1
Financial Constraints and Relative Performance Evaluation in CEO
Turnover
Faculty of Economics and Business
MSc Business Economics: Finance Track
August 2015
Supervisor: Dr. F.S. Peters
Alexander Ezedin
2 Statement of Originality
This document is written by Student [Alexander Ezedin] who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 Abstract
This paper looks into how industry and relative performance affect the probability of a forced CEO turnover in the United States for the time period 1993-2013. This analysis is then refined to separate firms into highly and lowly financially constrained firms by computing the Kaplan-Zingales index. It finds similar results as Jenter & Kanaan (2014), namely that the probability of a forced turnover rises as industry performance worsens. This suggests that CEOs are dismissed due to exogenous shocks, which should not be factored in board decisions according to traditional economic theory. Similarly, and to an even stronger effect, the probability of a forced turnover rises as relative performance worsens. This suggests that boards actively use other peers in the same industry as benchmarks in assessing the performance of a CEO and making turnover decisions. However, whether a firm is highly or lowly financially constrained does not seem to be significant in affecting the probability of a forced turnover. These findings are largely reproduced when considering total turnovers in lieu of forced turnovers.
Keywords: CEO turnover, Forced turnover, Corporate governance, Kaplan-Zingales index of financial constraints, Peer performance, Industry returns, Relative performance
4
Table of Contents
I. Introduction………...…..5
II. Literature Review………..8
III. Methodology………...13
IV. Data and descriptive statistics………....16
4.1 Data………16 4.4. Descriptive statistics………...…18 V. Results………...22 5.1 Hypothesis 1………...22 5.2 Hypothesis 2 (1) & (2)………...23 5.3 Hypothesis 2 (3)……….25
VI. Robustness Checks………...……….27
6.1 Hypothesis 1………...…....27 6.2 Hypothesis 2 (1) & (2)………...28 6.3 Hypothesis 2 (3)……….29 VII. Conclusion………...……….31 References………....32 Appendix………...34 Tables………...34
5 I. Introduction
This thesis is based upon the paper by Jenter & Kanaan (2014) on CEO turnover and relative performance evaluation. That paper found that the probability of a CEO turnover is significantly greater after poor industry and market performance. The extent is stronger in relation to poor industry rather than poor market performance.
The sample period is more extensive in this paper; it spans from 1993 until 2013. Namely, the data set used in this paper spans 2,352 voluntary and 926 forced CEO turnovers for a total of 34,878 firm-years. Also the research is expanded by considering if there is a difference in turnover between more
financially constrained firms and less financially constrained firms.
The key variables that are repeated in all models are industry returns and industry-adjusted returns (using both value-weighted and equal-weighted industry returns). Industry returns are not directly reflective of CEO performance, and typical economic theory dictates that the board of directors should not consider these effects in a turnover decision (Jenter & Kanaan, 2014). However, strong results are found in contrast to this, i.e. industry returns are significant in predicting the probability of a forced turnover. The relationship is negative, signifying that an increase in the performance of the industry will tend to decrease the probability of a forced turnover. This signifies that, considering uniquely this variable, CEO turnover is anticyclical, at least as far as industry-specific business cycles go. More CEOs are fired in bad times than good. In this sense this contrasts to papers such as Eisfeldt & Rampini (2008), which establish the general procyclicality of CEO turnover. This paper had a slightly
perplexing finding that more CEOs are fired in good times than bad. Nevertheless, on the aggregate macroeconomic level this procyclical relation is likely to hold. However, the important point here is that whereas boards play an important role in corporate governance, specifically in monitoring top management, i.e. CEOs (Weisbach, 1988), they nevertheless do not seem to disregard exogenous factors when making turnover decisions. The results suggest that CEOs are in jeopardy of losing their jobs due to factors outside of their control, and are not judged simply based on merit in the company. Luck on how the industry performs also factors in.
Meanwhile industry-adjusted returns, i.e. the difference between firm returns and industry returns (or relative performance) is also highly significant and negatively correlated with the probability of a forced CEO turnover. The magnitude for industry-adjusted returns is greater than for simply industry returns, but both hold relevance in affecting turnover. There are important implications for this, namely
6
that the performance of a firm tends to be benchmarked in relation to how peers in the same industry group are performing when the board is making a turnover decision. Importantly it is not only firm performance that matters, but relative performance; even though the firm is outperforming the economy the CEO may be at greater risk of losing their position if the firm is not outperforming its peer group. This seems to be a more rational finding considering that outperforming CEOs are rewarded with an increased probability of keeping their job, while those who fail to remain competitive are at peril of losing their jobs.
The reason for considering financial constraints stems from the presence of investment-cash flow sensitivity when investment opportunities are factored in (Fazzari et al, 1988). More financially constrained firms may miss out on valuable investment opportunities, which may result in poorer firm performance. As in the previous passage, relative performance is significant in affecting the probability of a forced turnover, thus financial constraints may prove significant. Kaplan & Zingales (1997)
produce an index to measure financial constraints (which is reproduced in this paper). Their findings are different to Fazzari et al in that financially constrained firms exhibit a lower sensitivity of
investment to cash flow than unconstrained firms. In any case, the two papers demonstrate that the degree of financial constraints in a firm is significant. However, this paper is unable to establish a significant difference between financially constrained and financially unconstrained firms (as separated above and below the median Kaplan-Zingales index value). This result holds equally for forced
turnover and total turnover.
Thus overall results are largely reminiscent of Jenter & Kanaan (2014). Industry and relative
performance are significant in affecting the probability of a forced (and aggregated) turnover, whereas factoring in financial constraints produces insignificant results for the added terms. Testing for the difference equally yields insignificant results.
This thesis begins with a literature review discussing previous pivotal papers in the area. It presents papers that have previously considered industry and economy-wide conditions in predicting the probability of CEO turnover. Smaller passages are devoted to the role and effectiveness of board monitoring and decision-making. Some theory and previous findings on how financial constraints affect firms is also discussed.
The literature review is followed by a description of the methodology, which separates the thesis essentially into two sections based on the two principal hypotheses. The first one considers the impact
7
of industry and relative performance on the probability of a forced CEO turnover. The second one expands upon this by adding financial constraints. The second hypothesis is further divided into two subsections; the first splits the sample in two based on whether the Kaplan-Zingales (KZ) index value is above or below the computed median of the sample. A seemingly unrelated estimation is used to test whether there is a significant difference in the two samples. In the second subsection an interaction term is created, i.e. industry returns are multiplied by a dummy for firms with a KZ value greater than the median.
This is followed by a depiction of the data used plus descriptive statistics. Data uses hand-collected forced CEO turnover data, executive information from Execucomp in Compustat, further firm data on company fundamentals also from Compustat, data for monthly-stock returns from CRSP, and Fama French 48 industry information. Year fixed effects are computed using general economic data from the Federal Reserve plus S&P 500 market data.
Next the results are given. A total of three tables are constructed (which are presented in the appendix) using probit regressions with the dependent forced turnover variable taking the value of 1 if the CEO was dismissed, and conversely 0 if the CEO was not fired. This is followed by robustness checks by replicating the previous three tables, but for total turnover instead of forced (defined as the sum of forced and voluntary turnovers). Finally the paper concludes.
8 II. Literature Review
There are a few notable papers that have concerned CEO turnover and the business cycle, when considering the economy-wide level. Though this paper considers the industry-wide level instead, it is good to have an idea of the overall economy also. The paper by Eisfeldt & Rampini (2008) finds that CEO turnover and compensation are highly procyclical, and that the two are strongly correlated to one another. They find that these are due to countercyclical agency costs. To elaborate, they conclude that substantial bonuses need to be paid to relatively unproductive CEOs in order for them to be
incentivized in showing that capital ought to be reallocated away from them. However the correlation is stronger with compensation than turnover as measured against output (0.9 for compensation compared with 0.82 for turnover). Most critically however, they find the correlation of CEO turnover and GDP to be 0.818 and significant at the 5% significance level. However, the focus of these authors was
predominantly on external turnover caused by mergers and acquisitions.
Moving on to the main paper used for this paper, Jenter & Kanaan (2014), who focus on CEO turnover and relative performance evaluation, find that CEOs are vastly more likely to be dismissed after bad industry and bad market performance, i.e. by factors that are beyond their control, contrary to which standard economic theory would suggest in order to determine the quality of a manager. However, they also find that this is markedly different with previous empirical literature depicting that corporate boards do not act due to exogenous shocks when making CEO dismissal decisions. The main result of their paper is that CEOs are substantially more likely to be dismissed in the event of negative
performance shocks affecting their industry group. This accentuates the idea that the industry in which the firm operates matters, as does the performance of peers in determining the decision of whether to fire a CEO. It should be noted nevertheless that while there is evidence that boards analyze industry and market performance in determining the aptitude of their CEO, this is done at a superficial level and the effect of these actions are very weak. Their research uses 875 forced CEO turnovers for the period 1993-2009.
The three possible explanations they give for this result that more CEOs are dismissed when their peer group is performing on a subpar level include that managers may affect peer performance, boards might have better quality information concerning their CEOs at times of worse performance in the industry, or alternatively that boards act in a suboptimal manner thus misattributing exogenous performance elements to CEO performance. This is true also for CEOs with longer tenures or more
9
experience in the business. Also they point, on the theoretical level, that able boards do not dismiss more CEOs in bad than in good times.
Additionally, Jenter & Kanaan (2014) highlight the significance of industry recessions on CEO turnover; especially in periods of recession CEOs are more likely to be dismissed if they are underperforming rather than in booms. This marks a contrast to the general procyclical findings of CEO turnover by e.g. Eisfeldt & Rampini (2008) by suggesting an anticyclical pattern of turnover based on the industry-specific business cycle.
Also, Peters & Wagner (2014) accentuate that CEOs of companies that are experiencing more unstable industry conditions have a higher probability of being dismissed. Namely, a 0.1 increase in industry volatility is associated with a rise in the probability of a turnover by 1%. Additionally, CEO turnover risk is significantly positively related to compensation; there is roughly a 7% higher subjective compensation for every one percentage point increase in turnover risk. However, according to Bushman et al (2010) for specifically retained CEO, pay levels actually decrease the probability of a turnover.
Similarly, Eisfeldt & Kuhnen (2013) point that overall industry performance is critical in the decision for a forced CEO turnover, although the probability of a turnover is more likely if performance is mediocre compared to competitors and average levels. Indeed, they find that CEOs are more likely to be forced out of their position in the event of poor industry performance. They construct a model based on the assumption that industry conditions affect upon the choice of the ideal manager, which affects managerial turnover. Their other main conclusions are that both absolute and relative performance have an impact upon CEO turnover. Also turnovers are concentrated in certain industries and periods of time. It is important to point also that both firm and CEO characteristics in addition to the wider
industry conditions affect turnovers. However, in contrast, Kaplan & Minton (2008) find that boards do not appear to refer to the industry or market when making decisions of CEO turnover.
In a slight contrast, in so far as considering risk more generally, Bushman et al (2010) find in their paper that CEO turnover is increasing in idiosyncratic risk and decreasing in systematic risk, when controlling for firm performance. This signifies that it is more difficult for the board to distinguish the current CEO’s abilities during a downturn, i.e. where there is more systematic volatility possibly outside their control. Resultantly the underlying source of risk is important when making decisions and
10
it may not be as straightforward to fire CEOs especially in downturns. Furthermore, systematic risk does not necessarily influence the likelihood of a turnover.
Kaplan & Minton (2008) find that internal turnover is significantly related to three aspects of firm stock performance, which are performance relative to industry, industry performance relative to the overall market, and the performance of the overall stock market. Equivalently both forced and unforced turnover appear to be sensitive to the three types of subpar performance. They also find that CEO turnover is higher in the period 1998-2005 by 17.4% in their study from 1992-2005.This is somewhat striking, since this was a period of generally good economic performance in the US. In contrast external turnover, which is moreover related to mergers & acquisitions is not significantly related to any of the three stock performance measures.
It is also critical to point that Kaplan and Minton (2008), who concentrated on the general patterns in turnover and compensation, found that the period during which compensation increased substantially matches the period of higher turnovers. The important point they make is that there has been an
inherent rise in riskiness in CEO jobs. This is caused partially by increased regulation, such as Sarbanes Oxley legislation. Critically, they state that boards have become more sensitive to the performance of the CEO.
Results are similar for both forced and voluntary turnover. There is also evidence that increases in turnover are related board independence. On the other hand, external turnover is not significantly related to any of the aspects of stock performance.
Mikkelson & Partch (1997) concentrate more on external turnover and separate two periods when the takeover market was more active (1984-1988) and less active (1989-1993). They find that turnover and performance are related only in the active takeover period, and that takeover activity affects the
intensity of managerial discipline. They also find that only age of the CEO is reliably related to
turnover when examining the relationship with management turnover, whereas company size, financial leverage, board size and board composition are not.
Faleye et al (2011) focus on the effects of the intensity of board monitoring on the effectiveness of directors in conducting their monitoring and advising duties. They find that the quality of monitoring is ameliorated when the bulk of independent directors serve on a minimum of two of the three key
11
monitoring committees. However, this improvement in monitoring quality comes at the expense of poorer performance and short-sightedness of managers. There is thus resultantly poorer performance and a decline in corporate innovation. Thus increasing monitoring can in essence be likened to a double-sided sword where the drawbacks potentially outweigh the benefits.
Additional points that they make include that independent directors serving on several monitoring committees enables the directors to have a more complete view of the entire firm, which can aid in making better decisions. However, intense monitoring is costly to implement, and can actually leave directors with less time and information, which would contradict the entire purpose of monitoring. Therefore monitoring effectiveness is by itself not enough as a proxy for good corporate governance. However, empirical tests do not conclude any relationship between board structure and firm value. They do not find any significant effect for board size and board independence. It is also critical to note that Faleye et al (2011) only consider the period 1998-2006.
It is important to add that Bushman et al (2010) also consider the skillset of a potential replacement CEO in relation to the current one in assessing whether to dismiss the current CEO. Thus innate talent is also critical. However, in many cases, arguably, the CEO is dismissed with no clear succession plan or suitable replacement already in mind.
Other points include that authors such as Jenter & Kanaan (2013) signal to other factors that potentially influence CEO turnover in the case of worse performance, namely corporate culture. They find that turnover increases with a culture for control, but decreases with a culture leaning towards collaboration. They place special emphasis on performance measures such as ROA. In particular, the authors find a strong negative relationship between the firm-specific ROA and the probability of a CEO turnover. Similarly, Kaplan & Minton (2008) consider ROA as a measure of operating performance, and find similar results.
As an added note, according to Schoar & Zuo (2011), a CEO’s path and career are also influenced by the economic conditions in which they start. CEOs starting in recession times tend to be more
conservative in approach. This may also impact turnover through the riskiness of actions pursued by CEOs and subsequent board action. This rests upon the idea that CEOs have individual management styles and that they greatly affect the performance operations of the firms they direct.
12
Finally papers analyzing financial constraints are discussed. While the paper by Malmendier & Tate (2005) is involved with behavioral reasons for subpar corporate investment decisions, the manner with which they construct the Kaplan-Zingales Index is used as a model for this paper. The Index utilizes more straightforward ways to measure financial constraints, and illustrate in their paper that financially constrained firms exhibit a lower sensitivity of investment to cash flow than financially unconstrained firms (Kaplan & Zingales, 1997). While Fazzari et al (1988) also show that financial factors affect investment, and underline the connection between financial constraints and how investment varies by the category of firm, they find an opposite result. To elaborate, more financially constrained firms are more sensitive to changes in cash flow than firms which are able to pay relatively high dividends. Despite of the contrasting results in the papers, financial constraints do appear to have a significant impact upon investment activities of a firm, which is likely to lead to a decrease in the potential growth of a company. It is reasonable to assume according to standard economic theory that a company that is free to participate in more NPV positive projects can perform better than a firm that is more limited in ability. This directly relates to the idea of relative performance in the sense that more successful companies (i.e. firms outperforming the industry) experience less turnovers, as in Jenter & Kanaan (2014).
13 III. Methodology
This thesis involves collection of data from multiple sources, namely Compustat, CRSP, hand collected data on forced CEO turnovers and Fama French industry returns. The data will be collected for the period 1993-2013. Throughout this paper probit models will be used to test the effect of peer returns, industry adjusted returns, and financial constraints on the probability of a forced CEO turnover. Also, clustered standard errors will be used throughout this paper.
The first main hypothesis tests how peer (rpeer
it) and industry-adjusted firm performance (rit - rpeerit ) affect the probability of a forced CEO turnover. This is similar to Jenter & Kanaan (2014). Industry-adjusted firm performance is defined as the difference between 12 month rolling firm returns and 12 month rolling industry returns. Returns are compounded over the past 12 months to arrive at the correct figures. Both value-weighted and equal-weighted industry 12 month rolling returns will be constructed in testing this hypothesis. Whereas value-weighted returns may be more biased towards larger
companies, equal-weighted returns give the same weight for all companies in the data set. However, this may also give a slightly distorted view of the market. Consequently, the results are expected to be similar for the two, although some variation is possible for this reason. As a reference point, Jenter & Kanaan (2014) obtained significant and negative results for both coefficients for the period 1993-2009.
Hypothesis (1): The probability of a forced turnover is inversely related to both peer and industry-adjusted firm performance.
Pr(ForcedCEOTurnover = 1 | X0,…,X2) = Φ(β0 + β1rpeerit + β2(rit - rpeerit ))
It is anticipated that an increase in peer performance will decrease the probability of a forced CEO turnover; as peer performance ameliorates, it is likely that the average firm is less likely to experience a forced CEO turnover. This suggests that the industry is performing well overall. A similar, and possibly stronger, (inverse) effect is anticipated for industry-adjusted returns. A firm outperforming its peers is presumably less likely to experience a forced CEO turnover, as the CEO is effectively in charge of the
14
strategy of the firm. The board is less likely to fire a CEO under whom the firm is reporting good performance.
This hypothesis will also be tested by using year fixed effects. Year fixed effects will include
macroeconomic data from the Federal Reserve, namely US annual GDP figures and growth rates. Also annual market performance of the S&P 500 index will be considered.
The second principal hypothesis differs from Jenter & Kanaan (2014) in that it considers also how financially constrained a firm is, and whether this has an effect on the probability of a forced turnover. A firm that is highly financially constrained is less likely to be able to access external finance and potentially cannot respond as efficiently to available positive NPV projects. This may also have an impact on financial performance of the firm and also the probability of a forced CEO turnover.
Hypothesis (2): Forced turnover differs between financially constrained firms and financially unconstrained firms.
To test this hypothesis the Kaplan-Zingales (1997) Index will be constructed using Compustat data. The Index is defined in the following manner, as in Malmendier & Tate (2005):
KZit = -1.001901*(CFit/Kit-1 )+ 0.2826389*Qit + 3.139193*Leverageit – 39.3678*(Dividendit/Kit-1 ) -
1.314759*(Cit/Kit-1)
In this formula KZ is the index, CF is cash flow, K is the lagged (one year) value of property, plant and equipment, Q is Tobin’s Q, Leverage is financial leverage of the firm, Dividend is the sum of ordinary and preferred dividends of the firm, and C is capital.
The data sample will be split into two according to the median KZ value. Firms with a KZ value greater than the median will be divided into the upper half (and be defined as financially constrained firms), and firms with a KZ value less than the median will be divided into the lower half (and be defined as financially unconstrained firms). First the following probit regressions will be constructed in
15
(1) Pr(ForcedCEOTurnover = 1 | X0,…,X2) = Φ(β0 + β1rpeerit + β2(rit - rpeerit)) KZ>median
(2) Pr(ForcedCEOTurnover = 1 | X0,…,X2) = Φ(β0 + β1rpeerit + β2(rit - rpeerit)) KZ<median
Reminiscent of hypothesis 1, ( rpeerit ) is the industry return and (rit - rpeerit) is the industry-adjusted return. Also similar to hypothesis 1, industry and industry-adjusted returns will be calculated using both value-weighted and equal-weighted returns. The difference in the two approaches to calculating returns is again expected to be small.
Next, the following regression is constructed that includes an interaction term between the dummy for a
financially constrained firm (i.e. KZ>median) and the industry return (i.e. rpeer
it * FCit) (for value- and equal-weighted returns).
(3) Pr(ForcedCEOTurnover = 1 | X0,…,X3) = Φ(β0 + β1rpeerit + β2(rit - rpeerit) +β3( rpeerit * FCit))
A significant result in the interaction term would suggest that a highly financially constrained firm has a (significantly) different probability of a CEO turnover than that of a firm not financially constrained. Finally a test will also be constructed that will involve a seemingly unrelated estimation in ascertaining whether there is a difference in forced turnover between financially constrained and unconstrained firms. This will be more conclusive in determining exactly whether or not there is a significant difference in the two.
Specifically, equation (1) and (2) will be used. Equation (1) is defined by firms with a high KZ value (i.e. greater than the median), whereas equation (2) is defined by firms with a low KZ value (i.e. lower than the median). The test will involve analyzing whether highKZ = lowKZ. A significant result would confirm this, whereas an insignificant result would not support a difference between the two.
16 IV. Data and descriptive statistics
In this section the data used in the analysis is described. 4.1 Data
Firstly, information is acquired on CEOs from the Execucomp database. The database includes information on the main executives for firms in the S&P 500, S&P MidCap, and S&P SmallCap indexes. This is conducted by restricting observations to where CEOANN (Annual CEO Flag) equals CEO. This is for the period 1993-2013, and produced 37,172 observations. The key variables included in the search for company information are Company Name and Company ID Number (i.e. gvkey). For executive information the following variables were used: Executive ID number (i.e. execid), First Name, Last Name, Date Joined Company, Date Left as CEO, and Reason Left Company. Also CEO age and SIC Code (for industry classifications later on) are collected at this stage.
This data was then merged with the hand-collected data provided kindly by Dr. Peters for forced CEO turnover from 1993-2010. This was augmented by data from 2011 and 2012 by determining reasons for CEO turnover using news sources, especially LexisNexis conducted by myself and a colleague. The forced observations were then added to the hand-collected data by Dr. Peters. CEO turnover is
determined as forced similarly to Peters and Wagner (2014), i.e. forced if the CEO was below a certain age at the time of departure. Then the aggregated forced turnover data from 1993-2013 is merged with the Execucomp data. It should be noted that some data up until 2014 is also in this data; however this year is not complete (and is not included in the analysis). Data is merged on gvkey and year.
Next individual returns of firms are merged from CRSP. Monthly stock returns (item RET) were downloaded from CRSP for the time period January 1990- December 2013 for all firms. This resulted in a total of 2,181,637 observations across all firm years in the database. 12-month rolling firm returns are computed from the monthly returns. The critical variables here are the firm identifier (permno), month of the fiscal year (month), and ret12m (rolling 12 month firm returns).
To be able to merge the above data from CRSP to the Compustat data, the permno and date of the fiscal year end (datadate) needed to be added to the main file. This information is found from Fundamentals Annual in the CRSP-Compustat Merged database. Before this step gvkeys were matched with the appropriate permno. Finally the individual returns are merged to the main file by permno and month. The variable month is thereafter defined as the month of the fiscal year end if there is no forced
17
turnover, and the month of turnover minus three if there is a forced turnover. Observations are dropped for which there is no datadate or no values for a 12 month return.
The next step is to add data on industry returns. Data for value-weighted and equal-weighted Fama-French 48 industry returns are used for the appropriate time period. It follows the Fama Fama-French (1997) classification. 12 month rolling returns are computed in a similar way as for firms above, but for the respective industry instead. Similar as above, observations are dropped for which there is no 12 month return. Data is downloaded on Compustat for firm sic codes to enable classification into the 48
industries. This list is seen to be a comprehensive choice for industry classifications. Finally, 12-month industry returns (both value and equal-weighted) are merged to the main file on month and FF48 group. Finally, the Kaplan-Zingales (1997) index of financial constraints is constructed similar to Malmendier & Tate (2005) using Compustat data (specifically Fundamentals Annual) and merged to the main file. This is to check whether the coefficients of the regressions vary for financially constrained firms relative to others. The relevant variables are income before extraordinary items (ib), depreciation and amortization (dp), total assets (at), common shares outstanding (csho), price close – annual – fiscal (prcc_f), common equity (ceq), deferred taxes (balance sheet) (txdb), total debt in current liabilities (dlc), total long-term debt (dltt), stockholders equity – parent (seq), dividends common/ordinary (dvc), dividends – preferred/preference (dvp), cash and short-term investments (che), and finally property, plant and equipment – total (ppent). Cash flow is defined as (ib+dp), Tobin’s Q is (at+(csho*prcc_f)-ceq-txdb)/at), financial leverage is (dlc+dltt)/(dlc+dltt+seq), dividends is (dvc+dvp), cash is (che), and capital is ppent in the previous year. This data is then merged to the main file on gvkey and year. Voluntary and aggregated turnovers are also identified in the data. Voluntary turnovers are classified as events when the CEO of a firm in Execucomp changes, and that turnover is not forced. An (aggregated) turnover is simply when a CEO changes, whether forced or voluntary. The main regressions in this paper will be repeated with aggregated turnovers in the robustness checks section.
Year fixed effects are computed using economic data for the United States and market data for the S&P 500 Index. GDP data for the United States is found from the US Bureau of Economic Analysis,
specifically the Federal Reserve Economic Data of the St. Louis Fed. It is collected on a quarterly basis and subsequently annualized. Specifically quarterly figures are collected as real gross domestic product in billions of chained 2009 dollars using a seasonally adjusted annual rate. Standard and Poor’s 500
18
Index performance (i.e. S&P value) is on a monthly basis from January 1993 until December 2013. This is annualized similarly as GDP.
4.2 Descriptive statistics
After all duplicates and relevant missing values have been deleted the total amount of observations is 34,878 with 926 forced observations across all firm-years, as seen in Panel A. Voluntary turnovers amount to 2,352, giving a total of 3,278 turnovers in the sample. In percentage terms the proportion of firm-years with a minimum of one forced, voluntary, and aggregated turnover are 2.65%, 6.74%, and 9.40%, respectively.
Panel B, on the other hand, illustrates average and median values of various pertinent firm characteristics. Also, critically, the individual components of the KZ index are shown. All firm
characteristics depicted are book values. It is interesting to compare the value of firm characteristics for three turnover outcomes, i.e. when there is a forced, voluntary, and no turnover. Mean total assets are larger for companies experiencing a forced turnover, at 13,477 million dollars in comparison to those with a voluntary or no turnover at 9,865 and 10,011 million dollars, respectively. It should also be noted that the median is considerably smaller than the average in all three turnover outcomes, especially so in the case of a forced turnover. Namely the median is 1,051, 1,685, and 1,333 million dollars, respectively. This suggests that more very large companies have more forced turnovers than other smaller companies.
Similarly, companies experiencing a forced turnover appear to have greater liabilities on average than companies experiencing a voluntary or no turnover. Namely, the average values are 10,858, 7,105, and 7,469 million dollars, respectively. Again the median is considerably smaller at 561, 889, and 687 million dollars, respectively.
Differences in stockholder’s equity are not that great among the three outcomes at 2,570, 2,668, and 2,470 million dollars, respectively. However, income before extraordinary items is notably lower in the case of a forced turnover when compared to a voluntary and no turnover. Average values are 109, 334, and 318 million dollars respectively, whereas median values are lower at just 3.88 million dollars for firms with a forced turnover, 61 million dollars for firms with a voluntary turnover, and 55 million
19
dollars for a firm with no turnover. Thus the company seems to be performing less well, on average, in the case of a forced turnover, which is to be expected.
Moving on to the KZ index and its individual components also points to some notable differences between the different turnover outcomes. Similar to income before extraordinary items, also cash flow is diminished for a firm with a forced turnover in comparison to ones with a voluntary or no turnover. Namely, the average values are 352, 622, and 558 million dollars, respectively. Median values are 37, 122, and 106 million dollars, respectively.
The Tobin’s Q is reminiscent of the above results. A value greater than 1.0 signifies a market value of assets that exceeds its recorded assets (Dimand, 2014, p.214). A higher value indicates that the market values assets more highly than recorded. The average and median values of Tobin’s Q are lower than in the case of voluntary and no turnovers. Specifically, averages are 1.58, 1.90, and 1.97, and median values 1.23, 1.42, and 1.49, respectively. A company whose assets the market values less favorably is more likely to experience a forced turnover.
Financial leverage is higher on average for firms with a forced outcome; however the median is similar to the average and median values of firms with voluntary and no turnovers. This suggests that the average company with a forced turnover outcome has more debt, but the ratio is still fairly low, which should not indicate insurmountable indebtedness. Namely, average values are 0.69, 0.35, and 0.33, while median values are 0.34, 0.34, and 0.32, respectively.
Also dividends are lower in the case of a forced outcome, with the average value at 109 million dollars and the median at an even 0. This contrasts to average values of 149 and 117 million dollars, and median values of 9.24 and 4.35 million dollars for firms with voluntary and no turnovers, respectively. However, cash and short term investments appears to be actually at the highest levels for firm with a forced turnover outcome, on average, at 1,950 million dollars. Nevertheless, the median value is just 82 million dollars. This compares to average values of 926 and 1,088 million dollars, and median values of 96 and 99 million dollars, respectively, for voluntary and no turnovers.
Capital is lowest in the case of firms with forced turnovers, with average and median values of 1,693 and 175 million dollars, respectively. This contrasts to firms with voluntary and no turnovers with higher average values of 2,254 and 1,869 million dollars, and higher median values of 331 and 234 million dollars, respectively.
20
Finally, the KZ index appears to show relatively healthy results for all companies, with negative results for each suggesting a relatively low degree of financial constraints. There are differences however, with companies with no turnover demonstrating a very healthy KZ value of -21.82, on average. This
contrasts to -7.07 for forced, and -7.35 for voluntary turnovers. Median KZ values, on the other hand, do not show as great differences. They are the highest for firms with forced turnover outcomes at -0.73, and lowest for firms with no turnovers at -1.38. Firms with voluntary turnover outcomes have a median KZ value of -1.17.
Panel C is the final panel in table 1, and it gives a picture of average and median firm plus industry performance by CEO turnover outcome. As in Panel B, the three turnover outcomes are the same, i.e. forced, voluntary, and no turnover. Standard errors of the mean are reported in brackets below the average values.
The first item of interest is a firm’s stock return in the 12 months prior to a CEO turnover. The average firm has a 12 month rolling stock return of -12.78% in the case of a forced turnover. Similarly the median firm has a 12 month rolling stock return of -16.54% in the same scenario. This suggests that a firm involved with a forced turnover is more likely to have been experiencing poor performance in the 12 months before deciding to fire the incumbent CEO.
The opposite is true for firms experiencing voluntary or no turnovers. The average values are 11.90% and 19.34%, whereas the median values are 5.84% and 11.90%, respectively. These are sizable positive returns, especially in the case of firms with no turnovers. This follows standard economic theory, as a firm that is performing well should be less likely to experience a turnover than one that is performing poorly. It should also be noted that the standard error of the mean is much smaller for firms with no reported turnover at 0.38%, as compared to -1.55% for forced and 1.22% for voluntary turnovers. Rolling 12 month industry stock returns are computed using both value and equal-weighted returns. Equal-weighted returns are consistently higher than value-weighted returns, and conversely also have larger standard errors.
Industry returns are consistently positive for forced, voluntary and no turnovers. However, they are the lowest for firms with forced turnovers. Namely, firms with forced CEO turnovers have an average and median return of 8.65% and 10.68% for value-weighted returns, and 11.03% and 9.36% for
equal-21
weighted returns, respectively. Standard errors of the means are 0.83% for value-weighted and 1.07% for equal-weighted returns.
Firms with voluntary turnovers exhibit higher industry performance. Specifically, the average and median value-weighted returns for these firms is 10.47% and 11.01%, respectively. The standard error is 0.52%. Conversely for equal-weighted returns the values are 14.12% and 12.78%, with a standard error of the mean amounting to 0.70%.
Finally, and to reiterate, firms with no turnover have notably higher returns than their forced and voluntary counterparts. For value-weighted industry returns the relevant average and median figures amount to 12.33% and 12.72%, with a standard error of 0.15%. The same for equal-weighted industry returns is 15.45% and 13.18%, respectively, with the standard error of the mean coming to 0.21%.
22 V. Results
5.1 Hypothesis 1
Table 2 presents findings using the first hypothesis specified in the methodology. Namely, the effect of value and equal-weighted 12 month rolling industry returns, and value plus equal-weighted industry-adjusted returns is considered upon the probability of a forced CEO turnover. Year fixed effects are used in models [3] and [4]. Average marginal effects are computed to determine numerically the impact of a 1% increase of the independent variables on the dependent variable.
Model [1] considers value-weighted industry returns. Both the coefficients for the 12 month rolling industry return and industry-adjusted return is negative and significant at the 0.1% level. Specifically, a 1% increase in the industry return is expected to decrease the probability of a forced CEO turnover by 2.35%. Thus an elevation in the performance of a company’s peers is expected to diminish the
likelihood of a forced turnover of a firm in the same industry. This is significant, so boards appear to also consider industry-wide performance when making decisions on a turnover. This is similar to findings in Eisfeldt & Kuhnen (2013), in that industry performance matters. However, overall industry performance may not be reflective of individual CEO skill or benefit to the firm they direct. Forced turnover appears also to be anticyclical, in contrast to Eisfeldt & Rampini (2008), at least at the industry level.
Similarly, a 1% increase in the industry-adjusted return (which was simply the difference in the firm’s and industry’s performance) is expected to decrease the probability of a forced turnover by 4.68% on average. Notably forced turnover is more sensitive to the industry-adjusted return than the industry return, which is to be expected. Though the industry may be doing well, it is also important to consider the performance of the individual firm in question, which the boards appear to do based on this. This also suggests that industry performance is an important proxy that influences boards in turnover decisions. Though a firm may be performing admirably in relation to the economy at large, its competitiveness with its own industry is critical. Similarly, Kaplan & Minton (2008) accentuate the importance of performance relative to the overall market.
Model [2], on the other hand, consider equal-weighted industry returns. However, results are widely similar to the value-weighted counterparts with negative coefficients for both industry and industry-adjusted returns that are significant at the 0.1% level. Specifically, a 1% increase in the 12 month
23
rolling industry return is expected to decrease the probability of a forced CEO turnover by 2.86% on average. Also, a 1% increase in the industry-adjusted return is expected to decrease the probability of a forced turnover by 4.76%. The effects are predicted to be slightly greater using equal-weighted than value-weighted returns.
Model [3] is identical to [1], but also considers year fixed effects. The results also parallel those of [1]. A 1% increase in the value-weighted industry returns is expected to decrease the probability of a forced turnover by 2.49%. Conversely, a 1% increase in the industry-adjusted returns is expected to decrease the probability of a forced turnover by 4.87%. The two variables are both significant at the 0.1% level. Likewise, Model [4] mimics [2], but adds year fixed effects. While both coefficients are again
significant at the same level, the estimated impact of industry returns is more pronounced here.
Namely, a 1% increase in the equal-weighed industry returns is attributed with a 3.29% decrease in the probability of a forced turnover. A 1% increase in the industry-adjusted returns is attributed with a 4.82% decrease in the probability of a forced turnover, which is similar as in the other three models.
<<Insert Table 2 Here>> 5.2 Hypothesis 2 (1) & (2)
The following tables are based on variations of the second principal hypothesis, i.e. they analyze the significance of financial constraints on predicting a forced CEO turnover. Table 3 separates the sample in two based on firms with a KZ index value greater than the median and firms with a KZ index value lower than the median. Firms with a KZ index value greater than the median are considered to be relatively highly financially constrained. Correspondingly, firms with a KZ index value lower than the median are consider to be financially unconstrained. Models [1] and [2] deal with value-weighted returns, whereas models [3] and [4] consider equal-weighted returns.
To commence, model [1] considers only firms with a computed KZ value greater than the median. For these firms the 12 month rolling value-weighted industry return and industry-adjusted return are both negative and significant at the 0.1% level. A 1% increase in industry returns is expected to decrease the probability of a forced turnover by 2.75%, whereas an identical increase in industry-adjusted returns is expected to decrease the probability of a forced turnover by 5.06%.
24
In contrast model [2] takes into account firms with a computed KZ less than the median. Coefficients are identical in sign and significance as before. Computing marginal effects yields proportionally smaller results for less financially constrained firms than for highly financial constrained firms.
Empirically, a 1% increase in the industry returns is attributed with a 1.76% decrease in the probability of a forced CEO turnover. This is nearly one percentage point lower than in the case of financially constrained firms. Similarly, a 1% increase in industry-adjusted returns is expected to decrease the probability of a forced turnover by 4.05%. Again, this figure is minutely over one percentage point smaller than in the case of a financially constrained firm. These results suggest that highly constrained firms are more sensitive to both industry and industry-adjusted returns than non-financially constrained firms. However, this will be later (after model [4]) empirically verified using a seemingly unrelated estimation. In any case, these results give further credence to findings that forced turnovers are more probable in the event of poor firm performance in relation to the industry as a whole, than to merely industry returns, though also these are significant, but at a lesser magnitude.
Moving on to equal-weighted returns instead of value-weighted ones in the next two models, model [3] is otherwise identical in specification to [1] by considering only firms with a KZ index value greater than the median. Coefficients are again identical to those in previous models in terms of sign and significance. This model estimates that a 1% increase in industry returns is associated with a 3.16% decrease in the probability of a forced turnover. Conversely, a 1% increase in industry-adjusted returns is associated with a 5.21% decrease in the probability of a forced turnover. These results are slightly greater than in the case of value-weighted returns.
On the other hand, model [4] considers firms with a KZ index value less than the median. Equal-weighted industry returns and industry-adjusted returns are both negative and significant at the 0.1% level. The model attributes a 1% increase in industry returns with a 2.34% decrease in the probability of a forced turnover. Equivalently, a 1% increase in industry-adjusted returns is associated with a 4.05% decrease in the probability of a forced turnover. As in the first two models, highly financially constrained firms appear to be more sensitive to industry and industry-adjusted returns than not financially constrained firms. Proportional changes are around one percentage point less for the two coefficients of interest between the two samples.
25
However, a seemingly unrelated estimation is mandated to empirically test whether there is a difference between highly financially constrained and not financially constrained firms. This is tested for both value-weighted and equal-weighted industry returns.
In the case of value-weighted industry returns the test reports insignificant results, with a χ2 value of
0.676. When repeated for equal-weighted industry returns the test reports a higher χ2 value of 0.817.
However, this finding is again insignificant. Thus it is impossible to conclude that a firm that is highly financially constrained has a greater probability of a forced CEO turnover than a firm that is not very financially constrained. The results are not significantly different between the two groups of firms.
5.2 Hypothesis 2 (3)
Table 4 follows specification (3) of the second principal hypothesis. It differs from earlier
specifications by adding an interaction term between the industry return and a dummy for a firm with a high KZ value (i.e. greater than the median). This is a different approach of attempting to test whether the interaction term is significant, and consequently if there is a difference between financially
constrained and unconstrained firms. Models [1] and [3] consider value-weighted industry returns, whereas models [2] and [4] consider equal-weighted industry returns. Models [3] and [4] are otherwise identical to [1] and [2], respectively, but add year fixed effects.
In model [1] the value-weighted industry return and industry-adjusted return are both negative and significant at the 0.1% level. However, the interaction term, i.e. the product of industry returns and the high KZ dummy, is not significant. Analytically, a 1% increase in industry returns is associated with a 2.30% decrease in the probability of a forced turnover. Additionally, a 1% augmentation in the
industry-adjusted returns is associated with a 4.68% decrease in the probability of a forced turnover. These results widely follow findings in table 2.
In model [2] the equal-weighted industry return and industry-adjusted return are again both negative and significant at the 0.1% level. A 1% increase in industry returns is associated with a 2.95% decrease in the probability of a forced turnover. Similarly, a 1% rise in the industry-adjusted returns is associated with a 4.77% decrease in the probability of a forced turnover. As in [1] the interaction term is not significant (at any level).
26
Model [3] is largely reminiscent of [1] in regards to variables, sign, and significance. Namely, a 1% increase in industry returns is associated with a 2.45% decrease in the probability of a forced turnover. Similarly, a 1% rise in the industry-adjusted returns is associated with a 4.87% decrease in the
probability of a forced turnover.
Equivalently, model [4] closely resembles [2] in regards to variables, sign, and significance. Namely, a 1% increase in industry returns is associated with a 3.34% decrease in the probability of a forced turnover. Similarly, a 1% rise in the industry-adjusted returns is associated with a 4.82% decrease in the probability of a forced turnover.
Overall (and again) the findings in this table give further evidence that there is not a significant difference in the probability of a forced turnover between financially constrained firms and non-financially constrained firms. However, industry returns and industry-adjusted returns are critical in affecting a forced turnover decision by the board. Whereas industry-adjusted returns are consistently higher in magnitude than industry returns in affecting a forced turnover, the difference between the two is more pronounced in the case of value-weighted returns than equal-weighted returns.
27 VI. Robustness Checks
6.1 Hypothesis 1
As a means of checking for robustness, tables 2 to 4 are replicated using total (aggregated) turnover as the dependent variable instead of forced turnover. Otherwise the specifications are identical to those presented in the results. Total turnover sums up forced and voluntary turnover. Whereas voluntary turnover is likely to stem from different reasons than forced turnover, the overall results should nevertheless be similar for total turnover as in the case of forced turnover.
Table 5 is thus identical to table 2 with the exception of total turnover as the independent variable. On a general level, the significance and sign of all coefficients in all models is identical with the exception of value-weighted industry returns when year fixed effects are considered. Here, this variable is still significant, but at the 5% level instead of the 0.1% level, as in table 2. Average marginal effects are also similar; as before a 1% increase in industry-adjusted returns have a greater (inverse) effect on the probability of a turnover than solely industry returns. Average marginal effects are listed in
completeness in the following passage.
In model [1] a 1% increase in industry returns is associated with a 3.41% decrease in the probability of a turnover. Similarly, a 1% increase in value-weighted industry-adjusted returns is associated with a 4.37% decrease in the probability of a turnover. It is notable that the probability of a turnover responds more greatly to industry return than in the case of a forced turnover (being nearly one percentage point higher). However, the effect of an industry-adjusted return on the probability of turnover is smaller than for a forced turnover, though by a lesser amount than the difference in industry returns with the two dependent variables (0.31 percentage points compared to 1.06).
Results are similar in model [2] considering equal-weighted returns instead. A 1% increase in industry returns is associated with a 2.83% decrease in the probability of a turnover. Similarly, a 1% increase in value-weighted industry-adjusted returns is associated with a 4.37% decrease in the probability of a turnover. The results here are closer to their forced turnover counterparts than in [1] with industry returns almost identical and industry-adjusted returns 0.39 percentage points lower.
In model [3], considering year fixed effects, a 1% increase in industry returns is associated with a 1.95% decrease in the probability of a turnover. Conversely, a 1% increase in value-weighted industry-adjusted returns is associated with a 4.50% decrease in the probability of a turnover. In this model the
28
magnitude of industry returns is reduced, whereas the impact of industry-adjusted returns in affecting a turnover is more pronounced.
Meanwhile, in model [4], the magnitude of equal-weighted returns is slightly more pronounced again compared to [3]. Here, a 1% increase in industry returns is associated with a 2.57% decrease in the probability of a turnover. Conversely, a 1% increase in value-weighted industry-adjusted returns is associated with a 4.52% decrease in the probability of a turnover. These figures are again slightly lower than in the forced turnover models. However, taking into account voluntary turnovers signifies that the general pattern is similar as before, but not as pronounced. It is also noteworthy that adding year fixed effects tends to diminish especially the magnitude of industry returns, whereas the proportional change of industry-adjusted returns does not alter by much.
Thus overall the results of table 5 point out to the significance of the returns of peers and outperforming the industry in reducing the probability of a turnover, just as in the case of forced turnovers. Boards place importance not solely on firm performance, but relative performance as benchmarked in comparison to industry standards. However, it also helps to be in a booming industry rather than an industry experiencing a downturn. To elaborate, the stage of the business cycle the industry is in is also critical.
<<Insert Table 5 Here>>
6.2 Hypothesis 2 (1) & (2)
Table 6 mimics table 3 with, again, total turnover as the dependent variable. Models [1] and [3] are identical in sign and level of significance as before. However, models [2] and [3] (with a sample of firms possessing a KZ value less than the median), while identical in sign, have industry return coefficients (both value and equal-weighted), which are significant only at the 5% level. Industry-adjusted returns, on the other hand, are the same as before being significant at the 0.1% level. This difference is notable (though still all variables are significant), as in the case of forced turnovers the difference between the high and low KZ samples was non-existent.
In model [1] a 1% increase in value-weighted industry returns is attributed with a 4.31% decrease in the probability of a turnover, whereas an identical increase in industry-adjusted returns is attributed with a 5.29% decrease in the probability of a turnover. Both values are higher than in the case of the forced
29
turnover specification, especially so for industry returns, which here are just shy of one percentage point under industry-adjusted returns.
Meanwhile in model [2] a 1% increase in industry returns is attributed with just a 2.21% decrease in the probability of a turnover, whereas an identical increase in industry-adjusted returns is attributed with a 3.19% decrease in the probability of a turnover. These results are for a sample with low KZ values, and are considerably less than in the case of the sample with only firms possessing high KZ values.
Subsequently, in model [3] a 1% increase in equal-weighted industry returns is attributed with a 3.49% decrease in the probability of a turnover. Conversely, a 1% increase in industry-adjusted returns is attributed with a 5.52% decrease in the probability of a turnover.
Finally, in model [4] a 1% increase in equal-weighted industry returns is attributed with a 1.90% decrease in the probability of a turnover. Conversely, a 1% increase in industry-adjusted returns is attributed with a 3.01% decrease in the probability of a turnover. The results are again considerably lower than in the case of only high KZ firms.
A seemingly unrelated estimation is also done for total turnover. The results are insignificant, as in the case of forced turnover, further cementing that there is not a significant difference between highly and lowly financially constrained firms. However, it is interesting to note that, while not significant the results are much stronger than in the case of only a forced turnover. Namely, using value-weighted
industry returns the χ2 value is 3.93 (Prob > χ2 = 0.1401). Using equal-weighted industry instead yields
a χ2 value of 4.41 (Prob > χ2 = 0.1103).
<<Insert Table 6 Here>>
6.3 Hypothesis 2 (3)
Table 7 is the final table to be replicated with total turnover to check for robustness. It is based on table 4. Most of the results in this table parallel closely those in table 4 with one exception. When adding year fixed effects the coefficient of the value-weighted industry returns is not significant at any level. This is somewhat surprising as the same coefficient is significant at the 0.1% level in the identical model without year fixed effects. This coefficient was also consistently significant in all other tables.
30
Here again the interaction term between the industry returns and the high KZ dummy is insignificant in all models. A description of average marginal effects follows.
In model [1] value-weighted industry returns and industry-adjusted returns are significant at the 0.1% level, whereas the interaction term is not. The sign of all coefficients is negative. It is estimated that a 1% increase in the industry returns will decrease the probability of a turnover by 3.25%, whereas an identical increase in industry-adjusted returns will decrease the probability of a turnover by 4.37%. Similar findings are present in model [2], which like model [1] does not consider year fixed effects. The model predicts that, of the statistically significant terms, a 1% increase in the equal-weighted industry returns will decrease the probability of a turnover by 2.90%, whereas an identical increase in industry-adjusted returns will decrease the probability of a turnover by 4.38%.
Moving on to adding year fixed effects in model [3] produces some striking results, at least for value-weighted industry returns. The term is not statistically significant, which is somewhat contradictory since it has exhibited high statistical significance before, and in the following model equal-weighted industry returns is also highly statistically significant. Only industry-adjusted returns are described here as statistically significant (but at the 1% level instead of the 0.1% level), with a 1% increase being associated with a 4.50% decrease in the probability of a turnover.
Finally in model [4] the average marginal effects are as follows. A 1% increase in equal-weighted industry returns is attributed with a 2.57% decrease in the probability of a turnover. Conversely, a 1% increase in industry-adjusted returns predicts a 4.52% decrease in the probability of a turnover. Thus overall the results are similar for total turnovers as they are for forced turnovers. There are some differences in the magnitudes of the coefficients, but in virtually all cases the same levels of statistical significance are maintained. The proportional change of industry-adjusted returns is consistently greater than that of industry returns (whether value or equal-weighted). There is also no ascertainable difference in the probability of a turnover between highly financially constrained firms and lowly financially constrained firms.
31 VII. Conclusion
This paper has looked at forced (and aggregated) CEO turnover in the United States for the period 1993-2013, considering how it is affected by industry performance, relative performance, and financial constraints. As in Jenter & Kanaan (2014) industry and industry-adjusted returns are significant and negative in affecting the probability of a forced (and aggregated) turnover. However, whether a firm is highly constrained or unconstrained is not significant in altering the probability of a forced CEO turnover. In other words, there is no significant difference in this regard between financially
constrained and non-constrained firms. The results hold true when considering both value-weighted and equal-weighted returns.
Industry-adjusted returns have a greater proportional impact upon the probability of a (forced) turnover than industry returns. Using equal-weighted returns tends to yield slightly higher results for especially industry returns than when using value-weighted returns. However, arguably value-weighted returns better take into account the size of companies and present a more realistic picture of the aggregated market.
Boards of directors clearly do seem to factor in exogenous elements when making turnover decisions instead of filtering them out. Industry returns, which are outside of control of CEOs, affect job security, as in Eisfeldt & Kuhnen (2013). Industry based turnover is anticyclical; more CEOs are fired in bad times than good. Additionally performance of the firm is not enough, even though positive, if it falls short of its peers. Relative performance matters; a firm must outperform its industry if the CEO desires to uphold its position. This follows Kaplan & Minton (2008). Relative performance is also more critical than industry performance, though both matter in a turnover decision.
Finally, though financial constraints may impede upon the ability of a firm to benefit from all
investment opportunities, this nevertheless should not have a significant impact upon the probability of a turnover. Rather, exogenous elements in the form of peer performance and endogenous elements insofar as how the CEO is managing the company matter in affecting the turnover decision. CEO skill and luck thus seems to matter more than a firm’s financial constraints.
32 References
Bushman, R., Dai, Z. & Wang, X. (2010) Risk and CEO turnover. Journal of Financial Economics, Vol. 96, Issue 3, pp. 381-398
Dimand, R. (2014) The Palgrave Macmillan James Tobin, Palgrave Macmillan, London
Eisfeldt, A. & Kuhnen, C. (2013) CEO Turnover in a Competitive Assignment Framework. Journal of Financial Economics, No. 109, pp. 351-372
Eisfeldt, A. & Rampini, A. (2008) Managerial Incentives, Capital Reallocation, and the Business Cycle. Journal of Financial Economics, Vol. 87, pp. 177-199
Faleye, O., Hoitash, R. & Hoitash, U. (2011) The Costs of Intense Board Monitoring. Journal of Financial Economics, Vol. 101, pp. 160-181
Fama, E. & French, K. (1997) Industry Costs of Equity. Journal of Financial Economics, Vol. 43, pp. 153-193
Fazzari, S., Hubbard, R. & Peterson, B. (1988) Financing Constraints and Corporate Investments, Brookings Papers on Economic Activity 1988, pp. 141-195
Jenter, D. & Kanaan, F. (2014) CEO Turnover and Relative Performance Evaluation. NBER Working Paper No. 12068
Kaplan, S. & Minton, B. (2008) How Has CEO Turnover Changed? NBER Working Paper No. 12465
Kaplan, S. & Zingales, L. (1997) Do Investment-Cash Flow Sensitivities Provide Useful Measures of Financing Constraints? The Quarterly Journal of Economics, February 1997
33
Malmendier, U. & Tate, G. (2005) CEO Overconfidence and Corporate Investment. The Journal of Finance, Vol. 60, Issue 6, pp. 2661-2700
Mikkelson, W. & Partch, M. (1997) The Decline of Takeovers and Disciplinary Managerial Turnover. Journal of Financial Economics, Vol. 44, pp. 205-228
Peters, F. & Wagner, A. (2014) The Executive Turnover Risk Premium. The Journal of Finance, Vol. 69, No. 4, pp. 1529-1563
Schoar, A. & Zuo, L. (2011) Shaped by Booms and Busts: How the Economy Impacts CEO Careers and Management Styles. NBER Working Paper No. 17590
Weisbach, M. (1988) Outside Directors and CEO Turnover. Journal of Financial Economics, Vol. 20, January-March 1988, pp. 431-460
34 APPENDIX
Table 1
The following table provides summary statistics on CEO turnover. Panel A gives a brief overview of the sample; it depicts the size of the sample, and the number plus percentage of forced, voluntary, and aggregated turnovers, respectively. Panel B illustrates firm characteristics by CEO turnover outcome (forced, voluntary, and no
turnover). It gives average and median values of various book values of the firm. Also all individual components of the KZ index are depicted. Panel C gives a picture of firm and industry performance by CEO turnover
outcome.
Panel A: Frequency of voluntary and forced CEO turnovers
Number of firm-years Number of forced CEO turnovers Number of voluntary CEO turnovers Percentage of firm-years with at
least one CEO turnover Percentage of firm-years with at least one forced CEO turnover Percentage of firm-years with at least one voluntary CEO turnover 34.878 926 2.352 9,40 % 2,65 % 6,74 %
Panel B: Firm characteristics by
CEO turnover outcome
Turnover type
Forced Voluntary None
Mean Median Mean Median Mean Median
Total assets ($m) 13.477 1.051 9.856 1.685 10.011 1333
Total liabilities ($m) 10.858 561 7.105 889 7.469 687
Stockholder's equity ($m) 2.570 410 2.668 659 2.470 548
Income before extraordinary
items ($m) 109 3,88 334 61 318 55
Cash flow ($m) 352 37 622 122 558 106
Tobin’s Q 1,58 1,23 1,90 1,42 1,97 1,49
Financial leverage 0,69 0,34 0,35 0,34 0,33 0,32
Dividends ($m) 109 0 149 9,24 117 4,35
Cash & st investments ($m) 1950 82 926 96 1088 99
Capital ($m) 1693 175 2254 331 1.869 234
35
Panel C: Firm and industry performance by CEO turnover
outcome
Turnover type
Forced Voluntary None
Mean Median Mean Median Mean Median
Stock return in the 12 months prior to a CEO turnover -12,78 % -16,54 % 11,90 % 5,84 % 19,34 % 11,90 % [S.E.] [-1,55%] [1,22%] [0,38%] VW industry stock return in the 12 months prior to a CEO turnover 8,65 % 10,68 % 10,47 % 11,01 % 12,33 % 12,72 % [S.E.] [0,83%] [0,52%] [0,15%] EW industry stock return in the 12 months prior to a CEO turnover 11,03 % 9,36 % 14,12 % 12,78 % 15,45 % 13,18 % [S.E.] [1,07%] [0,70%] [0,21%]
36 Table 2
This table illustrates how the probability of a forced turnover is affected by industry and relative performance for the entire sample from 1993-2013 using annual data. Industry performance in encapsulated by 12 month rolling industry returns, whereas relative performance uses the difference between 12 month rolling firm returns and 12 month rolling industry returns (which is the industry-adjusted return). Industry and firm returns are calculated separately using value-weighted returns (VW) for [1] & [2] and equal-weighted returns (EW) for [3] & [4]. Models [3] and [4] take year fixed effects into account. Robust t-statistics are reported in parentheses. All standard errors are clustered at firm level. *, **, and *** indicate significance of 5%, 1%, and 0.1%, respectively.
Dependent variable: Forced turnover dummy
[1] [2] [3] [4] VW industry return -0,398*** -0,408*** (-6,51) (-4,30) VW industry-adjusted return -0,794*** -0,799*** (-11,54) (-11,13) EW industry return -0,474*** -0,547*** (-8,29) (-6,45) EW industry-adjusted return -0,788*** -0,800*** (-11,90) (-11,27) Constant -1,951*** -1,946*** -2,152*** -2,168*** (-105,86) (-107,61) (-21,54) (-21,42) Observations 34.878 33.154 33.365 33.154
