CEO SUCCESSIONS AND FIRM PERFORMANCE:
Internal Promotion or External Recruitment?
Shirley Brandt 10216545
Finance & Organization: Organizational Economics Silvia Dominguez Martinez
January 2015
ABSTRACT
This paper studies the difference in firm performance between CEO successor types: internally promoted, or externally recruited. Provided that a successor is externally recruited, we make a split up between industry insiders, and industry outsider. The sample used in this study contains 45
successions, which took place in large US firms in the year 2001. Both operating return on assets (OROA), as well as Tobin’s q are used as performance measures. A seven-year period centered around the succession event is used to study the effect of the CEO succession on firm performance. We compute the differences in firm performance between pre-succession, and post-succession performance for each successor type. Afterwards, the seven-year period is split-up in two periods, 1998-2000, and 2000-2004. The analyses for Tobin’s q, and OROA yield different results.
Both unadjusted Tobin’s q, as well as Tobin’s q, adjusted for industry- and pre-succession performance, provide evidence that firms that appoint insiders, experience an increase in firm performance in the years prior transition. When the change in performance from 2000 to 2004 for Tobin’s q is studied, external candidates seem to outperform their internal counterpart. When we separate external candidates by successor origin, we see that industry outsiders are outperforming industry insiders. On the other hand, both unadjusted, and adjusted OROA provide evidence that firms that appoint outsiders, are outperforming firms which appoint internal candidates in the years prior succession. Between 2000 and 2004, firm insiders are outperforming firm outsiders, and industry insiders are outperforming industry outsiders. However, differences are not significantly different.
1 Statement of Originality
This document is written by Student 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.
2 Table of content Statement of Originality 1 Table of content 2 1. Introduction 3 2. Theoretical background 5
2.1 CEO, managerial – and CEO successions 5
2.2 Executive turnover 6
2.3 Successors: internal versus external 7
2.4 Industry insiders versus industry outsiders 12
3. Methodology and Sample 13
3.1 Sample 13 3.2 Performance measures 14 3.3 Method 15 3.4 Test 17 4. Results 18 4.1 Sample 18 4.2 Matching procedure 19 4.3 Results 21
4.3.1 Successors: internal versus external 21
4.3.2 Industry outsider versus industry outsider 25
5. Discussion and conclusion 26
5.1 Sample 28
5.2 Matching 29
5.3 Insider versus outsider 29
5.4 Industry insider versus industry outsider 32
5.5 Conclusion 33
Reference list 35
3 1.Introduction
The CEO is the one who has the ultimate responsibility of an organization (Dalton & Kesner, 1985). At some point, however, organizations need to find a successor, to whom this power and authority can be passed. This obviously is a critical decision, because the future of an organization can be largely influenced by who is appointed as successor (Kesner & Sebora, 1994). Existing literature about succession choice, and succession context focuses on several dimensions. This paper will focus on the relation between succession type, and firm performance. Greenwald (1986) mentions the employer’s risk-aversion as a reason to promote an internal candidate as opposed to hiring an external candidate. Many studies address the positive effect of an external successor on firm performance and shareholder gains (Borokhovich et al., 1996; Ferris et al., 2007; Helmich, 1974; Huson et al.,2004;
Murphy & Zábojník, 2004). Which type of successor is, in the end, more beneficial to a firm’s performance? A brief outline of several studies, and their limitations, will be described now.
Denis, and Denis (1995) study executive turnover events, in which they use a sample from 1985 to 1988. They do not make a distinction between CEO turnover events, and other executives’ turnover events. The result of their study is that preceding forced resignations of top executives, operating performance declines. Afterwards, large improvements in operating performance are present. Denis, and Denis’ (1995) study has some limitations. First, only unadjusted and industry-adjusted measures are used. Since there is no correction for pre-event performance, mean reversion might be a problem. It is possible that the pure
succession effect on performance is influenced by other factors than the executive turnover event. Only controlling for industry performance may deteriorate the reliability of the results, as compared to results in which is controlled for both industry- and pre-event performance. The second drawback is the choice to use operating return on assets as performance measure, since this could be subject to manipulation (Péréz-González, 2006). Huson, Malatesta, and Parrino (2004) study firm performance surrounding successions as well. As opposed to Denis and Denis (1995), Huson et al. (2004) limit their focus to CEO successions only. Their sample consists of large public US firms, in which a succession took place
between 1971 and 1994. Operating return on assets, and stock performance are used to measure the firms’ operating performance. This is done for the seven-year period centered around the succession event. They use Barber and Lyon’s (1996) matching firm technique, to correct for both industry, and pre-event performance. Huson et al. (2004) find that the degree of the firm’s operating performance and stock performance is more higher for firms that promote external candidates, rather than promoting internal candidates. Again, the used operating performance measure, OROA, is a drawback. It could be subject to manipulation, and misleading, since it only displays current firm performance (Pérez-González, 2006). Both
4
papers do make a distinction between company insiders, and company outsiders. However, the industry origin of the company outsider is not taken into account. The importance of the successor’s industry-origin is stressed by Parrino (1997), and Jalal and Prezas (2012). Parrino (1997) studies CEO turnover, in which he makes a distinction between industry insiders, and industry outsiders. His main result is that intra-industry CEO appointments increase with the homogeneity of the industry. Jalal and Prezas (2012) mention the lack of academic literature covering successors’ industry origin. They limit their scope to a
successor’s industry origin only, firm performance of firm insiders are not taken into account. This paper will focus on the difference in firm performance surrounding CEO successions, making a clear distinction between internal promotion and external recruitment. It contributes to the already existing literature by some small, but useful combinations of studies, and improvements. To overcome the problem of OROA as performance measure, which can be subject to manipulation, Tobin’s q will also be included as performance measure in this paper. Tobin’s q is a measure of market valuation and investment opportunities, and is an indicator for a company’s growth potential (Jalal & Prezas, 2012; Klapper & Love, 2004). For each company outsider, we make a distinction between industry insiders, and industry outsiders. At last, we will use Barber, and Lyon’s matching method (1996) to compose a relevant control group for each sample firm. The goal of this paper is to study the effect of the type of CEO successor on firm performance: is there a difference in firm performance
between internal promotion and external recruitment?
We will study differences in firm performance, during a seven-year period centered around the CEO succession. Firms which experienced a CEO turnover event in 2001 are included in the sample. The difference between pre-succession, and post-succession performance will be used to analyze the performance of the successor types. First, the three-year average performance prior is compared to the three-year average performance following the
succession event. Afterwards, the seven-year period is split-up in two smaller periods: 1998-2000 to study the change in performance prior succession, and 1998-2000-2004 to study the change in firm performance surrounding the CEO succession event. Non-parametric tests are used to test these differences in firm performance. The analyses of results for OROA, and Tobin’s q yield different results. Both unadjusted, and Tobin’s q adjusted for industry, and pre-event performance provide evidence that firms that appoint insiders, experience an increase in firm performance in the years prior transition. When the change in performance from 2000 to 2004 is studied, external candidates seem to outperform their internal
counterpart. When we separate external candidates by successor origin, we see that industry outsiders are outperforming industry insiders. On the other hand, both unadjusted, and adjusted OROA provide evidence that firms that appoint outsiders, are outperforming firms
5
which appoint internal candidates in the years prior succession. Between 2000 and 2004, firm insiders are outperforming firm outsiders, and industry insiders are outperforming
industry outsiders. However, we fail to find significant evidence for the outperformance of one successor type, or the other.
This paper will continue in the following way. First, a theoretical background will be provided. In section 2, the existing literature covering executive, and CEO turnover will be provided. This literature will be discussed, and hypotheses that will be tested in this paper will be derived. Section 4, the methodology, will focus on both the sample composition, and the research method. This will give insight in the way this study is conducted. In section 4, the results of this paper are shown, and analyzed. This paper will conclude with section 5. There, we will compare our results to the results found in previous literature. We will pay attention to differences, and similarities that may arise. Furthermore, reasons for differences, or
similarities that appear will be discussed.
2.Theoretical Background
This section starts with a review of the existing academic literature surrounding this topic. First, it will stress the importance of a CEO within a company, followed by the importance of managerial successions, and CEO successions in specific. The different types of
successors, accompanied by their pros and cons will be considered. Finally, the main findings of the extant literature will be mentioned, and the hypotheses will be derived.
2.1 CEO, managerial – and CEO successions
The Chief Executive Officer is ultimately responsible, and accountable for many factors within an organization. The organization’s performance, together with her strategy, structure, and environment, are factors the CEO has to decide about. As mentioned by several
authors, an organization could be seen as the reflection of the organization’s top managers and their decisions (Kesner & Sebora, 1994; Hambrick & Mason, 1984). The CEO is the one who directs a company towards its goals. At some point, however, companies need to find a successor to whom power, and authority could be passed. The succession event has a major effect on organization’s environment, both internal, and external. Within the organization, employees may experience insecurities regarding the organization’s future and the
accompanying influence this may have on their job or tasks. For external parties, for example shareholders, suppliers, and customers, the CEO is the face of the organization. The CEO’s tasks are hard to define, and it is even harder to structure their day-to-day work. This makes it even more difficult to find a successor who is capable to successfully run a company. The exact criteria that have to be fulfilled are not clear. The CEO succession event is often seen
6
as a signal about the future of an organization (Kesner & Sebora, 1994). This future is in the hands of the Board of Directors. The Board is responsible for the succession decision, which is one of the most important choices the Board has to make (Huson, Parrino, & Laura, 2001). The Board of Directors is in a small, or bigger fraction, composed of outside directors.
Outside directors are, in general, relatively unfamiliar with the firm and its activities. The above mentioned factors combined – CEO’s role in the organization, the CEO job’s nature, and the persons in charge of the succession decision – makes the CEO succession decision different from the turnover decision at lower levels. In short: the CEO succession choice has major implications for the company, influencing both the short-run and the long-run
investment, operational and financial decisions made.
2.2 Executive turnover
In this section, the executive turnover in general will be discussed, together with the effect on firm performance. Over time, the amount of CEO positions available has increased, reflecting the increasing the demand for executive talent (Evans, Luo, & Nagarajan, 2014). Not only the demand for executive talent, but also the quality of CEO replacement decisions has
improved over time. Huson, Malatesta, and Parrino (2004) find that managerial quality that follows from the succession event has increased over time. They split their sample of large US firms into two: the first part covers the 1971-1982 period, the second part 1983-1994. They find that the increase in firm performance following the turnover event is significantly higher in the second part, than in the first part. One conclusion may be that the replacement decisions’ quality has improved over time. Investors also seem to believe that executive turnover is a positive event in a firm’s lifecycle. Several studies find that stocks yield a positive abnormal return in the days following a turnover announcement. These abnormal stock returns will most likely be followed by an improved firm performance (Bonnier & Bruner, 1989; Huson et al., 2004; Weisbach, 1988). This is especially the case with forced, rather than voluntary resignations. Forced resignations are preceded by large and significant declines, and followed by large improvements in operating performance. For normal resignations, this is only a small increase (Denis & Denis, 1995).
There are several factors that influence the CEO turnover decision. The composition of the Board of Directors, the amount of stocks owned by the CEO, the availability of a capable successor and the ability to accurately monitor and evaluate CEO performance are some examples (Parrino, 1997; Schleifer & Vishny, 1989; Weisbach, 1988). Besides the factors just mentioned, there are other factors forecasting a CEO succession event. The CEO’s age is important in explaining CEO turnover (Murphy, 1999). Once a CEO reaches a certain age, he retires and a successor has to be found. A last factor determining CEO turnover is firm
7
performance. This relation is addressed in several papers. We first mention the scapegoat hypothesis. This hypothesis holds that poor performance does not arise from CEO quality, but is the sole result of bad luck. The successor will, according to this hypothesis, not possess better skills, and knowledge than his or her predecessor. The replaced candidate could therefore be seen as a scapegoat (Huson et al., 2004). However, managerial
successions are triggered by incredibly bad firm performance, resulting from chance. As a result, the expected change in firm performance surrounding turnover events, is positive. Stock returns are indicators of a firm’s financial performance. Warner, Watts, and Wruck (1988) find that firms are more likely to change their top management, when stock returns are low. This negative relation between stock returns, and CEO successions is found more frequently (Coughlan & Smith, 1985; Weisbach, 1988). Furthermore, when a CEO
succession took place, the likelihood of turnover of other executives increases (Coughlan & Smith, 1985). Other studies focus not only on stock returns as a firm’s performance indicator, but study firm performance in general. Their general consensus is that CEO turnover often follows a period of poor firm performance (Brickley, 2003; Evans et al., 2014; Murphy, 1999). Contrary to the scapegoat hypothesis you have the increased management hypothesis (Huson et al., 2004). This hypothesis says that managerial quality has the tendency to increase following forced management turnovers. As a result, future firm performance is expected to increase. This expected increase in firm performance follows from two factors: the improved managerial quality, and the reversion of the executive’s luck to a normal level. To conclude: a CEO succession is often the result of a period of poor firm performance.
2.3 Successors: internal and external
The Board of Directors has the choice between different types of CEO successors. One option is to promote an internal candidate, who is already active within the organization. The other option is to recruit an external candidate to the CEO position. This paragraph is
especially aimed at outlining the main benefits, and drawbacks for each type of successor.
Internal successors are the most common type of CEO successors. Agrawal et al. (2006) find that, between 1974 and 1995, 80 percent of large US companies appointed an internal candidate to the CEO position. On the other hand, however, there is an increasing trend in external hires over the last decades (Huson et al., 2001). In the 70’s, only 15 percent of the CEO positions was occupied by external candidates, whereas in the 90’s, this was already 26 percent (Murphy & Zábojník, 2004). Outsiders have served the company for less than a year, before being promoted to the CEO position. The increasing trend in external hires has to do with the increasing importance of general skills, relative to firm-specific skills (Murphy & Zábojník, 2004). The importance of general skills, for example the ability to manage a large
8
company, is increasing. This explains the rise in external hires (Murphy & Zábojník, 2004). Nowadays, specific information on a firm’s buyers, suppliers, and competitors does not have to be obtained during years of service within a firm. Executives now have easy access to huge amounts of information, since this is widely available in the firm’s digital system, and on the internet. Firm-specific information has, therefore, lost importance.
Succession literature focuses on several elements surrounding CEO successions. For example the company size – successor-type relationship, and the amount, and composition of the executives’ compensation package. Regarding the relation between company size and the type of successor, mixed evidence is found. Some studies provide evidence that smaller firms tend to have a stronger preference for external successors (Dalton & Kesner, 1983; Helmich & Brown, 1972), whereas others suggest that smaller firms are more likely to hire an internal candidate (Schwartz & Menon, 1985). Therefore, the relation between company size, and successor type remains ambiguous. Literature also compares the CEO compensation of insiders with outsiders’ compensation. There is a general trend in literature that externally hired CEOs receive a higher compensation than their internal counterpart (Murphy &
Zábojník, 2004; Ferris et al., 2007). External candidates might have several job opportunities, before they choose to become the CEO of a certain company. The Board of Directors has to give the external candidate a considerably attractive offer, in order to appoint him or her. This is the result of the better bargaining position outsiders have, relative to that of the Board of Directors (Elsaid & Davidson, 2009). Regarding the composition of the CEO compensation packages for the different types, there is mixed evidence. One study finds that the
compensation for outsiders consists of more fixed pay, relative to risk-dependent pay (Elsaid & Davidson, 2009). Other authors find that outsiders receive less fixed pay, and more
incentive pay. This higher fraction of incentive pay appears to lead to a better operating performance (Ferris et al., 2007).
The choice to either promote an internal candidate, or to recruit one from outside the company, also has to do with the type of resignation of the former CEO. This former CEO could have been forced to leave, or might have left voluntary. Although a forced departure could be the result of many factors, one of the main reasons is poor firm performance. Most forced departures are followed by the appointment of an external successor (Brickley, 2003; Huson et al., 2001; Parrino, 1997). The choice to hire an outsider has, most likely, to do with the general view that outsiders have the ability to change firm policy. Insiders are often too tight with the existing firm policy, which makes it hard for them to break with current firm policy. Relatively well, or well performing companies might be, in this light, leaning to the appointment of insiders, who are likely to retain firm strategies. Poorly performing firms, as said before, may well choose an external candidate to constitute change (Dalton & Kesner,
9
1985; Hambrick & Mason, 1984; Helmich & Brown, 1972). Since external candidates are not that familiar with current firm policy, it is easier for them to break with the past. However, the argument that outsiders are able to constitute change need some caution. The reason for this, is the potential unmet main assumptions underlying the relation between poor performance and outside replacements (Dalton & Kesner, 1983). The first underlying assumption is that externally recruited successors are able to change firm policy. Changing firm policy, however, is not only in the hands of the CEO. There are many people involved in policy changes, and organizations are complex. The internal, and external stakeholders could constrain the successor’s ability to make successful changes in firm strategy. The second underlying assumption has to do with the ability of a company to attract a capable successor. Qualified external candidates may be unwilling to become the head of a failing company. It could damage the high reputation the CEO might have established, which many candidates are not willing to risk. Furthermore, the appointment of an external candidate comes along with risk for the company, due to uncertainty concerning the exact ability of this candidate (Allen, Panian, & Lotz, 1979; Helmich, 1976; Dalton & Kesner, 1983). At last, there is the fear of the top executives losing their job, after an outsider has been appointed.
Turnaround strategies are likely to come along with the replacement of top managers. The Board of Directors might choose to ignore external candidates in the replacement decision, in order to save their own job (Dalton & Kesner, 1985). Outsiders might face a difficulty in CEO successions, as compared to insiders. This handicapping of outsiders implies that the
likelihood of the appointment of an insider to the CEO position, is greater than for an outsider (Agrawel, Knoeber, & Tsoulouhas, 2006; Chan, 1996). This handicapping of outsiders will be described in more detail later on in this section.
Zajac (1990) addresses the principal-agent relationship, to explain the difference in firm performance between both types of successors. In his paper, he uses an agency perspective to study the CEO succession event. In a principal-agent relationship, more extensively addressed by Jensen and Meckling (1976), the principal contracts the agent to take actions on his behalf. The agent’s actions are not perfectly observable. The principal-agent
relationship also applies to the relationship between the Board of Directors, and the CEO. Two fundamental characteristics underlying the relationship are goal incongruence, and information asymmetry. We focus here on the latter one, information asymmetry. The Board of Directors is, ceteris paribus, more likely to know the characteristics of the insiders, rather than the outsiders’ characteristics. Therefore, the principal-agent problem will be less severe in the BoD-insider relationship, than in the BoD-outsider relationship. This implies that, focusing on the relation between the principal and the agent, the decision to hire an external CEO will be, on average, a bad decision (Zajac, 1990).
10
Whereas outsiders are appointed to induce change, internal successors are often appointed for the continuity they provide (Chan, 1996; Kesner & Sebora, 1994). Insiders are likely to adopt a maintenance strategy. The impact of the succession event on the firm’s employees, and other stakeholders will be limited in this case. Another benefit of hiring an insider, is the fact that they possess firm-specific skills and knowledge. These skills, and knowledge are obtained by the inside candidates during their years of service within the firm. Information about the firm’s products, buyers, and suppliers are useful while running the company. Acquiring this firm-specific knowledge will not take as much time as it used to take in the past, due to the existence of the digital information systems. However, outsiders still need time before being completely familiar with the firm’s specific products, services, etcetera. Same is true for the firm’s social networks with all kinds of stakeholders. Internal candidates already established those networks, while an external candidate did not. At last, there is less uncertainty involved with the appointment of internal candidate, rather than an external candidate (Greenwald, 1986; Zajac, 1990). Whereas an outsider has the possibility to hide some of his characteristics, the characteristics of the internal candidate are, mostly, already known. The internal CEOs already exposed his skills and knowledge throughout the
employment at the specific firm. Summarizing: insiders possess firm-specific skills,
knowledge, and the social network. Inside candidates provide continuity within the firm, and cause less uncertainty regarding their skills, as opposed to external candidates.
Besides the drawbacks of hiring external successors, outside successions have been proved successful. Many studies provide evidence for the better performance outside candidates yield over inside candidates. Both operating performance, as well as stock market
performance are studied. Post-succession abnormal stock returns are significantly positive for external successions, in comparison with internal successions (Adam & Mansi, 2009; Borokhovich, Parrino, & Trapani, 1996). Therefore, the appointment of an external candidate is perceived to be more beneficial to investors than inside appointments. Not only stock market performance, but also operating performance seems to be higher for successions in which outsiders are appointed. The increase in operating performance measures is
significantly higher when an external candidate, rather than an internal candidate is
appointed (Borokhovich et al., 1996; Ferris et al., 2007; Helmich, 1974; Huson et al., 2004; Murphy & Zábojník, 2004).
The main hypothesis that will be tested in this paper, will be derived from the findings of the most relevant literature on the relation between firm performance and successor type. These findings will be outlined below.
11
Chan (1996) mentions in his paper the difficulty outsiders face in CEO successions. The insider-outsider decision is analyzed within the framework of an economic contest. He finds that in a competition framework, outsiders are only recruited if they prove to be significantly better than insiders. This results in a negative incentive effect for the firm’s current
employees. Therefore, the firm might want to mitigate this negative incentive effect of the open competition. A firm can do this by awarding a competitive handicap to inside
candidates. This means: a raise in the probability of internal candidates to win the contest As a result, in order for an outsider to be appointed, a greater increase in managerial quality of an outsider is required. This implies better post-succession firm performance for outsiders, than for insiders.
Performance changes surrounding turnover events are studied by several authors. Denis, and Denis (1995) study more than 900 management turnover events. The data stems from a period covering the years 1985 to 1988. They do not make a clear distinction between CEO – and other executives’ successions. Operating return on assets is used as performance measure, to quantify firm performance for the three years before, and three years after the succession event. Denis and Denis find that the average, and median industry-adjusted OROA increase between the year preceding and the two, or three years following the turnover event. This increase in firm performance is higher when the manager has been forced to leave, rather than quitted service voluntary. Péréz-González (2006) studies the impact of inherited control on a company’s performance. His sample consists of more than 300 firms, where a succession took place in the year 1994. A clear distinction is made between family successors, and unrelated successors. An incoming CEO is regarded as a family successor, when he is related by blood or marriage to his predecessor CEO, to a company’s founder, or to a large shareholder. Stock market performance, operating return on assets, and market-to-book ratios are used here to study the difference in firm performance between the two types of successors. The analysis of OROA and M-B ratios is done, among other things, by comparing differences of means. Besides unadjusted measures of firm performance, also performance measures corrected for industry and pre-event performance are regarded. Each firm’s two-digit SIC code serves as the identifier to compose a relevant industry-control group. The median of the relevant industry is subtracted from the unadjusted OROA, to end up with an industry-adjusted performance measure. The control group used here consists of firms that operate in the same industry. Performance-adjusted measures are created by sorting the firms into deciles, performance is measured in the year prior transition. The unadjusted measures are corrected by the median of this control group. The three-year average performance prior and following the turnover event is studied. There is a split-up made between T= -3 to T= -1, and T= -1 to T= +3. The year 0 is the year in which the
12
succession takes place, T= – 3 meaning three years before transition. This split-up is made to control for differential pre-transition trends, although it appears that this trend is not significantly present. Péréz-Gonzáles (2006) concludes that firms that promote family-type successors, underperform firms that promote unrelated successors.
The basis of this paper, is the study of Huson et al. (2004). They also study firm performance surrounding CEO successions. Their sample consists of 1344 CEO succession, which took place in the period 1971-1994 at large US public firms. OROA and stock performance are used to measure the firms’ operating performance, which is done for the seven-year period centered around the succession event. Performance results are corrected, for industry and pre-event performance, using Barber and Lyon’s (1996) matching firm technique. They find that external recruitment has a more positive impact on firm performance, compared to internal promotion.
For this, our main hypothesis follows:
H1 Firm performance increases following a CEO succession event, and outside
successors are outperforming inside successors 2.4 Industry insiders versus industry outsiders
The choice to either promote an internal candidate, or recruit an external candidate, is discussed in the previous section. Besides that, some attention has also been given to the choice to recruit either an industry insiders, or an industry outsider. However, the amount of research in this field is quite small. Parrino (1997) shows that CEO appointments within the same industry are less costly when the industry is homogenous. In order to reduce the likelihood of costly errors, most outsiders who replace fired CEOs, have some industry experience. This industry experience reduces the learning time they need, to become familiar working in a certain industry (Parrino, 1997). Jalal and Prezas (2012) stress the importance of the successor’s industry origin. They mention the lack of academic literature that makes the distinction between industry insiders, and industry outsiders. Jalal and Prezas (2012) use data of slightly more than 500 CEO succession between 1993 and 2009. Successions in which firm insiders are appointed, are not taken into consideration. Their reason for this exclusion is the fact that the appointment of firm-insiders could be driven by more than pure an candidate’s quality. Jalal and Prezas (2012) focus on the difference between firm
performance for intra- and inter-industry appointments. To identify the industry-origin of the CEO successor, the GICS (Global Industry Classification Standard) is used, in which 68 industries are classified. They find that the positive results, immediately after the
announcement of the CEO succession, are higher for intra-industry hiring firms than inter-industry hiring firms. This is, at least in part, the result of the inter-industry-specific experience the
13
intra-industry hired CEOs have, so that they do not have a significant learning period. In the long-run (5 year post-succession period), however, firms hiring from a different industry appear to have higher profitability, growth potential and stock returns than their intra-industry hiring counterparts. At first, the market seems to be not so positive about the new CEO, hired from a different industry, but once new firm policies are introduced and firm performance has increased, their opinion turns into a more positive direction.
Since we test firm performance in the first years following the CEO turnover event, the second hypothesis tested in this paper is:
H2 Provided that the successor is externally recruited, industry insiders will outperform
industry outsiders 3. Methodology and Sample
In this section, we will first discuss our sample. Hereafter, our research method will be described: the performance measures used, the research method, and the actual test.
3.1 Sample
The sample used in this paper contains CEO-successions of large US firms in the year 2001. A seven-year period centered around the succession event will be studied here. Since a more recent sample would include data from crisis years, the year 2001 has been chosen. Moreover, a more recent sample would prevent us to evaluate the performance during a three-year period following the turnover event. Therefore, we focus on successions in 2001. The executives’ information are retrieved from ExecuCom, financial information from
COMPUSTAT. Firms must satisfy the following conditions, in order to be included in the sample:
There has been only one CEO succession in the sample period
The necessary information on the CEO, and its successor must be available: date when agent joined the company, date when agent became CEO, date when agent left as CEO
Financial information is available for the seven-year period, centered around the succession event:
book value of assets book value of debt
market value of shareholders’ equity operating income
14
Successions will be manually identified by comparing the columns date left as CEO and date became CEO for every company. When one CEO leaves, and another one is appointed in the year 2001, we identify this as a CEO succession. We check whether or not this was the only CEO succession event between 1998 and 2004. When another CEO succession has taken place in this period, the firm will be excluded from the sample. Otherwise, the firm’s performance cannot be studied in a reliable way. Firm performance will be influenced, in this case, by the presence of another succession event, and the other CEO(s). Furthermore, financial information must be available for the three years preceding, and following the turnover event. This in order to calculate the firms’ financial performance. In the appendix, table 4, total assets are presented for both the firms included in the sample, as well as for the firms which did have a CEO succession, but did not satisfy the conditions mentioned above. Main reason for the exclusion of firms from the sample is the occurrence of more than one CEO succession.
The cutoff-point between an internal and an external successor is ‘one year of service’. A successor will be identified as ‘external’ when the length of service in the specific company is less than a year (Huson et al., 2004; Shen & Cannella Jr., 2002; Wiersema & Bantel, 1993). To identify whether an external successor is an industry insider, or an industry outsider, the two-digit SIC code is used. Executive information retrieved from ExecuCom will be used to find out each successor’s career history. For some CEO successors, former career
information is not available in ExecuCom. This could be the case when they used to work outside the United States, for which no information is available is this database. For these successors, we will manually identify their industry origin. The ExecuCom database is available from 1992 on. Therefore, we identify an external successor as industry outsider, when he has served a company in a different industry in the period 1992-2001.
The sample contains 45 successions: 25 internal, and 20 external successions, of which 4 are industry insiders, and 16 industry outsiders.
3.2 Performance measures
To study the effect of successor type on firm performance, performance measures have to be selected. Two performance measures will be used in this study and are described in more detail below.
𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑖𝑛𝑐𝑜𝑚𝑒
𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 = 𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑎𝑠𝑠𝑒𝑡𝑠 (𝑂𝑅𝑂𝐴)
Operating return on assets (OROA). OROA is a widely used and well-understood measure in succession studies, and is not affected by changes in capital structure
15
(Barber & Lyon, 1996; Denis & Denis, 1995; Huson et al., 2004; Pérez-González, 2006; Zajac, 1990). Operating income is a quite clean measure of the productivity of operating assets, and scaled to total assets to be able to compare performance across different firm. The nominator consists of operating income excluding interest expense, special items, income taxes and minority interest. The denominator can consists of the average of total assets at the beginning and ending of the period of interest, but in most of the literature the end-of-period value of total assets is used. According to Barber and Lyon (1996), this difference does not affect results. In this paper, end-of-period value of assets will be used to scale operating income.
𝐵𝑜𝑜𝑘 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑑𝑒𝑏𝑡+𝑚𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑠ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑒𝑟𝑠′𝑒𝑞𝑢𝑖𝑡𝑦
𝑏𝑜𝑜𝑘 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 = 𝑇𝑜𝑏𝑖𝑛
′𝑠 𝑞
OROA can be misleading since it only displays current firm performance and executives can severely manipulate it (Pérez-González, 2006). To overcome this problem, Tobin’s q will be included as additional performance measure. Tobin’s q is a measure of market valuation and investment opportunities, and is an indicator for a company’s growth potential (Jalal & Prezas, 2012; Klapper & Love, 2004).
As already mentioned, the time-span for which the performance will be measured is three years before, and three years after succession. The year in which the succession takes place is year ‘zero’, and will not be taken into account in the performance analyses (Denis & Denis, 1995; Evans III et al., 2014; Huson et al., 2004; Pérez-Gonzales, 2006).
3.3 Method
We will conduct this study by analyzing pre-succession, and post-succession firm
performance. The change between those two will be used to identify the succession-effect. We perform these analyses for all successions, as well as we split-it up internal, and external successions. As mentioned before, OROA and Tobin’s q will be used as performance
measures. We distinguish a treatment group, and a control group. The treatment group consists of firms with succession in the year 2001. The control group contains firms which are similar to the firms in the treatment group, regarding industry- and pre-event
performance, but in which no CEO succession took place between 1998 and 2004.
In this paper, we correct for industry-, and pre-event performance. This is done in order to reduce the mean reversion problem. Mean reversion is a theory that suggests that prices and returns move back towards the (historical or industry) mean over time. The mean reversion problem could have a large influence on the reliability of results. Without matching, and
16
therefore without correcting for industry- and pre-event performance, the test results could be biased. Large firms might have the tendency to choose outsiders, and therefore the firm performance for outside successors could be higher. This higher firm performance is in this case not attributable to the type of successor, but to the company size. Furthermore, biases may arise which are attributable to the firm’s industry. It might be that a specific industry has the tendency to appoint a certain type of CEO successor, instead of another. This preference for a specific successor type could influence results, when unadjusted results are analyzed. For example: a certain, well-performing industry, might have a preference for an internal candidate. This could lead to the suggestion that internal candidates are outperforming the external candidates. In this case, however, this is the result of industry performance, instead of the internal candidates’ performances (Barber & Lyon, 1996).
Barber and Lyon (1996) introduce a method, which entails that each sample firm is matched to one or more control firms. We will follow their method to correct the unadjusted results of this study. Several types of matching are studied: matching based only on two-digit SIC codes, four-digit SIC codes, pre-event performance or a combination of the former three. Barber and Lyon (1996) show that when sample firms are not accurately matched to control firms with the same pre-succession performance, a misspecification of results will arise. Matching based on pre-event performance, and the two-digit SIC code, is proved to provide the most successful matching result. Therefore, we will use both pre-event performance, and the two-digit SIC code to perform company matching. Barber and Lyon (1996) match in their paper on both firm performance in the year prior succession, as well as the average pre-event performance of the three years prior the succession pre-event. Matching based on the three-year average does not lead to more accurate results than the matching based on the one-year prior event matching. Therefore, pre-event performance matching will be based on firm performance, OROA in specific, in the year prior succession, 2000. The control firm’s OROA must lie within a 90 to 110 percent range of the treatment firm’s OROA. The same control group will be used, for both OROA and Tobin’s q analyses. The reason for this is the ability to compare results. Control firms will be chosen from a list, which contains all firms in which no succession took place between 1998 and 2004, and for which the necessary financial ratios could be computed.
Sometimes, more than one control firm for a sample firm can be found. In this case, the median of this control group is used to correct the unadjusted result. Sometimes, no control firm can be found, based on the pre-event performance and two-digit SIC code requirements. Removing those firms from the sample, leads to a selection bias. To overcome this bias, some theoretical rules will be followed for alternative matching. First, the two-digit SIC code requirement will be relaxed to the one-digit SIC code, while holding the pre-performance
17
requirement. When this still does not lead to a match, the industry requirement will be dismissed entirely. Pre-event performance is more important in matching firms, than the industry in which they are operating (Barber & Lyon, 1996). At last, when no control firm is found with performance within the 90-110% range, the firm with performance closest to the firm in question will be used. After matching sample firms to control firms, the median of the control group’s firm performance will be subtracted from the accompanying unadjusted firm performance.
In related literature, corporate governance characteristics are also taken into account in the analysis of changes in operating performance. However, Péréz-González (2006) shows that board ownership has no significant effect on changes in operating performance, and Huson et al. (2004) only find weak evidence for the existence of a reliable relation between board composition and institutional shares. For this reason, and the fact that corporate governance characteristics does not belong to the focus of this study, these characteristics are not taken into account here.
3.4 Test
As mentioned above, the hypothesis is tested by comparing differences in performance changes between insiders and outsiders. We based our method on the method used by Péréz-González (2006). The intuition is that, unless drastic events occur, negligible
differences in OROA, and Tobin’s q should be expected surrounding CEO turnover events (Péréz-González, 2006). We will present unadjusted, and industry- and performance adjusted results. The median of the control group’s performance will be subtracted from the unadjusted performance. Result is firm performance, being either OROA or Tobin’s q, adjusted for performance- and industry effects.
The time-span for which we test differences in performance, is the seven-year period
centered around the succession event, being 2001 in this paper. We compare the three-year average performance prior succession, to the three-year average performance following transition. A split-up is made in the seven-year period centered around succession (Péréz-González, 2006). Pre-event differences in performance will be compared, and tested, for the period T= -3 to T= -1. This means: three years prior, to one year prior succession. To
examine the change in firm performance surrounding the CEO-succession, performance changes for the period one year prior, to three years after succession, T= -1 to T= +3, will be calculated (Pérez-González, 2006). This is done for all types of successors combined, internal successions only, and external successions only. Afterwards, the external successions will be split-up in industry insiders, and industry outsiders.
18
First, we calculate the difference between pre-event and post-event performance. These differences will be tested by the Wilcoxon signed-rank test. It tests the equality of two related observations. In this case, the firm’s the three-year average performance prior, and the three-year average performance following turnover. The Wilcoxon signed-rank test is the non-parametric version of the paired-sample t-test. It does not assume that the difference between two variables is normally distributed. Since we have a small sample, this non-parametric test yields to more powerful and reliable results, than a simple t-test (Barber & Lyon, 1996). The next step is to compare these performance differences for the different types of successors. To test this difference-in-differences, we again use a non-parametric test, but this time for independent samples. This is the two-sample Wilcoxon rank-sum test, also called the Mann-Whitney test. This test is also used to compare industry insiders to industry outsiders.
4. Results
In this section, we will provide the results of our study. We start with an extensive sample description, followed by the result of our matching procedure. In the third paragraph, we will present the results of our data study.
4.1 Sample
First, we will present a brief overview of the firms present in the sample. Table 4, in the appendix, shows the amounts of total assets for our sample firms, and the firms in which a CEO succession took place, but did not satisfy the imposed criteria (relevant information available, only one CEO succession). It follows from this table, that the amount of total assets in our sample is smaller than the amount of assets of the excluded firms. The main reason for exclusion of these firms is the fact that more than one CEO succession took place. The difference of these two groups indicates that our sample firms differ in size,
measured by total assets, from other firms in which a succession took place in the year 2001.
Table 1 shows summary information on the firm’s company size, providing information on the average and median amounts of total assets, number of employees, Tobin’s q, and OROA. All numbers relate to the three-year average prior the CEO succession event. Numbers based on the year prior turnover, may anticipate on the occurrence of the turnover event. In the appendix, table 5, the separate numbers for each of the three years prior succession are depicted for the amount of total assets, and number of employees. From this, it follows that, in our sample, the separate years do not show a bias towards the succession event.
Therefore, using the year 2000 to match the control firms does not cause a bias, when compared to the three-year average.
19
The summary information in table 1 is the starting point from which we analyze the insider-outsider decision. As follows from this table, the average company size, both in total assets and number of employees, is larger when the successor is internally promoted. This also holds for the median amount of total assets, which is also larger in case of internal
promotion. The amount of average total assets for firms that appoint insiders, is more than three times as high, than this amount for firms that appoint outsiders. Same holds for number of employees. Our control group displays somewhat lower numbers than our treatment group, regarding total assets, and number of employees. The median number of Tobin’s q, and OROA, are relatively comparable for our treatment and control group. Table 6, which can be found in the appendix, provides summary information on each sample firm
individually. Each company’s type of successor, both by firm origin, and – provided that they are outsiders - their industry origin.
We continue our sample description with the sample firms’ industry distribution. We will present the most striking results, a total overview of the industry distribution can be found in the appendix, table 7. It shows every industry’s two-digit SIC code, the accompanying industry description, and number of firms in each industry. The two-digit SIC code of each individual firm can be found in the appendix as well, table 6. As follows from table 7, in the appendix, most of the industries are represented by only one, or two firms. These industries are therefore not subject to discussion here. The most represented industries will be
mentioned hereafter. The two most represented industries in the sample are electronic and other electrical equipment and components, and business services. Both industries are represented by five firms, which equals both eleven percent of the total distribution. The first category, identified with SIC code 36, contains four internal successions, and only one external succession. The second category, business services, identified with SIC code 73, contains three external -, and two internal successions. SIC code 28, primary metal
industries, is represented by four firms. All four firms promote internal candidates to the CEO position. Industrial and commercial machinery and computer equipment, SIC code 35, is represented by three firms in our sample, all hiring external candidates.
4.2 Matching procedure
This paragraph will provide the result of the matching procedure. We were not able to match all sample firms on the two matching criteria mentioned before (two-digit SIC, and pre-event OROA in a 90-110 percent range of sample firm). Alternative matching criteria are used
20
Table 1 - summary information on company size, by successor type TYPE OF SUCCESSION ALL (N=45) INTERNAL (N=25) EXTERNAL (N=20) INDUSTRY INSIDER (N=4) INDUSTRY OUTSIDER (N=16) CONTROL GROUP (N=113) BEST MATCH (N=26) TOTAL ASSETS (IN MILLIONS) AVERAGE OF THREE YEARS PRIOR SUCCESSION 13995.84 19424.5 6097.66 9155.08 5404.92 12810.51 6244.94 MEDIAN THREE-YEAR PRIOR SUCCESSION 1467.16 1519.54 1201.64 2006.46 538.76 1240.23 1571.92 EMPLOYEES (IN THOUSANDS) AVERAGE NUMBER OF EMPLOYEES THREE-YEAR PRIOR SUCCESSION 22.92 28.32 15.68 26.91 14.67 14.93 16.11 MEDIAN NUMBER OF EMPLOYEES THREE-YEAR PRIOR SUCCESSION 7.431334 8.858333 7.4 9.789333 7.39 5.59 9.79 TOBIN’S Q AVERAGE OF THREE-YEAR PRIOR SUCCESSION 3.19 2.73 3.78 2.90 4.00 2.76 3.14 MEDIAN THREE-YEAR PRIOR SUCCESSION 1.74 1.51 2.72 2.75 2.72 1.97 1.60 OROA AVERAGE OF THREE-YEAR PRIOR SUCCESSION .09 .099 .08 .02 .10 0.8 .13 MEDIAN THREE-YEAR PRIOR SUCCESSION .10 .09 .10 .00 .12 0.10 .10
instead. In the end, we were able to match 58% of the sample firms on basis of the two primary matching criteria. Furthermore, 20% of the sample firms are matched to control firms on basis of pre-event performance and the same one-digit SIC code, another 20% on basis of only pre-event performance. The remaining 2% is matched to a control firm, of which
Notes: summary information on company size, classified by type of succession. A succession is classified as ‘internal’, if the CEO successor has served the company for more than a year before being appointed to the CEO position. Similarly, a succession is classified as ‘eternal’, if the CEO successor has been employed by the company for less than a year.
21
performance is closest to that of the sample firm. This result is graphically showed in the appendix, figure 1, as well. The average control group contains 2.7 firms per sample firm. The control group’s median performance is subtracted from the unadjusted performance, to end up with performance, correct for industry-, and pre-event performance. A list with every sample firm’s control group is presented in the appendix, table 8. The control group’s observables are presented in table 1.
4.3 Results
Differences in pre-event, and post-event firm performance are calculated, and tested. Both unadjusted firm performances, as well as industry-, and performance-adjusted firm
performances are studied. Results are shown in table 2. Four panels are used to separate unadjusted Tobin’s q, adjusted Tobin’s q, unadjusted OROA, and adjusted OROA.
4.3.1 Successors: internal versus external
We start with the analysis of Tobin’s q, unadjusted. This paper studies the difference in performance between internal, and external successors. The unadjusted difference between the three-year average performance prior, and following the turnover event is showed in Panel A’s first row. The coefficients show that, in general, firm performance declines following the CEO turnover event. This holds for both internal, as well as for external
successors. The difference-in-difference coefficient in column 4, 0.01333, is tested by means of the Mann-Whitney test. Since this value is not significant, we can not, at this point,
conclude anything on the difference in performance between the two successor types. For both the lack of significance of this results, as well as to control for pre-event performance, we split up the seven-year period into two different time-spans. First, we focus on the three-year period prior the succession event. It covers the three-years 1998 to 2000, and is shown in panel A’s second row. This time-span is used to check whether there are pre-event differences in firm performance. This appears to be the case. As follows from panel A’s second row, companies which appoint insiders, experience an increase in performance in the years prior successions. Opposite is true for firms that appoint outsiders, which suffer a 93.2 percent decline in performance between 1998 and 2000. In general (column 1) prior a CEO turnover event, firm performance increases by 6.2 percent, which is significant at a ten-percent significance level. To see whether there is a difference in firm performance surrounding the CEO succession events, we calculate the difference in firm performance between 2000 and 2004. The coefficient for this period for internal successions (column 2), is negative, while the external successor’s coefficient (column 3) is positive. This indicates that external successors are outperforming internal successors, surrounding succession events. However, these coefficients, as well as the difference-in-difference coefficients, are not
22
significant. Therefore, we fail to conclude that external successors are outperforming internal successors.
In panel B, industry- and performance adjusted results for Tobin’s q are presented. Since the unadjusted results in panel A does not provide us with significant results, results are adjusted for industry- and pre-event performance. This is done to see whether this provides us with evidence to conclude on our hypothesis. In panel B, as well as in panel D, unadjusted results are corrected for industry, and pre-event performance, by subtracting the median of the matching group’s performance. In this matching group, which serves as a control group, no CEO succession took place. This indicates that the difference in firm performance
surrounding the turnover event, must be caused by the CEO succession event. The general coefficients of change in firm performance, which are presented in panel B’s first row, move in the same direction as the unadjusted coefficients for Tobin’s q. Same holds for the other two time-spans. Again, firms promoting internal candidates have a positive difference coefficient in the years prior successions, while a negative difference coefficient arises for firms recruiting external candidates. Surrounding turnover, between 2000 and 2004, external candidates are outperforming their internal counterparts. These results are, however, not significant. We do find a significant increase, 44.1 percent, between the year prior, and the third year after succession.
Since existing succession literature uses OROA as performance measure, this measure of operating performance is also included in this paper. Panel C’s first row shows that
unadjusted firm performance, measured by the subtraction of the three-year average after, from the three-year average performance before succession, is declining. This corresponds to our findings in the first two panels. In general, companies who have experienced a turnover, suffer a 5.13 percent decline in OROA, which is significant at a five-percent level. Our analyses of the different time-spans (second and third row), do not correspond to our finding in panels A, and B. While for Tobin’s q firms that promote internal candidates experience an increase in firm performance between 1998 and 2000, they suffer a decline when OROA is used as performance measure. Opposite results, as compared to Tobin’s q are also found for the period 2000-2004. These results lack significance. The difference-in-differences coefficient for the unadjusted OROA is 0.1597, which indicates that internal candidates are outperforming external candidates by almost 16 percent.
Adjusted results for OROA, shown in panel D, contradict our unadjusted OROA findings in panel C. Firm performance, for both internal, and external recruiting firms, is increasing following turnover. Prior succession, all columns show negative coefficients of difference.
23
Table 2 - differences in firm performance surrounding successions
Type of succession External by industry origin All (1) Internal (2) External (3) Difference (4) Industry insider (5) Industry outsider (6) Difference (7) Panel A - Tobin’s q [3y average after] – [3y average before] -.0302065 (0.1923) -.0195387 (0.3967) -.0435411 (0.3703) 0.0133346 (0.7665) -.0697275 (0.1072) .0414253 (0.9588) -0.1111528 (0.4480) T= -3 to T = - 1 (1998 to 2000) .0622256 (0.0736)* .8572586 (0.2418) -.9315655 (0.1354) 1.7888241 (0.4789) .4419823 (0.0677)* -.6260834 (0.5014) 1.0740657 (0.4789) T = -1 to T = + 3 (2000 to 2004) -.0631263 (0.1469) -1.077687 (0.4758) 1.205074 (0.1354) -2.282761 (0.5993) -.8875479 (0.3812) 1.431138 (0.1627) -2.3186859 (0.6019) Panel B - Industry- and performance adjusted Tobin’s q [3y average after]– [3y average before] -.3845146 (0.9955) -.2841718 (0.8612) -.5099432 (0.8813) 0.2257714 (0.7840) -0.430012 (0.6891) -.3020503 (0.6051) -0.1279619 (0.4769) T = -3 to t = -1 (1998 to 2000) 0.0378953 (0.6557) 1.247976 (0.6571) -1.474706 (0.2180) 2.722682 (0.2534) 0.6804665 (0.7704) -1.126765 (0.6417) 0.6425712 (0.9056) T = -1 to T=+3 (2000 to 2004) 0.4407694 (0.0572)* -.7530807 (0.1919) 1.933082 (0.1913) -1.1800013 (0.4917) -.5361156 (0.1766) 2.211373 (0.1788) -2.7474886 (0.4917) Panel C - OROA [3y average after] – [3y average before] -.0513162 (0.0266)** -.0212414 (0.1036) -.0889097 (0.1259) 0.0676683 (0.1259) -.0230837 (0.0744)* -.1024877 (0.1627) 0.079404 (0.7399) T = -3 to t = -1 (1998 to 2000) -.030071 (0.1667) -.0543305 (0.0480)** .0002534 (0.7089) -0.0545839 (0.1098) -.0608359 (0.0179)** .0256902 (0.2553) -0.0865261 (0.0201)** T = -1 to T=+3 (2000 to 2004) -.0511729 (0.8170) .0198223 (0.7366) -.139917 (0.4330) 0.1597393 (0.3149) .0164145 (0.9397) -.1736753 (0.4691) 0.1900898 (0.4201) Panel D - Industry- and performance adjusted OROA [3y average after] – [3y average before] .215555 (0.4394) .150921 (0.5812) .2963474 (0.6274) -.1454264 (0.8016) .1190288 (0.0744)* -.1024877 (0.1627) 0.2215165 (0.5220) T = -3 to t = -1 (1998 to 2000) -.061255 (0.9955) -.0673898 (0.8612) -.0535864 (0.7369) -0.0138034 (0.6478) -.0702486 (0.6420) -.044954 (0.6417) -0.0252946 (0.5855) T = -1 to T= +3 (2000 to 2004) -.0590847 (0.4461) .0279176 (0.1094) -.1678376 (0.6274) 0.1957552 (0.2729) -.0590847 (0.4461) -.1863263 (0.7564) 0.1272416 (0.6523)
Notes: this table presents the difference in performance, prior and surrounding the CEO succession event. ‘All’ represents the coefficients in change for all firms which experiences a CEO succession, regardless of the type of successor. ‘Internal’ represents the firms that appointed a successor to the CEO position, who has served the company for a year or more. Similarly, ‘external’ represents all sample firms that appointed a candidate, who has served to company for less than a year. Panels B and D show industry, and pre-event performance adjusted differences in firm performance: median of the control group’s performance is subtracted from the unadjusted firm performance. A seven-year period centered around succession is studied, in which a split-up is made between the three year period prior succession (T= -3 to T= -1), and surrounding
succession (T= -1 to T= +3). Year 0, the year of succession, is omitted in all analyses. P-values in parentheses. *significant at
10% significance level ** significant at 5% level. The difference-in-differences (columns 4 and 7) are tested by the Mann-Whitney test. All other difference coefficients are tested by the Wilcoxon signed-rank test.
24
The difference-in-differences coefficient for the periods 1998-2000, and 2000-2004 (second and third row) correspond to these coefficients in panel C. Moreover, both unadjusted, and adjusted OROA, present that in general (column 1), turnover yields to a decline in firm performance. This holds for both the period prior succession, 1998-2000, as well as for the period surrounding succession, 2000-2004. Panel D does not show any significant results.
It follows from table 2 that performance changes, when split up for the different time-spans, are large. This especially holds for the results when Tobin’s q is used as performance measure. For example, a 124.8 percent increase in adjusted Tobin’s q between 1998 and 2000 (panel B, second row), and a 147 percent decrease for firms appointing insiders, and outsiders, respectively. These large performance changes are attributable to extreme values of Tobin’s q. We studied the origin of the extreme values. We found that these extreme results are driven by the companies’ shares prices, which are used to calculate the amount of shareholders’ equity. Research on the origin of these extreme share prices does fails to answer the question what the extreme share prices drives. It is unpredictable when, and for which kind of companies, these extreme share prices do arise. Therefore, we present an additional analysis, in which an adjusted sample is used. This sample excludes six firms, which have outliers in one or more observations for Tobin’s q. A list of the excluded firms is presented in the appendix, table 9. Results of the tests with the adjusted sample are shown in table 10, appendix. Aim of this analysis is the to see whether our results for Tobin’s q change when these extreme values are excluded. Does performance change in the same direction as when the original sample is used?
As follows from table 10, appendix, most coefficients are less extreme than the ones in table 2. Panel A’s second row shows that the coefficient for all successions, is negative. Without the exclusion of the six firms, this coefficient was positive. The coefficient for internal successions is 0.044. It follows from this, that Tobin’s q has increased by 4.4 percent,
between 1998 and 2000, while this was a 85.7 percent increase in the original sample. In this period, the external coefficient is negative, for both unadjusted, as well as adjusted Tobin’s q. This indicates that for firms which appoint outsiders to the CEO position, Tobin’s q has decreased in the years prior turnover. This corresponds to our findings presented in table 2. Again, we find that firm performance in general, suffers a significant decline in the years prior succession. Regarding the change in the sub-periods, 1998-2000, and 2000-2004, we find similar results as in the analysis of the original sample. In the years prior succession, firms that promote internal successors seem to outperform external successors. This corresponds to our results in table 2. Similarly, surrounding succession (20002-2004), external successors seem to outperform internal successors.
25
Regarding the change in performance in the periods 1998-2000, and 2000-2004, we do see a difference in the coefficients’ signs between all successions in table 2 and the same column in table 10, appendix. The average unadjusted Tobin’s q for all successions has increased by 2.1 percent in the sample that excludes firms with outliers, while this was a 6.3 percent decline when all firms were still in the sample. However, both coefficients are not significant.
One more analysis is conducted. Table 11, appendix, presents results for the conducted analysis, in which a second adjusted sample is used. This sample only contains the 58 percent of the companies which are perfectly matched to one or more control firms. No distinction is made here between internal, and external successors. Aim of this analysis is to study whether perfect matching improves the significance of our results. Results are
presented in the appendix, table 11. The unadjusted OROA has decreased by 3.7 percent, when pre-succession, and post-succession performance are compared. This value is significant at a five-percent level. None of the other results is significant. From this table follows that, generally speaking, significance has not increased by excluding imperfectly matched firms.
4.3.2 Industry outsider versus industry outsider
We continue our analysis to answer the second hypothesis. Internal successions are left aside here. The focus here is on external successions only. A distinction is made between industry insiders, and industry outsiders. Again, we start with analyzing the results for Tobin’s q. Results are shown in table 2. At this point, a cautious note must be made. In our sample, twenty firms recruited an external successor. From these firms, only four appointed an industry insiders, and industry outsiders are appointed by sixteen firms. This small sample size, in combination with the lack of significant results, restrain us from drawing significant conclusions.
Similar to our approach in the firm insiders versus firm outsider analysis in section 4.3.1, we start with the analysis of unadjusted Tobin’s q. The general change in firm performance, as shown in the first row, is negative for firms which appoint industry insiders, while it is positive for firms which appoint industry outsiders. This follows from both columns 5 and 6, as well as from the negative difference-in-differences coefficient in column 7. Firms that appoint
industry insiders, experience a 44.2 percent - significant - decline in the years prior transition. Surrounding succession, industry outsiders seem to outperform industry insiders. However, the coefficients are not significant, and the sample is small, so this conclusion should be interpreted with care. Results are again adjusted for industry-, and pre-performance, to check whether this improves the significance of our results. Regarding the direction of the change,
