Did CEO-narcissism decrease Return On Assets during the financial crisis of 2008-2012?

(1)

Did CEO-narcissism decrease Return

On Assets during the financial crisis of

2008-2012?

Melvin J. H. Welsink

Amsterdam Business School – Accountancy & Control University of Amsterdam

Final Version January 11, 2016

Abstract

This empirical study tests the relationship between CEO narcissism and the narcissistic CEO firms’ ROA. Tests are conducted using linear regression 105 firms in the high-tech industry, listed on the S&P 1500 in the period of financial crisis between January 2008 and December 2012. The CEO narcissism variable is based on both firm-specific accounting-based narcissism criteria and on the more personal option-delay of exercisable options based narcissism criterion.

The results indicate that CEO narcissism has a negative impact on the ROA of the narcissistic CEO firms and that volatility of these firms’ ROA is higher compared to the firms with a non-narcissistic CEO. The negative relationship between CEO narcissism and ROA found within periods of economic crisis is important for both the firms of the narcissistic CEOs and its (potential) investors.

Author: Melvin J. H. Welsink Student ID: 0582948 First supervisor: S. Bissessur

(2)

Statement of Originality

This document is written by student Melvin Welsink who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

I. Introduction and research question

In this article, the relationship between Chief Executive Officer (CEO) characteristics and the financial performance of the CEO’s firm is analyzed. Executives, and CEOs because of their central role within the executive team in particular, are responsible for the organizational strategies of a firm and these organizational strategies are related to the financial performance of a firm. The upper echelon theory developed by Hambrick and Mason (1984), based on the theory of Child (1972), presumes that executives’ perceptions, experiences, cognitions and values reveal themselves in the decision making process. As Rijsenbilt states “Personal characteristics influence executives’ rational and irrational beliefs and actions” (Rijsenbilt, 2011) or as Hambrick and Mason argue “organizations become reflections of their top executives” (Hambrick and Mason, 1984). CEOs have a unique position within their company and within the companies’ executive team and their personal characteristics will therefore have an impact on organizational performance (Daily and Johnson, 1997; Norburn, 1989). The proxy used for CEO characteristics in this research is “narcissism” and the goal of this research is to improve the understanding of the influence of characteristics of narcissistic CEOs on financial performance.

Since the introduction of narcissism in psychology by Havelock Ellis (1898) in the 19th century, a lot is said about this characteristic. Ellis used narcissism to describe people absorbed in self-admiration, the well-known neurologist Sigmund Freud argued that narcissism is a state in which people wish to act out of desire to pursue a self-ideal (1914/1957). In the decades hereafter, researchers’ thoughts on narcissism changed from a clinical disorder (Kernberg, 1975), to a social trend (Lasch, 1978), to a psychological construct (Raskin and Terry, 1988; Vazire and Funder, 2006). According to the “Diagnostic and Statistical Manual of Mental Disorders” (DSM V, 2013), narcissism is a pervasive pattern of grandiosity, need for admiration and a lack of empathy. DSM V labels high levels of narcissism as a Narcissistic Personality Disorder (NPI) which could lead to dysfunctional behaviour. In literature, the narcissism characteristic of CEOs has been described both as necessary, productive, healthy or constructive and as negative and destructive for the CEO’s firm. In Maccoby, 2000, both positive and negative characteristics of a narcissistic CEO are listed. Their great vision and their talent to attract followers can help them in creating value for the firm. However, their sensitiveness for criticism, poor listening skills and their lack of

(6)

2

empathy should also be taken into consideration. Also, the excessive optimism and confidence of narcissistic CEOs could lead to excessive risk taking (Shapira, 1995).

In this empirical research the financial performance of firms with narcissistic CEOs are compared with non-narcissistic CEO-firms within the same industry. The narcissism proxy is measured by a combined measure of industry-adjusted excess investments, acquisitions, debt-to-equity and the use of risky debt. A “delayed-option-exercise” measure is added to this four-term-proxy in an alternative regression.

The accounting based measure Return On Assets is used as proxy for financial performance. ROA is an indicator of a firm’s profitability relative to its total assets, it provides an indication of the efficiency of the management in using the firm’s assets to generate earnings. The use of ROA instead of other financial performance measures is preferred, since ROA avoids potential distortions created by financial strategies, such as growing debt leverage, stock splits and stock buybacks, whereas other financial performance measures as Earnings Per Share and Return On Equity are sensitive for these strategies (John Hagel III, 2010).

The relationship between narcissistic CEOs and ROA will be tested in a specific time frame, from 2008 till 2012, during the economical recession. In times of economical recession, if a relation between the narcissistic characteristic and ROA is (still) found, it could provide clarity on the cause of the relationship found. If firms of narcissistic CEOs have better financial performance within periods of crisis, better management of real and operational activities are a reasonable explanation. This is unlikely to be the case if ROA is lower for narcissistic-CEO firms. This suggests higher risk-taking (less risk-averse because of overconfidence) leads to higher ROA (Olsen et al., 2013; Schrand and Zechman, 2012) in years of booming economy compared to non-narcissistic CEO firms and consequently leads to lower ROA in years of crisis.

Testing the narcissistic personal characteristic of CEOs and its possible relation to ROA is important for investors for two main reasons. First of all, it could clarify the objectives of narcissistic CEOs. How well are they looking after the investors’ interests, in other words, is there (good) stewardship by the CEO? Secondly, knowing the consequences of CEO narcissism on financial performance could improve the awareness of potential risks or opportunities, especially in times of economic crisis. When there is a negative relationship between narcissistic CEOs and ROA (due to relative higher risk), investors could take multiple actions; (1) They could choose to invest their money in a different company with a non-narcissistic CEO, (2) they could (together) replace the CEO for a non-narcissistic CEO, and (3) they could also, because the positive relationship with ROA in prior research, choose

(7)

3

to maintain the narcissistic CEO and trust on the positive characteristics belonging to a narcissist. A possibility is to combine the narcissistic CEO with a non-narcissistic, trusted sidekick as Maccoby (2000) suggests. By adding a sidekick who understands the CEO and what he is trying to achieve, but at the same time keeps the CEO grounded and helps focusing on the right point and prevents him from making disaster decisions. The full potential of the narcissistic CEO can be used while the risks are limited to a certain extent. It is a way of exercising increased monitoring on narcissistic CEOs.

Even though research on narcissistic CEOs increased in the last two decades, the specific relationship with ROA has not often been studied in the present literature. This research contributes to the present literature because prior research was performed within years of thriving, or at least stable economies, while this research focuses on the most recent financial crisis of 2008-2012. Analyzing the relationship between narcissism and ROA within these years could increase our understanding of this relationship and its causes.

The research question of this empirical study is “Did CEO-narcissism decrease Return On Assets during the financial crisis of 2008-2012?”.

In other words, is ROA lower for narcissistic-CEO firms, compared to non-narcissistic-CEO firms in the same industry during the crisis of 2008-2012?

The remainder of this paper is organized as follows: the next section, section II, contains a literature review; in section III the research methodology is explained; section IV gives an overview of the results and the conclusion of the research can be found in section V.

II. Literature review

2.1 CEO influence on financial performance

The executive team of a firm is responsible for the organizational strategies of a firm and these organizational strategies are related to the performance of a firm (Crossland and Hambrick, 2007; Child, 1972). Based on the theory of Child (1972), Hambrick and Mason (1984), developed their upper echelon theory. This theory assumes that the perceptions, experiences, cognitions and values of executives reveal themselves in the decision making process. The executives make decisions based on their personal interpretations of the faced

(8)

4

situation, and these interpretations are based on the experiences, values and personalities of the executives, in other words, on the executives’ personal characteristics. Hambrick and Mason therefore argue that “organizations become reflections of their top executives”. Consistent with this thought, Shapira (1995) mentions that a set of human factors causes decision makers to differ in their risk-taking behaviour, some individuals are less risk-averse than others. In terms of risk taking by executives, making acquisitions and invest in new capacity for example, it is likely to have various (performance) outcomes when individuals differ in risk-appetite. Chatterjee and Hambrick (2011) mention that narcissism can be expected to play an important role in CEO personality factors that enter into risk taking. Although the upper echelon theory is based on executives with the CEO as a “normal” member of the executive team, the individual characteristics of the CEO may be reckoned to have influence on the financial performance. Finkelstein (1992) confirms the power of the team of executives, but also mentions that power may emanate from a managers’ personality, that the CEO is usually “the most powerful member of this group” and the dominance of the CEO within the team. Norburn (1989) considers the CEO to have a unique position within the firm by mentioning the CEO as “THE corporate leader” who will influence organizational performance. The influence on this performance is bigger for US-firm CEOs than for CEOs in other advanced economies such as Germany and Japan (Crossland and Hambrick, 2007). The reason provided by Crossland and Hambrick for the bigger influence in the US is that levels of constraints on CEOs differ per country and CEOs in countries with lower constraints could better shape the firm’s strategy to their own desire.

Other researchers found that CEO characteristics can predict strategic actions and orientations (Gupta and Govindarajan, 1984; Miller et al., 1982) and Nohria et al. (2003) found that 15% of the total variance in the profitability of a firm is due to CEO influence. Rijsenbilt (2011) states that because of the importance of a CEO within an executive team, the CEO is a central determinant influencing the upper echelon and this upper echelon affects the organizational economic performance outcomes. Summarized, CEOs can be considered as executives with a central role within the executive team and therefore can influence the financial performance of the firm.

2.2 Financial performance: ROA

Financial performance measures are used by both external and internal users. External users make use of the performance measures for making investing-, financing- and benchmarking

(9)

5

decisions and internal users, managers and executives for instance, to improve, grow, reward and learn. Aliabadi et al. (2013) explain that performance measures are only useful when they are “value relevant”. It must have a predictive association with equity values.

By using ratios instead of absolute measures the usefulness of financial performance measures increases since ratios provide more clarity about efficiency and are easier to compare with each other (Aliabadi et al., 2013). The most common accounting based ratios are Return On Sales (ROS), Return On Investment (ROI), Return On Equity (ROE) and Return On Assets (ROA).

The most common market based ratios are Price to Book, Market To Book, Price To Earnings, Dividend Yield and Tobin’s Q. The use of market based ratios can be doubtful because their values reflect share prices and these share prices may be based more on expected performance and market imperfections than on true performance (Aliabadi et al., 2013). Stock prices in imperfect markets can be influenced by managers’ decisions not to disclose all available information to investors (Bettis, 1983).

Focusing on the accounting based measures reveals some disadvantages for some of the ratios as well. ROI does not represent the likelihood of costs and returns matching predictions and could lead investors to rely solely on the internal rate of return without looking at the total return (for example, prefer a $2 return on a $1 investment to a $20.000 return on a $13.000 investment) (Guzman, n.d.). ROI could also neglect the increased leverage from invested debt financing. Tosi et al. (2000) mention that ROE together with ROA are the most explanatory with respect to a firm’s financial performance. ROE focuses on the return to the firm’s shareholders. The potential harm of ROE is that firms could use financial strategies to maintain a healthy ROE and hide financial problems. ROE is sensitive for strategies such as growing debt leverage, stock splits and stock buybacks through accumulated cash (John Hagel III, 2010). ROA is a financial performance measure which shows a firm´s profitability relative to its total assets. It provides an indication of the efficiency of the management in using the firm’s assets to generate earnings.

ROA avoids the possible distortions created by these financial strategies and also is a better metric for financial performance than ROS since it is measuring the ability of the firm to generate return given the assets rather than showing the robust ROS. Hagel III et al. (2010) mention that asset-heavy firms with require a higher level of net income to generate profits relative to asset-light firms. A disadvantage of ROA is that no indication about the financing of the assets is given, assets could have been paid through leveraging (Boundless, 2015).

(10)

6

There is no perfect measure for capturing financial performance, and ROA is not a perfect measure either. Keeping this in mind, ROA could be seen as the most effective, and broadly available, financial measure for measuring financial performance. The ratio is less vulnerable for financial (misleading) strategies compared to other accounting based measures and ROA is looking at both income statement performance and the assets required to generate profits (Hagel III et al., 2013).

2.3 Narcissism

The origin of the word narcissism comes from the Greek mythology. Narcissus was a hunter, known for his pride and his beauty who refused the love of a nymph because he believed he was so much better than anyone else. Aphrodite, the goddess of love, punished Narcissus with self-love. Narcissus fell in love with his own reflection in a pool and eventually killed himself because he could not have his object of desire (Greek Myths & Greek Mythology, 2015). Since the introduction of the term “narcissism” in psychology by Havelock Ellis (1898) in the 19th century, a lot is said about this characteristic. Ellis used narcissism to describe people absorbed in self-admiration, the well-known neurologist Sigmund Freud argued that narcissism is a state in which people wish to act out of desire to pursue a self-ideal (1914/1957). In the decades hereafter researchers’ thoughts on narcissism changed from a clinical disorder (Kernberg, 1975), to a social trend (Lasch, 1978), to a psychological construct (Raskin and Terry, 1988; Vazire and Funder, 2006). According to the “Diagnostic and Statistical Manual of Mental Disorders” (American Psychiatric Association, 2000), narcissism is a pervasive pattern of grandiosity, need for admiration and a lack of empathy. DSM IV labels high levels of narcissism as a Narcissistic Personality Disorder (NPD) which could lead to dysfunctional behaviour. Since narcissism is a personality disorder, the narcissism characteristics can be assessed as stable. In other words, narcissism is considered to be invariant across years. It is not possible to be narcissistic in a certain year, and non-narcissistic in another year (Chatterjee and Hambrick, 2007; Olsen, Dworking, Young, 2013). Previous research suggests that narcissists have very inflated views of themselves and are preoccupied with having those views continuously reinforced (Campbell et al., 2004). Narcissists are very socially skilled, although they are generally not interested in emotional closeness and intimacy. Their social skills and relationships serve the function of self enhancement, not to develop intimacy (Brunell et al., 2008). Narcissists developed skills at initiating relationships with others since they need these relationships to acquire and maintain

(11)

7

their inflated self-views. Their main motivation is applause and admiration and they engage in bold, attention-getting behaviour to gain this attention (Wallace and Baumeister, 2002) and when criticized they react with extreme anger and denial (Kernis and Sun, 1994). Other key elements of highly narcissistic personalities are: sense of entitlement, desire to be the centre of attention, arrogance based on a sense of superiority, and self absorption (Ames et al., 2006). Narcissists maintain these inflated senses of self-view, regardless of their actual accomplishments and their objective performance (Raskin et al., 1991; Rhodewalt and Eddings, 2002). Performance feedback does not influence their predictions about their (own) performances in the future (Campbell et al., 2004) and social praise is far more important than objective performance (Chatterjee and Hambrick, 2011; Vazire and Funder, 2006; Wallace and Baumeister, 2002).

Roll (1986) was one of the first who invoked the idea of (over)confidence having a major impact on executive risk taking with his “hubris hypothesis”: CEOs make acquisitions because they believe they have the ability to make better deals and to manage acquisitions better than their peers. Narcissism describes a larger set of characteristics than overconfidence does (Campbell et al., 2004), but overconfidence can be seen as an important part of narcissism. The primary difference between narcissists and pure overconfident individuals is that narcissists care little about others where overconfident individuals could not only be overconfident about their own abilities but also about the abilities of others; overconfident individuals do not need constant recognition and attention per se (Campbell et al., 2004; Ham et al., 2014). Narcissism is a significant predictor for overconfidence (Campbell et al., 2004) and narcissists are likely to be subject of overconfidence, but will also have other behavioural characteristics that will influence decision making (Ham et al., 2014). After Roll (1986) and his proven relationship between acquisitions and confidence, more evidence of the influence of confidence on risk taking appeared in literature. Besides the relation with acquisitions Hayward and Hambrick (1997) confirmed the results of Roll considering the more frequent acquisitions and showed that the acquisitions made by confident CEOs are also larger and overpaid compared to the acquisitions made by the peers. Shapira (1995) conducted interviews with executives and found some evidence for the role of confidence in risk taking. Executives’ confidence is a part of the basis for their decisions to take, or refrain from, certain actions (Shapira, 1995). Overconfidence, is assumed to generally cause individuals to engage in more risk taking than is objectively sensible (Malmendier and Tate, 2005; Simon and Houghton, 2003). Malmendier and Tate (2005) also found that overconfidence by CEOs is related to the holding of in-the-money stock options in their firm.

(12)

8

Maccoby (2000) states that “narcissism can be extraordinarily useful - even necessary” for CEOs, since the great vision of narcissists and their talent to attract followers can help them in creating value for the firm. However, their sensitiveness for criticism, poor listening skills and their lack of empathy should also be taken into consideration. Brunel et al. (2008) found that narcissism predicts emergent leadership; narcissists are more likely to emerge as leaders when no (strong) leader is present.

2.4 CEO narcissism and financial performance

When CEOs are narcissists, impact on organizational strategy, structure and processes can be expected (Chatterjee, 2009). Because of the effect on strategic decision processes and strategic choices, an impact on performance outcomes can be expected as well. Chatterjee and Hambrick (2011) found evidence that CEO narcissism limits the effect of recent firm performance on risk taking. Where non-narcissistic CEOs respond to positive recent objective performance with higher risk taking and to lower recent performance with lower risk taking, narcissistic CEOs are maintaining their views. Their predictions about the future performances are not influenced by objective measures and they consolidate their (higher) risk taking level (Chatterjee and Hambrick, 2011; Raskin et al., 1991; Rhodewalt and Eddings, 2002). This lower risk-averse narcissistic firm leaders make riskier decisions (Rosenthal and Pittinsky, 2006) which may succeed in the long term, but will “destroy the systems that they and others depend on to survive and thrive” in the long run (Campbell et al., 2011).

Ham et al. (2014) find a negative economically significant impact on firm performance with a negative relation between CEO narcissism and Return On Assets. Since narcissistic CEOs are more likely to disregard feedback on their prior decisions from both their own performance outcomes and by other people. Poor decisions will be reflected in overinvestment in less profitable investment options (relative to the non-narcissistic CEO firm peers) and lower firm financial performance measured by ROA (Ham et al., 2014). The negative effect of CEO narcissism on financial performance was observed to the greatest extent in periods of uncertain operating environments. Overconfident managers tend to misperceive ongoing negative Net Present Value (NPV) projects as value creating, since they overestimate the future cash flows of their investment projects and their own ability to accomplish good performance (Heaton, 2002; Malmendier et al., 2011; Malmendier and Tate, 2005). This could result in misperceiving negative NPV projects as positive NPV projects, which

(13)

9

eventually could lead to a market crash of the stock price (Kim et al., 2014). Schrand and Zechman (2012) state that overconfident executives are more likely to make an initial misstatement because of optimism, creating a positive bias, and are thus more likely to start misreporting because they want to “cover up” these misstatements. Eventually investors will be informed about, and financial performance will be negatively influenced by, this earlier misstatements.

Chatterjee and Hambrick (2007) do not find significant results for performance differences in a positive or negative way between narcissistic-CEO firms and their non-narcissistic peers, but they do find higher volatility in organizational performance for narcissistic CEO-firms. Their explanation for this relationship is that narcissistic CEOs are taking bold actions which attract greater attention and therefore larger gains and/or losses.

Olsen et al. (2013) document that narcissistic-CEO firms have higher Earnings Per Share (EPS) after controlling for factors related to the CEO, firm and industry; they suggest two possible explanations for this positive relationship. Firstly, the inclination of executives to manage key financial variables through accounting choices such as discretionary accruals, stock buybacks, or questionable reporting behaviour. Secondly, the inclination of executives to manage financial variables through real activities such as operational changes. In contrast, no evidence was found for a relation between CEO narcissism and accrual-related earnings management (Olsen et al., 2013). These findings suggest that narcissistic personality characteristics of top executives mainly affect financial performance measures through the executive’s decisions and influence over the firm’s operational activities rather than through accrual and accounting decisions. Evidence for a positive effect on innovation by CEO overconfidence has been found by Galasso and Simcoe (2011). They state that overconfident CEOs (especially in competitive industries) pursue more innovation and are more likely to take their firms in a new technological direction. Peterson and Luthans (2003) suggest that optimistic leaders (optimism is a characteristic of narcissists and overconfident individuals) have better worker morale, higher retention rates among their employees and work units are more profitable. Likeminded research concludes that overconfidence and optimism have a positive impact on investment policies (Gervais et al., 2002). Based on the productive narcissist theory of Maccoby (Maccoby, 2003, 2000), narcissistic CEOs result into positive financial outcomes. However, Maccoby (2000) mentions some destructive characteristics of narcissistic CEOs as well and suggests these characteristics should be “controlled” by better monitoring of the CEO and/or a trusted non-narcissistic sidekick of the CEO.

(14)

10

As described, opinions and evidence in literature about impact of narcissistic (and overconfident) CEOs on firms performances are divided and the focus of the research is often on different variables. In general, the research findings are inconsistent and complex in finding a relationship between narcissistic CEOs and firm performance. Some studies find positive relationships, another finds negative relationships and still another neither finds evidence for positive nor for negative relationships but an effect on performance volatility. Given these conflicting findings, this study could probably not distinct the “right” and “wrong” opinions of prior literature with decisive proof, but it could enhance the understanding of when and why a narcissistic CEO can have a positive or negative impact on the financial performance of a company.

2.5 The 2008-2012 crisis

According to the U.S. National Bureau of Economic Research, the latest U.S. recession started December 2007 and ended June 2009 (NBER, 2012). The recession was related to the financial crisis (2007-2008) and subprime mortgage crisis (2007-2009). However, persistent high unemployment lasted till December 2012, together with lower consumer confidence, continuing decline in house values and increases in federal debt and (personal) bankruptcies (Arellano and Kocherlakota, 2008; Michelle Hirsch, 2011; Reinhart and Rogoff, 2010; US CencusBureau, 2013; Wright, 2012).

Hassett, the director of economic policy studies at American Enterprise Institute, argues that the narcissistic leaders and their firms have been instrumental in fuelling the global financial crisis since these leaders took excessive risks and their firms had a lack of control (Hasset, 2009).

In times of economical crisis and subsequent poor financial performance, narcissistic CEOs will continue to retain their elevated self-concepts and their sense of correctness (Greenwald, 1980) and tend to ignore the significance of their objective performance (Raskin et al., 1991; Rhodewalt and Eddings, 2002). Because of the higher risk taking by narcissistic CEOs compared to their non-narcissistic peers (Rosenthal and Pittinsky, 2006), this excessive risk taking is likely to cause lower ROA in times of financial crisis. Another view is that these risky decisions could cause a faster turnaround and recovery. The results of testing the relationship of narcissistic CEOs and their firms’ ROA could give clarity in this dichotomy.

(15)

11

III. Research methodology

3.1 Hypothesis development

The research question of this empirical study is “Did CEO-narcissism decrease Return On Assets during the financial crisis of 2008-2012?”. In other words, is ROA lower for narcissistic-CEO firms, compared to non-narcissistic-CEO firms in the same industry during the crisis of 2008-2012?

This is expected to be the case because of overconfidence of the narcissistic CEOs, they are more likely to predict a positive Net Present Value (NPV) for investments while in reality these investment are not value creating (anymore). In times of stable economy these inaccuracies could persist for, and distributed over, many years without a direct negative effect on ROA. In times of crisis however, more conservative accounting will be present and pressure to adjust the predicted NPV will increase, resulting in a negative adjustment of Net Income and consequently a lower ROA. Therefore, the hypothesis of this research is as follows.

H1: ROA will be lower for firms with narcissistic-CEOs compared to non-narcissistic-CEO firms during the crisis of 2008-2012.

Figure 1 shows the used framework of the conceptual and the operational validity for this research.

Figure 1: Predictive validity Framework

Conceptua l Modera tor

Opera tiona l

Independent Va ri a bl e Theory Dependent Va ri a bl e CEO cha ra cteri s tics Upper Echel on Theory 2008-2012

cri s i s

Fi na nci a l performa nce

Proxy: Na rci s s i s m Regres s i on Model Proxy: ROA Mea s ures : _ROA_t_{= β}₀_{+ β}₁_CEO_NARCIS₊

β2CEOage + β3CEOmale +

β4CEOtenure + β5CEOchair +

β6Si ze + β7ROAt-1 + εt

Mea s ure:

-Exces s i nves tments Net Income

-Exces s a cqui s i tions Total As s ets

-D/E-ra tio

-CEO gender -CEO cha i rma n -Si ze

-Pri or yea r performa nce -Ri s ky debt

-Del a yed option exerci s e Control Va ri a bl es -CEO a ge

(16)

12

3.2 Data

The empirical approach of this study is archival research. The sample is based on CEO firm-years of (machinery, computer and other electronic) hardware firms (selected out of two-digits SIC codes 35 and 36), from now on called high-tech industries, and derived from the ExecuComp database within the period of 2006 till 2012. In this database, all current and past S&P 1500 index companies are covered. The data of 2006 and 2007 are used for computing the CEO narcissism variable, and the 2008-2012 data are used for the variables necessary for running the regression of ROA. The financial data for the sample have been obtained from the Compustat database. When data necessary for observing narcissism or computing ROA are not present in these databases, the observations will be deleted from the sample.

High-tech industries were chosen for several reasons. Firstly, a large number of firms of these industries are publicly held and therefore data for measuring the narcissism of CEOs is available in databases. Secondly, the firms within these industries are normally not subject to a high degree of external constraints and/or regulations (Chatterjee and Hambrick, 2011, 2007). Therefore, it is more likely that a CEO’s real narcissistic level is measured; when a CEO is restricted because of a lot of regulations his influence on important (investment) decisions will be less significant and the firm’s characteristics will be measured instead of the CEO’s characteristics. The same industries were mainly used in prior studies by Schrand and Zechman (2012) and Chatterjee and Hambrick (2011, 2007).

Looking at the share prices of the high-tech industries between January 1, 2008 and December 31, 2012, an average decrease in share price of 12.39% can be detected. This confirms that the selected industries were subject of the 2008-2012 crisis described before.

3.3 Variables

3.3.1 Dependent variable: Return On Assets

In this study, the effect of CEO narcissism on Return on Assets (ROA) is tested. The

dependent variable, the ROA of a firm, is an indicator of the profitability of a firm relative to its total assets; the ROA gives an indication of the efficiency of the management in using the firm’s assets to generate earnings, in formula:

(17)

13

The ROA per firm can be computed with information from the Compustat database. The Net Income (Compustat code NI) and total assets (Compustat code AT).

3.3.2 Independent variables: Narcissism variable

Most ideal, CEO narcissism is measured by the Narcissistic Personality Inventory (NPI) but in this study another approach to measure narcissism is used. Because interviews are and/or survey research is necessary for determining the NPI and CEOs of public companies are very reluctant to participate in survey research, it would yield in influenced answers and low participation (Cycyota, 2006). Finishing this study within a few months would not be possible. Therefore, there has been chosen for measuring CEO narcissism with objective indicators publicly available.

Accounting-based narcissism indicator

This indicator assumes that overconfident CEOs are consistently optimistic about (investment) decisions. The data required for estimating this variable is available in the Compustat database.

The variable takes the value of one, more likely to have a narcissistic CEO, if the firm meets at least three of the following four criteria and zero otherwise (Schrand and Zechman, 2012):

1) Industry-adjusted excess investment; to compute this excess investment a regression of total asset growth (increase in AT in Compustat) on sales growth (increase in SALE in Compustat) has to be assessed for a firm and from the resulting residual, the three-digit SIC code industries median residual has to be deducted. An excess investment for narcissistic CEOs is predicted because these CEOs are more likely to overestimate the NPVs of the investments.

2) Industry-adjusted net dollars of acquisitions made by the firm. When a firm’s acquisitions (code AQC) are in excess of the high-tech industries median (per year), narcissism is suggested to be present (Malmendier and Tate, 2008).

3) Industry-adjusted debt-to-equity ratio; this indicator is equal to the long-term debt (code DLTT) plus short-term debt (code DLC) divided by the firm’s total market value (market value of equity (code PRCC * code CSHO) plus book values long-term debt (code DLTT) and preferred stock (code PSTK), minus the high-tech industries median in the same year. A debt to equity ratio greater than the median high-tech industries median indicates narcissism (Malmendier et al., 2011).

(18)

14

4) Convertible debt / preferred stock; this criteria has a value of one if the firm uses either convertible debt (code DCVT) or preferred stock (code PSTK) and a value of zero if otherwise. Ben-David et al. (2007) state that narcissistic managers will choose risky debt and found evidence that narcissistic executives have longer debt duration as a measure of risky debt.

The (dummy) variable is displayed as the CEONARCIS-variable with value “1” when on average

three or more criteria are met in 2006 and 2007 and “0” otherwise.

Stock option-based narcissism indicator

A CEO who voluntarily retains in-the-money options is likely to be overconfident about his own ability and the firm’s prospects because the CEO’s human capital is already highly exposed to firm-specific risk (Kim et al., 2014). The indicator uses the timing of option exercises to identify the overconfidence characteristic of narcissistic CEOs. Overconfidence induces narcissistic CEOs to postpone option exercise because they overestimate the future returns of their investment projects (Malmendier and Tate, 2005).

For this study, the option-delay variable is estimated in the same way as Schrand and Zechman (2012) did in their study, who in their turn based their estimation method on the method used by Malmendier and Tate (2005, 2008). If the option delay is greater than the 3-digit industry (SIC score) median, the firm is more likely to have a narcissistic CEO and the variable equals 1. The data required for estimating this variable is available in the ExecuComp database, the value of the exercised exercisable options (code opt_unex_exer_est_val) is divided by the total holdings of the CEO in the firm (prccf*number of shares+ opt_val). The CEO’s total holdings in the firm, including options and stock holdings, is used since the delay in the option exercise could be related to the CEO’s total holdings in the firm (Schrand and Zechman, 2012).

The variable is created by taking the logarithm of this (relative value of the) unexercised, exercisable options. The (dummy) variable (CEOOPTION) will be used as replacement of the

earlier discussed accounting-based firm-specific CEONARCIS-variable in some of the regressions to test if the explanatory power increases when de stock option-based indicator is used as a proxy for CEO narcissism. The variable has a value of “1” when the average of the option delay in 2006 and 2007 is greater than the industry median and “0” otherwise.

(19)

15 Combined narcissism indicator

Besides running regressions with the single narcissism variables CEONARCIS and CEOOPTION,

another narcissism variable is used to measure CEO narcissism. This variable is created identically as the CEONARCIS before, the four accounting-based criteria will be compared with the industry median. But unlike the CEONARCIS variable, the new combined CEO narcissism variable has five instead of four criteria. As fifth criteria the stock-option based narcissism variable is added. By combining the four accounting-based criteria and the (more person-specific) option-delay variable, the new variable should be better in predicting the CEO narcissism. This combined indicator for CEO narcissism has a value of “1” if the firm meets more than three of the five criteria and “0” otherwise. The combined CEO narcissism variable is displayed as CEONARCIS5>3.

3.3.3 Control variables

Multiple control variables are used in the regression testing the possible relationship between narcissistic CEOs and their firms’ ROA.

To control for CEO characteristics, control variables for CEO age (CEOAge), gender (CEOMale, value of 1 when male), tenure (CEOTenure, per year, with n > 3) and for CEOs being chairman of the board at the same time (CEOChair) are constructed.

The last two control variables control for the firm characteristics and are the SIZE (computed as the natural log of total assets at the end of prior year, controlling for effect of firm size) and ROAt-1 (controlling for the effects of performance in the prior year) variables.

3.4 Regression

All the variables taken together, the main regression for tests and analyses is as follows:

𝑅𝑂𝐴𝑡 = 𝛽0+ 𝛽1𝐶𝐸𝑂𝑁𝐴𝑅𝐶𝐼𝑆+ 𝛽2𝐶𝐸𝑂𝐴𝑔𝑒,𝑡+ 𝛽3𝐶𝐸𝑂𝑀𝑎𝑙𝑒+ 𝛽4𝐶𝐸𝑂𝑇𝑒𝑛𝑢𝑟𝑒,𝑡 + 𝛽5𝐶𝐸𝑂𝐶ℎ𝑎𝑖𝑟,𝑡 + 𝛽6𝑆𝐼𝑍𝐸𝑡−1+ 𝛽7𝑅𝑂𝐴𝑡−1+ 𝜀𝑡

(2)

SPSS is used to run the regression and obtain the results, further explanation about the used variables can be found in Appendix B.

(20)

16

IV. Results

4.1 Descriptive Statistics

After merging all relevant data from the Execucomp and Compustat databases, a total of 105 firms of the high-tech industries meet the conditions of having enough information publicly available for testing their CEOs on narcissism and testing this composite variable. The names of these firms and their CEOs can be found in Appendix A. These 105 firms provide a sample of 448 CEO firm-years between 2008 and 2012 which are tested with regression (2). Of this sample, most companies are in the industry of “Electronic Components and Accessories”, the composition of the total sample per three-digit SIC-code is shown in Table 1 below.

Table 1. Total sample per thee-digit SIC-code

3-digit SIC code

Description CEO

Firm-years (N)

% of sample

351. ENGINES AND TURBINES 4 0.9

352. FARM AND GARDEN MACHINERY AND EQUIPMENT 14 3.1 353. CONSTRUCTION, MINING, AND MATERIALS HANDLING MACHINERY AND EQUIPMENT 34 7.6

354. METAL WORKING MACHINERY AND EQUIPMENT 5 1.1

355. SPECIAL INDUSTRY MACHINERY, EXCEPT METALWORKING MACHINERY 21 4.7 356. GENERAL INDUSTRIAL MACHINERY AND EQUIPMENT 24 5.4

357. COMPUTER AND OFFICE EQUIPMENT 35 7.8

358. REFRIGERATION AND SERVICE INDUSTRY MACHINERY 9 2.0 360. ELECTRONIC AND OTHER ELECTRICAL EQUIPMENT AND COMPONENTS, EXCEPT COMPUTERS 5 1.1 361. ELECTRIC TRANSMISSION AND DISTRIBUTION EQUIPMENT 4 0.9

362. ELECTRICAL INDUSTRIAL APPARATUS 19 4.2

363. HOUSEHOLD APPLIANCES 18 4.0

364. ELECTRIC LIGHTING AND WIRING EQUIPMENT 14 3.1

365. HOUSEHOLD AUDIO AND VIDEO EQUIPMENT 5 1.1

366. COMMUNICATIONS EQUIPMENT 42 9.4

367. ELECTRONIC COMPONENTS AND ACCESSORIES 179 40.0 369. MISCELLANEOUS ELECTRICAL MACHINERY, EQUIPMENT, AND SUPPLIES 16 3.6 Total: 448 100 Summary of the industries per 3-digit SIC code of all the CEO firm-years used and the total percentage per SIC-industry.

Table 2 reports descriptive statistics for all variables of regression (2). For the three dummy-variables the frequency of occurrence is shown and for all dummy-variables the mean, standard deviation (volatility) and quartiles are shown. Using the information from Table 2, it can be observed that 14.96% of the tested CEO firm-years have a narcissistic CEO, only 1.12% of the CEO firm-years have a woman as CEO, most of the CEOs are also Chairman of the Board and the average ROA is positive with 3.73% between 2008 and 2012.

(21)

17

Table 2. Descriptive Statistics for independent, control and dependent variables

N Frequency (Dummy) Mean Standard Deviation Quartiles 0 1 First (25%) Median Third (75%) CEONARCIS 448 381 67 0.1496 0.3570 0.0000 0.0000 0.0000 CEOAge,t 448 - - 57.1518 7.1681 52.0000 56.0000 61.0000 CEOMale 448 5 443 0.9888 0.1052 1.0000 1.0000 1.0000 CEOTenure,t 448 - - 11.4978 7.9371 6.0000 9.0000 14.0000 CEOChair,t 448 173 275 0.6138 0.4874 0.0000 1.0000 1.0000 SIZEt-1 448 - - 7.3299 1.4692 6.1380 7.1917 8.2835 ROAt-1 448 - - 0.0457 0.0932 0.0105 0.0549 0.0877 ROAt 448 - - 0.0373 0.1058 0.0137 0.0540 0.0862

CEONARCIS, proxy for CEO narcissism based on four criteria: 1) Industry-adjusted excess investment, 2) Industry-adjusted relative dollars of

acquisitions, 3) Industry-adjusted debt-to-equity ratio and 4) the use of convertible debt and/or preferred stock. If “1” the average sum of these four criteria is >2.5 in 2006 and 2007, ”0” otherwise. CEOAge,t: Age of the CEO in sample-year. CEOMale: “1” if CEO is male, “0” if

CEO is female. CEOTenure,t: Tenure of the CEO in sample year, minimum value of 4. CEOChair,t: "1" if CEO is chairman of the board in

sample-year, "0" otherwise. SIZEt-1:natural log of the total assets at the end of the prior year. ROAt-1: Return On Assets of prior year. ROAt:

Return On Assets of sample year. Extensive variable definitions are provided in Appendix B.

Since only one of the firms has a female as CEO, the CEOMale variable is excluded from the regression because the explanatory value of this variable would be low and could be misleading. The regression tested therefore is:

𝑅𝑂𝐴_𝑡= 𝛽₀ + 𝛽₁𝐶𝐸𝑂_{𝑁𝐴𝑅𝐶𝐼𝑆}+ 𝛽₂𝐶𝐸𝑂_{𝐴𝑔𝑒,𝑡}+ 𝛽₃𝐶𝐸𝑂_{𝑇𝑒𝑛𝑢𝑟𝑒,𝑡}+ 𝛽₄𝐶𝐸𝑂_{𝐶ℎ𝑎𝑖𝑟,𝑡} + 𝛽₅𝑆𝐼𝑍𝐸_𝑡−1+ 𝛽₆𝑅𝑂𝐴_𝑡−1+ 𝜀_𝑡

(3)

The correlations of regression’s (3) variables are reported in Table 3a (Spearman correlations) and Table 3b (significance of Spearman correlations). The correlation between the narcissism proxy variable CEONARCIS and CEOAge, CEOChair and SIZE are highly significant, however the correlation of the narcissism variable is not significant with ROA. The firm size variable is the only variable significant correlated with all other variables and the ROA is (positive) significant correlated with CEOs being chairman of the board, firm size and the ROA of the year before. Strong correlations between some explanatory variables suggest there could be multicollinearity. Multicollinearity leads to higher standards errors, which enlarges confidence intervals of the variables’ coefficients and therefore reduces the t-values. Consequently, coefficients have to be larger to be significantly different from zero.

Mason and Lind (1996) stated that correlations between -0.70 and 0.70 do not cause problems regarding multicollinearity. Testing for multicollinearity in SPSS by looking at the Variance Inflation Factor (VIF) of the variables gives no VIF-values above 1.69, where VIF values are considered to be highly multicollinear above 10, and multicollinear above 5 (Burns, 2006).

(22)

18

Based on both literature and testing in SPSS the assumption can be made that no significant multicollinearity is present between the used variables.

Table 3a. Spearman correlations

CEONARCIS CEOAge,t CEOTenure,t CEOChair,t SIZEt-1 ROAt-1 ROAt

CEONARCIS 1.000 CEOAge,t 0.156 1.000 CEOTenure,t -0.040 0.442 1.000 CEOChair,t 0.165 0.110 0.217 1.000 SIZEt-1 0.173 -0.171 -0.247 0.115 1.000 ROAt-1 -0.047 0.018 0.136 0.130 0.117 1.000 ROAt -0.047 0.000 0.007 0.120 0.201 0.513 1.000

Table 3b. Significance of Spearman correlations

CEONARCIS CEOAge,t CEOTenure,t CEOChair,t SIZEt-1 ROAt-1 ROAt

CEONARCIS CEOAge,t ,001** CEOTenure,t .397 ,000** CEOChair,t ,000** ,020* ,000** SIZEt-1 ,000** ,000** ,000** ,015* ROAt-1 .318 .697 ,004** ,006** ,013* ROAt .317 .993 .875 .011* ,000** ,000**

Correlations of the variables used in the regression of CEONARCIS on ROA. CEONARCIS: proxy for CEO narcissism based on four criteria: 1)

Industry-adjusted excess investment, 2) Industry-adjusted relative dollars of acquisitions, 3) Industry-adjusted debt-to-equity ratio and 4) the use of convertible debt and/or preferred stock. If “1” the average sum of these four criteria is >2.5 in 2006 and 2007, ”0” otherwise.

CEOAge,t: Age of the CEO in sample-year. CEOTenure,t: Tenure of the CEO in sample year, minimum value of 4. CEOChair,t: "1" if CEO is

chairman of the board in sample-year, "0" otherwise. SIZEt-1:natural log of the total assets at the end of the prior year. ROAt-1: Return On

Assets of prior year. ROAt: Return On Assets of sample year. Extensive variable definitions are provided in Appendix B.

**_{: significant at the 0.01 level.}*_{: significant at the 0.05 level.}

4.2 Accounting-based narcissism variable

4.2.1 Results

The empirical results in Table 4 report an adjusted R2 of 0.189 which means 18.9% of the variation in ROA is explained by the variation of the variables used in the regression. Table 4 further indicates (only) marginal negative significance for the coefficient of the narcissism proxy CEONARCIS. Although there is a marginal indication Hypothesis 1 could be correct, it should be rejected since evidence for a negative relationship between CEO narcissism and ROA is lacking and cannot be extracted from the regression.

(23)

19

Table 4. Regression of CEONARCIS on ROA

β

Standard

Error t-value Sign.

α (Constant) -0,134** .048 -2.762 .006 CEONARCIS -0.024 .013 -1.838 .067 CEOAge,t 0,001 .001 1.651 .099 CEOTenure,t -0.001 .001 -1.005 .316 CEOChair,t 0,022* .010 2.251 .025 SIZEt-1 0,011** .003 3.169 .002 ROAt-1 0,416** .049 8.444 .000 CEO firm-years 448 CEOs 77 R2 .200 Adjusted R2 .189

Results of the regression of CEONARCIS on ROA. CEONARCIS, proxy for CEO narcissism based on four criteria: 1) Industry-adjusted excess

investment, 2) Industry-adjusted relative dollars of acquisitions, 3) Industry-adjusted debt-to-equity ratio and 4) the use of convertible debt and/or preferred stock. If “1” the average sum of these four criteria is >2.5 in 2006 and 2007, ”0” otherwise. CEOAge,t: Age of the CEO in

sample-year. CEOTenure,t: Tenure of the CEO in sample year, minimum value of 4. CEOChair,t: "1" if CEO is chairman of the board in

Return On Assets of sample year. Extensive variable definitions are provided in Appendix B. **_{: significant at the 0.01 level.}*_{: significant at the 0.05 level.}

Judging from the results, the ROA is mainly affected by the ROA of the previous year, since there is a strong significantly positive coefficient for this control variable. Besides the prior ROA, the variables for firm size and CEOs being chairman of the board have a positive relationship with the ROA.

4.2.2 Robustness

The marginally significant coefficient of the narcissism variable in Table 4 suggests that a higher degree of narcissism could have a bigger impact on ROA. Up to now, the narcissism variable had a value of “1” when on average (in 2006 and 2007) three or more of the four narcissism criteria (industry-adjusted excess investment, industry-adjusted acquisitions, industry-adjusted debt-to-equity ratio and use of convertible debt and/or preferred stock) were met. In the next regression the same variables are used, but the independent variable CEONARCIS>3 has a value of “1” if on average more than three (so 3.5 or 4) out of the four narcissism criteria are met. By adjusting the cut-off value in this way, the number of narcissistic CEO firm-years (the frequency of a value of “1” for the dummy CEONARCIS>3) is 18, where it was 67 CEO firm-years in the previous regression (see Table D1 in Appendix D). In Appendix D the correlation tables (D2a and D2b) can be found, there is no significant multicollinearity and the narcissism variable CEONARCIS>3 is not significantly correlated with

(24)

20

ROA. Correlations between the other variables are exactly the same as in the correlation matrix, Table 3ab, described above.

Table 5 reports the variables’ coefficients and their significance of the regression on ROA when the adjusted narcissism variable CEONARCIS>3 is used. The R2 and the adjusted R2 both slightly increase compared to the previous regression, which means that the explanatory value of the used variables increased in this regression. Since only the narcissism variable changed, the conclusion can be drawn that the adjusted variable CEONARCIS>3 has greater explanatory value than the unadjusted narcissism variable CEONARCIS. Identical to the previous regression, the coefficient for ROAt-1, SIZEt-1 and the constant (α) are highly significant, and the coefficient for CEOChair is significant. Most importantly, a highly significant coefficient for the narcissism variable CEONARCIS>3 can be observed in Table 5, representing a strong negative relationship between narcissistic CEOs and their firms’ ROA. The results of the regression with the adjusted (cut-off values for the) narcissism variable indicate there is enough proof to state that Hypothesis 1 is correct at both a 5% significance level as at a 1% significance level. From 2008 to 2012, firms with narcissistic CEOs have an average ROA that is 7.0% lower compared to firms within the same industries without a narcissistic CEO.

Table 5. Regression of adjusted narcissism variable CEONARCIS>3 on ROA

β

Standard

α (Constant) -0,143** .048 -2.973 .003 CEONARCIS>3 -0,070** .023 -2.979 .003 CEOAge 0,001 .001 1.929 .054 CEOTenure -0.001 .001 -0.908 .364 CEOChair 0,020* .010 2.009 .045 SIZE 0,010** _.003 _3.090 _.002 ROAt-1 0,413** .049 8.427 .000 CEO firm-years 448 CEOs 77 R2 .210 Adjusted R2 .199

CEONARCIS>3, proxy for CEO narcissism based on four criteria: 1) Industry-adjusted excess investment, 2) Industry-adjusted relative dollars

of acquisitions, 3) Industry-adjusted debt-to-equity ratio and 4) the use of convertible debt and/or preferred stock. If “1” the average sum of these four criteria is >3 in 2006 and 2007, ”0” otherwise. CEOAge,t: Age of the CEO in sample-year. CEOTenure,t: Tenure of the CEO in

sample year, minimum value of 4. CEOChair,t: "1" if CEO is chairman of the board in sample-year, "0" otherwise. SIZEt-1:natural log of the

total assets at the end of the prior year. ROAt-1: Return On Assets of prior year. ROAt: Return On Assets of sample year. Extensive variable

(25)

21

4.3 Stock-option based narcissism variable

In this section a stock-option based narcissism variable (CEOOPTION) replaces the earlier used accounting-based narcissism variables. The stock-option based variable has been composed in the same way as Schrand and Zechman (2012) did in their study and measures the delay of exercising exercisable options by the CEO. The consequent regression is displayed below:

𝑅𝑂𝐴_𝑡 = 𝛽₀+ 𝛽₁𝐶𝐸𝑂_{𝑂𝑃𝑇𝐼𝑂𝑁}+ 𝛽₂𝐶𝐸𝑂_{𝐴𝑔𝑒,𝑡}+ 𝛽₃𝐶𝐸𝑂_{𝑇𝑒𝑛𝑢𝑟𝑒,𝑡}+ 𝛽₄𝐶𝐸𝑂_{𝐶ℎ𝑎𝑖𝑟,𝑡} + 𝛽₅𝑆𝐼𝑍𝐸_𝑡−1+ 𝛽₆𝑅𝑂𝐴_𝑡−1+ 𝜀_𝑡

(4)

4.3.1 Results

With the CEOOPTION variable, 201 of the 448 firm-years (44.9%) are predicted to have a narcissistic CEO (see Table D3 in Appendix D). This is a large increase compared to the number of narcissistic years predicted by the CEONARCIS (15,0%) and CEONARCIS>3 (4.0%) variables. The correlations of the variables of regression (4) can be found in Table D4a and D4b of Appendix D, no significant multicollinearity is observed and, in contrast to the previous narcissism variables, the CEOOPTION variable is highly correlated with the ROA of the firm. Thereby, the option-based narcissism variable has a highly significant correlation with the control variables of firm size, prior year ROA and CEO age, and a significant correlation with the variable measuring CEOs being chairman of the board.

In table 6 the results of regression (4) on ROA with the option-based variable for narcissism are shown.

The adjusted R2 shows an explanatory value of the variables of 18.5% which is slightly lower than the adjusted R2 observed with the accounting-based narcissism variables (18.9% and 19.9%). Apart from the constant, only the ROA of the previous year has a highly significant coefficient and the control variable of firm size has a significant coefficient.

The option-based narcissistic variable CEOOPTION has a minor negative coefficient, but this coefficient is not significantly different from zero.

(26)

22

Table 6. Regression of option-based narcissism variable CEOOPTION on ROA

β

Standard

α (Constant) -0,119** .048 -2.479 .014 CEOOPTION -0.012 .010 1.133 .258 CEOAge 0,001 .001 1.508 .132 CEOTenure -0.001 .001 -0.922 .357 CEOChair 0.019 .010 1.955 .051 SIZE 0,008* _.004 _2.310 _.021 ROAt-1 0,416** .050 8.400 .000 CEO firm-years 448 CEOs 77 R2 .196 Adjusted R2 .185

Results of the regression of CEOOPTION on ROA. CEOOPTION: proxy for CEO narcissism based on the delay of exercising exercisable options

by CEOs. “1” if the average option delay in 2006 and 2007 is greater than industry median delay, "0" otherwise. CEOAge,t: Age of the CEO

in sample-year. CEOTenure,t: Tenure of the CEO in sample year, minimum value of 4. CEOChair,t: "1" if CEO is chairman of the board in

Return On Assets of sample year. Extensive variable definitions are provided in Appendix B. **

: significant at the 0.01 level. *: significant at the 0.05 level.

With CEO narcissism measured by the described option-delay of exercisable options variable CEOOPTION it can be stated that Hypothesis 1 should be rejected. No negative relation is found between CEO narcissism and ROA between 2008 and 2012 for the high-tech firms. In fact, running regression (4) does not provide a significant relation at all between the narcissism and ROA variables.

4.3.2 Robustness

In Appendix C an overview is given of the values of the used narcissism variables. What stands out in this table is that none of the 45 CEOs marked as narcissistic by the CEOOPTION variable is marked as narcissistic by the CEONARCIS>3 variable, while both variables should measure the same characteristic. A deviation between the narcissistic CEOs when measured in different ways is realistic. Especially because CEONARCIS>3 focuses on the accounting-based narcissism and CEOOPTION focuses on the option-delay factor, which is more personal since only the CEO can choose to exercise his exercisable options. However, the absence of any overlap between the two CEO narcissism variables may indicate a questionable reliability of at least one of the used variables.

(27)

23

4.4 Combination narcissism variables

In the next regression CEO narcissism is measured by a combination of the previous used CEONARCIS>3 and CEOOPTION variables. By combining these measures both the accounting-based as the more personal-specific CEO narcissism are captured in one variable. The combined variable is displayed as CEONARCIS5>3 and has the value of “one” if more than three of the five criteria are met and “zero” otherwise. In this way, all the narcissistic CEO firm-years from the regression with CEONARCIS>3 are included, but also the CEO firm-years that only met two and a half or three out of the four firm-based narcissism criteria but have an option delay in exercising exercisable options compared with the industry median.

4.4.1 Results

The descriptive statistics are shown in Table D5 in Appendix D, with the combined narcissism variable, 77 of the 448 CEO firm-years (17.19%) are identified as firm-years with a narcissistic CEO. The results of the regression can be found in Table 7 below, the adjusted R2 shows that 19.1% of the movement in ROA is explained by the movement of the variables used in this regression. Table 7 reports highly significant coefficients for the constant and the control variables firm size and ROA in the prior year. Significant values for the coefficients of CEOs being chairman of the board at the same time and the combined narcissism variable CEONARCIS5>3 are shown. Based on this regression, CEO narcissism has a negative impact on the ROA of the narcissistic CEO’s firm and Hypothesis 1 should be accepted at a 5% significance level.

Table 7. Regression combined narcissism variable CEONARCIS5>3 on ROA

β

Standard

α (Constant) -0,142** .049 -2.904 .004 CEONARCIS5>3 -0,026* .012 -2.118 .035 CEOAge 0,001 .001 1.752 .080 CEOTenure -0.001 .001 -0.977 .329 CEOChair 0,021* .010 2.117 .035 SIZE 0,011** _.003 _3.342 _.001 ROAt-1 0,417** .049 8.493 .000 CEO firm-years 448 CEOs 77 R2 .202 Adjusted R2 .191

(28)

24

Results of the regression of CEONARCIS5>3 on ROA. CEONARCIS5>3: Combined proxy variable for CEO narcissism based on five criteria: 1)

Industry-adjusted excess investment, 2) Industry-adjusted relative dollars of acquisitions, 3) Industry-adjusted debt-to-equity ratio, 4) the use of convertible debt and/or preferred stock and 5) Industry-adjusted option delay of exercisable options by CEO. “1” if the average sum of these four criteria is >3 in 2006 and 2007, ”0” otherwise. CEOAge,t: Age of the CEO in sample-year. CEOTenure,t: Tenure of the CEO in

sample year, minimum value of 4. CEOChair,t: "1" if CEO is chairman of the board in sample-year, "0" otherwise. SIZEt-1:natural log of the

total assets at the end of the prior year. ROAt-1: Return On Assets of prior year. ROAt: Return On Assets of sample year. Extensive variable

definitions are provided in Appendix B. **: significant at the 0.01 level. *: significant at the 0.05 level.

4.5 Summarizing tables

Table 8 below summarizes the most important information previously mentioned in chapter IV. In this table the following information is shown; 1) the CEO firm-year frequencies of the values “0” and “1” per narcissism variable, 2) the percentages of narcissistic CEO firm-years of the total sample per narcissism variable, 3) the coefficient of each narcissism variable and its significance and 4) the adjusted R2 per regression.

The narcissism variables are all negative, but only the variables CEONARCIS>3 and CEONARCIS5>3 have significant coefficients. The adjusted R2 is highest for CEONARCIS>3 and lowest for CEOOPTION.

Table 8. Frequencies, percentages, regression coefficients and significance, adjusted R2 of the narcissism

variables

Summary of descriptive statistics and regression results of the four used narcissism variables. Extensive variable definitions are provided in Appendix B. **_{: significant at the 0.01 level.}*_{: significant at the 0.05 level.}

Table 9 reports the average and standard deviation of the dependent variable ROA, between 2008 and 2012, per regression. The average ROA is lower for firms of CEOs that were selected as narcissists in comparison with non-narcissistic CEO firms in the regressions with narcissism variables CEONARCIS, CEONARCIS>3 and CEONARCIS5>3. On the other hand, for the regression with the CEOOPTION narcissism variable, the average ROA is higher for firms of narcissistic CEOs compared to non-narcissistic CEO firms. The same distribution is observed for the volatility (measured by the standard deviation) of ROA. Volatility is higher for firms with narcissistic CEOs in the regression samples with the CEONARCIS, CEONARCIS>3 and CEONARCIS5>3 compared to non-narcissistic CEO firms. However, when narcissism is measured by the CEOOPTION variable, volatility is higher for non-narcissistic CEO firms.

0 1 CEONARCIS 448 381 67 14.96% -0.024 .067 .189 CEONARCIS>3 448 430 18 4.02% -0.070** .003 .199 CEOOPTION 448 247 201 44.87% -0.012 .258 .185 CEONARCIS5>3 448 371 77 17.19% -0.026* .035 .191 Narcissism variable used in regression β N Frequency Narcissistic CEOs % Adjusted R2 Sign.

Did CEO-narcissism decrease Return On Assets during the financial crisis of 2008-2012?