CEO overconfidence, firm value and the economic
environment
L. M. Nijenhuis
July 2015
Master's Thesis Finance
University of Amsterdam
Educational institute: Graduate School of Business
CEO overconfidence, firm value and the economic environment
Author: Lotte Maria Nijenhuis
Administration number: 6181843
Master’s programme: Business Economic: Finance
Date: 07-07-2015
Abstract
This study examines the effect of CEO overconfidence on firm value, as
measured by Tobin’s Q. In addition, it is investigated if this effect alters when the
economic environment is included. The empirical analysis, based on all firm included in
the Execucomp database, over a period of 10 years, shows a significant effect solely for
high overconfident CEOs. This implies CEOs who hold valuable options, which are
more than 67% in the money. In addition, a positive relation between overconfident
behaviour and innovativeness, with respect to firm value was found. Finally, solely
companies in innovative industries can benefit from a recession. Being overconfident
helps them to exploit growth opportunities, which are prevalent in an innovative
environment. As such, they will be able to grasp opportunities when others are reluctant
to take risk. A panel data analysis was used to control for unobserved characteristics.
However endogeneity has to be taken into account as this might have influenced the
results.
Table of Contents
Abstract ... III
Table of contents ... IV
1. Introduction ... 6
2. Literature review ... 8
2.1 The theory of overconfidence ... 8
2.2 Overconfidence and Corporate Finance ... 10
2.3 Beneficial implications of CEO overconfident ... 11
2.4 Overconfidence in the context of recession ... 16
2.5 An example from the field: The overconfident US vs. a reserved Europe during the financial crisis of 2007... 17
3. Model, variables and data ... 20
3.1 Model ... 20 3.2 Variables ... 21 3.2.1 Firm value ... 21 3.2.2 Overconfidence ... 21 3.2.3 CEO turnover ... 22 3.2.4 Innovativeness ... 23 3.2.5 Innovative industries ... 23 3.2.6 Recession years ... 24 3.2.7 Control variables ... 24
3.3 Data and sample ... 24
4. Methodology ... 25 4.1 Regression specification ... 25 5. Results ... 28 5.1 Descriptive variables ... 28 5.2 Correlations ... 30 5.3 Regression results ... 31 5.4 Robustness tests ... 36
6. Conclusion and discussion ... 37
6.2 Implications ... 38
6.3 Limitations and recommendations for future research ... 39
Bibliography ... 41
1. Introduction
Behavioural finance tries to provide clarification about financial phenomena by allowing models in which some agents are not fully rational (Barberis & Thaler, 2003). It allows for personal attributes and behavioural biases to affect information-provision incentives. An example of irrational behaviour is overconfidence. Overconfidence is often defined as taking excessive risk through underestimating volatility. By deviating from the traditional framework, focus on the impact of overconfident CEOs has increased extensively over the last 10 years. More specifically, the way behavioural biases have influenced corporate finance related decisions. Malmendier and Tate (2008) wrote a number of articles concerning this topic. They showed evidence of distortions associated with overconfident behaviour. An example of this behaviour is overpaying during M&A activities. In this scenario, overconfident CEOs have the tendency to overestimating the benefits of synergies and underestimating the probability of failure, which can result in deteriorating decisions.
One of the first results that become evident when investigating beneficial implications of overconfident behaviour is innovativeness. Current literature suggests that overconfident CEOs are better innovators, which is often measured by R&D expenditures (Hirshleifer et al., 2011). This innovativeness stems from the tendency to overestimate expected cash flows or underestimate potential risk. So, although risky, by exploiting this behaviour at their advantage overconfident CEOs could be value enhancing. Their craving to investments in innovations could facilitate the discovery of new opportunities. Also, the willingness to adapt to new circumstances and rapid decision-making has found to be successful in the past (Goel & Thakor, 2008).
Building on these findings, mainly on the paper of Hirshleifer et al. (2011) and Malmendier and Tate (2005a), this study will focus on the effect of overconfident CEOs on the value of the firm. The existing literature teaches us that there are multiple and ambiguous effects of overconfident behaviour. However, few have investigated what the eventual implications on firm performance are. As such, this study will contribute to the existing literature by provide insights on this matter. Hence, the first central question this study addresses will be: What is the effect of an overconfident CEO on
firm value?
Secondly, building on the paper of Hirshleifer et al. (2011) and Srinivasan et al. (2005), this study contributes by introducing an additional variable: the economic climate. More specifically, the economic state of recession will be under examination. So far, no research was conducted on the effect of CEO overconfidence while embedded in recession. As such, comparing the first and the second central question of this study will provide insight on the effect of overconfident CEOs on firm value and to what extent the economic environment influences this effect. By focussing on recession, some light will be shed on the effect of overconfident CEOs, when others are reluctant to take risk. Since, this effect has not been investigated, providing support for a positive effect of overconfident behaviour during a recession could stimulate companies to invest, and take risk, when embedded in recession.
In this study, the level of innovativeness measures the amount of risk the CEO is willing to take. Former research has examined recession and R&D spends broadly and generally the results showed a decreasing trend (Rafferty & Funk, 2008) (National Science Foundation [NSF], 2010). Often, when recession arises, budget cuts on R&D expenditure are found inevitable (Brockhoff & Pearson, 1998). However, there is also evidence of a contradicting theorem. Morbey and Dugal (1992) show positive long term effects of R&D spend during recession. Further supporting R&D investments are Filippetti and Archibugi (2011) who state that, under certain conditions, innovations could add value during an unfavorable economic climate. Finally, Srinivasan et al. (2005) conducted a similar research exploring the implications of proactive marketing during a recession. They found, through a survey of 154 senior marketing executives, that some firms derive benefit from investing aggressively. These firms view recessions as opportunities to strengthen their business.
As we have recently experienced a global financial crisis this will be included into the research. The overall level of confidence has declined during 2007/ 2008 (Claessens et al., 2010); therefore it would be interesting to see in what manner this would affect overconfident CEOs and their level of innovativeness. As such, the second central question to be addressed in this paper is: To
what extent do companies with overconfident CEOs benefit from a recession?
The sample of this empirical research consists of all companies included in the Execucomp data. These are solely North American companies. Information will be gathered from 2003 until 2013. Heirshleifer et al. (2011) used a similar sample with data form 1993 until 2003 to measure the effect of overconfidence on innovativeness. In addition, Malmendier and Tate (2005a) examined the effect of CEO overconfidence on corporate finance related decision using a sample before 2005. Therefore this study also contributes to the existing literature by examining if their results still hold, or if implications of overconfident behaviour have changed over the last couple of years.
The remainder of this paper is structured as follows. This study begins with an overview of existing literature on the effect of CEO overconfidence related to corporate finance related decisions. In addition, were this irrational behaviour stems from will be investigated. The literature will follow exploring beneficial implications of CEO overconfidence and how their excessive risk taking can be detained. This will be supplemented with theory on overconfident behaviour and innovativeness while being embedded in recession. Finally, based on the literature review a model will be composed, which will be followed by the results, the discussion of the results, a conclusion and limitation and recommendation for future research.
2. Literature review
2.1 The theory of overconfidence
The literature review starts by exploring the basic concept of overconfidence. To know were this behavioural trait stems from is valuable when investigating the consequences of overconfident behaviour. Hence, the statistical explanation elaborated in this chapter provides clarification on the cause of overconfidence. Also, the psychological implications, supported by several studies, and earlier research are discussed.
Overconfidence is a widely known phenomenon. The definition of the term confident, given by the dictionary, shows various descriptions. These definitions are “having strong believe or full assurance”; “having no uncertainty about one’s own abilities, correctness, and successfulness”; “sure of oneself”. Overconfidence is a form of irrational behaviour and one of the building blocks of Behavioural Finance. Research on the effect of irrational behaviour on business decision-making started with Roll in 1986. After Roll, this particular form of study extended rapidly, especially the last decade. The theoretical perspective on overconfidence implies that people put too much weight on their personal signal, therefore overlooking the public signal regarding the value of information. As a result they take too much risk, by focusing on their individual signal while underestimating the related variance. The theorem used to show the intuitiveness of overconfidence is Bayes’ rule for normal distribution. Bayes’ rule provides the probabilistically correct way to update beliefs in the light of new information (Gelman et al., 1995). In the standard model p(s) are rational beliefs based on a signal. In order to update those beliefs people apply true probability distribution and use Bayes' rule to update new information with the following formula:
Prob[A|B] = Prob[B|A]*Prob[A]/ Prob[B]
where Prob [A] is called the prior, Prob [A|B] is called the posterior and Prob [B|A] is the likelihood, measuring the reliability of the indicator B. Using this proper Bayesian updating the posterior, the believe and therefore level of confidence about an event, is measured as follows:
Posterior E[A|B] = Prior * wprior + Signal * wsignal
where wprior + wsignal = 1. This formula shows the origin of overconfidence. The prior is information that is publicly known, whereas the signal can be seen as personal information. This information is exclusive to the person who has gathered it and therefore unknown to other parties. The weight someone puts on their private signal influences his or her believe and level of overconfidence. Moreover, the formula used to calculate wsignal, in accordance with Bayes’ rules theory, is as follows:
wsignal =
For clarification, this formula is explained with an example of a stock price movement. As the true value of the posterior, or movement of the stock, is uncertain the signal is based on . These are the variance of the stock and the variance of the error term, which is an element of the regression analysis to measure the stock movement. A large error term suggests a signal that is inaccurate and one that contains a lot of noise. The intuition behind the theory is that one should put less weight on the signal when the variance of the error term is larger, hence when uncertainty is higher. However, overconfident people underestimate this risk. They tend to put too much weight on their private signal.
Additionally to the statistical explanation of overconfidence, various evidence of overconfidence can be found in the field of psychology. For example, Buehler et al. (1994) found that people underestimate the time they need to finish a project. They tested three different hypotheses, which were all found supported. These hypotheses were based on the time people need to complete a task and the impact of plan-based scenarios relative to relevant past experiences. They found significant result for over-optimistic behaviour as people tend to use future scenarios to base their predictions on when estimating the time they needed to complete a task. However, when instructed to use past relevant experiences the overall level of overconfidence declined. As such, Buehler et al. (1994) concluded that diminishing the uncertainty and impact of expectations about the future, declines overconfidence.
Other research was done by Cooper et al. (1988) who showed that entrepreneurs overestimate their chance of success, while expecting businesses of others to experience less optimistic prospects. The reason for starting their study was that entrepreneurs often have to evaluate their chances for success. These assessments bear on the current position and possibilities of the company, but also on predictions about future prospects. As such, they make business related decisions based, in part, on expectations. Contrary to former research, which has shown that entrepreneurs are only cautiously optimistic, Cooper et al. (1988) found evidence of very favourable expected prospects. 81% assigns themselves a change of seven out of ten or more of success. In addition, even more excessive are the 33% seeing odds of success of ten out of ten. Also, they found that these numbers do not differ for well-prepared entrepreneurs relative to moderately prepared ones. This implies that the level of preparation does not influence the level of overconfidence. Furthermore, final evidence of overconfidence from the field of psychology is a paper written by Svenson (1981). During his research he asked people about their driving skills and found that 93% believed that their skills were above the median. Obviously this is statistically impossible.
Overconfidence originates from the field of psychology. Several studies have demonstrated this behaviour trait and its implications. As discussed, the last decade, the implications with respect to
1 /σε 2 / (1 /σA 2 +1 /σε 2 ) σA 2 +σε2
corporate finance have evolved. Therefore, current literature regarding this development is described in the next chapter.
2.2 Overconfidence and Corporate Finance
The literature review continues by investigating implications of CEO’s overconfidence with respect to corporate finance related decisions. Malmendier and Tate (2005a, 2008) wrote a number of articles addressing this topic. Their studies and methodology of measuring overconfident behaviour, are key to many studies on overconfidence that followed (Hirshleifer et al., 2010; Campbell et al., 2010).
Regarding the impact of CEO overconfidence on firm characteristics, the existing literature focuses on four main effects; merger and acquisitions, capital structure, investment and trading activity. First, research examining the effect of CEO overconfidence on mergers and acquisitions is extensive. Starting with the hubris hypothesis from Roll (1986), followed by research from, among others, Malmendier and Tate (2008), overconfidence has shown several negative effects. Some of these effects are overpaying by acquirers (Roll, 1986) and undertaking value-destroying mergers (Malmendier & Tate, 2008). As a result these mergers and acquisitions worsen shareholders wealth. One of the reasons why companies with overconfident CEOs engage in such mergers is that synergies could be overestimated. They could overestimate cost as well as efficiency synergies. Furthermore, they could be subject to the winners curse. The winners curse is called a curse since winning the bid implies that everybody else expected the uncertain value of the target to be lower. The bidder will receive a noisy signal about the value of the target. This signal is based on due diligence and/or an M&A adviser. However, since overconfident CEOs have the tendency of overweight their individual noisy signal this could lead to overpaying. When, at the end, they have won the bidding race against other candidates the price they have to pay will probably vary from the true value of the target.
In addition, related to the effect of overconfidence on capital structure, research shows that overconfident CEOs exhibit a stronger pecking order when it comes to financing decisions. They choose cash and debt over equity (Malmendier & Tate, 2008; Oliver, 2005). The reason for preferring debt to equity is that, in their perspective, equity is undervalued. Therefore, they will perceive raising capital through equity financing as excessive costly (Malmendier & Tate, 2008). In addition, the level of CEO overconfidence could influence investment activity. Investments often extend as CEOs become more convinced of their own capacities and skills (Malmendier & Tate, 2008). Especially when a firm has extensive availability of cash. Excessive free cash provides overconfident CEOs the opportunity of investing as they wish. As such, high cash flow-investment sensitivity could signal an overconfident CEO. Finally, with regard to trading activity, research was done by Odean in 1999. He found evidence of the disposition effect, which also originates from irrational behaviour. The disposition effect is the tendency of investors to hold losing investments too long while selling winning investments too soon. Odean (1999) showed that this phenomenon results in extensive trading activities while overall trading results are negative. Current literature suggests that rational
reasons for trading are shocks to wealth, private information or different preferences of risk. However, looking at the high frequency of trading today it seems that other factors affect trading activity as well (Odean, 1998). For example, trades could be triggered by past performance. A past performance is a signal on which individual traders base their decisions. As turns out, this signal could be extremely noisy or negatively correlated with future performance leading to negative returns.
Further evidence of overconfidence comes from research done by Ben-David et al. (2010). When testing whether financial executives are miscalibrated, they asked CFOs, among others, to estimate the probability distribution of stock market returns. As a result, they showed that 58% of the executives estimated volatility below the lowest value realized in the past 60 years, which was 5.1%. Also, they found that realized returns are in the CFOs’ 80% confidence interval only 36% of the time. This evidence shows the deviating perceptions of CFOs relative to the actual volatility and risk in the market. The research of Ben-David et al. (2010) also provides evidence that these CFOs, which demonstrate the misconception of the actual probability distributions of the stock market return, also possessed deviating perceptions about their own firm. This could be an implication of overconfidence CFOs.
The effects of overconfident CEOs are multiple and ambiguous. According to several studies CEO overconfidence often results in value deteriorating choices (Malmendier & Tate, 2008). On the other hand, in taking more risk, results will often be extreme. This can lead to either positive or negative outcomes. As a result, benefits of overconfidence behaviour will be discussed in the next chapter.
2.3 Beneficial implications of CEO overconfidence
Previous working on the adverse consequences of CEO overconfidence raises the question of why firms hire overconfident managers or CEOs. While recent literature has yielded a number of various insights, this puzzling question often stays unanswered. In order to answer the central questions of this study, existing literature on the reason to hire overconfident CEOs and the beneficial effects will be discussed.
One of the studies that have addressed the question comes from Goel and Thakor (2008). The focus of their research is overconfidence and the selection process of CEOs. They found that with the selection of a CEO – whether from within or from outside the firm – the choice is made out of a pool dominated by senior executives who have survived multiple intrafirm tournaments. These tournaments are often decided based on the level of risk one is willing to take. Hence, especially in a tournament setting, taking risk often leads to promotion. Therefore, by being positively biased about their abilities overconfident managers are more likely to pursue careers as CEOs. Goel and Thakor (2008) also suggest that when an overconfident manager is introduced his probability of getting selected to become CEO is higher than rational counterparts. As a result, overconfidence is an attribute, which is more prevalent with CEOs and managers than the general public. In addition,
Gervais et al. (2011) show evidence that CEOs have the tendency to become even more overconfident after a successful deal or investment. This is because they attribute too much of their success to their own ability without incorporating outside factors (Daniel et al., 1998; Gervais & Odean, 2001). From a theoretical perspective this implies that, when a CEO has completed a successful task in the past, the probability exists that he will put even more weight on his private signal the next time an opportunity occurs. Consequently, the level of risk he perceives becomes even less, often resulting in value-destroying investments. This analysis complements work of Gervais et al. (2011) and Goel and Thakor (2008) who found that a moderate level of overconfidence benefits the firm, while an extreme form is detrimental.
In contrast to the existing literature, which suggests that beliefs and preference can’t be disentangled (Sandroni & Squintani, 2007), Goel and Thakor (2008) designed a model that permits separation between overconfidence and risk aversion. According to their results overconfidence affects the value of the firm in a non-monotonically way. The research was constructed as followed. First, they examined a risk-averse rational CEO. The results showed underinvestment in projects relative to shareholders’ optimum. Subsequently, they found that a risk-averse moderate overconfident CEO diminishes this underinvestment and increases firm value. The reason they suggested was that overconfident CEOs would overestimate the precision of their private information, therefore overestimate the amount of risk the information eliminates. The CEO overreacts to his signal and invests in scenarios where a rational CEO would not. Finally, they show that in the event of a risk-averse and sufficient overconfident CEO, overconfidence is detrimental. He generates overinvestments and decreases the value of the firm. Goel and Thakor’s (2008) conclusion is therefore that low ability and either excessively cautious or excessively overconfident CEOs are not beneficial for firm performance.
As discussed, personal attributes and behavioural biases, such as overconfidence, affect the information-provision incentives and corporate decisions of CEOs. Roll (1986) was the first to examine overconfidence in relation to an increasing number of corporate takeovers. His conjecture that successful investments require risk-taking is supported by Gervais et al. (2011). However, they add to the existing literature that overconfidence leads to overinvestment in case of a suboptimal contractual arrangement between firm and decision maker. Their analysis shows that overconfidence can be detected and that the bias can be controlled by offering appropriate contracts. As such, overconfidence doesn’t need to result in overinvestments. Even, assuming that firms can identify overconfidence, by offering a compensation-based contract the firm can benefit from the excessive beliefs of the CEOs in achieving successes while shifting the risks onto the CEO. Gervais et al. (2011) build on the work of Fershtman and Judd (1987) and Englmaier (2006) who included optimal incentives and showed that overconfidence commits managers to be more aggressive which could result in profitability. Additionally, Hackbatch (2008) supports these findings by showing that overconfidence can act as a commitment device. Finally, the finding that firms can benefit from
overconfidence as long as it can be detected is in alignment with research of Palomino and Sadrieh (2011).
Gervais et al. (2011) further extend their model by incorporating the competitive labor market. They found that firms have to compete to attract the manager of their choice. The type of contract suited to attract this manager depends on the degree of overconfidence and the type of firms that seeks to hire him. When firms compete for a mildly overconfident manager they rely on his modest level of confidence to suffice for his commitment to undertake valuable risky projects. A limited level of performance-based compensation would then satisfy to align incentives. However, to attract the manager firms have to increase the safer portion of his salary. This makes an overconfident manager better off than his rational counterpart and allows for a more efficient transfer of economic surplus. The risk will be shared more equally due to a flatter compensation schedule. On the other hand, in the case of an excessive overconfident manager it would be optimal of the firm to increase the performance-based part of his contract. In that event, the manager would overvalue this type of compensation resulting in a transportation of the risk onto the manager. In this perspective, Gervais et al. (2011) confirm with Goel and Thakor (2008) that modest overconfident CEOs can add value, while extremely overconfident CEOs destroy firm value. They finish by pointing out that mildly overconfident and rational managers more often end up working for diversified, safer, value firm, which offer them a relatively flat compensation package. While on the other hand excessively overconfident managers are attached by risky, focused, growth firms that offer them a contract with a large part of performance-based compensation (Ben-David et al., 2010). This phenomenon puts the most confident managers into those firms where overconfidence can radically influence strategy, investment choices, and survival. Also, they show that overconfident managers are more easily motivated to learn about risky projects, since they overestimate the potential benefits. As a result, overconfidence implicitly commits them to exert effort to gather information that improves the possibility of success. At the same time, the firm, the manager, or both encounter the benefits depending on the firm and the market. The commitment to gather information is in contrast to the study of Goel and Thakor (2008) who focus on CEOs instead of managers. They suggest that overconfident CEOs invests less in information acquisition due to the extensive believe in their private signal. With respect to overconfidence as a commitment device (Kyle & Wang, 1997) examined this effect in a Duopoly model. In contrast to the traditional view (Friedman, 1953) that rational beliefs are better to survive the market test, they show that the overconfident trader experienced higher utility and expected profit. The reason was found to be the aggressive reputation of the irrational agent. His rational opponent recognizes this behaviour and consequently acts to it. Hence, the survival over overconfidence in this event is due to the fact that overconfidence acts like a commitment device, which both the players are aware of. Perhaps this is also one of the reasons why experts are more prone to overconfidence than beginners, in the sense that they have their reputation to live up to (Griffin & Tversky, 1992).
Hirshleifer et al. (2012) have also examined beneficial implications of overconfident CEOs. They acknowledge the puzzling question why firms give overconfident CEOs leeway to follow their beliefs when it comes to corporate decision (Malmendier & Tate, 2008; Ben-David et al., 2010). Unbiased believes are expected to be preferable, therefore hiring overconfident CEOs to run the company is counterintuitive. Hirshleifer et al. (2012) provide a solution: overconfident CEOs are better innovators. According to them overconfidence can increase shareholders wealth by engaging in risky projects (Goel & Thakor, 2008; Gervais et al., 2011). They show evidence that firms who hire overconfident CEOs have higher return volatility, aspire innovations and as such obtain more patent citations. Successes resulting from research and development expenditure consequently exceed those of companies with less overconfident CEOs. Be that as is it may, they also found that these successes only get achieved in innovative industries. Being overconfident therefore helps to exploit growth opportunities, which are prevalent in an innovative environment. The finding of Hirshleifer et al. (2012) conform to research done by Goel and Thakor (2008) and Gervais et al. (2011) which state that the benefits of overconfidence are higher when such opportunities are presented. These industries contain a higher level of risky growth opportunities, enabling CEOs to accomplish successes. With respect to the studies of Goel and Thakor (2008) and Gervais et al. (2011), in this scenario they address the excessively overconfident CEOs. Hirshleifer et al. (2012) examined a period of 10 year, from 1993 to 2003 and provide evidence that their results are not due to private information about future prospects or being more risk-intolerant. By using an exogenous instrument for future growth they show that overconfident CEOs are better at translating external growth opportunities into value. Furthermore, this effect is especially strong in innovative industries (Ben-David et al., 2010).
Other interpretations regarding overconfident CEOs and innovations come from, among others, (Galasso & Simcoe, 2011). The theoretical framework they present suggests that overconfident CEOs innovate to provide evidence of their ability. They also state that overconfident CEOs are expected to be enthusiastic about uncertain, risky, new business opportunities. As innovative projects are associated with uncertainty and complexity, overconfident CEOs might see them as a change to showcase their skills. This theory is supported by Griffin and Tversky (1992) who state that people tend to be more overconfident about difficult rather than easy tasks. Furthermore, innovative projects can be viewed as an indication of superior strategic vision (Hirshleifer et al., 2012). Therefore innovative initiatives are likely to appeal to self-absorbed CEOs. Another characteristic of innovative projects is their long-term horizon when it comes to results. As such, a reason for overconfident CEOs to pursue innovations can be explained by evidence that overconfident behaviour tends to be more severe in situations with ambiguous perspectives and deferred feedback (Einhorn, 1980). Galasso and Simcoe (2011) conclude their study by suggesting that overconfident CEOs underestimate the probability of fail, and are therefore more likely to be innovative. Also, firms hiring overconfident CEOs are more likely to seek new technological direction and, in consistency with Hirsleifer et al.
(2012), they show this effect is more prevalent in certain industries. Hence, industry characteristics are expected to amplify or mute the impact of behavioural biases.
The relation between CEO characteristics and innovativeness through R&D spending has been examined thoroughly. The main reason is that R&D investments can lead to corporate competitive advantage. Since these investments are key decisions made by executives, it is interesting to understand which characteristics are positively correlated to these expenditures. Hayes and Abernathy (1980) and Porter (1990) initiated the debate by examining the connection between innovativeness and a managerial background. More specifically, a technical, operational or scientific background. Both studies showed significant results. Barker and Muller (2002) extended the research by focusing on the CEOs’ education, age, tenure and career. They also provided evidence of a positive correlation with respect to R&D intensity. Finally, Yunlu and Murphy (2012) examined the moderating effect of CEO characteristics and R&D spending, while being embedded in recession. To quote them “It is important that scholars begin to understand the linkages between environmental conditions and upper echelon characteristics”. Their findings show that, during recession, CEOs with a shorter career horizon decreased R&D spending more dramatically compared to their counterparts with a longer career path. The perspective of Yunlu and Murphy (2012), to include environmental conditions, has also inspired to investigate the relation to overconfident CEOs. As such, recession will be included in this research and further clarified in the following chapter.
After assessing the implications of CEO overconfidence on corporate finance related decisions, as well as other effects of this behavioural traits the following hypothesis was formulated:
Hypothesis 1: CEOs that are low or excessively high overconfident influence firm value negatively
compared to a moderate CEO overconfidence.
In addition, although overconfidence is positively correlated to innovations the question remains whether rewards from successful innovations are sufficient to offset unprofitable investment activities. As such, the following hypotheses will be examined:
Hypothesis 2a: Overconfidence CEOs are better innovators.
Hypothesis 2b: This positive effect is stronger for companies in innovative industries.
Finally, the influence of being innovative and overconfident will be measured. Therefore the next statement is hypothesized:
2.4 Overconfidence in the context of recession
This study examines the effect of CEO overconfidence during recession whereas most former research excludes this environmental condition. Thus far, economic researchers have focused on the impact of overconfidence on corporate decisions without taking the general level of confidence into account. It may be suggested that the lens should be applied such that overconfident behaviour is the cause of our recent recession (Kowalski & Shachmurove, 2011). Yet another perspective could be that organizations and CEOs become embedded in this economic situation and what varies across the organizations, at least within the focus of this study, is level of confidence the CEO has. Therefore this study examines whether overconfidence, and adapting rapidly through innovations, can be beneficial recovering from a recession. Expected is that especially during times of financial difficulty, overconfident CEOs will benefit and grasp opportunities since others are cautious.
Recession solely, but also combined with R&D spends, have been examined extensively (Yunlu & Murphy, 2012). The generally perspective of these studies is a decrease in R&D expenditure during a recession. This evidence is further supported by research of Barlevy (2007) and Rafferty and Funk (2008) who showed the pro-cyclicality of R&D growth. Whenever the overall economy is growing, R&D spending on average increase, whereas diminishing research and development expenditures support a stagnation of the economy. Further research also indicates that deep recessions are one of the primary reasons for large R&D budget cuts (Brockhoff & Pearson, 1998). On the other hand, when extending the view to a long-term horizon, Morbey and Dugal (1992) show that R&D spends, while being embedded in recession, have positive long-term effects. This result justifies the hypothesis that overconfident CEOs, by being innovative, can be value enhancing during a recession. Further support, regarding this perspective, comes from Filippetti and Archibugi (2011). They reveal that under certain conditions, innovations can be value enhancing in times of financial crisis. Innovations can lead to a faster recovery and discovering growth opportunities even when markets are down. Especially CEOs in innovative industries will benefit due to the enhancing effect of overconfidence on innovations. By making use of their ability to react quickly and adapt to a new situation they can seize opportunities during a general low-level of confidence.
Additionally, Srinivasan et al. (2005) have conducted research on the effect of marketing activities during recession. They found that proactive marketing, under certain circumstances, could be value enhancing. The results show three firms characteristics, which amplify these effects: strategic emphasis on marketing, an entrepreneurial culture and slack resource. Hence, firms that focus, before a recession occurs, on marketing activities could establish an advantage over their weaker competitors by investing aggressively. In the light of this study, these results could imply that the effect of CEO overconfidence will be stronger for companies in innovative industries. Experience in detecting innovative opportunities, or marketing opportunities in the case of the study of Srinivasan et al. (2005), could be a reason. Also, for companies that are operative in innovative industries, being innovative is essential for survival.
2.5 An example from the field: The overconfident US vs. a reserved Europe during the financial crisis of 2007/2008
Since little empirical research was conducted on the behaviour of CEOs during a recession, this chapter will provide insights on the results of overconfident behaviour by addressing an alternative practical example. As such, the reaction with respect to the financial crisis of 2007, of country that has known to be aggressive and overconfident compared to a reserved continent, will be examined.
Before further considering the relation between overconfident and recession, a short summary of the financial crisis starting in 2007 is given. The paper of author Markus K. Brunnermeier (2009) provides a widespread description of the events related to the financial crisis of the late 2000’s. It contains an overview of the foregoing occurrences, as well as the mechanisms amplifying the crisis:
The foundation of the financial turmoil in 2008 was created by the combination of financial innovations and a transformation in the traditional banking sector. One of these innovations was securitization. In an attempt to off-load risk and distributed it to those who were willing to bear it, banks started to pool loans, like credit card receivables, mortgages and corporate bonds. Subsequently, these pools of assets were repackaged and sold to financial investors as tranches with different credit ratings (Brunnermeier, 2009).
When people first started to default on the subprime mortgages, the uncertainty about the value of the structured products and the reliability of the rating started to increase. As a consequence interest in these securities fell and therefore liquidity dried up. Since banks were reluctant to lend amongst each other, the government had to intervene. Banks, who were primarily holders of structured products, had to write-down billions on mortgage-related securities and losses in wealth on the U.S. stock markets were enormous. These losses in the mortgages market exacerbated and spilled-over to other sectors, leading to a severe financial crisis and a low level of trust and confidence in the market (Brunnermeier, 2009).
When discussing the effect of overconfidence and innovation during a recession, it is interesting to examine the different policy responses of the United States and European countries. The U.S., who are known for their can-do mentality, reacted rapidly. Through an aggressive response the Federal Reserve helped to support employment and incomes during the first year of the crisis. Their first action was a drastic cut in the discount rate. In addition, the Federal Open Market Committee began to ease monetary policy, which was quickly followed by other measures (Bernanke, 2009). To quote Ben Bernanke (Chairman of the Federal Reserve at the time) during a speech at the London school of Economics “As indications of economic weakness proliferated, the Committee continued to respond,
bringing down its target for the federal funds rate by a cumulative 325 basis points by the spring of 2008. In historical comparison, this policy response stands out as exceptionally rapid and proactive. In taking these actions, we aimed both to cushion the direct effects of the financial turbulence on the
economy and to reduce the virulence of the so-called adverse feedback loop, in which economic weakness and financial stress become mutually reinforcing”. In the months to follow the Fed had no
other choice than to even further reduce the rate down to 0 to 25 basis points for the target federal funds rate (Dijkstra, 2013). At the same time, a number of short-term lending rates declined extensively, which reflected the monetary easing. This phenomenon was associated by restrictive lending standings since traditional funding sources found themselves short of capital. Therefore the US felt the need to expand their policy. The Federal Reserve accomplished this by support of the functioning of credit markets and by providing liquidity to the private sector. They stimulated the economy by buying an extreme amount of financial assets from commercial banks and other financial institutions, to increase prices and lower yields (Bernanke, 2009). As such, large corporations benefited of cheap funding and investors could easily obtain additional capital to invest. This regulatory policy got known as Quantitative Easing.
Assessing the response of Europe showed a different view. The main differences were the pace at and the extent to which both economies responded. The U.S. government responded more rapidly and more aggressive then the Euro area did. However there were also similarities. For example, the standard monetary policy to adjust the key interest rates downward in order to achieve price stability (Praet, 2013). In the case of Europe, the standard monetary policy to reduce interest rates however was impaired. The transmission mechanism, which facilitates interbank lending, was dysfunctional because the banking sector in some countries became dysfunctional themselves. As such, they were unable to perform their normal intermediating role. In response to this challenge, the European Central Bank engaged in a sequence of measures to improved financial integration. These monetary policy impulses were fixed-rate lending operations by the ECB or the IMF, longer maturity of liquidity and expansion of the set of assets that could serves as collateral for receiving liquidity (Praet, 2013). Further steps, as was also part of the US response, were the intervening in the security markets to correct mal-functioning of certain segments. Programs that helped supporting this approach were the Securities Markets Program (SMP) and the Outright Monetary Transactions (OMT) program (Praet, 2013). However, these initiatives were, in contrast the quantitative easing policy of the US, not successful.
As already noted, the US reacted more extreme compared to the euro area. Although there are several reasons to clarify this difference, for example the American can-do mentality, one explanation deserves additional attention. Within the euro area all countries have national as well as international priorities. On one hand, the establishment of the monetary union made it easier for countries with a deficit to finance imbalances from a cross the border. However, this mechanism requires shared responsibility for its preservation. The crisis unveiled that this mechanism, as good as it may have seen, caused for an extensive level of risk that became apparent in 2008. Therefore, the credit crisis converted into the European debt crisis and for example Greece, Ireland and Portugal had to borrow billions of euros from the IMF to prevent bankruptcy (Dijkstra, 2014).
The reason for elaborating these chain of actions from both nations is to see what can come from acting aggressively and overconfident. The US, who lowered their interest rate more rapidly than the ECB and supported their banks through a capital injection of 324 billion euro in 2008 (Dijkstra, 2013)), recovered faster from the recession. The reason is their drastic approach in crisis management. The results show an S&P 500 that has already exceeded the level of the Internet bubble in 2000. Their economy showed a growth percentage of 2.4 in the first quarter of 2014 compared the previous year (Dijkstra, 2014). Since the American banks received an extreme amount of capital they were able to fulfill their task and lend money to companies and consumers. In line with their overconfident mentality the Americans looked for innovative opportunities with respect to energy (Dijkstra (2013)). When comparing these results to the recovery of the euro area the difference is obvious. Due to the debt crisis, governments are reluctant to pay interest. Investors withdraw from the European market since the returns on investments are limited and, as such, the value of the euro decreases.
The effects of overconfidence are multiple and ambiguous. In taking more risk, the results will often be extreme which can either be positive or negative. Also, according former research overconfident CEOs will recover faster from a setback by taken more risk and grasping growth opportunities. As a consequence they will have a higher positive effect on firm performance while recovering from a recession. In the example of the U.S. policy response and the European policy response the results of overconfidence are evident. The US, by taking excessive risk, has realized a fast recovery of a higher value. Keeping this in mind, the following statement is hypothesized:
Hypothesis 4: Embedded in recession, overconfident CEOs will have a positive effect on firm
performance.
In addition, this research explores the impact of overconfident CEOs in innovative industries. Did overconfident CEOs in innovative industries gain additional from the global financial crisis, by grasping opportunities when others were cautious? As such, the amplifying effect between CEO overconfidence and the global financial crisis for companies in innovative industries is examined:
Hypothesis 5: There is a catalysing effect between CEO overconfidence and the global financial
H 2a: + H 1: + H 1 H 2b: + H 3: +
3. Model, variables and data
3.1 Model
Figure 1 provides an overview of the model that is tested in this study, in order to answer the main question. Since the effect of overconfident CEO’s on firm value is examined, firm value is the dependent variable and CEO overconfidence is the independent variable. To provide multiple insights, several sub questions will be examined. Hence, other independent variables are included. These variables are innovativeness, innovative industries, CEO turnover and recession years. The interaction of these variables will show if an amplifying effect on the value of the firm exists. In addition, the dependent variable, firm value, is affected by many factors. Therefore, control variables are included.
Figure 1. An overview of the hypotheses used to answer the central questions and their expected outcome.
Dependent variable Firm value Control variables Assets Leverage Profitability Industry Independent variable CEO Overconfidence Dependent variable Innovativeness Independent variable CEO overconfidence * Innovativeness H 1: + H 1 Independent variable CEO overconfidence * CEO
turnover Independent variable CEO overconfidence * CEO
turnover
(Sample: Recession years)
H 4: + H 1
Independent variable CEO overconfidence * Recession year dummy
H 5: + H 1 Independent variable CEO overconfidence * Innovativeness
3.2 Variables
3.2.1 Firm value
Tobin’s Q provides a measurement for firm value. Former research on firm performance shows that Tobin’s Q is a valid approximation (Tobin, 1969; Fang et al., 2009). Generally Tobin’s Q is defined as dividing market value of assets by the replacement costs of assets (Chung & Pruitt, 1994). However, since calculating replacement costs of assets requires many assumptions, and therefore is dubious, often the book value of assets is used. This variable is defined as Total Stockholders’s Equity minus Preferred Stock at Carrying Value, plus Deferred Taxes (Balance sheet). Since Preferred Stock at Carrying Value is always missing, its value becomes zero. Market value of assets on the other hand is determined by summing total assets and market equity and subtracting book equity value. In this scenario, market equity is determined as Number of Common Shares Outstanding times the Stock Price at Fiscal Year end. Finally, when dividing the market value of assets by the book value of assets, a simple approximation is given. Calculating Tobin’s Q as describe above is in line with the study of Malmendier and Tate (2005a).
3.2.2 Overconfidence
As many researchers have pointed out, it is hard to construct a proxy for overconfidence. Overconfidence is a biased belief that is hard to observe and measure. Nevertheless, several have made an attempt to develop a method in order to measure overconfidence. To begin with, as explained in section 2.1, some studies on overconfident behaviour use surveys. Svenson (1981) who examined personal beliefs on someone’s drivers’ ability and Buehler et al. (1994) who studied the estimation of the time necessary to finish a task are examples of these studies. Difficulty with these types of studies is collection data. This is often time consuming and not all qualified participants will be willing to actual participate. Therefore, this study is not carried out using a survey.
A second method to construct a proxy for overconfidence is by using firm characteristics. For example, excessive leverage and acquisition intensity are, according to former research, signs of overconfident behaviour (Malmendier & Tate, 2008; Oliver, 2005). As such, Hribar and Yang (2010) have utilized the level of leverage in order to identify overconfidence. This study, however, will not be using such characteristics. For the reason, that other factors besides managerial overconfidence, affect these characteristics as well. Therefore, it is harder to find support for a causal relation between overconfidence and firm value.
Existing literature has pointed out that the most influential proxy for overconfidence is constructed by Malmendier and Tate (2005a). Based on the execution of valuable options they found a classification for managerial overconfidence. In addition, they designed an alternative measure. This proxy is based on the assessment of press releases focusing on a specific CEO. When the word overconfident, or related words, is mentioned more often than conservatism, when discussing the CEO, this implies overconfident behaviour. However, this proxy requires many research and access to
such articles. Therefore, in this study, the proxy based on option execution is used. This proxy, which they called “holder 67” exploits the under diversification of CEO’s. As part of their compensation package CEO’s often receive company options and stocks. With respect to these securities, CEO’s are limited when it comes to trading possibilities. They cannot sell short, since their human capital is invested in the firm, but also other trade possibilities are limited. For example, the time they need to wait before they can sell their stocks. Therefore, in order to reduces its exposure to idiosyncratic risk, expected is that a risk averse CEO would executive its options early. As was explained in the literature review of this study, overconfident CEO’s on the other hand, have the tendency to see their companies stocks as undervalued. They expect the value of the stock to go up. As a result, Malmendier and Tate (2005a) constructed a measure based on unexercised valuable options. When a CEO’s holds options, which are more than 67% in the money, Malmendier and Tate (2005a) classified them as overconfident. 67% is a threshold that was chosen as a rational benchmark for option exercise based on a framework design by Hall and Murphy (2002).
The “holder 67” can easily be constructed by extracting data from ExecuComp. The average value, or moneyness, of the options of the CEO needs to be calculated. This is accomplished as follows. First, for each year, the Realizable Value per Options is calculated by dividing the Realizable Value of all Options by the Number of Exercisable Options held. Consequently, the exercise price needs to be constructed. This is done by subtracting the Average Realizable Value per Option from the Stock Price at Fiscal Year End. Finally, the moneyness of the options is determined by dividing the Stock Price at Fiscal Year End by the exercise price. This ratio minus one will give a percentage of the moneyness to the options of the CEO. As a result, the variable for overconfidence is a dummy that is equal to one if a CEO holds options that are equal to or more than 67% in the money.
Following Campbell et al., (2010), a second proxy for overconfidence is given. In essence, they apply the same method as Hirshleifer et al. (2011) but in addition they distinguish low, moderately and highly overconfident CEOs. In this scenario, moderately overconfident CEOs are the baseline group and are therefore excluded from the specification. A CEO is classified as low overconfident when he exercises stock options which are less than 30% in the money. Highly overconfident CEOs are those executives that hold their stock options while they are more than 100% in the money. Consequently, moderate overconfident CEOs exercises their stock options when they are between 30% and 100% in the money. The variables are calculated in a similar matter as the “holder 67” and again CEO that do not have unexercised exercisable options are excluded. If hypothesis 1 holds this would imply that low and excessively high overconfident CEOs would impair firm value whereas moderate overconfident CEOs would increase company performance.
3.2.3 CEO turnover
A change in the behavioural traits of the CEO, accompanied by a change in firm value could imply a connection between the two variables. Hence, in order to examine causality between CEO
overconfidence and the value of the firm, a variable needs to be admitted indicating a switch from a CEO, which is not identified as overconfident, to a CEO that is. Without including CEO turnover other elements could affect firm value, which makes it precarious to assign an alteration in firm value as a consequence of CEO overconfidence. When a new CEO has the behavioural trait overconfidence, while his predecessor did not and the firm value increase, this could indicate causality. As such, a dummy variable is required indicating a new CEO. The year of CEO turnover is defined by comparing the CEO identifier of the next observation ‘[_n+1]’ with the current observation. Equal firm identifiers, but unequal CEO identifiers indicate a CEO turnover. If this occurs, the dummy variable CEOturnover was given the value one. The regression was carried out only including firms with at least one CEO turnover during the observed period. Including firm without a CEO turnover would obscure the results.
3.2.4 Innovativeness
Innovativeness is often seen as one of the positive consequences of CEO overconfidence (Palomino & Sadrieh, 2011; Allayaniss & Weston, 2001). Expected is that by being innovative overconfident CEO can profit in times of recession. With respect to innovativeness, several measurements are used in former research. For example Hirschleifer et al. (2011) use the amount of patent and patent citations. However, this data is difficult to obtain. R&D expenses give another indication of innovativeness. R&D expenditure can be normalized in several ways. For example, by following Allayannis and Weston (2001) which calculated the level of Capital Expenditures or Advertising Expenses to Net sales. This study will use the same ratio by dividing R&D spends by Revenue.
3.2.5 Innovative industries
In order to calculate the effect of CEO overconfidence on firm value for firms in innovative industries, a specification of which industries are innovative is required. The identification for this variable is derived from a paper written by Audretsch and Feldman (1996). They provide an overview of the seven most innovative four-digit (SIC) industries. This list is as follows:
3573 Computers
3823 Process control instruments 3662 Radio and TV communications 3674 Semiconductors California 3842 Surgical appliances 2834 Pharmaceuticals
3.2.6 Recession years
Hypotheses 3 and hypotheses 4 require identification of recession years. To get as much data as possible the moment the financial crisis originated will be part of the sample. As such, 2007, 2008 and 2009 are recession years. This is in line with the research of Saunders and Allen (2010). As a result,
RY is the dummy variable that specifies if there is a recession or not; 1 is a recession year and 0 not.
3.2.7 Control variables
The control variables will be assets, leverage, profitability and industry. Leverage is defined as debt divided by total assets, where debt is calculated by combining long-term debt and current liabilities. In addition, profitability is defined as operating income before depreciation (EBITDA) divided by total assets. Below the abbreviations of variables discussed:
𝑇𝑄!"= 𝐹𝑖𝑟𝑚 𝑉𝑎𝑙𝑢𝑒 𝑂𝑉!"= 𝑂𝑣𝑒𝑟𝑐𝑜𝑛𝑓𝑖𝑑𝑒𝑛𝑐𝑒 𝐼𝑁𝑁!" = 𝐼𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑣𝑒𝑛𝑒𝑠𝑠 𝐴!" = 𝐴𝑠𝑠𝑒𝑡𝑠 𝐿𝐸𝑉!"= 𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 𝑃𝑅𝑂𝐹!"= 𝑃𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑆𝐼𝐶!"= 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝐼𝑁𝐼𝐷!"= 𝐼𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑣𝑒 𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑑𝑢𝑚𝑚𝑦 𝑅𝑌!" = 𝑅𝑒𝑐𝑒𝑠𝑠𝑖𝑜𝑛 𝑦𝑒𝑎𝑟 𝑑𝑢𝑚𝑚𝑦 𝐶𝐸𝑂𝑡𝑢𝑟𝑛𝑜𝑣𝑒𝑟!" = 𝐶𝐸𝑂 𝑡𝑢𝑟𝑛𝑜𝑣𝑒𝑟 𝑑𝑢𝑚𝑚𝑦
3.3 Data and sample
The sample consists of all companies in the WRDS ExecuComp database. Limiting the sample to only companies in, for example, the S&P 500 would result in a narrow sample and therefore less support for any recommendation. ExecuComp only includes North American companies and thus the sample will as well. The timeframe of the research will be 2003 – 2013, since the focus lies on finding the effect of overconfident CEOs during recession.
The method of measuring CEO overconfidence is similar to the method applied by Malmendier and Tate (2008). The data on CEO characteristics is obtained from ExecuComp. Also collected from ExecuComp is data on CEO turnover. In order to reduce endogeneity CEO turnover is
included. If an overconfident CEO succeeds a CEO who is not overconfident and this change is accompanied by an increase in firm value, this could imply causality. Although this is a suited way to measure the effect of CEO overconfidence on firm value, these changes do not occur often. The data presents 266 observations in which a company replaced their current CEO with an overconfident CEO. Therefore, even if a positive outcome is found this limitation has to be taken into account. Furthermore, CEO overconfidence is measure by unexercised option value. Therefore, CEO’s with no valuable options are excluded from the sample. If they would be preserved their value would become 0, which automatically implies a non-overconfident CEO, even though it is not possible to observe the level of overconfidence. If, however, there is data on one of the 10 years the observation is kept. Overconfidence is seen as a persistent behavioural trait, so data on one year would be enough to identify a CEO as overconfident.
The most important limitation of this study is endogeneity. Providing evidence that an increase in firm value comes from an overconfident CEO will be difficult, since an increase in firm value could be the result of unobserved variables. For example, experiences, the size of a company or company culture can influence firm performance positively. Using firm fixed effect and difference in difference models tries to control for these factors. In addition, 1 year lagged variables are included to measure the effect of overconfident behaviour on firm value after a period of 1 year.
In order to measure innovativeness, R&D expenses are used. The data show however, that R&D expenses are often missing or 0. This has to be taken into account when interpreting the results. Data on R&D expenses is obtained from COMPUSTAT. Additionally, data to calculate Tobin’s Q, innovative industries and the control variables are also collected from COMPUSTAT. Following Hirshleifer et al. (2010) firms and years with missing data on the dependent, independent or control variables are not included. As such, the final sample of panel data consists of 9513 observations, 8750 observations on overconfident behaviour and 1055 companies.
4. Methodology
4.1 Regression specification
OLS is used to measure the effect of overconfidence on firm value. The endogeneity problem has to be taken into account, since other (non-measurable) factors could also influence firm value. Therefore panel data analysis will be used, by including firm and year fixed effects, to control for omitted or unobservable characteristics. In addition, firm-level controls are added. The regressions specified below will be extended with lagged variables and an alternative proxy for overconfident behaviour. Including these variables provides additional insights with respect to the economic value of the results.
The hypotheses will be measured as follows:
Hypothesis 1: CEOs that are low or excessively high overconfident influence firm value negatively compared to a moderate CEO overconfidence.
𝑇𝑄!" = 𝑂𝑉!" + 𝐴!"+ 𝐿𝐸𝑉!"+ 𝑃𝑅𝑂𝐹!"+ 𝑆𝐼𝐶!"+ 𝜀!"
A second econometric model will provide insights on the effect of changes in the level of CEO overconfidence on the changes in Tobin’s Q. This however can only be measured when a change in CEO has occurred. By applying the difference in difference methodology, where the treatment is a change from a CEO who is not overconfident into an overconfident CEO, this effect can be measured. As such, T is a dummy variable indicating whether i belongs to the treatment group (=1 for treated firms even before the treatment) and post the dummy variable indication whether an observation belongs to the post-treatment period (=1 even for non-treated firms). In that scenario, the regression analysis will be as follows:
𝑇𝑄!" = 𝑇!"+ 𝑝𝑜𝑠𝑡!"+ 𝑇!"∗ 𝑝𝑜𝑠𝑡!"+ 𝐴!"+ 𝐿𝐸𝑉!"+ 𝑃𝑅𝑂𝐹!"+ 𝑆𝐼𝐶!"+ 𝜀!"
Hypothesis 2a: Overconfident CEOs are better innovators.
In line with the study of Heirshleifer et al. (2011) the effect of overconfident behaviour on innovativeness is measured. They however used two proxies: R&D expenses and patent citations. Since patent citations are not part of the sample their results could differ from the result in this study. Similar to hypothesis 1 these regressions will be executed using firm and year fixed effects:
Hypothesis 2b: This positive effect is stronger for companies in innovative industries
𝐼𝑁𝑁!" = 𝑂𝑉!"+ 𝐼𝑁𝐼𝐷!"+ 𝑂𝑉!"∗ 𝐼𝑁𝐼𝐷!" + 𝐴!"+ 𝐿𝐸𝑉!" + 𝑃𝑅𝑂𝐹!"+ 𝜀!"
Hypothesis 3: By being innovative, overconfident CEOs will have a positive effect on firm value.
Expected is that being innovative will have a positive effect on firm value. Therefore, an analysis is included measuring the enhancing effect between overconfidence and the level of innovativeness of CEO’s.
𝑇𝑄!" = 𝑂𝑉!"+ 𝐼𝑁𝑁!"+ 𝑂𝑉!"∗ 𝐼𝑁𝑁!"+ 𝐴!"+ 𝐿𝐸𝑉!"+ 𝑃𝑅𝑂𝐹!" + 𝑆𝐼𝐶!" + 𝜀!"
Hypothesis 4: Embedded in recession, overconfident CEOs will have a positive effect on firm performance.
The regression will run only using recession years. When the results of this regression are compared to the outcome of hypothesis 1 the effect of CEO overconfidence during crisis measured. In order to answer hypothesis 4 the following specification is used:
𝑇𝑄!" = 𝑂𝑉!"+ 𝐶𝐸𝑂𝑡𝑢𝑟𝑛𝑜𝑣𝑒𝑟!"+ 𝑂𝑉!"∗ 𝐶𝐸𝑂𝑡𝑢𝑟𝑛𝑜𝑣𝑒𝑟!"+ 𝐴!"+ 𝐿𝐸𝑉!"+ 𝑃𝑅𝑂𝐹!"+ 𝑆𝐼𝐶!"+ 𝜀!"
Hypothesis 5: There is a catalysing effect between CEO overconfidence and the global financial crisis for companies in innovative industries.
To see if there is an enhancing effect between CEO overconfidence and the global financial crisis panel data analysis is utilized to control for unobserved characteristics. In addition, an interaction variable is included with a dummy variable indicating if there was a crisis during that year or not. The regression analysis will be as follows:
𝑇𝑄!" = 𝑂𝑉!"+ 𝑅𝑌!"+ 𝑂𝑉!"∗ 𝑅𝑌!!+ 𝐴!"+ 𝐿𝐸𝑉!"+ 𝑃𝑅𝑂𝐹!"+ 𝑆𝐼𝐶!"+ +𝜀!"
The regression analysis to measure the effect of overconfidence on firm performance for companies in innovative industries will use a comparable regression. However, to measure this effect a sample of only companies in innovative industries will be use. In this scenario the variable SIC will be excluded.
