Firm remoteness, the CEO turnover decision and
CEO compensation
Amsterdam Business School
Name: Shirley Brandt
Student ID: 10216545
MSc in: Business Economics
Specialization: Finance
Supervisor: Dr. Torsten Jochem Completion: July 4, 2016
Abstract: This study aims to examine differences between central and remote firms, surrounding CEO turnovers. Firstly, I expect that remote firms are less likely to force CEOs to leave, accompanied by a less strong turnover-to-performance sensitivity, compared to central firms. However, I fail to find evidence for this statement. Moreover, due to the smaller pool of candidates remote firms are faced with, I expect that these firms rely more on internal candidates to replace an outgoing CEO. Results seem to support this finding. Finally, I find that central firms provide their CEOs with a more generous compensation package than remote firms. This compensation package is also composed of a larger fraction of equity compensation, which suggests both a central overall pay premium, as well as a central equity pay premium. This may be explained by the fierce competition central firms face while attracting a qualified candidate, while remote firms do not face that much competition. A difference in peer group, or in the type of candidate the two firms are aiming at, could also explain this difference.
2
Statement of Originality
This document is written by Shirley Brandt 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 Content
Table of Content ... 3
1. Introduction ... 4
2. Literature ... 7
2.1 Remoteness and agency costs ... 8
2.2 Remoteness, the executive labour market and CEO turnover ... 9
2.3 CEO compensation ... 12
3. Data and methodology ... 14
3.1 Data description ... 15
3.1.1 Departure- and replacement types ... 16
3.1.2 Firm-, board-, ownership-, and CEO characteristics ... 16
3.1.3 CEO compensation ... 17
3.1.4 Firm performance ... 18
3.2 Methodology ... 19
3.2.1 Remoteness ... 19
3.2.2 Firm performance ... 19
3.2.3 Forced CEO turnover ... 19
3.2.4 Inside promotion versus external recruitment ... 21
3.3 CEO compensation ... 21
3.4 Endogeneity ... 22
4. Results ... 23
4.1 Forced CEO turnover ... 23
4.2 Internal promotion versus external recruitment ... 27
4.3 CEO compensation ... 31
4.3.1 Total compensation ... 31
4.3.2 CEO compensation structure ... 35
4. 4 Robustness checks ... 39
4.4.1 Remoteness ... 39
4.4.2 Equity compensation: stocks and options ... 40
5. Conclusion and discussion ... 41
5.1 Conclusion and discussion ... 41
5.2 Limitations and suggestions ... 43
Table section ... 45
References ... 60
4 1. Introduction
The labour market for executives refers to the process of matching the supply of, and demand for executives. If this market functions efficiently, it becomes a disciplining mechanism for executives’ behaviour (Fee & Hadlock, 2003). If executives know that they can easily be replaced, and that good performance will be rewarded, it is likely that they will make decisions that increase their appeal to others (Fama, 1980; Fee &
Hadlock, 2003). From a macroeconomic point of view: better matching may improve welfare. There are, however, some frictions which cause the executive labour market to work inefficiently. Besides the most obvious frictions such as asymmetric information, uncertainty about the managerial skills of the candidate and frictions due to corporate governance, there are more (Rajgopal, Taylor & Ventachalam, 2012). The remoteness of firm location being one of them (Knyazeva, Knyazeva & Masulis, 2013). Although
geographic location might have become less of an issue nowadays, due to technical innovation and the ability to travel relatively easy, remoteness still plays a role in many aspects (John, Knyazeva & Knyazeva, 2011). Firms’ geographic locations have been studied with respect to shareholders (Coval & Moskowitz, 1999; Loughran & Schultz, 2005), corporate decisions and results (John et al., 2011; Becker, Ivkovic & Weisbenner, 2011), the supply of corporate directors and board independence (Knyazeva et al., 2013) and bank lending characteristics (Knyazeva & Knyazeva, 2012). These studies do find evidence that geographic location plays a role in many corporate decisions and outcomes. So far, to the best of my knowledge, little or no research has been done on the effect of the remoteness of a company’s location on the CEO turnover decision and the effect on the CEO compensation package. Addressing these topics will create additional insight into one of the frictions the executive labour market faces: remoteness. Results of this paper could be interesting for boards of directors that face the challenge of
attracting a CEO and aligning the firm’s and shareholder incentives with these of the CEO. Finally, firms might want to use the results of this paper when they decide on the location of their headquarter, in case they seek for the ‘right’ location. The research question I will address in this paper is the following:
“Does the remoteness of a firm’s location influence the (forced) CEO turnover decision and the level of CEO compensation – and if so, how?”
5
I use a sample that includes S&P1500 firms and covers the period 1992-2005. I will use information collected by Eisfeldt and Kuhnen (2013) on the type of CEO departure and CEO succession, that broadly covers the same time period. Consistent with the method used by Parrino (1997), CEO departure types are classified as either ‘exogenous’, which refers to a planned retirement, as ‘forced’ if it is known that the firm initiated the turnover, or as ‘unclassified’ otherwise. This could range from an unanticipated retirement, a switch to another position at a different firm, or simply if there is no information available on which party initiated the turnover (Eisfeldt & Kuhnen, 2013; Parrino, 1997). A CEO successor is defined as either ‘firm insider’ or ‘firm outsider’, the latter classified as either ‘industry insider’ or ‘industry outsider’. I use information on forced departures to study whether a board’s decision to fire a CEO differs between central and remote firms. Knyazeva et al. (2013) find empirical evidence that it is easier, and less costly to attract a competent, independent board member when the pool of potential candidates is larger. Moreover, they find that remoteness lowers the willingness of a potential candidate to serve a board of directors. While they focus on attracting independent board members, I will focus on attracting qualified CEO successors. I believe that their findings regarding the smaller pool of candidates also includes to potential candidates for the CEO position. Specifically because of their sample, which is composed of executives in charge of nearby firms, since they are most likely to serve board positions. Competent, independent candidates are scarce, and it takes time and effort to find them. In the end, however, it is not the board, but the CEO who has to define firm strategy, and who is ultimately accountable for firm result (Kesner & Sebora, 1994). Accordingly, CEO turnover events are among the most important events in a firm’s lifecycle (Weisbach, 1995). Also, CEO compensation has received considerable attention from proponents of the optimal contracting view (Hall & Murphy, 2003; Murphy, 1999), as a way to align the CEO’s and shareholders’ incentives. On the contrary, opponents of the optimal contracting view (Bebchuk & Fried, 2003), believe that executive compensation is not a solution to agency problems, but is part of the problem. As a result of the importance of both topics, CEO turnover and CEO compensation, I would like to extend remoteness analysis performed by Knyazeva et al. (2013) from the supply of corporate directors and board independence, to the CEO turnover event, and CEO compensation packages.
6
I argue that remote firms are provided with a smaller pool of potential candidates than central firms. It might therefore be easier for central firms to find a qualified successor for the CEO position. On the other hand, it may be possible that remote firms face less competition in the quest for a qualified CEO. In any way, as a result of prior literature (Becker et al., 2011; Coval & Moskowitz, 1999; John et al, 2011; Knyazeva & Knyazeva, 2012; Knyazeva et al., 2013) it is reasonable to assume that board decisions surrounding CEO turnover events vary between remote and central firms. More specificly, I believe that, surrounding a forced CEO turnover event, the sensitivity to performance is different for central and remote firms. Secondly, I argue that due to the smaller pool of candidates remote firms are faced with, these firms are more likely to appoint a firm insider to the CEO position. In this light, I believe that remote firms are more likely to engage in succession planning. CEO succession planning involves the process in which a board, often together with the incumbent CEO, prepares for the CEO turnover event (Shen & Canella Jr., 2003). A firm insider, appointed by the board, is promoted to – often – the Chief Operating Officer position, in which he or she can be trained for the Chief Executive Officer position if the incumbent CEO leaves office. So, I argue that remote firms are more likely than central firms to appoint an internal candidate to the CEO position. Current literature merely focuses on the relation between successor type and firm performance (as outlined in Brickley, 2003), board- , firm- and ownership characteristics (Dalton & Kesner, 1983; Helmich & Brown, 1972; Huson, Parrino & Starks, 2001; Parrino, Sias, Starks, 2003; Schwartz & Menon, 1985; Weisbach, 1988) and executive compensation (e.g. Elsaid & Davidson, 2009). I will extend this scope by comparing compensation packages of remote and central firms. As suggested by Fernandes, Ferreira, Matos and Murphy (2013), I focus not only on the difference in the level of executive compensation, but also on the structure of the compensation package. Fernandes et al. (2013) compare CEO compensation packages of US firms with those of their non-US counterparts. They find that, after controlling for several firm-, board-, ownership- and CEO characteristics the US pay premium is not excessive, and can be explained by, among other things, a difference in risk profile. I use the setup of their paper to study the difference in the level and structure of the CEO compensation package between remote and central firms.
7
After performing all analyses, I find ambiguous evidence on the difference in the forced CEO turnover decision between central and remote firms. Secondly, results suggest that, after controlling for firm-, board-, ownership- and CEO characteristics, remote firms are more likely to promote insiders to the CEO position, as compared to central firms. This seems to confirm the statement that remote firms, faced with a smaller pool of potential candidates, compensate the lack of a large pool of external candidates with promotion from inside the firm. While training the heir apparent for the CEO position, the firm reduces uncertainty surrounding the turnover event, which is in best interest of the shareholders (Shen & Cannella Jr., 2003). Thirdly, I find evidence that seems to confirm that, after controlling for firm-, board-, ownership- and CEO characteristics, a central compensation premium exists. This premium seems to be driven by the fraction of equity compensation, measured as the sum of stock and option compensation to total compensation. This central pay premium could, among other things be explained by the fierce competition central firms face in the quest for a qualified CEO successor (Fernandes et al., 2013; Gabaix & Landier, 2008). Remote firms may face less competition, and therefore have to provide their CEOs with less excessive compensation packages. Results could give guidance in the design of the optimal contract between the firm and the CEO.
The remainder of this paper is structured as follows: section 2 provides an outline of the existing literature on the effects of a firm’s geographic location, CEO turnover, and CEO compensation. Section 3.1 provides information on the sample used in this paper, including summary statistics, continued by the methodology presented in section 3.2. Results on the forced CEO turnover decision, the insider versus outsider choice and the CEO compensation package are outlined and linked to prior literature in section 4. Robustness checks conclude this section. Finally, a discussion and conclusion is presented in section 5.
2. Literature
This paper is part of a broader stance of literature on the effects of a firm’s geographic location, CEO turnover and CEO compensation. This section discusses existing literature on these topics and links them to firm remoteness. Moreover, expectations based on previous findings are translated into three hypotheses in order to answer the research question.
8 2.1 Remoteness and agency costs
During the last two decades, literature continuously address the effect of a firm’s geographic location on financial decisions (Coval & Moskowitz, 1999; Grinblatt & Keloharju, 2001; Huberman, 2001; Ivkovic & Weisbenner, 2005; John et al., 2011; Knyazeva & Knyazeva, 2012; Loughran & Schultz, 2005; Loughran, 2008), in the light of analysts and venture capitalists (Bae, Stulz & Tan, 2008; Lerner, 1995; Malloy, 2005), the location of a firm’s annual meeting and a firm’s financial performance (Li & Yermack, 2015), and the supply of local directors and the fraction of independent directors (Knyazeva et al., 2013).
John et al. (2011), among others, argue that despite modern (communication-) technology and ever improving infrastructure, a firm’s geographic location still affects corporate decisions and the degree of the agency problem. Jensen (1986) outlines the principal-agent shareholder conflict that exists between firm executives and the firm’s shareholders. This conflict is the result of two frictions: a conflict of interest and incomplete contracting. To start, the interests of shareholders and firm executives are not well-aligned, which results in a conflict of interest. Whereas shareholders are solely interested in firm value, and thus want firms to only invest in positive net present value projects, executives may have their own agenda. They may engage in empire building, which is often not in the best interest of the company and its shareholders. Second, there is information asymmetry between firm executives and outsiders, that causes shareholders to fail to prevent the firm from making non-optimal choices. As argued by John et al. (2011), the information asymmetry problem and the ability to monitor worsens when geographical distance increases. There are many ways to mitigate agency costs. Jensen (1986) was one of the first ones to mention that leverage can reduce agency issues. The probability that executives waste money is more likely if the firm has a large free cash flow. Instead of only investing in positive NPV-projects, a CEO can engage in, for example, empire building, which is certainly not in best interest of the shareholders. Leverage commits a firm to make regular interest payments, which help to reduce an executive’s ability to misbehave and increases monitoring by debtholders. This may explain the general finding that rural firms take on more leverage than urban firms (Arena & Dewally, 2012; Lougran, 2008).
9
2.2 Remoteness, the executive labour market and CEO turnover
Many authors have analysed what role career concerns play in managerial decisions and incentives, and how these incentives should be contracted (Fama, 1980; Gibbons & Murphy, 1992; Gompers & Lerner, 1999). These authors provide empirical evidence consistent with the hypothesis that managerial decisions are affected by reputational- and job concerns in some specialized labour market. The magnitude of this threat is likely to be lower for CEOs in charge of remote firms. Knyazeva et al. (2013) focus on the supply of independent directors, and find empirical evidence that it is easier, and less costly to attract a competent, independent board member when the pool of potential candidates is larger. Competent, independent candidates are scarce, and it takes time and effort to find them. For this reason, I believe that remote firms are provided with a smaller pool of possible candidates, which makes it even harder for them to find a suitable candidate. Moreover, remoteness has a negative effect on the willingness of potential candidates to serve a board (Knyazeva et al., 2013). Therefore, I argue that remoteness likely negatively affects the willingness of potential candidates to serve the CEO position. The potential lack of suitable candidates for the CEO position might influence the decision of a board of directors to force a CEO to leave. On the one hand, it might be that remote firms are less likely to fire a CEO, since their pool of potential successors is smaller than the pool of their central counterparts. On the other hand, it might also be true that remote firms face less competition than central firms in the quest for highly capable candidates, and change their forced CEO decision accordingly (Fernandes et al., 2013; Gabaix & Landier, 2008). Moreover, Knyazeva et al. (2013) find that smaller and less visible companies face relatively more issues in the quest for potential candidates, than their larger and more visible counterparts.
Extensive literature exists on the general relation between firm performance and CEO turnover (Fee & Hadlock, 2003; Hazarika, Karpoff & Nahata, 2012; Huson et al., 2001; Warner, Watts & Wruck, 1988). One of the most important tasks of a board of directors is what they should do with a CEO after poor firm performance (Jenter & Kanaan, 2015). On the one hand, they can force a CEO to leave, on the other hand, they can choose to retain the CEO. Parrino (1997) finds evidence that the decision to retain or fire a CEO is largely influenced by the presence of a qualified external candidate.
10
Among others, Huson et al. (2001) provide evidence that CEOs are more likely to be fired after bad firm performance. Eisfeldt and Kuhnen (2013) argue, in line with findings by other authors, that a board is more likely to force a CEO to leave if the firm performance, corrected for industry performance, is bad. Moreover, Huson et al. (2001) emphasize that differences in governance mechanisms, such as the fraction of independent directors or the fraction of shares held by institutions, and the external control market result in differences surrounding the forced CEO turnover event. Improved monitoring often leads to increased forced turnovers in case a company is performing poorly, in order to protect shareholders interests. Moreover, their evidence shows that while the executive takeover market became more competitive – which is supposed to increase monitoring - over the years, the sensitivity of a CEO turnover event to firm performance did not increase. Also, it indicates that an increasingly competitive takeover market does not result in more effective internal monitoring (Huson et al., 2001). Prior literature studies the relationship between monitoring characteristics and CEO turnover decisions (Goyal & Park, 2002; Huson et al., 2001; John & Senbet, 1998; Parrino, Sias & Starks, 2003; Weisbach, 1988; Yermack, 1996). They generally agree on the positive relationship between the fraction of independent directors and the effectiveness of the monitoring mechanism (Borokhovich, Parrino & Trapani, 1996; Fama & Jensen, 1983; Weisbach, 1988). Conversely, it is argued that a larger board reduces effectiveness (Core, Holthausen & Larcker, 1999). In this light, a larger fraction of independent directors and a smaller board would increase the magnitude of the negative relation between firm performance and forced executive turnover events. The fraction held by institutional investors is often related to the effectiveness of monitoring. Among others, Huson et al. (2001) argue that a larger fraction of institutional ownership strengthens the relationship between executive turnover and firm performance. Although boards are supposed to assess a CEO based solely on the factors he can influence, Kaplan and Minton (2006) and Jenter and Kanaan (2015) provide contradicting evidence. Eisfeldt and Kuhnen (2013) also argue that, while controlled for industry performance, a CEO is more likely to be fired if the industry their firm operates in is performing poorly.
Despite the fact that other factors, mentioned above, influence whether a CEO is retained or fired, there is broad evidence that earnings are an important and significant indicator
11
of CEO turnover (Engel, Hayes & Wang, 2003; Fee & Hadlock, 2004; Hermalin & Weisbach, 1998; Huson et al., 2001; Murphy & Zimmerman, 1993; Weisbach, 1988). To test the effect of remoteness on the CEO turnover decision, I will test the sensitivity of remote and central firms of firm performance on the forced CEO turnover decision.
H1: The sensitivity of firm performance on the forced CEO turnover decision is different for
central and remote firms.
There is also extant literature on CEO successor origin (e.g. Dalton & Kesner, 1985; Eisfeldt & Kuhnen, 2013; Hambrick & Mason, 1984; Helmich & Brown, 1972; Harris & Helfat, 1998; Schwartz & Menon, 1985; Shen & Cannella Jr., 2003; Zhang & Rajagopalan, 2010). CEO succession planning refers to the process of a board preparing for a CEO turnover. CEO turnovers are important events in the context of stock performance (Shen & Cannella Jr., 2003). Shen & Canella Jr. (2003) use an event study surrounding a CEO turnover event, and find that stock performance increases when a heir apparent is promoted to the CEO position, or when an external CEO is hired. A negative abnormal return is noticed when a heir leaves the company. The potential CEO successor, the heir apparent, is often promoted to the Chief Operating Officer (COO) position. From this position, the heir apparent can be trained and prepared to be promoted to the CEO position (Shen & Cannella Jr., 2003; Zhen & Rajagopalan, 2010). Without succession planning, boards of directors face a large risk that stock price will decrease (Zhang & Rajagopalan, 2010). As mentioned by Harris and Helfat (1998), succession planning can help to mitigate agency problems, for example managerial entrenchment. Regarding successor origin, there is general consent among authors that the majority of forced departures is followed by the appointment of a firm outsider to the CEO position (Brickley, 2003; Dalton & Kesner, 1985; Helmich & Mason, 1984; Helmich & Brown, 1972; Huson et al., 2001; Parrino, 1997). While inside candidates are most likely not to break with current firm policy, outside candidates are not tied to current firm policy and therefore more likely to change firm strategy (Dalton & Kesner, 1985; Eisfeldt & Kuhnen, 2013; Hambrick & Mason, 1984; Helmich & Brown, 1972; Parrino, 1997). Parrino (1997) also finds that the probability that an external candidate is hired varies from one industry to the other, and is higher in industries in which firms are more comparable to each other. The relation between firm size and successor type is ambiguous, according to prior literature (Dalton & Kesner, 1983; Parrino, 1997; Pérez-González, 2006;
12
Schwartz & Menon, 1985). Regarding board composition, a larger fraction of inside directors is argued to decrease the probability that an outsider is being appointed to the CEO position (Boeker & Goodstein, 1993). Inside directors may fear that an external successor replaces them, as a result of a change in firm strategy (Helmich and Brown, 1972).
As mentioned before, it is easier and less costly to attract a competent, independent board member when the pool of potential candidates is larger. Moreover, remoteness has a negative effect on the willingness of potential candidates to serve a board (Knyazeva et al., 2013). I argue that due to the smaller pool of candidates remote firms are faced with, they invest more time and money in succession planning. As a result, remote firms are more likely than central firms to promote internal candidates to the CEO position. This is translated in the following hypothesis:
H2: The CEO successor for remote firms is more likely to be an insider, rather than an
outsider, as compared to centrally located firms 2.3 CEO compensation
The executive labour market is often argued to work as a disciplining mechanism for CEOs (Fama, 1980; Gibbons & Murphy, 1992; Gompers & Lerner, 1999). Same holds for executive compensation. There are two opposing views regarding the relation between executive compensation and the principal-agency problem (Bebchuk & Fried, 2003; Hall & Murphy, 2003; Murphy, 1999). Proponents of the optimal contracting view claim that executive compensation is a way to better align shareholders’ and executives’ incentives (Hall & Murphy, 2003; Murphy, 1999). The optimal contracting view is often argued to explain the generous CEO compensation packages. Proponents of the managerial power hypothesis argue that compensation packages are excessive, and part of the agency issues itself, since the executives themselves have control over their compensation package (Bebchuck & Fried, 2003). According to the managerial power view, compensation packages will be more generous and/or characterized by a weaker pay-for-performance sensitivity. Compensation packages are less generous when a board does not monitor effectively (Bebchuck & Fried, 2003). A board’s monitoring effectiveness is often argued to decrease in the number of board members, the number of boards the directors are serving and the CEO being also chairman of the board (Bebchuck & Fried, 2003; Core et al., 1999; Yermack, 1996). Also, a larger fraction of
13
independent directors has been argued to increase monitoring efficiency and thereby reducing the level of compensation (Bebuck, Fried & Walker, 2002). The presence of a large outside shareholder is argued to increase monitoring, which should lead to less managerial power, and thus a less generous compensation package (Bebchuk & Fried, 2003). On the contrary, Guthrie, Sokolowsky and Wan (2012) find no evidence that a higher level of independent directors affects the level of CEO compensation. A high level of institutional ownership is also associated with increased monitoring and thus decreased agency issues. As a result, a higher level of institutional ownership strengthens pay-for-performance sensitivity and to decrease the level of CEO compensation (Hartzell & Starks, 2003). Fernandes et al. (2013) however, argue that it is reasonable to assume that the compensation package becomes more generous, as a result of this increased pay-for-performance sensitivity. General consent exists regarding the positive relationship between the level of CEO compensation and firm size (Conyon, Core & Guay, 2011; Conyon & Murphy, 2000; Core et al., 1999; Fernandes et al., 2013, Rosen, 1981, 1982). Hermalin (2005) also studies corporate governance characteristics in relation with CEO turnover and CEO compensation. His findings suggest that a larger probability of a forced turnover should correspond to a higher level of compensation. Other studies find similar evidence (Peters & Wagner, 2014). Bizjak, Lemmon and Naveen (2008) argue that using peer groups also increase the level of CEO compensation. Central and remote firms might use different peer groups, due to the picking of companies based not only on industry, but also on its geographic location. As a result, their compensation packages might differ. Furthermore, as earlier mentioned, John et al. (2011) find that agency costs increase when a firm has a remote headquarter. This observation, combined with the optimal contracting view (Hall & Murphy, 2003; Murphy, 1999), causes me to believe that compensation packages may vary across locations as well. The compensation premium, however, could also run the other way around. CEOs might demand a higher level of compensation to compensate for the fact that they should move to or stay at a remote place in order to fulfil the position (T. Jochem, personal communication, June 16, 2016).
Furthermore, previous literature also investigates the structure of the compensation package. The compensation structure depends on the CEO’s level of risk-aversion. While risk-averse CEOs prefer a larger amount of fixed compensation and a smaller amount of
14
variable compensation, risk-seeking CEOs prefer a larger amount of variable compensation in form of stocks and options (Hall & Murphy, 2002). To compensate for a riskier compensation package, i.e. more equity-compensation, a risk-averse CEOs will require a pay premium (Hall & Murphy, 2002). Also, prior literature finds that equity-based compensation helps to align shareholders’ and executives’ incentives (Murphy, 1999). Fernandes et al. (2013) argue that the structure of the compensation package should be taken into account while comparing compensation packages, not only the level of the CEO pay. They find that the firm-, ownership-, board and CEO characteristics positively related to the level of CEO compensation, are in fact positively related to the fraction of variable compensation. For example, a positive relation is found between sales-level and equity-compensation (Core & Guay, 1999; Fernandes et al., 2013), while a negative relation is found between the fraction of executive ownership and the fraction of equity compensation (Hartzell & Stark, 2003; Mehran, 1995). A high level of CEO ownership can cause a CEO not to be motivated by their level of compensation, but by their level of ownership in the firm. The board of directors seem to take this into account while designing the compensation package, in terms of a lower fraction of equity compensation (Mehran, 1995). Likewise, institutional holdings and equity compensation are negatively related, since monitoring by these investors can be used a substitute for equity compensation (Mehran, 1995). Since a larger fraction of independent directors is associated with increased monitoring quality, empirical evidence is found to increase the fraction of equity in the compensation package (Knyazeva et al, 2013; Mehran, 1995; Ryan Jr. & Wiggings, 2004). Prior literature thus argues that equity compensation could be used to decrease agency issues. Recall that John et al. (2011) find that agency costs increase when a firm has a remote headquarter. To address the increased agency issues associated with the remote locations, I argue that remote firms increase the fraction of equity compensation in their compensation package. Therefore, the third hypothesis is:
H3: The level of CEO compensation is higher for CEOs in charge of central firms, while the
fraction of equity compensation is higher for CEOs in charge of remote firms. 3. Data and methodology
In this section, the sample and methodology will be discussed. First, an outline of the sample used in this study will be presented, followed by summary statistics. Thereafter, the method used to test the hypothesis will be discussed.
15 3.1 Data description
The sample that will be used in this study consists of S&P1500 firms headquartered in the US and covers the period 1992-2005. I retrieve stock return information from the CRSP database, and annual company fundamentals from Compustat. Corporate governance information, such as CEO compensation, age and board size is retrieved from the ExecuComp database. Information such as the independence of board members, institutional holdings, number of shares held by the board, and CEO-Chairman duality are available in the ISS Directors Legacy Database, formerly known as RiskMetrics. CEO turnover information is gathered for a previous study by Eisfeldt and Kuhnen (2013). This dataset is publicly available and presents data on CEO turnovers in US firms over the period 1992-2006. To make sure all necessary information and control variables are available and comparable, the sample used in this thesis covers a slightly smaller period: 1992-2005. Remoteness is defined using each firm’s geographic location, based on the longitude and latitude of the firm’s headquarter ZIP code. A firm is classified as ‘centrally located’ if headquartered in one of the top 10 MSAs: New York City, Los Angeles, Chicago, Washington-Baltimore, San Francisco, Philadelphia, Boston, Detroit, Dallas, or Houston, or in a 150 miles range from the centre of these places. A firm is classified as ‘remote’ otherwise.
I use the Eisfeldt and Kuhnen (2013) dataset, in which CEO turnovers are classified as either ‘exogenous’, ‘forced’ or ‘unclassified’. This classification method used corresponds to the methods used by Parrino (1997) and Jenter and Kanaan (2015). A turnover is classified as exogenous if the CEO retires, and the retirement is reported half a year before the turnover event, or if it is the result of a detailed reported health issue. A turnover is classified as ‘forced’ when the media reported a forced turnover, when the CEO left office due to different opinions regarding company policy, or as a result of board- or shareholder pressure. Turnovers in which it is unknown which party proposed the departure are classified as ‘unclassified’ turnovers and cover the majority of the turnovers in the sample (Eisfeldt & Kuhnen, 2013). Examples of the latter category are unanticipated retirements or switches to other positions. Moreover, the type of replacement is documented. A candidate is classified as ‘insider’ if he used to be employed by a firm before being appointed to the CEO position of the same firm. The CEO successor is classified as ‘firm outsider, industry insider’ if he used to be employed at another firm, but within the same industry, or as ‘firm outsider, industry outsider’ if
16
he used to be employed at a company active in another industry (Eisfeldt & Kuhnen, 2013). Industries are classified according to the Fama-French 48-industry classification scheme, but results do not change when the two-digit SIC classification scheme is used (Eisfeldt & Kuhnen, 2013). Throughout the paper, I will use the two-digit SIC classification scheme, often used in prior literature (Denis & Denis, 1995; Pérez-González, 2006). I end up with a sample of about 14.000 central firm-year observations, and 8400 remote firm-year observations.
3.1.1 Departure- and replacement types
Table 1 presents summary statistics on the departure- and replacement types of the turnovers included in the sample. From column (6), it follows that more turnover events took place at central firms as compared to remote firms. Panel A shows that for central firms, about 16 percent of the CEOs was forced to leave, whereas this was 14.7 percent for their remote counterparts. Panel B shows that remote firms promote an internal candidate to the CEO position in 73.69 percent of the cases, whereas central firms do this in ‘only’ 69.07 percent of the turnover events. This difference is not statistically significant. Panel B also shows that remote firms are less likely to appoint industry outsiders to the CEO position. This difference is significant at a five percent significance level.
[ insert table 1 here ]
The forced CEO turnover decision will be analysed in section 4.1, the insider succession decision of the board will be analysed in section 4.2.
3.1.2 Firm-, board-, ownership-, and CEO characteristics
Table 2 presents summary statistics on firm-, board-, ownership-, and CEO characteristics. Measured at means, central firms are significantly larger in terms of assets, sales, and market- and book value of equity. This negative relation between firm size and remoteness is also found by Arena and Dewally (2012). Leverage, on the other hand, is significantly higher for remote firms at both median and mean. This is consistent with findings in prior literature, agreeing that remote firms take on significantly more leverage than central firms, attempting to decrease agency costs (Arena & Dewally, 2012; Jensen, 1986; Loughran, 2008). Arena and Dewally (2012) find a 28 percent leverage ratio for rural firms and a 25 percent leverage for central firms. Similarly, Loughran (2008) find a 28.5 and 23.5 percent leverage ratio for central and
17
remote firms, respectively. This is slightly larger than the 22.8 and 25.1 percent leverage ratios I find for central and remote firms, respectively. This difference however, can be explained by the shorter time-span and an adjusted remoteness-threshold used in my sample.
[ insert table 2 here ]
Moreover, the percentage of shares held by institutions is significantly larger for central firms, than for remote firms . This observation is in line with prior literature. John et al. (2011) also find a higher percentage of block holdings for central firms, and argues that this is attributable to the proximity of institutions to the companies, for these firms. The fraction of shares held by the board and CEO is not significantly different across both types of firms. Average board size is significantly larger for remote firms than for central firms. This may be surprising, since remote firms may face a smaller pool of potential candidate. Note, however, that remote firms have a larger board on average, but this board can be of lower quality (Core et al., 1999). The median fraction of independent board members is similar for both firms. On average, however, a slightly larger fraction of independent directors is present at central firms’ boards than at remote firms’ boards. This is consistent with evidence found by Ang, Cole and Lin (2000), who find evidence for a negative relation between the fraction of independent board members and agency costs.
Directors at central firms serve significantly more public boards than directors at remote firms. This could be explained by the fact that firms are concentrated when located in large cities, and therefore have to travel less to serve these board positions. Finally, summary statistics show that CEO age and tenure are comparable across the two types of firms. The average CEO is 55 years old, and serves the CEO position at a specific firm on average 3.6 years.
3.1.3 CEO compensation
Table 3 presents summary statistics on the CEO compensation package. It immediately follows from the table that central firms reward their CEOs with a more generous compensation package than remote firms. The difference is significant at one percent for the level of total compensation, salary, equity compensation, option compensation and bonus. This difference is insignificant for stock compensation. These significant
18
differences can be partly explained by the difference in firm size. The size of the pool with potential candidates can also be a factor explaining the difference in CEO compensation (Knyazeva et al., 2013). Equity compensation and the size of the total compensation package will be used to test the third hypothesis. Results will be presented in section 4.3.
[ insert table 3 here ] 3.1.4 Firm performance
Table 4 presents sample stock returns in the year preceding the CEO turnover event. The ‘complete’ sample includes all firms listed in the S&P1500 database between 1992 and 2005. The ‘turnover’ sample consists of all firms experiencing a CEO turnover in the sample period, according to the data gathered by Eisfeldt and Kuhnen (2013). The ‘forced turnover’ sample only includes stock returns of firms that have forced a CEO to leave during the sample period, according to information gathered by Eisfeldt and Kuhnen (2013). Unadjusted stock returns are positive for the complete sample, the turnover sample and the control sample. Unadjusted stock returns are negative in the year preceding a forced turnover. This corresponds to the summary statistics on returns found by Engel et al. (2003). Stock returns are corrected for the median-industry stock performance using the two-digit SIC classification, and are presented in each second column. Corrected returns are negative in all samples, but appear to be most negative in the forced CEO turnover sample, as expected. Remote firms experience a significantly worse stock performance preceding a forced CEO turnover event than central firms. This may indicate that remote firms react not as sensitive to stock performance as central firms while considering a forced turnover. Hypothesis 1 will be tested in section 4.1.
[ insert table 4 and table 5 here ]
Table 5 presents sample firm performance measured by Tobin’s Q. Again, numbers are measured at the year preceding the CEO turnover event, and are corrected for the median industry performance. As follows from the table, most numbers are positive, but are again smallest for firms in which a forced CEO turnover took place. As with stock performance, remote firms perform significantly worse in the year preceding the CEO turnover event, suggesting that remote firms react not as sensitive to firm performance as central firms while considering a forced turnover. Results will be presented in section 4.1.
19 3.2 Methodology
Hereafter, the remoteness threshold will be discussed first, followed by a description of the performance indicators. This section will be concluded by a description of the regressions that will be performed to test my hypotheses.
3.2.1 Remoteness
To study the influence of the remoteness of a firm’s location, I use the firms’ headquarter location as a proxy for their location. Although the location of the headquarter might be subject to bias, due to for example regulation, the headquarter location is most often used in prior literature (Knyzeva et al., 2013; Li & Yermack, 2014). Since the analysis to be performed is the first study in this specific field, I believe the location of the headquarter can serve as a good proxy. I identify firms as ‘centrally located’ if headquartered in one of the top ten MSAs. These are: New York City, Los Angeles, Chicago, Washington-Baltimore, San Francisco, Philadelphia, Boston, Detroit, Dallas, or Houston. Moreover, firms in a range of 150 miles, measured from the centre of abovementioned areas, are also classified as central. Firms are classified as remote otherwise.
3.2.2 Firm performance
Both firm performance (Tobin’s Q), as well as stock performance measures will be used to measure firm performance. Stock returns and Tobin’s Q are measured over a 12-month-window (Fernandes et al., 2013; Jenter & Kanaan, 2015). The decision to measure firm performance by Tobin’s Q, as opposed to measures such as (operating) return on assets, or return on investment is the result of these latter measures being subject to manipulation (Pérez-González, 2006). Moreover, where measures such as ROA only reflect current performance, Tobin’s Q also takes investment opportunities into account and can be used as a proxy for future growth (Fernandes et al., 2013; Jalal & Prezas, 2012). Firm performance is corrected for the median industry performance, using the firms’ two-digit SIC code. If not stated otherwise, corrected firm performance is used in the analyses. To get rid of outliers, Tobin’s Q is winsorized at a 1 percent level.
3.2.3 Forced CEO turnover
By testing the first hypothesis whether a difference in the sensitivity of firm performance is present between central and remote firms, that result in a forced CEO turnover.
20
One of the methods commonly used in (executive) turnover literature is a logit or a probit regression. Many authors use a logit regression model of the form turnover =
ϕ(performance, firm type, firm type*performance, Wi)1, in which authors test the
sensitivity of firm performance for each type of firm on the occurrence of a turnover event. They mostly focus on the sign and significance of the interaction term, to test the influence of performance on the likelihood of a turnover event for each type of firm (Powers, 2005). Powers (2005) mentions in his paper that focussing on the interaction term in non-linear models can be misleading, and demonstrates that this issue can be solved by analysing marginal effects instead of the ‘standard’ interaction term in the logit regression output. However, among others, Ai and Norton (2003) extensively discuss the difficulties associated with the use of interaction terms in non-linear regressions (Dhar & Weinberg, 2015; Greene, 2010).2 The authors mention that using and interpreting an interaction term’s coefficient in a linear model does not raise concerns (Ai and Norton, 2003; Dhar & Weinberg, 2015; Greene, 2010). For this reason, I will perform analyses using a linear probability model. In the linear probability model, the dependent variable is also a binary variable. One main drawback of this model is the fact that, due to the design of the model, the probability can have a value smaller than zero, or larger than one (Stock & Watson, 2012). This should, however, not lead to large issues in the analyses to be performed. The regression that will be tested is:
(1) Forced Turnoveri =
αi + β1Firm performancei + β2Remotei + β3 Firm perf..*Remotei, + Wi)+ εi.
in which forced turnover is a binary variable that equals one if a forced turnover took place, and zero otherwise. Firm performancei is a continuous variable that quantifies pre-turnover firm performance. Firm performance is measured by corrected stock return and Tobin’s Q in the year preceding the turnover event. Remotei equals one if a firm is located on a remote location (see section 3.2.1 for a more detailed description of the remoteness threshold). Firm performance*remotei is an interaction term, and my variable of interest. First, I expect to find a negative coefficient, indicating that bad firm performance and forced turnovers are negatively correlated. Moreover, a significant coefficient on this interaction term would indicate that the sensitivity to firm
1See Powers (2005) for a complete list of articles using this method.
21
performance is different for remote firms, compared to central firms. Finally, Wi is a vector of control variables. I will generally follow the groups of control variables introduced by Fernandes et al. (2013). They divide the control variables into four groups: firm characteristics, ownership characteristics, board structure characteristics and CEO characteristics. A detailed description of these variables can be found in the appendix, table A1. I follow literature guidance and control for firm size, measured by the natural logarithm of assets, sales and market value of equity, growth opportunity measured by the M/B ratio, and the leverage ratio. Moreover, I control for several board- and ownership characteristics: board size, board ownership, institutional ownership, fraction held by CEO, fraction independent directors, a dummy for CEO-chairman duality. Finally, I control for CEO age and CEO tenure (Campbell et al., 2011; Fernandes et al., 2013; Li & Yermack, 2014).
3.2.4 Inside promotion versus external recruitment
To test the second hypothesis, I use information gathered by Eisfeldt and Kuhnen (2013). They distinguish firm insiders from firm outsiders, in which they split the latter group into industry insiders and industry outsiders. I test whether remote firms are more likely than central firms to appoint an inside candidate to the CEO position. Note that the complete sample of turnovers is used here, not only forced turnovers. I will perform a linear probability model, using the four previously mentioned groups of control variables: firm characteristics, ownership characteristics, board structure characteristics and CEO characteristics. The dependent variable equals one if the CEO successor is promoted from inside the firm, while it equals zero when the successor is a firm outsider, regardless of the industry the candidate used to work in.
(2) Inside successori =
α i + β1 (Remote Location Dummy)i + β2 (Firm Characteristicsi)
+ β3 (Board- and ownership Characteristicsi)+ β4 (CEO Characteristcsi)+ εi.
3.3 CEO compensation
Hypothesis 3 is tested according to a method used by Fernandes et al. (2013), with a minor adjustment. Whereas they use both Tobit and OLS regressions, I only use OLS to test the third hypothesis.
(3) Ln(Total Compensationi) =
α i + β1 (Remote Location Dummy i) + β2 (Firm Characteristicsi)
22 εi.
Ln(Total Compensationi) is the natural logarithm of total compensation. Total
compensation is retrieved from the ExecuComp database, and is composed of salary, bonus, other annual compensation, total value of restricted stock granted, total value of stock options granted (using the Black-Scholes method), long-term incentive pay-outs and all other total compensation. Remote location dummy is a dummy variables that equals one if it concerns a remote firm and zero if it concerns a central firm. Firm characteristicsi is a vector of firm characteristics: logarithm of sales, logarithm of assets, leverage, Tobin’s Q and stock return. Ownership Characteristicsi is a vector consisting of board ownership and institutional ownership. Board Characteristicsi is a vector of board characteristics: board size, fraction of independent directors, a CEO-chairman dummy and average number of board positions held by board members. CEO Characteristicsi is a vector of CEO characteristics: age, tenure and the number of boards the board currently serves. A more detailed description of these variables can be found in the appendix, table A1.
I do not only want to test the level of the compensation package, but also the difference in structure of the compensation package for remote and central firms. I test this using a similar regression, in which the dependent variable is Equity Compensationi/Total
Compensationi.
Equity Compensationi is retrieved from the ExecuComp database, using restricted stock
and value of options awarded. The Black-Scholes method is used to value option awards.
Total compensationi and all other variables are similar to the ones in the regression on
total compensation.
3.4 Endogeneity
As in most finance(-related) studies, and turnover studies in particular, endogeneity is always an issue to take into account. So far, I failed to find a exogenous and relevant instrument, so I will include control variables to control for endogeneity as much as possible. All variables proved to be important in explaining CEO turnover and CEO compensation, based on prior literature are included in all regressions (Campbell et al., 2011; Fernandes et al., 2013; Li & Yermack, 2015). A detailed description of these variables can be found in the appendix, table A1.
23 4. Results
In this section, I present regression results used to test firm performance sensitivity to the occurrence of a forced CEO turnover event. Since the pool of potential candidates is likely to be smaller for remote firms, the boards of these firms may adjusted their turnover decision accordingly. Moreover, I test whether remote firms are more likely than central firms to appoint firm insiders to the CEO position, also as a result of the smaller pool of candidates. Finally, I compare CEO packages in terms of compensation level and compensation structure across the two types of firms.
4.1 Forced CEO turnover
Hereafter, I discuss the results of the linear regressions presented in tables 6 and 7. As mentioned before, remote firms may have more issues finding a qualified successor for a departing CEO. (Knyazeva et al., 2013). It would therefore be possible that remote firms, which are provided with a smaller pool of candidates, have a weaker turnover-to-performance sensitivity than central firms. Remote firms may have more difficulty finding a qualified CEO successor, and choose for this reason not to fire an underperforming CEO. Table 6 presents regression output using the linear probability model. Stock return in the year preceding the CEO turnover event, corrected for the median industry return, is used as performance measure. Regression (1) does not contain control variables, regressions (2)-(6) contain several firm-, board-, ownership- and CEO-characteristics. Additionally, regressions (5) and (6) contain industry-fixed effects and industry- and year-fixed effects, respectively. Robust standard errors are clustered at firm-level and presented between parentheses. The interaction term stock return*remote is included to measure the sensitivity of firm type to firm performance, and is my variable of interest. The coefficient on this interaction term is negative in each regression, which suggests a negative relationship between stock return and the probability of a forced CEO turnover, for a remote firm. For remote firms, an increase in stock price indicates a decrease in the probability of a forced turnover.
[ insert table 6 and 7 here ]
While summary statistics suggest that central firms are more likely than remote firms to fire a CEO, results presented in table 6 do not present any convincing evidence on the first hypothesis. The coefficients on remote are mostly positive, and significant in regressions (4)-(6). This suggests that CEOs in charge of remote firms face a 0.9 percent
24
larger probability of being replaced, compared to central firms. My variable of interest, stock return*remote, fails to be significant in all regressions, except for regression (2). The negative coefficients on this interaction term indicate a stronger turnover sensitivity to performance for remote firms than for central firms. This contradicts my expectation, since I expected remote firms to react less sensitive than central firms. However, since five out of six coefficients is not significant, I fail to give this finding economic meaning. Results presented in table 7 do also not provide evidence to support my first hypothesis. Coefficients on remote are positive in regressions (3)-(6), but this coefficient is only significant in regression (4). Again, results suggest that CEOs in charge of remote firms have a 0.6 to 0.7 percent larger probability of being fired than their central counterparts. The coefficients on the interaction term are very close to zero and lack significance. Contradicting the negative remote coefficients in table 6, the corresponding coefficients are positive in table 7. This indicates a weaker turnover sensitivity to performance for remote, than for central firms. Although this finding meets my expectation, I fail to give it economic meaning, since none of the coefficients is significant.
The coefficients presented fail to provide evidence to support the hypothesis that (bad) firm performance is not as important in predicting forced CEO turnover for remote firms, as it is for central firms. It may be possible that I fail to find convincing evidence for the first hypothesis, due to for example endogeneity issues. However, it may also be possible that no significant difference exists. It may be true that CEOs are self-selecting. If a manager is risk-averse, he chooses to be in charge of a firm in a sleepy town, where the probability of being fired is low. In contract, a risk-seeking CEO might choose to lead a dynamic firm in a large city, in which the probability of being fired is much higher (T. Jochem, personal communication, June 16, 2016). This self-selection may explain the lack of significant differences in the obtained results. The ever-decreasing influence of geographic locations may also be a factor that causes this lack of difference in sensitivity between central and remote firms. Modern (communication-)technology and a constantly improving infrastructure may cause a difference between these firms in terms of a forced CEO turnover decision to be almost non-existent (John et al., 2011). Moreover, the lack of difference is in line with findings by Huson et al. (2004) that a
25
difference in monitoring, as a result of an increasingly competitive takeover market, does not affect sensitivity to firm performance.
Coefficients found on control variables are, in general, comparable to these found in turnover literature. In prior literature, there is a general consensus that the probability of a forced turnover increases if firm performance decreases (Coughlan & Schmidt, 1985; Eisfeldt & Kuhnen, 2013; Fee & Hadlock, 2003; Huson et al., 2004; Warner et al., 1988; Weisbach, 1988). Fee and Hadlock (2003) and Eisfeldt and Kuhnen (2013) find a negative coefficient on industry-corrected stock return, similar to what I find on the coefficients on stock return in regressions (1)-(3), presented in table 6. Regressions (4)-(6) present a positive coefficient on corrected stock return, but these lack significance. Similarly, prior literature finds negative coefficients on their proxy for firm performance: industry-corrected return on assets (Denis & Denis, 1997; Eisfeldt & Kuhnen, 2013). This is similar to what I found on the coefficients on Tobin’s Q in all regressions, although five out of six coefficients fail to be significant. Moreover, I find positive coefficients on the logarithm of assets, as a proxy for firm size, as expected and found by prior literature (Eisfeldt & Kuhnen, 2013; Parrino 1997). I find solely negative coefficients on leverage in my regressions. This contradicts findings by Marshall, McCann & McColgan (2014) and Frank and Goyal (2007), who find a positive relation between leverage and turnover. Referring to Huson et al. (2004), forced turnovers are preceded by poor firm performance, which is consistent with high leverage ratios. Marshall et al. (2014) argue that higher leverage is accompanied by better monitoring by banks, which in turn leads to an increased likelihood of turnover. The negative relation I find is odd, and I fail to find a credible explanation for this finding. However, since all coefficients lack significance, I do not want to attach great importance on this issue. Except for board ownership, none of the board- and ownership characteristics appears to be significant. The negative relation between board ownership and probability of forced turnover contradicts the statement that board ownership and monitoring are positively related. As with leverage, a higher fraction of board ownership is argued to increase monitoring effectiveness (Jensen & Meckling, 1976). This in turn would suggest that a CEO is more likely to get fired if the firm performs poorly. An explanation could be that boards owning a larger fraction of the shares are more attached to the company. In turn, they are more concerned with their job and afraid to
26
lose their job. This is also argued by Dalton and Kesner (1985), stating that boards are sometimes unwilling to replace a CEO, since they fear losing their own position. Coefficients on other board- and ownership characteristics are insignificant and close to zero. These characteristics thus appear not to be as important in explaining forced CEO turnover as firm characteristics. This is consistent with prior literature, finding that there exists no, or only weak evidence for the relevance of corporate governance characteristics (Huson et al., 2004). A larger fraction of independent directors, however, is often referred to strengthen the quality of board monitoring. I therefore expect a positive relation between the fraction of independent directors, and the probability of a forced CEO turnover (Borokhovich et al., 1996; Huson et al., 2004 ; Weisbach, 1988). This is what I find regressions (5) and (6) presented in table 6 and 7, although none of the coefficients is significant. Finally, CEO characteristics fail to be significant in any of the regressions, indicating that these do not influence the forced CEO turnover decision.
In their paper, John et al. (2011) find evidence that corporate decisions differ between central and remote firms. Remote firms alter their corporate decisions in order to reduce agency issues. My lack of evidence on the difference in the forced turnover decision between central and remote firms do not support the findings of John et al. (2011). One of the reasons for this lack of significant results may be the endogeneity issue, outlined in the limitations section. It may be that there are more factors underlying the CEO turnover decision, besides the factors proved to be important by prior literature (Campbell et al, 2011; Fernandes et al., 2013; Huson et al, 2001; Li & Yermack, 2014). As mentioned before, it might also be true that there exists practically no difference regarding the performance sensitivity of remote and central firms on the forced CEO turnover decision. This relates to a finding by Huson et al. (2001) that the sensitivity of a CEO turnover event to firm performance did not increase, while monitoring increased due to an increasingly competitive takeover market. Also, geographical factors may be of less importance, due to constantly improving infrastructure, and modern communication technologies (John et al., 2011). It may also be true that CEOs are self-selecting. If a manager is risk-averse, he chooses to be in charge of a firm in a sleepy town, where the probability of being fired is low. On contract, a risk-seeking CEO might choose to lead a dynamic firm in a large city, in which the probability of being fired is much higher. This self-selection may lead me failing to find a
27
significant difference in the obtained results (T. Jochem, personal communication, June 16, 2016).
All in all, I fail to conclude evidence to support the first hypothesis, which could be the result of several factors outlined above.
4.2 Internal promotion versus external recruitment
While the forced CEO turnover decision by the board of directors seems not to be influenced by the expected scarcity of qualified CEO successors, summary statistics suggest that remote firms are more likely to appoint a firm insider to the CEO position, as compared to central firms (73.69 percent versus 69.07 percent, respectively). This difference is not corrected for any firm-, board-, ownership- or CEO characteristics. In this section, I present results after performing linear probability regressions, using an extensive set of control variables.
In table 8, the dependent variable equals one if the CEO successor is promoted from inside the firm, while it equals zero when the successor is a firm outsider, regardless of the industry the candidate used to work in. Robust standard errors are clustered at firm-level, and presented between parentheses. Regression (1) does not include any control variables, regressions (2)-(4) contain different control variables, but no fixed effects. Regressions (5) and (6) contain industry-fixed effects and industry- and year-fixed effects, respectively.
[ insert table 8 here ]
My coefficient of interest, coefficients on the dummy variable remote, is positive in all regressions. This suggests that remote firms are 0.4 to 1.7 percent more likely to appoint firm insiders to the CEO position, as compared to central firms3. This is in line with findings by Knyazeva et al. (2013). In a recent study on the effects of remoteness on board composition, they find that firms provided with a smaller pool of candidates are more likely to appoint grey and inside directors. Likewise, remote firms are expected to be provided with a smaller pool of executives than central firms. For this reason, remote firms likely depend more on internal promotion, as compared to external recruitment. Therefore, it seems to be true that remote firms spend more time and money on
3 Results are comparable if a logit regression model is used instead of a linear probability model.
28
succession planning, compared to their central counterpart. Without some kind of succession planning, a board of directors face a large risk if they cannot find a suitable CEO successor. This can result in a drop in stock price (Zhang & Rajagopalan, 2010). Concluding, the positive, but insignificant coefficients on the remote dummy variable suggest that remote firms are more likely to promote internal successors, compared to central firms.
The first two firm characteristics, stock performance and Tobin’s Q present either negative values, or values extremely close to zero. A general finding in previous literature is the appointment of a firm outsider to the CEO position following poor firm performance. The idea behind the appointment of a firm outsider in that situation, is the expectation that a firm outsider is better able to change firm strategy than their inside counterpart (Dalton & Kesner, 1985; Eisfeldt & Kuhnen, 2013; Hambrick & Mason, 1984; Helmich & Brown, 1972). Likewise, I expect positive signs on the firm performance variables, which would indicate that a decrease in firm performance would result in a decrease of inside promotion. Results presented in table 8 however, fail to confirm this finding. However, almost all negative coefficients lack significance, so I fail to draw any conclusions from these coefficients. The coefficients on the logarithm of assets, used as a proxy for firm size, are negative but insignificant in regressions (3)-(6), but positive and significant in regression (2). Since this latter regression does not contain the full set of control variables, it seems that larger firms are more likely to appoint external candidates to the CEO position. This is in line with findings by prior literature, although evidence regarding the relation between firm size and successor type vary (Dalton & Kesner, 1983; Helmich & Brown, 1972; Schwartz & Menon, 1985). Larger firms may be more difficult to manage, so that boards want an experienced outside candidate to replace the current CEO. Generally, internal candidates lack CEO experience and boards therefore have a preference for outside candidates (Dalton & Kesner, 1983; Helmich & Brown, 1972; Schwartz & Menon, 1985). On the other hand, larger firms have more potential qualified candidates to promote to the CEO position (Dalton & Kesner, 1983). Coefficients on the sales, market value of equity variable and market-to-book ratio are insignificant and close to zero, and are therefore appear to be unlikely to explain the choice for either type of successor. I find negative coefficients on leverage, which are significant at a five percent significance level in regressions (3) and (4). This negative
