University of Amsterdam, Amsterdam Business School
MSc Finance Corporate Finance
Master Thesis
The effect of succession planning on the change of
firm performance
Fehér, Viktória
July, 2018
Statement of Originality
This document is written by Student [Viktória Fehér] who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are 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.
1. Introduction
What happens to a company when the most important person, the leader of the organisation, the CEO leaves? How does the performance react? Does it react differently when everything is already clearly planned and this plan is disclosed? If yes, how?
It is a widely known fact that the CEO is a key character in any companies’ life. If she has to or wants to leave, it can cause a stir if there is not a clear plan, who will success the position or how this succession will happen.
An anecdotal example for this is the case of Bank of America that shed light on the importance of succession planning.
Bank of America’s CEO, Ken Lewis announced on 30th September 2009 that he wanted to retire.
The board was unprepared and for several months it was unable to find a proper successor for the position. Multiple successor candidates rebuffed the bank’s approach that worsened the situation even further. Just before a few weeks of Lewis’ departure, on 16th December was the Board able
to name the successor. (The Financial Times (2009) and Bloomberg (2009)).
However, there are also positive examples for succession planning, too, with a smooth turnover, such as the case of Goldman Sachs in 2006. Its current CEO at that time, Henry Paulson was promoted to the U.S. Treasury Secretary, thus he had to leave the CEO position. His successor was Lloyd Blankfein, who was already named as an heir of Paulson in 2003. (New York Times, 2007)
However, in 2010 Heidrick & Struggles conducted an extensive survey about the CEO succession practices among public and large private firms in the US and Canada. They found that more than half of the companies are not prepared for an unexpected event resulting in the leaving of the current CEO. (Heidrick and Strugless, 2010) It means they could not name a successor instantly after that event.
Reflecting to the succession failures of several companies, such as Bank of America, the US Securities and Exchange Commission (SEC) issued a guideline in 2009 to stimulate firms’ disclosure of their CEO succession planning. (SEC, 2009)
The report of Strategy& shows the financial importance of succession planning. It proves that big companies, which were forced in a CEO turnover, lost $1.8 billion per company potential shareholder value because of the lack of succession planning. (Financial Times, 2015)
This thesis examines whether firms benefit from the succession planning, more specifically, from the disclosure of the succession planning. It studies how the disclosure of succession planning of a CEO affects the change in firm performance after the turnover. The event window for the turnovers is between 2010 and 2015, inclusive. Performance was chosen as the main interest of this thesis, because performance is important not only for the CEO, but also for the board and the investors.
Furthermore, succession planning can influence performance through various channels. It minimizes the organizational disruption with decreasing the likelihood of forced CEO turnovers and the likelihood of the resignations of the other executive officers, and with also decreasing the possibility of an interim CEO, who holds that position for less than a year. (Cvijanovic et al, 2017) Thus it saves the high transaction costs. Moreover, as the firm specific human capital is positively related to succession planning (Cvijanovic et al, 2017), thus further smoothening the turnover process and hence, resulting in higher post-turnover performance.
The author uses the operating return on assets (OROA) as the main measure for performance. This measure is commonly used in the CEO turnover literature, such as Denis and Denis (1995), Huson et al. (2004), Pérez-González (2006), Bennedsen et al. (2007). Following Zhao (2015), this thesis also uses adjusted OROA, adjusted for industry. This is needed to control for the trends in the given industry.
The author conducts hand collection of data to capture succession planning. The regulatory filings (8K and DEF-14A) of the companies, which had a turnover during the sample period are investigated whether they mention anything about a succession plan for the CEO position.
Succession planning does not have a long research history, especially not in the financial literature, however management turnovers are well researched. Thus this thesis contributes both to the financial and management literature of succession planning in several ways. First, it proposes a new way to capture CEO succession planning, that is not used in the literature yet. This method is that the determination of a succession plan is based on the companies’ regulatory filings. If they
mention CEO succession planning at least one year prior to the turnover, this thesis considers them as firms with a succession plan, if they do not mention this, this thesis considers them as firms without a succession plan. This method is new to the literature as only one working paper by Cvijanovic et al (2017) uses it, but without the one-year-criteria. Secondly, according to the best knowledge of the author, this is the first paper that measures the effect of succession planning captured by this new method on the change of the firm performance around the turnover. As Cvijanovic et al (2017) focuses on volatility and the efficiency of management turnover. Furthermore, other previous studies, such as Shen and Canella (2003) or Zhang (2015), that measure performance change effect capture succession planning as relay succession, and Rivolta (2018) and Behn et al (2006) uses delay in naming a new CEO as a succession planning measure.
The rest of the thesis is as follows. Section 2 describes the literature, section 3 outlines the methodology, section 4 presents the data and descriptive statistics, section 5 shows the results of this thesis, section 6 demonstrates the robustness tests, and section 7 concludes.
2. Literature Review
There are several articles about the relationship of the management turnover and firm performance.
The classic paper, which is mentioned in every other article published later about management turnover and/or performance is Denis and Denis (1995).
Denis and Denis (1995) show that before a forced management turnover the operating performance of a firm significantly decreases and after the turnover it increases significantly. Furthermore, following the turnover, these companies shrink their operations significantly regarding capital expenditures, employment, and total assets. They also find that normal retirements do not face with these swings in the performance, only a small rise is noticeable after the turnover. They conclude these results by analysing the operating rate of return on total assets (OROA) over seven years, in which centre the year of the management change stands. They also prove that although forced turnovers increase shareholder value by increasing the performance, however this raise cannot be the results of effective board monitoring. This have three reasons: forced resignations are not common events, they are mostly not initiated by the board, and they happen mostly after bad performance.
Huson et al. (2004) extends the research of Denis and Denis (1995) because they argue that the concluded performance improvements could be the result of the mean reversion of the accounting performance time series and not the turnover. Thus to tackle this, Huson et al. (2004) uses a longer time period and a different method, which is a performance-based control group matching. They find that a firm performs worse before a CEO turnover and the market reacts positively for the announcement of the turnover. Furthermore, they also find that the average change in the unadjusted OROA for their sample is negative after the turnover. However, there are several factors, which can positively affect the operational performance in case of a turnover, such as board composition, institutional shareholdings, takeover pressure, and outside CEO successor.
Peréz-González (2006) continues the investigation of CEO succession on the performance of publicly traded US firms. He analyzes the effect of family CEOs, who are blood- or marriage – related to the departing CEO. He concludes that the companies, which appointed family CEOs significantly underperform compared to the ones, which appointed an outsider CEO. Just like Denis and Denis (1995) and Huson et al (2004), he also uses OROA as the main indicator for performance. Furthermore he evaluates it with market-to-book ratios and net income to assets, too.
Bennedsen et al (2007) goes on with the performance effect of managerial turnover. They also shed light on the problem of family succession among Danish firms. They use difference-in-differences methodology and instrumental variables to prove that the firm performs worse after a family related insider succession than after an outsider succession, which is not family-related.
As it can be seen the management turnover and its effect on performance is very broadly studied, like the performance effect on management turnover. This one is investigated by for example Eisfeldt and Kuhnen (2013), and Jenter and Kanaan (2015). They prove that CEOs are fired after poor firm performance caused by factors, which were independent from them. However, Taylor (2010) says that CEOs are not often fired, because of four reasons. First, it costs too much for the shareholders to fire the CEO. Second, if a bad and a good CEO are similar, it makes no sense to change her. Third, it takes time for the board to decide if the CEO is capable enough, thus an untalented CEO can be unnoticed for several year. Fourth is CEO entrenchment, when a board does not fire a CEO, but the shareholders wants her to be fired. The reason for this could be personal ties, or the ignorance of the board towards the shareholder value. Taylor (2010) concludes that
replacing the CEO costs at least $200 million for the shareholders, but these costs primarily originates in entrenchment than in the actual costs to shareholders.
The literature about succession planning itself is much narrower. One of the reason for this could be the difficulty to capture succession planning.
Naveen (2006) studies the effect of human capital considerations on the process of CEO succession. Her paper’s main focus was relay succession, which is a type of succession planning referring to a selection of an heir. One of her result is that plenty of firms uses relay succession. However, she identifies relay succession with the existence of a COO and or President position hold by a non-CEO. This can be flawed based on Zhao (2015), who states that based on Execucomp data during 1992-2013, only 30% percent of the people who were a COO or President became a CEO at some point. Anyways, Naveen (2006) states that firms, which are larger, more diversified and operate in heterogeneous industries are more likely to have relay succession. Furthermore, with succession planning there is a higher probability of inside and voluntary succession and lower probability of outside and forced succession. Moreover, lower horizon problems in case of a relay succession.
Shen and Canella (2003) investigates the effect of relay succession on performance measured by cumulative abnormal returns and return on assets in the year of the succession. They conclude that there is a positive wealth and performance effect for promoting or appointing an heir for a CEO, but promoting an insider, who is not an heir has a negative effect on the performance. Furthermore, they also find an outside succession is paired with significantly higher cumulative average abnormal returns than insider relay succession. However, there are some issues with this article too. This article has the same main drawback as Naveen (2006) regarding the identification of relay succession.
Behn et al (2006) tackle the above mentioned issue in a different way. They measure a delay in naming of the new CEO after the death of the previous one as an indicator for an existing succession plan and analyze its effect on performance. They conclude that the delay in naming a successor and outsider succession have affect the firm’s financial performance and equity value negatively. Thus the longer the waiting time is, the more the future firm net income performance declines. Furthermore, firms that are able to name the next CEO immediately tend to outperform those
companies, which defer it. However, they also lists several factors in their own research that limit the generalization of their results. These issues are related to the research design and methodology.
Rivolta (2018) has other results as Behn et al (2006). She also investigates the connection between the delay in replacing the CEO and the firm performance. She finds that this delay has a positive effect on the change in stock and firm performance in several cases, especially for firms without succession planning. The delay is valuable, when the given firm is highly R&D intense, has a better pre-turnover stock performance, is a family firm, and when the successor is an outsider. Yet, in some other cases, such as greater firm risk or greener pasture, delay is disadvantageous for performance. She explains this that the current uncertainty around the turnover gets amplified with the delay, thus resulting in worse performance. Like Behn et al (2006) also Rivolta (2018) lists some issues, which can bias her results.
Zhao (2015) proposes a new way to capture relay succession and thus the likelihood of being promoted to a CEO. This is a logit regression dependent on a number of executive attributes. He analyzes the effect of relay length of the incoming CEO on the firm’s performance after the turnover and volatility. Relay length is meant as the number of years the future CEO spends as an heir. He proves it increases the performance the highest in the year of the turnover and its effect disappears over time. Furthermore, he states that relay decreases volatility the highest also in the year of the turnover and its effect disappears over time, too.
Zajac (1990) used a survey among firms to capture succession planning. He finds that firms, which have someone to nominate as a CEO in case of a turnover, are higher performing ones. However, his sample is very limited consisting answers only from 92 CEOs. Furthermore, since the investigated time period (1979-1986) several changes have happened.
Mehrotra et al (2013) analyzes a very special type of succession among Japanese firms and its effect on the performance. This special type is when the patriarch of a family firm adopts an adult giving this adoptee his own family name and names him as a successor of the firm. Mehrotra et al (2013) finds that firms run by these adoptees outperform the firms run by blood-relative successors and also the ones, run by professional CEOs (sarariman). Although their results are meaningful that a designated heir can have a positive effect on performance, however it cannot be generalized because of the special circumstances of Japanese firms and their corporate governance.
But all of the above mentioned article take into consideration only one type of succession planning, which is relay succession. However, the following is stated in the New York Stock Exchange Corporate Governance Standards among Corporate Governance Guidelines
“Succession planning should include policies and principles for CEO selection and performance review, as well as policies regarding succession in the event of an emergency or the retirement of the CEO”
Hence, suggesting that succession planning is a much broader process, than naming an internal candidate to follow the incumbent CEO in its position. Furthermore, Pearl Meyer, one of the biggest compensation consultancy firms defines the scope of ideal CEO succession planning process as identifying a few candidates within the company (2 or 3), and also keeping an eye on the possibilities outside the company regarding quality and/or availability of possible successor for multiple time horizons (emergency, near-term, mid-term, long-term).
Thus, Cvijanovic et al (2017) is very innovative regarding the measurement of CEO succession planning, as it states that it is the first paper, which uses such a measure, which refers to a firm’s formal disclosure of succession planning in its 8K filing or DEF 14A proxy statements. Based on their prior researches, they formulated three hypothesis in their paper. First, firms with succession plans are more likely to hire CEOs permanently and less likely to experience changes in the management after the turnover. Second, firms with succession plans experience lower stock return volatility around the turnover and it returns to its normal pace faster. Third, firms with succession plans are less likely to fire their successor CEOs in case of an industry shock. Furthermore, their findings show strongly that succession planning smoothes out management turnover. It decreases the possibility for non-CEO executives’ resignations. Because of the lower stock return volatility, there is a lower risk around the turnover and quicker reduction in the uncertainty with the new CEO’s tenure. Furthermore, succession planning is also negatively associated with CEO entrenchment, but positively connected to firm-specific CEO human capital. Moreover, firms with succession planning are less prone to choose an interim successor.
Based on the above mentioned literature, it is visible that there is a gap in the literature that analyzes the effect of the succession planning on post-turnover firm performance, if the succession planning
is captured the same way as Cvijnanovic et al (2017). This thesis attempts to fill it in. Based on the following arguments, it seems clear that succession planning can affect performance several ways.
Firstly, succession planning is negatively connected to CEO entrenchment (Cvijanovic et al, 2017) and Taylor (2010) proves that it costs around $200 million a CEO turnover for the investors. This sum is mostly originates in CEO entrenchment. Furthermore, Cvijanovic et al (2017) also finds less likelihood for an interim CEO. Thus, an existing succession planning should increase post-turnover performance.
Secondly, by decreasing the chance of the resignations of other executes, an existing succession planning reduces the organizational disruption around the CEO turnover. (Cvijanovic et al, 2017). This will minimize the total number of turnovers around the CEO turnover that, without succession planning, could be quite high if the current management is not satisfied with their new CEO. Hence, the huge transition costs can be saved, which would have been spent otherwise on the replacement of these other executives. This possibly can increase the post-turnover performance compared to the firms without succession planning.
Thirdly, as firm-specific human capital is positively related to succession planning (Cvijanovic et al, 2017), a new CEO can be better skilled and have a better general ability to run that given firm. This fact contributes to smoother turnover and hence, a higher post-turnover performance.
Based on these arguments, the following hypothesis is developed and investigated in this thesis.
Hypothesis 1 (H1). All else equal, a firm with a pre-defined and disclosed succession planning has a higher change in its performance around the CEO turnover.
3. Methodology
The aim of this thesis and its methodology is to test the effect of succession planning on the change of the firm’s performance around the CEO turnover.
As one can see in the literature part, during the years several attempts were done to capture a good proxy for succession planning in the academic literature. Mostly these proxies focused on relay succession, which is however only one type of succession planning. Cvijanovic et al (2017) are the
only ones, who takes into consideration that CEO succession planning can include not only an internal heir, but also searching for external candidates. This thesis follows the same way to capture a proxy for succession planning.
This thesis uses Compustat public firms excluding utilities as the main sample that are located in the United States and had a CEO turnover between 1st January 2010 and 31st December 2015. The sample is divided into two groups: firms with and without succession plans.
This thesis defines a firm with succession plan as the following: the given firm has to formally disclose an existing succession plan in its DEF14A or 8K filing at least one year but not more than four years before the turnover.
3.1 Determinants of Succession Planning
Given the limited prior literature about succession planning, especially about that type of proxy for succession planning used throughout this thesis, the results part start with inspecting the determinants of succession planning.
The methodology follows Cvijanovic et al (2017). A simple OLS regression is conducted to determine, which factors have an effect on the likelihood of succession planning. Cvijanovic et al (2017) writes that “adoption of a succession plan as a function of the interplay between the power of the CEO, seen through the lens of both CEO entrenchment and CEO ability, and the monitoring of the board” (Cvijanovic et al, 2017:13).
Cvijanovic et al (2017) refers back to Hermalin and Weisbach (1998) to argue that the more the CEO is entrenched, the lower the likelihood that the firm adopts succession planning.
In this thesis CEO entrenchment is measured in several different ways, provided by the data is available and sufficient for create a proxy for it. Thus, based on Berger et al. (1997), one of the proxy used is CEO tenure, following the idea that the longer the CEO has in position the more time she has to gain control over internal monitoring systems. Furthermore, fixed compensation used another measure as suggesting the CEO ability to extract resources from the company. As a third proxy, CEO pay slice is used based on Bebchuk et al. (2011). They describe that the higher the gap between the CEO’s and other managers’ compensation, the more entrenched the CEO is.
Based on these, it is expectable that all the three proxies are negatively associated with the likelihood of having a succession plan on hand.
Thus, three equations are tested empirically:
(1) Succession planningi=αi+β1 Ln(Assets)i+β2 Firm agei+β3ROAi+β4 Market-to-book
ratioi+β5 Independent boardi+β6 Log(CEO tenure)i+β7 (Log(CEO tenure)x
Independent board)i+εi
(2) Succession planningi=αi+β1 Ln(Assets)i+β2 Firm agei+β3ROAi+β4 Market-to-book
ratioi+β5 Independent boardi+β6 Log(Fixed compensation)i+β7 (Log(Fixed
compensation)x Independent board)i+εi
(3) Succession planningi=αi+β1 Ln(Assets)i+β2 Firm agei+β3ROAi+β4 Market-to-book
ratioi+β5 Independent boardi+β6 CEO pay slicei+β7 (CEO pay slice x Independent
board)i+εi
Most independent variables are lagged by one year.
Although Cvijanovic et al (2017) uses several other measures for CEO entrenchment and general ability, it is not possible to conduct those regression because of the lack of data and/or the difficulty to gain access to those databases.
3.2 Performance Effect of Succession Planning
Based on Bennedsen et al (2007) and Zhao (2014) an easy way to measure the effect of succession planning is to estimate the difference in the company’s performance around the CEO turnover and measure the method in which the company results alter due to the turnover. But it comes with disadvantages. One of these disadvantages is namely that it does not take into account the aggregate changes in performance that occur for example because of industry or aggregate trends.
Following Barber and Lyon (1997), Huson et al (2004), and Pérez-González (2006) the measurement for firm performance, OROA, is adjusted using industry benchmarks to tackle this issue, and then using diffence-in-differences analysis relative to a control group. For this thesis, the
changes in the unadjusted and industry-adjusted OROA are compared between firms that have a predefined succession plan and firms that do not have. The equation is the following
(4) yi = αi+β1 Succession planningi+β2 Posti+ β3 (Post x Succession planning)i+εi,
where Succession planningi is an indicator variable equal to one, if the firm has a succession plan
and zero otherwise. Under the null hypothesis, that succession planning has no effect on performance, β3 would not be different from zero.
However, this analysis does not include any control variables. To separate the succession effects from other confounding, the following regression analysis is applied:
(5) yi = αi+β1 Succession planningi+γ control variablesi+εi,
where yi is the difference in firm performance around the CEO turnover, while Succession
planningi is an indicator variable equal to one, if the firm has a succession plan and zero otherwise.
Following Bennedsen et al (2007) and Zhao (2015), the following control variables are used: firm size, measured as the natural logarithm of the total assets, firm age, industry adjusted OROA one year before the turnover, CEO pay slice, a dummy variable indicating for an outsider CEO, and board independence.
For the firm size, firm age, board independence, outsider CEO a positive relation is expected towards y, while for the other control variables, a negative association is expected towards y.
4. Data and descriptive statistics
4.1 CEO turnover and CEO succession
As a starting point CEO turnovers are identified based on the ExecuComp database between 1st January 2010 and 31st December 2015 in the United States. 2010 is the starting point to avoid biasing the sample with the financial crisis. The method of identification is the following: first to exclude those people from the database, who never become a CEO as the database contains other executives too, such as CFOs. Secondly drop those, which left the CEO position either earlier or later than the observation window. After that, with the help of the Compustat database, SIC codes
are ordered to the current observations, thus it becomes possible to exclude the utility firms (SIC 4900-4999). Lastly, after dropping also the duplicate observations, the sample yields 1,097 CEO turnovers among US public firms.
Determining whether a firm has a CEO succession planning or not, is more subjective. An initial base for this is CapitalIQ database, filtered for public firms located in the US that have a CEO succession planning. However, this option filters for firms, which have succession planning at the present and not the historical data or when this succession planning was introduced. One can assume that a company, which does not have a CEO succession planning currently, it did not have in the research window of this thesis either. After manually comparing this filtered dataset with the previously described CEO turnover dataset, one can see that around 40 % of the turnovers can be eliminated from the next, hand-collection part – but be kept in the total sample – as those firms do not have a succession plan based on the previous assumption. Thus, 659 firms’ DEF14A proxy statements and 8K filings is necessary to check, if they describe a certain way of CEO succession planning. However, one firm can introduce a succession planning process just months before the turnover, if for example the board wants to lay off the CEO already. This “succession plannings” have more an ad-hoc feature than a properly planned feature and they could cause endogeneity, thus as a requirement for this thesis, it is necessary that a firm should have a described succession plan at least one year before the turnover, but not more than four years. As an example, let us take a look at Fluor Corporation. Fluor had a CEO turnover in 1st February 2011. From the data
collection point of view, it means that the 8K filings and DEF14A proxy statements, which were issued between 1st February 2010 and 1st February 2007 should disclose a statement about CEO succession planning to confirm that Fluor had a succession plan at the time of its turnover.
This statement about the CEO succession planning can be either just a short hint that the firm has some sort of plan for an unexpected (and also expected) turnover or a well described scenario. An example for the short suggestion could be the following sentence in Carpenter Technology Corporation’s DEF14A proxy statement:
“Review succession plans for the CEO and the Company’s executive officers.”
This implicitly suggests that Carpenter Technology has already an introduced succession plan at the time of the issuance of the DEF14A in question. Mostly, this short descriptions are typically
listed as one of the responsibilities of the Board or the Compensation (and/or Nomination) Committee.
On the other hand, some other firms describe their succession plans in a more detailed way or they even have a Succession (Planning) Committee. A good example for a thorough disclosed succession plan can be Carnival Corporation with the following:
“Our boards believe that planning for the succession of our Chief Executive Officer is an important function. Our decentralized structure enhances our succession planning process. At the
corporate level, a highly-skilled management team oversees a collection of separately managed cruise brands. Each of our brands is led by locally-based executives who are driven to grow and
optimize their brands, which fosters an ownership-oriented attitude that is not common in an organization of our size. At both the corporate and brand levels, we continually strive to foster the professional development of senior management. As a result, Carnival Corporation plc has developed a very experienced and strong group of leaders, with their performance subject to
ongoing monitoring and evaluation, as potential successors to our Chief Executive Officer”
One can find further examples for succession planning definitions in the Appendix.
This way, a new variable is created in the dataset for succession planning. This is a binary variable taking one, if the given firm discloses any information about the existence of its CEO succession planning as how it is defined above (either a short or a thorough descriptions, guidelines etc.), and zero otherwise. The judgement if it is a proper description or suggestion about an existing plan is highly dependent on the opinion of the author of this thesis.
After the hand-collection part, the CEO turnover dataset with the succession plan variable is merged to the Compustat dataset. The firms with missing sales or total assets variable are dropped from the final dataset, as it is not possible to calculate the main variables of interest if these key variables are missing.
4.2 Performance
Based on the CEO turnover literature from Denis and Denis (1995), Huson et al (2004) Pérez-González (2006), and Bennedsen et al (2007), performance is measured with the operating return on assets (OROA). It is the ratio of the operating income (EBIT) and the book value of total assets. The advantage of this measure is that it is unaffected by the firms’ capital structure decisions, unlike the equity based return on equity (ROE) or the net-income based return on assets (ROA).
However, this unadjusted OROA does not capture the industry-wide trends. Thus, to control for these trends, industry adjustment is performed following Barber and Lyon (1997), Huson et al (2004), and Pérez-González (2006). The industry adjusted OROA is calculated by subtracting the mean OROA of the relevant industry and year. For industry, two-digit SIC codes are used based on Huson et al (2004) Pérez-González (2006).
4.3 Descriptive statistics
There were 1,097 turnovers for the research window.
Table 1 shows the frequency of the turnovers in different years for the total sample, for the firms with succession planning and for the firms without succession planning.
Table 1 Summary Statistics of CEO Turnover and Succession planning
This table reports how many CEO turnover happened in a certain year, what percentage of the total turnovers it is. This is categorised into three groups: firms without succession planning, with succession planning, and the total sample. Number of observations are in parentheses
(1) (2) (3)
Without succession planning
With succession planning
For the total sample
2010 18.01% 6.37% 14.68% (141) (20) (161) 2011 15.58% 16.24% 15.77% (122) (51) (173) 2012 17.88% 17.52% 17.78% (140) (55) (195) 2013 18.65% 17.20% 18.23% (146) (54) (200) 2014 14.94% 20.06% 16.41% (117) (63) (180) 2015 14.94% 22.61% 17.14% (117) (71) (188) Total 100% 100% 100% (783) (314) (1,097) Observations 783 314 1,097
As it is seen in the table, 314 turnovers happened with a succession plan, while 783 without. This means that around 29% of the total CEO turnovers happened at a firm, which have already had a predefined succession plan. The table shows an obvious pattern, how CEO succession planning gains popularity with time. While 14.68% of total turnovers happened in 2010, only 6.37% of the total succession-plan turnovers happened in that year, and 18.01% of “no-succession-plan” turnovers.
However, the percentage of total turnovers happened in 2015 did not significantly increased compared to the ones in 2010, but already 22.61% of the “succession-plan” turnovers come from 2015, that means a substantial increase.
This pattern keeps the same if one considers another point of view. In 2010, 20 CEO turnovers were “succession-plan” turnovers out of 161, which means 12.4%, while the remaining 87.6% of turnovers in that year happened without succession plans. For 2011, this “succession-plan” turnover increases to 29.47%, then in 2012 it becomes 28.2%, in 2013 27%, in 2014 35%. Furthermore, in 2015, already 37.77% of the turnovers happened with a succession plan.
This huge development may suggest that succession planning becomes more and more important for the Board, which ultimately decides about the introduction of it. This development is consistent with Cvijanovic et al (2017) findings. Furthermore, it may also imply the lack of academic literature about this topic, as succession planning was so sparse in the past. But as it becomes more and more important, thus it makes it relevant to further investigate its effect.
Table 2 shows some interesting results. It compares the means of different firm and board characteristics in the same year, when the turnover happened between the sample with and without succession planning.
Table 2 Summary statistics
Table 2 reports summary statistics for firm and board characteristics of firms with and without succession in the year of the turnover. The mean and the standard deviation (in parentheses) are reported for the two sample of firms. Column 3 shows the result of the t-test of the difference between column 2 and column 1. All variables are defined in the Appendix. *** denotes significance at 1% level; ** denotes significance at 5% level; * denotes significance at 10% level
(1) (2) (3)
Without succession planning
With succession planning
t-test results (column 2 – column 1)
Return on Assets 0.00844 0.0289 2.735**
(0.111) (0.101)
Market to book ratio 1.663 1.754 1.358
(0.944) (0.913) Ln(Assets) 7.701 8.146 3.786*** (1.744) (1.614) Firm age 23.88 29.81 4.475*** (18.45) (21.84) Leverage 0.615 0.595 -1.138 (0.258) (0.250) Short-term leverage 0.0257 0.0212 -1.553 (0.0435) (0.0357) Dividend payer 0.512 0.681 5.033*** (0.500) (0.467) Return on Equity 0.0344 0.0526 0.574 (0.462) (0.446) Q 1.096 1.207 1.629 (0.981) (0.927) Log(Fixed compensation) 6.136 6.302 1.945* (1.299) (1.100)
CEO pay slice 0.268 0.259 -0.790
(0.157) (0.149)
CEO-Chairman duality 0.233 0.266 1.111
(0.423) (0.442)
Employee growth rate 0.0209 0.0260 0.456
(0.163) (0.152)
Sales growth rate 0.0429 0.0229 -1.645*
(0.183) (0.155)
As it can be seen three variables show significant differences at 1% level between firms with and without succession plans. One variable shows significant difference at 5% level, and two variables show 10% significance level.
The difference in return on assets suggests that firms with succession plans perform better at 5 % significance level in the year of the turnover than the firms without. But as it does not show the change in the performance, one cannot make any conclusions for the outcome of the main interest of this thesis.
Furthermore, from table 2, it can be also concluded that the firms with succession planning are older and bigger at 1 % significance level, as size is measured by the natural logarithm of the assets. Moreover, these firms are also more likely that they are dividend payers in the year of the turnover. Additionally, also the Log (Fixed compensation) variable almost reaches the threshold of 5% significance level, with 1.945 value for t-statistic. While regarding sales growth rate the firms with succession plans score worse compared to the ones without at 10% significance level.
These results are not fully consistent with the findings of Cvijanovic et al. (2017) as their paper shows very high absolute t-statistics for all the above described variables, thus also implies that the firms with a succession plan are significantly different from the ones without in several aspects in their sample. Although, there are differences between the results of this paper and Cvijanovic et al (2017) paper especially for the significance of the differences, but the sign of the t-test results mostly comply with Cvijanovic et al (2017).
5. Results
5.1 Determinants of succession planning
When Cvijanovic et al (2017) investigates the determinants of succession planning, they define as the interaction between the board and the power of the CEO, where the latter is measured with the CEO entrenchment. They motivate this with Hermalin and Weisbach (1998), who describes that the more the board appreciates the current CEO, the more reluctant the board is to strongly monitor her and the lower are the minimum requirements for the CEO. Thus, Cvijanovic et al (2017) points
out that the more entrenched the CEO is, the more likely she will be against the adoption of a succession plan.
CEO entrenchment is measured with three variables in this thesis. Based on Berger et al. (1997), CEO tenure is applied to capture the idea that the longer the CEO stays at her position, the more time she has to gain power over the board and other internal monitoring mechanisms. CEO fixed compensation is variable calculated as the sum of salary and bonus payments the CEO receives. So, the higher fixed compensation shows a stronger power for the CEO to extract resources from the company. Following Bebchuk et al (2011), the larger the gap between the pay of the CEO and the others of the management team, the higher the possibility that the CEO is entrenched. Thus, also CEO pay slice is added to the analysis that is the fraction of the aggregate compensation (TDC1) of the top five executives paid to the CEO from the ExecuComp database.
To measure board monitoring capability, independent board variables are used, that is the fraction of the board members, who have not been co-opted by the CEO. This data is from Naveen’s website and Coles, Daniel, and Naveen (2014) defines it as the residual tenure-weighted (TW) co-option is subtracted from one. TW co-option is “Sum of tenure of co-opted directors divided by the sum of tenure of all directors”. (Coles et al, 2014, p.1782)
Based on Cvijanovic et al. (2017), the author conducts a regression to investigate the determinants of CEO succession planning. Thus, the dependent variable is an indicator showing, whether the firm formally disclosed a CEO succession planning at least one year before the turnover. Control variables are lagged by one year. Table 3 shows the results of the regressions.
Table 3 Determinants of succession planning
This table shows estimates of OLS regressions of succession planning. The main independent variables are proxies of CEO entrenchment – Log (CEO tenure), Log (CEO fixed compensation), CEO pay slice) – and their interactions with Independent board, which is defined 1- residual tenure-weighted (TW) co-option as in Coles, Daniel, and Naveen (2014). The dependent variable is a dummy indicating if a firm has a succession plan in a given firmyear (in which case, it is 1) or not (in which case it is 0). Control variables are lagged by one year. Every variable is defined in Appendix A. The mean and the standard deviation (in parentheses) are reported. *** denotes significance at 1% level; ** denotes significance at 5% level; * denotes significance at 10% level
(1) (2) (3) Succession planning Succession planning Succession planning Ln(Assets) 0.0143 0.0159 0.0166 (0.014) (0.014) (0.014) Firm age 0.0029** 0.0029** 0.0029** (0.001) (0.001) (0.001) Return on Assets -0.2356 -0.2025 -0.2434 (0.287) (0.281) (0.282)
Market to book ratio 0.0265 0.0231 0.0276
(0.027) (0.026) (0.026)
Independent Board 0.3605 -0.3243* 0.2508
(0.268) (0.174) (0.163)
Log(CEO Tenure) 0.0012
(0.038) Log (CEO tenure) x Independent board -0.0846 (0.116)
Log(Fixed compensation) -0.0285*
(0.016) Log(CEO fixed compensation) x Independent board 0.0774***
(0.027)
CEO pay slice -0.0912
(0.214)
CEO pay slice x Independent board -0.3509
(0.547)
Constant 0.0610 0.2321 0.0711
(0.138) (0.152) (0.142)
Observations 440 456 459
The results shows a negative association between succession planning and entrenchment, measured as CEO fixed compensation and the pay slice, but if it measured with CEO tenure the sign changes to positive. However, the results are only significant in the case of the fixed compensation at 10 % level. The positive sign for CEO tenure is not significant, so it is not advisable to make any conclusion about its effect on succession planning.
The results are only partly what they are expected. Cvijanovic et al. (2017) receives results that entrenchment is negatively associated with the introduction of succession planning, with high significant rates of 5 % for tenure and fixed compensation, the results of this thesis are not as clear as theirs. Although they also find insignificant negative result for the pay slice. Furthermore, also the results for the independent board’s significance comply with their paper, too.
Based on the differences already in the descriptive statistics between this thesis and the work of Cvijanovic et al. (2017), it could have been expected that there will be differences for the determinants too.
Overall, these table suggests weak results that CEO entrenchment is negatively related to succession planning. They are weak, because of the significance issues.
5.2 Difference-in-differences results for the performance
To analyse the effect of succession planning on firm performance, difference-in-differences method is used. Table 4, which is based on Bennedsen et al. (2007) and Zhao (2015), presents the results.
Table 4 CEO succession planning and changes in firm performance around the turnover I.
Table 4 presents the changes in pre- and post-turnover for the turnovers during 2010-2015. Panel A reports changes in the unadjusted OROA. Panel B reports changes in industry-adjusted OROA, which is calculated by subtracting the mean OROA of the relevant industry and year from the unadjusted OROA. Pre is a proxy for the average performance of the two years prior to the turnover. Pre to 0 is the change from proxy pre and the turnover year’s performance, pre to 1 is the change of performance between the pre proxy and one year after the turnover, pre to is the change of performance between the pre proxy and two year after the turnover. The mean and the standard deviation (in parentheses) are reported. In column (4) *** denotes significance at 1% level; ** denotes significance at 5% level; * denotes significance at 10% level.
(1) (2) (3) (4) All With succession planning Without succession planning Difference
Panel A: Unadjusted OROA
Pre-turnover 0.075 0.0946 0.0667 0.028*** (0.1124) (.00533) (0.0045) (0.0077) Pre to 0 -0.0166 -0.0212 -0.0147 -0.0064 (0.1552) (.0104) (0.0054) (0. 0107) Pre to 1 -0.0071 -0.0098 -0.0059 -0.0039 (0.1265) (0.005) (0.0055) (0.0088) Pre to 2 -0.0092 -0.0144 -0.0068 -0.0076 (0.1086) (0.007) (0.0041) (0.0077)
Panel B: Industry-adjusted OROA
Pre-turnover -0.0024 0.0105 -0.0078 0.018*** (0.0895) (0.0048) (0.0034) (0.0061) Pre to 0 -0.0101 -0.0146 -0.0081 -0.0065 (0.1025) (0.0084) (0.0029) (0.0070) Pre to 1 0.0012 -0.0001 0.0018 -0.002 (0.0717) (0.0043) (0.0028) (0.005) Pre to 2 -0.0013 -0.0063 0.0011 -0.0075 (0.084) (0.0049) (0.0034) (0.0059) Observations 1016 301 715 1016
This table comperes the performance change around the turnover between the firms with and without succession planning. Year 0 marks the year of the turnover. “Pre” is a benchmark denoting
the two years average performance before the turnover. Although Zhao (2015) excludes the year -1, it was not necessary in this case, as turnover is captured with a different ExecuComp variable in this thesis compared to the paper of Zhao (2015), thus it made possible not to exclude year -1. Definition of turnover and succession planning elaborated in section 4.1.
In panel A unadjusted OROA is used as the dependent variable. As one can see, the performance decreases compared to the pre benchmark to -0.015 and -0.021. The magnitude of the decrease is higher for the firms with succession planning. However, there is a very low statistical significance ordered to the difference, only 0.598. This is even under the 50% significance level. The picture is not developing, if one looks at the change between the pre benchmark and one and two years after the turnover, one can see, that the effect of succession planning is negative, but the significance is still very low with a t-statistic of 0.44 and 0.99, respectively. Thus the difference between the performances of the pre benchmark and the year 2 is significant at 40% confidence level.
Panel B uses industry – adjusted OROA. The first row describes that there are significant performance differences between the two groups of firms, even when the industry effect is controlled. For pre to 0, pre to 1, and pre to 2, column (4) still shows negative, insignificant results. The t-statistics for them are 0.93, 0.4, and 1.27, respectively. This suggest that a firm with succession plan performs worse at 30% significance level two years after the turnover, than a firm where the turnover happened without succession planning. However, finance literature does not consider 40%-30% significance levels sufficient to make any conclusions, thus it is not possible to deduce any performance effecting consequence of succession planning. Hence, table 4 does not support H1.
However, the above described method does not have any control variables. Thus, these significance issues can be caused by other confounding effects.
5.3 Regression results
Table 5 CEO succession planning and changes in firm performance around the turnover II.
Table 5 shows the effect of succession planning on the difference of the performance between different years related to the turnover. Pre is a proxy for the average industry-adjusted performance of the two years prior to the turnover. Post is a proxy for the average industry-adjusted performance of the two years after to the turnover. Year 0 denotes the turnover year, year1 is one year after the turnover, and year2 is two years after the turnover. Dependent variable for column (1) to (3) is Post – Pre, for column (4) is Year0 – Pre, for column (5) is Year1 – Pre, for column (6) is Year2 – Pre. The main independent variable is Succession Planning, while the others are control variables. The mean and the t statistics (in parentheses) are reported. *** denotes significance at 1% level; ** denotes significance at 5% level; * denotes significance at 10% level
Dependent variable (Pre,Post) (Pre, Post) (Pre, Post) (Pre,0) (Pre,1) (Pre,2)
(1) (2) (3) (4) (5) (6) Succession Planning -0.0045 0.0020 -0.0000 0.0057 0.0067 -0.0059 (-0.9054) (0.4285) (-0.0055) (1.1995) (1.2128) (-0.8492) Ln(Assets) 0.0035*** 0.0028* 0.0021 0.0036** 0.0016 (2.7879) (1.7742) (1.5155) (2.2487) (0.7812) Firm age -0.0000 0.0001 0.0001 0.0000 0.0003* (-0.1777) (1.1165) (0.5525) (0.3245) (1.8020) Industry-adjusted OROA, t=-1 -0.3082*** -0.3493*** -0.0979*** -0.3668*** -0.3367*** (-12.9484) (-9.3494) (-3.1451) (-9.8271) (-7.1675)
CEO pay slice -0.0122 -0.0094 -0.0151 -0.0070
(-0.7023) (-0.6378) (-0.8682) (-0.3201) Independent Board 0.0149 -0.0074 0.0201** 0.0097 (1.5131) (-0.8813) (2.0454) (0.7909) Outsider -0.0112* 0.0018 -0.0033 -0.0190** (-1.9345) (0.3689) (-0.5765) (-2.5833) Constant 0.0011 (0.3911) Observations 959 925 468 490 468 452 R-squared 0.001 0.162 0.186 0.041 0.198 0.138
Year FE No Yes Yes Yes Yes Yes
Following Bennedsen et al. (2007) and Zhao (2015), different regressions are conducted with different control variables for different dependent variables.
Column (1) has the difference in industry-adjusted OROA between the two years average after the turnover and two years average prior the turnover as a dependent variable, but it does not have any
control variables or fixed effects, and the coefficient of succession planning is not significant and the R-squared is very low. Column (2) follows Bennedsen et al. (2007), who added control variables, such as firm age, size, and the industry-adjusted OROA one year prior the turnover. Adding these control variables results in the change of the sign of succession planning on the performance difference. However, the t statistics becomes lower.
For column (3) more control variables are added, following the logic of Zhao (2015). But the succession planning loses its predictive power. Thus, the author unwraps the “Post” benchmark to see how succession planning explains the differences between the different years around the turnover.
In column (4), where the dependent variable is the difference in industry adjusted OROA between the “Pre” benchmark and the year of the turnover, it can be first observed that the coefficient of succession planning largely gains its t statistic, to 1.1995, while also having a positive sign. Its magnitude grows higher for column (5). This indicated that firms, with succession planning have a higher positive change in performance one year after the turnover at almost 20% significance level compared to firms without succession planning.
The coefficient of succession planning becomes negative again in column (6) with a low t statistics.
Overall, the results of succession planning effect are ambiguous and do not support H1. Although the performance change is positive in case of several columns, just how it is expected, but the significance level of the succession planning variable is too low. Based on the highest significance level of 10% used in finance literature, one can conclude that changes in firm performance around the CEO turnover are not influenced by an existing succession plan.
There could be several reasons, why these results are ambiguous. First, it is possible that a firm, which says in its filings that it has a succession plan, but in the reality, it does not have one. For example, New York Stock Exchange requires the firms listed there to disclose CEO succession planning. Thus some firms may disclose a short notice about it just to comply with the requirement of NYSE, but they do not put succession planning into practice. Or also, the opposite can happen, when firms do not formally disclose their CEO succession planning arguing with the fact that it
might deteriorate their competitiveness, but in practice, they might have a well-developed succession planning in place.
Secondly, it is also imaginable, that succession planning has an importance from a different point of view, but not from performance. For example, its importance may come from only its volatility reducing effect, or management turnover smoothening effect, but succession planning itself may not have an effect on performance change.
Thirdly, based on the nature of the topic and its data collection intensity, it might happen that some firms were wrongly determined to have or not to have a succession plan.
Thus, maybe these reasons one by one would not affect the results seriously, however, if they add up, they might have serious consequences on it, leading to the experienced insignificance issue.
6. Robustness tests
6.1 Instrumental variable regression
To tackle the possible endogeneity regarding the data collection for succession planning, this part of the thesis uses two possible instrumental variables. One of them is compensation consultant and the other one is the Institutional Shareholder Services Directors Data’s “succession committee” variable. The economical meaning of the latter one is self-explanatory to instrument succession planning with it. However, having a compensation consultant as an instrument may need some further explanation to connect it to succession planning.
The best way to see whether compensation consultants and succession planning are economically related is to read the opinion of the compensation consultant firms about the topic. Thus, first the author researched the biggest of these firms based on their market share in 2015. According to the list of Equilar, the firms, which were both on the Fortune 1000 board engagements and management engagements lists are Pearl Meyer, Frederic W. Cook & Co., Meridian, Haygroup, Towers Watson, Pay Governance, and Compensation Advisory. After visiting the website of these companies, one can notice that in some form, all are in favor for succession planning.
Haygroup has a separate consulting service dedicated only for succession planning. One can find 449 results of articles, press releases etc. from the United States for the term “succession planning” on the website of Towers Watson. Pearl Meyer refers to succession planning as it has critical importance, and Meridian refers to it as “an increasingly hot topic in the boardroom and a focus of outside investors”. Frederic W. Cook & Co. organized an event in 2013 to facilitate succession planning. Although, Compensation Advisory Partners and Pay Governance do not pay a lot of attention to the topic, one can learn from Pay Governance’s website that 43 out of 100 companies delegate the succession planning responsibilities to the compensation committee. The author also noticed during the data collecting that the compensation committee and also the nomination committee are quite often the ones that oversee the succession planning process.
Based on this research about these consultant firms, one can conclude that there might be a positive relation between the succession planning and hiring a compensation consultant.
After explaining the economic reasons for the chosen instruments, the following tables describe the results of the instrumental variables regression.
Table 6 Instrumental variables regression
Table 6 Panel A shows the first stage results. The dependent variable is the succession planning indicator. The main independent variables are Succession Committee and Compensation Consultant, while the others are control variables. Panel B shows the effect of succession planning instrumented by succession committee or compensation consultant on the difference of the performance between different years related to the turnover. Pre is a proxy for the average adjusted performance of the two years prior to the turnover. Post is a proxy for the average industry-adjusted performance of the two years after to the turnover. The dependent variable is the difference of the proxy post and proxy pre. The main independent variable is Succession Planning, while the others are control variables. The mean and the t statistics (in parentheses) are reported. *** denotes significance at 1% level; ** denotes significance at 5% level; * denotes significance at 10% level
Dependent variable Dependent variable: succession planning
Part A. First Stage (1) (2) (3) (4) (5) (6)
Succession Committee 0.2108*** 0.1671*** 0.1088* (7.03) (5.05) (1.85) Compensation Consultant 0.1216*** 0.0768** 0.1300** (4.11) (2.09) (2.37) Ln(Assets) 0.0097 0.0012 0.0094 -0.0178 (1.07) (0.09) (0.90) (-1.10) Firm age 0.0021*** 0.0028** 0.0025*** 0.0025** (2.65) (2.47) (3.22) (2.23) Industry-adjusted OROA, t=-1 0.1723 0.1646 0.2225 0.0127 (1.21) (0.63) (1.55) (0.05)
CEO pay slice -0.1403 -0.1797
(-0.95) (-1.15)
Independent Board 0.1799** 0.2015**
(2.16) (2.37)
Outsider -0.1048 -0.0852
(-1.17) (-0.95)
Log (Fixed Compensation) 0.0126
(0.85)
Constant 0.1781 0.0794 0.1654 0.2463*** 0.138* 0.2667**
(7.55) (1.17) (1.37) (11.61) (1.88) (1.97)
Observations 959 925 468 959 925 465
Dependent variable: differences in industry adjusted OROA around the CEO turnover (two years average after the turnover – two years average
before the turnover)
Part B. Second stage (1) (2) (3) (4) (5) (6)
Succession Planning (=Succession Committee) -0.0193 0.0038 0.0941 (-0.69) (0.10) (0.89) Succession Planning (=Compensation Consultant) -0.0061 0.002 0.0116 (-0.13) (0.02) (0.16) Ln(Assets) 0.0039** 0.0002 0.004 0.0017 (2.16) (0.06) (1.60) (0.70) Firm age -0.0000 -0.0002 -0.0000 0.0000 (-0.17) (-0.44) (-0.09) (0.31) Industry-adjusted OROA, t=-1 -0.2375*** -0.2413*** -0.2371*** -0.2128*** (-8.47) (-4.51) (-6.93) (-4.64)
CEO pay slice 0.0013 0.0055
(0.04) (0.18)
Independent Board -0.0029 0.0156
(-0.11) (0.82)
Outsider 0.0013 -0.0055
(0.06) (-0.33)
Log (Fixed Compensation) -0.0061**
(-2.24) Constant 0.0035 -0.0353*** -0.0326 -0.0005 -0.0352** 0.0094 (0.39) (-2.68) (-1.00) (0.03) (-2.37) (0.36)
Methodology IV-TSLS IV-TSLS IV-LIML IV-TSLS IV-LIML IV-LIML
Anderson-Rubin p-value 0.3162 0.9823 0.8741
Observations 959 925 468 959 925 465
Table 6 shows the results of the regressions with instrumental variables. Part A describes the first-stage results. As one can notice both the compensation consultant and the succession committee variables are strong instruments only without controls. However, it is difficult to accept that any of these instruments would not influence succession planning or performance through other channels,
hence, control variables are necessary in the equation. Thus, in column (2) and (4) the author uses the controls as Bennedsen et al. (2007) and succession committee still keeps as a strong instrument, with an F-statistic of 25.45. When further control variables are added the F-statistic drastically drops. The compensation consultant instrument also loses its strength when the author adds any control variables to the equitation.
Thus, according to Stock & Watson (2015), there are two options. First is to find another, stronger instrument. However, as succession planning is not widely researched, one could have a hard time finding one. Hence, the author chose the second option, which is to continue with the weak instrument, but not with the two stage least squares (TSLS) method. Following the book’s advice, the limited information maximum likelihood estimator (LIML) is used in the second stage in case of column (3), column (5), and column (6).
Taking a look on part B of the table, one can see that succession planning stays similar as in table 5 even for the instrumental variable approach. However, the t-statistic slightly increases, it gives still insignificant results. Furthermore, also the Anderson-Rubin test’s p-value is insignificant. The best significance has column 3, which shows a positive result, that firms with succession planning have a higher average performance change after the turnover at 32% significance level (based on the AR p-value) compared to firms without succession planning.
6.2 Excluding financial firms
The next robustness test focuses on excluding financial firms as it is common in the literature. Taking a look on the database, it is noticeable that financial firms do not have a succession plan. Out of the 157 turnovers, which happened at financial firms, only one was with a succession plan. Thus, it is imaginable that the exclusion financial firms can increase the significance level and give better estimates for the performance change effect of the succession planning.
Table 7 Regression results excluding financial firms
Table 7 shows the effect of succession planning on the difference of the performance between different years related to the turnover, excluding financial firms. Pre is a proxy for the average industry-adjusted performance of the two years prior to the turnover. Post is a proxy for the average industry-adjusted performance of the two years after to the turnover. Year 0 denotes the turnover year, year1 is one year after the turnover, and year2 is two years after the turnover. Dependent variable for column (1) to (3) is Post – Pre, for column (4) is Year0 – Pre, for column (5) is Year1 – Pre, for column (6) is Year2 – Pre. The main independent variable is Succession Planning, while the others are control variables. The mean and the t statistics (in parentheses) are reported. *** denotes significance at 1% level; ** denotes significance at 5% level; * denotes significance at 10% level
Dependent variables (Pre,Post) (Pre, Post) (Pre, Post) (Pre,0) (Pre,1) (Pre,2)
(1) (2) (3) (4) (5) (6) Succession Planning -0.0045 0.0007 0.0038 0.0034 0.0088 0.0017 (-0.6834) (0.0982) (0.4168) (0.5588) (0.6252) (0.2030) Ln(Assets) 0.0060*** 0.0023 0.0033 0.0033 0.0005 (2.8683) (0.7414) (1.6177) (0.6902) (0.1631) Firm age -0.0001 0.0001 0.0001 -0.0001 0.0003 (-0.3670) (0.2744) (0.4785) (-0.2885) (1.6082) Industry-adjusted OROA, t=-1 -0.2637*** -0.2856*** -0.0967*** -0.2812*** -0.2880*** (-8.7342) (-5.2191) (-2.7342) (-3.3314) (-5.7726)
CEO pay slice -0.0235 -0.0111 -0.0406 0.0025
(-0.8004) (-0.5782) (-0.8966) (0.0936) Independent Board 0.0081 -0.0162 0.0216 -0.0053 (0.4784) (-1.4446) (0.8304) (-0.3443) Outsider -0.0122 -0.0013 -0.0144 -0.0092 (-0.7091) (-0.1135) (-0.5435) (-0.5888) Constant -0.0003 (-0.0679) Observations 809 785 404 424 404 389 R-squared 0.001 0.095 0.076 0.038 0.037 0.105
Year FE No Yes Yes Yes Yes Yes
Against what is expected, the significance further dropped in most of the columns. However, in column (3), where the two year average post performance is compared to the two year average pre performance, and where all the control variables added, the sign of succession planning beta
changes to positive and the absolute t-statistics increases from 0.0055 to 0.4168 compared to table 5.
Furthermore, also the R-squared drops in all the equitation compared to the original results. If one compares the instrumental variable regression with and without the financial firms, one will notice a drop both in significance and the strength of the instruments.
6.3 Reduced time period
To further investigate the results, the author conducted the main analysis for a reduced time. This period contains the turnovers, which happened from 2013 to 2015. It is best to investigate this reduced period as 60 % of the turnovers, which happened with a succession plan occurred in this time frame.
Table 8 Regression results for reduced time period
Table 8 shows the effect of succession planning on the difference of the performance between different years related to the turnover for the period of 2013-2015. Pre is a proxy for the average industry-adjusted performance of the two years prior to the turnover. Post is a proxy for the average industry-adjusted performance of the two years after to the turnover. Year 0 denotes the turnover year, year1 is one year after the turnover, and year2 is two years after the turnover. Dependent variable for column (1) to (3) is Post – Pre, for column (4) is Year0 – Pre, for column (5) is Year1 – Pre, for column (6) is Year2 – Pre. The main independent variable is Succession Planning, while the others are control variables. The mean and the t statistics (in parentheses) are reported. *** denotes significance at 1% level; ** denotes significance at 5% level; * denotes significance at 10% level
Dependent variables (Pre,Post) (Pre, Post) (Pre, Post) (Pre,0) (Pre,1) (Pre,2)
(1) (2) (3) (4) (5) (6) Succession Planning 0.0021 0.0024 0.0133 0.0069 0.0237 0.0024 (0.2422) (0.2615) (0.8133) (0.9536) (0.8532) (0.1745) Ln(Assets) 0.0064** -0.0024 0.0021 -0.0019 -0.0036 (2.4239) (-0.4794) (0.9454) (-0.2252) (-0.8498) Firm age 0.0001 0.0004 0.0004** 0.0002 0.0007** (0.3466) (1.0592) (2.0707) (0.2470) (2.0716) Industry-adjusted OROA, t=-1 -0.2113*** -0.1668* -0.0556 -0.1907 -0.1350* (-4.5622) (-1.8114) (-1.3919) (-1.2137) (-1.7437)
CEO pay slice 0.0186 -0.0079 -0.0042 0.0365
(0.3328) (-0.3167) (-0.0438) (0.7782) Independent Board 0.0254 -0.0281** 0.0465 0.0053 (0.8825) (-2.2343) (0.9479) (0.2200) Outsider -0.0219 -0.0079 -0.0309 -0.0140 (-0.5189) (-0.4441) (-0.4292) (-0.4046) Constant -0.0071 (-1.2952) Observations 496 483 193 204 193 184 R-squared 0.000 0.051 0.032 0.088 0.019 0.048
Year FE No Yes Yes Yes Yes Yes
Table 8 shows the results of the analysis conducted on the reduced time period. The t-statistic of succession planning still keeps low and its sign stays positive in all the six columns. The R-squared values are lower than in any other previous tables. If one uses the instrumental variable approach with the compensation consultant and succession committee instruments for this shortened time
window, one will notice similar reaction as in the previous subchapter: the t-statistics of succession planning further decreases, just like the instruments’ strength.
To sum up, one can see, that the t statistics of the succession planning variable do not change much. Thus, the results of table 5 are robust for a different methodology, for the exclusion of financial firms, and for a shorter time period, too.
7. Conclusion and Discussion
In this thesis the author used a unique, hand-collected dataset to investigate the effect of succession planning on the operational performance change of a firm around the CEO turnover. The research focuses on the turnovers, which have happened from 2010 to 2015 among US public firms.
First, the author identified these turnovers based on the ExecuComp database, then she conducted a hand-collection to decide, whether a given turnover happened with a succession plan. This decision is built upon the fact, whether a firm discloses succession planning in its DEF 14A or 8K filings at least one year, but not more than four years prior to the turnover. This is innovative in the literature of succession planning, as only one working paper captured this variable in a similar, but not the exact same way.
Thus, in the first part of the results chapter focuses on the determinants of the succession planning. Succession planning shows a negative association with CEO entrenchment, if it is measured as CEO pay slice and the logarithm of CEO fixed compensation. However, the result is only significant for the fixed compensation. This is not consistent with the findings of Cvijanovic et al. (2017), but this could originate from the fact they did not use the one year criteria on the data collection of succession planning.
Next, the difference-in-differences analysis gives mixed, insignificant results about the effect of succession planning both on the change of unadjusted and industry-adjusted OROA. It gives negative results for the difference, but only at 30%-40% confidence level.
Because of the possible confounding effects biasing the previous results, the author conducted an analysis with control variables. This gives again insignificant results for the effect of succession planning. One can conclude, that having a succession plan increases the change of the industry
