Do Overconfident CEOs Affect Default Risk? It Depends

(1)

Do Overconfident CEOs Affect Default Risk? It Depends

Using an option based proxy for CEO overconfidence (Malmendier & Tate 2005, 2008) and the distance-to-default, Ohlson’s (1980) O-score and Altman’s (1968) Z-score as default risk measures, I find that over the 2006-2016 period S&P1500 firms with an overconfident CEO do not have a higher or lower risk to default. However, when focusing on leverage,

profitability and Tobin’s Q, I find that overconfidence can influence default risk in different directions. A potential limitation of this study could be that matching effects results in simultaneous causality.

Master’s Thesis Finance

Author: Sander Blijham

Student number: 11420995

Supervisor: dr. Rafael Almeida da Matta Date of publication: 28-06-2017

(2)

Statement of Originality

This document is written by Sander Blijham 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 has 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)

1. Introduction

Do overconfident CEOs affect their firms’ default risk? Bertrand & Schoar (2003) show that a significant portion of corporate decisions can be explained by the so-called manager fixed effects. The authors interpret these findings as evidence that individual characteristics affect corporate policies. In the behavioral finance literature, one of these individual characteristics is overconfidence. The overconfidence concept stems from the notion of the ‘better-than-average’ effect, which is well documented in the psychology literature (see e.g., Weinstein, 1980). The ‘better-than-average’ effect means that agents overestimate their own qualities and abilities relative to others and consider themselves to be above the average at a particular skill (Alicke, 1985). In the behavioral finance literature, overconfident CEOs are often modelled to overestimate future firm performance.1

That overconfidence can have a substantial impact on corporate decision making was probably first documented in Roll’s (1986) “hubris hypothesis” of corporate takeovers. The “hubris hypothesis” is a theoretical approach to explain the phenomenon of value destroying mergers and acquisitions. Although, the term “overconfidence” is not explicit mentioned in his work, the “hubris hypothesis” suggests that managers are too confident about the benefits emanating from mergers and acquisitions, resulting in overbidding for target firms. This effect is empirically shown by Malmendier & Tate (2008); overconfident CEOs tend to overestimate their ability to generate returns. This is caused by the fact that they are overconfident about their signals of the value of the firm that they are taking over. That means, that on the margin, those CEOs undertake mergers that destroy value. Since Malmendier & Tate (2005) created a proxy for overconfidence, CEO overconfidence is a rapidly growing area in the field of economics and finance.

Hackbarth (2008), for example, suggests that overconfident CEOs are more disposed to debt finance because they believe that the firm is more profitable and/or less risky. Hence, in their view, the firm is less likely to experience financial distress. Malmendier, Tate & Yan (2011) show that in addition to the traditional theories about financing decisions, managerial

1_{The definition for overconfidence and optimism is sometimes misunderstood in the literature. Therefore,} these terms are often used interchangeably. Someone is classified as optimistic if he believes that good outcomes are more likely than they really are: Ê[X] > E[X]. Overconfidence means that someone believes that the information he possesses is more precise than it really is: !"# $ < !"#[$]. In the behavioral finance literature, overconfidence CEOs are often modeled to overestimate future firm performance: the expected future firm performance is better than the realized future firm performance. Following the definitions, optimism would be a better term than overconfidence. However, because most authors use overconfidence instead of optimism, I stick to their terminology.

(4)

characteristics have significantly power in explaining corporate financing decisions. Overconfident CEOs see external financing as unnecessarily costly, because they

overestimate their firms’ future cash flow. As a result, overconfident CEOs believe that their firms are undervalued by the market. Since equity prices are more affected than debt

regarding differences in opinions about future cash flows, overconfident CEOs prefer debt to equity when using external financing, which (other things being equal) leads to higher leverage observations. According to Huang, Tan & Faff (2016) overconfident CEOs believe that they have positive information, which the market does not know yet. Positive private information about the firm’s prospects leads to mispricing of its securities. Since this

mispricing is more severe on long term debt than short term debt, overconfident CEOs tend to adopt a debt structure with a high proportion of short-term debt (due within 12 months). The high liquidity risk associated with such a financing strategy does not discourage them.

Overconfident CEOs also tend to invest more than rational CEOs, since they are more confident about their management skills (Hirshleifer, Low, & Teoh, 2012). This indicates that overconfident CEOs overestimate the returns to their investment project. Due to their

reluctance to access external financing overconfident CEOs tend to overinvest when free cash is available, but underinvest when they need external financing (Malmendier & Tate, 2005).

The vast majority of researchers emphasize that overconfident CEOs make irrational and value destroying decisions. Nonetheless, there are many firms that hire overconfident CEOs. In order to explain that phenomenon, some authors have suggested positive roles for overconfident CEOs. According to Hirshleifer, Low & Teoh (2012) overconfidence results in greater innovation within innovative industries, both on the input side (R&D) and on the output side (patent and citations). Furthermore, for a given level of expenditure on R&D, overconfident managers achieve greater innovative success. Therefore, CEO overconfidence does not necessarily harm firm value and profitability. Moreover, they find little evidence that overconfidence might even have a positive effect on firm performance. Another advantage of a risk averse manager’s overconfidence is that it makes him less conservative. Therefore, the willingness of such a CEO to engage in risky projects can be valuable to the firm (Gervais, Heaton, & Odean, 2011).

Given above findings, the question arises whether and to what extent such traits affect the default risk of firms. Default risk is the probability that a firm is unable to meet the required payments of their debt obligations. Although a lot have been written about the concept and impact of overconfidence in financial decision-making, its influence on default risk is, to my knowledge, never examined before. With this research, I aim to fill this gap. It

(5)

contributes to both distress risk and behavioral finance literature by investigating if, besides accounting- and equity based measures, individual characteristics can affect the firm’s default risk.

In this thesis, I assess CEO’s overconfidence following the rationale of Malmendier & Tate (2005, 2008) and Malmendier, Tate & Yan (2011). The distinction between

overconfident and non-overconfident CEOs is based on the CEO’s option exercise decision. CEOs that exercise their options in the last year before expiration, despite the fact that the options were already deep in the money at the end of the previous year, are classified as “overconfident”. The reasoning is that CEOs receive large grants of stock and options as compensation, which means that they are overexposed to the idiosyncratic risk of their firms. According to Hall & Murphy (2002) risk averse CEOs should exercise stock options well before expiration due to the suboptimal concentration of their portfolio in company-specific risk.

In order to measure default risk, I use three different approaches. The first approach is called the “distance-to-default”, which is based on the Black & Scholes (1973) option pricing model and the structural default model of Merton (1974). The other two measures are

accounting based models of bankruptcy prediction developed by Altman (1968) and Ohlson (1980). Both models have been widely used in other research over the last decades.

The remainder of the thesis is organized as follows. In section 2, I derive empirical predictions linking overconfidence to default risk. Section 3 describes the data used in this research and the descriptive statistics. Section 4 discusses the methodology used in this study. Section 5 presents my main results. Section 6 tests the robustness of my main results. Finally, section 7 concludes and provides some recommendations for future research.

2. Hypotheses

Default risk is the probability that firms will be unable to meet the required payments on their debt obligations. That means that the risk of default increases with leverage

(Campbell, Hilscher, & Szilagyi, 2008). Overconfident CEOs overestimate the probability of future success of their firms. Due to this overestimation, overconfident managers think that their firm is more profitable/less risky than it actually is (Hackbarth, 2008). As a result, firms with overconfident CEOs tend to adopt more and riskier leverage (Malmendier, Tate, & Yan, 2011; Huang, Tan, & Faff, 2016). Therefore, my first hypothesis is as follows:

(6)

Hypothesis 1: the effect of leverage on default risk increases with CEO overconfidence. As a result, firms with overconfident CEOs have higher default risk.

However, several researchers shed a light on the positive sides of overconfident CEOs. According to Hirshleifer, Low & Teoh (2012) a possible solution to the puzzle why so many firms hire overconfident CEOs is that within innovation industries, greater confidence results in greater innovation. Furthermore, a risk averse overconfident CEO can be valuable to a firm, because it increases the willingness to engage in risky but profitable projects. A rational, risk averse CEO underinvest in projects relative to the shareholders’ optimum. Moderately overconfident risk-averse CEOs overestimate the precision of their private information and overreact to it. Thus, they invest in a project even when positive information about the project is such that they would not invest if they were rational (Gervais, Heaton, & Odean, 2011; Goel & Thakor, 2008). Therefore, overconfidence does not necessarily harm, but can even increase firm performance (Hirshleifer, Low, & Teoh, 2012). Since better firm performance reduces default risk (Campbell, Hilscher, & Szilagyi, 2008), my second hypothesis is:

Hypothesis 2: the effect of firm performance on default risk increases with CEO overconfidence. As a result, firms with overconfident CEOs have lower default risk.

The last hypothesis tests the overall effect of overconfidence on default risk. If there is a stronger leverage effect than firm performance effect, CEO overconfidence might increase default risk and vice versa. Therefore, the third hypothesis is:

Hypothesis 3: default risk increases/decreases with CEO overconfidence.

3. Data and Descriptive Statistic

3.1 The Data

I use several databases to construct my sample. Thomson Reuters insider filings provides information on the CEO’s exercise decisions and I use this information to construct my CEO overconfidence measure. Personal CEO information, like tenure, gender and age, is gathered from Execucomp. Stock returns are from CRSP and all accounting data are from Compustat.

(7)

Firms included in the sample have to correspond between Execucomp, Compustat and CRSP. Firm-years with missing data on either control variables or dependent variables are deleted. Financial firms (SIC 6000 – 6999) and regulated utilities (SIC 4900 – 4999) are dropped since they are heavily subjected to regulations and therefore the manager might manage the earnings differently (Malmendier & Tate, 2005). Furthermore, I drop all firms that are not listed on the S&P1500. The final sample consists of 683 CEOs and 4,742 firm-year observation between 2006 and 2016. Of these 683 CEOs, 307 are classified as

overconfident.

To test my hypothesis that overconfident CEOs affect the default risk of their firms, as dependent variables I use the distance to default (DD), Ohlson’s (1980) O-score and Altman’s (1968) Z-score. The measurement of CEO overconfidence, default risk and the associated control variables are discussed below.

3.2 Overconfidence

To determine CEO overconfidence, I follow the rationale of Malmendier & Tate (2005, 2008) and Malmendier, Tate & Yan (2011). They distinguish overconfident CEOs and non-overconfident CEOs by looking at their option exercise decisions. A CEO is classified as overconfident (dummy variable Longholder is equal to 1) if he ever holds, during the sample period, an option until the year of expiration even though the option is at least 40% in the money entering its final year. Holding options until its final year, even when it is deeply in the money, means that the CEO has been consistently overconfident about the company’s

prospects. The intuition behind this approach is that, regardless of the exact level of risk-aversion, diversification and wealth, a CEO is expected to exercise his stock options early if they are deep in the money in order to reduce his exposure to company-specific risk (Hall and Murphy, 2002; Huddart and Lang, 1996).

Note that the use of options in executive compensation packages drastically changed after the 1980s and early 1990s. Both the absolute level of option compensation and its relative level (percent of compensation paid via options) increased substantially from about 1995 to 2000 (Murphy, 2013). Consequently, the increase results in significantly more

opportunities to identify a CEO as overconfident nowadays than in the 1980s and early 1990s (Malmendier & Tate, 2015). In more recent literature the proportion of overconfident CEOs is around 40%, which is substantial higher than the 20% when using 1980-1994 Hall-Liebman data. Because the data on exercise decisions used by Malmendier & Tate (2005, 2008) is

(8)

outdated, Malmendier, Tate & Yan (2011) constructed four alternative overconfidence measures that correspond as closely as possible to their core measure. They compare insider trading data from Thomson Reuters insider filings database and personal portfolio data from Compustat’s Execucomp. Theoretically, both variables should exactly match, but actually they do not. However, there is no unequivocal answer to use one database in favor of the other. Since Execucomp does not provide information about transaction dates, it is not possible to exactly observe CEOs holding an option until the last year before expiration. Therefore, in this thesis, I construct my CEO overconfidence measure in the same way as Otto (2014) using the Thomson Reuters insider filings database.2

Thomson Reuters insider filings database is a collection of forms that are filled out by corporate insiders. Form 4 is the relevant source of information as it indicates changes in an insider’s ownership position. This could be the sale of stocks, purchase of stocks, an option grant or exercise or any other transaction that changes the ownership position. I start with all the Form 4 observations between January 2006 and December 2016 for all the S&P500 firms that have been listed for at least three consecutive years and I focus only on the exercise of stock options. To get rid of noisy data, I only keep observations with the cleanse indicators R, H, C, L, or I. Furthermore, I drop observations when the following items are missing: the person ID that identifies each CEO, the transaction date, the exercise date and the expiration date of the options. The time to expiration is calculated as the difference between the

expiration date of the options and the transaction date. I merge this database with stock data from the CRSP database so I can calculate the extent the option is “in-the-money” at the moment it enters its last year as the difference between the daily closing stock price a year prior to expiration and the exercise price divided by the exercise price.

Then, I assign each observation with the value one if the options were exercised within one year of their expiration date and were at least 40% in the money entering its final year. Otherwise, the observation takes the value zero. For each CEO, I take the sum of all observations. If this sum is larger or equal to one a CEO is classified as overconfident.

3.3 Default Risk

Previous research in the area of distressed firms assess default risk with different measures. Widely used proxies are accounting models like Altman’s (1968) Z-score and

2_{Despite the fact that both databases do not yield the exact same measures for overconfidence, the empirical}

(9)

Ohlson’s (1980) O-score. However, Vassalou & Xing (2004) argue that using Merton’s (1974) model to determine the “distance-to-default (DD)” is a more accurate approach. Their reasoning is that accounting models imply that firms with similar financial ratios will have similar likelihoods to default. In contrast to Merton’s model, where firms with similar levels of debt and equity, can have very different likelihoods to default if the volatility of their assets differ. Because in my opinion all models are proper and I want to robust my results, I use all three measures to assess default risk.

3.3.1 Altman’s Z-score and Ohlson’s O-score

Altman (1968) applied a discriminant analysis on a sample of 66 firms, half of which had filed for bankruptcy. He came up with a formula which may be used to predict the probability that a firm will go into bankruptcy within two years (Altman, 1968). The Z-score is a linear combination of five common business ratios, weighted by coefficients and looks as follows: (_),+ = 1.201),+ 23_),+ + 1.4 67_),+ 23_),++ 3.3 79:2_),+ 23_),+ + 0.6 =7_),+ 9>_),+ + 1.0 ?_),+ 23_),+

Where WC is working capital, TA is total assets, RE is retained earnings, EBIT is earnings before interest and taxes, ME is market value of equity, BL is the book value of liabilities and S stands for sales. Firms with a Z-score lower than 1.81 are in the “distress-zone” and firms with a Z-score higher than 2.99 are in the “safe-zone”. A Z-score in between indicates that a firm is in the “grey-zone”.

Ohlson (1980) developed an alternative scoring method for measuring bankruptcy. Where Altman (1968) used a sample of 66 firms, the O-score is based on a much wider pool of just over 2000 companies. As a result, it is significantly more accurate than the Z-score. The O-score uses nine linear combinations of coefficient-weighted business ratios and can be obtained as follows: @_),+ = −1.32 − 0.407 ln 23_),+ + 6.039>),+ 23_),+− 1.43 01),+ 23_),+ + 0.0757 1>),+ 13_),+− 1.72$ −2.37F:),+ 23_),+− 1.83 HH@_),+ 9>_),+ + 0.285I − 0.521 F:_),+− F:_),+JK F:_),+ + |F:_),+JK|

(10)

Where WC, TA and BL are equal to the values used in the Z-score. CL is current liabilities, CA is current assets, FFO is funds from operations and NI is net income. X and Y are dummy variables, where X takes value one if BL>TA and zero otherwise and Y equals one if the firm has a net loss for the last two years and zero otherwise. This O-score is then used to evaluate the probability of a firm’s failure (PF):

N_OP)QRST = UVJWXYST (1 + UVJWXYST₎

A PF score above 0.5 suggest that the firm will default within two years.

In order to calculate the Z-score, I need seven different variables. All accounting variables are gathered from Compustat. Total assets, book value of liabilities, retained earnings and sales are at, lt, re and sale, respectively. WC is calculated as the difference between the firms’ current assets (act) and current liabilities (lct). EBIT is the sum of net income (ni), interest and related expenses (xint) and income taxes (txt). The product of the number of shares outstanding (csho) and the year-end stock close price (prcc_f) is taken for ME. To construct the O-score, the only extra variable I need is FFO (fopo). After calculating the O-score for every firm-year, I determine the PF.

3.3.2 Distance to Default

The third measure uses the insights of Merton’s (1974) model, where the equity of a firm is viewed as a call option on the firm’s assets. The strike price of the call option is the book value of the firm’s liabilities with time to expiration equal to T. Because equity holders are residual claimants, the value of equity is zero when the value of the firm’s assets is less than the strike. Therefore, the market value and volatility of the firm’s equity will be given by the Black & Scholes (1973) formula for call options:

!_\ = !_]F ^_K − $UJS__{F ^}

` (1) and

a_\ = !]

(11)

Where

^

_K

=

bc de f g Sg

h ijei _

j_e _

, ^

`

= ^

K

− a

]

2,

!_\ is the market value of equity, !_] the firm’s asset value, a_\ the volatility of the firm’s equity, a_] the volatility of the firm’s asset, X the book value of liabilities, r is the risk-free rate and N is the cumulative density function of the standard normal distribution.

The probability of default is the probability that the market value of the firm’s assets will be less than the book value of the firm’s liabilities by the time debt matures (Crosbie, 1999). That is:

k_lTO,+ = k#mn ln !_{],+g_} ≤ ln $₊ |!_],+

And after rearranging it looks like:3

k_lTO,+ = F −ln !_],+ $₊ + p −a] ` 2 2 a_P 2 . (3)

Where p is the expected return on the firm’s asset, also called the drift rate. Different authors use different methods to calculate the drift rate. Vassalou & Xing (2004) estimate the drift rate as the mean change in ln (!_]). If the drift rate is negative, they replace it with the risk-free rate. According to Campbell, Hilscher & Szilagyi (2008) it is better to use a common expected return than a noisily estimated stock-specific number. I follow the method of Campbell, Hilscher & Szilagyi (2008), where the drift rate is estimated as p = # + 0.06. The DD is then defined as follows:

qq =ln !_],+ $₊ + p −a] ` 2 2 a_P 2 . (4) 3_{For detailed calculations see e.g. Crosbie (1999) or Vassalou & Xang (2004)}

(12)

In order to compute the DD, I need to estimate the asset value and volatility for each firm-year, neither of which are directly observable. I construct measures of these variables by solving equation (1) and (2) simultaneously.

To do so, I start with computing all observable variables using data from Compustat and CRSP. !_\ is computed by taking the product of the number of shares outstanding (csho) and the shares daily closing price (prc) at the end of each year. To estimate a_\, I take the standard deviation of the daily stock returns during the year. The book value of liabilities (lt) at the end of each year is taken for X. The risk-free rate is the 1-year T-bill rate observed at the end of each year. Following convention in the literature on the Merton model (Crosbie 1999, Vassalou and Xing 2004), I assume 2 = 1. What remains is two equations, (1) and (2), with two unknowns, namely !_] and a_]. As starting values for these unknowns, I use:

!_] = !_]+ $ and

a_] = !\ !_]a\

I iterate until I find values for !_] and a_] that are consistent with the observed values for !_\, X and a_\ (Campbell, Hilscher, & Szilagyi, 2008). Finally, I plug these values in equation (4) to calculate the DD for each firm-year. Default occurs when the ratio of the value of assets to debt is less than 1, or its log is negative. The DD tells us by how many standard deviations the log of this ratio needs to deviate from its mean for default to occur (Vassalou & Xing, 2004).

3.4 Control Variables

To minimize the possibility that my main results are driven by omitted variables I use a set of control variables based on the findings of Campbell, Hilscher & Szilagyi (2008). According to Campbell, Hilscher & Szilagyi (2008) firms with higher leverage, lower profitability, lower prices per share, more volatile past stock returns, lower cash holdings, lower market capitalization and lower Tobin’s Q are more likely to be financially distressed. All control variables are constructed using Compustat and CRSP data. Most authors measure total assets at book value. However, taking the market value of total assets as measure gives better explanatory power than the book value of assets. The market value of asset is obtained by adding the book value of liabilities (lt) to the market value of equities (ME) (Campbell, Hilscher, & Szilagyi, 2008). Profitability is measured as net income (ni)

(13)

relative to total market value of assets. Market capitalization is the product of the number of shares outstanding (csho) and the closing share price (prc). Leverage is measured as total liabilities relative to the market value of assets. As measure for cash holdings, I take the company’s ratio of cash and short-term assets (che) and market value of assets. Tobin’s Q is the ratio of the market value and the book value of assets. Two market based variables are based on the company’s stock performance. The volatility of the past stock returns is the standard deviation of each firm’s daily stock return over the past year in percentages. Finally, I calculate each firm’s log price per share, truncated above at $15. The truncation is to capture the tendency of distressed firms to trade at low prices per share, without reverse-splitting to bring price per share back into a more normal range (Campbell, Hilscher, & Szilagyi, 2008). In order to eliminate outliers, all the control variables are winsorized at a 5/95 percentile interval.

3.5 Descriptive Statistics

Table 1 provides descriptive statistics for the variables used in this study. The table is divided in a summary of the full sample and summaries of the overconfident and

non-overconfident subsamples. First interesting observation is that more non-overconfident CEOs have significantly higher mean and median DD and Z-score and significantly lower mean and median PF. Overconfident CEOs have a mean DD of 8.29, while non-overconfident CEOs have a mean DD of 7.81. The mean Z-score and PF for overconfident (non-overconfident) CEOs is 4.83 (4.44) and 0.15 (0.17), respectively. My first hypothesis is based on several papers that state that overconfidence is associated with higher leverage observations

(Malmendier, Tate, & Yan, 2011; Hackbarth, 2008; Ben-David, Graham, & Harvey, 2007). Surprisingly, in my sample the mean and median leverage are significant lower for

overconfident CEOs. However, Hirshleifer, Low & Teoh (2012) also observe lower leverage for overconfident CEOs in their sample and argue that this univariate relation is not

necessarily causal. Consistent with my second hypothesis I find a higher mean and median profitability and Tobin’s Q for overconfident CEOs.

I do not find a significant difference mean for liquidity. This changes when I use book value of assets instead of market value of assets. In that case, more overconfident CEOs have significant higher mean liquidity, which is consistent with the findings of Hirshleifer, Low & Teoh (2012) and Malmendier & Tate (2005). As discussed earlier, I choose to take the market value of total assets, because it gives better explanatory power than the book value of assets.

(14)

With respect to the other control variables, overconfident CEOs manage firms with greater Tobin’s Q, higher prices per share, higher market capitalization and lower stock volatility. The observations of Tobin’s Q and market capitalization are consistent with the findings of Hirshleifer, Low & Teoh (2012). However, the literature is not unambiguous about

overconfident CEOs and stock volatility. For example, Hirshleifer, Low & Teoh (2012) find a higher mean stock volatility for overconfident CEOs, in contrast to Otto (2014). That CEO overconfidence is associated with longer tenure corresponds with all the above-mentioned papers.

4. Method

A panel data regression is used to test my main hypotheses. In this research,

Longholder is equal to one if a CEO is classified as overconfident. The dependent variable is default risk. To robust my results, I use three different measures for default risk: DD, Z-score and PF. To avoid that the results are driven by omitted variables, I also include accounting and equity-market variables which are proven to affect firms’ default risk. In line with previous research on this topic, I include Fama-French 12 industry fixed effects and year fixed effects to control for time invariant industry level determinants and effects of hidden macroeconomic shocks. I do not include firm fixed effects, because my sample does not contain enough cases of overconfident and non-overconfident CEOs in the same firm, which leaves insufficient time-series variation to identify the effect of overconfidence. Another drawback for including firm fixed effects is the potential sample selection bias. By including fixed-firm effects, I am examining only those firms with multiple short-tenured CEOs (Malmendier & Tate, 2005; Huang, Tan, & Faff, 2016). All standard errors are clustered at firm level.

The first hypothesis is that the effect of leverage on default risk increases with CEO overconfidence. To test this hypothesis, I estimate the following regression:

6rst uU"sv#U_),+ = w + x_K>y_),++ x_`>y_),+ ∗ >U{_),+ + x_|>U{_),++ x_}1m~#mÄs_} + Å_),+

Where Risk measure is DD, PF or Z-score, LH is the Longholder variable and Lev is the market leverage of the firm. I predict x_` to be negative for the risk measures DD and Z-score and positive for PF if the leverage of an overconfident CEO has a more pronounced effect on default risk. Analogous to the first hypothesis, I test my second hypothesis as follows:

(15)

6rst uU"sv#U_),+ = w + x_K>y_),+ + x_`>y_),+∗ k#m_),+ + x_|k#m_),++ x_}1m~#mÄs_}+ Å_),+

and

6rst uU"sv#U_),+ = w + x_K>y_),++ x_`>y_),+∗ Ç_),+ + x_|Ç_),+ + x_}1m~#mÄs_}+ Å_),+

Where Pro and Q are firm performance measures and stand for profitability and Tobin’s Q, respectively. Profitability is measured as net income divided by market value of assets. Q is the ratio of the market value and book value of assets. Since better firm performance is associated with lower default risk, I predict the opposite effect of the first regression. That is, x_` is positive for the risk measures DD and Z-score and negative for PF if the effect of

profitability on default risk is more pronounced for overconfident CEOs. Finally, I test if CEO overconfidence affects default risk. To do so, I estimate the regression:

6rst uU"sv#U_),+ = w + x_K>y_),++ x_}1m~#mÄs_} + Å_),+

x_K shows if, and to what extent, CEO overconfidence influences firm’s default risk.

With respect to the control variables, profitability, Tobin’s Q, liquidity, price per share and market capitalization are predicted to have a positive sign. The predicted sign for leverage and stock volatility is negative. Note that these signs are the opposite for PF, since a lower PF indicates a lower default risk.

5. Empirical Analysis

I have hypothesized that the (negative) effect of leverage and the (positive) effect of firm performance on default risk increases with overconfidence. This section discusses the empirical results regarding these hypotheses. Finally, I check whether and to what extent overconfidence affects default risk.

5.1 Overconfidence, Leverage and Default Risk

Prior literature suggests that overconfidence results in higher and riskier leverage observations (Malmendier, Tate, & Yan, 2011; Huang, Tan, & Faff, 2016; Hackbarth, 2008). To test whether leverage of overconfident CEOs increase the default risk, I use three different

(16)

default measures as dependent variable, namely DD, PF and Z-score. Table 2 reports the estimation results from the regressions. The variable of interest is the interaction between overconfidence and leverage. The estimated coefficients on this interaction term show that the effect of leverage on default risk significantly increases with overconfidence, lending support to the first hypothesis. The coefficients on the interaction term are significant at the 5% level in column (1) and significant at the 10% level in column (2) and (3). The coefficient of -1.956 in column (1) is almost half of the single leverage effect, meaning that the effect of

overconfidence is substantial. The overconfidence effect in columns (2) and (3) is smaller; 17.81% and 22.60%, respectively.

Overall, the results suggest that, all else equal, an increase in leverage has a larger effect on default risk for firms with overconfident CEOs than for firms with

non-overconfident CEOs. As a result, default risk is on average more sensitive to leverage for firms with an overconfident CEO.

5.2 Overconfidence, Firm Performance and Default Risk

Hirshleifer, Low and Teoh (2012) wondered why so many firms hire overconfident CEOs, despite the usual presumption that overconfidence is undesirable. Their results suggest that overconfident CEOs not necessarily harm firm value and profitability. Moreover, they even find little evidence that overconfidence increases firm performance. Another potential benefit of overconfident CEOs is that they are less conservative, which makes it cheaper for firms to motivate him to pursue risky valuable risky projects (Gervais, Heaton, & Odean, 2011). Since better firm performances decreases default risk (Campbell, Hilscher, & Szilagyi, 2008), my second hypothesis is that the effect of firm performance on default risk is stronger for overconfident CEOs. As measures for firm performance I use profitability and Tobin’s Q. Table 3 shows that the effect of profitability on default risk is stronger for

overconfident CEOs. The estimated coefficients on the interaction term are positive for column (1) and (3) and negative for column (2), suggesting a stronger decrease in default risk for overconfident CEOs. The results are significant at the 1% level for DD and 5% level for Z-score. The coefficient on PF is insignificant. Overall, these results suggest that, all else equal, an increase in profitability has a larger effect on default risk for firms with an overconfident CEO.

Table 4 indicates that default risk also decreases with Tobin’s Q and again this effect is stronger for overconfident CEOs. The results are significant at the 10% level for DD and

(17)

PF and at the 1% level for Z-score, meaning that, all else equal, an increase in Tobin’s Q decreases default risk and this effect is significantly larger for overconfident CEOs. Both tables support the hypothesis that, on average, the positive effect of firm performance on default risk is larger for firms with an overconfident CEO.

5.3 Overconfidence and Default Risk

Above findings show opposite effects. On the one hand, the negative effect of leverage on default risk increases with overconfidence. On the other hand, the positive effect of firm performance on default risk also increases with overconfidence. If one effect appears to be stronger than the other, overconfidence increases (decreases) default risk, which is stated in my third hypothesis.

The results are reported in Table 5, where Longholder is the variable of interest. None of the estimated coefficients are significant, suggesting that overconfidence does not have a direct impact on default risk. The control variables are strongly significant, except for Log(price per share). Stock volatility is only significant in column (1). The coefficient signs of the control variables are as predicted aside from the Tobin’s Q in column (2) and

Log(market capitalization) in column (3).

The results are quite surprising, because the vast majority of literature on this topic emphasize the undesirable effects of overconfidence. However, the results in Table 5 suggest that on average overconfident CEOs do neither increase nor decrease default risk of their firms. Apparently, the downsides of overconfidence are offset by its benefits. This is consistent with the idea that overconfident CEOs have both advantages and disadvantages (Hirshleifer, Low, & Teoh, 2012). Moreover, above results could suggest that overconfidence can be (un)desirable in combination with specific firm-characteristics. I explore this

implication in the next section.

6. Robustness Checks

In this section I examine several robustness checks. First I divide the sample in leverage, profitability and Tobin’s Q tertiles. By doing so, I can check whether the results of section 5 are an average effect or if they are driven by specific subsets. Next, I consider an alternative interpretation that can explain the results.

(18)

6.1 Default Risk and Firm-Characteristics Tertiles

Table 6 reports the estimation results of the different leverage tertiles for each measure of default risk. Columns (1), (4) and (7) show the results for firms with the lowest leverage and columns (3), (6) and (9) show the results for firms with the highest leverage. The coefficients on the interaction term show significant estimations in column (3) and (6), with significant levels at 10% and 5%, respectively. The results indicate that overconfident CEOs do not affect default risk differently for firms with low levels of leverage, but when leverage is high they do. This makes intuitive sense, an increase in leverage will always affect the default risk of a firm, but when leverage is low, the consequences of overconfidence are not strong enough to affect the default risk differently compared to non-overconfidence. The overconfidence effect only comes in when firms already observe a high level of leverage. Table 7 shows the estimation results analogous to Table 6, but now the sample is divided in profitability tertiles. The coefficients on the interaction terms are all insignificant for columns (1)-(6). However, they become significant in column (8) and (9) at a significance level of 5% and 10%, respectively. The overconfidence effect of profitability on default risk from Table 3 looks more like an average effect than that it is explained by specific tertiles. Though, the effect on Z-score is mainly explained by the two highest tertiles.

Finally, Table 8 shows the results of the sample with Tobin’s Q tertiles. The

distribution of the significant tertiles is very smooth; with respect to the DD measure, only the first tertile gives a significant coefficient at the 5% level, in case of the PF measure, the first and second tertile gives significant effects at the 1% and 5% level respectively and for the Z-score only the highest tertile is significant at the 5% level. These outcomes suggest that the overconfidence effect of Tobin’s Q on default risk is an average effect and not dominated by firms with high or low Tobin’s Qs.

Dividing the sample in different tertiles gives a better understanding how

overconfidence affects default risk. The default risk of highly leveraged firms suffers the most from overconfidence. On the contrary, there is little evidence that the default risk of high profitability firms benefits from overconfidence. That could imply that overconfident CEOs can be valuable for firms with low leverage and high profitability. This implication is tested in Table 9. I have split the main sample in a subsample containing firms with low leverage and high profitability. Interestingly, Table 9 provides little evidence that overconfidence might result in lower default risk. The coefficient in column (3) is positive at the 10% significance level, meaning that on average the Z-score for firms with overconfident CEOs, low leverage and high profitability is 0.246 higher. I do not find that overconfidence

(19)

significantly increases default risk for firms with high leverage and low profitability, meaning that the effect is not two sided.

In summary, not all evidence is strongly significant, but there is no reason to assume that overconfident CEOs increase default risk. Moreover, there is even little evidence that in combination with specific firm-characteristics overconfidence can reduce default risk.

6.2 Alternative Interpretations

This section considers an alternative interpretation of the findings associated with endogenous matching between CEOs and firm characteristics. A possible concern about the results is the possibility that simultaneous causality is present. For example, better performing firms may sooner hire an overconfident CEO than one that is more conservative when it comes to taking risk (Graham, Harvey, & Puri, 2013). Also, Goel & Thakor (2008) find that boards of firms are more likely to end up with a pool of overconfident managers from which to choose a new CEO. Consequently, better firm performance might be the result of

overconfidence or overconfidence is more often associated with firms that are better performing.

One reason to believe that the results are not solely driven by matching is that all tests control for industry fixed effects (Hirshleifer, Low, & Teoh, 2012). To further address

matching effects, I follow the method of Hirshleifer, Low & Teoh (2012). I restrict my sample to focus on a subset of firm-years for which matching is likely to be less important. Any matching effects between CEO overconfidence and firm characteristics should be strongest when the CEO is first appointed. Therefore, I eliminate all the firm-years for which the CEOs tenure with the firm was 3 years or less and reexamine the overconfidence effects. Table 10 only shows the interaction term coefficients. The control variables are the same as in the corresponding earlier tables.

The coefficients on the Longholder * Leverage variable are comparable to the one found in Table 2, except that the coefficient in column (3) is not significant anymore. That means, that after deleting short-tenured CEOs, the overconfidence leverage effect is still there. The interaction between overconfidence and profitability is very similar to the results showed in Table 3. The significance level of column (1) changed in 5% instead of 1%, but the positive effect of profitability on default risk for overconfident CEOs is still observable. The coefficients of the interaction between overconfidence and Tobin’s Q are close to the one in Table 4 except for the coefficient in column (2), which has become insignificant. Another

(20)

possible limitation with respect to Tobin’s Q is that holding options, that are deeply in the money, might be related to a high Tobin’s Q. The market can pick up CEOs that refrain from exercising their options as a favorable signal, resulting in a higher Tobin’s Q. As a result, both variables could be related leading to omitted variable bias (Campbell, Gallmeyer, Johnson, Rutherford, & Stanley, 2011).

In summary, these tests are consistent with causality from CEO overconfidence on default risk, but I cannot exclude the presence of matching effects with certainty. Also, the results regarding Tobin’s Q might be upward biased as a result of omitted variables. A

possible solution regarding Tobin’s Q could be to use a different overconfidence measure that is not based on CEO’s option exercise decisions.

7. Concluding Remarks

In the recent past CEO overconfidence is a rapidly growing research area. Until now, the vast majority of researchers emphasize the negative roles of overconfident CEOs.

Overconfident CEOs tend to overestimate the returns they generate on their investments resulting in value destroying investments and acquisitions (Malmendier & Tate 2005, 2008). Also, they are more inclined to debt financing, because they think that their firm is less risky/more profitable than it really is. As a result, firms with overconfident CEOs are on average higher and riskier leveraged (Hackbarth, 2008; Huang, Tan, & Faff, 2016). On the other hand, some researchers shed a light on the positive roles of overconfidence.

Overconfident CEOs can be valuable to a firm, because they are better innovators and less conservative to engage in profitable but risky projects (Hirshleifer, Low, & Teoh, 2012; Gervais, Heaton, & Odean, 2011). However, if such traits result in higher or lower default risk is still a gap in the literature.

In an attempt to fill this gap, this thesis uses a proxy for CEO overconfidence based upon exercise decisions and three different measures for default risk, distance-to-default, O-score and Z-O-score. I find that over the 2006 – 2016 period, overconfident CEOs do not directly affect the default risk of their firms. However, when focusing on specific aspects that are related to default risk, the results show that overconfidence can influence default risk in different directions. While default risk is more sensitive to leverage of firms with

overconfident CEOs, default risk also appears to be more responsive to firm performance of overconfident CEOs. It seems to be that both effects offset each other such that on average overconfidence has no direct effect on default risk.

(21)

The overconfidence leverage effect is the strongest for highly leveraged firms. Also, there is little evidence that the overconfidence profitability effect is the strongest for high profitability firms. Given a low leveraged and high profitability firm, overconfidence might even result in a reduction of default risk. In any case, I find no sign that overconfident CEOs increase default risk.

There is a chance that the results are biased due to simultaneous causality. It is hard to exclude matching effects, meaning that firm-characteristics can play a role in appointing an overconfident CEO. I have tried to minimize the matching effects by measuring the effects of a subsample, where the first 3 years of an CEOs tenure are excluded. The effects of the subsample are similar to the effects of the main sample, which implies that there is at least some causality.

Nevertheless, the findings in this thesis are interesting and encouraging with respect to the question why so many firms hire overconfident CEOs. They are consistent with the idea that overconfidence has advantages as well as disadvantages. Furthermore, the results lay a good foundation for future research. Firstly, due to data restriction, the overconfidence proxy used in this thesis is limited to option exercise decisions. It can be interesting to evaluate the results with other overconfidence measures based on press-mentions or analyst’s earnings per share forecasts (Hirshleifer, Low, & Teoh, 2012; Otto, 2014). Secondly, in line with

Hirshleifer, Low & Teoh (2012) the results provide little evidence that there are specific firm-characteristics for which overconfidence can be valuable. Future studies can further examine if there are matching criteria such that overconfidence can be desirable. To deal with

endogeneity problems, one could consider using another methodology. For example,

instrument variables or an event study could give new insights. Finally, more research could focus on the positive effects of overconfidence. A lot is known about how overconfidence drives leverage decisions, but little is known about how profitability is driven by

overconfidence. It can be captivating to examine why overconfident CEOs have a positive influence on profitability.

(22)

Bibliography

Alicke, M. D. (1985). Global self-evaluation as determined by the desirability and controllability of trait adjectives (Vols. 49(6), 1621). Journal of personality and social psychology. Altman, E. I. (1968). Financial ratios, discriminant analysis and the prediction of corporate bankruptcy (Vols. 23(4), 589-609). The journal of finance. Ben-David, I., Graham, J. R., & Harvey, C. R. (2007). Managerial overconfidence and corporate policies (Vol. No. w13711). National Bureau of Economic Research. Bertrand, M., & Schoar, A. (2003). Managing with style: The effect of managers. The Quarterly Journal of Economics. Campbell, J. Y., Hilscher, J., & Szilagyi, J. (2008). In search of distress risk (Vols. 63(6), 2899-2939). The Journal of Finance. Campbell, T. C., Gallmeyer, M., Johnson, S. A., Rutherford, J., & Stanley, B. W. (2011). CEO optimism and forced turnover (Vols. , 101(3), 695-712). Journal of Financial Economics. Crosbie, P. (1999). Modeling default risk. Gervais, S., Heaton, J. B., & Odean, T. (2011). Overconfidence, compensation contracts, and capital budgeting (Vols. 66(5), 1735-1777). The Journal of Finance. Goel, A. M., & Thakor, A. V. (2008). Overconfidence, CEO selection, and corporate governance (Vols. 63(6), 2737-2784.). The Journal of Finance. Graham, J. R., Harvey, C. R., & Puri, M. (2013). Managerial attitudes and corporate actions (Vols. 109(1), 103-121). Journal of Financial Economics. Hackbarth, D. (2008). Managerial traits and capital structure decisions (Vols. 43(04), 843-881). Journal of Financial and Quantitative Analysis. Hall, B., & Murphy, K. (2002). Stock options for undiversified executives (Vols. 33, 3–42). Journal of Accounting and Economics . Hirshleifer, D., Low, A., & Teoh, S. (2012). Are overconfident CEOs better innovators? (Vols. 67(4), 1457-1498). The Journal of Finance. Huang, R., Tan, K. J., & Faff, R. W. (2016). CEO overconfidence and corporate debt maturity (Vols. 36, 93-110). Journal of Corporate Finance. Huddart, S., & Lang, M. (1996). Employee stock option exercises: an empirical analysis (Vols. 21, 5–43). Journal of Accounting and Economics . Löffler, G., & Posch, M. P. (2011). Credit risk modeling using Excel and VBA. John Wiley & Sons. Malmendier, U., & Tate, G. (2005). CEO overconfidence and corporate investment (Vols. 60(6), 2661-2700). The Journal of Finance. Malmendier, U., & Tate, G. (2008). Who makes acquisitions? CEO overconfidence and the market's reaction (Vols. 89(1), 20-43). Journal of Financial Economics. Malmendier, U., & Tate, G. (2015). Behavioral CEOs: The role of managerial overconfidence. (Vols. 29(4), 37-60). The Journal of Economic Perspectives. Malmendier, U., Tate, G., & Yan, J. (2011). Overconfidence and early-life experiences: the effect of managerial traits on corporate financial policies (Vols. 66(5), 1687-1733). The Journal of Finance. Merton, R. C. (1974). On the pricing of corporate debt: The risk structure of interest rates (Vols. 29(2), 449-470). The Journal of finance. Murphy, K. J. (2013). Executive Compensation: Where We Are, and How We Got There (Vol. Vol. 2A). Handbook of the Economics of Finance.

(23)

Ohlson, J. A. (1980). Financial ratios and the probabilistic prediction of bankruptcy (Vols. 109-131). Journal of accounting research. Otto, C. A. (2014). CEO optimism and incentive compensation (Vols. 114(2), 366-404). Journal of Financial Economics. Roll, R. (1986). The hubris hypothesis of corporate takeovers (Vols. 197-216.). Journal of business. Vassalou, M., & Xing, Y. (2004). Default risk in equity returns (Vols. 59(2), 831-868). The Journal of Finance. Weinstein, N. D. (1980). Unrealistic optimism about future life events (Vols. 39(5), 806.). Journal of personality and social psychology.

(24)

Appendix A

This appendix defines the variables used in the study. Accounting data is from Compustat, stock return data from CRSP, CEO personal information and exercise decisions from Thomson Reuters insider filings and Execucomp.

Variables Description Dependent variables Distance to default (DD) qq =ln !],+ $₊ + p −a] ` 2 2 a_P 2

The DD tells us by how many standard deviations the log of this ratio needs to deviate from its mean for default to occur Altman Z score (Z-score) ₍_),+ _{= 1.2}ÉÑÖ,Ü

_]Ö,Ü + 1.4 á\Ö,Ü _]Ö,Ü+ 3.3 \àâ_Ö,Ü _]Ö,Ü + 0.6ä\Ö,Ü àãÖ,Ü + 1.0 åÖ,Ü _]Ö,Ü

The Z-score gives an indication of the default risk of a firm. The higher the Z-score the lower the probability a firm will default. Probability to failure (PF)

N_OP)QRST = UVJWXYST (1 + UVJWXYST₎

PF is derived using Ohlson’s O-score. A PF above 0.5 indicates that a firm will default within two years

Independent variable

Longholder Options-based measure of CEO

overconfidence. Indicator variable equals to one if a CEO holds options until the last year before expiration and that are at least 40% in-the-money entering its last year, and zero otherwise.

Control variables

Leverage Ratio of book value of liabilities and market value of assets

Profitability Ratio of net income and market value of assets

Market capitalization The product of the number of shares outstanding and the closing stock price Liquidity Ratio of cash holdings and market value of

(25)

Tobin’s Q Ratio of market value of assets and book value of assets

Price per share The firm’s closing stock price, truncated at above at $15

Stock volatility Standard deviation of daily stock returns over the year, in percentage

Other variables

Gender Variable that equals one if a CEO is a male and zero if a CEO is a female

Age Age of the CEO in years

(26)

Table 1. Summary statistics

The table gives the means and medians of the variables used in this study. For the full sample, I also included the minimum and maximum values. The sample consists of all non-financial and non-utility S&P1500 firms in Execucomp from 2006 to 2016. To be included in the sample, firms are required to have accounting data from Compustat and stock returns from CRSP. A CEO is defined as being overconfident if he ever holds an option until the year of expiration and the option is 40% in the money entering its last year. T-tests (Wilcoxon-Mann-Whitney tests) are conducted to test for differences between the means (medians) for the firms with overconfident CEOs and firms with non-overconfident CEOs. *, **, and *** measure significance at the 10%, 5%, and 1% level, respectively.

Full Sample Non-overconfident

CEOs

Overconfident CEOs

(N=4742) (N=2408) (N=2334)

Variables Mean Median Min Max Mean Median Mean Median

Dependent variables DD 8.04 6.81 2.19 20.05 7.81 6.57 8.29**** 7.05*** Z-score 4.63 3.87 1.17 12.86 4.44 3.78 4.83**** 3.94*** Probability of failure 0.16 0.12 0.013 0.53 0.17 0.12 0.15**** 0.11*** Control variables Leverage 0.12 0.10 0 0.36 0.13 0.11 0.12**** 0.10*** Tobin’s Q 2.05 1.76 0.99 4.63 1.95 1.67 2.15**** 1.85*** Profitability 0.066 0.065 -0.06 0.18 0.064 0.062 0.069**** 0.066*** Market cap 12.393 2.627 20.99 382.421 11.960 2.268 12.837**** 3.033*** Liquidity 0.076 0.055 0.005 0.24 0.077 0.056 0.074**** 0.054***

Price per share 46.67 35.47 15 741.8 45.16 33.53 48.21**** 37.36***

Stock volatility 2.47 2.17 0.47 14.36 2.54 2.22 2.39**** 2.12***

Personal information

Gender 0.98 1 0 1 0.98 1 0.98**** 1***

Age 61.72 62 44 86 61.59 62 61.84**** 61***

(27)

Table 2. Regressions default risk, CEO overconfidence and leverage

The table presents results of regressions of default risk on the interaction of CEO overconfidence and leverage. Default risk is measured as DD, PF or Z-score. Longholder is an indicator variable equals to one if a CEO holds options until the last year and that are at least 40% in-the-money entering its last year.

Leverage is measured as the ratio of book value of debt to market value of assets. Definitions of the variables are described in Appendix A. All regressions include year and industry fixed effects, based on Fama-French 12 industry classification. Intercepts are not reported. Standard errors, clustered at the firm level, are reported in parentheses. *, **, and *** measure significance at the 10%, 5%, and 1% level, respectively. (1) (2) (3) DD PF Z-score Longholder 0.283* -0.0233** 0.418* (0.154) (0.010) (0.249) Longholder * Leverage -1.956** 0.127* -2.190* (0.834) (0.071) (1.320) Leverage -3.946*** 0.713*** -9.690*** (0.729) (0.060) (1.203) Profitability 2.831*** -0.720*** 13.21*** (1.069) (0.087) (1.567) Tobin’s Q 0.854*** 0.0588*** 1.145*** (0.084) (0.006) (0.131) Liquidity 2.275*** -0.328*** 3.226*** (0.732) (0.061) (1.143) Log(market capitalization) 0.106*** -0.0165*** -0.357*** (0.040) (0.003) (0.046)

Log(price per share) 0.145 0.00224 -0.044

(0.110) (0.009) (0.132)

Stock volatility -1.964*** -0.00265 -0.0438

(0.096) (0.003) (0.052)

Observations 4,667 4,516 4,354

R-squared 0.847 0.457 0.666

Industry fixed effects Yes Yes Yes

(28)

Table 3. Regressions default risk, CEO overconfidence and profitability

The table presents results of regressions of default risk on the interaction of CEO overconfidence and profitability. Default risk is measured as DD, PF or Z-score. Longholder is an indicator variable equals to one if a CEO holds options until the last year and that are at least 40% in-the-money entering its last year. Profitability is measured as the ratio of net income and market value of assets. Definitions of the variables are described in Appendix A. All regressions include year and industry fixed effects, based on Fama-French 12 industry classification. Intercepts are not reported. Standard errors, clustered at the firm level, are

reported in parentheses. *, **, and *** measure significance at the 10%, 5%, and 1% level, respectively.

(1) (2) (3) DD PF Z-score Longholder -0.280** -0.00106 -0.238 (0.126) (0.0122) (0.188) Longholder * Profitability 4.877*** -0.108 5.858** (1.833) (0.123) (2.382) Profitability 0.503 -0.665*** 10.40*** (1.401) (0.103) (1.889) Leverage -4.897*** 0.772*** -10.76*** (0.527) (0.048) (0.908) Tobin’s Q 0.855*** 0.0582*** 1.149*** (0.083) (0.006) (0.129) Liquidity 2.259*** -0.323*** 3.232*** (0.726) (0.061) (1.136) Log(market capitalization) 0.104*** -0.0164*** -0.360*** (0.040) (0.003) (0.045)

(0.109) (0.009) (0.132)

(0.095) (0.003) (0.052)

R-squared 0.847 0.456 0.668

(29)

Table 4. Regressions default risk, CEO overconfidence and Tobin’s Q

The table presents results of regressions of default risk on the interaction of CEO overconfidence and Tobin’s Q. Default risk is measured as DD, PF or Z-score. Longholder is an indicator variable equals to one if a CEO holds options until the last year and that are at least 40% in-the-money entering its last year. Tobin’s Q is measured as the ratio of market value and book value of assets. Definitions of the variables are described in Appendix A. All regressions include year and industry fixed effects, based on Fama-French 12 industry classification. Intercepts are not reported. Standard errors, clustered at the firm level, are reported in parentheses. *, **, and *** measure significance at the 10%, 5%, and 1% level, respectively.

(1) (2) (3) DD PF Z-score Longholder -0.381* 0.0224 -0.767*** (0.216) (0.019) (0.292) Longholder*Tobin’s Q 0.208* -0.0150* 0.453*** (0.120) (0.009) (0.167) Tobin’s Q 0.755*** 0.066*** 0.916*** (0.104) (0.008) (0.171) Market leverage -4.826*** 0.769*** -10.63*** (0.524) (0.048) (0.869) Profitability 2.795*** -0.718*** 13.31*** (1.062) (0.086) (1.545) Liquidity 2.258*** -0.329*** 3.386*** (0.721) (0.061) (1.120) Log(market capitalization) 0.103** -0.016*** -0.368*** (0.040) (0.003) (0.045)

(0.110) (0.009) (0.132)

(0.095) (0.003) (0.051)

R-squared 0.847 0.458 0.670

(30)

Table 5. Overconfident CEOs and default risk

The table presents results of regressions of default risk on CEO overconfidence. Default risk is measured as DD, PF or Z-score. Longholder is an indicator variable equals to one if a CEO holds options until the last year and that are at least 40% in-the-money entering its last year. Definitions of the variables are described in Appendix A. Note that the predicted signs are the opposite for PF. All regressions include year and industry fixed effects, based on Fama-French 12 industry classification. Intercepts are not reported. Standard errors, clustered at the firm level, are reported in parentheses. *, **, and *** measure significance at the 10%, 5%, and 1% level, respectively.

(1) (2) (3)

Predicted sign DD PF Z-score

Longholder +/- 0.0428 -0.00831 0.148 (0.086) (0.008) (0.127) Leverage - -4.867*** 0.771*** -10.69*** (0.541) (0.049) (0.920) Profitability + 2.737** -0.715*** 13.08*** (1.078) (0.087) (1.551) Tobin’s Q + 0.868*** 0.0579*** 1.161*** (0.084) (0.006) (0.131) Liquidity + 2.143*** -0.321*** 3.101*** (0.737) (0.061) (1.139) Log(market capitalization) + 0.108*** -0.0165*** -0.356*** (0.040) (0.003) (0.046)

Log(price per share) + 0.132 0.00304 -0.0576

(0.110) (0.009) (0.134)

Stock volatility - -1.961*** -0.00290 -0.0394

(0.096) (0.003) (0.052)

R-squared 0.846 0.455 0.665

(31)

Table 6. Regressions default risk, overconfidence and leverage tertiles

The table presents results of regressions of default risk on CEO overconfidence and profitability. Default risk is measured as DD, PF or Z-score. Longholder is an indicator variable equals to one if a CEO holds options until the last year and that are at least 40% in-the-money entering its last year. Leverage is measured as the ratio of book value of debt to market value of assets. Definitions of the variables are described in Appendix A. Sample is divided in tertiles, where the first tertile (columns 1, 4 and 6) contains the firms with the lowest leverage and the last tertile (columns 3, 6 and 9) the firms with highest leverage. All regressions include year and industry fixed effects, based on Fama-French 12 industry classification. Intercepts are not reported. Standard errors, clustered at the firm level, are reported in parentheses. *, **, and *** measure significance at the 10%, 5%, and 1% level, respectively.

(1) (2) (3) (4) (5) (6) (7) (8) (9)

DD DD DD PF PF PF Z-score Z-score Z-score

Longholder 0.214 0.580 0.726* -0.0212* -0.0453 -0.0721** 0.589 0.241 0.317 (0.233) (0.373) (0.418) (0.011) (0.032) (0.036) (0.357) (0.328) (0.233) Longholder*Leverage -5.243 -4.147 -2.868* 0.115 0.407 0.297** -12.02 -1.347 -1.135 (7.440) (3.414) (1.557) (0.378) (0.270) (0.141) (9.164) (2.571) (0.886) Leverage -14.44*** -2.866 -1.898* 0.977*** 0.555** 0.538*** -40.87*** -12.42*** -4.400*** (5.143) (2.320) (1.111) (0.336) (0.223) (0.120) (7.328) (1.942) (0.712) Profitability 4.677*** 4.016*** -0.674 -0.756*** -0.774*** -0.834*** 18.88*** 12.88*** 8.135*** (1.510) (1.386) (1.459) (0.126) (0.106) (0.130) (2.418) (1.105) (0.675) Tobin’s Q 0.721*** 0.767*** 0.572*** 0.0245*** 0.0934*** 0.161*** 1.257*** 0.437*** -0.0432 (0.110) (0.103) (0.182) (0.008) (0.007) (0.024) (0.156) (0.095) (0.133) Liquidity 3.015** -1.574* 0.551 -0.392*** -0.153* -0.388*** 3.469 0.206 0.354 (1.280) (0.930) (0.908) (0.088) (0.081) (0.114) (2.162) (0.852) (0.548) Log(market capitalization) 0.110 0.144*** 0.285*** -0.00991** -0.0186*** -0.0305*** -0.300*** -0.241*** -0.119*** (0.0807) (0.044) (0.057) (0.004) (0.004) (0.005) (0.110) (0.036) (0.031)

Log(price per share) 0.271 0.0656 0.323** 0.0121 -0.00356 -0.0126 0.397 0.117 0.0176

(0.235) (0.111) (0.129) (0.001) (0.011) (0.015) (0.272) (0.109) (0.096) Stock volatility -2.075*** -2.524*** -1.368*** 0.003 0.000 -0.003 -0.143* -0.139*** -0.0496*

(0.197) (0.104) (0.106) (0.004) (0.004) (0.005) (0.081) (0.041) (0.027)

Observations 1,556 1,556 1,555 1,540 1,533 1,443 1,373 1,525 1,456

R-squared 0.833 0.886 0.850 0.353 0.386 0.475 0.582 0.553 0.588

Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes

(32)

Table 7. Regressions default risk, overconfidence and profitability tertiles

The table presents results of regressions of default risk on CEO overconfidence and profitability. Default risk is measured as DD, PF or Z-score. Longholder is an indicator variable equals to one if a CEO holds options until the last year and that are at least 40% in-the-money entering its last year. Profitability is measured as the ratio of net income and market value of assets. Definitions of the variables are described in Appendix A. Sample is divided in tertiles, where the first tertile (columns 1, 4 and 6) contains the firms with the lowest profitability and the last tertile (columns 3, 6 and 9) the firms with highest profitability. All regressions include year and industry fixed effects, based on Fama-French 12 industry classification. Intercepts are not reported. Standard errors, clustered at the firm level, are reported in parentheses. *, **, and *** measure

significance at the 10%, 5%, and 1% level, respectively.

(1) (2) (3) (4) (5) (6) (7) (8) (9)

DD DD DD PF PF PF Z-score Z-score Z-score

Longholder -0.147 -0.436 0.0940 -0.00369 0.0455 -0.00324 -0.120 -1.019* -0.922 (0.112) (0.518) (0.679) (0.013) (0.035) (0.033) (0.182) (0.571) (0.610) Longholder*Profitability 2.932 9.167 0.470 -0.0933 -0.816 -0.0189 3.634 18.04** 9.284* (3.193) (8.012) (5.593) (0.311) (0.526) (0.253) (4.995) (8.879) (5.039) Profitability 0.617 8.171 2.276 -0.788*** -0.0415 -0.278* 10.43*** -2.081 8.756** (1.946) (5.965) (3.677) (0.188) (0.431) (0.164) (2.587) (6.694) (3.396) Leverage -3.884*** -5.637*** -10.07*** 0.665*** 0.765*** 1.220*** -8.189*** -11.19*** -21.59*** (0.583) (0.707) (1.174) (0.062) (0.062) (0.082) (0.758) (0.931) (1.576) Tobin’s Q 0.329*** 0.901*** 0.697*** 0.0778*** 0.0624*** 0.0379*** 0.784*** 1.263*** 1.317*** (0.122) (0.112) (0.093) (0.013) (0.010) (0.006) (0.223) (0.186) (0.119) Liquidity 1.328 3.396*** -1.755 -0.371*** -0.308*** -0.414*** 2.778** 5.702*** 3.458** (0.893) (1.011) (1.409) (0.098) (0.066) (0.075) (1.231) (1.452) (1.654) Log(Market capitalization) 0.143** 0.133*** 0.0706 -0.0144** -0.0183*** -0.0199*** -0.302*** -0.360*** -0.345*** (0.061) (0.051) (0.064) (0.006) (0.004) (0.003) (0.058) (0.046) (0.064)

Log(Price per share) 0.352** -0.0583 -0.0736 0.00176 0.000296 0.00489 0.0240 0.0874 -0.0531

(0.138) (0.111) (0.202) (0.015) (0.010) (0.011) (0.202) (0.136) (0.183)

Stock volatility -1.194*** -2.489*** -2.421*** 0.00294 0.00154 -0.0106*** -0.193*** -0.131* 0.0910

(0.110) (0.106) (0.202) (0.005 (0.004) (0.003) (0.068) (0.069) (0.060)

Observations 1,555 1,552 1,560 1,478 1,505 1,533 1,449 1,467 1,438

R-squared 0.858 0.879 0.845 0.379 0.440 0.563 0.504 0.595 0.715

Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes

Do Overconfident CEOs Affect Default Risk? It Depends