1
Master Thesis
The influence of CEO narcissism on the cost of equity
Jeffrey Karamat 2341670
Supervisor – Y. Karaibrahimoglu MSc. Organizational and management control
Faculty of Economics and Business University of Groningen
9 March 2017
2 1. Introduction
The aim of this study is to examine the relation between the cost of equity and CEO narcissism. With the cost of equity being defined as the return a firm theoretically pays to its shareholders, to
compensate for the risk they undertake by investing their capital. And CEO narcissism being the degree of narcissism that a CEO has in his personality. The measurement of CEO narcissism is done with help of unobtrusive measures of narcissism that was pioneered by Chatterjee & Hambrick (2007). And the approximation of the cost of equity will be done by looking at the beta and the leverage ratio. Which are two risk variables that a CEO can have an effect on and they both could influence the risk value that a shareholder would attach to a stock. Thus affecting the theoretical return that a shareholder would have to expect in the face of more risk. With consequently, the cost of equity going up.
The reason for addressing the relation between CEO narcissism and the cost of equity is because as of now, there are more studies that used the unobtrusive measures of narcissism made by
Chatterjee & Hambrick (2007), to examine the effect of CEO narcissism on financial measures or other outcomes (Rijsenbilt & Commandeur, 2013; Olsen et al., 204). But seeing the results of two of these paper together leaves a gap that may be worthwhile to address. To go into specifics, Olsen et al. (2014) found that CEO narcissism had a positive effect on the earnings per share and share price of a company. Furthermore, they looked at whether this was due to operational decisions or accrual and accounting decisions, such as buying shares back to increase both the share price and EPS. They found that narcissistic CEOs tended to perform operational decisions to increase these measures.
And Chatterjee & Hambrick (2007) found that CEO narcissism led to more extreme firm
performance. And that leads to ambiguity, and thus a gap that warrants clarification. On the one hand, investors accept that risk has to be taken to generate value, on the other hand, taking too
u h isk fo ot e ough pa off e o es p o le ati . Said diffe e tl , Chatte jee & Ha i k s (2007) finding that CEO narcissism leads to volatile performance and Olsen et al. (2014) their finding that CEO narcissism leads to higher earnings could actually be explained. Quite simply too, as taking on more risk will lead to potentially higher earnings than there would be with less risky endeavours.
But, these two findings cannot help in examining whether the amount of risk taken leads to sufficient extra value.
This paper tries to aid in this by looking at the leverage ratio and the beta, which captures risk for shareholders and tries to relate that to CEO narcissism. This will be done with help of the agency theory and efficient contracting theory which are explained in the theoretical framework (Jensen &
Meckling, 1976; Subramanian, 2003). With the agency theory describing a principal-agent
relationship. In which the principal, hires the agent to perform a task on behalf of the principal. In this case that means the shareholders that hire the CEO to run a company in their behalf in exchange for compensation. But this relationship can have some problems due to information asymmetry and differing interests. These problems can lead to costs and those costs can be seen as risks to the shareholder. With the efficient contracting theory being a response to those agency costs and showing some strategies to deal with this (Subramanian, 2003).
After the theoretical framework there will be an elaboration of narcissism and consequently its effect on CEOs that are narcissistic. What the previous papers on narcissism (Chatterjee & Hambrick, 2007; Rijsenbilt & Commandeur, 2013 & Olsen et al., 2014) have in common is that they argue that narcissistic people will be more willing to take risky decisions. It is argued that this comes to be due to the fact that narcissistic people have an inflated self and find themselves superior to others, but simultaneously feel insecure. Consequently, they have a never ending need of having their own sense of self affirmed by others (Buss and Chiodo, 1991). That can lead to narcissists doing attention grabbing or other risky things which could eventually lead to the potential desired praise and self-
3 affirmation (Wallace and Baumeister, 2002). Supporting evidence was found by Campbell et al.
(2004), as he found that narcissists are overconfident people that tend to value their own abilities more than others, even though there may not be an actual basis for that. And thus are more willing and confident in risky decision making and the odds of those relevant payoffs. These manifestations of CEO narcissism will be examined with the theoretical framework to show how it can increase risk for shareholders from an organizational process perspective. After that, the two measures that capture risk for shareholders will be connected to CEO narcissism by drawing from corporate finance literature (Hackbarth, 2008; Ben-David). With the two hypotheses being that CEO narcissism is positively related to the leverage ratio and is also positively related to the beta.
The results of the paper show that there is indeed a positive and statistically significant relation between the financial leverage ratio and CEO narcissism. With the caveat that this is the case when a CEO is succeeded by a more narcissistic one. There was no significant relation between the leverage ratio and CEO narcissism over the tenure of the same CEO. Regarding the relation between the beta and CEO narcissism, there was no evidence to back this hypothesis up. The test results were
insignificant.
So, this research will try to address the gap by looking at if volatile performance changes the share price. If the volatile performance is reflected in fluctuation of the share price, this would lead to a higher beta, which is the movement of a share price in comparison to a benchmark. Where the benchmark is the S&P 500 index in this study. If this would not lead to a higher beta, then the findings of Olsen at el. (2014) and Chatterjee & Hambrick (2007) would also be interesting, since that would mean that the riskiness of a stock is not necessarily higher, but the earnings per share and share price would still be higher. So in that case, one could say that a degree of narcissism in the personality of a CEO would not be a bad thing at all in the eyes of investors. The findings of this study indicates that the latter is more likely.
The rest of the paper will be organized as followed, in the following part there will be a theoretical framework, in which the agency theory and efficient contracting theory will be looked at. Followed by the literature review. In this literature review there will be an examination of the literature regarding narcissism. It starts with literature on narcissism itself, which is then related to CEOs. Then CEO narcissism will be connected to the theoretical framework and lastly, CEO narcissism will be used to generate hypotheses. Subsequently, there will be a discussion and explanation of the data sample followed by some descriptive statistics. After that, the results will be shown and analysed.
Then there will be a discussion and conclusion. And at the end there are some limitations to the paper and avenues for future research.
2.1 Theoretical Framework
Before explaining and relating narcissism to, lastly, the cost of equity, this section will be used to bring forward two relevant theories that can be used as a framework to examine the relation between narcissism and the cost of equity. The first one that will be looked at is the agency theory, followed by a section about efficient contracting theory. After that there will be a literature review on narcissism and consequently narcissism in CEO s ill be looked at. After that, CEO narcissism will be examined with the agency theory and the efficient contracting theory. Which provides an organizational process perspective that is used to justify the generated hypotheses on CEO narcissism with regard to the financial leverage ratio and the beta.
Agency theory
The agency theory describes a principal-agent relationship where one part, the principal, hires another party, the agent, to make decisions on behalf of the principal (Jensen & Meckling, 1976).
This theory is used to provide a theoretical premise with which organizational processes can be
4 examined from a principal-agent perspective. In this case the principal-agent relationship means the relation between shareholders (principal) and the CEO (agent).
The relevance for using the agency theory to understand the relation between narcissism and the cost of equity is because the agency theory holds two key assumptions that create friction in the principal-agent relationship. And this friction arises because two important assumptions of the agency theory are that there is asymmetric information and different interests between the principal and the agent. Leading to possible conflict, because the agent holds more information than the principal, as he is more involved in the day to day operations of the firm. Coupling this with different interests between both parties and situations could come to be where the agent may know of activities that are useful to the principal, but not to the benefit of the agent. An example of this would be where there is an action that could be taken that would be good for the long term future of a company, but would decrease financial measures in the short term that are directly linked to the compensation of the agent. Thus an agent may choose to never disclose this activity due to the conflict of interests. Another issue arises when certain things that an agent does would be costly for him if the principal would observe it and consequently chooses not to disclose it. An example of this would be that an agent chooses to take riskier options because of the higher potential payoff whilst not bearing the costs of those risks himself. This problem where the risk is borne by others is called moral hazard. All in all, it is important to note that the principal-agent relationship is impacted by the opportunistic and individualistic tendencies of both parties. Taking the fact that agents have
different interests, where these interests are presumed to at least partly be influenced by the age ts pe so alit , the the deg ee of a issis of a CEO e o es ele a t i this elatio ship.
Because narcissism leads to certain needs that have to be fulfilled which can lead to risky behaviour.
But the elaboration of this idea will be done after the literature review on narcissism itself.
For now, it is important to emphasize that there is reason to believe that personalities have an influence on the opportunistic and individualistic tendencies that one can have. Therefore,
so eo e s pe so ality will be a factor in the principal-agent relationship and the possible problems that could arise in such a relationship. This idea is backed by other literature that deals with
personality and aptitude to take risks. For example, Zuckerman & Kuhlman (2000) found that certain genes are inherited that will increase a pe so s tole a e to take isk. The e tio t o
components of this increased tolerance to risk. One of them is impulsivity, which is described as follo ed: the te de to e te i to situations, or rapidly respond to cues for potential rewards, without much planning or deliberation and without consideration of potential punishment or loss of
e a d Zu ke a & Kuhlman, 2000). And the other component they talk about is sensation seeking, explaining it as followed: a t ait defi ed the seeki g of a ied, o el, o ple a d intense sensations and experiences, and the willingness to take physical, social, legal and financial
isks fo the sake of su h e pe ie e Zu ke a , 99 . The go on to explain that high sensation seekers tend to attach a lower risk value to activities than low sensation seekers, even if they have no prior experience with the activity (Horvath & Zuckerman, 1993). Also, they experience less anxiety than low sensation seekers in unknown situations (Zuckerman, 1979). In their 2000 study (Zuckerman & Kuhlman) they found that impulsivity and sensation seeking had a significant influence on general risk taking. Where general risk taking was based on 6 risky behaviours (gambling, drugs, drinking, smoking, sex and driving).
Ultimately, the personalities of individual people can indeed have an influence on the individualistic and opportunistic tendencies that will play a role in the conflict that may emerge in the principal- agent relationship.
Efficient contracting theory
Contracting theory can be seen as a response to the problems that could appear between the principal and the agent, due to their different interests. And consequently the possible misuse of the asymmetric information that exists in their relationship. Now, contracting theory emphasizes that a
5 contract between the principal and agent can be seen as an incentive system (Hensher et al., 2015).
Or in other words, a contract can be used to incentivize the agent to behave in a certain way that is desirable to the principal. Ideally leading to a situation where cooperation leads to beneficial outcomes for both parties.
In classical contracting theory (Coase, 1937; Alchian & Demsetz, 1972 and Jensen & Meckling, 1976), a contract is described as the tool with which principals and agents can balance the sharing of risk and stipulate the effort incentives for the agent. They also assume that the agent will be risk averse and that the principal will be risk neutral. This is because a contract can be totally fixed and with that the pa off to a age t is al a s the sa e. O , a age t s pa off a e pa tl o totall dete i ed by (a) certain output(s). Some key assumptions to this situation are that the output of the firm is observable and can be contracted upon. Furthermore, it should be possible to indicate the output in monetary terms over a single period. Also, another assumption is that the agent is free to choose his own behaviour and the principal is not able to fully observe this behaviour. And lastly, there may be events outside of the control of the agent that can influence the measured output. Therefore both the agent and the principal hold risk if the payoff for the agent is not totally fixed.
In this classical theory it is therefore so that the principal sets the contract conditions in the hope that the age t ill put i o e effo t to eate o e e efits to the p i ipal. Whe e the age ts objective is to maximize his own payoff by choosing the optimal level of effort knowing the terms of the contract that are set by the principal. Even if these assumptions of risk preferences do not hold it is still important to not discount the fact that a contract is still useful to find an agreeable balance between sharing risk and providing incentives (Jensen & Meckling, 1976).
Finally, what is important about the efficient contracting theory for this paper is that there are costs associated with trying to find and creating the optimal contract for the principal and the agent (Subramanian, 2003). With costs being borne by the principal. These costs are called agency costs and consist of three components, namely: monitoring costs, bonding costs and residual costs. Where monitoring costs are incurred through monitoring the agent and spending on information systems or other oversight processes that can aid in reducing and checking for opportunistic behaviour. More specifically, monitoring can be done through external or internal audits. Where annual reports, for example, can be used to assess whether the agent is performing up to par and the other way around, an agent can showcase its capability and credibility to the principal. Reducing agency costs as both parties will experience less friction then (Anderson et al. 1993; DeFond, 1992). Another monitoring mechanism is budgeting, which is not only used to set performance targets but can also be used to restrict and monitor agent behaviour (Indjejikian, 1999).
Now, bonding costs are associated with finding ways to make the agents i te ests alig ith that of the principal, which, for example, could be finding a reward structure that would minimize the loss to the principal in the case of opportunistic behaviour (Subramanian, 2003). But they could also be efforts from the agent to smoothen cooperation in the principal-agent relationship (Jensen &
Meckling, 2006). By allowing for contractual restrictions or for example, accepting a public auditor for the financial accounts. And regarding residual costs, they occur as a result of the discrepancy between the principal and the agent their interests, despite monitoring and bonding processes.
2.2 Literature review
In this literature review there will first be a review on some literature about narcissism itself. Then a issis ill e elated to a issis i CEO s and its manifestations will be evaluated in the context of the theories described in the theoretical framework. To see what effect CEO narcissism can have on organizational processes that can explain how the principal has to possibly deal with
6 more risk. And after that CEO narcissism will be used to generate hypotheses about the financial leverage and beta, which are measuring risk that can affect the cost of equity.
Narcissism
The origin of the word narcissism is based on a Greek myth. The myth tells a story about Narcissus, a handsome and proud but emotionally detached Greek who refuses to love anyone. He is punished for this by the goddess of love, Aphrodite, and is forced to love himself. Consequently, Narcissus catches his own reflection in a pool and becomes enamoured with it. He then tries to grasp his own reflection, but by doing so he falls into the pool and drowns in his own image.
The word narcissism dates back to the end of the nineteenth century, as Ellis (1898) coined the term to describe a perverse self-love. It influenced the thinking of Freud who was instrumental in making narcissism a fruitful avenue for research. Also, Freud (1957) found some other characteristics of narcissism, such as self-admiration, self-agg a dize e t a d seei g othe s as a e te sio of o e s self. Following other literature that deals with narcissism (Raskin & Hall, 1979; Emmons, 1984 &
Campbell, 2004), the definition of narcissism is made with use of the Diagnostic and Statistical Ma ual of Me tal Diso de s DSM a d is said to e a pervasive pattern of grandiosity (in fantasy or behavior), need for admiration and a lack of empathy, beginning by early adulthood and present in a variety of o te ts .
What is important to note here is that till the 1980s narcissism was often merely seen as personality disorder, whereas nowadays academics see it as a personality characteristic that is present in everyone and can be measured (Emmons, 1987; Raskin & Terry, 1988). Only in extreme cases of narcissism is it seen and diagnosed as a disorder (Campbell, 2004). This difference in perception was made possible due to the creation of a psychometric scale in relation to narcissism by Raskin & Hall (1979). They made the narcissistic personality inventory (NPI): a 54 item forced choice survey based on the 220 item DSM II. This NPI was further probed into by Emmons leading to a conceptualization of narcissism into four components. (1) authority/leadership (I like to be the center of attention); (2) superiority/arrogance (I am better than others); (3) self-admiration (I am preoccupied with how extraordinary and special I am); (4) entitlement (I insist upon getting the respect that is due to me).
Furthermore, Emmons (1987) also underlined the fact that these four factors work as a solitary personal construct, as in, narcissism coherently relates all four of these factors to a degree, and not just a few or one of them. This has been supported by other studies (Raskin & Terry, 1988; Watson &
Biderman, 1993; Judge et. al, 2006), as they also show that narcissism is a coherent yet multifaceted personality characteristic that tells something about the degree to which an individual has an inflated sense of self and the willingness to act in ways thatwill reinforce that self-view.
On a cognitive level, narcissism leads to the belief that one has superior qualities, as narcissists rate themselves significantly higher than others on things like intelligence, competence, creativity and leadership skills (Judge et al., 2006). Yet they lack self-confidence and try to compensate that by doing things that will make them feel better than others. This lack of self-confidence was shown by Kernis & Sun, 1994) when they found that anger and aggression towards negative feedback was positively related to narcissism. Even though Emmons (1987) found NPI scores are negatively connected to the self and ideal-self score, meaning that narcissists see little room for improvement.
Their sense of self may be shining in their own eyes, but it sits on a shaky foundation. This paradox is the fuel behind the fact that narcissists are constantly craving admiration and can lead to a reckless pursuit of self-enhancing opportunities (Campbell et. al, 2000; Wallace & Baumeister, 2002).
CEO narcissism
Relating this to CEOs and the degree to which they are narcissistic, there are several implications.
One of them has to do with that every CEO is going to be faced with choices and narcissism can
7 affect the subjective probabilities that one attaches to various outcomes. And that is because the inflated sense of self will lead to over optimism and therefore bad outcomes will be seen as likely to happen and good outcomes will be more likely to happen (Shapira, 1995; Sanders,
2001). This effect is strengthened in the situation of bold or unconventional actions, as it aligns with the personal desires of narcissistic individuals. In other words, the outcomes and payoffs of decisions that may lead to attention are seen as a lot more favourable opposed to non-narcissistic people (Postman et al, 1948; Molden and Higgins, 2005).
This could lead to risky operational decisions, such as firm acquisitions, to make the financial measures look better, which is what Olsen et al. (2014) found. These tendencies are fuelled by the
a issist s eed to have themselves be praised by other and thus get their self-image cast in a positive light. Furthermore, when Rijsenbilt & Commandeur (2013) examined whether there was a relationship between narcissism and fraud, they found a positive one. And they also argued that that may be due to the fulfilment of narcissistic tendencies, resulting in them choosing risky decisions to get potential praise and attention. So all in all, the common thread that is to be found here is that narcissists believe themselves to be superior to others and have a constant need to have that idea of themselves acknowledged by others. Leading to less risk adversity than others would have and a clouding of judgement, in the sense that they could act on a certain risky decisions fully believing that they are able to propel themselves towards favourable outcomes. Even when common sense would advise caution or to refrain from choosing that option altogether.
CEO narcissism related to the agency theory and efficient contracting theory
And this phenomenon ties in with the studies done by Zuckerman & Kuhlman (1979, 1994, 2000) where he finds that personalities do indeed have an influence on the risk aptitude of people.
Zuckerman mentioned sensation seeking and impulsivity, where sensation seeking was defined as: a trait defined by the seeking of varied, novel, complex and intense sensations and experiences, and the willingness to take physical, social, legal and finan ial isks fo the sake of su h e pe ie e (Zuckerman, 1994). And impulsivity was described as: the te de to e te i to situatio s, o rapidly respond to cues for potential rewards, without much planning or deliberation and without consideration of pote tial pu ish e t o loss of e a d . Now, sensation seeking behaviour is explicitly mentioned by Chatterjee & Hambrick (2007) in their paper dealt with narcissism in relation to firm performance. And they used sensation seeking in the sense of that they are often activities that will catch the limelight and will therefore give narcissists the necessary attention and praise that they need. Even though they can be risky and could lead to extreme firm performance. Which, Chatterjee & Hambrick (2007) found evidence to. About impulsivity, there is a lack of literature that connects narcissism to impulsivity, but based on the mentioned literature, it may not be that narcissists do not do an internal cost/benefit analysis of their actions, but they attach different values to things opposed to non-narcissistic people. Leading to them taking riskier decisions than others, but that does not imply impulsivity, as in not thinking about their actions beforehand.
Nevertheless, this increased tolerance to risk and likeliness to pursue risky activities will have the principal hold more risk.
Also, going back to the fact that narcissists rate themselves better on a cognitive level than non- narcissistic people (Judge et al., 2006). This also has a consequence in the scope of the agency theory. As one of the assumptions of the agency theory is that there is information asymmetry. For this finding that is important because the principal will not be able to be fully able to judge the capabilities of the agent. This is called adverse selection, where the agent misjudges their capacity to perform and the principal is not able to tell at, for example, the moment of hiring or green lighting a certain risky project (Subramanian, 2003). This lack of information and insight also increases the risk that the principal has to bear.
8 Furthermore, as mentioned before, narcissists have different priorities/interests opposed to non- narcissistic people. This, as mentioned before, have to do with activities that can lead to them getting more attention (Postman et al, 1948; Molden and Higgins, 2005). These activities can be unconventional/bold and could be seen as risky. They can also be beneficial to the needs of the narcissist (the agent), but are not necessarily beneficial to the principal (shareholders). Leading to a moral hazard problem that makes the principal hold more risk due to a misalignment in interest, whereas the agent knows that they are not (fully) bearing the risk (Subramanian, 2003).
This also refutes the risk preference assumption that is made in the classical contracting theory (Coase, 1937; Alchian & Demsetz, 1972 and Jensen & Meckling, 1976), as that says that the agent is risk averse and the principal is risk neutral. And thus, the contract is used to make the agent exert more effort, which will be beneficial to the principal. But in this case, the agent may be risk seeking, necessitating a shift in focus for the principal in setting up a contract that provides the right balance between incentivizing effort and sharing risk. Instead of only putting in incentive schemes to do something more, a principal may want to consider including covenants, helping to deter an agent from taking too much risk. Knowing that the agent may be more risk seeking than the principal can also lead to higher agency costs, through monitoring and bonding costs (Subramanian, 2003). Since there is more need to monitor the agent, knowing that the agent is more risk willing than the principal. And including the right incentive schemes in a contract can be difficult, but including covenants in a contract, which are punishments for doing something unwanted, can be even more difficult to have both parties agree upon. As covenants can also signal distrust to the earnestness of the agent, increasing friction and thus agency costs.
All in all, CEO narcissism can increase agency costs, and that could be reflected in the cost of equity.
As the cost of equity is the theoretical return on equity that shareholders rationally expect. This will go up if shareholders assume that there are more risks involved in holding the shares of a company that is run by a narcissistic CEO. Higher expected costs should lead to a bigger return demand and following that, the cost of equity should be increasing. This is related to two variables, namely the leverage ratio and the beta, that capture risk for shareholders.
CEO narcissism related to cost of equity through financial leverage and beta
The consequences of CEO narcissism may have an influence on the financial leverage. This can be explained by drawing from corporate finance literature, as Hackbarth (2008) looked at the capital structure decisions of overconfident and/or optimistic managers. He starts with the Idea that optimistic managers will overestimate the growth rate of their earnings, terming that as growth perception bias. And overconfident managers will underestimate the riskiness of their earnings (risk perception bias). What this leads is that managers with growth perception bias believe their firms to be more profitable than they are and are less likely to approach financial distress. For managers with risk perception bias it means that they see their firms as less risky than they are, which also makes financial distress. Therefore, they are more willing and more tolerant to higher debt levels than managers that do not have these biases. Other studies have come to the same conclusion,
managerial overconfidence leads to higher levels of debt (Barros & Silveira, 2008; Fairchild, 2009 &
Ben-David et al., 2013).
Debt leverage influences the cost of equity because debt payment takes precedence to equity gains.
Of course, there is an upside to financial leverage too, as long as the debt that is taken will generate more value than it costs. But it also increases the downside risk in the event that the generated return will be lower than the cost of debt. Leading to more risk for shareholders if they hold the shares of strongly leveraged firms. Because there is more need for monitoring by shareholders to ensure themselves of that the extra of cost debt will be outweighed by the increase in return on assets. Which is what validates the extra financial leverage to equity holders.
9 The personal biases that are described in the studies on managerial confidence overlap with the manifestations of narcissism, even though narcissism is distinct to managerial overconfidence.
Because overconfidence describes an individual that overestimates the probability of desirable outcomes, based on the skills, experiences and successes one has had. In a decision making context that is. With narcissism this extends to every aspect of their life. Nevertheless, based on the
literature on narcissism and its findings, one could find enough common ground to exchange managerial overconfidence for narcissism in this context. Hence, CEO narcissism will lead to higher debt levels, as their confidence in themselves will make them discount the risk attached to their actions. Also, Campbell & Green (2008) found that narcissists are more motivated by reward than by punishment. And that could lead to or strengthen growth perception bias. Which will also make financial distress risk seem less likely.
H1: CEO narcissism is positively related to financial leverage
And this is relevant to the main question: does CEO narcissism increase the cost of equity? Because, the more leveraged a company, the more risk an investor will bear. This is due to the fact that using more debt makes returns for equity more volatile. Since using a higher ratio of debt to equity makes fluctuations in return have a stronger effect on equity. Since debt is guaranteed a fixed return and its payment takes precedence over equity. For example, if an investment of 1000 euros is fully financed by equity then a 10% decrease or increase falls solely on the equity holder. If one uses 50% debt and 50% equity, a 10% increase or decrease in the overall investment actually means a 20% (1100- 500/500=1.2) increase or 20% (900-500/500=0.8) decrease for equity holders. This effect is happening because only debt has a fixed claim on the assets of a company. Any extra value that is generated that is beyond the cost of debt will become value for equity holders. Naturally, a more strongly leveraged company will have to generate a higher rate of return in the eyes of investors, and thus will have a higher cost of equity. With the cost of equity entailing the theoretically required rate of return one should return to its equity investors to compensate for the incurred risk of
investing their capital. Since the cost of debt is bigger, the returns on investments need to be bigger for equity holders to gain value out of an investment.
Now, moving on by regarding the cost of equity in relation to the beta. This could be affected since the cost of equity is normally calculated with this formula: 𝑅 + 𝐵 𝑡𝑎 𝑅𝑚 − 𝑅 . And the cost of equity is seen as the theoretical return a firm should pay to its shareholders, as was mentioned before. Since investors want to be compensated for the fact that their money is unavailable for a period of time and also want compensation for bearing risk in making the investment. The agency theory and efficient contracting theory are relevant here, since the associated agency costs would create the need for a higher theoretical return. As more agency costs means that the principal (shareholders) would have to bear a higher risk.
Going back to the formula, Rf = the risk free rate, and that would be the interest one would attain on totally risk free investments. But that is impossible in practice, so the proxy that economists use for the risk free rate is an investment which is normally not subject to strong changes in value, so for example, short dated government bonds. Rm stands for the average market rate of return, where an often used proxy is the historical return of an index for, such as the S&P 500, the index that is also used here. And the most important part of the equation in this context is the beta. As the beta is used to measure the volatility of a share in comparison to the market as a whole. It is used to estimate the riskiness of a stock, as it tells to what extent and which direction a stock moves whilst taking the market into account. So, a beta between 0 and 1 would mean that the stock fluctuates less than the market as a whole, 1 means that it tracks the market perfectly and more than 1 would mean that stock is more volatile than the market.
10 This relates to CEO narcissism because Chatterjee & Hambrick (2007) conducted a study that
attempted to measure CEO narcissism and hypothesized that company performance would be more volatile. They connected characteristics of narcissism, such as the need for attention and sense of supe io it to a tio s that ould e high isk high e a d. O i thei o ds: it is likel fo narcissistic CEOs to engage in certain types of strategic actions: bold, quantum, highly visible initiatives, rather than incremental elaborations on the status quo. Given this, narcissistic CEOs will te d to deli e e t e e a d flu tuati g pe fo a e fo thei o ga izatio s. Fo e a ple, the used the superior sense of self and need for attention to argue that firm acquisition could be related to CEO narcissism. Since an acquisition is a high profile event and there is a belief that one could do better than the incumbent managers.
Since, as mentioned before, the beta is a number that one assigns to signify the degree to which the stock is subject to market risk, it should increase with the degree of CEO narcissism that is involved.
Since Chatterjee & Hambrick (2007) found that narcissistic CEOs tend to have more volatile performance, thus their beta, the sensitivity to market risk should be higher compared to non- narcissistic CEOs.
H2: CEO narcissism is positively related to the strength of beta
Through the beta and financial leverage it should be the case that the cost of equity of companies with narcissistic CEOs should be higher than the companies of non-narcissistic CEOs. Which makes it so that shareholders have higher return expectations of such firms. Otherwise, they are not willing to hold the shares in their portfolio. Furthermore, if a firm becomes so leveraged that it faces distress risk, the cost of debt will also increase, although that is out of the scope of this paper.
3. Methodology
Sample construction and data collection
The sample selection process was started by selecting all CEOs of S&P 500 companies available on ExecuComp database. Our initial sample consists of 4,755 CEO-years observations from 920 CEOs and 500 firms over the 2005-2014 period. I exclude 867 CEO-years observations from financial firms (SIC 6000-6799) and 72 CEO-years observations from firms with ambiguity regarding CEO
information. I obtained financial data from Compustat, the Morningstar database, the Google Finance database, the AmigoBulls databse and the Yahoo! Finance database. During this process, I eliminate 1459 observations with missing firm-specific variables. Our final sample consists of 2288 firm-year observations from 447 CEOs and 369 firms over the 2005-2014 period.
The S&P 500 is the index that tracks 500 large, publicly traded US companies. Due to their size, in the sense of high market capitalization, they can be seen as high profile firms. Consequently, they give opportunity for narcissistic needs such as attention and self-affirmation to be fulfilled. This index is also used by Rijsenbilt & Commandeur (2013), whereas Chatterjee & Hambrick (2007) focused on software and hardware companies and Olsen et. al (2014) decided to use Fortune 500 companies.
The CEOs that will be examined are the ones that lead a company in the S&P 500. With the filter that the CEO needs to have at least four years of tenure and this tenure count starts from the year 2005.
(Chatterjee & Hambrick, 2007; Rijsenbilt & Commandeur, 2013 & Olsen et al., 2014). The reason for the start after 2005 is because the thesis group data collection had the time frame of 2005 till 2014.
Regarding the four years of tenure, the first year after the appointment of a new CEO can be quite chaotic and was therefore regarded as a transitional year (Chatterjee & Hambrick, 2007; Rijsenbilt &
Commandeur, 2013 & Olsen et al., 2014). The next two years allow for the narcissistic tendencies of a CEO to shine through and year four (for the dependent variable, so the cost of equity) and after can be used to test the effect of narcissism. In other words, year two and three are used for defining
11 the narcissism measure and year four and after are used for the debt leverage and beta. This lagged design was done by the three papers (Chatterjee & Hambrick, 2007; Rijsenbilt & Commandeur, 2013
& Olsen et al., 2014) to solidify the fact that narcissism is a stable personality disposition. It is considered constant over the years and is measured prior to the measurement of company
dependent numbers. And thus eliminating any circular or recursive relationship between narcissism and the dependent variable.
What is required therefore is identifying CEOs with a tenure of at least four years, which is done by looking through the ExecuComp database to find these CEOs. Furthermore, annual reports of the second and third year of tenure have to be available, since the narcissistic measure is based on these two years. And these annual reports are found by looking at the respective company websites and if not found there they were attained by consulting the Morningstar database. After these
requirements are satisfied, a sample will have been made that can be used to measure narcissism and test its effect on the dependent variables.
Dependent variables
To find the degree of financial leverage, the most common method is used here, which entails that the total amount of debt will be divided by the total amount of equity. This data is provided by the Morningstar website/database by looking at the key ratios section. It will only be calculated if all required data for the narcissism score is available and this calculation will start for year four and later. The same holds for the calculation of beta. Whether there is a relation between the leverage ratio and composite narcissism score is tested through panel data regression analysis clustered by CEOs, since the narcissism score is used to check whether it has any influence on the leverage ratio, the dependent variable. And the reason to use panel data regression is because the data sample is dealing with measurements over time and therefore it is a panel data sample that needs to be analysed as such. Also, fixed effects will be used over random effects. Where the random effects assumption (made in a random effects model) is that the individual specific effects are uncorrelated with the independent variables. The fixed effect assumption is that the individual specific effect is correlated with the independent variables. In other words, these assumptions basically entail that there must be no covariance between any independent variable included in a model and the unobserved variable that captures all that is not being explained by the model. In the case of using random effects that is. This assumption is very likely to not be fulfilled by this model, therefore the safer option here is to avoid using random effects and go for fixed effects. To ensure that this was the right choice a Hausman test was done and gave a probability of 0.006. That outcome backs the choice to go for fixed effects. Because the null hypothesis in the Hausman test is that the use of random effects is appropriate, but since it is rejected, fixed effects will be used.
Regarding the cost of equity (𝑅 + 𝐵 𝑡𝑎 𝑅𝑚 − 𝑅 ), the risk free rate can be based on a treasury bill with a maturity of three months and the treasury bill rates are made publicly available by the U.S. department of the treasury. The market rate of return is based on the yearly historical
performance of the S&P 500 and by using that info to get to an average rate of return of the market.
But these two things are not necessarily the most important things, since they are out of reach for CEOs of a single company. The focus in this paper is more on two components that have an influence on the height of the cost of equity. As the financial leverage ratio should normally make
performance more volatile and thus increase the beta. And the beta itself is included in the formula and will directly increase the cost of equity if it is higher. To get a number on the beta, the volatility of a stock compared to the market, I will attempt to capture that by using the closing price of the stock over a relevant period of time and pitting that against the closing price of the S&P 500 index.
The closing price data on the stocks of the S&P 500 companies and the index itself are publicly available on the Yahoo! Finance database. In the moments where these stock prices were not available on the database I resorted to the Google Finance database. And finally, if that that was
12 necessary to make the calculation was not available then the company got omitted. This calculation will be done by first defining the change in the closing price of a stock in percentages. And then doing the same for the index. Having both those percentages, Excel can be used to calculate the covariance between the two stocks and then divide it with the variance of the index itself. Which captures how the stock and index move together relative to how the index moves on its own. The result is the beta of a stock. Then the beta of the stock and the composite narcissism score will be analysed, with beta as the dependent variable, through panel data regression analysis with fixed effects. To find out whether there is a relation between the two and if it is positive and significant. If so, then the degree of narcissism increases the beta of a stock.
CEO control variables that could have an influence on risk taking are CEO age, CEO duality, CEO tenure. Regarding CEO age, Rolison et al. (2014) found in their study that risk taking in the financial domain decreases for men as they age, but not for women. In this context, with a data sample that is 97% male (429 out of 444 CEOs are male in this sample), the age of a CEO can be an interesting control variable. This data on CEO age was provided by the ExecuComp database. For CEO duality the reasoning is that if the CEO is also the chairman of the board, then they will have more power and thus are more able to push their own ideas into fruition than CEOs that are not in this position.
Jensen (1993) argues that agency problems could arise in these kind of situations. To include this in the data, I looked at the titles that CEOs had, provided by the ExecuComp database and if they had the title of hai a I put a i thei espe ti e o fo the du a ia le of CEO dualit . Looking at CEO tenure, this one could be interesting since Miller (1991) found that CEOs with longer tenure would be less likely to tune their strategy to the environment in uncertain situations, which is what risky decisions are. To include the CEO tenure length I looked at the date at which a CEO started in its function and then counted from 2005 (or later) and on.
Firm controls for the beta could be the firm size and firm size is given by the total assets of a company in billions. These numbers were taken from the Amigo Bulls database as they provided a clear overview of the balance sheet numbers over the years for every company. This is a useful control variable since Binder (1992) found that beta is negatively related to firm size. In relation to leverage, firm size is often used in corporate finance literature as a control variable when leverage is dealt with, yet the relation between the two seems to be complex and rarely one dimensional.
Nevertheless it could be an important factor here too, so it is included for the debt equity ratio as a control variable. Importantly, the debt equity ratio affects the beta, so that should be included as a control variable as well. Moreover, another common firm control variable has been the firm age when the beta is studied. As Berger & Udell (1990) and James & Wier (1990) hypothesized that the fi s age and corresponding longer financial track record may decrease investor uncertainty coupled with a better market position, resulting possibly into lower systematic risk. Therefore the fi s age a de ease the eta. Rega di g the le e age atio a d fi age, the e see s to e o plausible link and there are no arguments found in the literature available. So, the firm age is a control variable that will only be used for the beta.
Another control variable could be the industry type, since different industries could have different levels of competitions and for example, with higher competition one could argue that more risk has to be taken in an attempt to be better than the other competitors. So that could possibly lead to an increase in the beta. Or an industry as a whole could have more growth potential opposed to other industries. Requiring more financing and could thus increase the debt equity ratio. This distinction between industries on two digit SIC codes. And they were provided in the ExcecuComp database, although they were four digits there and therefore I omitted the last two digits to prevent the industries to become too fragmented. Which is in line with what Olsen at al. (2014) did.
13 The last control variable that is considered here is the year variable, with each year having a dummy variable. This variable is included since the economic crisis is in this data sample period, therefore it would be wise to assume that this could have an effect on the two dependent variables that could say something about the riskiness of a company. This effect is obviously very multifaceted of itself and especially now in the case of a data sample that includes 370 firms. But one could argue that the global economic landscape in which most of these companies work would probably be changed in a way. Therefore there may be reason to think that the debt equity ratio and beta could be affected during or in the aftermath of the crisis.
Therefore the models become: De t e uit atio = β + β Na issis s o e+ β CEO age+ β CEO dualit + β CEO te u e+ β Fi size+ β6 I dust + β7 Year + ϵ
Beta = β + β Narcissism score+ β CEO age+ β CEO dualit + β CEO tenure+ β Fi size+ β Fi age + β7 Debt equity ratio + β8 Industry+ β9 Year + ϵ
Measuring narcissism
For measuring narcissism, a commonly used measure is the Narcissistic Personality Inventory (NPI) (Raskin & Terry, 1979, 1988) and the used approach is an attempt to mimic this test through the use of proxies. A necessity, because CEOs of major companies are unsurprisingly not all that eager to fill out the NPI through a survey. Therefore Chatterjee & Hambrick (2007) decided to use an approach that was based on unobtrusive indicators of narcissistic tendencies of CEOs. They had two main
ite ia, the fi st o e as that ea h i di ato had to efle t the CEO s ill. As in, to make sure that the i di ato as sho i g the CEO s pe so alit , the CEO had to ha e o t ol o e that i di ato . Their second criteria was that every included indicator had to be connected to one or multiple aspects of the narcissism personality characteristic. In identifying these indicators they were looking for manifestations that had resemblance to the four features of narcissism that Emmons (1987) through a factor analysis. These four factors were: superiority/arrogance,
exploitativeness/entitlement, self-absorption/self-admiration and leadership/authority.
The CEO s a issis is easu ed ith a o posite easu e o p isi g of the CEO s elati e ash pay to the second-highest paid e e uti e, the CEO s elati e o -cash pay to the second-highest paid e e uti e a d the size a d o positio of CEO s pi tu e i the a ual epo t. What is i po ta t to note here is that these three indicators are distinctive to narcissism and have nothing to do with overconfidence, which is a trait that is implicitly used here to explain the hypothesized outcomes.
Since narcissistic people want to feel superior and important in comparison to others and possess the need to have that self-view reinforced. But that does not apply to overconfidence (Campbell et al., 2004). These three indicators show narcissism, but do not show overconfidence in non-
narcissistic people. The cash pay, non-cash pay and statistics are found in the ExecuComp database and the CEO picture statistic will be collected by going through annual reports. The data will be taken from tenure year two and three and then averaged. The cash and non-cash pay reflect the fact that CEOs have considerable influence over their compensation packages and that of other
executives (Bebchuck & Fried, 2006) and can thus use it as display of power and importance. It is calculated by taking the cash compensation of the CEO in the Execucomp database and dividing it with the cash compensation of the second highest paid executive. Regarding the relative non cash payment, it is calculated by taking the total compensation (TDC1 in ExecuComp) of the CEO and then subtracting the cash compensation. Subsequently the same is done for the second highest paid executive and lastly, the CEO non-cash payment is divided with the non-cash payment of the second highest paid executive.
The same idea goes for the presence of the CEOs picture in an annual report, as the annual report provides a chance to report on the company and its progress, but also as an opportunity to present
14 oneself at the helm of the company. Even though CEO photographs are common to see in annual reports, they are not always presents and also, the size and thus prominence of pictures vary too.
Therefore the inclusion of a picture in an annual report can be seen as taking a chance to bask in the spotlight and making clear that they are the most important person in the company. Furthermore, following in the footsteps of Olsen et al. (2014), the prominence of the pictures will be categorized and will have a score assigned to them. The categories are:
(1) No photograph of the CEO;
(2) The CEO was photographed with other executives;
CEO s photog aph as of hi o he alo e a d o upied less tha half the page;
CEO s photog aph as of hi o he alo e a d o upied more than half of the page with text taking up some space on the page; and
CEO s photog aph as of hi o he alo e a d o upied the hole page.
Where option 1 gives 1 point and option 5 gives 5 points. Together with the cash and non-cash pay score this will be used to make a composite narcissism score. Descriptive statistics will be provided that shows the amount of observations, means, standard deviation, minimum and maximum. And correlation between the three indicators will be shown.
Furthermore, a factor analysis was conducted to confirm that the three components are capturing the same construct. It showed that the three components loaded on a single factor. And the factor weightings of the three components on that factor were used to make a composite narcissism score.
With the photograph points having a factor weight of 0.263 relative cash pay has a factor weight of 0.446 and relative non cash pay having a weight of 0.445.
Descriptive statistics
Panel A: descriptive statistics
n Mean Standard
deviation
Minimum Maximum
Photograph points
892 2.26 1.12 1 5
Relative cash pay
892 1.74 0.82 0 10.65
Non relative cash pay
892 2.46 1.74 1 26.44
Narcissism score (normal average)
892 6.46 2.48 1 16.27
Narcissism score (factor average)
892 2.47 1 0.26 11.22
Panel A shows the descriptive statistics on the three components that are used to make the
narcissism score, namely: photograph points, relative cash pay and non-relative cash pay. The mean for the photograph points is 2.26, with a standard deviation of 1.12. This standard deviation and mean indicate that most photographs included in the annual reports are a small one of the CEO himself, a picture with other executives or there is no photo included at all. Confirmed by panel C, which provides a distribution of the photograph points and its five categories. And that distribution shows that only 8% of all photographs were of only the CEO and bigger than half the page or filling the whole page. In regard to relative cash pay, the mean is 1.74, with a standard deviation of 0.82, a minimum of 0 and a maximum of 10.65. This minimum of 0 was from a CEO that did not have
15 compensation in the form of a salary, but only had compensation in the form of for example, share options, bonuses or perquisites. For non relative cash pay the mean was 2.46, the standard deviation 1.74, the minimum 1 and the maximum 26.44. This shows that non relative cash pay differences between the CEO and the second highest paid executive is more substantial than the difference in cash pay. Which is in line with the survey that Frydman & Jenter (2010) did on CEO compensation. As they examine why CEO compensation rose substantially in the last few decades. In their study they noted that this increase was not coming from an increase in salary, as salary had a decreasing share in total CEO compensation. The increase had more to do with CEOs having more pay in the form of options and stocks. The narcissism score (factor average), which is used for the regression analysis, has a mean of 2.47, with a standard deviation of 1, a minimum of 0.26 and a maximum of 11.22.
Panel B: correlations
Photograph points Relative cash pay Non relative cash pay Photograph points 1.00
Relative cash pay 0.10*** 1.00
Non relative cash pay 0.09*** 0.20*** 1.00
*** Correlation is significant at the 0.01 level (two tailed).
Panel B shows the correlation between the three components that make up the narcissism score. It shows that all three components are significantly correlated to each other at the 0.01 level. So, a higher photograph point score leads to a higher relative cash and non-cash pay. And a higher relative cash pay is associated with an increase in relative cash pay and vice versa. Noteworthy, since
Chatterjee & Hambrick (2007) decided to include the unobtrusive measure of photographs in the annual reports after consulting two corporate communication executives and an external
communications consultant. And they indicated that CEOs are very involved in the content and design of these reports. Furthermore, they also have strong opinions and control over how they themselves are portrayed in the annual report. So, these CEOs that are using their influence to show themselves (more prominently) also manage to get themselves more compensation than their counterparts that do not use their influence like this. This can be seen as an instance where powerful executives can use that power to influence their own pay packages. And that is in line with the managerial power theory of Bebchuk & Fried (2004) in which they argue that due to weak corporate governance and acquiescent boards, there will be an unequal power distribution between the board and the CEO. Therefore, CEOs have enough power to be able to determine their own pay (at least somewhat).
Panel C: distribution of photograph size
Photograph points Frequency Percentage of CEOs
1 308 35%
2 158 18%
3 352 39%
4 30 3%
5 44 5%
Total 892
Panel D: year frequency distribution
Fiscal year Frequency Percentage
2005 128 5.59%
2006 160 6.99%
2007 192 8.39%
2008 238 10.4%
2009 247 10.8%
16
2010 267 11.67%
2011 267 11.67%
2012 264 11.54%
2013 267 11.67%
2014 258 11.28%
Total 2288 100%
Panel D shows the year frequency distribution. It shows that there is an increase in observations from 2008 and on. Which can be explained by the filtering process that was used by this and other papers that measured narcissism through unobtrusive measures (Chatterjee & Hambrick, 2007;
Rijsenbilt & Commandeur, 2013 & Olsen et al., 2014). In this process the first year of tenure was seen as a transitional year and the second and third year of tenure were used for the components that make up the narcissism score. After that, only the third years of tenure and after were left in the data sample. So, a data sample that starts from 2005 will only include CEOs in the year 2005 if they started in 2002 or earlier. And so, a spike in observations naturally occurs in 2008.
Panel E: Industry dummy variable descriptive statistics
Code number SIC code description Firms Percentage
of sample
10 Metal mining 3 0.82%
12 Coal mining 1 0.27%
13 Oil and gas extraction 21 5.72%
14 Mining and quarrying of non-metallic minerals (no fuels) 2 0.54%
15 General building contractors 3 0.82%
16 Heavy construction, except building 2 0.54%
17 Construction and electrical work 1 0.27%
20 Food and kindred products 21 5.72%
21 Tobacco products 2 0.54%
22 Textile mill products 2 0.54%
23 Apparel and other textile products 5 1.36%
24 Wood related products 2 0.54%
25 Furniture and fixtures 2 0.54%
26 Paper and allied products 2 0.54%
28 Chemicals and allied products 34 9.26%
29 Petroleum and coal products 6 1.63%
30 Rubber and misc. plastic products 2 0.54%
31 Leather products 1 0.27%
33 Primary metal industries 1 0.27%
34 Fabricated metal products 4 1.09%
35 Industrial machinery and equipment 23 6.27%
36 Electronic and other electronic equipment 22 5.99%
37 Transportation equipment 11 3.00%
38 Instruments and related products 24 6.54%
39 Mics. manufacturing industries 3 0.82%
40 Railroad transportation 4 1.09%
42 Trucking and warehousing 2 0.54%
44 Water transportation 2 0.54%
45 Transportation by air 5 1.36%
47 Transportation services 3 0.82%
48 Communication 15 4.09%
17
49 Electric, gas and sanitary services 36 9.81%
50 Wholesale trade – durable goods 5 1.36%
51 Wholesale trade – nondurable goods 4 1.09%
52 Building materials, hardware, garden supply and mobile 4 1.09%
53 General merchandise stores 7 1.91%
54 Food stores 2 0.54%
55 Automotive dealers and service stations 6 1.63%
56 Apparel and accessory stores 5 1.36%
57 Furniture and home furnishings stores 3 0.82%
58 Eating and drinking places 4 1.09%
59 Miscellaneous retail 7 1.91%
70 Hotels 2 0.54%
72 Personal services 1 0.27%
73 Business services 37 10.08%
75 Auto repair services and parking 1 0.27%
78 Motion picture and video rental services 1 0.27%
79 Entertainment services 1 0.27%
80 Health services 6 1.63%
87 Engineering services 2 0.54%
99 Non-operating establishments 2 0.54%
Total firms 367
Panel F: descriptive statistics other variables
Variable Observations Mean Standard deviation
Min Max
Beta 2278 1.06 .40 -1.28 6.68
Debt equity ratio
2286 1.68 12.50 -313.5 174.44
CEO tenure 2288 9.52 6.33 4 52
CEO duality 2288 .67 .47 0 1
CEO age 2288 61.88 6.64 31 85
Firm age (logarithm)
2288 3.86 0.87 .69 7.01
Total assets in billions (logarithm)
2288 2.39 1.22 -1.8 6.68
Panel F shows the descriptive statistics of the dependent and control variables. Something that may need clarification here is the fact that the minimum and maximum of the debt equity ratio are so extreme. This is because the debt equity ratio is divided by debt / equity, which makes an equity value of below 1 or -1 have a very high debt equity ratio. There were 36 instances in the data set where the debt equity was below -20 or above 20. Which is 1.6% of the total amount of observations on the debt equity ratio that were used for the regression analysis. Also, the reason to use
logarithms for firm age and firm size is because it is not expected that these two control variables have a linear relationship to the beta and the leverage ratio, which are used to capture risk. More specifically, a company age difference of 20 years matters more if one company is a start-up and the other one is 20 years old. Compared to a company that is 50 years old and one that is 70 years old.
The risk difference between the first two companies is assumed to be higher here than the latter two companies. Which makes a logarithm of the control variable suitable, since it condenses the values. The same argument goes for the firm size.
18 4. Results and analysis
Panel G: CEO narcissism and financial leverage: panel regression results with robust fixed effects clustered by companies
Variable Coefficient Robust std. error t-statistic p-value
Narcissism score .97** .49 1.97 .05
Total Assets (log) .01 .24 .07 .94
CEO age -.01 .04 -.18 .86
CEO tenure -.07 .05 -1.32 .19
CEO duality -.19 .19 -1.03 .30
Industry SIC digits****
- - - -
Years Included Included Included Included
Number of observations
2248 Companies 369 Within R² 0.0213
* Significant at the 0.10 level
** Significant at the 0.05 level
*** Significant at the 0.01 level
**** Omitted due to collinearity
H1: CEO narcissism is positively related to financial leverage
Based on the Hausman test, giving a probability of 0.006, the decision was made to use fixed effects for the panel data regression. Panel G shows the result of the panel data regression clustered by companies and indicates that the narcissism score is positive and significant in relation to the financial leverage ratio, which is estimated with the debt equity ratio (coefficient = 0.97; p = 0.05).
This tells that when there is a difference in narcissism scores in a single company, there will be an increase in financial leverage. With the data set at hand it means that when a more narcissistic CEO takes over, the company will become more leveraged. Because CEOs get assigned a single narcissism score. Therefore, the only fluctuation in narcissism scores are due to different people leading the company in the time frame of 2005 – 2014. Furthermore, it shows that the other control variables are not able to say something about the financial leverage. With the exception of the year dummies, as the years 2006, 2007, 2010 and 2013 were significant at the 0.10 level and had negative effects on the degree of financial leverage. Moreover, the industry dummy control variable was omitted in this fixed effects test due to collinearity. Which can be explained by the fact that a regression analysis with fixed effects uses an estimator which is time invariant. It is like running a regression analysis with a dummy variable that is grouped and does not change over time, which is the exact same as the industry dummy variable if the grouping is done by company. Making the industry dummy variable collinear to the fixed effects estimator.
Also, un-tabulated results on financial leverage and CEO narcissism with clustering on CEOs instead of companies do not give a significant result for the narcissism score in connection to the financial leverage (p = 0.40). This means that there is no support on the idea that more narcissistic CEOs will perpetually increase the leverage of the company they run during their tenure.
19 Finally, to answer the first hypothesis, it is indeed the case that CEO narcissism is positively related to financial leverage (p = 0.05). Hence, replacing a CEO with a more narcissistic one can indeed lead to a more leveraged firm. Although this effect occurs in the beginning of the tenure and does not last over the tenure of the succeeding CEO.
Panel H: CEO narcissism and beta: panel regression results with robust fixed effects clustered by companies
Variable Coefficient Robust std. error t-statistic p-value
Narcissism score -.01 .06 -.19 .85
Debt equity ratio .007* .004 1.76 .08
Total Assets (log) -.06*** .01 -4.05 .00
Firm Age (log) .009 .06 1.44 .40
CEO age .009 .006 1.44 .15
CEO tenure .003* .002 1.67 .10
CEO duality .03 .03 1.05 .29
Industry SIC digits****
- - - -
Years Included Included Included Included
Number of observations
2240 Companies 369 Within R² .0485
* Significant at the 0.10 level
** Significant at the 0.05 level
*** Significant at the 0.01 level
**** Omitted due to collinearity
H2: CEO narcissism is positively related to the strength of beta
Panel H shows the test results of the regression analysis with robust fixed effects on the relation between the narcissism score and the beta. It reports that the p value of the narcissism score is .85, making it statistically insignificant at any significance level. Furthermore, it tells that the financial leverage ratio has a marginal positive (coefficient = 0.007) and significant effect at the 0.10 level (p = 0.08). Confirming the idea that financial leverage can have an effect on the beta It also shows that firm size, measured with total assets, has a negative (coefficient = -0.06) and statistically significant effect on the beta at the 0.01 level (p=0.00). This is in line with what Binder (1992) found. As he connected the beta to the firm size in his study and argued that bigger firms are more efficient in production, making the firm value more certain and less subject to fluctuation. Culminating in a lower beta. What is also shown in the test is that CEO tenure has a marginal positive effect (coefficient = .003), which is statistically significant at the 0.10 level (p = 0.10). Making it
incompatible with the study of Miller (1991), where he argues that CEOs with more tenure are less likely to alter their strategy to uncertain situations in that process make risky decisions. Regarding the year dummy variable, the years 2007 and 2008 are statistically significant at the 0.01 level.
Unsurprising as those years were the start of the global crisis in which stock markets over the world were going through rough times, therefore strong fluctuations in the beta can be related to that.
Un-tabulated results on the relation between the narcissism score and the beta with robust fixed effects clustered by CEOs also does not give a statistically significant result either (p = 0.31).
Therefore the answer to the second hypothesis is that there is no evidence to conclude that there is a relation between CEO narcissism and the beta.
