Behavioral
elements
of capital
structure
Bachelor Thesis-
Finance
05-07-2013
Wesley van der
Vijgh-10012001
Supervisor-Dr. D.A. Salzman
2
1. Table of contents
2. Introduction ... 3
3. Literature ... 4
Overoptimism and Overconfidence ... 4
Growth perception bias and risk perception bias ... 5
Ability ... 6 Gender ... 7 Age ... 7 4. Data description ... 7 5. Research Methodology ... 9 6. Results ... 12
7. Discussion and concluding remarks ... 19
3
2. Introduction
An increasing amount of literature talks about personal characteristics and character traits of managers that influence the decision making process in a firm. These papers make models or perform empirical research on the influence of one or a few characteristics or character traits on among other things the capital structure of a firm. This goes against classical financial theory that talks about rational managers who make decisions in a perfect market (Modigliani and Miller, 1958). In the papers that have been written, statements are made about the influence of certain behavioral variables on the capital structure of a firm. But no paper has been written before that collects the variables that are found to have an influence on capital structure and empirically finds out to what extend these variables together influence capital structure. Another fact that has never been researched is if one of these variables influences capital structure to a greater extend, than the others. This paper answers to the research question:
To what extend do personal traits and characteristics of the CEO influence the capital structure of a firm?
To answer this question, this paper collects the characteristics and traits and includes them in one OLS regression to see if and to what extend all of these variables actually influence capital structure. The variables included in the regressions are overconfidence measured in unexercised but exercisable option holdings of a CEO, gender, age, ability and CEO tenure.
If the term CEO is used, the CEO is CEO of a firm that is part of the Standard and Poors (S&P) 1500 or was part of the S&P 1500 but is still listed.
Section 1 describes the related literature of this field. Section 2 describes the dataset that is used, section 3 describes the results, section 4 combines the literature with the results and section 5 discusses the paper and makes some concluding remarks.
4
3. Literature
In 1958, Modigliani and Miller wrote about the capital structure of a firm and the first theories in the field of corporate finance were formed. This classical theory of corporate finance created a market in which capital structure showed to be irrelevant for the value of a firm. This market is called the perfect market (Modigliani and Miller, 1958). A perfect market is defined as a market in which every investor acts fully rational, possesses all information that exists in the market and there are no transaction costs or taxes. These assumptions cause debt and equity to be perfect substitutes. In 1963 Modigliani and Miller published another theory that built on the theory of 1958, which showed the influence taxes could have on the value of the firm. This theory is called the tradeoff theory and included tax benefits of debt and bankruptcy costs caused by increasing debt (Modigliani and Miller, 1963).
These theories where succeeded with papers that empirical find variables that influence the capital structure of a firm. There is found that many factors influence the capital structure of a firm, Rajan and Zingales (1995) state that size, profitability, investment opportunities, tax shields, volatility of earnings and the probability of bankruptcy influence the capital structure of a firm. Bowen et al. (1982) found that capital structure is also industry specific.
Today’s literature adds some assumptions in the form of a behavioral element to that theory, which causes the theory on capital structure to resemble reality more closely. The variables of interest that the literature focused on are discussed in the sections below.
Overoptimism and Overconfidence
The paper of Baker, Ruback and Wurgler (2007) writes about the irrational managers approach, which contradicts the assumptions of Modigliani and Miller(1958). In this irrational managers approach managers show behavioral biases, which cause them to depart from rational expectations, and do not succeed in maximizing expected utility. This theory also stated that limited corporate governance causes the managers to not be constrained in making rational decisions. The paper views these behavioral elements from a psychological angle. The behavioral elements that influence manager behavior are overoptimism and overconfidence. To describe optimism from a psychological angle, Baker, Ruback and Wurgler (2007) cite Weinstein (1980). The
5
paper of Weinstein (1980) states that people have the tendency to believe that it is more likely than average that a positive future life event happens to them (like having a mentally gifted child) and less likely than average that a negative life event happens to them (like divorcing after a few years of marriage). This shows the natural human optimism bias. The angle of overconfidence is covered by a study of Svenson (1981), in which Swedish- and US people are asked to indicate their level of driving skills in comparison with the other participants. 93 percent of the US-sample considered themselves more skillful than the median driver, and for the Swedish-sample in this research it was 69 percent.
Overconfidence and optimism are also showed in investment decisions. In case of startup firms, 95 percent of the entrepreneurs indicate that they think their chance of success is 5 out of 10 or better, 81 percent estimate their success to be 7 out of 10 or better (Cooper, Woo Dunkelberg, 1988). This article also states that 68 percent of the entrepreneurs perceive that their odds for success are better than for other start-up firms. Actual odds turn out to be much lower, the authors state that 67 percent of start-ups do not survive the first four years of business.
Growth perception bias and risk perception bias
The paper of Hackbarth (2008) assembles a model that incorporates managerial traits in the capital structure decision. The managerial traits the author talks about are growth perception bias and risk perception bias. Managers with a growth perception bias tend to overestimate the future earnings, you can also define them as optimistic. Managers with a risk perception bias tend to underestimate the riskiness of earnings, they could also be defined as overconfident. Hackbarth states that biased managers perceive the firm as more profitable or less risky and therefore prefer to finance with debt. The paper concludes that the model shows that managers with the growth- and risk perception biases choose higher levels of debt and issue debt sooner than identical but unbiased managers. Due to these managerial traits, there can be significant variation in capital structure of firms with the same characteristics that operate in the same industry.
A paper that contributes to the findings of the paper of Hackbarth, is the paper of Malmendier and Tate (2005). This paper also talks about the influence of an overconfidence bias on making financing decisions. It states that CEOs tend to overestimate the future returns of the projects they invest in. If these CEOs have
6
enough cash on their hands, they will overinvest in projects that are relatively not profitable. In case these types of overconfident CEOs do not have enough cash on their hands, they are unwilling to issue equity to finance these investment projects. This unwillingness to issue new equity is caused by the fact that these CEOs think their equity is undervalued by the market. Because a lack of rationality in valuation of the firm, just like the managers in Hackbarth (2008), the managers in the paper of Malmendier and Tate tend to overestimate this value. This overvaluation due to irrationality causes these managers to make different capital structure decisions then rational managers would make.
Ability
The above literature describes the fact that due to overestimating their own skill set or the future situation managers in reality tend to make different decisions as perfectly rational managers would do. The influence of these biases on decision making, causes the managers to choose a different capital structure as rational managers would do. Another string of literature talks about the ability of a CEO that influences decision making.
In the paper of Bhagat, Bolton and Subramanian (2011) a model is built in which manager specific characteristics and incentives are analyzed on capital structure. Bhagat, Bolton and Subramanian expected that an increase in manager’s ability increases the output of a manager, which causes their expected payoff to rise. A manager prioritizes the continuation value of the firm above his initial payoff. This causes the manager to pick a lower level of long-term debt, because a higher level of debt increases the probability of bankruptcy. And an increased probability of bankruptcy causes the expected payoff of the manager to be lower. The regressions showed that when managerial ability rises, managers tend to choose lower levels of short-term debt. This fact also holds for long-term debt. This shows that an increasing ability causes the overall debt levels to be lower. The regressions they ran also showed that managers tend to choose lower levels of long-term debt when their equity ownership increases. This shows that when a manager is increasingly personally exposed to the idiosyncratic risk of the company, he will try to minimize this risk by choosing lower levels of long-term debt.
7 Gender
Another trait that possibly has an influence on the decision making of managers is gender. The article of Huang and Kisgen (2012) talks about the influences of gender differences on various decision-making processes in a corporate setting. This is the first paper that studies gender differences in this field. The paper states that males and females have different attitudes toward risk and confidence levels. It empirically finds that the difference in growth of a firm is significantly smaller in the tenure of a female CEO, compared to the tenure of a male CEO. In this paper growth is operationalized as the percentage of change in the total assets. This smaller growth in assets is for about 20 percent explained by the fact that women tend to make fewer acquisitions than men. Besides this finding, there is also found that female CEOs issue less debt than their male counterparts. In addition to this finding there is also found that in case of a female CEO, there is a lower component of debt in the capital structure.
Age
The paper of Bertrand and Schoar (2003) talks about managerial fixed effects that influence the way decisions are made within a company. This means that managers personally influence firm behavior and performance due to their own ‘management style’. From this paper it becomes clear that the age of a CEO influences the capital structure decision of a firm. The authors base their idea that age changes decision making in a firm on the fact that older generations of managers are more conservative in choosing the amount of leverage in their capital structure. They empirically find that older generations of CEOs choose a lower level of debt in the capital structure. If the year of birth of a CEO increases by 10 years, the debt component in the capital structure increases with approximately 2,5 percent.
4. Data description
The dataset that is used in this paper is retrieved from the WRDS database. The data on option holdings, the data that enabled me to calculate CEO tenure and the data on age are obtained from the ExecuComp database. This database contains data on firms of the S&P 1500, combined with firms that were once in the S&P 1500, but were removed and are still traded in the market. This database is used as the bases for selecting companies. The range of years that was selected on the time that this data
8
was retrieved from the ExecuComp database, was set as wide as possible. This wide range was chosen to be able to remove companies on which there was no data available in the other databases used. The data on Debt and Equity from the selected companies was retrieved from the capital IQ database. From this database a total debt and equity level was calculated and then merged with the data from the ExecuComp database. The data on assets, sales and earnings was retrieved from the Compustat database.
The companies that remained in the data selection are companies on which there was data to calculate CEO tenure, on which there was data on option holdings to calculate a confidence level of the CEO and on which there was data on debt and equity holdings. More on the calculations of confidence levels and the definition of overconfidence is included in the section about the research methodology. The dataset that remained after this selection contains 2149 data points, which include 458 firms. 1,6 percent of the CEOs in this dataset are female, which is tabulated in Table 2. Table 2 also shows the average amount of days that a CEO remains in his position as CEO, which is 3478 days with a standard deviation of 2677 days. The mean age of a CEO is 52 year, with a standard deviation of 7.8 years.
9
5. Research Methodology
To see if there is a significant relationship between the variables of interest and the dependent variable capital structure and to see which variables have the greatest influence on capital structure, multiple OLS regressions were performed on the variables confidence level, gender, ability and age and the dependent variable capital structure. First single regressions on these variables are performed to see if there is a significant relationship between the variable and the capital structure. To see what the influence of this single behavioral variable is on the capital structure, the adjusted R² of these regressions is analyzed. After this, regressions are performed with the control variables and a single behavioral variable included. By analyzing the adjusted R², a statement can be made about which variable influences capital structure the most. I expect that the hypotheses this model should prove is that CEO ability has the greatest influence on capital structure. This should be clear from the greater increase in R² when adding proxies for ability.
Then a benchmark regression is performed to see what the R² is when only the control variables are included. To see to what extend the combination of behavioral variables influence capital structure, the behavioral variables of interest are included to see how the R² changes when these variables are included. Due to correlation between the variables of interest, not all the variables can be included in one regression. The regression below is the starting point. The hypothesis that should be clear from the regressions of this paper, is that the behavioral variables influence capital structure to a great extend.
D/E-ratio = βo + confidence β + age β + gender β + CEO tenure β
+ CEOTenure/Age β + Assets β + Sales β + ROA β + dummies Industry β
Capital structure is in this regression model operationalized as D/E ratio, which is calculated as total debt divided by total equity.
The confidence level of the CEO is based on the option holdings of the CEO. The way a confidence level is determined in this regression model is derived from the paper of Malmendier and Tate (2011). They base their confidence measures on the fact that CEOs get a large amount of stock options and equity. The CEOs cannot
10
diversify due to the fact that the options cannot be traded for a while and the sale of equity holdings is only possible on restricting terms. Besides this, firms make it even more difficult to diversify by banning short selling. This causes an increased exposure to firm-specific risk. In this paper the authors state a CEO as overconfident if he holds 67 percent or higher in-the-money stock options during the fifth year. The five years limit is based on the vesting period of the options of the CEOs in the dataset of Malmendier and Tate. After the vesting period, the five years, a CEO is allowed to exercise his option. If he fails to do so and the option is 67 percent or higher in the money, he is defined as overconfident. The analogy behind this measure of overconfidence is that due to the high amount of confidence in the handpicked future investments by the CEO, the CEO overestimates his future earnings and thinks that the value of the options he holds will increase. This will cause the CEO to postpone the moment of exercising his options. Due to the limitations of the databases I have access to, I could not define a CEO as overconfident in exactly the same way. My definition is derived from the definition of Malmendier and Tate. In my regression, a CEO is stated as overconfident if he holds 67 percent or more of total options holdings that are exercisable, but he chooses to not exercise these options. This is based on the same idea that an overconfident CEO believes the value of his exercisable options will rise in the future due to the decisions he has made today. The percentage of CEOs that are labeled as overconfident according to this definition is 60 percent. The division of the confidence levels of the CEOs is tabulated in the histogram in Table 1.
Ability is defined in the same manner as in the paper of Bhagat, Bolton and Subramanian (2011) in which one of the proxy’s of ability is CEO tenure to age. The authors state that it is plausible that a younger CEO, who is CEO as long as an older counterpart, has greater ability. This assumption is based on the fact that the CEO was able to rise to the same rank at an earlier age, and was able to stay in his position as long as his older counterpart. Other things held constant, Bhagat, Bolton and Subramanian (2011) call it plausible that this younger CEO has a greater skill set, which is called ability. Another proxy for ability in the paper of Bhagat, Bolton and Subramanian is CEO tenure. CEO tenure is calculated as the amount of days that a CEO remains in his function as CEO. It is assumed that a CEO with a longer tenure has greater ability. The greater ability causes better firm performance, which enables the CEO to remain in his position. According to Bhagat, Bolton and Subramanian
11
(2011), literature shows that CEO turnover rises when a CEO is underperforming, this supports the idea that greater ability causes longer tenure.
Table 1 Division of confidence levels
Table 2 Summary statistics
Variable Observations Mean Std. Dev. Minimum Maximum
Confidence 2149 .5601 .2602 0 0.9975 Age 2149 52.10 7.832 30 81 Dummy male 2149 .9837 .1266 0 1 Dummy Female 2149 .0163 .1266 0 1 CEO tenure 2149 3478 2677 39 15857 CEO tenure/Days 2149 65.56 45.63 .6724 278.2 Assets 2149 6374 17930 .0000 345977 Sales 2149 12714 59654 11.63 1351520 ROA 2149 2.475 21.56 -10.98 599.2 Dummy 1 Industry 2149 .0665 .2493 10.00 55 Dummy 2 Industry 2149 .0726 .2595 0 1 Dummy 3 Industry 2149 .1587 .3655 0 1 Dummy 4 Industry 2149 .2215 .4154 0 1 Dummy 5 Industry 2149 .0316 .1751 0 1 Dummy 6 Industry 2149 .0973 .2964 0 1 Dummy 7 Industry 2149 .1019 .3026 0 1 Dummy 8 Industry 2149 .1745 .3796 0 1 Dummy 9 Industry 2149 .0121 .1094 0 1 Dummy 10 Industry 2149 .0633 .2435 0 1
12
6. Results
To control for other factors that influence the capital structure of a firm (the debt- equity ratio) I included some control variables in the regression. The list of control variables is not exhaustive, but should enable me to filter out some endogeneity from the regression. The control variables are selected from the papers of Bhagat, Bolton and Subramanian (2011), from Raghuram, Bowen et al. (1982) Rajan and from Zingales (1995). As proxy for size Assets and Sales are added to the regressions. As proxy for profitability the Return On Assets (ROA) is added, which is calculated as EBITDA/Total Assets. To control for industries the first two digits of a Global Industry Classification Standard (GICS) code are added to the database. These two digits divide the firms in sectors. Every sector is assigned a dummy and these dummies are included in the regressions. Because of the high correlation between Assets and Sales, these two variables cannot be included both in one regression. For continuation reason, there is chosen to leave out Assets from the preceding regressions.
To see the relationship between the variables included in the total regression and the dependent variable and to see the R² of the regression with only the single variable included, regressions were performed on these variables and the dependent variable separately. Table 4 shows the separate regressions of the variables that are included in the full regression. Due to the statistical biases that influence this regression, not much can be concluded from these single regressions. These single regressions show there is a significant negative relationship between the amount of days a CEO is in his tenure and capital structure, a significant negative relationship between the CEO tenure to age of a CEO and capital structure and a significant positive relationship between being a man and capital structure. However the relationship between the confidence level of a CEO and the capital structure is not significant in this single regression and the relationship between age and capital structure is not significant either. By analyzing the R² in this table, it seems that the proxy for ability calculated as CEO tenure/age explains by itself relatively the most
13 Table 3 Correlation between control variables
14
Dependent variable D/E 1 2 3 4 5
Confidence 1.7802 (5.1322) CEO tenure -.0010** (.0005) CEO tenure/age -.0653** (.0292) Dummy male 22.5243** (10.5381) Age .2667 (.1704) R² .0001 .0019 .0023 .0021 .0011 Adjusted R² -.0004 .0014 .0019 .0017 .0007
Table 4 Single regressions of variables of interest
*=significant at 10% level, **=significant at 5% level, ***=significant at 1% level, standard errors in parentheses. 2149 Obs. included.
After this, multiple regressions were performed to find out if there is a statistical significant relationship between the variables of interest and capital structure, with the control variables included. In addition to this, by analyzing the adjusted R², the regressions show which variable explains the most variation in the dependent variable capital structure. Table 5 shows the results of these regressions.
The first regression of table 5 shows that the significant negative relationship between the amount of days a CEO is in his tenure and capital structure remains with the control variables included in a multiple regression. The positive significant relationship between gender and capital structure remains as well. With the control variables included a significant negative relationship is found between age and capital structure. The relationship between the variable CEO tenure to age is also negative, but is not significant with the control variables included. Also the relationship between the confidence level of the CEO and the capital structure remains insignificant. When analyzing the R² of these regressions it is showed that the regression with the highest adjusted R², which is the regression that explains the most
15
variance of the capital structure, is regression 4. This regression includes the variable age, which is shown to relatively have the most influence on capital structure. This analysis is shown more clearly when taking regression 1 as the basic regression and compare the adjusted R² of the regressions with 1 additional variable to this basic regression with only control variables. This is done in Table 6, which shows a relatively negative percentage change in the adjusted R² for the regression with the variable confidence level included. This could be due to the fact that the relationship of confidence level with capital structure is not significant. The variable that shows the largest relative positive change is age. After this the proxies for ability show to cause the highest relative percentage change and after that gender and confidence level.
In the regressions performed in the section above, there is not found a significant relationship between the confidence level of a CEO and capital structure and between the CEO tenure to age and capital structure. To see if a significant relationship can be found, multiple regressions were performed with all the control variables and as much variables of interest included.
Do to the high correlation between the variables CEO tenure to age, CEO tenure and age, which is shown in Table 7, the variables CEO tenure and age are left out of this regression. First I ran regressions in which I included a dummy for every year. These dummies could control for different periods in time, which can include something like the IT bubble. These periods in time can also influence the capital structure in several industries. The addition of these dummies caused the regression to show some strange and insignificant results, which are left out of this section. The next step to find a significant relationship between the confidence level of a CEO and capital structure, like the literature states, was to scale the confidence levels with increments of 25 percent. In this way you are able to see if the relationship between capital structure does hold for higher amounts of confidence levels. The managers are now grouped to their levels of confidence, group 1 holds 25 percent exercisable options, group 2 holds between 25 and 50 percent exercisable options, group 3 holds between 50 and 75 percent exercisable options and group 4 holds over 75 percent of exercisable options. Group 4 only includes overconfident managers, according to the definition. For the groups 1, 2 and 3, there is no significant relationship found. In regression 4 of Table 8 the regression is shown, which includes group 4, the group with only overconfident managers. For this group a significant negative relationship is
16
found. For CEO tenure to age there is also found a significant negative relationship. Instead of Sales as control variable, the variable Assets is now included in the extended regressions. Sales caused some strange interference which caused the extended regression to show strange insignificant results and the variable Assets did not do that.
When analyzing the R² of the regressions of Table 8, it is shown that the R² rises after including a variable. To be able to answer the research question to what extend the behavioral elements influence the capital structure of a firm, the adjusted R² of this extensive regression (regression 4) is analyzed and compared to the basic regression with only the control variables included (regression 1). It is found that the extensive regression 4 causes the R² to increase with 3,7 percent.
When comparing the significant relationships found in the regressions with literature, several remarks can be stated. The significant negative relationship between CEO tenure to age and CEO tenure, which both proxy for ability and D/E structure found in the regressions corresponds to the findings of Bhagat, Bolton and Subramanian (2011). The significant positive relationship between being a man and capital structure, found in the regressions, corresponds to the findings of Huang and Kisgen (2012). The significant negative relationship found between age and capital structure corresponds to the findings of Bertrand and Schoar (2003). However the significant negative relationship that is found on group 4, the group with only overconfident CEOs according to the definition, and capital structure contradicts the literature that is written about overconfidence and the capital structure of a firm.
17
Table 5 Regressions of variables of interest with control variables included *=significant at 10% level, **=significant at 5% level, ***=significant at 1% level, standard errors in parentheses. 2148 Obs. included.
Table 6 % change of R² when adding variable to basic regression
Dependent variable D/E 1 2 3 4 5 6 Sales .0004*** (.0000) .0005*** (.0000) .0005*** (.0000) .0004*** (.0000) .0004*** (.0000) .0004*** (.0000) ROA -.0287 (.0536) -.0288 (.0536) -.0268 (.0536) -.0280 (.0535) -.0268 (.0536) -.0280 (.0536) Dum Industry 1 -15.4532*** (5.7872) -15.2778*** ( 5.8058) -14.9672*** (5.7909) -14.4697** (5.8019) -15.2500*** (5.7885) -15.8497*** (5.7894) Dum Industry 2 -2.6030 (5.6368) -2.5109 (5.6428) -3.4487 (5.6542) -2.0246 (5.6392) -3.3010 (5.6617) -2.7000 (5.6346) Dum Industry 3 1.6243 (4.6776) 1.7137 (4.6841) 1.081 (4.6853) 2.1811 (4.6815) 1.1378 (4.6919) 1.2313 (4.6812) Dum Industry 4 -13.5234*** (4.4172) -13.4680*** ( 4.4204) -13.4959*** ( 4.4150) -13.4085*** ( 4.4141) -13.5497*** (4.4166) -13.3308*** ( 4.4167) Dum Industry 5 -22.3185*** (7.3937) -22.4442*** (7.4022) -23.5724*** (7.4239) -22.5731*** (7.3889) -23.2864*** (7.4304) -22.6907** (7.3937) Dum Industry 6 -26.8257*** (5.1896) -26.8217*** (5.1906) -27.4859*** (5.2004) -27.5679*** (5.1977) -27.2637*** (5.2000) -26.8425*** (5.1873) Dum Industry 8 -30.9773*** (4.6273) -30.9301*** (4.6300) -31.6380*** ( 4.6400) -32.4793*** (4.6793) -31.3150*** (4.6339) -31.2094*** ( 4.6273) Dum Industry 9 -38.5061*** (11.0320) -38.6593*** (11.0411) -39.5052*** (11.0409) -40.0308*** (11.0475) -39.1248*** (11.0406) -38.8571*** (11.0290) Dum Industry 10 1.3958*** (5.8452) 1.2269 ( 5.8624) .3504 (5.8721) 1.5175 (5.8409) .5622 (5.8800) 1.010 (5.8471) Confidence -1.7505 (4.4803) CEO tenure -.0007* (.0004) Age -.3215** .1541 CEO tenure/age -.0333 (.0257) Dummy Male 15.4735* ( 9.122) R² .2655 .2656 .2666 0.2670 .2661 0.2665 adjusted R² .2617 .2615 .2625 0.2629 .2620 0.2624
Variable % change adj
R² Age 0,46 Days CEO 0,31 CEO tenure/age 0,11 Dummy male 0,07 Confidence -0,08
18 Table 7 Correlation table
Dependent variable D/E 1 2 3 4 Assets .0006*** (.0000) .0006*** (.0000) .0006*** (.0000) .0005*** (.0000) ROA .0353 (.0596) -.0362 (.0596) -.0326 (.0596) -.0327 (.0595) Dum Industry 1 -34.9191*** (6.3656) -35.4184*** (6.3753) -35.0890*** (6.3702) -35.5321*** (6.3713) Dum Industry 2 -22.6254*** (6.1978) -22.8490*** (6.1990) -24.2645*** (6.2220) -24.3292 (6.2187)*** Dum Industry 3 -18.6820*** (5.1230) -18.8597*** (5.1237) -19.8216*** (5.1349) -20.2356*** (5.1370) Dum Industry 4 -33.9303*** (4.8314) -34.0249*** (4.8310) -34.0112*** (4.8259) 23.1803*** (8.2305) Dum Industry 5 -55.3163*** (8.7519) -55.0079*** (8.7532) -56.7028*** (8.7741) 56.9051*** (8.7700) Dum Industry 6 -45.2335*** (5.7243) -45.0338*** (5.7252) -45.7660*** (5.7278) -45.7273*** (5.7247) Dum Industry 8 -51.3434*** (5.0504) -51.3190*** (5.0495) -51.9020*** (5.0504) -52.1392*** (5.0492) Dum Industry 9 -47.8704*** (12.3443) -47.7845*** (12.3421) -48.9700*** (12.3397) -49.2931*** (12.3341) Dum Industry 10 -15.5317*** (6.4591) -15.0025** (6.4700) -16.4369** (6.4920) -16.8480*** (6.4923) .75<Confidence<1 -3.9475 (2.9368) -5.4006* (2.9989) -5.3889* (2.9972) CEOtenure/age -.0682** (.0292) -.0679** (.0292) Dummy Male 18.5722* (10.1367) R² 0.0913 0.0921 0.0944 0.0958 Adjusted R² 0.0867 0.0870 0.0889 0.0899
Table 8 Extended regressions of variables of interest with control variables included *=significant at 10% level, **=significant at 5% level, ***=significant at 1% level, standard errors in parentheses. 2148 Obs. included. Confidence CEOtenure /age CEO tenure Age Dummy male Confidence 1.0000 CEOtenure /age 0.2252 1.0000 CEO tenure 0.2424 0.9592 1.0000 Age 0.2059 0.1746 0.3900 1.0000 Dummy male -0.0013 -0.0178 -0.0005 0.0730 1.0000
19
7. Discussion and concluding remarks
This paper shows that the personal traits and characteristics of the CEO influence the capital structure of a firm to a great extend. Inclusion of the variables of interest causes the adjusted R² to rise with almost 4 percent. In addition to this, it is found that the variable age influences capital structure the most. After that from the most influence to the least influence respectively the variables that proxy for ability, the variable gender and the variable which shows the level of confidence follow.
The fact that the significant relationship that is found in the regressions on overconfidence contradicts the literature could be due to the alternative operationalization of overconfidence. The databases we have access to, do not include information on how far a stock option of a CEO is in the money. The alternative operationalization that followed from this limitation caused the operationalization to be more like the operationalization that states a CEO as overconfident if he holds additional stock of a company. This way of operationalization is also stated in Malmendier and Tate (2011), but it is less accurate. Besides this, the paper of Bhagat, Bolton and Subramanian (2011) states that if the equity holdings of a manager increase, he decreases the long-term debt of the firm.
The downside of this thesis is the limited access it has to data. In addition to the shortcomings of the alternative operationalization of overconfidence another example of this limitation is the lack of data on mergers and acquisitions that a company goes through. If there was data on these mergers and acquisitions, overoptimism could be included as one of the personal characteristics of CEOs. Optimism would then have been operationalized as the amount of mergers and acquisitions a CEO undertakes in his tenure just like Malmendier and Tate (2003) did in their research.
20
8. References
Baker, M., Ruback, R. and Wurgler, J. (2007). Behavioral corporate finance: A survey." In The Handbook of Corporate Finance: Empirical Corporate Finance, edited by Espen Eckbo. New York: Elsevier/North Holland.
Bertrand, M. and Schoar, A. (2003). Managing with style: The effect of managers on firm policies. The Quarterly Journal of Economics, 118(4), 1169-1208.
Bowen, R. M., Daley, L. A., and Huber Jr, C. C. (1982). Evidence on the existence and determinants of inter-industry differences in leverage. Financial Management, 10-20.
Bhagat, S., Bolton, B., and Subramanian, A. (2011). Manager characteristics and capital structure: Theory and evidence. Journal of Financial and Quantitative Analysis, 46(6), 1581-1627.
Hackbarth, D. (2008). Managerial traits and capital structure decisions. Journal of Financial and Quantitative Analysis, 43(4), 843-882.
Huang, J. and Kisgen, D. J. (2012). Gender and corporate finance: Are male executives overconfident relative to female executives?. Journal of Financial Economics.
Malmendier, U., and Tate, G. (2005). CEO overconfidence and corporate investment. The journal of finance, 60(6), 2661-2700.
Malmendier, U. and Tate, G. (2008). Who makes acquisitions? CEO overconfidence and the market's reaction. Journal of Financial Economics, 89(1), 20-43.
Modigliani, F., and Miller, M. H. (1958). The cost of capital, corporation finance and the theory of investment. The American economic review, 48(3), 261-297.
21
Rajan, R. G. and Zingales, L. (1995). What do we know about capital structure? Some evidence from international data. The journal of Finance, 50(5), 1421-1460.
Svenson, O. (1981). Are we all less risky and more skillful than our fellow drivers?. Acta Psychologica 47(2), 143-148.
Weinstein, N. D. (1980). Unrealistic optimism about future life events. Journal of Personality and Social Psychology 39(5), 806-820.
