Master Thesis
CEO and CFO turnover in large U.S. listed companies –
An empirical study on the determinants of top executive turnover events By Wiebe Jan de Boer
MSc Finance
Faculty of Economics and Business University of Groningen
Author: Wiebe Jan de Boer Student number: 1958526 Date: January 14, 2016
Supervisor: Prof. dr. C.L.M. Hermes
CEO and CFO turnover in large U.S. listed companies –
An empirical study on the determinants of top executive turnover events
Wiebe Jan de Boer
Abstract
This thesis studies CEO and CFO turnover in a sample of 200 large U.S. listed firms for the period 2006-2014. The thesis examines to what extent firm performance, firm size and the executive’s age determine the likelihood of a turnover event. The thesis finds that firm performance and firm size are negatively associated with the likelihood of forced CEO and CFO turnover, whereas the executive’s age is positively associated with the likelihood of unforced CEO and CFO turnover. I find that that CEO and CFO turnover events are strongly associated with one another.
Key words: CEO turnover, CFO turnover, firm performance, firm size, executive age.
JEL classification: G3, G34,
1. Introduction
Deciding whether or not to replace top executives are among the most important decisions a firm’s board of directors has to make. Top management turnover (TMT) events of large, publicly listed companies receive a great deal of attention from investors, governments and the public. TMT events carry a great deal of (symbolic) value, as the replacement decision potentially has long-term implications for a firm’s investments, operations and financial decisions (Huson et al, 2001).
Researchers have taken an interest in the determinants and consequences of TMT events since the 1980s (e.g. Coughlan and Schmidt, 1985; Warner et al, 1988;
Weisbach, 1988). The majority of the TMT literature has focussed on the Chief Executive Officer (CEO), as this executive is widely regarded as “the final unquestioned decision-maker and the person to whom all eyes would turn in a crisis”
(Tulimieri and Banai, 2010, pp. 241). Much less emphasis has been placed on top
executives other than the CEO. This is a shortcoming, since the top executives just
below the CEO (in terms of hierarchy) are likely to also have a major impact on the
firm. In particular, it is important to increase our knowledge of the determinants and
consequences of Chief Financial Officer (CFO) turnover events. Over the years, the
CFO function has grown in importance. To the more traditional CFO tasks (e.g.
financial reporting, financial risk management and financial planning), several responsibilities beyond the financial domain have been added.
Today, CFOs must take on dynamic leadership roles in four important areas of the business. First, they have exemplary strategic management capabilities. Second, they are able to provide line management with detailed, real‐time information that improves the quality of strategic decision‐
making and execution. Third, they transform the traditional investor‐relations function into a source of competitive advantage. And fourth, their leadership transcends the finance function and carries over into all areas of the company. (Favaro, 2001, pp. 4)
In the role as “custodian of the corporate treasury, the protector of the margin and the pillar of the corporate conscience” (Tuliemieri and Banai, 2010, pp. 240), the CFO is nowadays seen as the “key strategic partner” and “chief lieutenant” to the CEO (Collins et al., 2009). In a recent international survey
1among firms with revenues in excess of 500 million dollar, 549 CEOs were asked about their view on the role and importance of the CFO within their firm. According to the respondents, the CFO is the most important top executive (next to the CEO) and thought to become increasingly important in the future.
Tulimieri and Banai (2010) argue that although the CEO and CFO functions are complementary in nature, they expect increasing overlap in the functions in the nearby future. Conducting business and leading a company is becoming increasingly complex. The authors envision a “partnership of equals” paradigm where CEOs need a true partner to cope with the size and complexity that come attached with leading future firms.
The global corporation to which business is evolving will be lead by a duopoly: Two equal partners, with equal accountability, authority and access to the people, processes and technology of the business will now run the organization. They would be simply the obverse and reverse of the same coin – they would be interchangeable and complementary parts. They would speak with the same voice to the stakeholders and be equally charged with achieving the corporation’s strategic objectives. (Tulimieri and Banai, 2010, pp. 243)
Given the growing importance of the CFO, it is therefore of interest to better understand the causes and consequences of CFO turnover and to draw a comparison with CEO turnover events.
This paper therefore studies the determinants of CEO and CFO turnover events.
Using a sample of 200 large U.S. listed firms for the period 2006-2014, I estimate logistic regression models that predict the likelihood of CEO and CFO turnover events. Although the literature has identified multiple factors as possible determinants
1
“The view from the top. CEOs see a powerful future for the CFO. Are CFOs ready for the challenge?”
KPMG.com, last modified November 23, 2015, https://assets.kpmg.com/content/dam/kpmg/pdf/2015/11/view-
from-the-top.pdf
of TMT events, due to data constraints and time considerations this paper focuses on a subset of these factors: firm performance, firm size and the executive’s age. I examine (1) to which extent these 3 variables are associated with the likelihood that a CEO or CFO turnover event occurs and (2) how CEO and CFO turnover events compare in terms of the strength of these determinants.
The first part of this thesis studies CEO and CFO turnover in isolation from one another. The second part of this paper however examines CEO and CFO turnover events from an interdependence perspective. The literature (e.g. Fee and Hadlock, 2003) reports that turnover events of top executives within the same firm are often correlated (i.e. happen together or shortly after one another). I explore this phenomenon and examine how this affects the turnover mechanisms found in the first part of this paper. This thesis therefore tries to answer the following research question:
To which extent is firm performance, firm size and executive age associated with CEO and CFO turnover events and how interdependent are CEO and CFO turnover events?
The results in the first part of this paper indicate that CFOs experience higher annual turnover frequencies than CEOs. This result is largely driven by a higher frequency of unforced CFO turnover events. I find no statistically significant difference with respect to forced CEO and CFO turnover events. I find that the likelihood of a TMT event is negatively associated with firm performance. Forced TMT events are strongly associated with negative cumulative abnormal stock returns.
I find that this association is stronger for CEOs than CFOs, however the difference is smaller than previously reported. This finding might be a reflection of the changed nature of the CFO function, where the CFO is held responsible in a manner similar to the CEO. I also find that unforced TMT events are strongly associated with the executive’s age. With respect to firm size, the results indicate that firm size is positively associated with forced TMT events. Given the fact that I study a sample of large U.S. listed firms, it might be that the high press and investor coverage of this type of firm drives this result.
When examining the interdependent nature in TMT events, I find evidence of a
team nature in TMT events. The likelihood of a forced CFO event increases during
periods in which the CEO is forced to leave the office. The likelihood of an unforced
CFO event increases during periods in which unforced CEO turnover events take
place. Once the model controls for forced CEO turnover, the negative performance- turnover association for CFOs becomes much weaker, suggesting that much of this association is driven by whether or not the CEO is removed. The results have implications for future research as they signify the importance to take into account a team perspective in TMT events. Studying TMT events on a stand alone basis is likely to miss important drivers in TMT events. Given the changed role and increasing importance of the CFO, it is likely that this team nature in TMT events stays present and becomes increasingly influential in future TMT events.
The remainder of the paper is organized as follows. Section 2 presents the theoretical background and hypotheses development. Section 3 describes the data sampling and discusses the properties of the sample. Section 4 tests the different models. Section 5 concludes, discusses the limitations of the paper and provides future research directions.
2. Literature review
2.1. Monitoring and evaluating executives
In general, large companies are organised in such a way that the decision-making executives do not bear a major share of the wealth effects of these decisions. This separation of ownership and control has potential problems: the top executives will not always act in the best interest of the owners of the firm. This potential divergence in interests is the classical principal-agent problem as described by Jensen and Meckling (1976). To mitigate this problem, the principal (often a board of directors acting on behalf of the owners of the firm) needs to (1) design a contract that stimulates the agent to act in the principal’s interest, (2) monitor the agent’s actions and (3) discipline when necessary. Early principal-agent literature (e.g. Hölmstrom, 1979) suggests that the board needs to define performance metrics and performance targets.
The degree and the manner in which these performance targets are achieved determine the evaluation of executives. We would expect that a top executive has an increased likelihood to be turned over when he does not meet the performance targets or misbehaves in another area.
Over the years CEO turnover frequencies have increased. Scholars (e.g. Jensen et
al, 2004; Huson et al, 2001; Fee and Hadlock, 2004; Jenter and Kanaan, 2015) report
rising annual turnover frequencies. In the 1970s, annual CEO turnover frequencies
were around 10%. Between 2005 and 2011, the average CEO TMT frequency was
14.1% among the world’s 2500 largest public companies (PWC, 2011). A combination of increased oversight, job difficulty and job pressure is likely to have caused this trend.
CEOs are required to master a broader range of skills than in the past, when top executives might have climbed the ranks with just one discipline. Companies are bigger, more global and increasingly complicated, and there’s accelerating competition in countries such as China, India and Brazil. Executives must also adapt to quicker technological change [..]. “The pressure is getting tougher and tougher” [..]. (Bloomberg, 2013)
Due to the fact that the increased importance of CFOs has only in more recent years been recognized by researchers, aggregate CFO turnover frequencies prior to 2000 are not (at least not free of charge) available. Fee and Hadlock (2004) provide an indication however. They report for a sample of large U.S. firms during 1993-1998 that the turnover frequency was on average 15.4% among the 5 top executives just below the CEO. For more recent years, there are TMT frequencies available specifically focussed on CFOs. It appears that CFOs have experienced TMT trends similar to those of CEOs, albeit at somewhat higher turnover rates. CFO turnover frequencies among the top 500 U.S. companies were on average 12.2% between for the period 2010-2012, with a peak of 19% in 2013 (Wall Street Journal, 2015).
2.2. The performance-turnover association
Standard economic theory suggests that with respect to the performance metrics, the board should ignore components of firm performance that are caused by factors beyond the executive’s control. It is the relative performance of a firm by which an executive’s efforts and ability should be measured and evaluated. According to Hölmstrom (1982), relative performance evaluation stimulates the executive while insulating him from exogenous shocks over which he has no control. Boards need to define an appropriate benchmark (e.g. the industry or market index). When a firm underperforms its benchmark, one would expect that the likelihood of a TMT event increases. The literature finds that firm performance is indeed negatively associated with the likelihood of CEO and CFO turnover events (e.g. Warner et al, 1987; Parrino, 1997; Mian, 2001; Brookman and Thistle, 2009). Although the performance-turnover association varies considerably across the different studies, the general picture arises that bad firm performance increases the likelihood of a TMT event. I therefore test the following hypothesis:
H1a: Relative firm performance is negatively associated with TMT.
Over the years, a continuous stream of corporate governance codes, principles and requirements has been implemented. A good example of a major change in U.S.
corporate governance legislation is the Sarbanes-Oxley Act of 2002 (hereafter called SOX). SOX has been designed with two major purposes in mind:
• To increase the personal accountability of CEOs and CFOs. Since SOX, CEOs and CFOs are required to officially certify and approve the quality and integrity of formal firm documents such as annual reports and other financial statements. Top executives that violate the SOX requirements risk fines and even imprisonment.
• To increase the quality of the monitoring role of the boards of directors, in order to prevent corporate scandals such as Enron from happening again. Since SOX, the fraction of independent directors on a board has to be at least 50% in the U.S.
The literature often regards the performance-turnover sensitivity as a measure of board performance quality. If the stream of corporate governance regulations such as SOX has indeed improved board performance, one would expect to see that in general the performance-turnover sensitivity has increased over the years. Huson et al (2001) research large publicly listed firms between 1971 and 1994 and find that the performance-turnover association does not change significantly during these years. In more recent years however, the association seems to have intensified. Kaplan and Minton (2012) study CEO turnover for a sample of U.S. firms over the period 1992 to 2007. The authors find that the performance-turnover sensitivity has increased since 2000. This suggests that in more recent years, CEOs are held more accountable for firm performance.
The question arises how the performance-turnover association compares between CEOs and other top executives. Fee and Hadlock (2004) find in a sample of large U.S.
listed firms for the period 1993-1998 higher annual turnover frequencies for non-CEO top executives than for CEOs. The authors find however that the strength of the performance-turnover association is substantially higher for CEOs. Hazarika et al (2012) study the effect of aggressive earnings management on the likelihood of CEO and CFO turnover in a sample for the period 1992-2004. When controlling for firm performance, the authors find that the firm performance coefficients are about 2 times more negative for CEOs than for CFOs.
Although the literature suggests that the performance-turnover association is higher
for CEOs than CFOs, I wonder whether this difference is still present in more recent
years. One could argue that the changed nature of the CFO function as described in the introduction of this paper has increased the similarity in the personal accountability of both CEOs and CFOs. The performance-turnover sensitivity difference might have diminished accordingly. Also, the majority of the data of Fee and Hadlock (2004) and Hazarika et al (2012) covers a time period prior to important regulation changes.
Regulation such as SOX might also have increased the similarity between CEOs and CFOs in terms of their respective performance-turnover sensitivities. I therefore test the following hypothesis:
H1b: This relative performance-turnover association is stronger for CEOs than for CFOs.
If the difference is found to be insignificant or non-existent, this is as indication that CEOs and CFOs have become more similar with respect to their performance-turnover sensitivities than the literature reported previously.
2.3. Other factors associated with TMT events
The executive’s age plays a large role in TMT events. Scholars (e.g. Kim, 1996) find that the probability of a turnover event increases sharply when the CEO reaches the age of 64 or 65. Murphy (1999) finds that a CEO who is 64 years of age or older has a 30 percent higher probability to turn over than a CEO who is younger. The positive association appears to be stronger in unforced TMT events, which seems logical. The fraction of natural retirements will be higher in unforced TMT events compared to forced TMT events. Hayes et al (2006) study CEOs and the top five executives directly below the CEO. They find that the age category variable has similar effects on CFO turnover. I therefore test the following hypothesis:
H2: The executive’s age is positively associated with the likelihood of a TMT event.
Several scholars find that the frequency of CEO TMT events tends to be somewhat higher in large firms (Warner et al, 1988; Jensen and Murphy, 1990). This could be due to the fact that larger companies are subject to more scrutiny from analysts and the media (Agrawal and Cooper, 2016) than smaller firms. The literature does not provide evidence that this association would be any different for CFOs. I therefore test the following hypothesis:
H3: Firm size is positively associated TMT. The association is similar for CEOs and
CFOs.
2.4.1. Team nature in TMT events
Several papers report that TMT events within the same firm are correlated (i.e.
happen together or shortly after one another). Mian (2001) for example finds that abnormally high CEO turnover precedes CFO turnover. Fee and Hadlock (2004) report that non-CEO turnover is strongly associated with the occurrence of a forced CEO turnover event, prior or just after the non-CEO turnover event. Hilger et al.
(2013) find for a sample of German listed firms evidence that preceding unforced CEO turnover increase the likelihood that a CFO either makes an upward move within the company or leaves the firm for a similar position elsewhere. They also confirm the findings that preceding CEO turnover increases the likelihood that a CFO is forced to leave the firm.
Apparently there exists a team nature in TMT events. However, the research of Hilger et al. (2013) suggests that it depends on the context whether the team turnover phenomenon has a forced or unforced nature for both executives. How does the literature explain the association that unforced CEO turnover precedes unforced CFO turnover? When a CEO leaves a firm under unforced circumstances, it seems less likely that bad firm performance was a reason for the CEO to leave. Changes are that the CEO and his management team performed reasonably well. In order to stimulate the continuation of the current operations, it might then make sense for a firm to choose a new CEO out of the current top executive team. A CFO is than a likely candidate, with considerable experience in different important areas of the firm. One could also think of a CFO that is of a similar age of the CEO. A CEO that chooses to retire might trigger the CFO to retire as well, especially in case when an outsider becomes the new CEO. Mian (2001) argues that this phenomenon can also be consistent with the notion that when the CFO is passed over for promotion to the CEO function, the probability that the CFO becomes CEO in the nearby future has significantly decreased. The CFO might consequently choose to leave the firm to try his luck elsewhere.
How does the literature explain the association that forced CEO turnover precedes
or surrounds forced CFO turnover? It is likely that bad firm performance or some sort
of a scandal induces forced TMT events. It then might make sense for a board of
directors to punish multiple top executives together. Together they top executives
failed to deliver good performance. Shen and Canella (2002) argue that failing to
deliver good performance is an indication of a failure to formulate and implement an
effective strategy to deliver firm value. If strategic change is needed, it might then make sense to remove multiple top executives from their function. This might also be consistent with the findings of Fee and Hadlock (2004). The authors examine this team nature phenomenon in different firm performance quartiles. They find that moving from the best to the worst quartile hardly alters the increase in the non-CEO turnover probability. Once their models control for CEO turnover, the performance-turnover association for non-CEOs becomes weak in a statistical and economical sense. This suggests that bad performance from a stand alone basis is not the only reason to remove a CFO, but also an indication of a need for strategic change. Either the board of directors can induce the CFO removal or the CEO when he feels he needs a new CFO in order to deliver expected firm performance.
I therefore formulate the following 2 hypotheses.
H4: Unforced CEO turnover is positively associated with unforced CFO turnover.
H5: Forced CEO turnover is positively associated with forced CFO turnover.
3. Data sampling and descriptive statistics 3.1. Sample selection
This paper uses the Standard and Poor’s ExecuComp database. This database tracks since 1992 executive compensation for companies listed on the S&P1500 index. Per fiscal firm year, ExecuComp provides detailed information on the identity, compensation and function of the 5 highest paid executives within the respective firm.
I focus on the firms listed on the S&P500 index during the period 2006-2014. The S&P500 index consists of 500 large firms (market capitalizations in excess of 5.3 billion USD) with common stock listed on the NYSE or NASDAQ. I focus on this subset of firms due to the expected high press coverage of these large firms. This high press coverage will be helpful when examining the nature of the turnover events.
For the resulting set of companies, CEO and CFO TMT events are recognized for
each year in which the person who held the respective function changed compared to
the previous fiscal year observation. I then use a variety of online sources (e.g. Lexis-
Nexis, Bloomberg, 10-K reports, proxy statements and company press releases) to
determine the exact turnover announcement date and the reason for the TMT event. To
control for the different reasons under which an executive can leave a company, the
TMT literature often determines whether a TMT event had a forced nature or not. It
makes sense to make this distinction: for example, it is likely that firm performance
plays a much smaller role in unforced TMT events than it does in forced TMT events.
Scholars (e.g. Parrino, 1997; Fee and Hadlock, 2004) confirm this. I follow Parrino (1997), as applied by Jenter and Kanaan (2015) to determine whether a TMT event was forced or unforced:
1. All successions for which the information source explicitly indicates that the executive is fired, forced out, or retires or resigns due to policy differences or pressure are classified as forced. All other departures for top executives above and including age 60 are classified as unforced.
2. TMT events of executives below age 60 are reviewed further and classified as forced when the press does not report the reason as death, poor health, or the acceptance of another position (elsewhere or within the firm) or when the press reports that the executive is retiring but the announcement is made within 6 months of the actual succession date.
3. The circumstances surrounding TMT events defined as forced in the previous step are further investigated by searching the business and trade press for relevant articles. A reclassification can take plan when the article(s) convincingly explain the departure as due to personal or business reasons that are unrelated to the firm’s activities.
4. I exclude turnovers related to M&As, spin-offs, or major corporate ownership changes. All TMT events in which the executive serves less than 12 months or where the executive finishes his interim job are excluded as well.
Next, I manually collected the missing data with respect to the dates when the
executive became CEO or CFO within the firm. All necessary accounting information
was obtained from the Compustat Industrial Annual files. To study the performance-
turnover association, I obtain stock return information from the monthly stock files
from the Centre for Research in Security Prices (CRSP). Following Fee and Hadlock
(2004) I calculate the firm’s cumulative buy-and-hold stock return over the 12-month
period prior to the start of the firm fiscal year observation. For example, if the fiscal
year from the observation in ExecuComp runs from January through December 2006,
the stock return is calculated over the period January through December 2005. See
appendix A for more details on the performance measure. To measure the relative
performance of a firm, I subtract from the individual firm performance the mean
benchmark industry stock return over the same period. Industries are defined using the
Fama and French (1997) classification of firms into 12 industries (see appendix B).
3.2. Descriptive statistics on the reasons for TMT events
The final sample consists of 200 firms
2and covers a total of 3161 person-year observations during the period 2006-2014. In total, the sample contains 317 TMT events. Table 1 provides descriptive statistics on the distribution of the TMT events across the two executive groups and summarizes the reasons for the TMT events. The table indicates that the annual turnover frequency is significantly higher for CFOs than CEOs. In my sample, it appears that it is the relative high proportion of unforced CFO turnover events that drives this difference.
Table 1
Descriptive statistics on TMT events
The table presents descriptive statistics on CEO and CFO turnover events in a sample of 200 S&P500 firms from 2006 to 2014. Each turnover event is assigned a single reason for why the event occurred, based on the information interpretation of online sources such as Lexis-Nexis or Bloomberg. The forced and unforced classification follows Parrino (1997). The figures in the “reasons for turnover section” are calculated as a percentage of total TMT events of the corresponding executive group.
All CEOs CFOs p-value for
difference Sample characteristics
Total observations 3161 1597 1564
Turnover events 317 129 188
Annual turnover frequencie 10.03% 8.08% 12.02% 0.0001
Forced turnover 17.98% 21.71% 15.43% 0.8312
Unforced turnover 82.02% 78.29% 84.57% 0.0001
Reasons for turnover
Fraud, misconduct or outright bad performance 3.75% 5.99% 2.21% 0.0209
Resignation 0.57% 0.95% 0.32% 0.3273
Death 0.26% 0.63% 0.00% 0.1616
Founder retiring 0.96% 1.89% 0.32% 0.0623
Personal reasons or pursuing other interests 3.45% 1.58% 4.73% 0.0220
Health 0.63% 0.63% 0.63% 0.9833
Retirement 12.84% 15.46% 11.04% 0.1468
Becoming chairman or president within the same firm 5.02% 8.20% 2.84% 0.0047 Other role within top management team 8.42% 0.00% 14.20% 0.0000 Moves to executive committee/board or senior advisor 2.38% 1.26% 3.15% 0.0997 Becoming CEO within the same firm 1.25% 0.32% 1.89% 0.0549 Becoming CEO or president elsewhere 2.44% 0.95% 3.47% 0.0291
Similar position elsewhere 6.43% 0.63% 10.41% 0.0000
No reason given 3.33% 2.21% 4.10% 0.1638
2