1
The influence of business cycles on the corporate financing
decisions of overconfident CEO’s
Jasper Stumpel 6126553
Faculty Economics and Business (FEB) MSc Business Economics, Finance track Master Thesis
July 2015
2
Statement of Originality
This document is written by Student Jasper Stumpel who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Abstract
In this thesis the relation between business cycles and the corporate financing decisions of
overconfident CEO’s is discussed. There already exists evidence that overconfidence helps in better explaining the corporate financing decisions of firms. This research finds that overconfident CEO’s are more likely to issue equity than their rational peers. This contradicts the economic theory, which suggests that CEO’s with irrational beliefs exhibit a stronger pecking order. It is impossible to draw any conclusions from the relation of business cycles to the likelihood that an overconfident CEO’s issues relatively more equity. Furthermore, the findings suggest that an overconfident CEO issues less debt than a rational CEO during years of low growth.
4 Table of contents 1. Introduction 5 2. Literature review 6 2.1. Capital structure 7 2.2. Overconfidence 8
2.3. Consequences of overconfidence on firm characteristics 10
2.4. Measures of CEO overconfidence 12
2.5. Business cycles 14
3. Hypotheses 15
4. Methodology 17
5. Data and descriptive statistics 19
6. Empirical results 25 7. Additional tests 31 8. Discussion 33 9. Conclusion 35 10. Literature 37 11. Appendix 40
5
1. Introduction
Behavioural finance is a topic with many different research areas. Within behavioural finance the phenomenon overconfidence has always been a topic of interest. The behaviour of overconfident CEO’s is studied with respect to many areas within finance. There also exists a relation between overconfidence among CEO’s and capital structure. In the paper of Malmendier, Tate and Yan (2011) the findings of this relation are discussed. Overconfident CEO’s tend to issue less equity relative to their peers. However, the evidence of the economic theory is not yet very strong. Furthermore, the relation between the macroeconomic circumstances, the capital structure and these irrational beliefs and is never examined. This study tries to fill that gap.
Several different factors help determine the current capital structure of a firm. Myers and Majluf (1984) discover a preference by companies which funds to use to finance their investment decisions. Due to information asymmetries there exists a so called pecking order. Companies prefer to use internal funds over external funds. Next to that, they prefer debt over equity. However, Leary and Roberts (2010) conclude that only half of the observed financial decisions are explained by the pecking order theory. Malmendier, Tate and Yan (2011) find that the financing decisions of firms can be better explained by incorporating the effect of overconfidence among their CEO’s. They show that overconfident CEO’s exhibit a stronger pecking order and therefore have a higher preference for debt. Another effect of overconfidence is that overconfident CEO’s overinvest in prosperous times and underinvest in economic downturns (Jaimovich and Rebelo, 2007). This can negatively influence the performance of a firm in recessions. So business cycles have an influence on investment decisions and firm performance. What happens to the corporate finance decisions of overconfident CEO’s in economic downturns has, however, never been studied.
This research is valuable in trying to make the influences of overconfidence more robust. Next to that, it tries to expand the explanatory power by overconfidence on the capital structure by incorporating business cycles. It is possible that the business cycles capture an important time variant factor in explaining long term capital structures. It is useful to know how the behaviour of rational and irrational CEO’s is affected by the changing macroeconomic circumstances. Furthermore it is important for boards, among others, to know how CEO’s with irrational beliefs behave and how that behaviour effects financing decisions. By understanding these influences it is possible to partly constrain overconfident CEO’s implementing these, most likely non-optimal, policies. This can be beneficial for the company and their shareholders on the long term.
The question answered in this thesis is as follows. What can be concluded about the relation of business cycles and the effect of overconfidence on the capital structure decisions? This research
6
question is answered in two parts by means of an empirical study. First, the influence of
overconfidence on capital structures is examined. To answer this question a logistic regression is used with as dependent variable new equity issues. A dummy variable whether a CEO exhibits overconfident behaviour or not is included along with CEO, firm and return controls. The
overconfidence measure, developed by Campbell et al. (2011) is an adjusted version of the option based measure by Malmendier and Tate (2005a). CEO’s are classified as overconfident if they hold an option package that is more than 67% in the money. The empirical study contains panel data of 2,143 firms from the United States between 1992 and 2012. The second part tries to capture the influence of business cycles on this effect. A dummy is created to identify the years in which the GDP growth of the United States is relatively low, or less than 2%. This dummy variable, along with the interaction term of this dummy and the overconfidence variable is included in the main regression. The results of this interaction term will clarify if in economic downturns overconfidence among CEO’s causes different behaviour with respect to corporate financing decisions.
The structure of this paper is as follows. In section 2, the economic theory about
overconfidence, capital structure and business cycles is discussed. Then in section 3, the hypotheses are constructed and explained. Section 4 describes the methodology in this study to test the
hypotheses. In section 5, the data and datasets are clarified. Next to that, this section also provides descriptive statistics about the data. Section 6 describes the empirical results of the regression analyses. In section 7, additional tests are executed to test the robustness of the findings in this paper. Section 8 encloses the discussion, or the main limitations of this study. Finally, in section 9 the conclusions of this paper and recommendations for future research are discussed.
2 Literature review
This section describes the theoretical foundations of the capital structure, CEO overconfidence and business cycles. First, the capital structure and the preferences for certain funding options are discussed. Thereafter in paragraph 2, the economic theory about overconfidence among CEO’s is presented, followed by a discussion of the consequences of overconfidence among CEO’s in paragraph 3. In the next part there are several ways described to measure CEO overconfidence. Finally, in paragraph 5 theories about the effect of business cycles on capital structure and overconfidence are described.
7
2.1 Capital structure
In this section, the most common factors that influence capital structures are discussed. Modigliani and Miller (1958) were one of the first who wrote about the capital structure. Capital structure is the combination of equity, debt and hybrid securities by which a company finances its assets. In their first study they show that the choice between debt and equity has no effect on the value of the firm. Later on they state that the capital structure does matter and debt if preferable due to the
deductibility of taxes (Modigliani and Miller, 1963). However, increasing the level of debt raises the costs of financial distress (Myers, 1984). Firms with a high level of debt have a higher possibility that they are unable to meet their debt obligations (Myers, 1984). This is called the trade-off theory. The optimal capital structure is determined by weighting the costs of financial distress against the benefits of the tax shield (Myers, 1984). In the studies that followed after these theories, it turned out that there are other factors besides the tax shield and the financial distress costs that could leads towards an optimal capital structure.
According to Myers and Majluf (1984), companies have a preference for which funds to use to finance investment decisions. This is called the pecking order. It suggests that companies prefer to use internal funds, then debt and as a last resort option equity. They assume that there exist
information asymmetries, where management has more information than investors about the real value of the firm (Myers and Majluf, 1984). Investors only see the financing decisions the firm makes and interpreted those decisions accordingly. They value issuing equity as a wise thing to do when a company is overvalued and for that reason new equity issues will be seen as a bad signal (Myers and Majluf, 1984). In certain situations, companies will refuse to issue stocks and because of this, the company will not invest in profitable investment opportunities (Myers and Majluf, 1984). By doing this, firms try to minimize the costs of adverse selection.
The core factors that determine the capital structure of a firm are Market to Book ratio, Profitability, Tangibility and Size (Frank and Goyal, 2009). Companies with a high Market to Book ratio have a lower leverage ratio according to Frank and Goyal (2009). The reason for this is that the Market to Book ratio is a proxy for growth and growth reduces leverage. Profitability has a negative relation with leverage. On the one hand, profitable firms use more debt because they experience less financial distress costs and value interest tax shields more (Frank and Goyal, 2009). However, the effect of accumulated profits is higher according to Frank and Goyal (2009). Firms prefer to use retained earnings to finance their investments. Profitable firms have higher retained earnings and are therefore relatively less leveraged (Frank and Goyal, 2009). Firms with high Tangibility have a higher leverage. Tangible assets reduce the costs of financial distress and make issuing debt
8
favourable (Frank and Goyal, 2009). Larger companies are relatively more leveraged because they are more diversified and face less financial distress costs (Frank and Goyal, 2009).
The capital structure of a firm is explained by a lot of different characteristics as discussed before, but very likely also by some not widely accepted factors. Baker and Wurgler (2002) find that the current capital structure of a company is highly dependent on historical market values. They state that if a company has a high market to book ratio it is more likely to issue equity and contrary, if it has a low market to book ratio equity repurchases are more common. This is one aspect of the theory that equity market timing is important when considering their financing methods. Other evidence for market timing is based on the fact that equity is issued when the cost of equity is relatively low and repurchased when the cost is relatively high (Baker and Wurgler, 2002). Baker and Wurgler (2002) believe there is no optimal capital structure, but that the current capital structure is an outcome of an accumulation of historical financing decisions.
Leary and Roberts (2010) conclude that only half of the observed financial decisions are explained by the pecking order theory. They test the influence of factors of alternative theories on the new debt and equity issuances. Results show that with this improved model more than 80% of the observed financial decisions are explained. Thereby indicating that the pecking order theory only explains a portion of the financial decisions made (Leary and Roberts, 2010). Fama and French (2005) confirm this conclusion, as they prove that most firms issue or repurchase equity every year, thereby violating the assumptions of the pecking order of how often or in which situations firms make equity issuances. Lemmon, Roberts and Zender (2008) find that there is another unobserved effect that causes surprisingly stable capital structures. It is most likely determined by factors that tend to remain stable through time (Lemmon, Roberts and Zender, 2008). All these studies suggest that there are certain unknown factors that help explain the changes in the leverage ratios of firms. Including the phenomenon overconfidence might help to tackle this problem.
2.2 Overconfidence
There are two main definitions of overconfidence. Overconfidence is explained as an upward bias in the expectations of future events, also called overoptimism (Deshmukh, Goed and Howe, 2013). Furthermore, overconfidence is clarified as the overestimation of information one person receives and an underestimation of risk (Deshmukh, Goed and Howe, 2013). Overconfidence exists in several forms: the “better-than-average” effect, miscalibration and the illusion of control. In general, people are overconfident about their knowledge, their abilities and their future prospects (Barber and
9
Odean, 2001). Next to CEO overconfidence, CEO dominance is sometimes used in related research. CEO dominance is explained by Brown and Sarma (2007) as the ability of CEO’s to impose their overconfident views and ideas on the decisions of the company.
Overconfidence exists among all groups of people. The “better-than-average” effect occurs when people consider themselves to be above average, overestimating their own capabilities relatively to the average (Brown and Sarma, 2007). One of the first experimental evidences of this effect was shown in a psychological study performed by Svenson (1981). He demonstrates that a clear majority of his subjects was conveniently estimating their driving skills above median. This specific study was later on replicated in multiple countries, taking into account a diversity of IQ or skill related variables, to generalize the “better-than-average” effect for all kinds of behaviour.
Miscalibration is identified as the tendency to overestimate the information someone receives (Biais, Hilton, Mazurier and Pouget, 2005) or as the underestimation of the range of potential outcomes (Ben-David, Graham and Harvey, 2013). Biais et al. (2005) state that this misjudgement of a person explains why some investors fail to realize they are consequently losing and they are suffering from the winner’s curse while trading. People are often miscalibrated as they overestimate their own ability to predict future events. Next to that, they underestimate the volatility of random outcomes (Ben-David, Graham and Harvey, 2013).
People become overconfident due to biased self-attribution (Malmendier and Tate, 2005a). Let’s assume people obtain an own signal before a certain event takes place. At the moment the event takes place, a public signal is created. If their own signal is confirmed, their confidence will increase. However, if the signal is not confirmed their confidence will either not decrease or will decrease less than the increase in confidence that would have occurred otherwise. This means that people attribute good outcomes to their own actions (skills) and bad outcomes to bad luck (Miller and Ross, 1975), inducing overconfidence among people.
This study focuses on overconfidence among CEO’s. Overconfidence is more likely to apply to high skilled individuals, for example to high ranked executives. There are three main factors that underlie this statement. Malmendier and Tate (2005b) state that this is probably the case because of abstract reference points. People insufficiently take into account the skills of their comparison group. If a CEO compares himself to the average manager rather than to another CEO, he is referring to a wrong peer group and will therefore most likely conclude himself to be above average
(Malmendier and Tate, 2005b). Next to that, it is the complexity of the nature of the job a CEO has. A CEO is rarely able to make a good one-to-one comparison with others, since there is simply no other
10
CEO to compare with (Malmendier and Tate, 2005b). The decisions or investments a CEO makes are difficult to compare across other firms, what makes it hard to discover if a CEO is overestimating himself. The second factor is the illusion of control (Malmendier and Tate, 2005a). The CEO has the position to decide about large-scale investment decisions and is, therefore, more likely to believe that he is able to control the outcome. This results in underestimating the possibility of failure (Malmendier and Tate, 2005a). Finally, the CEO has a high degree of commitment to positive
outcomes (Malmendier and Tate, 2005b). A large component of the CEO’s compensation, stocks and options, is dependent of the result of the company. The CEO will receive more money if the
performance of the company increases. Besides, the value of the CEO’s human capital is also strongly related to the performance of the firm (Malmendier and Tate, 2005b).
Furthermore, overconfidence is a plausible explanation for excessive trading on financial markets. According to Barber and Odean (2001), overconfident investors believe the precision of their knowledge about certain securities is greater than their actual knowledge. They show that overconfidence leads to more trading, while doing so simultaneously reduces the expected returns. Psychological studies demonstrate that in areas such as finance, there exists a difference in level of overconfidence among gender. Men are in general more overconfident than women and therefore also trade more excessively (Barber and Odean, 2001). The research of Barber and Odean (2001) shows that men trade 45% more than women. For both men and women the expected return is reduced through trading. However, men reduce their return even 0.94 percentage points more than women.
2.3 Consequences of overconfidence on firm characteristics
There are various studies that try to explain the remaining variation in financial decisions by overconfidence among CEO’s. The study of Malmendier, Tate and Yan (2011), who examine the effect of overconfidence and other managerial traits on corporate financial policies, is of great relevance for this thesis. They find that overconfident CEO’s use less external finance than their peers, and issue less equity when nonetheless using external finance. Bertrand and Schoar (2003) agree that there are manager-fixed effects, or managers having different ‘styles’, that explain variations in financing and investment decisions. Where, for example, the birth cohort of an executive and having a MBA seem to have an effect on the level of aggressiveness of the strategies they follow (Barber and Schoar, 2013).
11
An overconfident CEO systematically overestimates the return he will receive on an
investment project (Malmendier and Tate, 2005a). If the CEO has access to sufficient internal funds to finance the investments he desires, he generally overinvests (Malmendier and Tate, 2005a). However, in the case that there are no sufficient internal funds available, an overconfident CEO is unwilling to issue new equity because he beliefs his company is undervalued (Malmendier and Tate, 2005a). Next to that, he knows that investors will probable negatively evaluate this decision. For this reason, the CEO will invest less and underinvests relatively to the market. This makes the investing policy of an overconfident CEO highly dependent of the quantity of excess cash a firm currently owns (Malmendier and Tate, 2005a).
Brown and Sarma (2007) endorse that CEO overconfidence is important in explaining the acquisitions decisions of a company. In economic theory there are three main reasons for a takeover: potential synergies, agency conflicts between managers and the shareholders of the company and managerial hubris (Brown and Sarma, 2007). Roll (1986) states that managers of the acquiring companies are in general too optimistic about the potential synergies in a possible takeover and therefore make valuation errors. These valuation errors result in overbidding for the target firm at the expense of their stockholders (Roll, 1986). Brown and Sarma (2007) find that overconfident CEO’s accomplish more acquisitions than rational CEO’s do, especially when it comes to diversified acquisitions. This is also acknowledged by the study of Malmendier and Tate (2008), who found that there is a 65% higher probability of making an acquisition if the CEO is
overconfident.
As discussed before, an overconfident CEO tries to avoid external financing, what it is costly in his view and invests more than a rational CEO. Therefore, it would make sense to hoard more cash for future investments by lowering the dividend payout (Deshmukh, Goel and Howe, 2013).
Deshmukh, Goel and Howe (2013) indeed find that the level of dividend payout is lower in companies with overconfident CEO’s compared to companies with rational CEO’s. Their results suggest that the dividend payout provides information about the level of CEO overconfidence. According to Hirshleifer, Low and Teoh (2012), overconfidence of CEO’s is beneficial for
shareholders. This is because those CEO’s are more likely to undertake risky investments. In their research they suggest that being overconfident is positively related with innovative growth opportunities and greater return volatility. However, only in innovative industries there will be greater benefit of innovations (Hirshleifer, Low and Teoh, 2012). Hirshleifer, Low and Teoh (2012) suggest that the reason why so many overconfident CEO’s are hired is perhaps because
12
Another way how an overconfident CEO can create value for the company is by the fact that he receives less compensation compared to rational CEO’s (Otto, 2014). Next to that, he receives smaller stock options grants and fewer bonus payments (Otto, 2014). This finding suggests that companies take into account whether their CEO is overconfident or not and set the compensation payments accordingly. A possible explanation for this phenomenon is that an overconfident CEO overestimates the probability of good outcomes (Otto, 2014). He therefore needs for example less incentive claims or incentive payments to make the investment.
Overconfident behaviour of CEO’s ensures behavioural distortions like overinvestment or non-rational information provision incentives (Malmendier and Tate, 2005). However, an
overconfident manager has a higher probability to become CEO than a rational manager (Goel and Thakor, 2008). Goel and Thakor (2008) state that a moderate degree of overconfidence of the CEO is actually beneficial for the shareholders because it reduces the possibility of underinvestment. On the other hand, when the board of directors discovers that the concerning CEO is too overconfident there is a high likelihood that the CEO will be fired (Goel and Thakor, 2008). These conclusions create a paradox, being overconfident as a manager increases the chance of becoming CEO, however being (highly) overconfident as a CEO increases the chances of being fired. Campbell et al. (2011) confirm that CEO’s with a high level of overconfidence are more often fired than rational CEO’s. In their view the initial hiring decisions of overconfident CEO’s are imperfect and are later on corrected by firing those CEO’s. As an overconfident CEO may destroy firm value, an independent board of directors is important to curb the concerning CEO (Brown and Sarma, 2007). This raises the question of how the level of overconfidence of a CEO is measured.
2.4 Measures for CEO overconfidence
It is difficult to find an objective way to measure if a CEO is overconfident or not. Is this research a CEO is qualified as overconfident or rational. The existing literature elaborates on different ways to measure CEO overconfidence. In the studies of Malmendier and Tate (2005, 2008, 2011),
overconfidence of CEO’s is measured in two different ways, by a press-based and an option-based measure. Managers or CEO’s are classified as overconfident if they overinvest their personal funds in their own company (Malmendier and Tate, 2008). The option-based measure of Malmendier and Tate is well known and often used in other research papers (Campbell et al., 2011) (Hirshleifer, Low and Teoh, 2012). An adjusted version of this option-based measure is used in this thesis. Next to that, two alternative measures of overconfidence are shortly discussed
13
The option-based measure of Malmendier and Tate (2005a) is based on the idea that an overconfident CEO holds on to his stock options longer than a rational CEO. A risk averse CEO will exercise his stock options early, the moment they are sufficiently deep in the money, to reduce his exposure to idiosyncratic risk. Therefore, holding on to stock options after the moment that exercise becomes permissible is considered evidence for optimistic beliefs about the prospects of the firm (Malmendier and Tate, 2005a). Malmendier and Tate (2008) find evidence that suggests that late exercise of options by CEO’s does not reflect inside information about the future performance of their companies. They also find that CEO’s do not earn abnormal returns due to holding those exercisable options for a longer period. Therefore Malmendier and Tate (2008) conclude that CEO’s with exercisable options are systematically overestimating the future performance of their
companies. Malmendier and Tate (2005a) distinguish two different option-based measures, LONGHOLDER and HOLDER67. LONGHOLDER is an indicator that equals one if a CEO holds options until the expiry date and the options are at least 40% in the money. HOLDER67 is an indicator that equals one if a CEO holds options with 5 years remaining duration and is at least 67% in the money. However, the dataset that Malmendier and Tate (2008) use is obtained from Hall and Liebman (1998) and Yermack (1995). The problem with this dataset is that it is not publicly available and that there exists no database that contains similar data.
Campbell et al. (2011) and Hirshleifer, Low and Teoh (2012) also use the option-based measure in their study. They use an adjusted version of the one from Malmendier and Tate (2008), which is created out of data that is publicly available. Next to that, it is possible to use datasets with a higher number of CEO’s over a longer and more recent time period. They create the average percent moneyness by dividing the realizable value per option by the average exercise price of the options (Campbell et al., 2011). This measure of overconfidence from Campbell et al. (2011) is used in this thesis to study the effect of overconfidence on the capital structure of the firm. The downturn of this method is that it is not possible, due to data restrictions, to take into account the duration of the option packages as Malmendier and Tate (2005a) do.
Another often used measure of overconfidence by Malmendier and Tate (2005a) is the press-based measure. This measure counts for example the words in press articles related to a CEO and overconfidence (like “overconfident”, “overoptimistic” etc.). Contrary, the words related to a rational or cautious CEO are counted. Based on those word counts a CEO is classified as
overconfident or not (Malmendier and Tate, 2005b). This method requires certain adjustments to control for, among others, a bias in the media for positive stories, or a connection with the word “overconfident” and historic stock returns or innovations (Hirshleifer, Low and Teoh, 2012). The
14
problem with this measure is that it is very time consuming and therefore unrealistic to examine a large dataset.
A relatively new method to measure overconfidence among CEO’s is used by Otto (2014). This measure is based on the earnings per share (EPS) forecasts. For this end, the forecasted EPS by the firm and the realized EPS are compared. The intuition behind this approach is that an
overconfident CEO overestimates the future performance of the company and is therefore more likely to publish EPS forecasts that are higher than the actual EPS will turn out to be (Otto, 2014). It is unlikely the firm would release EPS forecasts that are not in line with the beliefs of the concerning CEO. Despite of the fact that companies are not obliged to provide EPS forecasts, 70% of all firms actually does (Otto, 2014). This measure of overconfidence is however not yet commonly used in the economic theory about overconfidence and possible downsides or correlations with other CEO characteristics are not widely evaluated. All discussed measures of overconfidence have no effect on the sign of the results (Malmendier and Tate, 2005a) (Campbell et al., 2011) (Otto, 2014).
2.5 Business cycles
In this section the influence of business cycles on the capital structure is discussed. Furthermore, a possible different relationship between overconfidence and capital structure in periods of low growth is elaborated. As discussed before, Lemmon et al. (2008) suggest that a large part of the variation in capital structures is determined by an unobserved effect that tends to remain stable through time. Recent studies show that after taking into account the firm fixed effects, there are no significantly additional variables known to explain the capital structures of firms. Akhtar (2012) considers four different stages of business cycles: peak, contraction, trough and expansion. He finds that, when including the four different business cycles in the regression analysis, the explanatory power of the model of Lemmon et al. (2008) is improved. There exists a partial time invariant
characteristic in the different phases of a business cycle (Akhtar, 2012). This indicates that there may exist a combined trend between leverage ratios and business cycles towards certain equilibriums (Akhtar, 2012). A business cycle is time related and can help explain capital structures determined on basis of time related factors (Akhtar, 2012).
Mueller and Brettel (2012) study the effect of overconfidence among German CEO’s on firm performance during different business cycles. They include the different business cycles in their regression analysis, along with the interaction term of the overconfidence measure and business cycles. In economic downturns there are often a high number of write-offs. This is suggesting that
15
there were overinvestments and biased expectations beforehand (Mueller and Brettel, 2012). Overconfidence is helping to explain these non-rational decisions. According to Audia, Locke and Smith (2000), past success leads to greater strategic persistence among CEO’s after substantial environmental changes. This strategic persistence, or CEO’s becoming more overconfident, causes a decline in firm performance (Audia, Locke and Smith, 2000). The higher and earlier investments caused by overconfidence that take place in the beginning of a business cycle, result in lower firm performance during an economic recession (Mueller and Brettel, 2012). Contrary, those investments lead to a higher stock performance in the middle phases of a business cycle. This suggests that overconfidence may contribute to business cycle volatility (Mueller and Brettel, 2012).
The different phases of a business cycle together with the financial flexibility of a firm help explain the capital structure of that firm. Companies that do not experience high financing
constraints show counter-cyclicality in their leverage ratios (Levy and Hennessy, 2007). Those firms substitute debt for equity during economic contractions. Vice versa, they substitute equity for debt during economic expansions. Contrary, the leverage structure remains constant over the business cycle for firms with high financing constraints (Levy and Hennessy, 2007). Considering the pecking order theory, leverage should increase during contractions since internal funds decrease in recessions (Frank and Goyal, 2009). If profits decrease during economic downturns, information asymmetries increase and the agency problems between managers and shareholders are more severe. Therefore, companies should issue less equity during economic downturns (Frank and Goyal, 2009). What happens to the effect of overconfidence on the capital structure during recessions has never been studied. Kruger and Burrus (2004) and Fuster, Hebert and Laibson (2010) state that beliefs are excessively pessimistic in bad circumstances. According to Jaimovich and Rebelo (2007), an overconfident CEO will put too much weight on the signals he gets. Therefore, it could be possible that an overconfident CEO overestimates the bad signal or values it more than a rational CEO does. This thesis further elaborates this possible effect.
3. Hypotheses
Malmendier, Tate and Yan (2011) find that whether or not a CEO is overconfident affects the corporate financing decisions of the firm the CEO works for. An overconfident CEO prefers internal finance as he sees external finance as relative costly. However, when internal finance is not sufficient, overconfident CEO’s tend to issue less equity than rational CEO’s and prefer debt. This
16
research examines if this effect is still visible. Therefore a more recent time period is used than the study of Malmendier, Tate and Yan (2011). Next to that, the effect is investigated over a longer time period and a different, easier constructible measure for overconfidence is used.
Hypothesis 1: Overconfident CEO’s choose less often equity financing than rational CEO’s.
The idea behind this hypothesis is that overconfident CEO’s believe that risky debt is too costly compared to internal finance or riskless debt. The CEO overestimates cash flows in default states and therefore beliefs the interest rates demanded on the risk debt are too high (Malmendier, Tate and Yan, 2011). However, the overconfident CEO believes that equity financing is even more
mispriced than risky debt. This is because the CEO’s perception of mispricing of equity is relevant for all possible situations and not only the default states (Malmendier, Tate and Yan, 2011). He is convinced of mispricing believing the stocks of his company are undervalued. Therefore he is unwilling to issue new equity because it will harm the current shareholders (Myers and Majluf, 1984), (Malmendier, Tate and Yan, 2011).
There is evidence that overconfidence among CEO’s causes underinvestment in recessions (Jaimovich and Rebelo, 2007). Next to that, overconfidence can cause lower firm performance during economic downturns (Mueller and Brettel, 2012). So different business cycles have an effect on investment decisions and firm performance in firms with overconfident CEO’s. This thesis examines if there exists also such an effect on the capital structure.
Hypothesis 2: The preferences of overconfident CEO’s to use less equity will be stronger during
economic slowdowns.
As an overconfident CEO overestimates the precision of the signal he gets, he will put too much weight on the signal (Jaimovich and Rebelo, 2007). During a recession or an economic downturn he will most likely overestimate the bad signal or values it more than a rational CEO does. According to Kruger and Burrus (2004) and Fuster, Hebert and Laibson (2010), beliefs are excessively optimistic in good times, but excessively pessimistic in bad circumstances. As discussed before, firms are
17
CEO’s may overvalue the bad signals and the economic contraction and substitute relatively more equity for debt than rational CEO’s. This probably enhances the general preference of overconfident CEO’s to use relatively less equity.
4. Methodology
As the current capital structure is an outcome of many different factors and probably multiple CEO’s and not only the current CEO, only new debt issues and new equity issues will be used to answer the research question. Financial firms and regulated utilities are excluded from the sample due to different capital structures. Fama and French (1992) argue that financial firms have a higher level of debt and that their leverage serves a different purpose than that of non-financial firms. Regulated utilities experience constraints to adjust their leverage structure through which it is difficult, if not impossible for the CEO to express his personal view on the financing decisions of the firm. This concerns companies with SIC codes 4900 to 4999 (utilities) and 6000 to 6999 (financial firms).
To determine if a CEO is overconfident the binary variable Longholder is used. To create the variable Longholder option moneyness is required and calculated in the following way, similarly to Campbell et al. (2011) and Hirshleifer, Low and Teoh (2012). The total realizable value of the
exercisable options is divided by the number of exercisable options, what will give the realizable value per option. To get the average exercise price of the options, the realizable value per option is
subtracted from the stock price. The average percent moneyness is then the realizable value per
option divided by the average exercise price of the options. If the CEO holds an option package that is
more than 67% in the money he is classified as overconfident as from that year and for the remaining of his tenure. The restriction for a CEO to be classified as overconfident is to show this behaviour at least twice during his tenure, following Malmendier and Tate (2005a). The 67% cut-off is determined by Malmendier and Tate (2005a) using a study of Hall and Murphy (2002) who studied executive stock option holding and exercise decisions. At the robustness section different cut-offs are discussed.
New equity issues are measured on the basis of an increase in common shares outstanding. Beforehand, common shares outstanding are adjusted for stock splits and stock dividends. If the increase in common shares outstanding is greater than 5% it is qualified as a new equity issue. The threshold of 5% is chosen to make sure the company experiences a significant equity issuance. This to account for errors, minor stock issues or stock remunerations that are not incorporated in the
18
adjustment factor. The binary variable new debt issues equals one if the increase in long term debt is greater than 5%. A hybrid issues is an issue of convertible debt or preferred stock and combines the elements of equity and debt. A hybrid issue is determined by an increase of at least 5% in
convertible debt and preferred stock.
The independent variable Longholder is used along with Stock Ownership and Vested Options in the regression analysis. Stock Ownership and Vested Options are included to control for the incentive effect and agency problems associated with the ownership in their own firms
(Malmendier and Tate, 2005a). They also mention the fact that delaying option exercise or
purchasing additional company stocks increases their holdings automatically. Stock Ownership and Vested Options are included as CEO controls and are expected to have no significant effect on equity issues (Malmendier, Tate and Yan, 2011).
The standard firm controls, Market to Book ratio (MtB), Profitability, Tangibility and the natural logarithm of Sales (Size) are included to capture the effect of the known factors for changes in leverage (Rajan and Zingales, 1995) (Frank and Goyal, 2009). The Market to Book ratio is included because firms with a high Market to Book ratio have a high proxy for growth and growth reduces leverage according to Frank and Goyal (2009). On the one hand, Profitability is added to the regression due to the effect of accumulated profits. On the other hand, firms with high profits face less costs of financial distress and value the tax shield more (Frank and Goyal, 2009). Weighting out both these influences, Profitability is expected to have a positive effect on new equity issues. Tangibility is included to control for the suggestion that firms with a have value of PPE have a higher leverage (Frank and Goyal, 2009). Therefore, the effect of Tangibility on equity issues is expected to be negative. Size, the natural logarithm of Sales, is expected to have a negative influence on new equity issues. Larger firms experience less financial distress costs and have thus a higher preference for debt (Frank and Goyal, 2009). Book leverage is used in the regression to control for systematic differences in the ability to attract new debt (Malmendier, Tate and Yan, 2011).
Next to the standard firm controls, return controls over the prior five years are added to control for past stock performance. Malmendier, Tate and Yan (2011) suggest that the measure of overconfidence used in this thesis may mix information about CEO characteristics with information about historic firm performance. They claim that good past performance of the firm can also cause CEO’s to hold options longer after they become exercisable. If return controls over the prior five years are included in the regression, the assumption that Longholder is a good variable to measure overconfidence is more reliable (Malmendier, Tate and Yan, 2011).
19
To test the first hypothesis, the effect of being overconfident as a CEO on equity issues, the following logit model is used. A logistic model is estimated with the use of Maximum Likelihood. The dependent variable equals one if there is a new equity issue under the condition that there is at least one public security issue by the firm in the specific year.
= + B1* + B2* + B3* B4* + B5* + B6* + B6* + B7* +
B7* + B8* + B9* + B10* +
For the examination of the second hypothesis the growth in the GDP of the United States is used. The dummy Low Growth is created and equals one if the GDP growth is lower than 2% in the specific year. This dummy is included in the main regression along with the interaction dummy of Longholder and Low Growth. The effect of Low Growth is expected to be negative on new equity issues. The reason for this is that firms tend to substitute equity for debt during economic downturns (Frank and Goyal, 2009). The interaction dummy is included in the regression to investigate if the effect of overconfidence on new equity issues is different during economic downturns. The prediction is that the interaction term has a negative sign. This means that overconfident CEO’s prefer to issue
relatively less equity in economic downturns. The S&P 500 (Market Index) is also included in the logit regression to control for the fluctuations of the stock market.
5. Data and descriptive statistics
The data is collected out of several databases for the sample period 1992-2012. The variables concerning the CEO characteristics and option packages to create the variable LONGHOLDER are collected from ExecuComp. Stock Ownership is shares owned (excluding options) divided by common shares outstanding. Vested Options is the number of unexercised exercisable options divided by common shares outstanding. This number is multiplied by 10 to ensure that the means of Stock Ownership and Vested Options are of the same magnitude (Malmendier, Tate and Yan, 2011). Observations with a value higher than one for either Stock Ownership or Vested Options are
dropped. Moneyness is calculated as described in the methodology section. Observations with missing values necessary to create the variable Moneyness are dropped.
20
Compustat is used to obtain all the necessary data about the public security issues and the firm specific data. Book Equity is total assets (at) minus total liabilities (lt) minus preferred stock liquidating value (pstkl) plus deferred taxes and investment tax credit (txditc) plus debt convertible (dcvt). Preferred stock liquidating value is replaced by the first available of preferred stock
redemption value (pstkrv), preferred stock total (prstk) or set zero if missing. Deferred taxes and investment tax credit and debt convertible are set to zero if missing. Book Debt is total assets (at) minus Book Equity. Market Equity is common shares outstanding (csho) times the stock price (prcc_f). Book Leverage is Book Debt divided by total assets (at). Market to Book ratio (MtB) is total assets (at) minus Book Equity plus Market Equity. Profitability is operating income before
depreciation (oibdp). Tangibility is property plant and equipment (ppent). All three variables are normalized by total assets. Size is the natural logarithm of total sales (sale). Net Equity Issues is the difference between Book Equity of year t and year t-1 minus the difference in retained earnings (re) of year t and year 1. Net Debt Issues is the difference between total assets (at) of year t and year t-1 minus the difference in Book Equity of year t and year t-t-1. Both Net Equity Issues and Net Debt Issues are normalized by total assets. Long term debt (dltt) and convertible debt and preferred stock (dcpstk) are replaced by zero if missing.
In Table I the summary statistics are displayed. The summary statistics are provided for the full sample and for the sample that includes only overconfident CEO’s separately. Panel A shows the Compustat data on the firm specific variables. When looking at the differences between the full sample and the Longholder sample, total assets are on average higher in the Longholder sample. There is also quite some variation as can be seen from the high standard deviation in both samples for total assets. This means that there are in general more overconfident CEO’s on the boards of firms with a high amount of assets. However, the difference in Size between the two groups is not that big. Still, the mean of Size is lower in the Longholder sample, but the means are located closer to each other. It is shown that Net Equity Issues are on average 0.08 in the full sample, facing 0.10 in the Longholder sample. This suggests that overconfident CEO’s issue on average more equity when looking at the differences in Book Equity divided by assets year by year. For Net Debt Issues there is no real difference visible between the two samples. Profitability and Tangibility are also quite similar between firms with rational and overconfident CEO’s. Only the Market to Book ratio is slightly higher for companies with an overconfident CEO.
21
Table Ι Summary Statistics
Net Equity Issues is the differences between Book Equity of year t and year t-1 minus the difference in Retained Earnings of year t and year t-1. Net Debt Issues is the differences between
Total Assets of year t and year t-1 minus the difference in Book Equity of year t and year t-1. Both Net Equity Issues and Net Debt Issues are normalized by assets. Profitability is operating income before depreciation, normalized by assets. Tangibility is Property, Plant and Equipment, normalized by assets. MtB (market to book ratio) is assets minus Book Equity plus Market Equity, normalized by assets. Size is the natural logarithm of Sales. Stock Ownership is the amount of shares owned by the CEO divided by Common Shares Outstanding. Vested Options is the amount of options owned by the CEO divided by Common Shares Outstanding. Vested Options are multiplied by 10 which ensures that the means of Stock Ownership and Vested Options are of the same magnitude. Longholder is a binary variable that equals one if the CEO at some point during his tenure held an option package that was more than 67% in the money. The Longholder dummy remains one for the rest of the tenure of the CEO from the moment it equals one. As a restriction the CEO must hold at least twice during his tenure an option package that was more than 67% in the money, otherwise the Longholder equals zero.
Panel A. Firm specific variables
Full Sample (Number of firms = 2143) Longholder Sample (Number of firms = 1513)
Variable Obs. Mean Median S.D. Min. Max. Obs. Mean Median S.D. Min. Max. Assets ($m) 9,219 6,539.85 1,227.31 26,567.32 3.43 795,337.00 5,596 5,045.33 1,072.72 17,814.17 3.43 437,006 Net Equity _Issues/Assets 9,166 0.09 0.02 0.21 -1.43 5.42 5,574 0.11 0.04 0.21 -.93 4.32 Net Debt _Issues/Assets 9,196 0.08 0.07 0.16 -2.66 0.78 5,588 0.08 0.07 0.17 -2.66 0.78 Profitability 9,198 0.13 0.14 0.12 -2.67 0.90 5,582 0.14 0.14 0.13 -2.67 0.90 Tangibility 9,192 0.30 0.23 0.23 0 0.98 5,575 0.28 0.21 0.23 0 0.98 MtB 9,219 2.17 1.61 2.22 0.25 78.56 5,596 2.44 1.76 2.62 0.25 78.56 Size 9,212 7.08 7.03 1.76 -2.72 13.00 5,592 6.92 6.90 1.73 -2.36 12.79
Panel B. CEO variables
Full Sample (Number of CEO’s = 3366) Longholder Sample (Number of CEO’s = 1762)
Variable Obs. Mean Median S.D. Min. Max. Obs. Mean Median S.D. Min. Max.
Age 9,183 52.30 52 7.36 25 83 5,583 52.46 52 7.49 27 83
Tenure 9,027 7.89 6 6.81 1 54 5,465 9.33 8 7.08 1 54
Stock Ownership 8,957 0.02 0.00 0.05 0 0.89 5,454 0.02 0.00 0.05 0 0.81 Vested Options 9,219 0.09 0.06 0.11 0.00 0.99 5,596 0.11 0.07 0.12 0.00 0.99
22
Panel B summarizes the characteristics of the CEO’s. As can be concluded from this table, roughly 52% of the CEO’s are classified as being overconfident. This means that more than half of the CEO’s hold at least two times during their tenure an option package that is more than 67% in the money. The average age of the CEO is the same for the two groups. This suggests that there is no indication that age is of influence on being overconfident or not. However, when looking at the tenure, there is a significant difference. The mean of overconfident CEO’s is 9.4 years compared to a mean of 8.1 years for the full sample. This is in line with the results of the study of Malmendier, Tate and Yan (2011) which even shows a greater difference in tenure. The amount of vested options owned by a CEO is higher in the Longholder sample compared to the full sample, with a mean of 10% and 8% respectively. This difference makes sense because CEO’s who are more optimistic about their future firm performance hold more options to capture the high prospects which they believe are going to happen.
Table II shows the growth rates of the GDP of the United States for all years in the sample period provided by the World Bank. For the years 2008 and 2009 the United States was in a recession, as the growth rates where negative in those years. The dummy Low Growth is created if the growth rate for the United States was lower than 2% in the specific year. As displayed in Table IV, this is the case for 6 out of 21 years of the total sample period. The threshold of 2% is chosen because in general economists agree the ideal GDP growth rate is more than 2%. Next to that, the average growth rate over the whole sample period is roughly 2.6%. Applying a threshold of 2% is significantly under the long term average and clearly indicating a low growth period.
Table II GDP Growth Rates
The growth in GDP of the U.S. measured in percentages for each year in the period 1992 – 2012.
Year GDP Growth (%) Year GDP Growth (%) 1992 3.56 % 2003 2.81 % 1993 2.74 % 2004 3.79 % 1994 4.04 % 2005 3.35 % 1995 2.72 % 2006 2.67 % 1996 3.80 % 2007 1.77 % 1997 4.48 % 2008 -0.26 % 1998 4.45 % 2009 -2.80 % 1999 4.79 % 2010 2.53 % 2000 4.09 % 2011 1.60 % 2001 0.98 % 2012 2.32 % 2002 1.78 % Average 2.63 %
23
The pairwise correlations of several firm and CEO characteristics along with the Longholder and growth measures are displayed in Table III. The correlation between Longholder and Equity Issue is positive and significant. This implies that there is a higher possibility that an overconfident CEO chooses to issue equity when doing a public security issue. The economic theory, Malmendier, Tate and Yan (2011) and the hypothesis of this thesis contradict this positive correlation and would expect a negative relation. They state that an overconfident CEO is less likely to issue equity
compared to a rational CEO. However, further examination and regression analysis in this thesis will determine if this correlations still holds when controlling for other factors. Also notable is the correlation between the age and tenure on the one hand and equity issues on the other hand. Both an increase in age as an increase in tenure of the CEO implies a significant lower willingness to issue equity. However, this negative correlation for age is much greater than for tenure. The Low Growth GDP factor has a positive significant correlation with the size of the firm, what implies that in a lower state of the economy firms become bigger in size. This seems illogical because if it the
macroeconomic circumstances worsen you would expect companies to experience a decrease in sales. The correlation between the Low Growth GDP factor and the Market Index is positive and significant, while the Market Index also is expected to decrease under low economic growth.
As discussed before, Malmendier, Tate and Yan (2011) argue return controls must be included in the regression analysis due to the correlation with the Longholder measure. Table IV shows the pairwise correlation between Longholder and the return controls over the prior five years. The correlation between Longholder and all five stock returns if positive and significant, what is in line with the findings of Malmendier, Tate and Yan (2011). This indicates that it is likely that the Longholder measure of overconfidence indeed mixes information about CEO characteristics with the historic performance of the firm. For this reason the prior five historic stock returns are included along the Longholder measure in the regression analysis.
24
Table ΙII
Correlation Main Variables
Correlations of all CEO, firm and Longholder specific variables for the sample period 1992 – 2012. Equity Issue is a variable that equals one if the increase in Common Shares Outstanding of year t and year t-1 is more than 5 percent. Common Shares Outstanding are adjusted for stock splits and stock dividends. Longholder is a binary variable that equals one if the CEO at some point during his tenure held an option package that was more than 67% in the money. The Longholder dummy remains one for the rest of the tenure of the CEO from the moment it equals one. As a restriction the CEO must hold at least twice during his tenure an option package that was more than 67% in the money, otherwise the Longholder equals zero. Stock Ownership is the amount of shares owned by the CEO divided by Common Shares Outstanding. Vested Options is the amount of options owned by the CEO divided by Common Shares Outstanding. Vested Options are multiplied by 10 which ensures that the means of Stock Ownership and Vested Options are of the same magnitude. MtB (market to book ratio) is assets minus Book Equity plus Market Equity, normalized by assets. Profitability is operating income before depreciation, normalized by assets. Tangibility is Property, Plant and Equipment, normalized by assets. Size is the natural logarithm of Sales. Market Index is the S&P 500 index. Low Growth is a binary variable that equals one if the growth of the GDP of the U.S. is less than 2%. Standard errors are clustered at the firm level. * indicates significant at a 10% level, ** indicates significant at a 5% level, *** indicates significant at a 1% level.
Equity Issue
Longholder Age Tenure Stock
Ownership
Vested Options
MtB Profitability Tangibility Size Book Leverage Market Index Longholder *LowGrowth Low Growth Equity Issue 1 Longholder 0.1073*** 1 Age −0.1400*** 0.0275*** 1 Tenure −0.0014 0.2624*** 0.3562*** 1 Stock Ownership 0.0397*** 0.0863*** 0.0324*** 0.2922*** 1 Vested Options 0.0821*** 0.1795*** −0.0256** 0.1726*** 0.1208*** 1 MtB 0.1816*** 0.1500*** −0.1457*** −0.0351*** 0.0298*** 0.0381*** 1 Profitability −0.1892*** 0.0465*** 0.0414*** 0.0173 −0.0064 −0.1050*** −0.0315*** 1 Tangibility −0.1295*** −0.0780*** 0.0840*** 0.0088 −0.0223** −0.1141*** −0.1744*** 0.1426*** 1 Size −0.3078*** −0.1115*** 0.2086*** −0.0369*** −0.1613*** −0.3075*** −0.2286*** 0.3206*** 0.0943*** 1 Book Leverage −0.2470*** −0.1578*** 0.1433*** −0.0677*** −0.0414*** −0.0956*** −0.2365*** 0.0217** 0.1134*** 0.4673*** 1 Market Index −0.0816*** 0.0828*** 0.0060 −0.0089 −0.0424*** 0.0551*** 0.0705*** −0.0094 −0.1281*** 0.1439*** −0.0284*** 1 Longholder*Low _Growth −0.0255** 0.3359*** 0.0439*** 0.1316*** −0.0156 0.0552*** −0.0456*** 0.0000 −0.0056 0.0446*** −0.0541*** 0.1069*** 1 Low Growth −0.0803*** -0.0051 0.0251** 0.0148 −0.0287*** -0.0012 −0.0845*** −0.0177* −0.0200* 0.1111*** 0.0023 0.1532*** 0.7303*** 1
25
Table ΙV
Correlation Return Controls
Correlation of Longholder and historic stock returns for the sample period 1992 – 2012. Longholder is a binary variable that equals one if the CEO at some point during his tenure held an option package that was more than 67% in the money. The Longholder dummy remains one for the rest of the tenure of the CEO from the moment it equals one. As a restriction the CEO must hold at least twice during his tenure an option package that was more than 67% in the money, otherwise the Longholder equals zero. Returnst are the natural logarithm of one plus annual stock returns (excluding dividends) from year
t-1 to t. * indicates significant at a 10% level, ** indicates significant at a 5% level, *** indicates significant at a 1% level.
Longholder Returnst-1 Returnst-2 Returnst-3 Returnst-4 Returnst-5
Longholder 1 Returnst-1 0.0897*** 1 Returnst-2 0.1010*** -0.1826*** 1 Returnst-3 0.0824*** -0.1056*** -0.1335*** 1 Returnst-4 0.0755*** -0.0834*** -0.1154*** -0.1589*** 1 Returnst-5 0.0773*** -0.0029 -0.0600*** -0.1425*** -0.1095*** 1 6. Empirical results
In Table V the frequencies of all different public security issues are displayed. Equity Issues are issues of common or preferred stock. Debt Issues are issues of long term debt and Hybrid Issues are issues of convertible debt or preferred stock. Longholder CEO’s issue 42% of the time equity when doing a public security. If the CEO is not classified as Longholder, he issues equity in 32% of the observation years. So overconfident CEO’s choose, with a difference of 10%, more often for equity issues than rational CEO’s. Overconfident CEO’s issue new debt in 74% of the firm years with public issues, contrary to rational CEO’s who make in 79% in the firm years debt issues. As a result, it can be concluded that overconfident CEO’s choose more often for equity and less often for debt issues compared to CEO’s with rational beliefs. There are no significant differences for new hybrid issues. These findings are not in line with the findings of Malmendier, Tate and Yan (2011). They find evidence that overconfident CEO’s have a preference for debt instead of equity.
In the second part of Table V the frequencies of the public securities are shown for years where the Low Growth factor equals one. That is for the years in which the growth of the GDP of the United States is below 2%. From these results can be concluded that overconfident CEO’s choose in 35% of the firm years for an equity issue, while rational CEO’s choose a new equity issue in 25% of the firm years. This is roughly the same difference for Longholder and non Longholder CEO’s as in the full sample. Next to that, an overconfident CEO chooses again less often to issue new debt compared to his peers. These findings imply that in Low Growth years the same relation regarding
26
Table V Public Issues
Longholder is a binary variable that equals one if the CEO at some point during his tenure held an option package that was
more than 67% in the money. The Longholder dummy remains one for the rest of the tenure of the CEO from the moment it equals one. As a restriction the CEO must hold at least twice during his tenure an option package that was more than 67% in the money, otherwise the Longholder equals zero. Equity Issue is a variable that equals one if the increase in Common Shares Outstanding of year t and year t-1 is more than 5%. Common Shares Outstanding are adjusted for stock splits and stock dividends. Debt Issue is a variable that equals one if the increase in Long Term Debt of year t and year t-1 is more than 5%. Hybrid Issue is a variable that equals one if the increase in Convertible Debt and Preferred Stock of year t and year t-1 is more than 5%. Only observation years with at least one public security issue are used.
Percent of issue years with
Equity Issues Debt Issues Hybrid Issues Full Sample Longholder = 0 31.6% 78.7% 12.8%
Longholder = 1 42.2% 74.1% 13.5%
Low Growth Longholder = 0 25.4% 81.8% 13.1% Longholder = 1 35.1% 76.1% 14.3%
the choice of public security issues exist between overconfident and rational CEO’s. However, the tendency for all firms is to issue less equity and prefer new debt issues than in years with a GDP growth of more than 2%. This is in line with the hypothesis stated before and the study of Frank and Goyal (2009). They claim that in economic contractions relatively more firms prefer to issue debt than in the expansion period of the business cycle.
In Table VI the results of the logit regressions to test the hypotheses are shown. The
dependent variable is a dummy that equals one if the firm issues at least one equity issue during the firm year under the condition that there is at least one public security issue that year. The
independent variable Longholder equals one is the CEO exhibits overconfident behaviour. The effect of Longholder on new equity issues is examined in column 1. Stock Ownership and Vested Options are included in the regression to control for agency problems (Malmendier and Tate, 2005a). The coefficient of Longholder is positive and significant. If a CEO is classified as being overconfident he is 49% more likely to issue equity than his peers. The percentage of stock and the number of vested options a CEO holds both have a significant and positive influence on new equity issues. In column 2 the standard firm controls are included in the regression. This concerns Market to Book ratio, Profitability, Tangibility, Size and Book Leverage. Including the standard firm controls lowers the effect of Longholder on Equity Issue slightly. All standard firm controls have a significant influence on the possibility of using equity financing. In column 3 the Return Controls over the prior five years are added to control for the influence of historic stock returns in the Longholder variable (Malmendier,
27
Tate and Yan, 2011). Adding the Return Controls seems to have little effect on all variables, except for Stock Ownership. Stock Ownership has a negative and significant effect on Equity Issue when including the Return Controls.
Table VI
Logit Regressions on Equity Issue
Equity Issue is a variable that equals one if the increase in Common Shares Outstanding of year t and year t-1 is more than
5%. Longholder is a binary variable that equals one if the CEO at some point during his tenure held an option package that was more than 67% in the money. The Longholder dummy remains one for the rest of the tenure of the CEO from the moment it equals one. As a restriction the CEO must hold at least twice during his tenure an option package that was more than 67% in the money, otherwise the Longholder equals zero. Stock Ownership is the amount of shares owned by the CEO divided by Common Shares Outstanding. Vested Options is the amount of options owned by the CEO divided by Common Shares Outstanding. Vested Options are multiplied by 10 which ensures that the means of Stock Ownership and Vested Options are of the same magnitude. MtB (market to book ratio) is assets minus Book Equity plus Market Equity, normalized by assets. Profitability is operating income before depreciation, normalized by assets. Tangibility is Property, Plant and Equipment, normalized by assets. Size is the natural logarithm of Sales. Market Index is the S&P 500 index. Low Growth is a binary variable that equals one if the growth of the GDP of the U.S. is less than 2%. Z-statistics are reported in parentheses and standard errors are clustered at the firm level. * indicates significant at a 10% level, ** indicates significant at a 5% level, *** indicates significant at a 1% level.
(1) (2) (3) (4) Longholder 0.398*** (7.29) 0.263*** (4.58) 0.267*** (4.11) 0.284*** (3.90) Stock Ownership 1.039* (1.66) -0.179 (-0.33) -1.834*** (-2.59) -2.032*** (-2.84) Vested Options 1.108*** (4.39) -0.426 (-1.61) -0.355 (-1.21) -0.124 (-0.44) MtB 0.213*** (7.90) 0.162*** (5.33) 0.167*** (5.41) Profitability -4.189*** (-9.37) -4.965*** (-9.28) -5.267*** (-9.55) Tangibility -0.404*** (-2.93) -0.226 (-1.48) -0.330** (-2.09) Size -0.244*** (-10.78) -0.203*** (-8.02) -0.169*** (-6.47) Book Leverage -1.280*** (-7.81) -1.331*** (-6.72) -1.533*** (-8.05) Return Controls √ √ Market Index -0.001** (-6.45) Longholder * Low _Growth 0.048 (0.37) Low Growth -0.279*** (-2.71) Observations 8,957 8,909 7,311 7,311 Number of Firms 2,119 2,111 1,779 1,779 Pseudo R2 0.012 0.120 0.102 0.110
28
The first hypothesis can be answered by looking at the results in Column 3. When all variables are added the sample for the logit regression contains 7,311 firm year observations and 1,779 firms. The Longholder variable is still positive and significant after including all controls. Overconfidence among CEO’s improves the chance of issuing equity with 31%. These results confirm the findings in the raw data as earlier described in this paper. Therefore, there exists an effect of overconfidence on equity issues. However, the effect is opposite to the hypothesis stated before. Overconfident CEO’s choose not less, but more often equity financing than rational CEO’s. This contradicts the economic theory, which in general claims that overconfident CEO’s prefer debt over equity. One of the main articles in this area, the paper of Malmendier, Tate and Yan (2011), also finds clear evidence of the preference for debt. A possible explanation for the different findings of the paper of Malmendier, Tate and Yan (2011) and this paper is the methodology. This thesis uses a different measure of overconfidence and a different method to detect new equity issues.
Furthermore, a more recent and longer time period is used in this research compared to the research of Malmendier, Tate and Yan (2011).
To test the second hypothesis, the coefficients of the dummy Low Growth and the interaction term of Longholder and Low Growth are included in the regression as displayed in column 4 of Table VI. Next to that, the S&P 500 is included as market index to control for the volatility of the stock market. The coefficient of Low Growth is negative and significant at the 1% level. This implies that in years with a GDP growth of less than 2% firms issue significantly less equity. The prediction that Low Growth has a negative effect on new equity issues is thereby confirmed. This suggests that in economic downturns firms substitute equity for debt. These findings are in line with the paper of Frank and Goyal (2009), in which they expect firms to substitute equity for debt during economic recessions. The coefficient of the interaction term is positive, but insignificant. So, it is not possible to draw any conclusion from this result. Therefore, the second hypothesis, the preferences of overconfident CEO’s to issue less equity will be stronger during economic slowdowns, cannot be accepted. On the one hand, there is no preference among overconfident CEO’s for debt, but for equity in general. On the other hand there is no evidence that the preferences of
overconfident CEO’s regarding financing decisions will be stronger or weaker during economic contractions. As discussed before, the influence of overconfidence on new equity issues is positive. Furthermore, the direct effect of Low Growth on new equity issues is negative. If the interaction term of Low Growth and Longholder would be significant, this implies that the positive influence of overconfidence on new equity issues is different during periods of low growth. The sign of the interaction term in Table VI is positive. This would mean, if it would be significant, that a CEO issues relatively more equity in a low growth year if he is marked as overconfident compared to a year with
29
a GDP growth of more than 2%. In the next section several robustness checks will be implemented to test the reliability of the main results in this thesis.
When looking at the control variables, all variables are significant except for Vested Options. Stock Ownership has a negative coefficient, which implies that the more shares a CEO possesses, the less chance he chooses to issue equity when doing a public security issue. This result is similar to the findings of Malmendier, Tate and Yan (2011). The effect of the Market to Book ratio on new equity issues is positive, as expected. This implies that companies with high Market to Book ratios are more likely to issue equity. This supports the findings in the study of Baker and Wurgler (2002), who suggest that the Market to Book ratio is important when investigating the financing decisions of firms. Profitability has a negative influence on Equity Issue, demonstrating that firms with high profits are more reluctant to issue equity and prefer debt more that the average company. This is contrary to the predictions earlier. However, as already discusses before, there are also arguments for a positive relationship between Profitability and leverage. The reason for this is that profitable firms experience less financial distress costs and value the tax shield more (Frank and Goyal, 2009). Tangibility seems to negatively predict equity issues. So, firms with a high value of PPE tend to issue less equity. This is in line with the economic theory, that more tangible assets reduce the costs of financial distress (Frank and Goyal, 2009). Size has a negative influence on new equity issues, which means that bigger companies prefer relatively less often equity. The reason for this effect perhaps lies in the fact that it can be easier for large firms to realize debt contracts than for smaller firms (Frank and Goyal, 2009). Book Leverage also negatively predicts equity issues, implying that firms that are already relatively high leveraged still prefer debt over equity in attracting new financing. The effect of the S&P 500 is negatively and significant, but very small in magnitude. This suggests that if the market index goes up firms prefer slightly more debt over equity.
In Table VII the results of a logistic regression on new debt issues are presented. The exact same logit regression is used as in Table VII, only Equity Issue is replace by Debt Issue. Debt Issue is a binary variable that equals one if the firm does at least one new debt issue during the firm year under the condition that the firm does at least one public security issue in the concerning year. As concluded before, Longholder firms issue more relatively more equity and therefore the question arises if those firms also issue significantly less debt. The suggestion is that firms with overconfident CEO’s indeed issue less debt as can be seen in Table V. In column 1 of Table VII the effect of
Longholder on new debt issues is negative and significant. However, only the controls Stock Ownership and Vested Options are included in the regression. When adding more controls, the effect becomes questionable and insignificant. In column 3 a logistic regression is shown, including
