The performance consequences of CEO succession type
in moderation of board composition
MSc International Economics and Business – Master thesis Warner de Jong – S2277670
Supervisor: Prof. Dr. H. van Ees Co-assessor: Dr. P. Rao Sahib
This thesis aims to assess the relationship between CEO succession types and board composition. Using data from MSCI ratings and COMPUSTAT databases about successions between 2005 and 2009, this research finds theoretical support for existing theory on CEO succession types and boards, contributing to the gap between the two fields. After outlier removal, the general effect of each succession and board type holds the predicted sign, given the context by prior CEO departure. Novel is the positive evidence on follower and outsider successions following CEO retirement, and the influence of board types following CEO retirement or resignation. The hypotheses on the interaction of boards and succession have not met the predicted signs, yet provide insights for future research.
Introduction 1
Research literature and hypotheses development 4
Methodology 21
Results 27
Discussion and conclusions 31
Limitations 35
References 36
Appendix A: Outlier removal 43
1
Introduction
In the words of Schnatterly and Johnson (2008): “CEO succession is a major organizational event that has implications for firm strategy and firm effectiveness (Datta & Guthrie, 1994) because organizations are often seen as reflections of their CEOs (Datta & Rajagopalan, 1998), and their “succession affects all types of constituents at virtually every level of the organization” (Kesner & Sebora, 1994, 328). CEO succession is therefore of major importance to any organization. However, what makes a good CEO succession? Research often associated the answer with successor origin, where boards hire successors from either the inside or outside of the firm. A rather simple distinction that trifled the literature with mixed findings. For instance, Thorsten Hein succeeded Research in Motion (Blackberry) and Satya Nadella succeeded Microsoft after their respective CEOs stepped down. Even though these new CEOs were both hired from within their companies, they had produced tremendously different results. Where Hein told analysts “to not expect any seismic changes” during his succession (Gilette, 2012), Nadella had pressed for change. What followed was that Blackberry further fell from grace whereas Microsoft changed course and improved strategy (Weinberger, 2016). That Nadella changed strategy whereas Hein maintained it shows an important difference between inside hires. Unfortunately, this difference between insiders has been of limited discussion in CEO origin until Ocasio (1994) and Shen and Cannella (2002). The authors find that under a power circulation theory of control, that an inside contender CEO advocates change whereas an inside follower CEO maintains the status quo. Prior to this distinction, both were seen as simply insiders. Likely giving reason to why the CEO succession literature before their time has been characterized as a “diffused and often chaotic research stream” (Kesner and Sebora, 1994, p. 327) as findings of one paper frequently conflicted with another.
2
outsider-dominated board and it might be that such boards interact better with successors pressing for change. Conversely, insider boards may interact better with CEOs that are reluctant to change. Which seems likely as insider board members may fear being replaced by pro-change CEOs (Friedman & Saul, 1991). It is therefore not surprising that insider boards prefer inside successors (Shen & Cannella, 2002a) as empirical research widely supports the view that outsider CEOs are more likely to initiate change (Berns & Klarner, 2017). Equally unsurprising is that outsider boards are found to prefer outside succession (Agrawal et al. 2006) as they share a more independent view on the firm. Yet, where does this leave contender CEOs that press for change? Or how does an insider board affect the performance of different types of CEOs?
To answer this question, I relate to the succession context. As not the succession itself but its “context is what affects post-succession firm performance” (Finkelstein & Hambrick, 1996: 193-201). In addition, boards are critical to the context of a succession event (Gordon & Rosen, 1981) and the origin of a CEO is found to reflect both the succession context and has significant implications for subsequent firm performance (Brady & Helmich, 1984). In conclusion, it is interesting to study the effect of different CEO types in moderation of the board, which may further the extant literature on CEO succession and boards, bringing together both fields of corporate governance, and provide practical relevance to both CEOs and boards.
Because the board is responsible for appointing a successor CEO, it is likely that their CEO of choice reflects the board´s expectations and goals, and subsequent implementation of strategy by the successor CEO. Also, as the board is tasked with monitoring the performance of the CEO on behalf of the firm its shareholders, the new CEO’s actions are likely to be in favour of the boards’ goals. This idea is based on the finding that the succession context and political activity during the succession process relate to post-succession firm activities (Welsh & Dehler, 1988), disruption, turnover and morale (Friedman & Saul, 1991). Meaning that how the board functions likely affects the CEO’s operational performance. In addition, board composition relates to the succession context and was found relevant to the operational performance of the firm (Müller, 2014), thus it may reflect the CEO preference of the board. Consequently, I use it in determining the moderating effect of the board. Furthermore as Giambatista et al. (2005) report that theoretical development has not always been robust in considering the role of the board of directors in CEO succession, this study may contribute.
3
database. The data has contributed to the following findings. Out of the sample, 47% of boards opt for follower succession, 19% for contender succession, and the remaining 34% for outside succession. As such, boards hire more frequently from inside of the firm (66%). Likely inferring that most boards compose an inside director majority, as such boards prefer inside succession. However, the opposite seems true as the data shows a vast majority of 90% of boards hold a majority of outside directors. The sample thus over represents outsider boards, which can be explained by 85% of boards adhering to strict U.S. board regulations, advocating majority outside director representation.
Further information on the data provides relevance to the results, and is as follows. 38% of boards were completely without inside directors, 30% had a lone inside director and the remainder had two or more inside directors. The average size of a board holds nine directors, composed out of two inside directors and seven outsiders, of which one is affiliated. The year with the most successions was 2008 with 50 successions, much higher than the average 26 yearly successions between 2005 and 2009. Thus, the financial crisis seemed to be of influence as 14 out of 24 contender successions took place in 2008, representing a greater desire for change as contenders are intent on changing firm strategy. In addition, another five contender successions occurred in 2009. As such, 79% of contender successions took place in the initial phase of the financial crisis. During the crisis, the rate of follower succession fell from an average 48% to 40% whereas outside succession remained stable around 40%. Thus, inside succession seems preferred to outsider succession, regardless of a crisis situation. Yet, the need for strategic change increases due to a crisis. The estimation method used is ordinary least squares (OLS) multiple linear regression. Although the method likely suffered from endogeneity bias, as is common when accounting for, board structure, it was perceived to account for interactions between boards and successions, given a short run time frame. Furthermore, due to constraints of time and the complexity of alternative estimation methods, it was deemed as a viable option.
4
Research literature and hypotheses development
In this thesis, two components of succession contexts are examined. Firstly, I consider the characteristics of CEOs to define whether a CEO successor is a follower, contender or outsider. According to Shen and Cannella (2002) the three types differ in their ability to manage change, firm-specific knowledge and the risk of adverse selection. Hence, CEO type allows inference about performance consequences and strategic expectations of the board and it’s CEO. Secondly, board composition as the composition of inside and outside directors, likely influences the strategic goal of the firm but also the perception of which CEO type is most desirable. Besides boards affecting succession types, succession types may also have important effects on board decisions, which can be quite different from effects on top management, as was researched previously (Shen and Cannella, 2002).
CEO succession types
5
In short, both the contender and outsider successor types characterize a situation where the incumbent CEO is forcefully dismissed, reflecting a situation of poor performance, and that a successor is hired to initiate strategic change. The difference between the two types relates to whether the successor originates from the inside (contender) or from the outside (outsider) of the firm. In contrast to the two types, a follower successor is defined as an inside successor that is intent on maintaining the strategic direction and status quo of the firm after retirement of the predecessor CEO. The maintenance by the follower therefore reflects strategic continuity under conditions of good company performance (Shen & Cannella, 2002; Friedman & Singh, 1989; Hambrick et al, 1993). To assess whether there is poor or good performance, the authors relate to CEO departure as a proxy measure. If the previous CEO is forcefully dismissed than an insider candidate is noted as a contender. In contrast, if the previous CEO has willingly retired than the insider candidate is noted as a follower, indicating that contestation against the previous CEO was unsuccessful. Together this holds true to contestation under the power circulation theory of control. As when a board disapproves of and fires its current CEO, it is unlikely to hire a CEO that is similar, such as a follower. In contrast, if the board approves of, and sees its CEO retire then contention was unsuccessful, by which the follower likely succeeds. Concerning the power circulation of control theory, an outside successor is not part of such contestation. Mainly because the theory is used to explain an internal contest for power, one where the outsider successor does not participate. As such, outside successors are only appointed in situations where a firm is in need of strategic change and has an unavailability of competent inside candidates (Cannella & Lubatkin, 1993; Fredrickson, Hambrick, & Baumrin, 1988; Ocasio, 1999).
6
that boards may select someone who is poorly suited for the job (Zajac, 1990). This level of adverse selection differs per succession type, as will be discussed. Second is that performance is noted here as operational performance, thereby laying focus on the operation of the firm instead of its market performance. Thus it is possible that different successors fit different performance measures. The decision for operational performance is discussed later on.
Followers
Follower successors are likely to succeed the CEO after retirement as heirs or candidates that have been groomed by the incumbent CEO to continue operations after departure. Followers can be seen as known individuals to the firm as they have worked within the firm, lowering the risk of adverse selection (Brady & Helmich, 1984) and are seen to bring minimal change to the strategy of the firm due to a grooming process. In addition to possessing organization-specific knowledge, internal successors have already established personal relationships within the firm (Cao et al., 2006; Helmich & Brown, 1972). Their close connections with predecessors (Fondas & Wiersema, 1997) and within-firm social networks (Shleifer & Summers, 1988) may constrain their ability to initiate change (Shen & Cannella, 2002), causing minimal disruption to the firm. Any change that is however brought about is small in its nature, and will not deviate largely from its predecessor. Therefore, the follower is alike the previous CEO, hence the idea of ceteris paribus largely holds. This seems preferable under the power circulation theory of control because contestation against the prior CEO was unsuccessful, signalling a need to continue in similar strategic direction. A follower thus succeeds a CEO whom retired after having delivered good performance, and left the company in good health, but also in hands capable of continuing similar strategy. This means that a follower will address matters in similar fashion to the predecessor. Judging from the idea that the prior CEO proved capable in facilitating good performance and that contestation was unsuccessful, I hypothesize that the effect of follower successions positively relates to operational performance.
Hypothesis 1: Follower successors will positively affect post-succession operational performance following the retirement of the previous CEO.
Contenders
7
exercised by the board (Alderfer, 1986; Lorsch & MacIver, 1989). The authors Shen and Cannella (2002) have tested the effect of contender succession on performance and found that the relationship was positive but insignificant, unless contender succession interacted with senior executive turnover. Meaning that a contender’s ability to affect performance may depend on whether the contender can change top management. The authors reason that contenders may find support both from its board and senior executives due to power contests against the prior CEO (Vancil, 1987) and that a greater power base facilitates the process of taking charge (Gabarro, 1987). The ability to initiate change may therefore be reliant on other actors, such as top management or the board.
Nonetheless, even if contender succession itself was previously found insignificant to operational performance, I provide two arguments that could infer a significant relationship. Firstly, the appointment of a contender successor signals that the board has validated a known individual as competent enough to initiate strategic change. Otherwise, the board could have opted for an outsider succession. Seeing as that contestation continues to take place under power circulation theory, it is likely that the contender was chosen in light of her competency in dealing with contesting senior executives. Furthermore, the level of contestation may have decreased or may be temporarily absent as contending senior executives find their chance at becoming the CEO foregone, mainly because forced removal by the board is disruptive. To avoid further disruption, another removal is unlikely to happen soon again. Secondly, not everyone may wish to become the CEO. There are ample cases where executives or directors became the CEO as it was considered necessary. For instance, a global Economist Intelligence Unit study on 685 business influentials has shown that 65 percent of North Americans and 60 percent of Europeans would reject the CEO position if offered (Matthews, 2005). Some directors may therefore express a dislike to the position. Hence, senior executives do not necessarily contest the CEO. Under these two arguments, a contender may simply work with a neutral top management. One that is not necessarily interested in contesting against the contender CEO yet knows and deems her as a valid successor. To then initiate strategic change, a successor’s organization-specific knowledge and personal relationships provides power (Cao et al., 2006; Helmich & Brown, 1972). Possibly sufficient power as empirical evidence shows that CEO turnover and discontinued operations are significantly positively related to cases of contender succession (Barron et al., 2011). Meaning that contenders can initiate strategic change as they dispose of assets from the previous leadership, a finding attributed to new CEOs with a mandate for change (Berger and Ofek, 1999). As such, I propose to further investigate the performance relationship of contenders and hypothesize that contenders bring a positive effect on operational performance.
8 Outsiders
Outsiders are typically portrayed as being hired from outside of the firm to provide a fresh look that will change the organization towards increased performance. However, both the concept of outside succession and its subsequent implications seem insufficiently established. I first briefly introduce the concept of outsider succession before going through its implications. Some authors argue that there are several levels of ‘outsiderness’ a CEO can experience (Karaevli, 2007) whereas others define outsiders simply as those completely disassociated from the firm before the hire (Shen & Cannella, 2002). In complement to these distinctions, Schnatterly & Johnson (2008) argue that hiring an outsider member on the board of directors can also be seen as an outsider succession. As such, the field is advancing in providing a better perspective on what should be defined as an outsider to provide more clarity over the contradictory findings of outsiders on performance. In light of such advances, I will adhere to the distinction of outsider CEOs being those that have no affiliation with the firm prior to their hire, as this construct is used most often. Nevertheless, I later investigate the situation where outsiders are hired after good performance and retirement, extending the concept of outsider implications. Furthermore, I do not control for board outsider succession because it is not the norm (Schnatterly & Johnson, 2008). As such I mainly follow the outsider successor type by Shen and Cannella (2002), whom further argue that the likelihood of outsider succession increases if the pool of capable insider candidates available lacks (Finkelstein & Hambrick, 1996; Cannella & Lubatkin, 1993). An outsider succession thus reflects the perceived lack of ability of inside candidates.
9
reorientation. In this aspect, outsiders are seen to come with a disadvantage to the firm as they lack a deep understanding of the internal operations and external environments of the firm, making it difficult for the successor to define and achieve strategic change (Gabarro, 1987; Kotter, 1982). This seems in line with the findings by Shen and Cannella (2002). Namely that outside succession is negatively associated with post-succession operational performance, yet senior executive turnover following outside succession was found to have an even stronger and significant negative impact. As such, the authors argue that the lack of firm-specific knowledge and resistance of top management puts outsiders at a strong disadvantage. In light of the finding that outsiders lack an understanding of the firm, have been increasingly fired and were found negatively related to performance, I hypothesize that outsider succession will negatively impact performance following the dismissal of the prior CEO, as the forced departure signals pre-succession poor performance.
Hypothesis 3: Outsider successors will negatively affect post-succession operational performance following the dismissal of the previous CEO.
For now, three different hypothesized effects are formulated. Follower succession was hypothesized positive under retirement and contender succession under dismissal. Outsider succession was hypothesized negative under poor performance. As a whole, it shows that pre-succession poor or good performance is attributed to the previous CEO (Graffin, Boivie, & Carpenter, 2013). Because under poor performance, strategic change is desirable (Guthrie & Datta, 1997), the contender successor is thus hypothesized to have a positive effect in facilitating change whereas the outsider successor will be at a strong disadvantage. As outsider successors lack firm-specific knowledge and personal relationships.
10
candidates to be more comparable to insiders (Agrawal et al., 2006). This concept of industry familiarity was once introduced by Birnbaum (1971) who argued that not all outsiders are alike. Outsiders that had social and training experience in similar institutions to their recruiting organizations were exposed to less post-succession conflict and provided greater organizational stability than previously thought. Therefore, if the prior CEO transitions to chairman, providing greater stability in the succession process, then outside succession may have a positive effect on performance. Allowing for hypothesis 3a which may contribute to Berns and Klarner (2017) who call for a greater understanding of how CEOs, transitioning to specific board roles, influence performance. To provide an overview of the hypothesized effects, Table 1 was composed.
Hypothesis 3a: Outsider successors will positively affect post-succession operational performance following the retirement of the previous CEO if she becomes the chairman.
Table 1 Hypothesized performance effects of CEO types following performance.
Succession type Poor performance
resignation Good performance retirement Follower succession - + Contender succession + - Outsider succession - +*
*under guidance of the chairman
Rough situation sketches
Because each CEO type has now been formulated in both its hypothesized positive and negative effect on post-succession performance, it is useful to note the ideal situation for hiring each succession type. As such, I provide rough situations that sketch contingencies, helpful at illustrating how each CEO type becomes more or less preferable. The sketches illustrate a simplified world where pre-succession performance is presumed either poor or good and where except CEO succession and CEO departure, ceteris paribus holds. This is because pre-succession performance and CEO departure are, as previously mentioned, related to the succession context (Gephart, 1978).
11
one that is deemed as sufficiently capable to initiate strategic change. In this case, successful contestation of the contender takes place, meaning that the contender successor incentivizes the board to remove the incumbent CEO and hire the contender CEO. The second situation shows that none of the inside candidates is found sufficiently capable of initiating strategic change, the candidates seem more likely to be fit for a follower succession as they express ideas to maintain strategic direction. In such a situation, the board has to consider hiring from the outside. Then the board has to either hire an unknown outsider successor at a greater risk of adverse selection, or it hires a follower successor. Regardless of that an outsider involves greater uncertainty, ceteris paribus; the follower will not initiate strategic change. Hence, this would identify the ideal outsider situation.
When considering a situation of high pre-succession performance, a board is likely to keep its CEO and maintain strategic direction. Firing the CEO would obviously be a bad decision as it would severely disrupt the organization. Nonetheless, a CEO might wish to retire. In such a scenario, it is ideal to hire a follower successor who will maintain the strategic direction of the predecessor. It is however possible that no sufficiently capable inside candidate is present, or that candidates express ideas to initiate strategic change, such contender candidates will be avoided to maintain strategic direction. Likewise, any outsider successor that wishes to initiate strategic change will be avoided. An option that could remain is that if the previous CEO transitioned to the position of chairman, than an outside candidate could be hired to maintain strategic direction. This option assumes that the previous CEO will guide the outsider successor during her tenure.
12
The board and the role of board composition
The board of directors is found to be critical in the context of a succession event (Gordon & Rosen, 1981) yet what the board exactly does and how the characteristics of the board influence succession are two different topics. Furthermore, other factors may also influence the board in the succession context. To accurately describe the relevance of the board within the succession context, I firstly explain what boards do within the succession context and subsequently explain how board characteristics affect succession behaviour. Afterwards, I address the implications that the board and its composition may have on succession.
What does the board of directors do?
The board of directors is the governing body of a company involved in the major decisions of the firm. It is a team that is involved in determining the strategic direction of the firm. This team typically holds a variety of roles that are of influence to the succession context. Mainly because the board controls the process by which executives are hired or fired, thereby including the appointment and dismissal of the CEO (Vancil, 1987). In addition, the board holds the responsibility of monitoring and assessing such executives. The board therefore largely serves as a sounding board to the CEO and top management (Mace, 1971). A board is thus supposed to be in charge of the strategic direction, the hire and fire of the CEO, and the monitoring on the CEO. These tasks indicate the influence of the board to the CEO succession event and subsequent performance.
13
has similar effects on different firms or their boards, presenting boards with similar governance issues, namely to hire their next CEO. Another issue however remains; boards of similar firms may choose different governance solutions to attain different strategic goals. This issue is a source of non-convexity (Adams et al., 2010). This means that different solutions to a problem may exist, which result in similar value. I reason that as boards determine the strategic direction of the firm, so too will they define the requirements and goals that a successor must fulfil. Thereby defining part of the succession context. A simple explanation helps clarify. A board of a poor performing firm will have similar strategic concerns to boards of other poor performing firms, and different strategic concerns than boards of good performing firms. Boards that face similar performance issues will thus formulate similar strategies and may therefore opt for a similar choice in CEO. As was mentioned in the contingencies of succession types; poor pre-succession performance raises the need for change initiating CEOs, in contrast to high pre-succession performance. The performance situation that a board is in therefore helps to deal with the non-convexity issue. In modelling this approach, where the board interacts between CEO succession types and post-performance, I respond to the call from Adams et al. (2010) whom remark such modelling as promising areas for theory in need of investigation.
14
negative signal about the CEO, and belong to a minority of the board, may risk losing their board seat. Together, these models show that a board not only monitors the ability of its CEO but that the degree, cost and risk of a signal may differ across the board and its directors. Hence, how a board monitors the CEO depends on the directors that monitor her. As such, I now move to the characteristics of boards.
How do the characteristics of the board influence succession behaviour?
As no board is likely to be an exact copy of another, so too may its functioning be different from any other. Nonetheless, theory has typically categorized the directors of the board as inside directors (executives) and outside directors (non-executives). The former are employees of the firm whereas the latter are not. The idea is that outside directors bring independence to the board, safeguarding the interests of the firm its shareholders. However, not all outside directors are completely independent, and are therefore seen as ‘grey’, ‘related’ or ‘affiliated’ directors (Hermalin & Weisbach, 1988). This independence is however critical as public pressure and regulations have forced many firms to have majority-outsider boards. For example, a study by Fich and Shidasani (2006) used a sample of 508 of the largest US corporations between 1989 and 1995 and found that, on average, outsiders make up 55 percent of directors, insiders 30 percent and affiliated the remaining 15 percent. To give the reader an indication of how this trend has evolved over time, this study used a sample of 129 of the largest US corporations between 2005 and 2009 where, on average, outsiders make up 68 percent of directors, insiders 20 percent and affiliated the remaining 13 percent. This development of greater outside director representation is largely related to a group of regulations, including the Sarbanes-Oxley act since 2002 and rules by the Securities and Exchange Commission (SEC) and New York Stock Exchange (NYSE) (Duchin et al., 2010). In conclusion, the representation of independent directors has been of severe importance to shareholders and institutions alike. Yet do the characteristics of inside and outside directors affect how a board proceeds with succession? Empirical evidence indicates so as board member characteristics were found to significantly affect succession choice (Zald, 1969). Moreover, in relation to board characteristics and succession, board composition was found significantly related (Helmich, 1975). As such, board composition likely affects CEO choice and possibly even the subsequent performance of the chosen CEO.
The implications of board composition on successions
15 How a board may perceive succession types
Hiring from the inside or outside of the firm is quite different to a board. As I previously mentioned, inside successors are known individuals that possess firm-specific knowledge, providing lower risk of adverse selection to boards. Hence, boards may prefer hiring insiders over outsiders as Mace (1977) argues that hiring outsiders is risky. Evidence by Agrawal et al. (2006) supports this idea as outsiders are only chosen if markedly better than the best insider. Hence boards can try to avoid outside succession by opting for succession planning, grooming a good selection of possible inside candidates. However, succession plans can still be undermined at critical points in the process (de Vries, 1988) as the capability of insiders may lack with regard to the strategic goals of the board. Hence, hiring from the outside remains a feasible choice with various advantages. As for instance Karaevli (2007) argues that outsider CEOs are valued for their more cognitively open-mindedness due to avoiding socialization processes. Their lack of commitment to the status quo may also enable them to better see a new course of action (Karaevli, 2007). However, as hiring outsiders remains as risky as they are unknown, so too is their potential effect, as their firm performance impact of strategic change was found much greater than compared to insider successions, both negatively and positively (Zhang & Rajagopalan, 2010). Nonetheless, the perception and implications of choosing inside or outside candidates may differ depending on the directors on the board.
The moderating implications of inside and outside directors on CEO succession types
16
Table 2 Possible interactions between board composition and CEO succession type
Succession type Poor performance
resignation
Good performance retirement
Follower succession Interaction
Contender succession Interaction
Outsider succession Interaction Interaction
The implications of inside directors
Boards with more inside directors have been found less likely than outsider boards to remove the CEO during poor stock performance (Weisbach, 1988) and may not willingly accept new policies and practices (Friedman & Saul, 1991). Furthermore, a firm may wish to promote managerial incentives to have managers invest in firm-specific skills. Building the idea that having a greater proportion of inside directors will lead to a greater likelihood of insider succession. Unsurprisingly, insider boards were also found to prefer inside successors (Shen & Cannella, 2002a). Likely because inside directors may fear being replaced by a new CEO (Friedman & Saul, 1991), as was shown through the Warther (1998) model. Mizruchi (1983) even argued that the endorsement of outside successors may jeopardize the careers of insider boards.
The implications of outside directors
As most boards have been forced to increase outside director participation, only a few boards remain without a majority of outside directors. Be that with or without including outside affiliated directors. Nonetheless, outsider dominated boards have been found to positively impact post-succession operational performance (Shen & Cannella, 2002) and are positively related to choosing outsiders following poor performance (Boeker & Goodstein, 1993) both in forced and voluntary departures (Borokhovich, Parrino & Trapani, 1996; Huson et al., 2001). Outside directors have also been found to initiate the exits of heirs apparent to diminish the impact of incumbent CEOs (Cannella & Shen, 2001). As such, greater outsider representation on the board seems to be associated with a greater preference towards outside successors. In addition, outside directors are generally viewed as playing a larger role in monitoring management than inside directors and are often regarded as “professional referees whose task is to stimulate and oversee the competition among the firm’s top management” (Fama, 1980, p. 294). Altogether, it seems that a greater proportion of outsiders on the board associates with greater monitoring of CEOs and greater competition among top management.
The implications of poor performance on succession types
17
operations (Weisbach, 1995) and top management turnover (Barron et al., 2011). Non-appointed insiders on insider boards are therefore likely to reduce their ownership and leave the firm because of negative career prospects (Boyer & Ortiz-Molina, 2008). Empirical research further supports this fact, as top management turnover is significant in case of contender succession (Shen & Cannella, 2002). Possibly, because the disruptive removal of the CEO and the poor performance prior to her removal signals that the board has failed to monitor its CEO properly, perhaps by offering too much slack. A painful signal and even more so if executives have to go against their preference and hire an outside successor. How do outsider boards then influence successors in poor performance? Seeing as that outside directors are viewed as directors with a greater monitoring role and do not only stimulate but also oversee the competition among management. I argue that outside directors likely provide a supply of capable inside contender candidates. Furthermore, outside directors will possibly have less fear towards change as it may not jeopardize their careers as much as inside candidates, for two reasons. Firstly, outside directors are not employees of the firm and secondly, the firm is likely obligated in maintaining a majority of outside directors. This may allow an outsider board to better support contender successors that are intent on initiating change. But also outsider candidates as they are typically found to initiate change. Outsider boards also prefer outsider candidates (Agrawal et al. 2006). Yet that does not necessarily imply a lower degree of monitoring. Mainly because outside successors were both increasingly hired (Karaevli & Zajac, 2013) but also fired (Kaplan & Minton, 2012) in a period where the representation of outside directors dominates. This indicates that outside directors are vigilant to both contender and outsider successors. In sum, it seems as if outside directors do well in facilitating successors that are intent on strategic change. Their difference can lie in the size of the effect as evidence shows that outside successors have a greater potential effect on performance than inside successors (Zhang & Rajagopalan, 2010). In light of the arguments presented, I hypothesize that, during periods of poor performance, the performance effects of contender and outsider successors are positively affected by outsider boards in contrast to insider boards.
Hypothesis 4a: Greater outside representation on the board positively affects the operational performance of contender successors following the dismissal of incumbent CEOs.
Hypothesis 4b: Greater outside representation on the board positively affects the operational performance of outsider successors following the dismissal of incumbent CEOs.
The implications of good performance on succession types
18
follower candidates of sufficient capability need to be available. In the case that candidates are found lacking, a board may opt for an outside succession, which opposes both the preference of the board and the managerial incentives that inside succession may have. In both cases, however, the board does not have to fear replacement. For instance, followers are restricted by their within firm networks to initiate change whereas outsiders are hired to continue strategy. A board may thus support an outside candidate as she poses less of a threat to inside directors.
Outsider boards may also affect inside successors. As previously mentioned, outside directors are found to initiate the exits of heirs apparent to lower the impact of the incumbent CEO (Cannella & Shen, 2001), which negatively relates to follower successions. Furthermore, outside directors are generally viewed as greater monitors (Weisbach, 1988) and stimulate competition among top management (Fama, 1980). Both of these points oppose the idea of a follower successor. As followers are not groomed to compete against the existing strategy but groomed to maintain it. Their networks within the firm also limit their ability to compete. Followers are thus likely to struggle with outsider boards. Outside successors may also struggle because of being new to the firm. In order to maintain the strategy and good performance of the firm, they require time to get to know the firm. However, outside directors are vigilant monitors and may monitor an outside successor too harshly. Taking these arguments together, I argue that in the case of good pre-succession performance, indicated by the retirement of the prior CEO, the board will negatively affect the post-succession performance of the follower and outside successor. Possibly explaining why outside successors have been fired at an increased rate.
Hypothesis 4c: Greater outside representation on the board negatively affects the operational performance of follower successors following the retirement of incumbent CEOs. Hypothesis 4d: Greater outside representation on the board negatively affects the operational performance of outsider successors following the retirement of incumbent CEOs.
19
previous CEO. Table 3 is composed to illustrate the hypothesized interactions whereas table 4 illustrates the main effects of boards.
Hypothesis 5: Outsider boards positively affect operational performance following the dismissal of incumbent CEOs.
Table 3 Hypothesized moderation performance effects of board composition.
Board type Poor performance resignation Good performance retirement
Follower and Outsider successions Contender and Outsider successions
Insider - +
Outsider + -
Table 4 Hypothesized main performance effects of board composition.
Board type Poor performance resignation Good performance retirement
Insider - +
Outsider + -
Firm performance
With regard to the review on performance measurement on succession events by Giambatista et al. (2005), firm performance has traditionally been measured using both market and operational measures. I stray from market measures as the authors argue that such measures “over a short window do not explicitly compare succession to non-succession events” (2005, p. 976). It is also argued that the board of directors uses accounting measures to set convenient targets to reach for management (Joskow, Rose, & Shepard, 1993) and more likely uses them to assess CEO compensation, in comparison to market valuation measures (Hambrick & Finkelstein, 1995; Jensen & Murphy, 1990). Furthermore, Grossman and Hoskisson (1998; Hambrick & Finkelstein, 1995) indicate that market valuation is often subject to forces beyond management control and that, in contrast, operational performance is more under management control. Boards of directors and CEO successors are therefore more likely to use operational performance measures instead of market valuation measures. Hence, I follow Shen and Cannella (2002) in measuring operational performance as a three-year average of post-succession return on assets (ROA). The average is taken over three years to limit the possibility of performance instability as deriving results from sudden performance fluctuations may cause erroneous findings.
Controls
20
Menon, 1985). For instance, outside succession was found to be more likely for smaller firms as they have less human capital to draw upon (Lauterbach et al., 1999), whereas insider succession was found more common in larger organizations (Friedman & Singh, 1989). Secondly, pre-succession performance because it matters how well the firm performed before succession. Thirdly, industry performance* at the 2-digit SIC level as the performance and volatility of performance differs per industry. Fourthly, year effects as external factors such as the financial crisis of 2008 may affect the performance of firms. The second set of controls relates to the board of directors. Firstly, board size because larger boards may provide a greater pool of insider candidates and decrease the risk of adverse selection. Secondly, previous CEO duality because past CEOs that also served as chairman of the board are argued to exert additional influence in the succession process as they hold more power than those without dual titles (Finkelstein & D’Aveni, 1994; Lorsch & MacIver, 1989) thus affecting the likelihood of follower succession. Thirdly, successor CEO duality as firms without CEO duality outperformed firms with CEO duality (Rechner & Dalton, 1991). Fourthly, and lastly, whether the previous CEO stays as the Chairman. This is because former CEOs that remain on the board were found to prevent a successor from implementing large changes (Quigley & Hambrick, 2010). These and other controls have been suggested to be influential on firm performance. The impact of some factors, such as firm size, governance structure and industry environment, may be particularly significant in the CEO succession context owing to their potential influence of managerial discretion (Finkelstein & Hambrick, 1996). By combining all the previously mentioned effects, figure 1 has been composed.
Figure 1: The hypothesized effects including the addition of control variables
*Dotted lines only hold during outsider successions e.g. the dotted line of previous CEOs transitioning to the position of chairman interacts only in cases of outsider succession.
*An earlier version of this thesis included industry type at a 2-digit SIC level as a control. It however caused
multicollinearity issues with industry performance. After inspection, models without industry type showed a better fit through the R-squared than models without the industry performance variable. Hence, industry type was removed. Multicollinearity allows two different predictors to predict largely the same for the dependent.
Succession type Contender Firm operational performance Follower Outsider Control variables - Board size
- Successor CEO duality - Previous CEO duality
Board composition
- Firm size
- Pre-succession performance - Industry performance
- Year effects - Previous CEO is chairman
Insider board Outsider board
21
Methodology
Sample and data collection
Data was primarily collected from three MSCI (Former GMI Ratings) datasets over the years of 2005 up until 2010. The datasets included company data, CEO data, and director data, which were combined using matching ticker codes. Deducting whether a CEO succession took place was done by firstly comparing the period of 2005 to 2009 for all firms in the compiled database. A year in which a CEO’s name was different from the previous year was coded as a 1, symbolizing a succession year. The data from the previous year was then taken to infer about the past CEO’s departure, and whether she transitioned to a new position on the board. However, the data from the three databases often showed conflicting years for when a succession took place, wrong CEO or director names or presented undocumented years in which successions could have taken place. Descriptive data on CEOs and their successors was used to largely overcome this error. The descriptive data was derived from reading online company statements, news sources such as PR News Wire, Business Wire, and bibliographies by Bloomberg. For firm performance, firm size and SIC codes, data from COMPUSTAT NORTH AMERICA – annual fundamentals - was taken for the years 2002 up until 2013. Data entries from both COMPUSTAT and MSCI were combined through matching ticker, CUSIP and CIK codes. The final sample consisted of 129 CEO successions.
Dependent variable: Post-succession operational performance (Post_ROA)
As operational and not market measures are chosen to measure performance, I follow several other scholars in using Return On Assets (ROA) as the dependent variable of choice (e.g. Shen & Cannella, 2002; Karaevli, 2004; Zhang & Rajagopalan, 2010; Karaevli & Zajac, 2013). ROA was calculated by dividing net income before extraordinary items and discontinued operations by net assets for each fiscal year. The items deduced from net income helped represent firm performance under normal circumstances. Starting from the succession year, ROA was averaged over three years to fully account for the effect of the succession event, whilst smoothening out sudden performance fluctuations (e.g. Shen and Cannella, 2002; Daily et al., 2000; Kesner & Dalton, 1994). More than three years were not included as a greater likelihood of other influential factors becomes present. The resultant variable was Post_ROA, a ratio variable for each succession.
Successor type (CEOtypei)
To test hypotheses 1 till 3 on the main effect of succession type to post-succession performance, the categorical variable CEOtype was created to represent CEO succession types. Its denominator i represents a follower succession for the value of 1, the value 2 represents a contender succession and 3 represents an outsider succession.
22
from the succession year and its prior year were compared. Afterwards, news sources were used to check if the successor was an employee to the firm, had re-joined the firm or had previously worked at the firm for over a year. If the successor was not employed by the firm before her succession, then CEOtype was coded as a 3 for outsider succession. In all other cases, the successor was either an inside follower or contender successor.
Separating these inside succession types relied on how the previous CEO departed. Followers are those that succeed after retirement whereas contenders succeed after the forced dismissal of the CEO. To infer about whether the previous CEO had resigned, both the MSCI data and descriptive news sources were checked. The data was assumed as more reliant as firms rarely report the true reasons of why a CEO is dismissed (Denis & Denis, 1995; Weisbach, 1988). In addition, an age cut-off at the age of 64 was used to determine departure. Although an age cut-off is a rather crude measure to determine a CEO’s dismissal, other researchers have used it in making this decision (e.g. Ocasio, 1994; Puffer & Weintrop, 1991; Shen and Cannella, 2002). This is reasonable as almost all inside successions were reported as retirements, leaving limited space for contender successors. This seems unlikely when considering the competition initiation of outside directors on top management, unlikelihood of follower succession during poor performance, or the larger pool of candidates that large U.S. stock-listed firms may have. Therefore, if the CEO was below the age of 64, and had left the firm according to the data and descriptive sources, than CEOtype was coded as a 2 for contender succession.
If news sources reported that a CEO had left her position effective immediately, possibly implying a resignation, than the data was checked against this finding. Mainly as CEOs often transition to the position of chairman or director, which can hardly be called a resignation (Shen & Cannella, 2002, Lorsch & MacIver, 1989). If the data showed that the CEO transitioned to another position on the board then a successor was labelled as a follower. In all other cases, where the successor was not a contender, CEOtype was coded as a 1 for follower succession. The process of deducing departure is depicted in Diagram 1.
23
Thus, all succession cases in the data were compared against descriptive sources, possibly providing a more robust result. However, three errors remain. Firstly, a number of CEOs may leave the firm before retirement due to pursuing a career elsewhere. Even though it is in the interest of this research to identify such departing CEOs, there is no such variable, which accurately defines such a departure. Using descriptive sources may also be unfruitful as the true reason of departure is often not conveyed to the general public. Hence, the distinction between a follower and contender successor remains crude. Secondly, prior CEOs might have left the firm due to health-related reasons or are diseased. Thirdly, a CEO that succeeds a temporary interim CEO could incorrectly infer a successor to be a follower. As interim CEOs only temporarily fill the CEO position and likely fill another role on the board after the succession event. Only in the latter two cases, the observations were removed from the sample as they were detectable through descriptive measures.
Now that the succession types have been categorized, hypotheses 1 till 3 can be tested, if each succession type is put in its proper context. In contrast to both follower and contender successions, outside successors do not automatically incorporate the departure of the CEO. Therefore, the dummy variable “Departure” was created, following the same method of separating follower and contender successions, yet including all successions. The dummy was coded a 1 if the previous CEO had resigned and a 0 if the previous CEO retired. The dummy is included in all models to provide a control on the categorization and effect of CEO departure. The dummy interacts solely with outsider successions, for which CEOtype is denominated with i=3, because Departure would correlate with inside succession types. As such, two models are devised. Model 1 tests hypothesis 1 for i=1 and hypothesis 2 for i=2. The predicted sign of CEOtype is positive for both hypotheses. Model 2 tests hypothesis 3 for i=3. The predicted sign of the dismissal on outside succession interaction is negative. Model 1 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃_𝑅𝑅𝑅𝑅𝑅𝑅 = 𝛽𝛽1 + 𝛽𝛽2 ∗ 𝐶𝐶𝐶𝐶𝑅𝑅𝑃𝑃𝐶𝐶𝐶𝐶𝑒𝑒𝑖𝑖+ 𝛽𝛽3 ∗ 𝐷𝐷𝑒𝑒𝐶𝐶𝐷𝐷𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝑒𝑒
Model 2 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃_𝑅𝑅𝑅𝑅𝑅𝑅 = 𝛽𝛽1 + 𝛽𝛽2 ∗ 𝐶𝐶𝐶𝐶𝑅𝑅𝑃𝑃𝐶𝐶𝐶𝐶𝑒𝑒𝑖𝑖+ 𝛽𝛽3 ∗ 𝐷𝐷𝑒𝑒𝐶𝐶𝐷𝐷𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝑒𝑒
+ 𝛽𝛽4 ∗ (𝐶𝐶𝐶𝐶𝑅𝑅𝑃𝑃𝐶𝐶𝐶𝐶𝑒𝑒𝑖𝑖∗ 𝐷𝐷𝑒𝑒𝐶𝐶𝐷𝐷𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝑒𝑒) + 𝜀𝜀
Simultaneous to the earlier checking for departure and continuity, it was recorded if the previous CEO transitioned to the position of chairman. If so, the variable PC_chairman was coded with a 1 and a 0 in all other cases. Hypothesis 3a is tested by allowing an interaction between CEOtype denominated with i=3, Departure and the control variable PC_chairman. This is because the guidance of the chairman is believed to moderate the relationship between outside succession and post-succession performance, if the previous CEO retired. This moderating situation provided model 3. The predicted sign of the interaction is positive. Model 3 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃_𝑅𝑅𝑅𝑅𝑅𝑅 = 𝛽𝛽1 + 𝛽𝛽2 ∗ 𝐶𝐶𝐶𝐶𝑅𝑅𝑃𝑃𝐶𝐶𝐶𝐶𝑒𝑒𝑖𝑖+ 𝛽𝛽3 ∗ 𝐷𝐷𝑒𝑒𝐶𝐶𝐷𝐷𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝑒𝑒 + 𝛽𝛽4 ∗ 𝐷𝐷𝑒𝑒𝐶𝐶𝐷𝐷𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝑒𝑒
24
Board type (Outside_pct)
Board type was measured as the ratio of independent outsiders on the board divided by total board size, creating the ratio variable Outside_pct. It is based on data about outsider representation on the board and board size in the succession year. This year was used as a successor would directly work with the board of that year. Due to the Sarbanes-Oxley act and other regulations by the SEC, some companies were forced to increase the number of outside directors on the board (Duchin et al., 2010). According to the MSCI data, 110 out of the 129 companies adhere to these regulations. The vast majority of boards within the sample thus consisted of outsider boards, as only three boards held a majority of inside directors due to past regulations. Nonetheless, both board types could be investigated as
Outside_pct is a ratio of outsider representation, where lower values represent a movement
towards insider boards, allowing to test inside director representation, which may provide similar results to testing insider boards. In testing hypotheses 4a till 4c, I build further on model 1 by adding an interaction with Outside_pct. Hypothesis 4a for contender succession predicts a positive interaction, hypothesis 4c for follower succession predicts a negative interaction. Again to avoid correlation, hypotheses 4b and 4d on outside successions build further on model 2, added with an interaction by Outside_pct. Hypothesis 4b predicts a positive interaction for outsiders under dismissal while hypothesis 4d predicts a negative interaction for retirement.
Model 4 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃_𝑅𝑅𝑅𝑅𝑅𝑅 = 𝛽𝛽1 + 𝛽𝛽2 ∗ 𝐶𝐶𝐶𝐶𝑅𝑅𝑃𝑃𝐶𝐶𝐶𝐶𝑒𝑒𝑖𝑖+ 𝛽𝛽3 ∗ 𝐷𝐷𝑒𝑒𝐶𝐶𝐷𝐷𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝑒𝑒 + 𝛽𝛽4 ∗ 𝑅𝑅𝐷𝐷𝑃𝑃𝑃𝑃𝑎𝑎𝑂𝑂𝑒𝑒_𝐶𝐶𝑐𝑐𝑃𝑃
+𝛽𝛽4 ∗ (𝐶𝐶𝐶𝐶𝑅𝑅𝑃𝑃𝐶𝐶𝐶𝐶𝑒𝑒𝑖𝑖∗ 𝑅𝑅𝐷𝐷𝑃𝑃𝑃𝑃𝑎𝑎𝑂𝑂𝑒𝑒_𝐶𝐶𝑐𝑐𝑃𝑃) + 𝜀𝜀
Model 5 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃_𝑅𝑅𝑅𝑅𝑅𝑅 = 𝛽𝛽1 + 𝛽𝛽2 ∗ 𝐶𝐶𝐶𝐶𝑅𝑅𝑃𝑃𝐶𝐶𝐶𝐶𝑒𝑒𝑖𝑖+ 𝛽𝛽3 ∗ 𝐷𝐷𝑒𝑒𝐶𝐶𝐷𝐷𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝑒𝑒 + 𝛽𝛽4 ∗ 𝑅𝑅𝐷𝐷𝑃𝑃𝑃𝑃𝑎𝑎𝑂𝑂𝑒𝑒_𝐶𝐶𝑐𝑐𝑃𝑃
+ 𝛽𝛽5 ∗ (𝐶𝐶𝐶𝐶𝑅𝑅𝑃𝑃𝐶𝐶𝐶𝐶𝑒𝑒𝑖𝑖∗ 𝐷𝐷𝑒𝑒𝐶𝐶𝐷𝐷𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝑒𝑒 ∗ 𝑅𝑅𝐷𝐷𝑃𝑃𝑃𝑃𝑎𝑎𝑂𝑂𝑒𝑒_𝐶𝐶𝑐𝑐𝑃𝑃) + 𝜀𝜀
Model 6 is about the unique effects of board types in the succession context. The dummy variable Departure was used to determine the context. Hypothesis 5 was tested when
Outside_pct interacted with Departure coded as a 1 for resignation. The interaction is
predicted with a positive sign for hypothesis 5a.
Model 6 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃_𝑅𝑅𝑅𝑅𝑅𝑅 = 𝛽𝛽1 + 𝛽𝛽2 ∗ 𝑅𝑅𝐷𝐷𝑃𝑃𝑃𝑃𝑎𝑎𝑂𝑂𝑒𝑒_𝐶𝐶𝑐𝑐𝑃𝑃 + 𝛽𝛽3 ∗ 𝐷𝐷𝑒𝑒𝐶𝐶𝐷𝐷𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝑒𝑒 +𝛽𝛽4 ∗ (𝑅𝑅𝐷𝐷𝑃𝑃𝑃𝑃𝑎𝑎𝑂𝑂𝑒𝑒_𝐶𝐶𝑐𝑐𝑃𝑃 ∗ 𝐷𝐷𝑒𝑒𝐶𝐶𝐷𝐷𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝑒𝑒)
Additional control variables
25
average prior to the succession year. ROA was calculated as net income before extraordinary items and discontinued operations divided by net assets. Industry performance (Ind_ROA) was calculated for each firm’s succession as the three-year average following the succession. Each year’s industry performance was the average ROA that accounted for all firms in its industry at the 2-digit SIC level. Board size (Boardsize) was taken from MSCI ratings as the total number of directors present on the board in the year of succession.
The CEO dummies are firstly, previous CEO duality (PC_duality) which was coded 1 if the prior CEO also held the position of chairman. Secondly, successor duality (Duality) was coded 1 if the successor held the position of chairman. Thirdly, the variable for if the previous CEO transitioned to chairman (PC_chairman) was included to account for the influence of a chairman on the successor, apart from the case of outsider succession as it was already included. Aside from CEO-related dummies, the categorical variable Year controlled for year effects as the 2008 financial crisis likely affected the performance of firms within the sample.
Models 1 till 7 were firstly taken without the additional controls to infer primary effects. Models 1x till 7x include the controls. This option was chosen as performing OLS means that including additional variables affects all previous variables in the regression.
Model #x 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃_𝑅𝑅𝑅𝑅𝑅𝑅 = "𝑀𝑀𝑃𝑃𝑂𝑂𝑒𝑒𝑀𝑀"
+𝛽𝛽4 ∗ 𝐿𝐿𝐿𝐿𝐷𝐷𝑀𝑀𝑒𝑒𝑃𝑃 + 𝛽𝛽5 ∗ 𝑃𝑃𝐷𝐷𝑒𝑒_𝑅𝑅𝑅𝑅𝑅𝑅 + 𝛽𝛽6 ∗ 𝐼𝐼𝑎𝑎𝑂𝑂_𝑅𝑅𝑅𝑅𝑅𝑅 + 𝛽𝛽7 ∗ 𝐿𝐿𝐿𝐿𝐷𝐷𝑀𝑀𝑒𝑒𝑃𝑃 +𝛽𝛽8 ∗ 𝑌𝑌𝑒𝑒𝐷𝐷𝐷𝐷 + 𝛽𝛽9 ∗ 𝐵𝐵𝑃𝑃𝐷𝐷𝐷𝐷𝑂𝑂𝑃𝑃𝑎𝑎𝐵𝐵𝑒𝑒 + 𝛽𝛽10 ∗ 𝐷𝐷𝐷𝐷𝐷𝐷𝑀𝑀𝑎𝑎𝑃𝑃𝐶𝐶
+𝛽𝛽11 ∗ 𝑃𝑃𝐶𝐶_𝑂𝑂𝐷𝐷𝐷𝐷𝑀𝑀𝑎𝑎𝑃𝑃𝐶𝐶 + 𝛽𝛽12 ∗ 𝑃𝑃𝐶𝐶_𝑐𝑐ℎ𝐷𝐷𝑎𝑎𝐷𝐷𝑎𝑎𝐷𝐷𝑎𝑎 + 𝜀𝜀
Statistical analysis
The hypotheses were tested using ordinary least squares (OLS) in a multiple linear regression (MLR). Meaning that, the linear relationship between (Post_ROA) was measured by multiple variables and interactions of continuous and categorical origin. The reason for using OLS will firstly be discussed, thereafter is the examination of the data and its adjustments to allow for OLS.
Statistical method
26
issues with other methods varied as it was unclear whether the method dealt with the endogeneity of interacting boards and succession, or was too complex to apply, given the available timeframe of this thesis. Furthermore, some methods were hard to infer the effects of succession from, or required certain data conditions that could not be met. In addition, others estimation methods required instrumental variables to allow for estimation and interactions of endogenous variables, as endogeneity issues are common in studying boards. For instance, board composition, CEO type and board size, are all reasonably endogenous (Adams et al, 2010; Wintoki, 2012). Mainly, as their resultant effect on performance, likely impacts the future choice of CEO, composition of the board, and size of the board. This may fall under one of three main sources of endogeneity that bias the dependent variable. Being, omitted variable bias, reverse causality or ‘simultaneity’. To address this concern, researchers can make use of instruments that correlate with an endogenous predictor, yet not with its dependent outcome, thereby accounting for possible bias. However, even if “instrumental variable regressions eliminate endogeneity, they require the identification of strictly exogenous instrumental variables, which is almost impossible in a corporate governance setting” (Guest, 2009, p.392; Wintoki et al., 2012). Thus, instruments for studying boards and successions are generally unavailable, and were not found. As such, OLS was chosen as it could regress a multitude of variables while potentially accounting for endogenous interactions. Furthermore, OLS is easier to infer effects from and was better understood than the other models considered. The method establishes a constant and slope to describe the linear fit of the dependent and its predictors, based on the unique contribution of each predictor included in the regression. Adding interactions to the regression can change the unique contribution of variables; hence, models were first tested with and without the controls, thereby checking the effect of additional controls to coefficients. The statistical programme used was Stata and the online book by Chen et al. (2003) served as the guide for conducting the analyses.
A first examination
27
Imputation methods were avoided as it was assumed better to use actual data. As such, to allow an investigation on successions in general, I tend to the remaining sample including 122 observations. I notify to the reader that by deleting seven observations, I forego the notion of implying significance from the analysis, as is proper conduct (Sweet & Karen, 2012). Nonetheless, the resulting sample allowed checking for a general analysis of relationships within the dataset and a better understanding of the sign and size of variables under investigation. More information about how the OLS assumptions for models 1x till 7x were met is included in Appendix B. The final sample consisted of 122 successions of which 60 (1 deleted) follower, 19 (5 deleted) contender and 43 (1 deleted) outsider successions.
Results
Table 5 reports the means, standard deviations and correlation coefficients of the variables used within the models. The variable CEOtype was split into Follower, Contender and
Outsider to better show the data and correlations on each succession type. Table 6 reports
28
were rightly met with their predicted signs. Allowing to generally state that followers positively affect the post-succession performance following retirement, and that outsiders positively affect performance under guidance of the prior CEO. Furthermore, contenders and outsiders have a positive effect following dismissal.
The board interaction results are as follows. After CEO dismissal, hypothesis 4a is predicted positive for contender succession yet does not hold under either model 4 (b = -.0413) or model 4x (b = -.192). Hypothesis 4c on outsider succession was predicted negative and does not hold for model 4 (b = .382) and model 4x (b = .171). Meaning that both hypotheses 4a and 4c do not hold their predicted signs under models 4 and 4x. After CEO retirement, hypotheses 4b is predicted positive for follower succession, yet is negative in both model 5 (b = -.781) and 5x (b = -.532). The predicted sign for hypothesis 4d does however hold in both models 5 (b = -.242) and 5x (b = -.0538). Seeing as models 4, 4x, 5 and 5x were given a much lower adjusted R-squared and higher, or even positive, criterion scores, it seems likely that these models suffer from misspecification due to endogeneity issues of boards. Furthermore, the findings seem highly unlikely as followers almost certainly do not positively affect post-succession performance after dismissal of the previous CEO.
Lastly, the main effects of boards were investigated. Hypothesis 5 holds the predicted sign under both models 6 (b = .601*) and 6x (b = .476) of which the former is significant. The results thus indicate that outsider boards positively affect post-succession performance after the previous CEO is dismissed, likely due to their preference of strategic change, which will be elaborated on later. In contrast, insider boards are more likely to positively affect performance when the CEO has retired, likely due to a preference for maintaining strategy. This finding relies on the opposite of hypothesis 5, where not outsider but insider boards positively affect post-succession performance. Assessing this hypothesis is possible as the opposite of the ratio variable for outsider representation is inside director representation. Further inspection of the MLR results shows that several control variables were significantly related. The coefficient for the unique effect of Departure was significant in model 1 (b = -.170**), 1x (b = -.169*), 4 (b = -.158*) and 4x (b = -.164*). These models investigated the inside distinction of follower and contender successors. That Departure is significant indicates its importance to defining CEO succession types. Model 3 on outside succession under retirement shows statistically significant coefficients for outside succession (b = -.171*) and PC_chairman (b = -.146*). Seeing as that the components of the interaction, but not the interaction itself is significant, an issue in the regression method may exist. Furthermore, Pre_ROA is significant in models 1x (b = .0191*), 2x (b = .0191.*), 3x (b =
29
Table 5 Mean, standard deviation, and correlations matrix of all variables
30
31
Table 8 Summaries of hypotheses relevant coefficient results of all MLR models
Hypothesis 1 2 3 3a 4a 4c 4b 4d 5a Predicted sign + + - + + - + - + Model 1 1 2 3 4 4 5 5 6 Outcome .0342 .144* -.117 .187 -.0413 .382 -.781 -.242 .601* Model x 1x 1x 2x 3x 4x 4x 5x 5x 6x Outcome .0404 .108 -.0683 .0410 -.192 .171 -.532 -.0538 .476 Difference .0062 .036 .0487 .146 .1507 .211 .249 .1882 .125
Lastly, when considering the correlations matrix in table 5, I look at the statistically significant correlations. PC_duality and PC_chairman correlate with all three succession types. However, only PC_duality correlates with outsider representation on the board. According to the data sample, 59% of boards had a CEO which held duality and in 47% of all successions, the CEO either remained or transitioned to the chairman position. As such, most boards were likely influenced by the duality or chairman power of the previous CEO, which helps explain why 47% of successions consisted of follower succession, especially whilst outsider boards overrepresented the sample. Furthermore, Departure correlates with
Post_ROA, Follower, Contender, PC_chairman and PC_duality. As CEO departure is used in
separating followers from contenders, it is likely that these correlations are significant. However, as departure also significantly correlates with post-succession performance, it is reasonable to state that CEO departure is essential in the succession context and subsequent performance. The variable Year correlates with follower and contender succession, likely because of growth prior to the financial crisis, where all 2007 coefficients are positive, and its subsequent negative effect giving rise to a need for change, where all 2008 coefficients were negative. However, the MLR results show no statistically significant coefficient for Year, likely because of endogeneity issues affecting the models. Nonetheless, the peak year of successions was 2008 with 50 successions taking place within that year, along with a fall in follower and rise in contender successions. In addition, Year significantly correlated with Departure and Ind_ROA suggesting a high rate of CEOs being replaced around the financial crisis. Lastly, the logarithm of sales correlates strongly with pre-succession performance, further supporting the idea that the volume of sales helps explain post-succession performance.
Discussion and conclusions
32
the effects of boards on performance (Bouaine et al, 2014). Moderated by board composition, as it is relevant to the characteristics and the actions of the board (Zald, 1969; Helmich, 1975), I have devised the following research question: how does board composition affect the relationship of different CEO types on performance? The predicted answer to this question was split into three CEO succession types and two board types, of which the effects were categorized through the departure of the previous CEO.
Succession types
The predicted effects of CEO successions hold, although most have not been found significant. When the previous CEO had retired, signaling good pre-succession performance, two succession types were investigated. A follower succession was predicted and found positive. An outsider succession was predicted and found positive if the previous CEO transitioned to the position of chairman. Both of these findings are novel and may contribute to the extant literature on CEO succession types. When the previous CEO had resigned, signaling poor pre-succession performance, again two succession types were investigated. Both the contender and outside succession types were predicted, and found, positive, of which contender succession was found significantly related to post-succession performance.
Board types
Both outsider and insider boards were estimated through the use of a single ratio variable about outside director representation. Higher ratio scores would infer greater representation of outside directors whereas lower scores would indicate greater inside director representation. Although such a variable is but an approximate measure of the effects of both board types, it was used over categorizing boards as either insider or outsider boards. This was done because the sample overrepresented outsider boards, which was due to regulations that increased outside director representation. The results are as follows, the predicted sign of outsider boards is positive after the resignation of the CEO and negative during retirement. The opposite holds for insider boards. The results show a significant correlation with departure, indicating that outside directors are preferable in situations of strategic change whereas inside directors are preferable to maintain strategic direction. This result may be due to inside directors being fearful of change, whereas outside directors may welcome change. As reflected by outside directors initiating the exits of heirs apparent (Cannella & Shen, 2001) and increased competition among top management (Fama, 1980). Both are aspects that facilitate strategic change. Although continuous improvement may be the maintained strategic direction, it is unlikely that inside directors would initiate such a strategy due to their inside successor preference.
The interactions of CEO succession types and board types
