Master Thesis
MSc BE, Finance track
University of Amsterdam
Are Overconfident CEOs More Likely to be Replaced
by Non-Overconfident CEOs?
Berit Säde
Student number: 10831223
Supervisor: Tolga Caskurlu
2
Statement of Originality
This document is written by Student Berit Säde who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Table of Contents
ABSTRACT ... 4
Introduction ... 5
1. Literature Review... 7
1.1 Managerial Hubris and Acquisitions ... 7
1.2 Empirical Predictions ... 11
2. Methodology ... 13
2.1 Measure of Overconfidence ... 13
2.2 Definition of a Forced Turnover and Replacement of a CEO ... 14
2.3 Regressions and Variables ... 14
2.3.1 Logistic regression ... 14
2.3.2 Event Study Methodology ... 15
2.3.3 Performance Measures ... 16
2.3.4 Research & Development expenses ... 16
2.3.5 CEO Characteristics ... 17
3. Sample and Descriptive Statistics ... 18
3.1 Sample ... 18
3.2 Descriptive Statistics ... 19
4. Empirical Results ... 21
4.1 CEO Replacement with Non-Overconfident CEOs ... 21
4.2 Innovation and Turnovers ... 26
5. Robustness Checks ... 30
5.1 Hazard Model ... 30
5.2 CEO Ownership and Replacement with Non-Overconfident CEO ... 32
5.3 R&D Expenditures and Accounting Variables ... 34
Conclusion ... 37
4
ABSTRACT
The paper investigates the relation between overconfident Chief Executive Officers (CEOs) and their possible forced replacement with non-overconfident CEOs after mergers and acquisitions. In a sample over years 1995-2014, it is found that overconfident CEOs who engage in value destructive acquisitions face higher probability of being replaced by a non-overconfident CEO than already non-non-overconfident executives. This indicates that boards pay extra attention to make sure that the replacement CEO is not overconfident, when they already have experience with an overconfident CEO. Therefore, the results seem to suggest that boards can distinguish between overconfident and non-overconfident CEOs, at least when they have previously worked with an overconfident one. Moreover, this probability increases with value-destructive acquisitions. Furthermore, in innovative sectors with high Research & Development (R&D) expenditures, the probability of being replaced by a non-overconfident CEO is higher than in less innovative industries.
5
Introduction
Previous literature (Malmendier and Tate 2008, Doukas and Petmezas 2007, Campbell et al 2011) provides evidence that managerial overconfidence can lead to making bad investment decisions. Moreover, it is shown that these investment decisions are not in the best interests of the shareholders and tend to decrease firm value. Malmendier and Tate (2008) provide examples of such bad investment decisions by showing that overconfident CEOs tend to engage in more mergers and acquisitions than rational CEOs. Moreover, authors show that overconfident CEOs are more likely to do diversifying acquisitions that are found to be particularly value-destroying. Thus, there is strong evidence that overconfident CEOs act in ways that destroy shareholder value. If the board of directors acts in the interests of shareholders, they should fire overconfident CEOs and try to hire non-overconfident ones.
In addition, Campbell et al (2011) show that overconfident CEOs have a higher probability of forced turnover than rational CEOs. Thus, overconfident CEOs are likely to be fired because of their overconfidence. However, previous literature has not investigated whether boards of directors can distinguish between overconfidence and non-overconfidence, especially before hiring a new CEO. If an overconfident CEO is simply replaced by another overconfident CEO, it provides little value to the shareholders. However, if board of directors can distinguish between the personality traits of CEOs, it should definitely lead to the replacement of overconfident CEOs with non-overconfident CEOs. Additionally, it seems plausible that boards of directors who already have experience with an overconfident CEO are better equipped to identify whether the replacement CEO is overconfident or not. Thus the current paper explores whether overconfident CEOs face a higher probability of being replaced by non-overconfident CEOs. Moreover, the paper examines if the probability increases with more value-destroying acquisitions.
In this Master thesis I hypothesize that, after a forced turnover, overconfident CEOs are more likely replaced by non-overconfident CEOs than CEOs who are not overconfident. This is consistent with the prediction that boards of directors act in the interest of shareholders and should replace a CEO that is not maximizing firm value with a one who is. Additionally, if this hypothesis is correct, it would provide evidence that the board of directors can distinguish between overconfident and non-overconfident CEOs, at least when they already have experience with overconfidence. This Master thesis also hypothesizes that the probability of
6
overconfident CEO being replaced by a non-overconfident CEO increases with value-destroying acquisitions. The results of the paper confirm the first two hypotheses and suggest that overconfident CEOs face higher probability of being replaced by a non-overconfident CEO than already non-overconfident CEOs. This probability increases with more value-destroying mergers or acquisitions. Also, the results indicate that boards of directors perceive overconfidence as a negative characteristic. The results reveal that overconfident CEOs are 15%-25% more likely to be replaced by overconfident CEOs than their already non-overconfident peers within five years after acquisitions. This probability further increases by 7%-17% when the acquisition is value destructive.
Existing literature argues that overconfidence can be beneficial in innovative industries (Hirshleifer et al 2012, Galosso and Simcoe 2011). Surprisingly, the results of the paper suggest that overconfident CEOs are more likely to be replaced by non-overconfident CEOs in innovative industries after bad acquisitions. Therefore, in high R&D sectors overconfident CEOs are actually more likely to be replaced by non-overconfident CEOs than in other sectors. This contradicts with the papers cited earlier and seems to suggest that overconfidence is not a favourable personality trait in innovative industries. The possible explanations for this surprising result are outlined in the results section.
In order to provide robustness to the results and deal with potential endogeneity issues, several regressions are estimated with extra control variables. The results are robust to using the hazard models or including Tobin’s Q, book leverage, sales growth and several corporate control mechanism measures as control variables in the regressions.
This paper contributes to the growing literature of forced turnovers by examining whether the succeeding CEOs are overconfident or not. Furthermore, the paper provides evidence that boards of directors have at least some ability to perceive if a CEO is overconfident or not. Moreover, when boards of directors have experience with overconfident CEOs, they prefer to hire new CEOs who are non-overconfident. The paper is structured as follows. The first section introduces several empirical and theoretical research papers that provide an overview of the hubris hypothesis and also the background on how overconfidence affects firm value. Furthermore, the first section describes the empirical predictions of this paper. Section two provides a detailed explanation of the research methodology, whereas section three presents the data and the descriptive statistics. Section four outlines the empirical results of the paper and section five deals with several robustness checks.
7
1. Literature Review
1.1 Managerial Hubris and Acquisitions
Existing theories and empirical papers provide evidence of several intriguing relationships between acquisition performance and overconfidence of Chief Executive Officers (CEOs). When it comes to mergers and acquisitions, an important characteristic to note is that they tend to be value destructive for the shareholders of the acquiring company. Rau and Vermaelen (1998) provide evidence that bidders underperform firms with similar portfolios when engaging in acquisitions. Moreover, Moeller et al (2003) examine U.S data from 1980-2001 and find that, on average, the shareholders of the acquiring firm lose $25.2 million on the announcement of the acquisition. Overall, acquiring firms often face negative impact on stock performance. Meanwhile, acquisitions tend to be good and value-enhancing for the target companies. The explanations of why acquisitions tend to have a negative effect on acquirers are controversial. Officer et al (2009) suggest that returns depend on how the acquirer finances the acquisitions. Masulis et al (2007) argue that corporate governance affects acquirer returns and Bargerona et al (2008) show that acquirers pay higher premiums for public targets than for private targets. Furthermore, it is remarkable that the negative effect on acquiring company is greater if the CEO of the bidder is overconfident (Malmendier and Tate 2008). Overly confident CEOs tend to allocate achievement to their abilities, without recognizing the many other factors leading to success. Billett and Qian (2008) show that CEOs who make successful acquisitions will likely become overconfident due to attribution of success to abilities. Moreover Billett and Qian (2008) argue that overconfidence likely increases with CEOs age.
Roll (1986) is one of the first authors to explore the effects of managerial overconfidence and his hubris hypothesis suggests that overconfident CEOs engage in acquisitions with an excessive optimism about their ability to create value. This optimism may lead to negative consequences. Several authors (Malmendier and Tate 2008, Doukas and Petmezas 2007) provide empirical evidence to hubris hypothesis and show that acquisitions of overconfident CEOs tend to be more value-destroying than acquisitions made by their rational peers. Malmendier and Tate (2008) explain this phenomenon by showing that overconfident
8
CEOs make more diversifying mergers and acquisitions, which are particularly value-destroying. Furthermore, the authors argue that in such diversifying acquisitions overconfident managers fail to estimate potential synergies and thus overestimate the future opportunities. In addition Rhodes-Kropf and Viswanthan (2004) suggest that overconfidence is even greater during bull markets and this leads to even higher probability of overestimating synergies and bidding too much.
If CEOs are too optimistic they may take excessive risk, which can lead to positive outcomes in a favourable business climate, but can also be extremely value destructive. Additionally, excessive risk taking seems to be associated with the age of CEOs. Serfling (2014) shows that risk-taking behaviour decreases with age. Therefore, younger CEOs tend to take more risk while their older peers are more cautious. This can possibly be caused by the lack of experience. Young CEOs, who have not faced failures yet, may underestimate the risk or overestimate their ability. However, with age and more experience they realize that excessive risk can be value destroying and therefore become more cautious about possible risks. However, becoming slightly more cautious still does not mean that the CEO is not overconfident. Evidence reveals that overconfident CEOs cannot realistically evaluate potential synergies and seek too much risk. As an example, Brown and Sarma (2007) find that CEO overconfidence is an important determinant in explaining the firm’s decision to engage in acquisitions. In addition, the authors suggest that overconfident CEOs are often dominant and thus have a remarkable impact on boards of directors. Malmendier and Tate (2008) also show evidence that overconfident CEOs are 65% more likely to make an acquisition. Moreover, they reveal that the market reaction to an acquisition announcement by an overconfident CEO is remarkably more negative than for a non-overconfident CEO. Therefore, markets seem to anticipate that overconfident CEOs make worse acquisitions.
Overall, overconfidence tends to have a negative impact on the performance and in most cases, overconfidence does not provide value for the shareholders. So, it is essential to know what actions are taken by the representatives of the shareholders - the board of directors - to eliminate the negative impact on the firm value. Since all companies want to create value in the long run, it is important to hire and keep CEOs who contribute to the value maximizing strategy. Therefore, assuming that the board of directors acts in the interests of the shareholders, they should try to hire non-overconfident CEOs. However, it is argued that in some cases overconfidence could be beneficial (Hirshleifer et al 2012, Galosso and Simcoe 2011) and saying that the board of directors always wants to hire non-overconfident CEOs is a very strong statement. However, if the board of directors has just fired an overconfident CEO,
9
they are clearly not satisfied with his performance and should be looking to replace him with a non-overconfident one. One way to identify cases in which CEOs are fired is to look at value destroying investment decisions. Value destruction is especially relevant in the context of acquisitions, which is one of the most important and substantial investments for a company.
Several authors have found that boards of directors punish poor performance by firing CEOs (Parrino 1997, Murphy and Zimmerman 1993, Huson et al 2003, Warner et al 1988). The relation between stock performance and non-routine CEO turnovers seems to be especially strong (Warner et al 1988, Lehn and Zhao 2006, Campbell et al 2011) and the probability of forced turnover increases when stock performance decreases. Lehn and Zhao (2006) have found that 47% of CEOs of acquiring companies are being replaced within 5 years after the value-destroying acquisition. However, this percentage also includes external control mechanisms, such as bankruptcy and hostile takeovers. Campbell et al (2011) have differentiated between overconfidence and non-overconfidence and show that when boards of directors truly act in the best interests of shareholders, then either too optimistic or too little optimistic CEOs have a higher probability of being replaced than their rational peers. Thus, there is strong evidence that CEOs are fired after bad decisions, such as value destroying acquisition, and that overconfident CEOs have a higher probability of being fired. However, existing literature has not investigated who succeeds the fired CEOs and whether the replacement CEOs are themselves overconfident or not.
There is also extensive research on the relationship between corporate governance and non-routine turnovers (Masulis et al 2007, Huson et al 2004). Masulis et al (2007) provide evidence that managers, who are pressured by several corporate control mechanisms, make better acquisitions. One important factor is ownership structure. Firms, which have high inside ownership, tend to have more motivated boards that align the interest of shareholders and managers (Dalton et al 2003). However, high level of ownership may lead to the increased power of an executive and this may enable them to become entrenched (Morck et al 1988).
Thus, there is evidence that bad bidders are fired, but there is no evidence if boards of directors distinguish between overconfident and non-overconfident CEOs when hiring a new CEO. If a CEO is fired and afterwards replaced by another overconfident CEO, then it shows that boards cannot perceive overconfidence. Moreover, it would have important implications for shareholders as replacing an overconfident CEO with another overconfident CEO does no good for the shareholders. If anything, the new overconfident CEO, who is not familiar with the firm, could destroy even more value.
10
Although, many authors (Malmendier and Tate 2008, Campbell et al 2011, Doukas and Petmezas 2007) suggest that executive hubris is harmful, it should be noted that overconfidence can, at least to some extent, be a positive trait. Galosso and Simcoe (2011) argue that overconfident CEOs expect higher rate of success and lower probability of failure and, therefore, they pay more attention to innovation. Moreover, overconfidence can mitigate the underinvestment problem that exists among rational CEOs because overconfident managers are more likely to identify innovative projects to invest in (Goel and Thakor 2008). Overconfident CEOs are more likely to see new opportunities or even increase firm value because they are willing to take risks. Moreover, Goel and Thakor (2008) argue that overconfident managers, who have made value destructive investments, still have higher probability of being promoted to CEO than rational managers. However, this holds for managers who are not excessively overconfident. Thus, there is evidence that in some cases, overconfidence can be beneficial.
Gervais et al (2011) argue that managerial overconfidence is more attractive for some companies, since overconfidence enables managers to dedicate more thoroughly on gathering information about projects. Furthermore, all industry sectors have different needs and requirements. Therefore, firms are hiring CEOs with different characteristics. Some companies are seeking for a CEO who is willing to take more risks than a rational CEO would. On the other hand, there are companies that strongly prefer rationality and little risk-taking attitude from their management. This type of preference is related to industry specific factors and growth potentials. Several authors show that overconfidence is a beneficial personality trait in rapidly growing innovative sectors (Galosso and Simcoe 2011, Tang et al 2012, Hirshleifer et al 2012). Innovative industries include riskier growth opportunities and therefore the risk-taking characteristic might be beneficial. Thus, it seems that innovative industries benefit more from overconfidence. In such case, it might be that boards of directors are willing to keep the CEO who occasionally makes value-destroying acquisitions as suggested by Goel and Thakor (2008). Alternatively, if overconfidence is beneficial in these industries but a CEO is still fired, the board of directors should try to make sure that also the replacement CEO is overconfident. Again, this should be more true for companies that already had overconfident CEOs as they should be better equipped to identify signs of overconfidence.
However, there is also another theory about overconfidence in innovative industries. Defond and Park (1999) suggest that in competitive and rapidly developing industries the frequency of CEO turnovers is higher. This might be because in these industries one wrong investment decision can lead to an instant loss in competitiveness, which may be hard to regain. In such cases, it is natural that boards become more risk-averse and want to replace an
11
overconfident CEO with a non-overconfident one. Although this is also a plausible theory, there seems to be more evidence that overconfidence is indeed a positive trait in innovative sectors.
1.2 Empirical Predictions
Existing theoretical and empirical studies do not investigate if boards of directors truly believe that overconfidence is a negative characteristic. Furthermore, there is still no evidence whether boards of directors can distinguish between overconfidence and non-overconfidence. If, after an acquisition, boards of directors decide to replace overconfident CEO with a non-overconfident CEO, rather than another non-overconfident one, it shows that the boards can perceive overconfidence.
All the aforementioned theoretical and empirical papers have led to a question: who replaces an overconfident CEO that is fired? The paper aims to find if overconfident CEOs are more likely to be replaced by non-overconfident CEOs than their non-overconfident peers. The paper uses acquisitions to identify value-destructive events that have led to the CEO being fired in the first place. Lehn and Zhao (2006) show that CEO turnovers occur after non-successful acquisitions. However, the paper does not differentiate between overconfidence and overconfidence. Campbell et al (2011) have distinguished between overconfidence and non-overconfidence. However, they do not examine the succeeding CEOs.
Furthermore, there is evidence in previous literature (Goel and Thakor (2008) that overconfident CEOs can help to solve the underinvestment problem in risky projects. Moreover, Hirshleifer et al (2012) and Galosso and Simcoe (2011) suggest that overconfident CEOs create more innovation growth and therefore perform better in innovative industries. This might indicate that boards prefer overconfidence in these rapidly growing industries.
The previously discussed theories and empirical papers lead to the three following hypotheses:
1. After making an acquisition, overconfident CEOs are more likely to be replaced by non-overconfident CEOs than their non-overconfident peers.
2. The probability of being replaced by a non-overconfident CEO increases with more value-destructive acquisitions as boards become more risk-averse.
12
3. Overconfident CEOs are less likely to be replaced by non-overconfident CEOs than their non-overconfident peers in innovative sectors after value-destructive acquisitions.
This paper contributes to the growing literature by studying the overconfidence of replacement CEOs. Firstly, this paper investigates whether boards of directors can identify overconfidence by replacing overconfident CEOs with non-overconfident ones. Secondly, the paper studies if the boards of directors become more risk averse after more value-destructive acquisitions. Finally, the paper examines whether there is evidence that overconfidence is actually perceived to be a positive characteristic in innovative sectors.
13
2. Methodology
2.1 Measure of Overconfidence
Overconfidence is not a easily observable characteristic of a CEO. Therefore, measuring the extent of overconfidence provides some challenges. Previous literature (Malmendier and Tate 2005, Hall and Murphy 2002, Core and Guay 2002) provides various models and approaches for measuring overconfidence. One of the widely used models is developed by Malmendier and Tate (2005), which is based on the execution of stock options. Malmendier and Tate (2005) have adjusted the model of Hall and Murphy (2002) and define that a CEO is overconfident when after the vesting period (usually four years) CEOs hold stock options that are at least 67% in the money. This indicates that CEOs hold stock options that are high in the money and should be executed when a CEO is not overly optimistic about the future stock performance. The aforementioned approach for measuring overconfidence has also been used in this paper. Another model that has been widely used is based on net stock purchases, however, the results are similar to the results of stock execution model (Malmendier and Tate 2008, Campbell et al 2011).
The database used by Malmendier and Tate (2005) and Hall and Murphy (2002) contains specific detailed information that is not publicly available. Therefore, many authors (Campbell et al 2011, Core and Guay2002) have used an alternative stock option methodology for measuring overconfidence. To compute the moneyness of stock options, data from COMPUSTAT Executive Compensation database is obtained. The database does not contain exercise prices of stock options, thus an alternative measure must be used. The value of a stock exercise price is estimated by using approximation method by Core and Guay (2002). According to the model, the total estimated value of unexercised exercisable stock options is divided by the number of unexercised exercisable stock options to obtain the estimated realizable value per option. Afterwards, the stock price at the fiscal year end is subtracted from the estimated realizable option value. This yields to an estimated average exercise price of the stock option. By dividing realizable value of an option with the estimated average price of the stock option, the average percentage of the moneyness is obtained.
14
A binary variable is created, which takes the value of one if a CEO holds stock options that are at least 67% in the money and zero otherwise. A CEO is classified as overconfident when he/she has been holding highly in the money stock options at least twice during the tenure. This decreases the possibility of other factors influencing the reason of holding stock options (for example inside information). A CEO, who is classified as overconfident, remains so throughout the sample period.
2.2 Definition of a Forced Turnover and Replacement of a CEO
CEO turnover is defined as the decision by the board of directors to fire the current CEO. In defining forced turnover of a CEO, the adjusted approaches of Lehn and Zhao (2006) and Campbell et al (2011) have been used.
CEO departure is regarded forced if: 1) the departure has not been announced at least six months in advance 2) it has been reported that the CEO is forced to resign 3) if the announcement does not specify that the reason of departure is related to health (including death) or accepting a new position 4) the CEO is under the age of 65 and the announcement of the retirement is reported less than six months in advance.
Data about the succeeding CEO is also obtained from LexisNexis or proxy statements. The succeeding CEO is classified as overconfident or non-overconfident by following the same approach and criteria that was discussed in previous section.
2.3 Regressions and Variables
2.3.1 Logistic regression
In order to explore if CEOs are replaced by non-overconfident CEOs after acquisitions, several logistic regressions are estimated. The population logistic model of the binary dependent variable Y with multiple regressors is:
15
where X is the vector of independent variables that for example include a dummy indicating CEO overconfidence, cumulative abnormal returns, Research & Development expenditures, performance measures, CEO characteristics and many more.
The dependent variable in the logistic regressions is binary that takes the value of 1 when a CEO has been replaced by a non-overconfident CEO within 5 years after the announcement of merger or acquisition and zero if the CEO is replaced by an overconfident CEO. The period of five years has also been used in similar studies (Lehn and Zhao 2006) that investigate forced turnovers.
2.3.2 Event Study Methodology
To measure abnormal performance, cumulative abnormal returns (CARs) around the announcement date of the merger or acquisition are conducted. The CARs for each acquisition have been calculated to the acquirer over the event window of three days [-1/1], where 0 indicates the day of the announcement. This window period should minimize noise and is used by several authors, for example Malmendier and Tate (2008) and Bouwman et al (2009).
In the sample there are firms that make multiple acquisitions, so using regular market-model event study would not reveal the real impact of a certain merger or acquisition. Therefore, by following the approach of Fuller et al (2002) and Malmendier and Tate (2008) the estimation period is eliminated from the event study and market returns are used as proxy for expected returns. Therefore, the approach assumes that α = 0 and β = 1
The abnormal returns are calculated by formula:
AR
it= r
it– r
t m(2)
where rit indicates firm i return on day t of the event window and rmt indicates the
market return on that specific day.
The coefficients on cumulative abnormal returns are expected to be negative, indicating that the acquisition is value-destroying for the shareholders of the company. This finding would also reveal that overconfident CEOs face higher probability of being replaced by
16 2.3.3 Performance Measures
By following the suggestions of Barber and Lyon (1997) and Lehn and Zhao (2006), market-adjusted abnormal buy-and-hold returns (Pre-BHAR (-2)) are computed over two years to 20 days before announcement of merger or acquisition. This approach deals with the possibility that the firm was already performing poorly before the announcement of the acquisition. For estimating firm performance after the merger or acquisition announcement, market-adjusted post acquisition buy-and-hold returns (Post-BHAR (+2) are calculated over two years after the acquisition. Murphy and Zimmerman (1993) have provided evidence that CEO turnovers are negatively correlated with changes in earnings. Therefore, industry adjusted (adjusted by three-digit SIC code) on-assets (Pre-ROA (-2), Post-ROA (+2) and return-on-equity (Pre-ROE (-2), Post-ROE (+2) profitability indicators are used to measure performance before and after the acquisition. Return-on-assets and return-on-equity are calculated by subtracting industry mean return from the firm return. These performance indicators are also measured two years before and two years after the announcement of merger or acquisition announcement. Values on return-on-assets and return-on-equity are winsorized at 1st and 99th percentile.
The coefficients on post-acquisition performance measures are expected to have negative signs in all regressions. This would provide evidence that decrease in earnings after the merger or acquisition increases the probability of overconfident CEO being replaced by a non-overconfident CEO. On the same time, coefficients on pre-acquisition performance variables are expected to have a positive sign, indicating that higher performance before the acquisition leads to a greater probability of forced replacement of overconfident CEO by non-overconfident CEO after value destructive merger or acquisition. Overview of data containing mean and median values of various variables that are used in the regressions can be obtained from Table I.
2.3.4 Research & Development expenses
To examine if overconfident CEOs are performing better in innovative industries, data on R&D expenditures is obtained. Many papers (Hirshleifer et al 2012, Galosso and Simcoe 2011, Coles et al 2006) use spending on Research & Development (R&D) to identify innovativeness of industries. Moreover, it is assumed that in innovative industries R&D expenditures are larger than in less innovative industries with low product cycles. Therefore, in
17
this paper industry average R&D expenditures are used to measure innovativeness. In regressions that include R&D expenditures, return-on-assets and buy-and-hold-returns are measured over one fiscal year before the announcement of the acquisition, as suggested by Hirshleifer et al (2012) and Huson et al (2003).
2.3.5 CEO Characteristics
By following several authors (Malmendier and Tate 2008, Campbell et al 2011, Hirshleifer et al 2012) CEO age, tenure, salary and bonus are included as control variables in the regressions.
Tenure. The effect of tenure on CEO replacement is controversial. It is possible that the
longer the CEO has been in the position, the more influence he has on the board, which may reduce the chance of forced replacement (Hermalin and Weisbach 1998). On the other hand, CEOs have more time to make bad decisions when their tenure is long and also their overconfidence may increase during that time. This would support findings of Billet et al (2008).
Age. Murphy and Zimmerman (1993) and DeFond and Park (1999) suggest that CEO
age is positively correlated with turnovers. The authors argue that the departing CEOs are significantly older than control firm CEOs. Therefore, the sign on the coefficient of age is expected to be positive.
Compensation. Boards of directors also try to provide incentives to CEOs by setting an
optimal compensation package. It might often be that the fixed salary is rather small, however the bonuses might be remarkable. This kind of compensation package should provide incentives to CEOs to perform well and reasoned, as suggested by Coughlan and Schmidt (1985). Moreover, it has been argued that higher compensation can reflect better skills, which may decrease the probability of forced turnover (Rose and Shepard 1997). Also, Campbell et al (2011) have found a positive relation between CEO salary and forced turnovers of overconfident CEOs. Therefore, the sign on CEO salary and bonus are expected to have positive signs.
18
3. Sample and Descriptive Statistics
3.1 Sample
The study focuses on the sample of U.S acquisitions and CEO compensation packages between years 1995-2014. The sample containing data about CEO compensation is obtained from COMPUSTAT Executive Compensation database. The sample from COMPUSTAT is based on the criteria that: 1) the number of unexercised exercisable stock options is available 2) the value of unexercised exercisable options are available 3) stock price at fiscal year is available.
The initial dataset from COMPUSTAT Executive Compensation database is merged with Thomson One Mergers and Acquisitions database. Criteria for the data are: 1) the status of the acquisition is completed 2) both the acquirer and target are publicly traded 3) the value of transaction is known.
In addition, it is required that all the sample firms are included to the Centre for Research in Securities Prices (CRSP) database. Another important criterion is that the data about turnovers is available for acquiring companies. For obtaining information about turnovers, LexisNexis database is used to search for various news reports that would provide information about the departure of a CEO. After hand-collecting this data, CEO departures are classified as forced or voluntary (exact specification is provided in the next section). If information is not available from LexisNexis database, proxy statements are examined to find the missing information. Data about Research & Development expenditures, institutional ownership and other accounting measures are obtained from COMPUSTAT database. CEO characteristics such as age, tenure, salary and bonus are obtained from COMPUSTAT Executive Compensation database.
This leads to a final sample of 474 forced turnovers out of which 440 took place within 5 years after the announcement of the acquisition.
19
3.2 Descriptive Statistics
Table I presents the sample descriptive statistics. The table contains CEO characteristics and also firm performance measures. CEO age and tenure are described in years. CEO salary and bonus are in thousands of U.S dollars. Pre-acquisition performance measures are calculated over two years before the announcement of merger or acquisition. Post-acquisition performance measures are calculated over two years after the announcement. The exact description of variables is provided in previous paragraph.
.
Table I
Sample Characteristics
CEO age and tenure are described in years. CEO salary and bonus are described in thousands of U.S dollars. Pre-acquisition buy-and-hold abnormal return (Pre-BHAR (-2)) is measured over two years to twenty days before the acquisition announcement. Pre-acquisition return-on-equity (Pre-ROE (-2)) and return-on-assets (Pre-ROA (-2)) are measured over two years before the announcement of acquisition. Post-acquisition buy-and-hold abnormal return (Post-BHAR (+2)), return-on-equity (Post-ROE (+2)), return-on-assets (Post-ROA (+2)) are measured over two years after the announcement of the merger of acquisition.
Overconfident Non-Overconfident
For firms with overconfident CEOs, the mean (median) CEO age is 56.1 (55), mean (median) CEO tenure is 8.4 (6) years, mean (median) CEO salary is 788.3 (721.2) thousand U.S Dollars and mean (median) CEO bonus is 1694.8 (564.8) U.S Dollars. From summary statistics, it is seen that CEO compensation packages are incentive-based, that is the packages consist of high bonuses and lower fixed salaries. The data also indicates that for firms with non-overconfident CEOs, the mean (median) CEO age is 54.3 (57) years, mean (median) tenure
CEO Characteristics Mean Median Mean Median
CEO age 56.113 55.000 54.278 57.000 CEO tenure 8.397 6.000 11.210 10.000 CEO salary 788.287 721.154 847.634 772.000 CEO bonus 1694.751 564.823 1577.495 945.010 Pre-acquisition performance Pre-BHAR (-2) 0.282 0.090 0.241 0.199 Pre-ROE (-2) 0.217 0.151 0.179 0.143 Pre-ROA (-2) 0.360 0.270 0.381 0.272 Post-acquisition performance Post-BHAR (+2) 0.046 -0.133 0.075 -0.130 Post-ROE (+2) 0.089 0.086 0.097 0.101 Post-ROA (+2) 0.044 0.007 0.050 0.009
20
is 11.2 (10) years, mean (median) salary is 847.6 (772) thousand U.S dollars, mean (median) bonus is 1577.5 (945) thousand U.S dollars. Therefore, in the sample, non-overconfident CEOs are on average slightly younger (around two years) than overconfident CEOs. On the other hand, non-overconfident CEO mean (median) tenure is 2.8 (4.0) years longer than for overconfident CEOs. This shows that firms with non-overconfident CEOs hire younger CEOs who on average are on the position for a longer time than overconfident CEOs.
By comparing compensation packages of CEOs, it is visible that mean (median) salary for non-overconfident CEOs is 59.347 (50.85) thousand U.S dollars higher. Mean bonus is 117.256 thousand U.S dollars higher for overconfident CEOs, however the median bonus is 380.187 thousand U.S dollars higher for non-overconfident CEOs. This indicates that in firms with overconfident CEOs there are some CEOs who have very large bonuses, while others have remarkably lower bonuses.
Descriptive statistics reveals that for overconfident CEOs, mean (median) buy-and-hold returns before the merger or acquisition are 83.7% (47.8%) higher than post-acquisition buy-and-hold-returns. Also, mean (median) pre-acquisition return-on-equity is 45.4% (43%) higher than post-acquisition equity. Moreover, mean (median) pre-acquisition return-on-assets is 47.8% (97.4%) higher than post-acquisition return-on-return-on-assets. Thus, the data shows a decrease in post-acquisition performance measures. The decrease in mean values is between 45% -84%, depending on the measure.
The mean pre-acquisition buy-and-hold abnormal return is 14.53% higher for firms with overconfident CEOs, however, the median is 54.8% higher for firms with non-overconfident CEOs. Also, mean (median) return-on-equity is 17.51% (5.3%) higher for companies with overconfident CEOs. However, mean (median) return-on-assets is 5.5% (0.7%) higher for companies with non-overconfident executives. This reveals that on average pre-acquisition performance measures are higher in firms with overconfident CEOs. This may be because overconfident CEOs are willing to take more risks and therefore these risks may pay off and increase profits.
Post-performance data reveals that mean (median) buy-and-hold abnormal return 38.6% (2.25%) is higher for companies with non-overconfident CEOs. Furthermore, mean (median) return-on-equity and mean (median) return-on-assets are 8.2% (14.9%) and 12% (22.2%), respectively, higher for companies with non-overconfident CEOs. Thus, in the sample post performance of acquisition seems to be better in firms with non-overconfident CEOs than in firms with overconfident CEOs. This shows that overconfident CEOs make acquisitions that tend to be more value destructive than acquisitions made by non-overconfident CEOs.
21
4. Empirical Results
4.1 CEO Replacement with Non-Overconfident CEOs
Several logistic regressions are estimated to examine the probability of overconfident CEOs being replaced with a non-overconfident CEO by boards of directors. The dependent variable in each regression is binary, which takes the value of one if a CEO is replaced with a non-overconfident CEO within five years after the announcement of merger or acquisition. Otherwise the variable takes the value of zero. A dummy indicating CEO overconfidence is also added to each regression, which take the value of one if a CEO is overconfident and zero otherwise. Other independent variables include cumulative abnormal returns with event window of three days (CAR [-1,1]), for measuring the announcement return. Several CEO characteristic variables are added to the regressions, such as age, tenure, salary and bonus. Also, as discussed in section 2.3, the logistic regressions include pre- and post-acquisition performance measures, such as buy-and-hold abnormal returns, assets and return-on-equity. The results of the logistic regressions are presented in Table II and the coefficients represent marginal effects on the probability of CEO forced replacement by non-overconfident CEO.
The coefficients of the overconfidence binary variable are positive and significant at 1% and 5% levels in all six estimated regressions. This reveals that after acquisitions overconfident CEOs face higher probability of being replaced by overconfident CEOs, than their non-overconfident peers. To put the coefficients in perspective, an non-overconfident CEO is 15%-25% more likely to be replaced by a non-overconfident CEO than his/her already non-overconfident peer. This suggests that boards who have experience with an overconfident CEO take extra care to make sure that the replacement is a non-overconfident CEO. Moreover, the coefficients on CAR [-1,1] are negative and significant at 1% and 5% levels in all of the regressions. The coefficients indicate that probability of forced replacement with a non-overconfident CEO increase by 21% to 61%, when there is a one percent decrease in abnormal returns. In efficient capital markets, the reaction to the announcement should provide an unbiased opinion if the acquisition is in the best interest of shareholders (Lehn and Zhao 2006). Thus, the results
22
suggest that the more value destructive an acquisition is perceived to be, the more likely it becomes that the board of director replaces the overconfident CEO with a non-overconfident one.
The coefficients of post-acquisition buy-and-hold-abnormal returns are also negative and significant at 1% and 5% levels, indicating that replacement with a non-overconfident CEO increases when the acquisition results in worse post-acquisition buy-and-hold returns. Moreover, the coefficients on pre-acquisition buy-and-hold abnormal returns are positive and negative at 5% level in all regressions, revealing that high stock performance before the announcement of merger or acquisition leads to a greater probability of overconfident CEO being replaced by a non-overconfident CEO. These findings are consistent with results of other forced turnover papers providing evidence that poor performance is punished by replacement of a CEO (Huson et al 2003, Lehn and Zhao 2006, Campbell et al 2011). Moreover, the results suggest that the board of directors becomes more vary of overconfident CEOs when firm performance before the acquisition was particularly good.
The coefficients of CEO tenure are positive and significant at 10% level in four out of five regressions. This finding is not very strong, but still indicates that the probability of overconfident CEO being replaced by a non-overconfident CEO increases with tenure. The positive sign of the coefficients is consistent with Billett and Qian (2008) who find that boards at first do not perceive negative characteristics of a CEO. When boards of directors finally perceive that a CEO is overconfident, he/she is being replaced and a non-overconfident CEO is hired instead. The coefficients of CEO age are also negative and significant at 10% level in two of the regressions with p-values of 0.075 and 0.076. This means that younger CEOs have a higher probability of being replaced with non-overconfident CEOs. The coefficients were expected to be positive since Murphy and Zimmerman (1993) and DeFond and Park (1999) show positive relation between CEO departures and age. However, the findings of this paper are in line with those of Serfling (2014) who argues that excessive risk taking behaviour decreases with age. The results seem to indicate that young CEOs take too much risk and show more optimistic behaviour about investment decisions. The older the CEO gets the less risk he/she is willing to take and thus the probability of being replaced by a non-overconfident CEO decreases.
23
Table II
Logistic Regressions of the Possibility of a CEO Being Replaced by a Non-Overconfident CEO within Five Years After Acquisition
The table presents results of the logistic estimates of CEO turnovers and replacement with non-overconfident CEOs. The dependent variable is binary, which takes a value of one if a CEO is involuntarily replaced by a non-overconfident CEO within five years after an acquisition. Otherwise the dependent variable equals zero. Overconfidence indicator takes the value of one when a CEO is overconfident and zero otherwise. Cumulative abnormal returns CAR [-1,1] are measured over the event window of 3 days, surrounding the announcement of acquisition. CEO tenure and age are in years. CEO salary and bonus are in thousands of U.S dollars. Pre-acquisition buy-and-hold abnormal returns (Pre-BHAR (-2)) are measured over two years to twenty days before the announcement. Pre-acquisition return-on-assets and return-on-equity are measured over two years before the announcement of the acquisition. Post-acquisition return-on-assets and return-on-equity are measured over two years after the announcement. More detailed description of variables is provided in section 2.3. The values represent the marginal effect. Numbers in parenthesis are p-values. Results are robust to heteroscedasticity.
(1) (2) (3) (4) (5) (6) Overconfidence Indicator 0.249*** 0.219*** 0.201*** 0.156** 0.152** 0.153** (0.001) (0.002) (0.006) (0.038) (0.041) (0.041) CAR [-1,1] -0.527*** -0.205** -0.218** -0.481*** -0.618*** -0.603*** (0.003) (0.022) (0.020) (0.006) (0.003) (0.004) CEO Tenure 0.006 0.007* 0.007* 0.007* 0.007* (0.146) (0.094) (0.062) (0.054) (0.055) CEO Age -0.006* -0.006* -0.005 -0.005 -0.005 (0.076) (0.075) (0.107) (0.116) (0.121) CEO Salary -0.001 -0.001 -0.001 -0.001 (0.170) (0.322) (0.392) (0.410) CEO Bonus 0.001 0.001 0.001 0.001 (0.212) (0.308) (0.214) (0.217) Post-BHAR (+2) -0.074** -0.076*** -0.075** (0.013) (0.010) (0.012) Pre-BHAR (-2) 0.032** 0.035** 0.034** (0.048) (0.029) (0.033) Pre-ROE (-2) -0.093 -0.094 (0.136) (0.146) Post-ROE (+2) 0.069 0.070 (0.164) (0.171) Pre-ROA (-2) 0.017 (0.852) Post-ROA (+2) -0.018 (0.811) Pseudo-R2 3.6 3.6 4.0 6.0 6.7 6.7 Number of observations 440 440 440 440 440 440
24
Other performance measures, such as return-on-assets and return-on-equity before and after the announcement of the acquisition are not significantly different from zero. This finding indicates that boards of directors tend to care more about stock performance than other profitability indicators. Warner et al (1988) and Lehn and Zhao (2006) also examine the relation between stock performance and management changes and suggest that the probability of management change increases when firm’s stock performance decreases. Thus, the findings of the paper are consistent with Warner et al (1988) and Lehn and Zhao (2006), suggesting that such relation also holds in more recent data. The coefficients of CEO bonus are positive and those of CEO salary are negative. However they are not significantly different from zero in the regressions. Therefore, CEO bonus and salary do not seem to be important determinants of the probability of being replaced by a non-overconfident CEO.
The results of the regressions contribute to the growing literature of forced turnovers relating to overconfidence. The findings provide support to papers by Campbell et al (2011) and Lehn and Zhao (2006) by revealing that boards of directors distinguish between overconfidence and non-overconfidence. Boards understand that overconfidence can lead to rather destructive investment decisions, which decrease firm value. In addition, the value destructive behaviour of overconfident CEO leads to his/her replacement with a non-overconfident CEO. Overall, boards of directors tend to act in the interest of shareholders. Therefore, when it comes to overconfidence, the board of directors tends to act in a way that maximizes firm value. The results also align with the first two hypothesis of the paper and thus show that overconfident CEOs are more likely to be replaced by non-overconfident CEOs than their already non-overconfident peers.
To further test if value-destroying acquisitions increase the probability of a CEO being replaced by a non-overconfident CEO, four extra regressions are run. The regressions include a dummy for CEO overconfidence, a dummy for indicating a value destructive acquisition, which takes the value of one if cumulative abnormal returns around the three day event window surrounding the announcement (-1,1) of the acquisition are negative and zero otherwise. Moreover, an interaction of value destructive acquisitions and CEO overconfidence is added to the regressions to test whether the relationship between overconfidence and probability of being replaced by a non-overconfident CEO only holds for bad or for all acquisitions. As control variables CEO age, tenure, salary and bonus are added. The results are presented in table 3.
25
Table III
CEO overconfidence and Value-Destructive Acquisitions
The table presents the results on the probability of a CEO being replaced by a non-overconfident CEO within five years after making an acquisition. The dependent variable is binary, which takes a value of one if a CEO is involuntarily replaced by a non-overconfident CEO within five years after an acquisition. Otherwise the dependent variable equals zero. Overconfidence indicator takes the value of one if a CEO is overconfident and zero otherwise. Value destruction takes the value of one if the announcement returns around the three day event window are negative for the bidding firm and zero otherwise. Value destruction x overconfidence indicator is an interaction term of value destruction dummy and overconfidence dummy. CEO salary and bonus are in thousands of U.S dollars. CEO age and tenure are in years. The coefficients represent marginal effects and p-values are in parentheses. The results are robust to heteroscedasticity.
(1) (2) (3) (4) Overconfidence Indicator 0.229*** 0.390*** 0.370** 0.270** (0.003) (0.010) (0.016) (0.048) Value Destruction 0.080* 0.260* 0.250* 0.149* (0.074) (0.088) (0.067) (0.070) Value destruction x Overconfidence Indicator 0.168* 0.154* 0.072** (0.079) (0.068) (0.048) CEO Salary -0.001 -0.001 (0.275) (0.222) CEO Bonus 0.001 0.001 (0.299) (0.321) CEO Age -0.006* -0.006* (0.064) (0.070) CEO Tenure 0.004 0.001 (0.390) (0.333) Pseudo-R2 3.0 3.2 3.8 4.2 Number of observations 440 440 440 440
*** indicates significance at 1% level, ** indicates significance at 5% level, * indicates significance at 10% level
The results reveal that the probability of being replaced by a non-overconfident CEO within five years after a merger or acquisition is 22%-39% higher for overconfident CEOs than for non-overconfident CEOs. This probability further increases by 7%-17% when the acquisition is value destructive as the interaction term of value destructive acquisitions and CEO overconfidence is positive and significant at 5% and 10% levels in all the estimated regressions. Moreover, the main effect of value destructive acquisitions on probability of a CEO being replaced by a non-overconfident CEO is positive and significant in all the regressions. The results provide further evidence to the regressions estimated previously and
26
reveal that overconfident CEOs are indeed more likely to be replaced by non-overconfident CEOs than non-overconfident CEOs within five years after acquisitions. Thus, the results support the hypothesis and reveal that the board of directors can distinguish between overconfident and non-overconfident CEOs before hiring them, when they already have experience with an overconfident CEO. The coefficient of CEO age is negative and significant at 10% level, just as in previously estimated regressions. Moreover, as in previous regressions CEO salary and bonus are not statistically significant from zero. In some of the previously estimated regressions CEO tenure is significant at 10% level. However, in the regressions which include interaction term, CEO tenure is not statistically different from zero.
Overall, the results are in line with the previous results and indicate that the probability of being replaced by a non-overconfident CEO within five years after mergers and acquisitions is higher for overconfident CEOs than for non-overconfident CEOs. Furthermore, the results suggest that this probability increases further after particularly bad acquisitions. These findings are also consistent with Campbell et al (2011) who find that after bad investment decisions overconfident CEOs are more likely to be fired than their rational peers. Additionally, the findings reveal that boards of directors who have experience with overconfidence prefer hiring a non-overconfident CEO. This indicates that board of directors can distinguish between overconfidence and non-overconfidence, even before hiring a CEO. The probability of being replaced by a non-overconfident CEO increases with more value-destructive acquisitions, as the board of directors is likely to become more risk averse and pay greater attention to making sure that the replacement CEO is non-overconfident.
4.2 Innovation and Turnovers
Table IV reports the results of the regressions that estimate the relation between Research and Development (R&D) expenditures and forced replacement of overconfident CEOs by non-overconfident CEOs. To examine this relationship, the sample is constrained to include only overconfident CEOs. Moreover, a dummy variable indicating good and bad acquisitions is included in the regressions. Five logistic regressions are estimated to see whether overconfident CEOs are replaced by non-overconfident CEOs in innovative industries within five years after value-destructive acquisitions. The dependent variable is binary, which takes the value of one if an overconfident CEO is being replaced by a non-overconfident CEO within five years after a merger or acquisition. To measure if an acquisition is value-destroying
27
(bad) or value-enhancing (good), a dummy (Value Destruction) has been created, which takes the value of one if cumulative abnormal returns around the announcement date (-1/1) of the acquisition are negative and zero if the returns are positive. Industry average R&D expenditures are added to the regressions to measure if a sector is innovative or not. Moreover, an interaction term of R&D expenditures and bad acquisitions is included to see if bad acquisitions in innovative industries are related to the probability of overconfident CEO being replaced by a non-overconfident CEO. As control variables CEO age, tenure, return-on-assets (ROA) and buy-and-hold abnormal (BHAR) returns are included. Return-on-assets and buy-and-hold abnormal returns are measured over one fiscal year preceding the announcement of merger or acquisition.
Table IV
Overconfident CEOs and R&D expenditures
The table present results of logistic regressions of overconfident CEO forced replacement with non-overconfident CEOs in innovative industries. The dependent variable is binary, which takes value of one when overconfident CEO is replaced by a non-overconfident CEO within five years after an acquisition and zero otherwise. Value destruction takes the value of one if cumulative abnormal returns around the announcement (three day window period, -1,1) of acquisition are negative and zero otherwise. Industry average R&D is calculated over one fiscal year preceding the acquisition and is measured in thousands of U.S dollars. Value destruction x Industry R&D is an interaction term of the two variables. CEO tenure and age are measured in years. Return-on-assets and buy- and hold abnormal returns are measured over one year preceding the announcement of acquisition. The coefficients represent marginal effects on the probability of CEO forced replacement with a non-overconfident CEO. P-values are in parentheses. The results are robust to heteroscedasticity.
(1) (2) (3) (4) (5)
Value destruction 0.094* 0.006 -0.031 -0.043 -0.186
(0.064) (0.922) (0.603) (0.472) (0.462)
Industry R&D 0.191** 0.004** 0.005* 0.005** 0.005**
(0.041) (0.050) (0.056) (0.043) (0.042)
Value destruction x Industry R&D 0.003* 0.003** 0.003** 0.001**
(0.063) (0.043) (0.037) (0.040) CEO Age -0.012 -0.012 -0.049 (0.001) (0.001) (0.003) CEO Tenure 0.002 0.002 0.008 (0.102) (0.101) (0.125) BHAR -0.009 -0.039 (0.572) (0.567) ROA 0.093 (0.761) Pseudo-R2 1.7 2.9 3.2 3.2 3.4 Number of Observations 189 189 189 189 189
28
In table IV the coefficient of the dummy variable indicating value-destruction is not significantly different from zero, indicating that a bad acquisition would not increase the probability of being replaced by a non-overconfident CEO if industry R&D expenditures equals zero. However, this is not likely to be the case in any industry and the interaction term included in the regressions is positive and significantly different from zero. Therefore, it can be inferred that a bad acquisition increases the probability of being replaced by a non-overconfident CEO. In addition, the main effect of industry R&D expenditures on the probability of an overconfident CEO being replaced by a non-overconfident one is positive and significantly different from zero. This reveals that the probability is higher in high R&D sectors even after a good acquisition. Furthermore, the interaction term is also positive and significantly different from zero. This reveals that bad acquisitions in high R&D industries further increase the probability of an overconfident CEO being replaced by a non-overconfident CEO. This finding rejects the third hypothesis of this paper, which states that innovative sectors appreciate overconfidence and therefore are willing to replace the overconfident CEO with another overconfident one. In fact, the evidence from the regressions suggests just the opposite. Hirshleifer et al (2012) and Galosso and Sicmcoe (2011) suggest that overconfident CEOs generate higher innovative growth and thus achieve success in innovative sectors. It can be argued that boards of directors might not immediately replace overconfident CEO in innovative sectors. However, when there is time to hire a new CEO they prefer non-overconfident executives. Furthermore, the results indicate that innovative sectors are more sensitive to overconfidence than less innovative sectors. One possible explanation for this finding might be that, in innovative sectors, after value-destroying mergers the competitive advantage might vanish rather quickly. Once the advantage is lost, the consequences may last for years before the company regains its competitiveness. However, it can be argued that in industries that require lower R&D expenditures and have lower product cycles, one mistake might not have such a huge impact on the competitiveness.
The coefficients of return-on-assets buy-and-hold abnormal returns and CEO tenure are not significantly different from zero. This reveals that return-on-assets and buy-and-hold returns one year before the acquisition are not important determinants of the probability of being replaced by a non-overconfident CEO. The coefficients on CEO age are negative and significant at 1% level, showing once again that younger overconfident CEOs face higher probability of being replaced by a non-overconfident executive. This is also in line with the results of the previous regressions.
29
Overall, it can be concluded that after value-destroying merger or acquisition overconfident CEOs face higher probability of being replaced by non-overconfident CEOs than their non-overconfident peers. It can be argued that board of directors somehow observe overconfidence and therefore can distinguish between overconfident and non-overconfident executives. Moreover, boards of directors seem to take extra care and attention in not hiring overconfident CEO again, when the leaving CEO is overconfident. This relation seems to be especially strong when firms’ stock returns were high before the merger or acquisition or when the acquisition was value-destroying. Therefore, boards perceive overconfidence as a negative characteristic and prefer non-overconfident CEOs.
Previous studies argue that innovative industries are the only industries where overconfidence is considered to be beneficial. However the results of this study provide evidence of just the opposite. Therefore the third hypothesis of this paper is rejected. Furthermore, the findings of the paper indicate that the probability of being replaced with a non-overconfident executive increases when firm performance was good before the acquisition
As in many forced turnover papers, possible endogeneity issues restrict from making conclusions that overconfidence causes forced turnovers and replacement with rational CEOs. Reverse causality is not a possible issue in this paper, as replacement by a non-overconfident CEO cannot cause an executive to not exercise his or her stock options in the past. However, there might exist omitted variable bias. Therefore, several other regressions are estimated in the next section to provide robustness to the presented results.
30
5. Robustness Checks
5.1 Hazard Model
CEO turnover studies typically use logistic regressions to analyse the determinants of forced turnovers. However, some authors (Campbell et al 2011, Shumway 2001) argue that if a CEO is in the sample for several years, it might violate the independence assumption of the logistic model. Therefore, this paper re-estimates the previous regressions using the Cox semi-parametric proportional hazard model.
Firstly, I perform Nelson-Aalen proportionality test (Campbell et al 2011, Lehn and Zhao 2006). The Nelson-Aalen test shows that overconfident CEOs (red line in Figure 1) face higher cumulative hazard of being replaced by non-overconfident CEOs than CEOs who are not overconfident.
In the Cox proportional hazards model (Cox 1972), the hazard is assumed to be:
hi(t|X) = h0(t) exp(X1β1 + · · · + Xpβp)
, (3)
where hi(t|X) is the time hazard of a CEO being replaced by a non-overconfident CEO i, h0(t) is
the hazard at time t, and Xi is the vector of covariate values, including cumulative abnormal
return CAR [−1,1], CEO tenure, salary, bonus, buy-and-hold stock returns, return-on-assets and
Figure 1. Nelson- Aalen cumulative hazards estimates
Overconfident
31
return-on-equity. Table V presents the results of the Cox proportional hazards model. The results from Cox regressions reveal the same results as logistic regressions estimated in previous section.
Table V
Cox Hazard Model Estimates of CEO Replacement
Table V presents Cox regressions modelling the probability of CEO being replaced by a non-overconfident CEO. The dependent variable is binary, which takes a value of one if a CEO is involuntarily replaced by a non-overconfident CEO within five years after an acquisition. Otherwise the dependent variable equals zero. Overconfidence indicator takes the value of one when a CEO is overconfident and zero otherwise. CEOtenure is measured in years. CEO bonus and salary are measured in thousands of U.S dollars. Cumulative abnormal returns are calculated over a window period of three days (CAR [-1,1]). Pre-acquisition buy-and-hold abnormal returns (Pre-BHAR (-2)) are measured over two years to twenty days before the announcement of merger or acquisition. Pre-acquisition return-on-equity (Pre-ROE (-2)) and return-on-assets (Pre-ROA (-2)) are measured over two years before the announcement of acquisition. Post-acquisition buy-and-hold abnormal returns (Post-BHAR (+2)), return-on-equity (Post-ROE (+2)) and return-on-assets (Post-ROA (+2)) are measured over two years after the acquisition. P-values are in parentheses.
(1) (2) (3) (4) Overconfidence indicator 0.916*** 0.775*** 0.734*** 0.739*** (0.001) (0.005) (0.008) (0.007) CAR [-1,1] -2.591** -3.135** -2.888** -2.833** (0.046) (0.012) (0.022) (0.027) CEO Tenure 0.001 0.002 0.001* 0.001* (0.171) (0.180) (0.096) (0.085) CEO Bonus 0.001 0.001 0.001 0.001 (0.486) (0.626) (0.576) (0.501) CEO Salary 0.002 0.001 0.001* 0.001* (0.290) (0.151) (0.086) (0.080) Post-BHAR (+2) -0.300*** -0.284** -0.273** (0.010) (0.015) (0.022) Pre-BHAR (-2) 0.083** 0.090** 0.091** (0.020) (0.013) (0.013) Pre-ROE (-2) -0.068 -0.098 (0.627) (0.503) Post-ROE (+2) -0.221 -0.237 (0.201) (0.150) Post-ROA (+2) -0.117 (0.548) Pre-ROA (-2) 0.179 (0.491) Number of observations 438 438 438 438
*** indicates significance at 1% level, ** indicates significance at 5% level, * indicates significance at 10% level
A dummy indicating CEO overconfidence is positive and significant at 1% level in all estimated regressions. Overconfident CEOs face 73%-91% higher hazard of being replaced by
32
a non-overconfident CEO. Moreover, the coefficients of cumulative abnormal returns CAR [-1,1] are negative and significant at 1% and 5% levels in all regressions. The coefficients of pre-acquisition buy-and-hold returns are positive and significant at 5% level. The coefficients of post-acquisition buy-and-hold abnormal returns are negative and significant at 1% and 5% levels in all five regressions. These results reveal that overconfident CEOs, who have conducted value-destroying acquisitions, face greater hazard of being replaced by a non-overconfident CEO. This confirms the results of logistic regressions estimated in previous section. Moreover, the coefficients of CEO tenure are positive and significant at 10% level in two regressions with p-values of 0.085 and 0.096. This again, confirms the earlier results.
The main difference between hazard model results and logistic results is the coefficients of CEO salary are positive and significant at 10% level in two of the hazard model regressions. Although, this relation is not very strong economically, it indicates that higher CEO salary increases the probability of being replaced by a non-overconfident CEO. Overall, the robustness results obtained from the Hazard model are very similar to the logistic results presented earlier. Just like in logistic estimates CEO bonus, post- and pre-acquisition return-on-equity and return-on-assets are not significantly different from zero. Therefore, this paper concludes that the results are robust to different estimation methods.
5.2 CEO Ownership and Replacement with Non-Overconfident CEO
Masulis et al (2007) provide evidence that managers, who are pressured by several corporate control mechanisms, make better acquisitions. Moreover, it has been argued by several authors (Masulis et al 2007, Cremers and Nair 2005) that one good corporate control mechanism is dispersion in institutional ownership. To test if ownership concentration does affect the probability of forced replacement, CEO stock ownership percentage of total equity is included to the regressions. Moreover, I control for insider ownership (percentage of stocks owned by executives) and CEO directorship.
Table VI presents the results of four logistic regressions that include CEO stock ownership (%), insider ownership (%) and a dummy indicating if a CEO is also a director. The dummy indicating CEO overconfidence is positive and significant at 1% and 5% levels. The magnitude of the coefficients remains the same as in previous section and reveals that overconfident CEOs face 15%-22% higher probability of being replaced by a non-overconfident CEO.
