Compensation benchmarking and the
influence of hedge fund activism
August 15, 2017
Abstract
Using a sample of firms listed on the S&P 1500 index between 2007 and 2015, the effect of shareholder activism on compensation benchmarking practices is examined using hand collected data from the SEC’s DEF 14A filings. The results of a multivariate
difference-in-differenceapproach show that shareholder activism decreases the mean CEO compensation of the peers in the compensation peer group. These results are verified using a propensity score matching approach. To rule out a potential pay trend in the compensation peer group, a test is performed to study the effect of shareholder activism on compensation peer turnover. The results show an increase in compensation peer turnover as a result of shareholder activism. By studying the influence of shareholder activism on the compensation peer selection process, a new perspective is offered on existing literature on rent-seeking behavior in compensation benchmarking practices.
Master Thesis
Name Laurens Hesse
Student nr. 10265759 Supervisor Dr. T. Jochem Specialisation Corporate Finance
University University of Amsterdam, Amsterdam Business School Faculty Faculty of Economics and Business
Statement of Originality
This document is written by Student Laurens Hesse 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.
Acknowledgements
The past few months have been intensive, but today is the day I finally hand in my master thesis. Handing in this last assignment indicates a personal end of an era. During the last few months I have learned a lot, both in terms of academics as well as on a personal level. I would like to use this opportunity to thank those who supported me during the last few months.
First, I would like to thank dr. Torsten Jochem for the inspiring lectures and tutorials on Corporate Governance, leading to the research topic of choice. I am grateful for your time and all the feedback you provided. Also, I would like to thank you for handing me the tools which allowed me to collect the required compensation peer data. Collecting the required compensation peer data was time consuming, but would have been impossible without the tools you provided.
Further, I would like to thank Arthur van Eeden and Stijn Smit. These fellow students and true friends have spent time discussing statistical approaches used in my thesis. I would like to thank them for their views and advice.
Last, I would like to thank dr. Jonathan Cohn, dr. Stuart Gillan and dr. Jay Hartzell for their help on retrieving a complete SharkWatch50 list.
Contents
1
Introduction
5
2
Literature review
6
2.1
Compensation Benchmarking
. . . .
6
2.2
Shareholder Activism
. . . .
12
3
Data
15
4
Methodology
17
4.1
Hypotheses . . . .
17
4.2
Models and dependent variable description
. . . .
19
4.3
Description explanatory variables
. . . .
22
4.4
Robustness tests
. . . .
23
5
Empirical Evidence
23
5.1
Results
. . . .
24
5.2
Robustness tests
. . . .
28
6
Limitations
33
7
Conclusion
34
References
37
1
Introduction
Between 1978 and 2014, the top 350 U.S Chief Executive Officer (CEO) inflation-adjusted compensation increased from $1.5 million to $16.3 million (Mishel & Davis, 2015). This 997% increase in CEO compensation is almost double the growth of the S&P 500 in the same time period. During that period, the average typical worker’s wage increased 10.9%. As a result of this unequal growth, the average CEO’s compensation is 300 times larger than that of the typical worker (Mishel & Davis, 2015).
The development of the compensation of U.S CEOs is highly debated. There are different theories to explain the increase in CEO compensation. The first theory is the “optimal contracting view”. Bebchuk & Fried (2003) explain the optimal contracting view as a possible remedy to the agency problem. According to this theory, boards construct manager incentive schemes in such way that shareholder wealth is maximized. The “managerial power view” takes a different approach. This theory does not only acknowledge the use of compensation to address the agency problem, but also sees compensation as part of the agency problem itself (Bebchuk & Fried, 2003). Various authors have studied if this increase in CEO compensation is explained by compensation benchmarking. These authors agree that the variation in CEO compensation is largely explained by the composition of compensation peers. However, there is an ongoing debate whether the observed increase in CEO compensation is caused by rent-seeking behavior in the compensation peer selection process or to attract, retain and reward managerial talent.
This study will contribute to the ongoing discussion by researching the relationship between shareholder activism and the compensation peer selection process. Prior studies have shown that the influence of activist shareholders affects firm policy. The observed shareholder activism effects include firms adopting new governance structure resolutions proposed by activist share-holders (Smith, 1996) and reducing CEO compensation (Hartzell & Starks, 2003; Brav et al., 2008; Klein & Zur, 2009). First, this research studies the effect of shareholder activism on the level of CEO compensation using a multivariate regression with a difference-in-difference setup. The obtained results show a negative relationship between shareholder activism and the level of CEO compensation. This result is verified using a propensity score matching approach (PSM) in order to increase causal evidence. As the results of Hartzell & Starks (2003), Brav et al. (2008) and Klein & Zur (2009) are verified, the observed results of the first regression functions as an introduction to study if activism causes a decrease in CEO compensation through
This relationship is studied by testing the effect of shareholder activism on the mean CEO compensation of peers in the compensation peer group, using a multivariate regression with a difference-in-differencesetup. The primary dataset includes hand collected data on compensation peers, retrieved from the SEC’s DEF 14A filings. The documented results provide evidence of shareholder activism causing a decrease in the mean CEO compensation of the compensation peers. To increase causal evidence, a second PSM test is performed which verifies the negative relationship between shareholder activism and the mean CEO compensation of the compensation peers. To rule out potential trends in CEO compensation, an additional test is performed where shareholder activism is tested on compensation peer turnover. The documented evidence shows that compensation peer turnover increases as an activist shareholder enters a firm. The results in this research provide evidence indicating that shareholder activism causes existing relatively high paid compensation peers to be replaced by less paid compensation peers. By studying the influence of shareholder activism on the compensation peer selection process, new insight can be given on existing literature that studied potential rent-seeking behavior in compensation benchmarking practices.
This study will first cover existing literature on compensation benchmarking and shareholder activism in Section 2, followed by the description of the data used in Section 3. The methodology is covered in Section 4. The empirical results and implications are discussed in Section 5. The limitations and conclusion of this study are presented in Section 6 and Section 7, respectively
2
Literature review
Section 2 will elaborate on existing literature on compensation benchmarking and shareholder activism. Literature on compensation benchmarking is studied in search of shareholder ac-tivism related variables. Literature on shareholder acac-tivism is studied to research the effect of shareholder activism on compensation practices. Subsection 2.1 elaborates on compensation benchmarking studies, followed by Subsection 2.2 which discusses literature on shareholder activism.
2.1
Compensation Benchmarking
Existing literature shows an ongoing debate regarding different views on the use of compensation benchmarking. The first view that is discussed describes rent-seeking behavior of top executives
in the compensation peer selection, which drives executive compensation upwards. However, the second view states that the compensation peer selection is based on economic factors in the managerial labor market.
Faulkender & Yang (2010) are the first to show that variation in CEO compensation is explained to a large extent by the composition of the compensation peers of firms. These authors research the role and the composition of compensation peers. Variables such as industry, size and the level of compensation of potential compensation peers are found to have a significant effect on the compensation peer selection. To justify high CEO compensation, firms are likely to choose highly paid peers. Faulkender & Yang (2010) find that this effect is in particularly strong when peer groups are smaller, when the CEO also serves as chairman on the board of directors, when the CEO has long tenure and when the board members are busier if they serve on the board of multiple companies. Using a propensity score matching approach, the authors show that the median firm in their sample chose a median peer that paid their CEO $470 thousand more than the non-selected, best-matched peer. The estimated difference in payment for the median firm implies a 3.3% to 4.5% annual increase of CEO pay. Bias towards highly paid compensation peers contributes to the increase of CEO pay over time. Moreover, firms could further increase CEO pay by selecting compensation peers that could not be justified based on the relative size of the firm. Faulkender & Yang (2010) illustrate this concept with an example: a firm that is ranked in the bottom quartile based on firm size, benchmarks against firms with median pay. This could potentially further increase CEO pay. Faulkender & Yang (2010) expect that due to increased transparency, through greater disclosure of compensation peer group members, should lead to increased shareholder and stakeholder analysis on the compensation peer selection process.
Using a similar methodology, Bizjak et al. (2011) study to which extent compensation benchmarking is used to inflate the CEO’s compensation level or whether the peer group selection is used to provide the board with a relevant market wage applicable to the CEO. The authors find that, on average, companies base their compensation peer group selection on economic factors that reflect managerial labor market conditions. These compensation peer groups contain peers from the same industry, are comparable in size and scope and show other commonalities related to labor. However, if firms deviate from the economic model for peer selection, firms tend to select higher paid and larger compensation peers. This effect tends to get stronger for firms that are smaller and less visible. Due to less visibility, management has the ability to use even more discretion in the compensation peer selection. Also, insufficient
evidence is found to relate the compensation peer group bias to weak corporate governance. Finally, Bizjak et al. (2011) show that the 2006 SEC peer group disclosure regulation has reduced the bias in peer group selection over time.
In the follow-up research of Faulkender & Yang (2012), the main objective is to determine if opportunistic benchmarking behavior documented in existing literature has continued after the mandate of the SEC to disclose a list of compensation peers in the firm’s proxy statements (Faulkender & Yang, 2010; Bizjak et al., 2011). The first research method used is a multivariate probit regression. In the authors’ model, the peer selection process is determined by various economic factors that capture similarities between a firm and its potential peer along with dimensions that are relevant for the CEO labor market: performance, industry, size, risk, visibility and CEO responsibility. Compared to prior research, Faulkender & Yang (2012) find similar results. Industry overlap and similarity in size are found to be important factors explaining choice of compensation peers. Also, authors find that compensation peers are significantly more likely to be included as compensation peer as potential peers have similar compensation structures, performance, risk and index notations. However, the authors are most interested if companies include highly-paid peers into their compensation peer groups. Therefore, the compensation peers’ total CEO compensation in the prior year is added to the multivariate regression. After controlling for similarities between the firms, the authors show that firms have continued to include higher-paid CEOs to their compensation peer groups. In economic terms, a one standard deviation increase in CEO compensation increases the likelihood of a potential compensation peer to be selected, for a company of similar size but different industry, from 5.85% to 7.84%. Furthermore, other results indicate that opportunistic behavior in peer selection has intensified over time. To validate the results of the multivariate probit analysis, Faulkender & Yang (2012) use a propensity score matching (PSM) approach in order to match each company to the closest unselected potential peer based on the CEO’s total compensation of the selected firm and that of the matched unselected firm. The difference in CEO total compensation between the actual chosen peer and its propensity score matched unselected peer. This difference is defined as the pay gap. Faulkender & Yang (2012) find that the pay gap is statistically different from zero and increases over time, validating the results from the multivariate probit analysis. Last, Faulkender & Yang (2012) study the effects how corporate governance affects the pay gap. An interesting measure used is the ownership structure. Hartzell & Starks (2003) show the importance of institutional investors and their monitoring role, especially in disciplinary
actions in compensation practices. Therefore, Faulkender & Yang (2012) expect that firms with high institutional holdings respond differently to the SEC’s compensation peer disclosure rule. Comparing the pay gap between samples of high and low institutional ownership, based on the sample median, the authors find that the peer selection bias is not eliminated for companies with high institutional holdings, but the bias does seem to be restrained from worsening over time.
The change in CEO compensation of firms joining the S&P 500 index and the role of com-pensation benchmarking is studied by Colak et al. (2017). These authors observe a significant increase in executive compensation after a firm is added to the S&P500 index. The authors find that the increase in executive compensation cannot be fully explained by weak corporate gover-nance or firm performance. The found increase in executive compensation, after a firm is added to the S&P 500, primarily takes the form of increased granted options. However, the cash-related part of executive compensation (cash and bonuses) also increases. The authors hypothesize that the increase in executive compensation results from compensation benchmarking practices. Supporting the hypothesis, Colak et al. (2017) find that after a firm is added to the S&P 500 index, it will include more relatively higher paid S&P 500 index firms in its compensation peer group. Moreover, the authors provide evidence that once a firm is added to the S&P 500 index, other firms of the index in the same industry as the newly added firm will include this firm in its compensation peer group. Colak et al. (2017) argue that this causes greater compensation levels of S&P500 index firms in that industry. However, when a firm is deleted from the S&P 500 index, the compensation level of firms in the same industry does not decline. Colak et al. (2017) conclude that compensation benchmarking has facilitated compensation contagion and that membership of the S&P500 index can induce an economy-wide artificial increase in executive compensation.
In contrast, Bizjak et al. (2008) find that compensation peer selection is not used to increase CEO pay due to rent-seeking behavior but rather for retention reasons. The authors show that the use of compensation peer groups is common practice in the U.S. They find that CEOs who receive pay below the median of their compensation peers receive substantial larger raises in comparison to CEOs who receive pay above the peer group median value. This effect seems to persist, even after controlling for factors that have been shown to affect compensation such as market and accounting performance, firm size and CEO tenure. Bizjak et al. (2008) note the criticism on compensation benchmarking regarding rewarding CEOs independently from firm performance. An alternative view is that compensation benchmarking could provide an
efficient approach to determine the reservation wage of the CEO and is a vital component to the compensation process. Bizjak et al. (2008) find evidence that the likelihood of CEOs that receive pay below the peer group median value, receive a pay increase that shifts the CEOs pay above the median value that is correlated with firm performance and used proxies for tighter labor markets. This finding is not systematically associated with weak corporate governance. Furthermore, the authors find that CEOs with pay below the median peer group value have significant risk in losing their job, which is not in line with the view that CEOs have captured the pay process.
Albuquerque et al. (2013) continue the research on compensation benchmarking, while taking into consideration the findings of Faulkender & Yang (2010), Bizjak et al. (2008) and Bizjak et al. (2011). Albuquerque et al. (2013) take a different perspective and hypothesize that the difference between CEO pay of selected peers and propensity-score-matched firms (“peer pay effect”) does not reflect self-serving behavior but is a reward for CEO talent, which captures successful managing a complex organization and achieving high levels of performance. Multiple methods using proxies are used to disentangle whether firms’ tendency to selects peer that are highly paid reflects self-serving behavior or if firms pay for CEO talent. The fitted value of a regression of peer pay effect on proxies for CEO talent is used as a proxy for talent. Following prior literature, Albuquerque et al. (2013) use the CEO’s historical abnormal returns and accounting performance, the market value of firms that the CEO has managed in the past and the amount the CEO is referred to in the business press as a proxy for talent. The authors proxy self-serving behavior by the fitted value of a regression of peer pay effect on proxies for weak corporate governance. The proxies for weak corporate governance are: board structure, number of anti-takeover provisions and ownership concentration. Albuquerque et al. (2013) find evidence that the peer pay effect can be mostly assigned to the need to pay CEOs more for their talent. The authors do find some evidence in favor of self-serving behavior, but this has less economic importance in explaining the peer pay effect than CEO talent. These results support the view that peer benchmarking on CEO pay is more consistent tight labor markets than with managerial entrenchment or weak corporate governance. In contrast with Faulkender & Yang (2010), Bizjak et al. (2011) and Colak et al. (2017), the results of this research provide evidence in favor of an efficient contract perspective.
The short to be published article of Denis et al. (2017) documents firms’ reactions to weak ”Say-on-Pay” (SoP) votes of companies in their compensation peer group. Prior researchers,
such as Ferri & Maber (2013), tried to answer a similar research question by studying how firms react when they themselves receive a weak SoP vote. However, the results in the research of Ferri & Maber (2013) are subject to bias as companies that receive a weak SoP vote often experience poor firm performance. As a result, boards could potentially change CEO compensation, even in the absence of a weak SoP vote. By studying firms’ reaction to weak SoP votes of companies in their compensation peer group, Denis et al. (2017) research is not subject to the above mentioned form of bias and can add to existing evidence on the impact of SoP and the role of compensation benchmarking. The results of Denis et al. (2017) show that when at least 10% of the self-selected compensation peers encounter a weak SoP vote, the members of the compensation committee respond by significantly reducing the compensation of its own CEO relative to control firms that do not experience a weak SoP vote at companies in their compensation peers group. This result is observed despite the fact that the firms with weak-vote compensation peers do not experience a weak SoP vote nor poor firm performance themselves. These firms are labeled ”primary firms”. The reduction of CEO compensation of primary firms is concentrated at those companies that have above median excess level CEO compensation. Further, the reduction of CEO compensation of these primary firms align pay-for-performance sensitivity of primary firms to that of control firms. These results also contribute to the debate if firms use compensation benchmarking for informational or opportunistic purposes. Denis et al. (2017) interpret the above described results in favor of informational benchmarking as boards have shown to reduce CEO compensation in response to a weak SoP of their compensation peers even though these firms do not experience a weak SoP vote nor poor firm performance themselves. Consistent with the informational benchmarking, primary firms have not been found to drop weak vote SoP compensation peers. However, a subsample of primary firms that are expected to have selected their compensation peers opportunistically do not show a decrease in CEO compensation after a compensation peer received a weak SoP vote.
In another short to be published article of Choi et al. (2017), presented at the 2017 annual meeting of the European Finance Association, discusses the impact of the 2006 SEC compensa-tion peer group disclosure rule. The main findings of their research include that CEO departure increases at companies with a higher number of peer citations and as peer citations increase, the CEO income increases with the equity based portion of total compensation is greater. Using an OLS model, the authors find that the mean CEO compensation increases by $680,000 when the number of citations increase by one unit. Similar significant results are observed using a
difference-in-differencemodel. Further, Choi et al. (2017) observe that the CEO compensation is the highest when firms are cited by relatively larger firms. The above mentioned results are robust to a difference-in-difference analysis where a subsample is tested of founder and non-founder CEOs around the 2006 SEC regulation change, providing more evidence of a causal relationship. Summarized, the documented results of Choi et al. (2017) give insight in the post 2006 SEC disclosure regulation regarding greater labor market transparency and the significant impact on executive compensation and retention incentives.
The articles mentioned above study the determinants of compensation benchmarking. Var-ious control variables, but nothing was to be learned on how activist shareholders affect the practice of compensation benchmarking. Faulkender & Yang (2012) have found interesting results how the level of institutional ownership affects the pay gap. However, Faulkender & Yang (2012) do not distinguish between institutional investors. Hartzell & Starks (2003) describe the difference between an activist block holder and an institutional investor. According to these authors, activist block holders have a clear intent to influence the decision-making process. This thesis will contribute to the literature by making a clear distinction how shareholder activism influences the selection of the compensation peer group.
2.2
Shareholder Activism
In the next section, existing research on the effect of shareholder activism on firm policy is discussed. Existing literature shows little conflicting results.
Smith (1996) studies firm characteristics that lead to shareholder activism and analyzes how activism affects the target firm’s shareholder wealth, governance structure and operating performance. The sample contains firms that are targeted by the California Public Employees’ Retirement System (“CalPERS”) between 1987-1993. According to the authors, CalPERS is regarded as a leader in activism in the U.S. If no effects are found using this sample, no other effects are expected to be found for other activists. Smith (1996) research shows that firm size and level of institutional ownership affect the probability that a firm is targeted by an institutional investor, after controlling for prior stock performance. During the time period of the research, 72% of the targets adopted new governance structure resolutions proposed by CalPERS or made changes sufficient to settle with CalPERS’ demands. Further findings include a significant stock price reaction when CalPERS successfully targeted a company whereas a negative stock price reaction resulted for an unsuccessful event. Increase in operating performance was not found to
be significantly different from the control group. Overall, the results from Smith (1996) research show that shareholder activism is successful in changing the governance structure of the targets and shareholder wealth is improved significantly. However, operating performance is not found to be significantly improved.
Other evidence on the effect of institutional activism is given by Hartzell & Starks (2003). These authors research the relationship between institutional ownership and compensation of firm’s executives. Their results show a show a strong positive relationship between the share of institutional ownership and the sensitivity of pay-for-performance of managerial compensation. Hartzell & Starks (2003) also find a negative relationship between institutional investor con-centration and the level of executive compensation. The authors control for firm size, industry, recent performance and investment opportunities. An increase of one standard deviation in the percentage of institutional holdings of the largest five institutional investors corresponds with a decrease in salary of 12% of the sample mean, and a decrease of 19% of total compensation. This corresponds with the hypothesis that institutional investors influence compensation practices.
Brav et al. (2008) find that shareholder activism reduces the CEO’s salary whereas performance-based-pay increases. These authors are the first to examine a hedge fund activism while using a large-scale sample between 2001 and 2006. The authors find a positive market reaction after announcement of hedge fund intervention. It is shown that the market reaction is consistent with improved post-intervention performance of the target, changes in payout policy and effect of interventions regarding the CEO pay. The authors find that the CEO pay post hedge fund activism is $1.21 million lower relative to the matched sample that did not experience hedge fund activism whereas pay-for-performance has increased.
The findings of Brav et al. (2008) are consistent with the view that agency costs can be reduced by informed shareholders monitoring at the firms that have been targeted. Unlike institutional investors, hedge fund activists are personally highly financial incentivized to increase the value of firms in their portfolio (Rock, 1990). Furthermore, hedge fund activists are not limited by regulatory or political barriers that do limit other investors. In general, hedge fund activism creates value. The value created does not originate from good stock picking, but because of the strong personal commitment of the hedge fund managers. The strong presence of hedge funds and the threat of disciplinary intervention puts pressure on the management of public firms to prioritize shareholder value.
campaigns of hedge funds and other private investors, where the latter consists of individuals, private equity funds, venture capitalist funds and asset management funds. The authors define confrontational entrepreneurial activism as: “instances in which an investor files a 13D filing after taking an initial stake of 5% or more in the company and clearly states in the filing’s “purpose” section that it intends to proactively influence management’s future decisions.” (p.225). The similarities between the two samples include market returns and the activist’s ability to accomplish its original objectives. The results of Klein & Zur (2009) indicate that, on average, activist hedged funds earn a 10.2% abnormal return around the initial 13D filing announcement whereas other activists earn a 5.1% abnormal return. Interestingly, these returns do not dissipate within a 1-year period after the initial 13D filing. In contrary, targets of activist hedge funds and other activist investors earn an abnormal return of 11.4% and 17.8% respectively within the first year of initial filing. The authors present evidence that these results are based on the, actual and threatened, use of activist proxy contests. There are also distinct differences between activist hedge funds and other entrepreneurial activists. According to Klein & Zur (2009), activist hedge funds target more financially healthy, high cash holding and profitable firms than other entrepreneurial activists. Hedge funds demand that the target initiates buying back company shares, reduce CEO pay and increase dividends. In contrast with hedge funds, other entrepreneurial activists frequently demand changes in operating strategies. Klein & Zur (2009) conclude that their results reinforce findings in prior studies (Smith, 1996; Brav et al., 2008), but contribute by examining the differences between the two types of activism that are researched in their study.
However, the beliefs in the effectiveness of activist hedge funds are somewhat controversial. Bebchuk et al. (2015) focus on the claim that activist shareholders are “myopic”. According to this claim, activist shareholders have short investment horizons and push for profitable actions in the short run but which are harmful in the long run for the company and other investors. Also, this claim accuses activist investors of “pump-and-dump” activities. According to these activities, negative abnormal long-term returns follow when activist hedge funds exit a target company. Last, according to the claim hedge activist shareholders left firms in a vulnerable position just before the financial crisis, contributing to the magnitude of the crisis. Bebchuk et al. (2015) claim that of all activist investors, especially hedge funds are accused of this. Studying 2.000 hedge fund interventions between 1993-2007, the authors do not find evidence in favor of the claim that activist hedge funds are myopic. In contrast, post-intervention long-term
improvements in performance, rather than declines, are observed. Bebchuk et al. (2015) do not find evidence that the initial positive spike of the stock price was followed by negative abnormal returns in the long-run. Also, the authors do not find evidence of “pump-and-dump” activities. Finally, Bebchuk et al. (2015) do not find evidence indicating that activist interventions left firms more vulnerable in the years preceding the financial crisis nor that target companies were more affected by the financial crisis.
The articles mentioned above have shown that shareholder activism has influenced firm policy. Most interesting, Brav et al. (2008), Hartzell & Starks (2003) and Klein & Zur (2009) find that CEO compensation is lower and pay-for-performance is higher at firms where an activist shareholder is present. These findings could indicate that activism leads to changes in the compensation peer group. In the next sections, the relationship between shareholder activism and changes in the compensation peer group is researched.
3
Data
The sample of firm used in this thesis contains data on S&P 1500 firms between January 2007 and December 2015. A useful setting is created due to the SEC’s disclosure ruling on August 29th, 2006 1. This disclosure rule requires all firms to list their compensation peers when compensation benchmarking is used to determine executive compensation. The S&P1500 index is chosen as it contains a wide variety of firms in terms of firm size and represents roughly 90% of the U.S market capitalization2. Still, all companies in the S&P 1500 index are covered by
important databases such as Execucomp, Compustat and CRSP. Before a subsample is taken of the S&P1500, companies with SIC codes ranging between 6000-6999 and starting with 49 are removed from the list. These industrial codes belong to financial companies and companies operating in the field of electric, water, gas and sanitary services. These industries are considered to be more regulated and are therefore removed from the sample. To ensure a true random sample, the S&P 1500 companies, as of January 1st, 2007, are placed in alphabetic order. Consecutively, every eighth firm is selected. To mitigate index survivorship bias, as described by Elton et al. (1996), the subsample of 132 initial S&P1500 firms determined on January 1st, 2007, will be use throughout this study. Meaning, the S&P 1500 index might change within the time frame of this research but the subsample created will not be changed if firms drop out or enter of
the index. Only firms that are acquired in M&A activity or that go bankrupt are removed from the sample.
The primary data set used in this research consists of hand collected firms that are included in the compensation peer group of the disclosing firms in the S&P 1500 subsample. Each company in this subsample is required to list its compensation peers. The companies in the compensation peer group are included in the SEC DEF 14A filings of the disclosing companies, which can be retrieved from the SEC’s EDGAR database. This hand collected data set is supplemented with measures for firm size, compensation, performance, leverage ratio, market-to-book ratio, a dummy variable indicating CEO chairman duality and CEO tenure. Data on these measures can be retrieved from Execucomp, Computstat and CRSP. Lastly, a dummy variable indicating the presence of at least one activist hedge fund is included. In order to create the activism variable, the owners of the companies in the subsample must be determined first using the Thomson Reuters Institutional Holdings database. Shareholders who own less than 1% are filtered out, providing a list of substantial shareholders of the companies in the S&P1500 subsample. As used in the study of Cohn et al. (2016), sharkrepellent.net’s ”SharkWatch50” list is used to identify activist hedge fund investors. Sharkrepellent.net defines the SharkWatch50 list as ”A compilation of 50 significant activist investors. SharkWatch50 is based upon a number of factors, including the number of publicly disclosed activism campaigns and the ability to elect change at targeted companies”. Brav et al. (2008) describe that unlike mutual funds and pension funds, hedge funds employ managers that have a strong personal financial incentive to increase the value of firms in their portfolio. Furthermore, hedge funds do not face the same regulatory and political barriers that pension funds or mutual funds face, which increases the effectiveness of hedge funds. These authors find that hedge funds generate value on average, not because they are good stock pickers, but as a result of strong commitment. Given these findings, the SharkWatch50 list is considered a valid proxy for shareholder activism and will be used to define activism throughout this study.
Table I shows the descriptive statistics of CEO compensation and firm characteristics of both disclosing firms and the firms in the compensation peer group in panel A and B, respectively. Panel A shows that the mean earned CEO salary of the disclosing firms is $890,000 and has a median value of $850,000. The difference between the median and mean of salary and bonuses shows larger differences compared to salary. The mean and median salary and bonuses are $2.19 million and $1.65 million, respectively. Total compensation has a mean of $6.01 million and a
median of $4.43 million. The summary statistics presented in Panel A are slightly lower but comparable to the findings of Faulkender & Yang (2010). The variation is most likely caused by the difference in sample selection as these authors only focus on S&P500 and S&P400 midcap firms whereas in this study a subsample of the entire S&P1500 is researched. Looking at the descriptive statistics in Panel B of Table I, the mean salary of the peer group is $950,000 and has a median of $910,000. The mean salary and bonuses of the peer group is $2.68 million and has a median of $2.29 million. Total compensation of the peer group has a mean value of $7.61 million and a median of $6.69 million. Comparing the mean and median compensation variables of the disclosing firms and the firms in the compensation peer group, it can be seen that the found values are similar. The size of the compensation peer group has a mean of 17.67 and median of 15. The size of the compensation peer group is comparable with Faulkender & Yang (2010), as these authors find a mean and median of 18.25 and 16, respectively.
4
Methodology
The aim of this thesis is to reveal if shareholder activism has a causal effect on the compensation peer selection process. This chapter will explain the methodology. Subsection 4.1 elaborates on the hypotheses. Subsection 4.2 presents the models used in this thesis and discusses the dependent variables. Subsection 4.3 reviews the explanatory variables. Subsection 4.4 discusses the robustness tests.
4.1
Hypotheses
Section 2 presented an overview of existing theories on compensation benchmarking and shareholder activism. The ongoing discussion on the increase in CEO compensation can be explained the optimal contract view and the managerial power view, as explained by Bebchuk & Fried (2003). Brav et al. (2008), Hartzell & Starks (2003) and Klein & Zur (2009) find that CEO compensation is reduced at companies which experience shareholder activism. Activism has shown to affect firm policy, indicating that shareholder activism could lead to changes in the structure of the compensation peer group. Furthermore, previous benchmarking studies have not yet studied the potential impact of shareholder activism.
Table I Descriptive statistics
Panel A: Descriptive statistics of disclosing firm
Mean Median Standard deviation Observations Salary (in millions of dollars) 0.89 0.85 0.41 1017 Salary and bonuses (in millions of dollars) 2.19 1.65 2.35 1017 Total compensation (in millions of dollars) 6.01 4.43 5.62 1017 Total sales (in millions of dollars) 7806.06 2113.40 16729.61 1028
ROA (%) 3.43 5.19 11.41 1028
Stock return (%) 10.93 5.46 79.19 1028
Leverage ratio 0.20 0.17 0.18 1026
Market-to book ratio 1.38 1.17 0.79 1026
CEO Duality 0.53 1.00 0.50 1017
CEO tenure 6.84 6.00 4.64 1017
Activism 0.20 0.00 0.40 1028
Panel B: Descriptive statistics of compensation peer group
Mean Median Standard deviation Observations Mean salary of peer group (in millions of dollars) 0.95 0.91 0.28 868 Mean salary and bonuses peer group (in millions of dollars) 2.68 2.29 1.52 868 Mean total compensation peer group (in millions of dollars) 7.61 6.69 4.36 868
Size 17.94 15.00 11.25 868
Compensation peer turnover (%) 17.67 3.33 32.13 732.00 Panel C: Sample data
Fiscal year Firms w/ activist Firms without activist Total firms
2007 29 103 132 2008 25 103 128 2009 21 102 123 2010 22 96 118 2011 25 91 116 2012 16 92 108 2013 19 86 105 2014 23 79 102 2015 26 69 95 Total sample 206 822 1028
Note: Table I shows the descriptive statistics. Panel A, Salary and bonuses equals Salary + Bonus + non-equity incentives. Total compensation (TDC1) is retrieved from the ExecuComp database. Firm characteristics are ROA, Stock return,Sales, Market-to-book ratio of assets and the Leverage ratio (total debt/market value assets). CEO Duality is a dummy variable which receives a value of one if the CEO is also the chairman of the company. Tenure is the number of years that the CEO has fulfilled the function as CEO of the company. Activism is a binary variable which receives value one if at least one Sharkwatch50 hedge fund owns a minimum of 1% of the shares outstanding of the disclosing firm. Panel B, Mean Salary, Salary and bonuses and Total compensation is the mean pay of the disclosing firm’s compensation peer group in the same year of the disclosing firm’s pay. Size is the number of peer companies in the disclosing firm’s compensation peer group. Compensation peer turnover is defined as the number of newly added peers in the contemporaneous period divided by the number of peers that have been used in the previous period and re-selected in the contemporaneous period, expressed as a percentage. Panel C shows the descriptive statistics on the annual number of firm observations. A distinction is made between firms that do and do not experience the presence of at least one activist hedge fund.
Based on these findings, the following hypotheses are tested:
Hypothesis 1 Shareholder activism decreases the median CEO compensation of companies in the compensation peer group
Hypothesis 2 Shareholder activism increases compensation peer turnover
The results obtained from this research contribute to the ongoing discussion between the optimal contracting view and the managerial power view by studying the role of shareholder activism in compensation benchmarking practices.
4.2
Models and dependent variable description
This subsection is used to elaborate on the first model, which is used as a stepping stone to the second model. Further, this subsection tests if the results of prior studies, indicating that shareholder activism causes a decrease in CEO compensation, can be verified (Hartzell & Starks, 2003; Brav et al., 2008; Klein & Zur, 2009). The first objective is to gain a better understanding on the effect of shareholder activism on the level of CEO pay. Using panel data, the level of CEO compensation is estimated among firms in the sample for which peer group information is available. Using firm and CEO characteristics that have been shown in prior literature (Core et al., 1999; Faulkender & Yang, 2010) and a binary variable indicating the presence of at least one activist investor, the level of CEO compensation is estimated. The following specification is used:
ln(CEOcompensationi,t) =β0+ β1Activismi,t−1+ β2ln Salesi,t−1+ β3ROAi,t+ β4ROAi,t−1
+ β5Stockreturni,t+ β6stockreturni,t−1+ β7leveragei,t−1
+ β8market − to − booki,t−1+ β9CEOdualityi,t+ β10CEOtenure
+ i,t (1)
The specification shown in Equation 1 is tested on a subsample of firms that experience the presence of an activist shareholder. The pre and post effect for firms that experience the presence of an activist shareholder can be tested in the two years around the activist shareholder entering a firm using a difference-in-difference setup. As used in the research of Core et al. (1999) and
Faulkender & Yang (2010), three measures for CEO compensation are used: (i) Salary, (ii) Salary and bonuses and (iii) Total compensation. Salary is the fixed annual compensation component, determined at the beginning of the year. Salary and bonuses equals salary plus bonus plus non-equity incentives. Total compensation is equal to TDC13 from Execucomp. Compensation
measures (i) and (ii) are also obtained from the Execucomp database. Given the findings of Brav et al. (2008), it is expected that the activism variable will have most significant effect on total compensation as this variable also includes equity based compensation. Also, the study of Colak et al. (2017) document a strong increase in equity based compensation of CEOs at firms after it is added to the S&P 500 index. As these authors find that this increase in CEO compensation seems unrelated to executive labor markets, shareholder activism is expected to reduce CEO compensation by reducing equity based compensation. As described in Subsection 2.2, Hartzell & Starks (2003), Brav et al. (2008) and Klein & Zur (2009) find that shareholder activism affects firm policy. Their results show that shareholder activism is negatively correlated with CEO compensation. This could indicate that shareholder activism influences CEO compensation through compensation peer selection. The next specification will test if shareholder activism affects the mean CEO compensation of companies in the compensation peer group of firms in the sample, whilst using the same control variables as used in Core et al. (1999) and Faulkender & Yang (2010) :
ln(M ean CEO compensation peer groupi,t) =β0+ β1Activismi,t−1+ β2ln Salesi,t−1+ β3ROAi,t
+ β4ROAi,t−1+ β5Stockreturni,t
+ β6stockreturni,t−1+ β7leveragei,t−1
+ β8market − to − booki,t−1+ β9CEOdualityi,t
+ β10CEOtenure + i,t (2)
As used in the first regression, specified in Equation 1, a subsample of firms is used that experience the presence of an activist shareholder. The pre and post effect for firms that experience the presence of an activist shareholder can be tested in the two years around the activist shareholder entering a firm using a difference-in-difference setup. The dependent variables in Equation 2 are comparable to the dependent variables used in Equation 1. In
3Salary+Bonus+Non-Equity Incentive Plan Compensation+Grant-Date Fair Value of Stock Awards+Value of Options Granted+Other Compensation
Equation 2, the dependent variables measure the mean CEO compensation of companies in the compensation peer group of firms in the sample. Using a framework similar to Core et al. (1999) and Faulkender & Yang (2010), the dependent variables are again split up in three compensation measures: (i) mean Salary, (ii) mean Salary and bonuses and (iii) mean Total compensation. Following Faulkender & Yang (2010), the compensation variables of Equation 1 and Equation 2 are winsorized at the 1st and 99th percentiles to mitigate skewness in the data. In addition, following Murphy (1999) and Faulkender & Yang (2010), a natural log transformation is applied to the compensation variables. Given the results of Brav et al. (2008) and Colak et al. (2017), the effect of activism is expected to be the most significant for the mean of total compensation of the compensation peer group of firms in the sample as this variable also includes equity compensation.
Using the specification of Equation 2, together with the results from the yet to describe robustness tests in Subsection 4.4, Hypothesis 1 can be either verified or rejected. Using the same subsample and difference-in-difference setup as used to test Equation 1 and Equation 2, Hypothesis 2 is tested using the following specification :
Compensation peer turnoveri,t =β0+ β1Activismi,t−1+ β2ln Salesi,t−1+ β3ROAi,t
+ β4ROAi,t−1+ β5Stockreturni,t
+ β6stockreturni,t−1+ β7leveragei,t−1
+ β8market − to − booki,t−1+ β9CEOdualityi,t
+ β10CEOtenure + i,t (3)
The same explanatory variables are applied in Equation 3 compared to Equation 1 and Equation 2. The dependent variable is compensation peer turnover. This variable is defined as the number of newly added peers in the contemporaneous period divided by the number of peers that have been used in the previous period and re-selected in the contemporaneous period, expressed as a percentage4. Prior studies by Hartzell & Starks (2003), Brav et al. (2008) and Klein & Zur
(2009) indicate that activism is negatively correlated with CEO compensation, and Faulkender
4The compensation peer turnover variable requires compensation peer data on the previous period, the first year that this variable can be determined is 2008. Prior to 2007, compensation peer disclosure was voluntarily (Faulkender & Yang, 2010). It is assumed that compensation peer turnover in 2007 is comparable to the observed
& Yang (2010) find that variation in CEO compensation is explained to a large extent by the composition of the compensation peers. Together, this provides a foundation for the hypothesis that activism causes an increase compensation peer turnover. Using the above mentioned specifications, stated in Equation 2 and Equation 3, allows to interpret the effect of activism on the compensation peer selection process.
4.3
Description explanatory variables
The models described in Subsection 4.2 include the same set of explanatory variables. The explanatory variable of interest is activism. This binary variable receives a value of 1 when at least one activist hedge fund on the SharkWatch50 list owns a minimum of 1% of the shares outstanding of a company in the sample, and 0 if not. Using the Thomson Reuters Institutional Holdings database, together with the SharkWatch50 list, the activism variable is created. The one year lagged form of activism is used in this research as it is assumed that activist hedge funds cannot always exert their influence on the board directly after entering a company, as proxy filings only occur once per year (Denis et al., 2017) and annually contracted CEO compensation cannot be changed within the year of signing. As activism is unlikely to be the only variable that explains variation in CEO compensation, nine control variables are added. The lagged form of some of the financial variables are used to account for the fact that the at the time of determining the CEO compensation, not all contemporaneous financial data would be available to the compensation committee. Lagged sales is used to reflect firm size, which is frequently used in the literature (Faulkender & Yang, 2010; Bizjak et al., 2011). To mitigate skewness of this variable, a log transformation is applied to lagged sales. Variables to control for profitability are return on assets (ROA) and annual stock return are added of both the contemporaneous and previous period. Further measures included are the lagged leverage ratio (total debt over total market value assets) and the lagged market-to-book ratio of assets. The five above mentioned control variables are retrieved from the CRSP-Compustat database. Last, CEO tenure and duality are added. Tenure measures the number of years that the CEO has been in function as CEO of the company. Duality is a binary variable which receives a value of one if the CEO serves as both chairman and CEO of the same company in the same year. Tenure and duality are both created using the Execucomp database.
4.4
Robustness tests
The regression results of Equation 1 and Equation 2 might be subject to selection bias. Activist hedge funds could potentially select into firms where the compensation committee would change the compensation peers regardless of activist intervention, resulting in a decrease in CEO compensation. Propensity Score Matching (PSM) is used to mitigate this form of bias and to contribute to the establishment of finding a causal effect of shareholder activism on compensation peer selection. PSM is a suitable way to correct for observable differences between groups that have and have not received the treatment as there may be differences in outcome between the two groups without activist intervention. Another advantage of PSM is that the propensity score matched control firms may then also be more comparable to firms that are subject to treatment along unobserved and uncontrolled for dimensions. In order to use PSM, a control group is formed using a propensity score, matching untreated firms to treated firm which creates a highly similar untreated control group based on industry, size, leverage ratio and market-to-book ratio. Now the counterfactual of what could have happened without the activist intervention can be studied. The average treatment effect on the treated (ATT) of activism is calculated to interpret the effect of activism in the models described in Subsection 4.2. As used in the previous regressions, a subsample of firms is used that experience the presence of an activist shareholder. The pre and post effect of firms that experience the presence of an activist shareholder can be tested in the two years around the activist shareholder entering a firm using a difference-in-differencesetup.
5
Empirical Evidence
Section 5 will first provide empirical evidence on the relationship between shareholder activism and the level of CEO compensation, using a framework similar to Core et al. (1999) and Faulkender & Yang (2010). Second, the relationship is tested between shareholder activism and the mean CEO compensation of the peers in the compensation peer group. Finally, the effect of shareholder activism on compensation peer turnover is tested. Subsection 5.2 provides the results of the robustness tests using a propensity score matching framework.
5.1
Results
This subsection will first discuss the regression results of the relationship between shareholder activism and CEO compensation between 2007 and 2015. This baseline estimation, similar to Core et al. (1999) and Faulkender & Yang (2010), is used to verify prior studies indicating a negative relationship between shareholder activism and CEO compensation, and using this as a introduction to the second model. The dependent variables are (i) Salary, (ii) Salary and bonuses and (iii) Total compensation. All dependent variables are log transformed. The explanatory variable of interest is activism, a binary variable indicating the presence of at least one activist hedge fund listed on sharkrepellent.net’s SharkWatch50 list. Three specifications are used per dependent variable, using different types of fixed effects. Year, industry and a combination of both year and fixed effects are applied. Year fixed effects are applied to account for the increasing significance of aggregate variables such as economic growth, inflation and compensation. Industry fixed effects are added as variation in CEO compensation is potentially industry specific and could vary across industries. As done by Denis et al. (2017), a combination of both year and industry fixed effects are applied to account for annual trends in compensation that are industry-specific. Together,these fixed effects are applied to account for potential spurious relations between the variables used in the regression. The emphasis of the analysis is placed on the observed results and implication of the relationship between activism and the compensation variables.
The results of the baseline estimation are shown in Table II. As shown in column 1, 2 and 3 of Table II, CEO salary is higher at firms that are larger, CEO tenure is longer (Column 1) and where the CEO also serves as chairman of the company. These results are consistent with findings in prior studies (Core et al., 1999; Faulkender & Yang, 2010), with addition of the significant CEO tenure variable. Activism shows no significant effect on CEO salary. This result is in line with findings of Klein & Zur (2009), as these authors find that out of 151 activist hedge funds campaigns between 2003 and 2006, only two campaigns had a purpose of decreasing CEO salary. From these two campaigns, one campaign was considered successful. A potential explanation for the non-significant relationship between activism and salary is explained by Dittmann et al. (2010), stating that base salaries are rigid. The results presented in column 4, 5 and 6 of Table II show that CEO salary and bonuses are higher at firms that are larger, experience higher return on assets and stock return, CEO tenure is longer and where the CEO also serves as chairman of the company (Column 4). The negative coefficients of the activism variable
can not be interpreted as the variable is not significant. As salary forms a large part of salary + bonuses (+- 50% of median value), the insignificance of activism could again be explained by the research of Dittmann et al. (2010), stating salary is relatively rigid. The results reported in column 7, 8 and 9 of Table II show that total compensation of the CEO is higher at firms that are larger, have shown positive stock returns and as CEO tenure increases. The leverage ratio has a significant negative coefficient. Most interesting, the activism variable is significant at a 5% level in columns 7, 8 and 9 and shows a value of -0.088, -0.095 and -0.098, respectively. Given that the dependent variables are log transformed, activism causes a decrease in total compensation of the CEO between -8.80% and -9.80%. Table I shows a median value of total compensation of $6.01 million. A 9.8% decrease corresponds to a decrease of CEO total compensation of $588,980. These results are statistically significant, but also economically significant. Also, above obtained results on the effect of shareholder activism on CEO compensation are comparable to the results reported by Klein & Zur (2009). As the negative coefficient of lagged activism is only reported significant in specification 8 and 9 where the dependent variable is total compensation and not on salary or salary and bonuses, the decrease in total compensation seems to be driven by a decrease in equity based compensation. This could be explained by the results of Colak et al. (2017) as these authors document a strong increase in equity based compensation as firms enter the S&P 500, which could not be explained by executive labor market conditions. Shareholder activism seems to reverse overcompensation in the form of equity compensation. The results of the robustness test, shown in Table V will be discussed in Subsection 5.2.
The results presented in Table II verify prior studies that have researched the effect of shareholder activism on the level of CEO compensation. These findings, together with results from prior studies (Hartzell & Starks, 2003; Brav et al., 2008; Klein & Zur, 2009), provide sufficient evidence to motivate research on how shareholder activism affects the choice of compensation peers. Table III presents these results.
Table II
Panel regression: CEO compensation
log of Salary log of Salary and bonuses log of Total compensation (1) (2) (3) (4) (5) (6) (7) (8) (9) Lagged activism 0.006 -0.000 0.006 -0.030 -0.041 -0.033 -0.088** -0.095** -0.098** (0.015) (0.015) (0.014) (0.040) (0.039) (0.039) (0.044) (0.044) (0.043) Ln lagged sales 0.026*** 0.020*** 0.013** 0.090*** 0.039** 0.034** 0.123*** 0.080*** 0.055*** (0.006) (0.006) (0.006) (0.015) (0.016) (0.016) (0.018) (0.019) (0.019) ROA (%) 0.000 0.000 0.000 0.008*** 0.007*** 0.007*** 0.003* 0.002 0.002 (0.001) (0.001) (0.000) (0.001) (0.001) (0.001) (0.002) (0.002) (0.001) Lagged ROA (%) 0.000 0.000 0.000 0.000 0.000 0.000 -0.002 -0.001 -0.001 (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.002) Stock return (%) 0.000 0.000 0.000 0.001*** 0.001*** 0.001*** 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Lagged stock return (%) 0.000 0.000 -0.000 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Lagged leverage -0.021 -0.053 -0.036 -0.075 -0.128 -0.115 -0.349*** -0.358*** -0.365*** (0.045) (0.046) (0.044) (0.115) (0.119) (0.119) (0.132) (0.136) (0.133) Lagged market-to-book 0.009 0.010 0.007 -0.010 -0.008 -0.014 0.018 0.029 0.001 (0.011) (0.010) (0.010) (0.027) (0.026) (0.027) (0.030) (0.029) (0.030) CEO Duality 0.045** -0.010 0.029 0.108** -0.044 0.016 0.060 -0.078 0.004 (0.020) (0.020) (0.020) (0.047) (0.051) (0.051) (0.056) (0.058) (0.058) CEO tenure 0.015*** 0.017*** 0.016*** 0.009** 0.014*** 0.012*** 0.012*** 0.017*** 0.014*** (0.002) (0.002) (0.002) (0.004) (0.004) (0.004) (0.005) (0.005) (0.004) Intercept -0.585*** -0.191 -0.198 -0.411*** 0.672* 0.624** 0.332** 1.388*** 1.573*** (0.060) (0.184) (0.183) (0.138) (0.345) (0.317) (0.166) (0.448) (0.463)
Year FE Yes No No Yes No No Yes No No
Industry FE No Yes No No Yes No No Yes No
Year x Industry FE No No Yes No No Yes No No Yes R-squared 0.297 0.768 0.778 0.372 0.628 0.640 0.352 0.691 0.704 Observations 902 902 902 899 899 899 902 902 902
Note: Table II shows the regression results of chief executive officer (CEO) compensation on firm, CEO and activism characteristics. The dependent variables of the disclosing firms are CEO log Salary in columns 1 through 3 , log Salary and bonuses in columns 4 through 6 and log Total compensation in columns 7 through 9. Lagged activism is a binary variable which receives value one if at least one Sharkwatch50 hedge fund owns a minimum of 1% of the shares outstanding of the disclosing firm in the previous year. Lagged sales is the one year lagged sales value. ROA is defined as net income / total assets in the contemporaneous year. Lagged ROA is the 1 year lagged value of ROA. Stock return and Lagged stock returnis the annual stock return in the contemporaneous and and 1 year lagged period. Lagged market-to-book ratio is defined as the one year lagged (Total debt + Market capitalization)/Total assets value. CEO Duality is a binary which assumes value one if the CEO serves as both CEO and chairman of the board. CEO tenure measures the number of years that the CEO has served at the company as CEO. To overcome skewness of the data set, dependent variables are winsorized at the 1st and 99th percentile and a log transformation is applied to the dependent variables as well as to the Lagged sales. Standard errors are clustered per firm and reported in parentheses. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.
Comparable to the first model, the dependent variables are compensation measures. In this model, the compensation measures are applied to the companies in the compensation peer group of the disclosing firms in the sample. The mean is taken of these variables. The dependent variables used are (i) mean Salary, (ii) mean Salary and bonuses and (iii) mean Total compensation. As used in the first model, three specifications are used per dependent variable where different fixed effects are applied: year, industry and a combination of both year and industry fixed effects. Columns 1, 2 and 3 show that the compensation of CEOs in the compensation peer group is higher when firms are larger, leverage increases and when the CEO also serves as chairman of the company (Column 1 & 3). The activism variable has significant negative coefficient with values of -0.025 and -0.024 in columns 2 and 3, respectively. These results indicate that activism reduces the mean compensation peer salary between 2.4% and 2.5%. The results in column 4, 5 and 6 show that the compensation of CEOs in the compensation peer group is higher when firms are larger, stock return increased in the previous period (Column 4 & 5) and in the current period (Column 5), the CEO also serves as chairman of the company (Column 1) and when CEO tenure increases (Column 5). Activism has a negative coefficient in columns 5 and 6 with values of -0.065 and -0.061, respectively. This indicates that activism reduces the mean salary and bonuses of the compensation peers between 6.1% and 6.5%. Columns 7, 8 and 9 show that the mean total compensation of CEOs in the compensation peer group increases as firms are larger and as CEO tenure increases (Column 8 & 9). As reported in specification 8 and 9, the mean total compensation of CEOs in the compensation peer group decreases by 6.2% and 6.5% when an activist shareholder is present at the disclosing firms. The lagged activism variable is significant in all specifications except in column 1, 4 and 7. Most interesting, the lagged activism variable is significant in all specifications where both year and industry fixed effects are applied, as this combination of fixed effects captures the most trends compared to the other specifications used. The results indicate that activism could affect the composition of the compensation peer group by selecting peer companies with lower CEO compensation, resulting in a decrease of the mean compensation of CEOs in the compensation peer group. These findings support Hypothesis 1. However, these results could potentially be subject to selection bias. In an attempt to mitigate the selection bias, a propensity score matching framework is applied to test for a causal relationship between activism and the mean peer group compensation. The results of this test are discussed in Subsection 5.2.
to study the effect of shareholder activism on compensation peer turnover. Table IV presents the results of testing the relationship between shareholder activism and compensation peer turnover. The effect of lagged shareholder activism is significant in specification 2 and 3, and shows a trend towards significance in specification 1. The coefficient of lagged activism shows a value of 32.681 and 35.212 in column 2 and 3 respectively, indicating that lagged activism increases the compensation peer turnover roughly between 32.7% and 35.2%. These results provide supporting evidence in favor of Hypothesis 2. Together with the findings presented in Table III, indicating a negative relationship between shareholder activism and the mean CEO compensation of the compensation peers, the documented results indicate that shareholder activism causes increased compensation peer turnover where relatively higher pay compensation peers are replaced by relatively lower paid compensation peers.
5.2
Robustness tests
As described in Subsection 4.4, the results shown in Table II and Table III might be subject to selection bias. In an attempt to test for a causal relationship between shareholder activism and CEO compensation (Equation 1) and shareholder activism and the mean CEO compensation of the compensation peers (Equation 2), a PSM approach is employed (Rosenbaum & Rubin, 1983; Armstrong et al., 2008; Faulkender & Yang, 2010) using the same difference-in-difference setup as used in the previous regressions. For the companies in the sample that encountered shareholder activism, a highly similar comparison group is created based on a company’s industry, size, leverage ratio and market-to-book ratio. Using the PSM approach, the counterfactual of an activist entering a firm can be tested. The results of the PSM approach are summarized in Table V and Table VI.
Table V shows the results of how activism affects CEO income, using an PSM approach. The dependent variable is split up in three compensation measures: (i) Salary, (ii) Salary and bonuses and (iii) Total compensation. The ATT of lagged activism is calculated on each compensation measure.
Table III
Panel regression: mean peer group compensation
log of Salary log of Salary and bonuses log of Total compensation (1) (2) (3) (4) (5) (6) (7) (8) (9) Lagged activism -0.017 -0.025** -0.024** -0.041 -0.065** -0.061** -0.042 -0.062** -0.065** (0.011) (0.011) (0.010) (0.027) (0.026) (0.026) (0.028) (0.028) (0.026) Ln lagged sales 0.036*** 0.023*** 0.015*** 0.112*** 0.049*** 0.041*** 0.093*** 0.050*** 0.026** (0.005) (0.005) (0.005) (0.011) (0.012) (0.012) (0.012) (0.013) (0.012) ROA (%) 0.000 -0.000 -0.000 0.001 0.001 0.000 0.001 0.001 0.000 (0.000) (0.000) (0.000) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Lagged ROA (%) 0.000 0.000 0.000 -0.001 -0.001 -0.001 -0.002* -0.001 -0.001 (0.000) (0.000) (0.000) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Stock return (%) -0.000 -0.000 -0.000 0.000* 0.000** 0.000* 0.000 -0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Lagged stock return (%) 0.000 0.000 -0.000 0.001*** 0.001*** 0.000 -0.000 0.000 -0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Lagged leverage 0.091** 0.084** 0.076** 0.016 -0.028 -0.047 -0.075 -0.082 -0.085 (0.037) (0.037) (0.035) (0.085) (0.089) (0.087) (0.089) (0.093) (0.086) Lagged market-to-book 0.007 0.005 0.002 0.035* 0.000 0.012 0.025 0.028 0.009 (0.009) (0.008) (0.008) (0.020) (0.019) (0.020) (0.022) (0.022) (0.021) CEO Duality 0.054*** 0.015 0.033** 0.087*** 0.000 0.025 0.065* -0.022 0.017 (0.015) (0.015) (0.014) (0.032) (0.035) (0.034) (0.035) (0.037) (0.035) CEO tenure -0.000 0.001 0.000 0.003 0.007** 0.005* 0.005* 0.009*** 0.007** (0.001) (0.001) (0.001) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Intercept -0.487*** 0.006 0.035 -0.213** 0.681*** 0.717*** 0.963*** 1.888*** 2.028*** (0.049) (0.120) (0.119) (0.104) (0.252) (0.263) (0.115) (0.293) (0.304)
Year FE Yes No No Yes No No Yes No No
Industry FE No Yes No No Yes No No Yes No
Year × Industry FE No No Yes No No Yes No No Yes R-squared 0.468 0.812 0.820 0.503 0.731 0.742 0.416 0.766 0.784 Observations 774 774 774 775 775 775 775 775 775
Note: Table II shows the regression results of CEO compensation on firm, CEO and activism characteristics. The dependent variables are log mean peer group salary in columns 1 through 3 , log mean peer group Salary and bonuses in columns 4 through 6 and log mean peer group Total compensation in columns 7 through 9. Lagged activism is a binary variable which receives value one if at least one Sharkwatch50 hedge fund owns a minimum of 1% of the shares outstanding of the disclosing firm in the previous year. Lagged sales is the one year lagged sales value. ROA is defined as net income / total assets in the contemporaneous year. Lagged ROA is the 1 year lagged value of ROA. Stock return and Lagged stock returnis the annual stock return in the contemporaneous and and 1 year lagged period. Lagged market-to-book ratio is defined as the one year lagged (Total debt + Market capitalization)/Total assets value. CEO Duality is a binary which assumes value one if the CEO serves as both CEO and chairman of the board. CEO tenure measures the number of years that the CEO has served at the company as CEO. To overcome skewness of the data set, dependent variables are winsorized at the 1st and 99th percentile and a log transformation is applied to the dependent variables, as well as to the Lagged sales. Standard errors are clustered per firm and reported in parentheses. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.
Table IV
Compensation peer turnover
Compensation peer turnover (%)
(1) (2) (3) Lagged activism 23.380* 32.681** 35.212** (13.174) (15.154) (15.205) Ln lagged sales -1.055 -3.942 -1.990 (3.343) (8.238) (8.366) ROA (%) -0.244 -0.392 -0.186 (0.453) (0.500) (0.508) Lagged ROA (%) 0.222 0.342 0.307 (0.517) (0.558) (0.561) Stock return (%) 0.013 -0.024 0.008 (0.053) (0.060) (0.062)
Lagged stock return (%) -0.126 -0.232** -0.191
(0.128) (0.116) (0.136) Lagged leverage 18.933 43.679 58.936 (31.135) (54.851) (55.162) Lagged market-to-book -3.259 11.472 6.148 (9.294) (12.392) (13.022) CEO Duality -9.915 -17.971 -19.101 (9.529) (15.401) (15.486) CEO tenure -0.373 -0.383 -0.336 (0.986) (1.334) (1.337) Intercept 88.476** 17.103 39.193 (34.579) (88.302) (90.284) Year FE Yes No No Industry FE No Yes No
Year × Industry FE No No Yes
R-squared 0.024 0.172 0.184
Observations 674 674 674
Note: Table IV shows the regression results of compensation peer turnover on firm, CEO and activism characteristics. The dependent variable Compensation peer turnover is defined as the number of newly added peers in the contemporaneous period divided by the number of peers that have been used in the previous period and re-selected in the contemporaneous period, expressed as a percentage. Lagged activism is a binary variable which receive value one if at least one Sharkwatch50 hedge fund owns a minimum of 1% of the shares outstanding of the disclosing firm in the previous year. ROA is defined as net income / total assets in the contemporaneous year. Lagged ROA is the 1 year lagged value of ROA. Stock return and Lagged stock returnis the annual stock return in the contemporaneous and and 1 year lagged period. Lagged market-to-book ratio is defined as the one year lagged (Total debt + Market capitalization)/Total assets value. CEO Dualityis a binary which assumes value one if the CEO serves as both CEO and chairman of the board. CEO tenuremeasures the number of years that the CEO has served at the company as CEO. Standard errors are clustered per firm and reported in parentheses. Statistical significance at the 1%, 5%, and 10% level is indicated by ***, **, and *, respectively.
