks. This is why it is likely that narcissistic leaders are appointed as CEO of the company. However, the longer the CEO is installed, the more negative tendencies are revealed. All these negative tendencies increase for instance the audit risk as explained above. However, the effect that a narcissistic CEO has can differ by company, depending on the monitoring of the CEO’s actions.

For the explanation of this argument the aforementioned agency problem is relevant as well. In general, there are several mechanisms to combat the agency problem. Commonly believed is that monitoring the agent, prevents the agent from behaving in ways that harms the investors. This monitoring can be executed by an auditor, who verifies that the agent reports truthfully about his results. The shareholders themselves can execute the monitoring as well. However, the effectiveness and willingness to monitor differs per shareholder.

According to Demsetz & Lehn (1985), shareholders would incur monitoring costs to monitor the principal. For many small shareholders the costs of monitoring exceed the return they can earn on their share and therefore these shareholders are not very likely to engage in monitoring. However, according to Anderson et al. (2003), family owners have more incentives to engage in active monitoring of the agent. First, the family owners are more interested in the long-term survival of the company compared to regular shareholders, since they often have larger undiversified portfolios in the company and they would like to pass the company on to their own children. In addition, firms with a large percentage of family ownership are more likely to install a CEO who is also a family member. According to the researchers, this CEO is more likely to have similar interests as the shareholders, which decreases the agency problems.

Moreover, according to Lawrence (1987), employee stock options are used by shareholders to align the interests of the agents with the interests of the principal. Stock options align the interests of the agent with the principals’, since the value of the company affects the agent as well as the principal. Therefore, the expectation is that these mechanisms can decrease the effect of a narcissistic CEO on audit risk, since the narcissistic CEO will have less incentives to engage in actions that will diminish shareholder value.

(16)

- 16 -

Therefore, the expectation is that firms with a percentage of employee/family shareholding will engage in additional monitoring and will decrease the incentives of the narcissistic CEO and thereby limit the impact of the CEO on the audit risk. For instance, the internal control systems. As explained, narcissistic CEOs have less effective internal control systems in place. However, it is likely that more active monitors would encourage the use of effective internal control systems and punish the CEO otherwise, which subsequently decreases the audit risk. The third hypothesis is therefore as follows:

Hypothesis 3: Employee/Family shareholding will moderate the relation between narcissism, audit risk and thereby also change the audit report lag.

In conclusion, the research model of this study is as follows:

Figure 1 – Research Model

H1 H2 H2 H3 Narcissistic CEO

Audit report lag

Employee/ Family Shareholding

(17)

- 17 -

3. Methodology

This section provides an overview of the data sources and the sample selection method. Secondly, the different variables used in this study are described. Finally, the methodology is discussed.

3.1 Sample and procedure

For this research, hand collection was used to obtain data from annual reports from the United Kingdom. Companies from the UK are chosen, because of their style of audit reporting, which is extended upon under the audit risk measure section. All public companies in the UK in the period 2010-2016 are examined. The data is collected from DataStream and Orbis. All financial companies are dropped from the sample, because of their significant differences in the nature and composition of their company’s statements. The sample initially consisted of 406 non-financial companies, with observations from 2010-2013 for the calculation of the narcissism score and with observations for the years 2013-2016 for the audit risk measure and ARL. Company’s annual reports are accessed through company’s websites or databases containing annual reports, such as Company.info. However, several companies are excluded from the final sample. The reasons for exclusion are;

1. CEO change in the period 2010-2013, which prevents correctly calculating their impact on ARL.

2. Missing annuals reports, which led to missing info on all variables.

3. Companies that did not switch to the new long form audit report, which led to missing audit information.

4. Audit reports based on the American 10-K filling, which led to missing audit risks and auditor names.

These companies are excluded from the dataset. Ultimately, two samples are created. The sample for the regression with ARL and narcissism consists of 898 observations. The sample used for the mediation and moderated mediation analysis consists of 645 observations. This is because, for several audit reports the audit date is stated, however the audit risks are not.

(18)

- 18 -

3.2 Measures

Independent variable

Narcissism measure. In the field of psychology, it is common to diagnose a narcissistic

personality disorder with the Narcissistic Personality Inventory (Raskin and Terry, 1988). However, direct data from CEOs is difficult to collect due to their unwillingness to cooperate. Despite this missing information, an unobtrusive measure can still be used, which is shown to be a credible alternative to self-reported measures (Webb and Weick, 1979). The measure of Chatterjee and Hambrick (2007), is based on public information and therefore not dependent on cooperation of CEOs. The validity of the measure of Chatterjee and Hambrick (2007) has been extensively tested in research and proven to be valid for a proxy of narcissism among CEOs, with available public data (Ingersoll et al.; 2017, Zhu and Chen, 2015; Buyl et al., 2017). Therefore, this measure is used to obtain narcissism scores for CEOs.

Narcissism will be measured on the basis of the following variables:

1. The prominence of the CEO’s photograph in the company’s annual report

According to Chatterjee and Hambrick (2007), the prominence of the CEO can be assessed by looking at the company’s annual report. The photograph prominence is scored on a scale of 1-4. Where a score of one is an absent photograph and four is a full-page photograph. The expectation is that more narcissistic CEOs have larger pictures in the annual report, to show that they are more important than the other employees are.

2. The CEO’s use of first-person singular pronouns in interviews

According to Raskin and Shaw (1988), the use of first-person singular pronouns reflect self-absorption and is an indicator of narcissism. First-person singular pronouns as I, me, mine, my, myself are counted. Expected is that more narcissistic CEOs use a higher number of single person pronouns.

3. The CEO’s relative cash and non-cash compensation

According to Tosi & Gomez-Mejia (1989), CEOs have a considerable impact on their own salary. According to Chatterjee and Hambrick (2007), the narcissistic CEOs believe that they are far more important than other employees working for the company are. Therefore, these CEOs prefer a bigger discrepancy between their salary and that of the second highest employee. Accordingly, the total pay of the CEO will be divided by the total pay of the second highest employee. O’Reilly et

(19)

- 19 -

al. (2014) confirm that narcissistic CEOs have larger discrepancies between their own salary and that of other member in their team.

The overall narcissism score will be calculated as follows; 𝑁𝐴𝑅𝐶𝑖 = ⁡(𝑠𝑐𝑜𝑟𝑒⁡𝑜𝑛⁡𝑝ℎ𝑜𝑡𝑜𝑔𝑟𝑎𝑝ℎ𝑖 + 𝑠𝑐𝑜𝑟𝑒⁡𝑜𝑛⁡𝑝𝑟𝑜𝑛𝑜𝑢𝑛𝑠𝑖 + 𝑠𝑐𝑜𝑟𝑒⁡𝑜𝑛⁡𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛𝑖) (1)

Dependent variables

Abnormal Audit Report Lag measure. According to Behn et al. (2006) and Blankley et

al. (2015), audit report lag represents the number of days elapsing between the end of the fiscal year and the completion of the audit for the current year for each individual firm. This means that the ARL measure represents the difference in days between the end of the fiscal year and the day the audit report is signed. This definition is used in this study to measure ARL.

The publication of the annual report of the listed financial companies of the sample are subject to regulation. In the UK, public companies have 6 months to publicize the annual report after the end of their fiscal year (Companies Act, 2006). This deadline has some implications for the ARL. This means that it is not very likely that the annual reports are published after the mandatory date required by law. However, companies are free to determine on which date within the six months they publish their report. Therefore, the expectation is that there is still enough variation within the ARL measure.

The measure of ARL has its own limitations. The biggest limitation is that the duration in hours of the audit cannot be interfered. It could very well be that the actual audit was planned much later in the year, and therefore the audit lag measured seems higher, when in fact the duration of the audit was shorter. The reason for using the end of the fiscal year is insufficient information about the real audit hours for each company. To assess the real number of hours spent on an engagement, audit firms would need to make this information public. However, audit firms keep their files confidential and therefore the ARL can only be determined based on the publication date of the report. Since there are no alternative measures in research, we use the publication date and control for many factors that have shown to have a significant impact on ARL.

This study follows the method of Blankley et al. (2014), Knechel and Sharma (2012) and Krishnan and Yang (2009) to create a measure for abnormal audit report lag. According to the researchers, the abnormal ARL variable is calculated with a two-stage approach. First, the ARL of

(20)

- 20 -

the sample is controlled for common determinants of (the logarithm of) audit report lag, these determinants are based on prior literature and available data. Second, the residuals become the abnormal audit report lag measure, which are used to further analyze the relationships between narcissism and audit report lag.

The measure for audit report lag is calculated as follows: 𝐿𝑜𝑔𝐴𝑅𝐿𝑖𝑡 = β0+ 𝛽1𝑆𝑖𝑧𝑒𝑖𝑡+ 𝛽3𝑆𝑖𝑧𝑒𝑖𝑡2 + 𝛽4𝐴𝑢𝐶𝑜𝑚𝑝𝑖𝑡 + 𝛽5𝐹𝑜𝑟𝑆𝑎𝑙𝑒𝑠𝑖𝑡+ 𝛽6𝐵𝑖𝑆𝑒𝑔𝑖𝑡+

𝛽₇𝑆𝑢𝑏_𝑖𝑡+ 𝛽₈𝐴𝑈𝐹𝐸𝐸_𝑖𝑡+ 𝛽₉𝐼𝑁𝐻𝑅𝑖𝑠𝑘_𝑖𝑡+ 𝛽₁₀𝐶𝑈𝑅𝑅_𝑖𝑡+ 𝛽₁₁𝐴𝐿𝑇𝑀𝐴𝑁𝑁_𝑖𝑡+ 𝛽₁₂𝐵𝑇𝑀𝑅_𝑖𝑡+ 𝛽13𝑁𝐼𝐶ℎ𝑎𝑛𝑔𝑒𝑖𝑡+ 𝛽14𝑅𝑂𝐴𝑖𝑡+ 𝛽15𝑁𝐼𝑛𝑐𝑜𝑚𝑒𝑖𝑡+ 𝛽16𝐵𝑈𝑆𝑌𝑖𝑡+ 𝛽17𝑅𝑒𝑠𝑡𝑎𝑡𝑒𝑚𝑒𝑛𝑡𝑖𝑡+

𝛽₁₈𝐴𝑈𝐶𝑆𝐼𝑍_𝑖𝑡+ 𝛽₁₉𝐵𝑂𝐴𝑆𝐼𝑍_𝑖𝑡+ 𝛽₂₀𝐶𝑂𝑅𝑃𝐺𝑂𝑉𝑆𝐶_𝑖𝑡+ 𝛽₂₁𝐴𝑈𝐶𝐼𝑁𝐷𝑃_𝑖𝑡+ 𝛽₂₂𝐴𝑈𝐶𝐸𝑋𝑃_𝑖𝑡+

𝛽23𝑌𝑒𝑎𝑟𝐷 + 𝛽24𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐷 + 𝛽25𝐶𝑜𝑢𝑛𝑡𝑟𝑦𝐷 (2)

This model includes several important determinants of audit report lag. Two measures for size are included, being 𝑆𝑖𝑧𝑒 and 𝑆𝑖𝑧𝑒2, since several researchers found that bigger companies tend to have significantly less ARL (Carslaw and Kaplan, 1991; Ng and Tai, 1994; Soltani (2002); Bonsón‐ Ponte et al., 2008). Furthermore, the audit company performing the audit can be of an importance, since research of Ng and Tai (1994), has found that larger (Big-N) audit companies have a significantly lower ARL than other audit companies. Therefore, a dummy variable for Big4 and non Big4 companies is included, being 𝐴𝑢𝐶𝑜𝑚𝑝. In addition, some measures for complexity are included, because McLelland and Giroux (2000) found that a more complex firm requires more audit hours and procedures. These measures are 𝐹𝑜𝑟𝑆𝑎𝑙𝑒𝑠,⁡𝐵𝑖𝑆𝑒𝑔, 𝑆𝑢𝑏, because firms with a big percentage of foreign sales, business segments or subsidiaries are more complex compared to firms with a low percentage, which increases the audit hours and therefore the audit lag. The 𝐴𝑈𝐹𝐸𝐸 measure is included, since Knechel and Sharma (2012) found a strong correlation between audit fees and ARL. Several measures are included to control for the risks companies face. The measure for⁡𝐼𝑁𝐻𝑅𝑖𝑠𝑘 is included because firms with a higher inherent risk require the auditor to increase the effort which increases the audit lag. The 𝐶𝑈𝑅𝑅 measure is included to control for the effect of the liquidity risk and the 𝐴𝐿𝑇𝑀𝐴𝑁𝑁 score, as a measure of financial distress risk. These determinants of audit lag are first found to have a significant impact by Whittred and Zimmer (1984). The 𝐵𝑇𝑀𝑅 (book to market ratio) and 𝑁𝐼𝐶ℎ𝑎𝑛𝑔𝑒 is included because Knechel and Sharma (2012) found the risk associated with growth, which is reflected in the measure, is an important determinant of audit report lag. Following Blankley et al. (2014), 𝑅𝑂𝐴 and 𝑁𝐼𝑛𝑐𝑜𝑚𝑒 are included

(21)

- 21 -

as measures of profitability, which can impact the effort the auditor should exert. Moreover, the timing of the audit report is important. Therefore, the measure⁡𝐵𝑈𝑆𝑌 is included, which is a proxy for a company having their fiscal year ending in December or January. Likewise, if the company reported a material misstatement for the year, the auditor preforms more work to correct the restatement and to prevent the restatement needs to be made again. Blankley et al. (2014) found that restatements can increase the audit lag significantly, therefore the measure 𝑅𝑒𝑠𝑡𝑎𝑡𝑒𝑚𝑒𝑛𝑡 is included. Lastly, some measures of corporate governance are included, because more recent research from Ika and Ghazali (2012), Abernathy et al. (2014) and Sultana et al. (2015) found that effective corporate governance mechanisms can decrease the ARL. The measures included are; 𝐴𝑈𝐶𝑆𝐼𝑍, 𝐵𝑂𝐴𝑆𝐼𝑍, 𝐶𝑂𝑅𝑃𝐺𝑂𝑉𝑆𝐶, 𝐴𝑈𝐶𝐼𝑁𝐷𝑃 and 𝐴𝑈𝐶𝐸𝑋𝑃. Which stand for audit committee size, board size, corporate governance score, audit committee independence and audit committee expertise. Finally country, industry and year dummies are included to control for fixed effects. The industry categorization is based on Ashbaugh et al. (2003), who identified industries which have a significantly different ARL. For company fixed effects, the model is clustered by company.

Audit Risk measure. Until to a few years ago, the audit report revealed little information

about the audit and contained several standard passages with little variation per company. According to Church et al. (2008), the audit report had little communitive value and mainly served as a symbol. Due to the limited information provided by the audit report, the audit risk could not be determined from the audit report directly and needed to be interfered from other measures such as, audit fees (e.g. Judd et al., 2017; Hay, 2006; Gul et al., 2003) or experiments (e.g. Johnson et al. 2012; Houston et al., 1999; Baldauf et al., 2015) .

However, regulators enforced a new audit standard, International Standard on Accounting (ISA) 700 for all public companies in the United Kingdom from 30 September 2013 onwards. This new standard requires auditors to include an overview of assessed risks of material misstatement, the materiality threshold used in the engagement, and the scope of the audit (ISA 700). In addition, the auditing standard encourages auditors to be company-specific and include explanations that directly link to a specific company. (Financial Reporting Council, 2013). Its goal is to provide greater transparency about the audit.

This study uses the key audit matters in the long form audit report, because this is a direct measure of the audit risks assessed by the auditor. According to ISA 701, auditors are supposed to

(22)

- 22 -

disclose significant audit risks in the audit report. These risks should be company-specific and should be related to material misstatements. This provides a direct measure of the audit risk and unlike other research there are no limitations due to a proxy used.

However, the measure for audit risk is not used in prior literature and researchers have not reached a consensus about its informativeness. According to research, the extra information provided with the new audit report generates mixed results. On the positive side, research finds a higher information quality of the audit reports, leading to different investment decisions (Christensen et al., 2014), and improvement in audit quality (Reid et al., 2016). On the other hand, Bédard et al. (2016) try to answer the question whether the new audit form is able to close the information gap. This information gap exists between shareholders who desire additional information on the financial statements of a company, for example about the audit risk. The idea of the long form audit report is that this information gap can be addressed, assuming that the auditor encloses this information in the new report. The conclusion after examining more than 30 studies is however, that the information gap is not closed by the new audit report. This is due to a decreasing number of key audit matters reported (Bédard, 2013) and key audit matters, that are instead of differentiated per company, rather generic (Bédard et al., 2016). Moreover, researchers find insignificant responses of investors to the changes in the audit report (Lennox et al., 2017) and no evidence of improved audit quality (Gutierrez et al., 2016). The limitations found in research on the long form audit report have direct consequences for the validity of the measure of audit risk. Given that the key audit matters are not found to have informational value to investors, the value of this long form audit report is limited for this study as well. Therefore, the results of this study must be interpreted with care. To obtain more certainty on the effect of narcissistic CEOs on the audit risks, a robustness test which includes a different measure for audit risk is included.

Audit risk is measured in the following way;

𝐴𝑈𝑅𝐼𝑆𝐾_𝑖𝑡 = 𝑛𝑢𝑚𝑏𝑒𝑟⁡𝑜𝑓⁡𝐾𝑒𝑦⁡𝐴𝑢𝑑𝑖𝑡⁡𝑀𝑎𝑡𝑡𝑒𝑟𝑠_𝑖𝑡 (3)

Moderating Variable

Employee/ Family shareholding. This variable is collected from the DataStream

database, under the code NOSHEM. According to Azevedo et al. (2014), Employee/Family Held Share (NOSHEM) is “the percentage of strategic holdings (of 5% or more) held by employees,

(23)

- 23 -

or those with a substantial position in the company’s shares that leads to a relevant voting power at annual general meeting (typically family members)”, p.5. This variable therefore refers to the percentage of shares not available to ordinary investors, but held by employees or family of the company. This variable is measured as follows;

𝐸𝑀𝐹𝐴𝑀𝑖𝑡 =

𝑠ℎ𝑎𝑟𝑒𝑠⁡ℎ𝑒𝑙𝑑⁡𝑏𝑦⁡𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠⁡𝑜𝑟⁡𝑓𝑎𝑚𝑖𝑙𝑦⁡𝑚𝑒𝑚𝑏𝑒𝑟𝑠

𝑡𝑜𝑡𝑎𝑙⁡𝑠ℎ𝑎𝑟𝑒𝑠⁡ _𝑖𝑡 (4)

3.3 Analyses

Before testing specific hypotheses the abnormal audit report lag will be determined with a two-stage regression approach. The two-stages from this analysis are explained separately.

First Stage

In the first stage, the ARL of the sample is regressed on the common determinants of ARL found in prior literature. Furthermore, controls for industry, year and country fixed effects are included. Also, to control for company fixed effects, this model is clustered by company. The following formula is used; 𝐿𝑜𝑔𝐴𝑅𝐿_𝑖𝑡 = β₀+ 𝛽₁𝑆𝑖𝑧𝑒_𝑖𝑡+ 𝛽₃𝑆𝑖𝑧𝑒_𝑖𝑡2 + 𝛽₄𝐴𝑢𝐶𝑜𝑚𝑝_𝑖𝑡 + 𝛽₅𝐹𝑜𝑟𝑆𝑎𝑙𝑒𝑠_𝑖𝑡+ 𝛽₆𝐵𝑖𝑆𝑒𝑔_𝑖𝑡+ 𝛽₇𝑆𝑢𝑏_𝑖𝑡+ 𝛽₈𝐴𝑈𝐹𝐸𝐸_𝑖𝑡+ 𝛽₉𝐼𝑁𝐻𝑅𝑖𝑠𝑘_𝑖𝑡+ 𝛽₁₀𝐶𝑈𝑅𝑅_𝑖𝑡+ 𝛽₁₁𝐴𝐿𝑇𝑀𝐴𝑁𝑁_𝑖𝑡+ 𝛽₁₂𝐵𝑇𝑀𝑅_𝑖𝑡+ 𝛽13𝑁𝐼𝐶ℎ𝑎𝑛𝑔𝑒𝑖𝑡+ 𝛽14𝑅𝑂𝐴𝑖𝑡+ 𝛽15𝑁𝐼𝑛𝑐𝑜𝑚𝑒𝑖𝑡+ 𝛽16𝐵𝑈𝑆𝑌𝑖𝑡+ 𝛽17𝑅𝑒𝑠𝑡𝑎𝑡𝑒𝑚𝑒𝑛𝑡𝑖𝑡+ 𝛽₁₈𝐴𝑈𝐶𝑆𝐼𝑍_𝑖𝑡+ 𝛽₁₉𝐵𝑂𝐴𝑆𝐼𝑍_𝑖𝑡+ 𝛽₂₀𝐶𝑂𝑅𝑃𝐺𝑂𝑉𝑆𝐶_𝑖𝑡+ 𝛽₂₁𝐴𝑈𝐶𝐼𝑁𝐷𝑃_𝑖𝑡+ 𝛽₂₂𝐴𝑈𝐶𝐸𝑋𝑃_𝑖𝑡+ 𝛽23𝑌𝑒𝑎𝑟𝐷 + 𝛽24𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐷 + 𝛽25𝐶𝑜𝑢𝑛𝑡𝑟𝑦𝐷 (2)

After this regression the residuals are predicted and these residuals are treated as abnormal ARL in further analyses. Thereby, the abnormal ARL is the part of audit report lag that cannot be explained by determinants of ARL found in prior literature. The abnormal ARL is used to determine whether the variables, narcissism and audit risk are able to explain some of the variation in the sample.

This method is used since it is computationally easier for the statistical program to include the control variables in the first stage and the other regressions in the second stage. Control variables, industry, year and country fixed effects are taken indirectly into account in the second

(24)

- 24 -

stage, since these effects are reflected in the abnormal audit report lag measure. Using two stage regressions is common in ARL literature (Blankley et al.; 2014, Knechel and Sharma, 2012; and Krishnan and Yang 2009).

Second stage

In the second stage, the analyses proceed with using the abnormal ARL, calculated in the first stage.

Test for hypothesis 1 – Relationship between Narcissism and Abnormal ARL

As explained in the hypothesis development, the expectation is that narcissistic CEOs are associated with higher abnormal ARL. This hypothesis is tested with OLS regression. The formula used is as follows;

𝐴𝐵𝑁𝐿_𝑖𝑡 = ⁡ 𝛽₀+ 𝛽₁ ∗ 𝑁𝐴𝑅𝐶_𝑖𝑡+ 𝜀⁡ (5)

Where ABNL is the abnormal ARL calculated in the first stage, other variables are extended upon in section 3.2.

Test for hypothesis 2 – Relationship between, Narcissism, Audit Risk and Audit Report Lag As explained under the hypothesis development, the expectation is that narcissism causes increased audit risk and this in turn causes higher ARL. This hypothesis is tested with mediation analysis.

Following Hayes (2013), the relationship between abnormal audit report lag and narcissism must be significant to continue with mediation analysis, since for mediation it must be established that increased narcissism leads to higher audit report lag. The presence of the relation between narcissism and ARL is established with the test of the first hypothesis. Furthermore, according to Hayes (2013), other relationships between the variables have to be statistically significant to conclude that mediation is present. Structural equation modelling is used to analyze the relationship between the variables and to estimate the direct effect and the effect mediated via audit risk, which is called the indirect effect.

(25)

- 25 -

These relationships are tested in the following way;

𝐴𝑈𝑅𝐼𝑆𝐾𝑖𝑡 = ⁡ 𝛽0+ 𝛽1∗ 𝑁𝐴𝑅𝐶𝑖𝑡+ 𝜀 (6)

𝐴𝐵𝑁𝐿_𝑖𝑡 = 𝛽₀+ 𝛽₁ ∗ 𝐴𝑈𝑅𝐼𝑆𝐾_𝑖𝑡+ 𝛽₂ ∗ 𝑁𝐴𝑅𝐶_𝑖𝑡+ 𝜀⁡ (7)

The variables are extended upon in section 3.2.

However, according to Preacher et al. (2007), simple regression analysis is not sufficient to be certain about the significance and extent of the indirect effects. To combat this problem, the researchers recommend to use bootstrapping as a resampling method. In the bootstrapping process multiple samples are taken from the original sample and subsequently tested. Since bootstrapping gives more certainty about the extent and significance of the indirect effects, this technique is used in the mediation analysis.

Test for hypothesis 3 – Relationship Abnormal ARL, Narcissism and Employee/Family shareholding

If employee/family ownership is able to moderate the relationship between narcissism, audit risk and ARL, moderated mediation is present. Within moderated mediation conditional indirect effects are present. Since, according to Preacher et al. (2007), conditional indirect effects are present when the moderator changes the relationship between the independent variable and the mediator.

This study follows the models of Hayes (2013) and Preacher et al. (2007) to test for moderated mediation. To test whether employee/family shareholding has a moderating impact, structural equation modelling is used to test the relationships and their significance. The following formulas are tested;

𝐴𝑈𝑅𝐼𝑆𝐾_𝑖𝑡 = ⁡ 𝛽₀+ 𝛽₁𝑁𝐴𝑅𝐶_𝑖𝑡+ 𝛽₂𝐸𝑀𝐹𝐴𝑀_𝑖𝑡+ 𝛽₃(𝑁𝐴𝑅𝐶_𝑖𝑡∗ 𝐸𝑀𝐹𝐴𝑀_𝑖𝑡) + 𝜀 (8)

𝐴𝑅𝐿_𝑖𝑡 = 𝛽₀+ 𝛽₁∗ 𝐴𝑈𝐷𝑅𝐼𝑆𝐾_𝑖𝑡+ 𝛽₂∗ 𝑁𝐴𝑅𝐶_𝑖𝑡+ 𝛽₃𝐸𝑀𝐹𝐴𝑀_𝑖𝑡 + +𝛽₄(𝑁𝐴𝑅𝐶_𝑖𝑡∗

(26)

- 26 -

After the estimation of the model the conditional indirect effects are calculated. Different levels of the moderator (being, mean- 1 std. dev, mean and mean + 1 std. dev.) are tested and these levels are multiplied with the coefficients obtained by the structural equation modelling procedure on the models mentioned above to obtain the conditional indirect effects (Hayes, 2013). This procedure provides the possibility to examine whether the change in the moderator significantly changes the relationships in the model used. Similar to the second analysis, bootstrapping is used to obtain more certainty about the presence and the extent of the conditional indirect effects.

(27)

- 27 -

4. Results

This section describes the descriptive statistics and the results of the analyses. The tables referred to in this section are presented in the appendix.

4.1 Descriptive Statistics

The descriptive statistics for the variables utilized in this study are presented in Table 4. The mean of ARL is 72 days. The standard deviation is 37 days, indicating considerable variation in ARL within the sample. Furthermore, unlogged assets have a mean of 7316 million, 95% of the companies are audited by Big 4 auditors, the mean of foreign sales is 46.87, suggesting that the sample consists of relatively large companies. Other variables are comparable to other studies that used large public companies form the United Kingdom.

Table 6 presents the Pearson Correlations among the variables utilized in this study. This matrix is included to examine the collinearity between the variables. The highest correlation is found between AUFEE and Size with correlation of 0.77. This is in line with previous research which finds that client size is a primary determinant of audit fee. The analysis shows that 18 of the 20 control variables are correlated to ARL, which provides initial evidence for the model employed.

4.2 ARL determinants

In table 7 the least square results of the regression of the determinants of audit report lag on the audit report lag of the sample are presented. 14 of the 19 control variables used show a significant association with the ARL of the sample. Especially the audit company, the number of business segments and the audit fee appear to be important variables in the estimation of ARL. The restatements, Altmann score, audit committee independence, ROA and the change in net income do not have a significant impact on audit report lag in the sample. The reasons for these insignificant control variables, could be the relatively small sample compared to other ARL studies or the selection of large companies where the variables lack sufficient variation to be a significant explanation factor.

(28)

- 28 -

Furthermore, the R-squared of the sample is 30.7%, which means that 30% of the variance in audit report lag is explained by the variables of the model employed. This is not in line with other studies, who find a higher R-squared. The expectation is that due to limited data availability important variables are missing from the ARL model used. The residuals of this regression are used in further analyses.

4.3 Hypothesis 1 - Narcissism and Abnormal ARL

Table 8 presents the results of the regression between abnormal ARL and narcissism. In line with hypothesis 1, narcissism is positively and significantly associated with abnormal audit report lag. The narcissism coefficient has a positive coefficient of 0.080 and is significant at the 1% level. This suggests that in line with the expectation, for CEOs with higher narcissism scores, higher abnormal report lag is expected. However, the total explanation value of a narcissistic CEO is low, with an R-squared of 1.3%.

4.4 Hypothesis 2 - Narcissism, Audit Risk and Abnormal ARL

Table 9 presents the results of the regressions between, ARL, narcissism and the mediator being audit risk. The associations are significant and positive, with p-values below the 5% and 1% level. The extent of the indirect effect can be calculated by (0.012/0.081) 14.81%, which is the percentage of the total effect mediated via audit risk. These results indicate that indeed some of the effects of narcissism on ARL are mediated via audit risk. Therefore, these results support hypothesis 2.

4.5 Hypothesis 3 - Narcissism, Audit Risk, Employee/Family held shares and Abnormal ARL

As explained in the methodology section, conditional indirect effects are tested by multiplying different levels of the mediator by the variables obtained via structural equational modelling. In Table 10 presents the results of the structural equation modelling procedure of narcissism, audit risk, EMFAM and ARL, on the basis of formula 7 and 8.

Table 11 presents the results of the different levels of the moderator and the calculation of the conditional indirect effects. In line with the expectations, evidence is found of conditional

(29)

- 29 -

conditional indirect effect slowly increases as the moderator increases. This indicates that the higher the employee/family shareholding is, the higher the moderating effect on the relationship between narcissism and audit risk. Therefore, the results suggest that employee/family

shareholding is able to moderate the effect a narcissistic CEO has on the audit risk and abnormal ARL, which supports the third hypothesis.

4.6 Robustness

The results of the robustness tests are presented in the appendix in table 13 to 18.

Dutch Sample

The results could be specific for the United Kingdom, therefore a sample from another country is examined, being the Netherlands. The Netherlands is chosen since it has similar regulation for audits as the UK. The data are obtained in a similar manner as the data for the main analyses. The conclusion is that the results of this study are only partially robust to the Dutch sample. For the Dutch sample, we find no evidence that narcissism increases abnormal ARL. However, evidence is found that narcissism increases the audit risk and that this relation is moderated by employee/family shareholding. This means that our results are partially robust to a sample including Dutch companies. However, these results need to be interpreted with care since due to lacking data, the score is based on the picture score and the use of pronouns only.

Propensity score Matching

The results show that companies with a narcissistic CEO have a significantly higher abnormal ARL. However, these results might be biased since companies with characteristics that cause higher ARL, might also be more likely to appoint a narcissistic CEO. This phenomenon is called the selection problem. To examine whether the selection problem has impacted the results of this study, propensity score matching procedure is used. With the propensity score matching

procedure, treatment groups are created, where one group receives the treatment and the other group does not (Caliendo and Kopeinig, 2008). For this study the treatment group is chosen on the basis of the narcissism score. Every company with a narcissism score above the mean will be used as the treatment group. The next step is to match companies of the treatment group with companies of the non-treatment group on the basis of common characteristics, which are selected

(30)

- 30 -

on the basis of the control variables. Subsequently, within every match the differences between the treatment group and the non-treatment group are evaluated. Therefore, this procedure is a solution to the selection problem, since similar characteristics are held constant where narcissism differs between the matches. In conclusion, most of the results are robust when using propensity score matching. However, the third hypothesis does not seem robust to propensity score

matching.

Audit fee analyses

There are several ways to measure audit risk, as explained this study used a new measure, being the key audit matters. Since this measure is not used in prior literature the measure might generate biased results. To test the robustness of this measure for audit risk, the results are examined with a more widely used measure for audit risk, being audit fee (Judd et al., 2017; Hay, 2006; Gul et al., 2003). When the measure for audit risk is replaced with audit fee, very similar results are found, which are in most cases even more significant than the measure of audit risk used, being the key audit matters. Therefore, the results are robust to another measure of audit risk.

Inverse Mills Ratio

For this study a sample of public companies is chosen with observable characteristics. This sample is not randomly chosen from all the companies in the market. Therefore the results might be biased, because certain firms are consistently excluded due to their size or exclusion from the market which makes their characteristics unobservable, which is referred to as the selection bias (Puhani, 2000). Heckman (1979) proposed a procedure to combat the problem of sample

selection. This selection model provides the opportunity to calculate the Inverse Mills ratio. The Inverse Mills ratio is included in the regression as a control variable to control for any sample selection effects. In conclusion, sample selection might biased the results of this study, since the results are not robust when the Inverse Mills Ratio is included in the analyses. Therefore the results of this robustness test suggest that it is likely that the results of this study only hold for firms included in the sample and that the results differ for other samples.

(31)

- 31 -

Auditor Fixed Effects

For the determination of the audit risk, the auditor and his style of auditing could significantly affect our results. Auditor fixed effects are included, since Francis et al. (2013) found that auditor can have a specific style of auditing. This means that some auditors are more likely to report key audit matters than other auditors. Therefore, we include a robustness test which includes auditor fixed effects to exclude the systematic differences in auditing. Since the incorporation of the new audit report, according to ISA 700, it is obligatory for auditors to sign the audit report with the name of the responsible audit partner, this provides the opportunity to create specific auditor dummies. It appears that by including auditor fixed effects, no differences in the results

regarding the impact of narcissism on ARL are found. However, it appears that the auditor style does play a role in the determination of the audit risk, since the audit risk is no longer a

predicting variable of ARL. Also, there is no evidence for conditional indirect effects. This suggests that the auditor style plays an important role in interpreting the results.

Other forms of monitoring

As shown in the section above, monitoring the narcissistic CEO can have a significant impact on the abnormal ARL. To examine whether the conclusions about increased monitoring and its impact on narcissism and audit risk can be extended to other forms of monitoring some

additional analyses are performed. Table 16 and 17 include similar regressions, instead of using employee/family shareholding, closely held shares and institutional ownership are used. For both these forms of ownership the expectation is that a large percentage of institutional ownership or closely held shares, will lead to increased monitoring of the narcissistic CEO, this changes the impact of the CEO on audit risk. The results of the robustness test suggest that other forms of monitoring provide similar results as employee/family shareholding. Therefore, the results are robust to other forms of monitoring.

Time flies when narcissistic CEOs have fun; the influence of narcissistic CEOs on audit report lag