1
CEO
OVERCONFIDENCE AND
LEVERAGE DURING THE FINANCIAL
CRISIS
:
E
VIDENCE FROM
U.S.
LISTED
FIRMS
Abstract
In this paper, I examine whether overconfident CEOs increase leverage during financial crises. I also examine what the consequences are if they do so. Using longitudinal study and statistical analysis, I find that CEO overconfidence increases the amount of leverage during a financial crisis without sacrificing profitability. My results appear to be sensitive to how leverage is defined, however this is not the case for profitability. Hence, I cannot conclude that overconfident CEOs borrow more during a crisis, but if they do so, it does not affect the firm performance negatively.
Fareed Alam UvA ID: 11103000
Supervisor: ShiveshChangoer
Faculty of Economics and Business Bachelor Thesis Economics and Business Specialization: Economics and Finance University of Amsterdam
2
Statement of Originality
This document is written by Fareed Alam, who declares to take full responsibility of the content of this document.
I declare that the text 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 content.
3 1. Introduction
During a financial crisis, debt levels are high, leading to higher default risk (Kumhof et al., 2015). According to Vassalou and Xing (2004), default risk is defined as the failure of an entity to service its debt obligations. Kumhof et al. (2015) suggest that further increasing leverage may result in unsustainable levels of debt and thus extremely high probability of default. Banks may also apply stricter policies to lending such as increased number of covenants. Therefore, I would expect CEOs to not increase debt levels during times of financial distress. Overconfident CEOs, however, may do so. The reason is that they may believe they can fulfill the criteria set by banks and that their firm would be able to perform well after the crisis.
In this paper, I examine whether overconfident CEOs increase leverage during financial crises. I also examine what impact such a decision can have on firm performance. My analysis is based on a sample of firms listed in the U.S. during a time period of 11 years from 2005 until 2015.
My results suggest that firms with overconfident CEOs borrow more money during a financial crisis in comparison to non-overconfidence CEOs. However, contrary to my expectations, firms that do so perform better in the post-crisis period. This is an interesting finding since most studies (see Ho et al., 2016) suggest that the actions of overconfident CEOs actually have negative consequences and destroy value. My results, however, are not robust in relation to changing the debt measure as it does not give the same results as the original analysis. I also test the robustness of the results by changing the profitability measure. In this case, the implications of the original analysis hold.
My research adds to the studies that examine the effect of a financial crisis on leverage (see Fosberg, 2012; Tsoy and Heshmanti, 2017). For instance, Fosberg (2012) observes a significant increase in debt levels, on average 5.1%, after controlling for other factors during the crisis. Additionally, Tsoy and Heshmati (2017) conduct a similar analysis for listed firms in Korea during the period 1985 to 2005. They find that for the 2007- 08 financial crisis, the debt levels increased significantly. Most studies (see Ho et al., 2016) find that management overconfidence leads to higher leverage levels. I contribute to the literature by providing an explanation for the increased leverage during the crisis. In order to achieve this, I incorporate personal CEO characteristics such as cognitive biases. Hence, this explanation is based on the overconfidence literature.
Prior research on overconfidence suggests that this cognitive bias can have a significant impact on corporate decisions (see Ben-David et al., 2007; Malmendier and Tate, 2005). According to Ben-David et al. (2007), overconfident executives are miscalibrated, meaning they form narrow intervals regarding their forecasts - hence, their predictions are inaccurate, resulting in decisions that may not maximize value. Moreover, Malmendier and Tate (2005) suggest that overconfident CEOs have an upward bias when estimating their ability to earn profits, thus making decisions that destroy firm value.
4 My paper is closely related to the work of Ho et al. (2016). They examine the impact of CEO overconfidence on the leverage decisions during a financial crisis and investigate the consequences for long-term performance. They observe that overconfident CEOs use more debt financing when compared to non-overconfident CEOs. Ho et al. (2016) also outline that the crisis already has a negative impact on profitability, with higher leverage further eroding profits in interest payments.
The main difference between my research and that of Ho et al. (2016) is that their analysis is from the perspective of banks. The contribution of this paper is that a similar analysis is done from the point of view of firms. It is important to focus on firms because banks and firms have different financial structures, and therefore, conclusions may differ. Also, banking is part of the service sector of the economy. It will be interesting to extend the analysis to other sectors such as manufacturing.
The remainder of the paper is organized as follows. Section 2 presents the literature review and discusses the financial crisis, leverage, overconfidence and other theories that explain the capital structure. In section 3, I discuss prior studies and develop my hypotheses. Section 4 describes the research design. In section 5, I discuss the results from univariate and bivariate analyses. Section 6 contains the sensitivity analysis. I conclude the paper in section 7.
2. Literature Review
The capital structure of a firm represents how firms choose between debt, equity or hybrid securities (Myers, 1984). Researchers have developed several theories to explain the capital structure of firms (see Modigliani and Miller, 1958; Jensen and Meckling, 1976; Myers, 1984). The most renowned are Capital Structure Theory, Agency Theory, Pecking Order Theory and Trade-Off Theory.
2.1.1 Capital Structure Theory
The capital structure irrelevance theory was developed by Modigliani and Miller (1958). According to this theory “the market value of a firm is independent of the capital structure and is given by capitalizing its expected return” (Modigliani, &, Miller, 1958) because if firm value was actually dependent on the relative proportions of debt and equity, investors would be able to benefit from arbitrage opportunities through trading between the cash flow streams.
A major limitation of this theory from Modigliani and Miller (1958) is that it is based on a number of unrealistic assumptions such as perfect competition, no taxes, no arbitrage possibilities and the absence of several costs, such as transaction costs, bankruptcy costs and agency costs. It is important to note that when some of these assumptions are relaxed, the value of the firms is no longer independent of capital structure. For instance, Modigliani and Miller (1958) suggest that when income is taxed the value of the firm increases with debt due to the fact that interest is tax deductible.
Although the capital structure irrelevance theory is based on unrealistic assumptions it gave birth to discussions surrounding capital structure and corporate finance theory, leading to the development of other major theories.
5 2.1.2 Agency Theory
Agency theory is based on the idea that, within each firm, there are there are “two (or more) parties: principals and agents, where principals are shareholders and agents (managers) are those appointed by the principal to take responsibility for the functioning of the firm, generating profits on their behalf (Jensen and Meckling, 1976).
According to this theory, agents are utility maximizers (Ross, 1973). Hence, they increase or reduce leverage only if such an action increases their expected value or utility (Ross, 1973). Managers may have an incentive to increase leverage in the case when increasing leverage is a value enhancing action (Kochhar, 1996). This is because it increases the value of their restricted stock and stock options which in turn increases their wealth (Cho, 1998). On the contrary, the managers will only reduce leverage when they believe that high levels of debt can result in increased probability of financial distress (Kochhar, 1996). Kochhar, (1996) explains that this is based on the reasoning that increased chances of financial distress result in higher default risk which can lead to the manager being fired.
However, there are certain limitations of the agency theory such as it being based on the interaction of two parties only. Also, Arthurs and Busenitz (2003) suggest that the variables used to test the agency theory need to be improved. They further state that even when there is a partial ownership stake of managers, they may still not behave in a manner the owner would.
2.1.3 Pecking Order Theory
The Pecking Order Theory, which was initially presented by Myers (1984), suggests that firms prefer internal funds such as retained earnings over external finances and debt over equity. The reason is that if firms issue stock, investors will assume that the stock is overvalued and hence no investor would be willing to buy (Myers and Majluf, 1984). As a consequence, the stock price will drop significantly due to information asymmetries (Myers and Majluf, 1984). To avoid such an undesirable situation, firms will first exhaust all internal funds, then issue debt - if the need for finances still remains, they issue equity (Myers, 1984). Myers (1984) further suggests that one major implication of this theory is that it cannot explain why firms may converge to a specific leverage ratio. 2.1.4 Trade-Off Theory
Contrary to the Pecking Order Theory, the Trade-off Theory results in an optimal capital structure (Tong, & Green, 2005). This theory posits an optimal debt ratio which is based on the trade-off between the benefits and costs of debt, with other things constant (Myers, 1984). The company tries to balance the value generated by the interest tax shield against the costs of financial distress or bankruptcy that are a result of borrowing (Myers, 1984).
Myers (1984), suggests that firms should keep substituting between equity and debt until they find a combination that results in the highest firm value. The substitution between equity and debt hence results in the trade-off (Myers, 1984). A graphical representation of the theory taken from Myers (1984) is in the appendix as Graph A2.
6 2.2 Effect of firm-specific characteristics on leverage
Prior research has tested the above-mentioned theories by examining the effect of firm-specific characteristics on leverage (see Myers, 1984; Titman and Wessels, 1988; Harris and Raviv, 1991; Rajan and Zingales, 1995). Titman and Wessels (1988) examine the effect of profitability on the leverage ratio. Rajan and Zingales (1995) investigate the relation of asset tangibility on the debt level. Harris and Raviv (1991) study the impact of growth opportunities on the leverage ratio whereas Wald (1999) analyses what effect firm size has on the amount of leverage a firm takes.
Titman and Wessels (1988) examine a sample of 469 firms in the U.S. during 1974 and 1982. They find a positive relation between asset tangibility and leverage ratio. They suggest that this is because tangible assets can be used as collateral to borrow money. This explanation is based on prior research which suggests that firms prefer secured debt over unsecured debt (see Myers, 1984). Therefore, this supports the positive relation between asset tangibility and leverage. Two other studies (Myers and Majluf, 1984; Harris and Raviv, 1991) also agree with this idea as agency costs of debt are reduced with higher collateral value.
Examining the same sample, Titman and Wessels further (1988) find that profitability is negatively related to the amount of debt issued. This finding is consistent with Myers (1984) and Rajan and Zingales (1995), and also in line with the Pecking Order Theory which states that firms prefer to use internally generated funds such as retained earnings over external sources, and debt over equity amongst the external sources.
Titman and Wessels (1988) also find a negative relation between growth opportunities, measured by the market-to-book value ratio, and the leverage ratio. They argue that firms with high growth opportunities act in an opportunistic and flexible manner with regards to their investment decisions. This can be linked to the agency theory as this type of investment behavior can increase agency costs. This is consistent with the findings of Wald (1999).
Rajan and Zingales (1995) examine the effect of firm size on leverage using a sample of 2583 firms based in the G-7 countries from 1987 to 1991. They find a positive relation between firm size and leverage. According to Rajan and Zingales (1995), this finding is expected because large firms are less likely to default, and should, therefore, issue more debt. Also, large firms have access to more debt due to fewer information asymmetries. The model presented by Harris and Raviv (1991) also supports the positive relation between leverage and firm size.
2.3 Effect of individual characteristics on leverage
Faccio et al., (2016) examine the impact of gender on the amount of leverage firms maintain. They find that firms run by female executives have lower debt levels and attribute this to the difference in attitude between the two genders towards risk. Female executives are comparatively more risk-averse, hence preferring lower debt levels in order to avoid financial distress.
Similarly, Serfling (2014) studies the effect of CEO age on the amount of leverage a firm takes. He finds that CEO age is negatively related to firm leverage, further explaining that the reason
7 for this is the higher risk-aversion of CEOs who are old. With age, CEOs prefer lower leverage to prevent financial distress caused by high level of debt.
Overconfidence
Overconfidence is a bias that results in the overestimation of the positive state (Ben-David et al., 2007). Such a bias can have strong effects on the financial decision of firms. The reason is that executives or managers who are overconfident forecast probability distributions that are extremely narrow and believe that randomly occurring events are very unlikely. They, therefore, predict the future to be favorable, believing that they possess correct knowledge about future events (Hackbarth, 2008).
Researchers have found that overconfidence can have a strong effect on the financial decisions of firms. For instance, Malmendier and Tate (2005) examine the effect of overconfidence on corporate investment decisions. Their study is based on the psychology literature. Building on this literature, they find that overconfident executives are overoptimistic and have the tendency to overestimate their skill in comparison to others. Malmendier and Tate (2005) also outline that overconfidence leads executives to believe that successful outcomes are a result of their actions and associate negative outcomes with misfortune, having an upward bias in estimating their returns on investments due to forecasting lower discount rates for their projects. They predict that such characteristics can negatively affect the firm. The findings of Ho et al. (2016) support this prediction.
Prior research has examined the impact of overconfidence on capital structure decisions. Overconfident executives underestimate the discount rate for their projects. The impact of the cognitive overconfidence has been researched in literature. Chen and Chen (2009) examine this relation and conclude their empirical research by asserting that CEO overconfidence has a significant positive impact on the on the leverage level. This is because executives who are overconfident tend to underestimate the chances of default and thus choose capital structures that consist of high debt levels (see Shefrin, 2001; Hackbarth, 2009).
3. Hypotheses
In this paper, I predict that overconfident CEOs will increase leverage during a financial crisis. This prediction is based on the observation that during a financial crisis, firms become credit constrained and internal funds are scarce due to lower profitability. The consequence is that firms require external funds for operations. These external funds are acquired through borrowing money, because equity issuance will signal to investors that the firm is overvalued and the stock price will drop. Such a scenario will be undesirable for CEOs as their stock options will become out of the money and lose value. The problem with borrowing during a financial crisis, however, is that banks increase the number of covenants and impose stricter conditions (Ivashina and Scharfstein, 2010). These covenants and conditions may be costly for the firm in the long-run and prevent it from taking first best decisions and maximizing. Overconfident CEOs are expected to accept such debt agreements due to their belief that they can overcome the negative consequences. Therefore,
non-8 overconfident CEOs will avoid increasing leverage, but overconfident CEOs may not do so. The reason is that they believe that they can meet the covenants and fulfill the conditions set by banks. This is because they underestimate the probability of the firm going bankrupt and have overoptimistic beliefs about the future success of their firm (Malmendier and Tate, 2005). Formally stated, the first hypothesis is:
H1: Firms with overconfident CEOs tend to increase leverage during the crisis.
My second prediction is that CEO overconfidence will have negative long-term consequences for the firm. I make this prediction based on the reasoning that increasing leverage during a crisis may be costly as firms may suffer from working capital problems, profitability may decrease and a fall in the stock price may be observed. Also, reviewing prior studies, there is no evident reason to believe that overconfident CEOs create more firm value than other CEOs (see Ho et al., 2016)
H2: Firms with increased leverage during crisis period observe negative long-term consequences on firm performance.
4. Research Design 4.1 Method
4.1.1 Testing H1:
To test my first hypothesis, I estimate the following model:
LEVERAGE1 = α + β1*OVERCONFIDENCE + β2*CRISIS +
β3*OVERCONFIDENCE*CRISIS +β4*ASSET TANGIBILITY + β5*PROFITABILITY + β6*FIRM SIZE + β7*GROWTH
OPPORTUNITIES + ε
(1)
In the above model, LEVERAGE1 is the dependent variable and is calculated by the ratio of total debt to total assets. I focus on total debt in order to capture the impact of both short-term and long-term debt. The dependent variable OVERCONFIDDENCE is defined the same way as in Malmendier and Tate (2005) and Ho et al. (2016). It is equal to 1 if the MONEYNESS is greater than 100% and zero otherwise.
MONEYESS =
𝑆ℎ𝑎𝑟𝑒 𝑝𝑟𝑖𝑐𝑒 𝑎𝑡 𝑦𝑒𝑎𝑟 𝑒𝑛𝑑
𝑆ℎ𝑎𝑟𝑒 𝑝𝑟𝑖𝑐𝑒 𝑎𝑡 𝑦𝑒𝑎𝑟 𝑒𝑛𝑑 − 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝑒𝑥𝑒𝑟𝑐𝑖𝑠𝑎𝑏𝑙𝑒 𝑜𝑝𝑡𝑖𝑜𝑛𝑠 𝑛𝑜𝑡 𝑒𝑥𝑒𝑟𝑐𝑖𝑠𝑒𝑑
9 CRISIS is a dummy that equals 1 for the years 2008 and 2009, and zero otherwise. The implicit assumption is that the crisis began in December 2007 and ended in the last quarter of 2009 as suggested by the NBER (2010).
OVERCONFIDENCE*CRISIS is the interaction between OVERCONFIDENCE and CRISIS. This interaction variable is the variable of my interest. If its coefficient is significantly positive, then this is evidence that overconfident CEOs increase leverage during the crisis – thus, hypothesis 1 will not rejected.
ASSET TANGIBILITY is the ratio of tangible assets at the end of the year to total assets at the end of the year (Ozkan, 2001; Rajan and Zingales, 1995). I include this variable because, according to Psillaki and Daskalakis (2009), it is an important determinant of capital structure and is positively related to leverage.
PROFITABILITY is measured by the ratio of current year’s net income to book value of total assets at the end of the year as defined by Titman and Wessels (1988) and Rajan and Zingales (1995). I include this variable because Rajan and Zingales (1995) observe a negative relation between profitability and the leverage of the firm. The Pecking Order Theory explains the negative correlation as firms with higher profits will utilize the internal funds first and require less external financing (Titman and Wessels, 1988).
FIRM SIZE is natural logarithm of current year’s total revenue, thus following Rajan and Zingales (1995). I include this variable because according to Titman and Wessels (1988), larger firms can operate with high leverage levels as they face lower risks such as bankruptcy.
GROWTH OPPORTUNITIES is the market value at the end of the year divided by the corresponding book value of the firm. I include this proxy for growth opportunities because Rajan and Zingales (1995) suggest a negative relation between leverage and growth opportunities due to the latter being associated with high distress costs.
4.1.2 Testing H2:
To test H2, I run the following model on the observations from 2010 to 2015 to represent the post-crisis period.
PROFITABILITY = α + β1*OVERCONFIDENCE + β2*INCREASE +
β3*OVERCONFIDENCE*INCREASE + β4*FIRM SIZE + β5*GROWTH OPPORTUNITIES+ β6*LEVERAGE1 + ε
(3)
This analysis is based on the years 2010 – 2015 to represent the post-crisis period. The dependent variable profitability is the return on assets and has been described above. The variable increase is a dummy and is equal to 1 if leverage was increased during the particular year. Other variables such as overconfidence, firm size, growth opportunities, and leverage are the same as explained above. I will focus on the coefficient of the interaction variable overconfidence*increase. If
10 the coefficient β3 is negative, it will suggest negative consequences on firm performance due to
increased leverage. Thus, the hypothesis will not be rejected. 4.2 Model Estimation
I estimate all regression models using the ordinary least squares. By doing so, I assume a linear relation and that the Gauss-Markov conditions hold i.e., the expected mean value of residuals is zero, absence of heteroscedasticity and there is no serial autocorrelation or multicollinearity. To control for heteroscedasticity, I cluster the standard errors by company and year.
Sample
The country under consideration for this research is the U.S. I focus on this country because relatively large and relevant amount of data is available. My sample period spans from 2005 to 2015. I exclude all years before 2005 in order to eliminate the effects of any other shock such as the dot-com crisis.
My sample period includes the financial crisis that started at the end of 2007 and lasted until 2009. The origins of this financial crisis come from the defaulting of subprime mortgages (Ivashina and Scharfstein, 2010). Loans were given to uncreditworthy borrowers who failed to pay back the borrowed money. As a result, large firms like Lehman Brothers and Washington Mutual went bankrupt since they were no longer able to pay their debt. Therefore, Ivashina and Scharfstein (2010) suggest that banks had to write-off loans causing large losses in the banking sector. Hence, banks had to borrow significantly more and the financial markets became extremely volatile, reaching levels rarely witnessed before (Ivashina, &, Scharfstein, 2010). The panic, originating in the U.S., eventually caused countries around the globe to enter into a severe crisis (Ivashina, &, Scharfstein, 2010). Lane (2012) suggests that economies with high exposure to the U.S. were severely affected. As an example he mentions the European Union, where this crisis had a significant impact and eventually resulted in the European sovereign debt crisis.
4.3 Data
To obtain data for my analyses, I use the widely known data sources Compustat Capital IQ and Execucomp, accessed through Wharton Research Data Services. Compustat allows me to collect data on company financials such as debt, assets, profitability, revenue and market value. Execucomp provides data for executive compensation relating to stock options. The databases initially contained more than 60,000 observations.
After collecting the requisite data, I first drop all observations with missing variables in both databases. I then merge the two databases using the company ID and fiscal year as the identifiers. The resulting sample comprises of 4063 observations with 550 unique firms.
The sample suggests that most firms have moderate debt ratios. The assets of the firms are mostly tangible. Also, the firms are relatively of similar size as suggested by the standard deviation. Profitability varies across firms and there are some firms that are highly unprofitable. Many firms also display prospects for growth opportunities.
11 Table 1: Summary statistics for dependent and independent variables
Variable Mean Std. Dev. Min P25 Median P75 Max
LEVERAGE1 0.252 1.902 0.000 0.100 0.202 0.316 120.943 LEVERAGE2 0.206 0.173 0.000 0.085 0.186 0.293 1.871 ASSET TANGIBILITY 0.785 0.186 0.182 0.658 0.824 0.950 1.000 PROFITABILITY 0.038 0.244 -13.670 0.021 0.054 0.088 1.626 FIRM SIZE 8.028 1.681 -3.540 6.951 7.954 9.163 12.449 GROWTH OPPORTUNITIES 1.168 0.901 0.001 0.590 0.945 1.493 10.532 EBITASSETS 0.113 1.431 -1.375 0.055 0.091 0.137 91.127 OVERCONFIDENCE 0.884 0.321 0.000 1.000 1.000 1.000 1.000 CRISIS 0.193 0.394 0.000 0.000 0.000 0.000 1.000 OVERCONFIDENCE *CRISIS 0.143 0.351 0.000 0.000 0.000 0.000 1.000 CRISIS2 0.285 0.451 0.000 0.000 0.000 1.000 1.000 OVERCONFIDENCE *CRISIS2 0.230 0.411 0.000 0.000 0.000 0.000 1.000 CRISIS3 0.375 0.484 0.000 0.000 0.000 1.000 1.000 OVERCONFIDENCE *CRISIS3 0.309 0.462 0.000 0.000 0.000 1.000 1.000 INCREASE 0.430 0.450 0.000 0.000 0.000 1.000 1.000 OVERCONFIDENCE *INCREASE 0.367 0.482 0.000 0.000 0.000 1.000 1.000
All variables are defined in the appendix under table A1.
The sample table suggests that most firms have moderate debt ratios. The assets of the firms are mostly tangible. Also, the firms are relatively of similar size as suggested by the standard deviation. Profitability varies across firms and there are some firms that are highly unprofitable. Many firms also display prospects for growth opportunities.
5. Results 5.1 Correlations
Table 2 contains the correlations among all the variables. It shows that the correlation between PROFITABILITY and LEVERAGE1 is high. This finding is as expected because high leverage results in high interest payments which in turn reduce profitability. Also, the correlations
12 Table 2: Correlations (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (1) LEVERAGE1 1.00 (2) LEVERAGE2 0.07 1.00 (3)ASSETTANGIBILITY 0.00 -0.16 1.00 (4)PROFITABILITY -0.89 -0.05 -0.05 1.00 (5) FIRMSIZE -0.09 0.12 -0.08 0.20 1.00 (6) GOWTHOPPORTUNITIES 0.09 -0.27 -0.01 0.051 -0.13 1.00 (7) EBITASSETS 0.00 0.00 0.00 0.02 0.00 0.01 1.00 (8) OVERCONFIDENCE -0.05 -0.09 -0.08 0.20 0.11 0.23 0.02 1.00 (9) CRISIS 0.03 0.00 0.02 -0.10 -0.03 -0.13 0.02 -0.21 1.00 (10) OVERCONFIDENCE*CRISIS -0.01 -0.03 0.00 0.01 0.00 -0.05 0.04 0.15 0.84 1.00 (11) CRISIS2 0.02 -0.01 0.03 -0.08 -0.03 -0.13 0.02 -0.16 0.77 0.65 1.00 (12) OVERCONFIDENCE*CRISIS2 -0.01 -0.03 0.00 0.02 0.01 -0.05 0.03 0.20 0.60 0.75 0.86 1.00 (13) CRISIS3 0.02 -0.01 0.02 -0.06 -0.02 -0.16 0.02 -0.14 0.63 0.53 0.81 0.70 1.00 (14) OVERCONFIDENCE*CRISIS3 -0.01 -0.04 -0.01 0.04 0.03 -0.07 0.03 0.24 0.46 0.61 0.67 0.81 0.86 1.00 (16) INCREASE 0.05 0.30 -0.04 -0.08 0.13 -0.16 -0.02 -0.08 0.03 -0.01 -0.01 -0.05 -0.01 -0.05 1.00 (16) OVERCONFIDENCE*INCREASE 0.01 0.22 -0.07 0.02 0.16 -0.05 -0.01 0.28 -0.06 0.04 -0.08 0.02 -0.07 0.03 0.88 1.00
13 between OVERCONFIDENCE with the interaction variable OVERCONFIDENCE*CRISIS are quite high suggesting that CEO overconfidence has a strong relation with the crisis.
As some of the correlations are high, it is important to check for multicollinearity. The vif scores are calculated and presented in table 3. Panel A shows the vif scores for model 1 while Panel B represents the vif scores for model 2.
Table 3: VIF Scores Panel A
Variable VIF 1/VIF
CRISIS 5.52 0.181 OVERCONFIDENCE 5.38 0.186 OVERCONFIDENCE*CRISIS 1.77 0.565 GROWTH OPPORTUNITIES 1.09 0.913 PROFITABILITY 1.09 0.920 FIRM SIZE 1.08 0.928 ASSET TANGIBILITY 1.01 0.987 Mean VIF 2.42
Focusing first on Panel A, it is observed that the mean vif score is 2.42 and the highest vif score is 5.52, which is well below the cutoff value of 10 as suggested by Robinson and Schumacker (2009). Thus, it appears that model 1 does not suffer from the problem of multicollinearity.
Table 3 Continued Panel B
Variable VIF 1/VIF
OVERCONFIDENCE*INCREASE 12.94 0.078 INCREASE 13.56 0.080 OVERCONFIDENCE 2.27 0.441 LEVERAGE1 1.16 0.862 GROWTH OPPORTUNITIES 1.12 0.894 FIRM SIZE 1.07 0.937 Mean VIF 5.19
Focusing now on Panel B, the mean score is 5.19 and the highest vif scores are for the variables INCREASE and OVERCONFIDNCE*INCREASE hinting existence of multicollinearity. A common solution is to remove variables with high vif scores. However, none of these two variables
14 can be removed from the model as they are the main variables. Moreover, Brambor et al. (2016) states that the problem of multicollinearity is overstated in relation to interaction terms omitting variables due to high multicollinearity may not be justified.
5.2 Regression Results
Table 4: Regression Results Dependent Variable = LEVERAGE1
OVERCONFIDENCE 0.466*** (8.93) CRISIS -0.274*** (-3.65) OVERCONFIDENCE*CRISIS 0.223*** (2.68) ASSET TANGIBILITY -0.226*** (-3.33) PROFITABILITY -7.286*** (135.41) FIRM SIZE 0.112*** (14.43) GROWTH OPPORTUNITIES 0.257*** (17.62) Constant -.889*** (-9.02) Number of observations 4063 Adjusted R2 82.23%
This table presents the regression results of the model that tests the H1 presented in the hypotheses section. All variables are defined in table A1. The t-statistics are contained in the parenthesis. ***represents significance at the 5% level for one-sided test. The standard errors are clustered by company and year.
Table 4 contains the results of regressing LEVERAGE1 on the independent variables. As per the expectation, I find a positive coefficient on OVERCONFIDENCE suggesting that CEO overconfidence has a positive impact on the amount of leverage firms take.
The coefficient on CRISIS is negative and statistically significant, suggesting that firms are unable to increase leverage during a financial crisis due to financial constraints such as stricter policies of lenders. Moreover, my variable of interest OVERCONFIDENCE*CRISIS is positive and statistically significant, suggesting that overconfident CEOs do increase leverage during a crisis. This finding is consistent with my first hypothesis.
15 The coefficients on ASSET TANGIBILITY and PROFITABILITY are statistically significant. ASSET TANGIBILITY has a negative coefficient which is contrary to my expectation, as tangible assets can be used as collateral for borrowing, while PROFITABILITY, as per my prediction, has a negative coefficient since profitable firms require less external funds.
Furthermore, the variables FIRM SIZE and GROWTH OPPORTUNITIES have statistically significant and positive coefficients. This suggests that larger firms operate with higher leverage which is consistent with my prediction. It further suggests that firms with higher growth opportunities also borrow more. This, however, is not consistent with prior research as such firms are expected to have higher distress costs.
Table 5: Regression Results
Dependent Variable = PROFITABILITY
OVERCONFIDENCE 0.035*** (3.52) INCREASE -0.089*** (-6.71) OVERCONFIDENCE*INCREASE 0.077*** (5.64) FIRM SIZE 0.016*** (13.72) GROWTH OPPORTUNITIES 0.029*** (12.93) LEVERAGE1 -0.049*** (-4.70) Constant -0.130*** (-9.94) Number of observations 1992 Adjusted R2 26.68%
This table presents the regression results of the model that tests the H2 presented in the hypotheses section. All variables are defined in the table A1. The t-statistics are contained in the parenthesis. ***represents significance at the 5% level for one-sided test. Standard errors are clustered by company and year.
Table 5 represents the results of regression with PROFITABILITY as the independent variable. Contrary to my prediction, the variable OVERCONFIDENCE has a positive coefficient suggesting that CEO overconfidence increases the profitability of a firm. The coefficient is statistically significant since the p-value is less than the significance level of 5%.
16 As per expectation, increasing leverage has a significantly negative impact on the profitability of a firm as can be inferred by the negative coefficient on INCREASE. This finding is expected as interest payments reduce profits. In addition, the variable of interest OVERCONFIDENCE*INCREASE has a positive and statistically significant coefficient. This contradicts my prediction that increased leverage due to overconfidence leads to negative consequences for firms and hence H2 is rejected. Moreover, FIRM SIZE has a positive coefficient since larger firms realize higher profits due to diversification benefits. GROWTH OPPORTUNITIES has a negative coefficient suggesting that firms that have a high market-to-book value are less profitable due to higher distress costs. These findings are in line with my prediction. Also, the variable LEVERAGE1 has a statistically significant and negative coefficient. As expected, and found in model 1, leverage results in interest payments that reduce profits.
6. Sensitivity Analysis
To test the robustness of my results from model 1, I conduct two checks. First, I examine whether my findings remain the same when I use long-term debt scaled by total assets as the dependent variable (LEVERAGE2). I conduct this analysis because certain financial theorists believe that short-term debt is taken in order to manage the working capital rather than part of the capital structure.
The coefficient on the interaction between OVERCONFIDENCE and CRISIS has the same sign as the original analysis but is no longer statistically significant. Thus, measuring leverage with long-term debt has a significant effect on the analysis as the interaction no longer has any impact on LEVERAGE2. All other the control variables are significant and have the same coefficient sign as in the main analysis except growth opportunities variable, which now has a negative coefficient.
Table 6: Regression Results
Dependent Variable = LEVERAGE2
OVERCONFIDENCE -0.033*** (-3.10) CRISIS -0.033 (-2.17) OVERCONFIDENCE*CRISIS 0.016 (0.94) ASSET TANGIBILITY -0.146*** (-10.59) PROFITABILITY -0.042*** (-3.87) (Continued)
17 Table 6 continued FIRM SIZE 0.009*** (5.61) GROWTH OPPORTUNITIES -0.048*** (-16.18) Constant 0.340*** (16.99) Number of observations 4063 Adjusted R2 10.98%
This table presents the regression results of the model similar to model 1 except the dependent variable leverage2 is calculated differently from leverage1. All variables are defined in table A1. The t-statistics are contained in the parenthesis. ***represents significance at the 5% level for one-sided test. Standard errors are clustered by company and year.
Next, I examine whether the results remain the same when I repeat the analysis by extending the crisis period by 1 year each time, until 2011. I conduct this test because the time for the end of the 2007-08 financial crisis is a debate. Most agree that it started in the last quarter of 2007, however, there may be a disagreement regarding its end (NBER, 2010). Some believe it came to an end in 2009 (NBER, 2010) whereas others suggest that its effects were observed until 2011.
Table 7 presents the results of the regression of LEVERAGE1 on the independent variables. The column (I) presents the results with CRISIS 2 as the independent variable for the crisis time period. This variable is 1 if the time period is from 2008 to 2010, and zero otherwise. Column (II) repeats the analysis with CRISIS 3 as the independent variable to represent the crisis. This variable is 1 if the time period is from 2008 to 2011, and zero otherwise.
When the crisis time period is extended until 2010 (see column I), the coefficients, their signs and the significance of the independent variables are identical to the original analysis. Moreover, when the analysis is further extended by 1 year (see column II), the crisis variable and the interaction variable both are still statistically insignificant suggesting that the financial crisis and CEO overconfidence during financial crisis do not affect leverage.
Table 7: Regression Results Dependent Variable =
LEVERAGE1 (I) (II)
OVERCONFIDENCE 0.470*** (8.51) 0.493*** (8.25) CRISIS2 -0.221*** (-2.99) (Continued)
18 Table 7 continued OVERCONFIDENCE*CRISIS2 0.197*** (2.46) CRISIS3 -0.127 (-1.70) OVERCONFIDENCE*CRISIS3 0.141 (1.77) ASSET TANGIBILITY -0.227*** (-3.33) -0.227*** (-3.33) PROFITABILITY -7.280*** (135.37) -7.273*** (-135.27) FIRM SIZE 0.112*** (14.38) 0.112*** (14.38) GROWTH OPPORTUNITIES 0.259*** (17.67) 0.261*** (17.73) Constant -0.894*** (-8.90) 0.931*** (-9.03) Number of observations 4063 4063 Adjusted R2 82.20% 82.17%
This table presents the regression results of the model similar to model 1 except the financial crisis time period is defined differently. (I) defines the crisis time period as 2007 – 2010 while (II) identifies the crisis time period as 2007 – 2011. All variables are defined in table A1. The t-statistics are contained in the parenthesis. ***represents significance at the 5% level for one-sided test. Standard errors are clustered by company and year.
The same analysis is repeated with LEVERAGE2 as the independent variable and the results are reported in table 8. The interaction between OVERCONFIDENCE and CRISIS2 has a positive coefficient (see column I) as expected, but it is not statistically significant. These results suggest that CEO overconfidence does not have on impact on the amount of leverage firms take during a financial crisis.
Column (II) shows that when LEVERAGE2 is regressed on CRISIS3 as the independent variable for the financial crisis, the variable of my interest (OVERCONFIDENCE*CRISIS3), it still has a positive coefficient implying that overconfident CEOs increase leverage during a financial crisis. This effect, however, it is not statistically significant. These results suggest that CEO overconfidence does not have an impact on leverage during crisis with the crisis time period extended beyond the year 2009.
19 Table 8: Regression Results
Dependent Variable =
LEVERAGE2 (I) (II)
OVERCONFIDENCE -0.036*** (-3.22) -0.030*** (-2.52) CRISIS2 -0.038*** (-2.56) OVERCONFIDENCE*CRISIS2 0.022 (1.37) CRISIS3 -0.029 (-1.89) OVERCONFIDENCE*CRISIS3 0.009 (0.55) ASSET TANGIBILITY -0.146*** (-10.57) -0.146*** (-10.56) PROFITABILITY -0.042*** (-3.86) -0.041*** (-3.78) FIRM SIZE 0.009*** (5.57) 0.009*** (5.55) GROWTH OPPORTUNITIES -0.048*** (-16.21) -0.049*** (-16.36) Constant 0.345*** (16.94) 0.343*** (16.42) Number of observations 4063 4063 Adjusted R2 10.89% 10.94%
This table presents the regression results of regressing the dependent variable LEVERAGE2 on the independent variables and the financial crisis time period is defined differently. (I) defines the crisis time period as 2007 – 2010 while (II) identifies the crisis time period as 2007 – 2011 All variables are defined in table A1. The t-statistics are contained in the parenthesis. ***represents significance at the 5% level for one-sided test. Standard errors are clustered by company and year.
To test the robustness of model 2, I conduct the analysis by changing the profitability measure. Instead of using net income scaled by the total assets as the dependent variable, I now use earnings before interest and tax (EBIT) divided by the total assets. The results are identical to the original analysis and the interaction between OVERCONFIDENCE and INCREASE has a positive and statistically significant coefficient. The difference, however, is that INCREASE no longer has a negative coefficient. Also, the variable LEVERAGE1 is not insignificant anymore. These results are expected as EBIT is calculated prior to making interest payments.
20 Table 9: Regression Results
Dependent Variable = EBITASSETS
OVERCONFIDENCE 0.033*** (3.16) INCREASE -0.060*** (-4.38) OVERCONFIDENCE*INCREASE 0.051*** (3.63) FIRM SIZE 0.018*** (14.59) GROWTH OPPORTUNITIES 0.038*** (16.43) LEVERAGE1 0.005 (0.44) Constant -0.124*** (-9.05) Number of observations 1992 Adjusted R2 25.27%
This table presents the regression results of the model that tests the H2 but with the dependent variable defined differently. All variables are defined in the table A1 The t-statistics are contained in the parenthesis. ***represents significance at the 5% level for one-sided test. Standard errors are clustered by company and year. Standard errors are clustered by company and year.
7. Conclusions
In this section, I conclude my research, the goal of which was to answer two questions: “Do overconfident Chief Executive Officers (CEOs) increase firm leverage during financial crises? And if they do so, what is the impact on firm performance?”As per the results of my statistical analysis, I find that overconfident CEOs increase leverage during the 2007-08 financial crisis. A possible explanation for this could be that overconfident CEOs overestimate the positive state and believe that they are able to outperform others. This idea is an interesting finding since, during a crisis, increasing leverage can bring the firm under further financial distress. Furthermore, I would assume that banks and other lenders would not be willing to lend during a financial crisis and impose stricter lending conditions since the probability of default is high. The results are contrary, however, showing that firms were indeed able to increase leverage.
What is more surprising is that overconfident CEOs report higher profits after the crisis. This finding is contradictory to the popular view that overconfident CEOs destroy firm value. Moreover,
21 taking on more leverage due to overconfidence should have affected the firm negatively through increased chances of default and reduced profitability due to interest payments. Yet, the results indicate that profitability, in fact, increases as a result of overconfident CEOs increasing leverage.
My study has certain limitations. First, many observations were dropped from the initial sample even if they were missing one of the variables. This sample selection bias can have a significant impact on the results presented. Moreover, the proxy used for overconfidence may not be an accurate representation as it focuses only on one aspect, namely, stock options. In addition, the analysis is based on a single crisis that occurred during 2007 – 2009. Also, only one country, the U.S., is analyzed during this study. Therefore, it is not clear if these findings can be generalized to the previous financial crises and other countries.
Further research in this domain will certainly contribute to existing literature. I would first suggest that the analysis may be extended to other countries and other financial crises in order to generalize the results. There has been similar research regarding the U.S. previously, but not for other countries. Moreover, a more comprehensive overconfidence measure can be developed that not only incorporates stock options, but also CEOs’ ownership stake in the firm and other psychological factors such as their beliefs. Furthermore, the firm performance measure used in this study is profitability. There may be firms that may target other measures of firm performance such as high cash and equivalents rather than profitability. Therefore, different proxies may be used for different types of firms. This will increase the precision of the research.
22 Bibliography
Arthurs, J. D., & Busenitz, L. W. (2003). The boundaries and limitations of agency theory and stewardship theory in the venture capitalist/entrepreneur relationship. Entrepreneurship
Theory and Practice, 28(2), 145-162.
Ben-David, I., Graham, J. R., & Harvey, C. R. (2007). Managerial overconfidence and corporate policies (No. w13711). National Bureau of Economic Research.
Brambor, T., Clark, W. R., & Golder, M. (2006). Understanding interaction models: Improving empirical analyses. Political analysis, 14(1), 63-82.
Chen, S., & Chen, L. L. (2009). The effect of managerial overconfidence on listed companies' capital structure and debt maturity structure. In Management Science and Engineering, 2009. ICMSE
2009. International Conference on (pp. 1348-1353). IEEE.
Cho, M. H. (1998). Ownership structure, investment, and the corporate value: an empirical analysis. Journal of financial economics, 47(1), 103-121.
Faccio, M., Marchica, M. T., & Mura, R. (2016). CEO gender, corporate risk-taking, and the efficiency of capital allocation. Journal of Corporate Finance, 39, 193-209.
Fosberg, R. H. (2012). Capital structure and the financial crisis. Journal of Finance and
Accountancy, 11, 1.
Hackbarth, D. (2008). Managerial traits and capital structure decisions. Journal of Financial and Quantitative Analysis, 43(4), 843-881.
Hackbarth,D. (2009). Determinants of corporate borrowing: A behavioral perspective. Journal of
Corporate Finance, 15(4), 389-411.
Harris, M., &Raviv, A. (1991). The theory of capital structure. the Journal of Finance, 46(1), 297-355.
Ho, P. H., Huang, C. W., Lin, C. Y., & Yen, J. F. (2016). CEO overconfidence and financial crisis: Evidence from bank lending and leverage. Journal of Financial Economics, 120(1), 194-209.
Ivashina, V.,&, Scharfstein, D. (2010). Bank lending during the financial crisis of 2008. Journal of Financial economics, 97(3), 319-338.
Jensen, M. C., &Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of financial economics, 3(4), 305-360.
23
Kochhar, R. (1996). Explaining firm capital structure: The role of agency theory vs. transaction cost economics. Strategic Management Journal, 713-728.
Kumhof, M., Rancière, R., &Winant, P. (2015). Inequality, leverage, and crises. American Economic
Review, 105(3), 1217-45.
Lane, P. R. (2012). The European sovereign debt crisis. Journal of Economic Perspectives, 26(3), 49-68.
Malmendier, U., & Tate, G. (2005). CEO overconfidence and corporate investment. The journal of finance, 60(6), 2661-2700.’
Modigliani, F., & Miller, M. H. (1958). The cost of capital, corporation finance and the theory of investment. The American economic review, 48(3), 261-297.
Myers, S. C. (1984). The capital structure puzzle. The journal of finance, 39(3), 574-592.
Myers, S. C., & Majluf, N. S. (1984). Corporate financing and investment decisions when firms have information that investors do not have. Journal of financial economics, 13(2), 187-221. Ozkan, A. (2001). Determinants of capital structure and adjustment to long run target: evidence from
UK company panel data. Journal of Business Finance & Accounting, 28(1‐2), 175-198. Psillaki, M., & Daskalakis, N. (2009). Are the determinants of capital structure country or firm
specific?. Small Business Economics, 33(3), 319-333.
Rajan, R. G., & Zingales, L. (1995). What do we know about capital structure? Some evidence from international data. The journal of Finance, 50(5), 1421-1460.
Robinson, C., & Schumacker, R. E. (2009). Interaction effects: centering, variance inflation factor, and interpretation issues. Multiple Linear Regression Viewpoints, 35(1), 6-11.
Rose, A. K., & Spiegel, M. M. (2010). Cross‐country causes and consequences of the 2008 crisis: International linkages and American exposure. Pacific Economic Review, 15(3), 340-363.
Ross, S. A. (1973). The economic theory of agency: The principal's problem. The American Economic Review, 63(2), 134-139.
Serfling, M. A. (2014). CEO age and the riskiness of corporate policies. Journal of Corporate
Finance, 25, 251-273.
The National Bureau of Economic Research. (2010). Business Cycle Dating Committee, National Bureau of Economic Research.
24
Titman, S., &Wessels, R. (1988). The determinants of capital structure choice. The Journal of
finance, 43(1), 1-19.
Tong, G., & Green, C. J. (2005). Pecking order or trade-off hypothesis? Evidence on the capital structure of Chinese companies. Applied economics, 37(19), 2179-2189.
Tsoy, L., &Heshmati, A. (2017). Impact of Financial Crises on Dynamics of Capital Structure: Evidence from Korean Listed Companies.
Vassalou, M., & Xing, Y. (2004). Default risk in equity returns. The journal of finance, 59(2), 831-868.
Wald, J. K. (1999). How firm characteristics affect capital structure: an international comparison. Journal of Financial research, 22(2), 161-187.
25 Appendix
Table A1: Measures for each variable
LEVERAGE1 This variable is the ratio of total debt to the total assets for the current year measured at year end.
LEVERAGE2 This variable is the ratio of long-term debt to the total assets for the current year measured at year end.
ASSET TANGIBILITY This variable is the ratio of total tangible assets to the total assets for the current year measured at year end.
PROFITABILITY This variable is the net income scaled by the total assets for the current year measured at year end.
FIRM SIZE This variable is the natural logarithm of total revenues for the current year measured at year end.
GROWTH OPPORTUNITIES This variable is the ratio of market value of firm to the total book value of assets for the current year measured at year end.
EBITASSETS This variable is the ratio of earnings before interest and tax to the total value of assets for the current year measured at year end.
OVERCONFIDENCE This variable is equal to 1 if the moneyness of the option held by a CEO is greater than 100%, and 0 otherwise.
CRISIS This variable is equal to 1 if the time period is 2008 – 2009, and 0 otherwise.
OVERCONFIDENCE*CRISIS This variable is an interaction between OVERCONFIDENCE and CRISIS. CRISIS2 This variable is equal to 1 if the time period is 2008 – 2010, and 0
otherwise.
OVERCONFIDENCE*CRISIS2 This variable is an interaction between OVERCONFIDENCE and CRISIS2. CRISIS3 This variable is equal to 1 if the time period is 2008 – 2011, and 0
otherwise.
OVERCONFIDENCE*CRISIS3 This variable is an interaction between OVERCONFIDENCE and CRISIS3. INCREASE This variable is equal to 1 if leverage was increase in that year.
OVERCONFIDENCE*INCREASE This variable is an interaction between OVERCONFIDENCE and INCREASE.
26 Graph A2
