Relative performance evaluation in CEO compensation
Master Thesis
Faculty of Economics and Business
MSc Finance
Track: Asset Management
Romy van As
(10186042)
July, 2017
Statement of originality
This document is written by student Romy van As 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.
Abstract
Although agency theory suggests that firms filter out market-wide effects from the
compensation package, there is little evidence to support this relative performance evaluation (RPE). This research examines the existence of relative performance evaluation in CEO compensation and the cross-sectional differences that can explain the extent to which companies use relative performance evaluation. Using data for S&P 1500 firms and peer performance based on companies of the same size and in the same industry, this study finds evidence for weak-form RPE. The findings suggest that companies operating in a more concentrated industry use less RPE, firms with higher growth opportunities are more likely to use RPE, large firms use less RPE, and controlled companies use less RPE. This study argues that controlled companies use less RPE since they have weaker corporate governance and thus allow more pay-for-luck.
Content 1. Introduction 5 2. Literature review 8 Hypothesis I 12 Hypothesis II 12 Hypothesis III 12 Hypothesis IV 14 3. Methodology 16
4. Data and descriptive statistics 21
5. Results 25
6. Robustness checks 30
7. Conclusion 37
8. References 40
1. Introduction
Agency theory suggests that if executive pay is linked to the performance of the firm, executives will take actions that are optimal for the firm (Aggarwal & Samwick, 1999). Compensation based on performance creates incentives for the executive to be productive, but this performance is often affected by random factors beyond the executive’s control (Gibbons & Murphy, 1990). When setting incentive compensation, relative performance evaluation (RPE) implies the use of the performance of peers in evaluating the executive’s performance (Rajgopal, Shevlin & Zamora, 2006). RPE partially insulates individuals from common risk and uncertainty, assuming that the performance of peers includes common exogenous shocks. The idea behind it is that a CEO should not be held responsible for risks and factors that are beyond his control, like stock market volatility and general economic conditions. In top-level management, the risk-sharing advantages of RPE are likely to exceed the costs.
Despite the theoretical appeal of RPE, prior research shows mixed evidence for the use of RPE. In fact, a common aspect of executive compensation is that executives are rewarded for increases in the stock price due to general market trends, also called pay-for-luck
(Bertrand & Mullainathan, 2001). The SEC’s new disclosure rules of 2006 on executive compensation caused several studies to examine the disclosures about peer group
composition. Besides this, several studies found a variety of characteristics of firms that affect the decision of firms to use RPE, such as industry concentration, CEO talent, firm size and growth opportunities. The aim of this research is to resolve the lack of empirical support for the use of RPE in CEO compensation, a fact that has been termed the “RPE puzzle”1, by using different factors to explain the use of RPE. The research will investigate if differences in these characteristics (i.e. in the cross-section) can explain a difference in the weight given to peer performance relative to the weight given to own firm performance in setting CEO compensation.
The research question that will be answered in this thesis is: “Is there relative performance evaluation in CEO compensation and which characteristics increase the likelihood that companies use RPE?”
This research tests for RPE in executive compensation in the following way. Four variables
1 See, for instance, Aggarwal & Samwick (1999), Gibbons & Murphy (1990) and Rajgopal, Shevlin & Zamora (2006).
are used that can increase the likelihood that companies use RPE (i.e. give a more negative weight to peer performance) in setting executive compensation. The four variables are industry competition, growth opportunities, company size and whether a company is controlled or not. Aggarwal & Samwick (1999) find that when an industry is more
concentrated (and thus less competitive), the weight on rival performance relative to own firm performance in setting compensation is more positive, which implies less RPE in more
concentrated industries. They argue that this is the case because RPE may encourage
destructive competition. They test if industry competition can explain the use of RPE and they base peer groups on industry. In this research, I will test if industry concentration
(competition) increases (decreases) the weight on peer performance relative to own firm performance. In addition to industry, peer groups will be also based on firm size. Previous research suggests that firms with high growth opportunities are more likely to use the performance of peers in compensation contracts, since internal measures provide CEO’s of firms with high growth opportunities strong incentives to smooth performance (Murphy, 2000). This paper will investigate if companies with high growth opportunities are indeed more likely to use the performance of peers in compensation contracts. The effect of firm size on the weight given to peer performance relative to own firm performance in setting
compensation will be investigated because firm size can capture CEO talent (Himmelberg & Hubbard, 2000). Rajgopal et al. (2006) found less RPE for more talented CEO’s, which makes it interesting to investigate the effect of firm size on the weight given to peer performance in setting CEO compensation.
The new aspect in this research is whether there is a difference in the presence of RPE between “controlled” and “non-controlled” companies. This has not been investigated before. A controlled company is a company in which a group or a person owns a majority of the voting stock or when that group or person is entitled to elect a majority of the directors of the firm. It is interesting to investigate because previous research argues that controlled
companies are weaker governed. Since Bertrand and Mullainathan (2001) find that there is less pay-for-luck in better-governed firms, this paper investigates if these weaker governed (controlled) firms are less likely to use RPE. The use of RPE will decrease this pay-for-luck since external shocks that are beyond the control of the CEO will be (partly) filtered out of the compensation. The expectation is that controlled companies use less RPE compared to non-controlled firms because companies with a weak governance structure will only use RPE when it is in the interest of the CEO to do so (Garvey & Milbourn, 2006).
Using more recent data than previous studies, a different peer group composition and a new independent variable that might explain the use of RPE, this research will help to resolve the mixed evidence of the use of relative performance evaluation in executive compensation.
The research question will be investigated by using a sample of S&P 1500 firms for the period 2010 to 2015. Ordinary Least Squares regressions will test if companies use RPE and an interaction regression model tests if the four variables can explain if these characteristics cause more or less use of RPE. I investigate whether absence of RPE is increasing for controlled companies since these companies are likely to have poor governance.
This paper is structured as follows. In section 2, the literature review, attention will be paid to a couple of things. The section will begin with a review of agency theory and the theory of relative performance evaluation. A selection of previous empirical research about relative performance evaluation and the variables that can explain the use of it will follow. After this, a review about controlled companies and why they can be important in explaining RPE will be discussed. The literature review ends with the theory about the selection of peers and the institutional background. Section 3, the methodology, outlines the method that will be used in this thesis. Section 4, the data section, first discusses the data sources that are used to find the data and then describes the data that will be used. Section 5, the results, analyses the data, reports the empirical results of the tests that have been used and will state the interpretations that became apparent. Section 6 will present the robustness checks. In section 7, the thesis will be concluded with an answer to the research question.
2. Literature review Agency theory
The separation of ownership and control in companies is an important feature of modern economic theories. An example of a principal-agent relationship is the relationship between the shareholders and the executive of a publicly owned company (Gibbons & Murphy, 1990). The decisions and actions of a firm’s executives can have a large impact on the shareholder wealth. Top executives are considered as being risk-averse, which implies that they want their compensation structured in a way that they bear less risk (Mehran, 1995). Besides this,
shareholders are considered risk-neutral because of their option to diversify firm-specific risk by holding a portfolio that is well diversified. Shareholders want the executive to make decisions that increase the shareholder wealth, whereas the executive wants to take actions that increase his expected utility. There has been considerable disagreement about whether Chief Executive Officer (CEO) compensation in large firms is set up so that the decision making of executives is directed toward maximizing the performance of the firm. In the interest of reducing their compensation risk, executives can take actions that reduce the firm’s risk, but adversely affect shareholder wealth (Mehran, 1995).
Agency theory (principal-agent theory) suggests that the primary means for
shareholders to make sure that an executive takes actions that are optimal for the firm is to link their pay to the performance of the firm (Aggarwal & Samwick, 1999). This will be tempered by the extent to which the firm performance is influenced by random shocks beyond the CEO’s control (Jenter & Kanaan, 2015). If these exogenous shocks are correlated across firms in the same industry, then the optimal compensation package compensates an executive on the performance of the firm relative to the performance of those other firms. In this way, CEO’s will be evaluated based on the firm-specific component of the performance of the firm only. Agengy theory predicts that this market-wide component of the returns of a firm should be excluded from the compensation package because executives are not able to influence the market and it is costly to bear the relative risks for the executive (Rajgopal et al., 2006).
Relative performance evaluation
The accurate evaluation of executives and the arrangement of appropriate compensation contract is an important matter. Compensation based on performance provides incentives for an executive to be productive, but this measured performance is often affected by factors that are beyond the executive’s control (Gibbons & Murphy, 1990). When setting incentive
compensation, relative performance evaluation (RPE) implies the use of the performance of peers in evaluating the executive’s performance (Rajgopal et al., 2006). It insulates
individuals from common risk, assuming that peer performance includes common exogenous shocks. The idea behind it is that an executive should not be held responsible for factors and risks that he cannot control like stock market volatility and general economic conditions. RPE filters out a common shock by giving a positive weight to the performance of the own firm and a negative weight to the performance of the industry (Aggarwal & Samwick, 1999). The negative industry pay-performance sensitivity implies that the compensation of the executive will be higher if the executives of other firms in the same industry have lower returns. Theory suggests that there are two forms of RPE: strong-from RPE and weak-form RPE (Jenter & Kanaan, 2015). The strong form version of the RPE hypothesis is that the market-wide component (i.e. the common factor) will be completely filtered out and that performance evaluation is a function of the unsystematic (i.e. firm-specific) component of the performance measure only. The weak form of this hypothesis is that there is RPE but that this common factor will not be completely filtered out (Rajgopal et al., 2006). For the weak form hypothesis, the performance evaluation can be a function of the unsystematic and also the systematic component of the performance measure.
Benefits and costs of relative performance evaluation in top-level management
Economic theory gives a rationale for RPE based on risk-sharing advantages. If compensation is based only on the individual performance, it provides incentives to be productive (Gibbons & Murphy, 1990). RPE can entail costs when it is expensive or hard to measure the output of workers or when it generates incentives to sabotage the performance of the reference group, to choose inappropriate peer groups or to collude with peers and shirk. According to
tournament theory, the heterogeneity among agents can weaken the advantages of RPE. To be more specific, RPE contracts can cause inefficiency when agents do not have an equal chance to win in a tournament, given the same effort level. An unequal contest like this may induce the disadvantaged agents to shirk (Gong, Li & Shin, 2011). Besides this, it can distort the risk choices of an agent, i.e. more capable agents become conservative with their risk choices to preserve their positions, whereas the agents that are less capable become too aggressive with their risk choices. However, in top-level corporate management, relative-performance contracts for executives are not expensive to administer because the performance of peer firms is available and because executives tend to have limited interaction with executives of peer firms, which makes sabotage unlikely (Gibbons & Murphy, 1990). Thus, in top-level
management, the advantages of RPE exceed the costs, which makes this a likely place to observe RPE. Besides this, it seems likely that industry and market shocks are important when setting top-executive compensation because filtering these shocks out can benefit risk-averse executives (Gibbons & Murphy, 1990).
Despite the theoretical appeal of RPE, the literature gives little and mixed empirical evidence of RPE, and suggests that a variety of factors affect the decision of firms to use RPE. There is an ongoing debate in literature about explaining the lack of empirical evidence supporting RPE. These studies are mostly focused on explaining the difference in the use of RPE between companies with different characteristics (i.e. in the cross-section) and on the argument that defining an appropriate peer group improves the power to detect RPE.
Variables that influence the use of RPE
There are multiple variables that can influence the use of RPE and previous studies found different empirical evidence for the use of RPE. Aggarwal and Samwick (1999) examine compensation contracts for executives in product markets that are imperfectly competitive, using data from 1993 until 1995 for executives of large corporations from the Standard and Poor's ExecuComp dataset. They test for the effects of strategic competition on RPE in compensation contracts. They demonstrate that the lack of relative performance-based incentives in which compensation decreases with rival firm performance can be explained by strategic interactions among firms. They argue that the more concentrated the industry, the more positive the weight on rival firm performance relative to own firm performance. This means that they find less RPE in more concentrated industries because RPE may encourage destructive competition. It might create an incentive for a manager to compete too
aggressively, which can hurt the profitability of the company. Rajgopal et al. (2006) tested for the hypothesis of less RPE in oligopolistic industries by using the Herfindahl-Hirschman Index as a measure for industry concentration and find no results that are consistent with that hypothesis.
Gong, Li and Shin (2011) examine the use of RPE in executive compensation contracts and the selection of peers to use, using data of 2006 for S&P 1500 firms. They suggest that when deciding to use RPE, firms consider both benefits and costs of RPE as an incentive mechanism. Their findings show that incorporating details of RPE contracts, for example the selection of peers, improves the significance to detect the use of RPE. Gong et al. (2011) find that firms are more likely to use RPE when they are exposed to a higher level of
common risk and when they operate in less concentrated industries.
Chen, Liang and Zhu (2012) examine RPE in executive compensation of Chinese listed companies with a focus on the difference of RPE use between state-owned enterprises (SOE’s) and non state-owned enterprises (non-SOE’s). Their hypothesis is that there should be a difference between SOE’s and non-SOE’s because the compensation models of these kind of firms differ. Overall, they find no significant evidence of an RPE effect. However, results show that RPE is more likely to be used in non-SOE’s than in SOE’s which can be due to the regulation of cash compensation, various forms of incentives and the multiple tasks of managers in SOE’s. Bertrand and Mullainathan (2001) argue that CEO pay has a positive correlation with exogenous shocks and call this pay-for-luck. They argue that CEO pay is more sensitive to good luck than to bad luck and find that this difference is bigger for weaker governed firms. They document that there is less pay-for-luck in better-governed firms. Because of these findings, ownership structure is an important variable for measuring the use of RPE.
Rajgopal et al. (2006) test if the sensitivity of CEO pay to systematic market-wide factors is increasing with CEO talent. They find that, based on observations from 1993 to 2001 of S&P 500 firms, compensation committees do not completely filter out industry-wide performance from the CEO’s compensation, that is, they do not practice strong-form RPE. Besides this, results show that RPE is optimally lower for more talented CEOs. Firm size is also an important variable for measuring RPE, because firm size could capture talent of CEO’s (Himmelberg & Hubbard, 2000). Since the supply of talented CEOs is relatively inelastic, it may be optimal to reward these talented CEOs for common shocks in the market if these shocks increase the market value of the firm and the outside employment
opportunities of the CEO (Rajgopal et al., 2006). Because Rajgopal et al. (2006) found that there is less RPE for more talented CEOs, it is interesting to see the effect of firm size on the effect of RPE.
Garvey and Milbourn (2003) find that there is little RPE for the average CEO, but find strong evidence of it for younger CEOs and for CEOs with less financial wealth. Murphy (2000) suggests that firms that have high growth opportunities are more likely to use external measures (like the performance of peers) in compensation contracts because internal measures (for example historical earnings growth) provide executives of companies with high growth strong incentives to smooth performance.
The variables that will be used to explain the use of RPE in this research are industry competition, growth opportunity, firm size and whether a company is controlled or not. The
expectation is that firms operating in a more concentrated industry use less RPE because RPE may encourage destructive competition. From this, the following hypothesis can be derived: Hypotheses I: Firms operating in a more concentrated industry use less RPE.
I expect that firms with high growth opportunities are more likely to use RPE, because these firms tend to be more likely to use external measures (for example peer performance) in setting compensation because internal measures provide incentives to smooth performance. The hypothesis is the following:
Hypothesis II: Firms with high growth opportunities are more likely to use RPE. Third, I investigate the difference in the use of RPE (i.e. the weight firms give to peer
performance in setting compensation contracts) between companies with a different size. Firm size could capture CEO talent and Rajgopal et al. (2006) found that firms use less RPE in setting compensation for more talented CEO’s. I expect that large firms use less RPE
compared to smaller firms since large firms are more likely to have talented CEO’s. The third hypothesis is:
Hypothesis III: Large firms use less RPE.
Controlled vs. non-controlled firms
In general, control is defined according to share capital, voting rights, board representation or a combination of those (Kamonjoh, 2016). A controlled firm is a firm in which a person or a group together owns a majority of the voting stock of a firm (also called blockholders), or when that person or group is entitled to elect a majority of directors. However, this definition of a controlled firm does not include shareholders with a link to the firm, such as family or founders, who can have control over voting outcomes. In addition, the definition does not take into account that there can also be owners of substantial (but not majority) stakes that allow these investors to control voting results. These persons are often executives or founders, whose stakes are intensified by their insider status. Because of these shortcomings in the definition of a controlled firm, in a study for the Investor Responsibility Research Center institute (IRRCi), Kamonjoh (2016) broadened the definition by including any group or person that owns 30 percent or more of the voting power of a firm. The IRRCi did research
(in 2016 but also in 2012) to investigate which companies of the S&P 1500 are controlled companies.
A common feature of a controlled firm is a capital structure in which the founders control a majority of the voting stock, while they hold a relatively small portion of the economic value of the firm. There exist two primary control mechanisms, namely ownership by a person or group of 30 percent or more of single class of stock and multiclass capital structures with unequal voting rights. The latter permits control of the firm through one or most of the time more classes of stock. These classes entitle the holder to enhanced voting rights. These control mechanisms may misalign interest between external shareholders and affiliated shareholders because it allows insiders to operate without the normal accountability mechanisms (Kamonjoh, 2016).
Most of the stock exchanges in the United States relax their basic governance listing requirements for controlled companies (Kamonjoh, 2016), which results in the fact that governance provisions (which provide protection for external shareholders) do not apply to these companies, which makes them weaker governed. Besides this, the IRRCi (2016) found that the average level of board independence is lower at controlled companies compared to non-controlled companies. The independence level of compensation committees of controlled companies is also lower than for non-controlled companies. Since Bertrand and Mullainathan (2001) find that there is less pay-for-luck in better-governed firms, it is an interesting issue is to investigate the difference in use of RPE between controlled and non-controlled firms. Well-governed firms limit the CEO’s ability to capture the pay process (Bertrand and Mullainathan, 2001). If the compensation of an executive increases based on performance while this
performance is driven by factors beyond the executive’s control, this can be called pay-for-luck. The use of RPE will decrease this pay-for-luck since external shocks will be (partly) filtered out of the compensation package. Moreover, Gong et al. (2011) found that firms with more independent board are more likely to use RPE, which is consistent with stronger
corporate governance encouraging the use of RPE. The expectation is that controlled firms use less RPE compared to non-controlled firms since they have weaker corporate governance. Because they are weaker governed, there may be more pay for luck and in turn less RPE. Kamonjoh (2016) found that in 2016, controlled companies underperformed non-controlled companies with respect to metrics that affect unaffiliated shareholders such as revenue growth, return on equity and shareholder returns. They also found that there are longer executive tenures, less frequent board refreshments and fewer financial experts in the boards of controlled companies. In addition, their study found that average CEO pay is higher
at controlled companies compared to non-controlled companies (especially in the S&P 500, so for large-cap firms). Appendix A contains a list with all the controlled companies in the S&P 1500, according to the report of the IRRCi. The list includes the companies that were controlled in 2012 and stayed controlled until 2016. If a company was controlled in 2012 but was not on the list of the IRRCi with controlled companies in 2016, this company is not on the list in Appendix A.
From the discussion above, the following hypothesis can be derived: Hypothesis IV: Controlled firms use less RPE compared to non-controlled firms.
Peer selection
Besides the costs of RPE mentioned above, another potential explanation for the lack of finding consistent evidence for the use of RPE is the inappropriate specification of peer groups. It is possible that previous research lacks power to find evidence for RPE because they chose the peer groups incorrectly (Albuquerque, 2009).
RPE peer groups are mainly used to filter out systematic risk in order to insulate compensation from exogenous shocks (Gong et al., 2011). If this is the case, RPE peer groups are chosen in a manner consistent with efficient contracting. The SEC’s new disclosure rules of 2006 on executive compensation (explained in Institutional Background) caused several studies to examine the disclosures about the compensation of peers. From the rent extraction perspective, executives can have incentives to choose an RPE peer group of which they expect to perform worse than their own firm. Worse performing firms are more likely to be selected as an RPE peer to boost the relative performance of the firm and hence the
compensation. This will create peer selection bias (Gong et al., 2011).
In choosing an RPE peer group, it is challenging to find a set of firms that have common exposure to exogenous shocks and share the same ability to react to those shocks (Albuquerque, 2009). If some firms in an industry are affected negatively by a shock and other firms in the same industry are affected positively by the shock, then the average industry performance does not capture the exogenous shock. Consequently, when firms in an industry are heterogeneous, then the index of an industry is not a good measure of peer performance. The research of Albuquerque (2009) identifies more appropriate peer groups when peers are matched on both industry and size. She argues that firms of different size face different constraints in reacting to a shock and are exposed to different shocks. She suggests that an ideal peer group includes firms that are similar in multiple characteristics, namely industry,
size, financing constraints, diversification, growth options and operating leverage. For example, firms that are more diversified can take advantage of multi-segment flexibility by moving production across business segments when reacting to industry shocks (i.e. smoothing industry shocks that influence the firms businesses differently). Second, firms that are less financially constrained can respond quicker to shocks. A negative shock to the liquidity of the market can force an executive of a small firm to drop a project (despite a positive net present value) because he fails to get access to credit, while an executive of a large firm can go on with the project because he can still find credit. Besides this, the degree of operating leverage of a firm can affect the profit sensitivity to industry demand shifts of a firm. It is difficult to create a peer group with firms that all include the above characteristics. Albuquerque (2009) argues that these characteristics are monotonically related to firm size. To give an example, small firms tend to have greater financing constraints, less operating leverage and are often less diversified. To conclude, the research of Albuquerque reports stronger evidence of RPE when firms are matched on both industry and size. That is why in this research, peer firms will be selected based on the same industry and the same size quartile.
Institutional background
In 2006, the Securities and Exchange Commission (SEC) introduced new disclosure rules on executive compensation. According to the new rules, companies have to provide
Compensation Discussion and Analysis (CD&A) report in their proxy statements. In this CD&A, it is mandatory to give a description of the process that is used by the firm to select performance targets. Besides this, it contains an evaluation of how these targets affect compensation. Consequently, firms that are using RPE are required to disclose the details about it in the CD&A. This may change the decision to use and the execution of RPE, for example the selection of peers (Gong et al., 2011). Because of the new disclosure rules, the research about RPE became more reliable because firms that do use it have to be able to justify their decision and execution since the information is publicly available.
3. Methodology
Different empirical methods are used in previous literature to test if there is relative
performance evaluation in executive compensation. Most of those researches use an implicit method to test for RPE (Aggarwal & Samwick, 1999; Garvey & Milbourn, 2003; Gibbons & Murphy, 1990; Himmelberg & Hubbard, 2000; Rajgopal et al., 2006). For example, Aggarwal and Samwick (1999) derive optimal compensation contracts in which managers are
compensated based on their own performance but also on the performance of peers. They derive the contracts for a Bertrand competition and also for a competition that is differentiated Cournot. They estimate the pay-performance sensitivity and use the percentile of the firm’s industry in the distribution of concentration ratios (i.e. how concentrated the industry is) to explain the use of RPE. Ordinary Least Squares (OLS) regressions are used to implicitly test the hypothesis that rival performance influences the compensation of executives. To test the effect of industry concentration on the use of RPE, they interact peer performance with the industry concentration ratio. Rajgopal et al. (2006) research RPE by testing the hypothesis that the sensitivity of CEO compensation to market-wide systematic factors is increasing in CEO talent. Besides this, they test whether RPE is increasing or decreasing for certain measures of corporate governance. They use a specification including shareholder wealth of the firm and shareholder wealth of the industry as independent variables and total CEO compensation as a dependent variable. To test for the impact of CEO talent and the corporate governance proxies on RPE, they include interaction terms of these variables with peer performance in the regression equation. Gong et al. (2011) used an explicit method to test for RPE. They obtained compensation data from ExecuComp, financial data from Compustat and stock return data from CRSP. In order to find out which firms explicitly used RPE and how these firms composited their RPE peer groups, they retrieved firms’ first proxy statements filed under the new executive compensation regime after the SEC introduced the disclosure rules in 2006.
In this research, I will analyze panel data to find out if S&P 1500 companies use RPE and if certain variables cause that RPE is used more (or less). The method of Albuquerque (2009) is used for the peer selection to avoid peer selection bias. This means that peers are selected per year on (2-digit SIC) industry and size quartile. These size quartiles are created per year based on the market capitalization of the companies. Following Albuquerque (2009), companies with less than two peers were dropped from the sample. OLS regressions test to what extent
(i.e. strong or weak form) companies use relative performance evaluation in executive
compensation and what factors affect the use of RPE. First, the research tests for the existence of RPE. Then, the study tests if different measures that can explain the use of RPE cause more or less RPE. From this, the differences in the use of RPE in the cross-section can be
discovered. The measures that will be used to explain the use of RPE are industry
competition, growth opportunities, firm size and corporate governance. This study contributes to the existing RPE literature by testing for the effect of being a controlled company on the use of RPE. The impact of corporate governance on the use of RPE will be tested by the difference in use of RPE between controlled and non-controlled companies.
In this research, the following model is estimated:
𝐶𝐸𝑂𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛! = 𝑐!+ 𝛼!𝐹𝑖𝑟𝑚𝑃𝑒𝑟𝑓!+ 𝛼!𝑃𝑒𝑒𝑟𝑃𝑒𝑟𝑓!+ 𝛼!𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!+ 𝛼!𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝜀! (1)
The subscript i indicates a CEO-firm pair. The dependent variable is the total compensation of the CEO’s, as in the study of Rajgopal et al. (2006). FirmPerfi and PeerPerfi are performance measures for firm i and the peer firms of firm i. Following the study of Gong et al. (2006), Albuquerque (2009) and Gibbons & Murphy (1990), firm performance and peer performance are 12-month buy-and-hold stock returns for the firm and it’s peer group. For the peer
performance, the performance is the buy-and-hold return on a value-weighted portfolio of firms (excluding the own firm) in the same two-digit SIC industry within the same size quartile. An endogeneity problem when estimating equation (1) is the simultaneous causality between CEO compensation and performance. If the performance increases, the compensation is expected to increase. However, if the compensation increases, it can stimulate the executive causing higher performance. When estimating the model, factors that absorb the correlation between performance and compensation have to be included. There will be controlled for industry, firm and CEO characteristics that might affect both compensation and firm performance. Following previous literature, the control variables include company size, growth opportunities, CEO tenure, CEO age, industry concentration and a dummy variable that equals one if the CEO served as a director during the fiscal year. Size and CEO age are included as control variables because previous research suggests that the pay for performance sensitivity is a function of CEO age and firm size. Gibbons and Murphy (1990) find that the pay for performance sensitivity increases with CEO age and Rajgopal et al. (2006) argue that the pay for performance sensitivity is firm size related because the CEO’s marginal product
varies with the size of the firm. Company size is measured by the market value of equity (Gong et al., 2006). Gong et al. (2006) find that a measure of growth opportunities is an important variable in explaining the use of RPE. The book to market ratio is used as an inverse proxy for growth opportunities. This indicates that if the book to market ratio increases, the growth opportunities decrease. Besides this, the research of Aggarwal and Samwick (1999) includes industry concentration because if rival firms in the same industry are strategic competitors, then it will alter an executive’s strategic product market choices if common shocks are filtered out of his compensation. Bertrand and Mullainathan (2001) include CEO tenure as a CEO specific variable because tenure can be a measure of corporate governance. Dummy variables for industry and year will control for differences in pay-levels across industries and time (Murphy, 2000).
RPE can be detected as follows. The coefficient of interest is the coefficient of the performance of the peer group. The weak-form test of RPE tests the null hypothesis 𝐻!: 𝛼! ≥ 0 against the alternative hypothesis 𝐻!: 𝛼! < 0. The expectation is that 𝛼! is negative but less than 𝛼! in absolute magnitude (Rajgopal et al., 2006). Rejecting the null hypothesis gives evidence that external shocks are partially filtered out from the performance of the own firm in the CEO compensation. For the strong-form test of RPE, the expectation is that 𝛼! is negative and equal in magnitude to 𝛼! (said differently, 𝛼! + 𝛼! = 0). The
prediction for this test is that the market-wide component will be filtered out completely when setting compensation.
To find out what impact the four variables mentioned earlier have on the weight given to peer performance in setting compensation, interaction terms are created of these variables with peer performance (Rajgopal et al., 2006; Aggarwal & Samwick, 1999). Such a model is called an interaction regression model. By creating an interaction term of a variable with peer
performance, 𝛼! will not be the only effect of peer performance on total compensation anymore. The effect of peer performance on compensation will now also depend on the coefficient of the interaction term and the value of the variable that is interacted with peer performance (Stock & Watson, 2012). To show this in formulas, we have to compare equation (1) with the following equation:
𝐶𝐸𝑂𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛! = 𝑐!+ 𝛼!𝐹𝑖𝑟𝑚𝑃𝑒𝑟𝑓!+ 𝛼!𝑃𝑒𝑒𝑟𝑃𝑒𝑟𝑓!+ 𝛼!𝑃𝑒𝑒𝑟𝑃𝑒𝑟𝑓! × 𝐹𝑎𝑐𝑡𝑜𝑟
In equation (1), the effect of peer performance on compensation is the coefficient 𝛼!. But in equation (1a), the effect of a change in peer performance on compensation is 𝛼!+ 𝛼!×𝐹𝑎𝑐𝑡𝑜𝑟. Thus, in this interaction specification, the effect of peer performance depends on the factor that is interacted with peer performance. If the factor is a dummy variable, the factor can be either zero or one. If the dummy variable equals zero, the effect of peer performance will be only 𝛼!. If the dummy variable equals one, the effect of peer performance on compensation is 𝛼!+ 𝛼!. This means that 𝛼! in equation (1a) is the difference in the effect of peer performance on compensation for observations where the dummy variable equals one versus observations where the dummy variable equals zero (Stock & Watson, 2012).
To estimate if industry competition has an effect on the use of RPE (i.e. the weight given to peer performance relative to own firm performance in setting compensation), I created an interaction term with the Herfindahl-Hirschman Index (HHI) and the peer performance. The HHI is an inverse proxy for industry competition and is calculated by using company sales within each two-digit SIC industry (per year). A higher HHI means more industry
concentration and thus less competition. To test Hypotheses I: Firms operating in a more concentrated industry use less RPE, we have to test if the coefficient of the interaction term is positive. The regression equation for testing Hypothesis I is the following:
𝐶𝐸𝑂𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛! = 𝑐!+ 𝛼!𝐹𝑖𝑟𝑚𝑃𝑒𝑟𝑓!+ 𝛼!𝑃𝑒𝑒𝑟𝑃𝑒𝑟𝑓!+ 𝛼!𝐻𝐻𝐼! × 𝑃𝑒𝑒𝑟𝑃𝑒𝑟𝑓!+ 𝛼!𝐻𝐻𝐼!
+𝛼!𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!+ 𝛼!𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝜀! (2) Since a higher HHI indicates a higher industry concentration and thus less industry
competition, a positive coefficient of the interaction term (𝛼!) will tell that firms in more concentrated (and thus less competitive) industries use less RPE. Vice versa, a negative coefficient indicates that the higher the industry concentration, the higher the use of RPE. An interaction term with the book-to-market ratio (inverse proxy for measuring growth opportunities) and peer performance will test Hypothesis II: Firms with high growth
opportunities are more likely to use RPE. The regression equation for this test is the following: 𝐶𝐸𝑂𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛! = 𝑐!+ 𝛼!𝐹𝑖𝑟𝑚𝑃𝑒𝑟𝑓!+ 𝛼!𝑃𝑒𝑒𝑟𝑃𝑒𝑟𝑓!+ 𝛼!𝐵𝑇𝑀! × 𝑃𝑒𝑒𝑟𝑃𝑒𝑟𝑓!+ 𝛼!𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!
+𝛼!𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝜀! (3)
A higher book-to-market ratio indicates fewer growth opportunities. Consequently, the interaction term would assume a positive coefficient 𝛼! if firms with fewer growth opportunities use less RPE. For Hypothesis II to be true, we expect that the coefficient is positive because this indicates that if the growth opportunities of a firm increase
(book-to-market ratio becomes smaller), the firm uses less RPE.
To test Hypothesis III: Large firms use less RPE, I created an interaction term consisting of the size of the company and the performance of the peers. This gives the following regression equation:
𝐶𝐸𝑂𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛! = 𝑐!+ 𝛼!𝐹𝑖𝑟𝑚𝑃𝑒𝑟𝑓!+ 𝛼!𝑃𝑒𝑒𝑟𝑃𝑒𝑟𝑓!+ 𝛼!𝑆𝑖𝑧𝑒! × 𝑃𝑒𝑒𝑟𝑃𝑒𝑟𝑓!+ 𝛼!𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!
+𝛼!𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝜀! (4)
Hypothesis III is true if the coefficient of the interaction term is positive. This would indicate that the larger the company, the less RPE. If the coefficient is negative, it tells us that larger firms use more RPE.
The contribution of this paper to existing research is the investigation of the difference in the degree of use of relative performance evaluation between controlled and non-controlled companies. To test the impact of corporate governance on the use of RPE, I will test if controlled firms use less RPE than non-controlled firms. To perform this test, I created a dummy variable that equals one if a company is controlled. As discussed in the literature review, Kamonjoh (2016) analyzed the S&P 1500 to find out which companies are controlled companies. Appendix A shows the list of companies that were controlled companies in 2012 and were still controlled in 2016. I manually assigned the data of these controlled companies into a group to create the dummy variable. To test Hypothesis IV: Controlled firms use less RPE compared to non-controlled firms, we have to look at the interaction term consisting of peer performance and the dummy variable that equals one if a company is controlled. I will estimate the following equation:
𝐶𝐸𝑂𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛! = 𝑐!+ 𝛼!𝐹𝑖𝑟𝑚𝑃𝑒𝑟𝑓!+ 𝛼!𝑃𝑒𝑒𝑟𝑃𝑒𝑟𝑓!+ 𝛼!𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑙𝑒𝑑! × 𝑃𝑒𝑒𝑟𝑃𝑒𝑟𝑓!
+𝛼!𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!+ 𝛼!𝐷𝑢𝑚𝑚𝑖𝑒𝑠 + 𝜀! (5) A positive coefficient of the interaction term says that if a company is controlled, the use of RPE will become less, which is why the expectation is that 𝛼! is positive. The coefficient 𝛼! in equation (5) is the difference in the effect of peer performance on compensation for controlled companies versus non-controlled companies (Stock & Watson, 2012).
To test for robustness, I will re-estimate these equations for three alternative industry definitions: one-digit, three-digit and four-digit SIC industry (Gibbons & Murphy, 1990).
4. Data and descriptive statistics
The data for this research were obtained from the CRSP-Compustat merged database,
ExecuComp and the Center for Research in Security Price (CRSP). The ExecuComp database contains data of total compensation for the top five executives at each of the firms in the S&P 1500. It includes measures of short-term compensation, like salary and bonus, and long-term compensation (for example long-term incentive plans, stock appreciation rights and restricted stock). The CRSP-Compustat merged database contains fundamentals and company data, so this database is used to find Standard Industry Classification codes (SIC), sales, shares outstanding, fiscal year closing price and stockholders equity. Stock return data are obtained from the CRSP monthly stock files. With these returns, the 12-month buy-and-hold stock returns for every year from 2010 until 2015 are calculated.
The sample that is used in this research contains yearly panel data from 2010 to 2015 of the companies in the S&P 1500. The initial dataset from Execucomp contains 63871 executive-firm years and 2197 companies. Observations for non-CEO’s and observations with missing compensation data were dropped from this dataset. After merging this dataset with the dataset from Compustat and CRSP, the observations with missing data from these databases were dropped too. Finally, observations with CEO tenure less than one year and observations with less than two peers were excluded from the sample. The final sample contains 9144 CEO-year observation for 1937 firms. Table I shows the sample selection process.
Table I
Sample selection
The starting sample of 2197 firms and 63871 executive-firm years with compensation data consists of executives of all S&P 1500 companies during the period 2010 to 2015 in Execucomp. After eliminating data for non-CEO’s, observations with missing data, observations with a CEO tenure of less than one year and observations with less than two peers, the final sample contains 1937 firms and 9144 CEO-firm years.
Firms Observations
2010 to 2015 Execucomp firm-years 2197 63871
Less: non-CEO executives 6 52196
Less: missing compensation data 129 1148
Less: missing Compustat data 23 106
Less: missing CRSP data 0 8
Less: observations with CEO tenure less than one year 29 590 Less: observations with less than two peers 73 679
Table II shows descriptive statistics for the compensation data, performance data, number of peers and other variables like CEO and firm characteristics. Panel A contains the summary statistics for the compensation data. Total compensation is comprised of salary, bonus, other annual, total value of restricted stock granted, total value of stock options granted, long-term incentive payouts and all other total compensation. The number of observations is 9144. The average total compensation for the CEOs in the sample is $6.16 million, with a median of $4.27 million. The average CEO receives a base salary of $855,000, which is less than 15% of the average total compensation of the CEOs. This indicates that the largest part of CEO compensation comes from performance-based pay. The mean of the bonus compensation is $218,000, with a median of $0. This means that about 50% of the CEOs does not receive a bonus. The dollar value of annual stock grant is on average higher than the dollar value of annual option grant ($2.517 million vs. $1.057 million). When the stock market is volatile, option grants can be valued less than the employee cost, which then makes them worthless (if the share price falls more than expected). Stock grants always retain at least some value because the employee did not buy the outright.
Panel B provides the performance data for the S&P 1500 companies and their peer firms. As said before, these returns are the 12-month buy-and-hold stock returns for every year from 2010 until 2015. For the peer firms, the performance is the average 12-month buy-and-hold stock return for the firms in the same industry and the same size quartile, excluding the own firm. As shown in Panel B, the average return over the years was 17%.
Panel C shows the number of peer firms in the industry for a two-digit, three-digit and four-digit SIC industry. The minimum number of peers in the 2-digit SIC industry is two because I dropped observations with less than two peers, following Albuquerque (2009). Panel C shows that the number of peers is the highest in the 2-digit SIC industry, which is logical since the number of companies in an industry becomes more when it is based on less digits. The average number of peers for the 2-digit SIC industry is 19. The maximum number of peers is 50, which belongs to the Business Services industry (SIC-code 73). Other
industries with a lot of peer firms are the Electronics industry and the Depository Institutions (SIC-codes 36 and 60 respectively).
Panel D shows firm and CEO characteristics. Market capitalization is calculated by multiplying common shares outstanding with the fiscal year closing price. The median of the market capitalization of the sample is $2.15 billion and the mean is $10.7 billion. Companies with a market capitalization of between $300 million and $2 billion are generally classified as small-cap companies. The lower quartile is $1.89 billion, indicating that more than 25% of the
Table II
Descriptive Statistics
This table presents the descriptive statistics for S&P 1500 firms for 2010 until 2015. Panel A presents the compensation data for Chief Executive Officers, obtained from ExecuComp. Total compensation from ExecuComp is comprised of salary, bonus, other annual, total value of restricted stock granted, total value of stock options granted, long-term incentive payouts and all other total compensation. All compensation variables are in thousands of dollars. Panel B provides performance data. The return data are obtained from the CRSP-Compustat merged database. Panel C shows the number of peers per industry and size quartile.
Panel D provides summary statistics for CEO and firm characteristics for the S&P 1500 firms. The CEO characteristics data are obtained from ExecuComp. Sales, common shares
outstanding, fiscal year closing price and book value of equity are taken from the CRSP-Compustat Merged database. Market capitalization (measured in millions of dollars) is calculated by common shares outstanding times the fiscal year closing price. The book-to-market ratio is calculated by the book value of equity divided by the book-to-market capitalization. The Herfindahl-Hirschman Index (HHI) is created with company sales for every 2-digit SIC industry. The higher the HHI, the more concentrated the industry. Exec/Dir is a dummy variable that equals one if the executive served as a director during the fiscal year.
Panel A: Compensation data ($ thousands)
Variable N Min p25 Median Mean St. dev p75 Max
Salary 9144 26.00 582.48 800.00 854.68 471.46 1,000.00 8,100.00 Bonus 9144 0 0 0 218.59 1,232.28 0 28,500.00 Other compensation 9144 1.29 18.80 63.54 239.45 964.37 190.52 46,347.77 Total compensation 9144 313.22 2,163.69 4,270.64 6,160.58 7,185.79 7,789.60 156,077.90 Stock awards 9144 0 119.25 1323.88 2,517.30 4.070.52 3,446.25 111,915.00 Option awards 9144 0 0 0 1,057.98 2,882.57 1,257.99 90,693.40
Panel B: Performance data
Variable N Min p25 Median Mean St. dev p75 Max
Firm performance 9144 -0.99 -0.05 0.13 0.17 0.41 0.33 8.51
Peer performance 9144 -0.79 0.01 0.15 0.17 0.25 0.30 4.23
Panel C: Number of peers per industry
Variable N Min p25 Median Mean St. dev p75 Max
No. of peers (2-digit) 9144 2.00 7.00 19.00 19.04 12.57 29.00 50.00 No. of peers (3-digit) 9144 0 2.00 5.00 9.25 10.23 14.00 39.00 No. of peers (4-digit) 9144 0 1.00 3.00 5.70 7.69 8.00 38.00
Panel D: Other Variables
Variable N Min p25 Median Mean St. dev p75 Max
Firm characteristics
Size (market cap.) 9144 1.89 789.40 2,147.51 10,699.30 32,961.29 7,024.24 626,550.40 Book-to-market ratio 9144 -0.47 0.27 0.47 0.54 0.40 0.74 2.10 HHI 9144 0.02 0.05 0.08 0.11 0.08 0.15 0.61 CEO characteristics CEO age 9144 28.00 52.00 57.00 56.69 7.16 61.00 95.00 CEO tenure 9144 1.00 3.67 7.00 9.05 7.50 11.99 61.00 Exec/Dir 9144 0 1.00 1.00 0.97 0.17 1.00 1.00 Interlock 9144 0 0 0 0 0 0 0 Shares owned (%) 9144 0 0.24 0.70 2.31 5.13 1.87 66.60
companies in the sample are small-cap companies. The book-to-market ratio is calculated by the book value of equity divided by the market capitalization. The mean of the book-to-market ratio is 0.54, so on average the book-to-market value of the companies in this sample is higher than the book value of the companies. The upper quartile of the book-to-market ratio is 0.74, indicating that for 75% of the firms the market value if higher than the book value. The data for the book-to-market ratio are winsorized at the top and bottom one percentile. The
Herfindahl-Hirschman Index (HHI) is created based on company sales for every 2-digit SIC industry. The higher the HHI, the more concentrated the industry. The HHI is on average 0.11, which tells us that on average there is quite a lot of competition.
The average CEO age is 56 years and the average CEO tenure is 9 years. The mean of Exec/Dir, which is a dummy variable that equals one if the executive served as a director during the fiscal year, is 0.97. This means that almost all CEOs in the sample also served as a director. The interlock variable (the proportion of the executive team subject to an interlocked relation) is zero for all observations and will therefore not be included in the regressions. The average CEO owns 2.31% of the total shares of the company.
5. Results
In this section, the results of the regressions will be reported. The section tests the five equations presented in the methodology.
Results of testing equation (1) are presented in Table III. It shows that the coefficient of firm performance is positive and significantly different from zero, which indicates that total compensation is positively related to firm performance as measured by the buy-and-hold rate of return on the stock. It implies that CEOs are rewarded if shareholder wealth increases. The firm performance coefficient is 0.039. Since the dependent variable is the natural logarithm of total compensation, this means that if buy and hold return increases by one, total
compensation increases by 3.9%. Said differently, a 10% percentage-point increase in the buy-and-hold stock return increases compensation with 0.39%. To test for weak-form RPE, we have to test the null hypothesis 𝐻!: 𝛼! ≥ 0 against the alternative hypothesis 𝐻!: 𝛼! < 0. The expectation of weak-form RPE is that the coefficient of peer performance is negative, but less than the coefficient of firm performance in absolute magnitude. The coefficient of peer performance is indeed negative and less in absolute value than the coefficient of firm performance and this would indicate weak-form relative performance evaluation. The
coefficient tells us that if peer performance increases by 1%, total compensation decreases by 0.005%. It can imply that companies decrease total compensation of the CEO when peer performance increases because they filter out a positive shock to the market that is beyond the control of the CEO. Vice versa, it can filter out a negative common shock by increasing total compensation if peer performance decreases. However, the coefficient of peer performance in Table III is not significant so there cannot be drawn a conclusion from this. Indicated by the positive significant coefficient, firm size has a positive effect on total compensation of CEOs of companies in the S&P 1500. I divided the market capitalization, initially measured in millions, by 100 so firm size is measured in 100 millions of market capitalization in the regression output. To make it easier to interpret the coefficient, I multiplied the coefficient of firm size by 100. The results indicate that if market capitalization goes up by 100 million, total compensation increases by 0.047%. Furthermore, the regression results show that the higher the book-to-market ratio and thus the lower the growth opportunities, the lower the total compensation. The degree of industry concentration also has a negative impact on CEO compensation. CEO tenure and age, on the other hand, are positively related to total
Table III – OLS regressions for testing RPE
The table shows the results from regressions of total compensation on performance measures, firm characteristics and CEO characteristics for the final sample, which consists of CEOs of all S&P 1500 firms during the period 2010 to 2015. Peer performance is based on companies in the same two-digit SIC industry and the same size quartile. The firm characteristics include the market capitalization and the book-to-market ratio. Industry concentration (measured by the Herfindahl-Hirschman Index) measures the degree of competition the firm faces. The CEO characteristics include CEO tenure, CEO age, the percentage of shares owned by the CEO and a dummy variable that equals one if the executive served as a director during the fiscal year. Total compensation and peer performance are in logarithms. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, ** and *, respectively.
Dependent Variable: Total Compensation Independent Variables Predicted sign (1) Firm performance + 0.039** (0.018) Peer performance - -0.005 (0.006) Control variables Firm size 0.047*** (0.004) Book-to-market ratio -0.220*** (0.031) Industry concentration -0.666*** (0.216) CEO tenure 0.007*** (0.002) CEO age 0.004** (0.002) Exec/Dir 0.084 (0.065) Number of observations 6968 R2 0.24
Year dummies Yes
Industry dummies Yes
Test strong RPE: 𝛼!+ 𝛼! = 0 (p-value) 0.046
that if the CEO served as a director during the fiscal year, it increases total compensation with 8.4%. However, this coefficient is not significant.
Table IV presents the results of the OLS regressions of equations (2), (3), (4) and (5). From these regressions, there can be investigated if certain variables increase or decrease the weight given to peer performance relative to own firm performance in setting compensation (i.e. RPE). Column (2) tests equation (2) and this equation adds an interaction term of the Herfindahl-Hirschman Index (HHI) with peer performance to equation (1). The table shows
Table IV – OLS regressions for testing RPE: four explanations
The table shows the results from regressions of total compensation on performance measures, firm characteristics and CEO characteristics for the final sample of 9144 CEO-years, which consists of CEOs of all S&P 1500 firms during the period 2010 to 2015. Peer performance is based on companies in the same two-digit SIC industry and the same size quartile. The firm characteristics include the market capitalization and the book-to-market ratio. Industry concentration (measured by the Herfindahl-Hirschman Index) measures the degree of competition the firm faces. The CEO characteristics include CEO tenure, CEO age, the percentage of shares owned by the CEO and a dummy variable that equals one if the executive served as a director during the fiscal year. Equation (2), (3), (4) and (5) include interaction terms of peer performance with respectively the Herfindahl-Hirschman Index, the book-to-market ratio, firm size and a dummy variable that equals one if a company is
controlled and zero otherwise. Total compensation and peer performance are in logarithms. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, ** and *, respectively.
Dependent variable: Total Compensation
Independent variables (2) (3) (4) (5) Firm performance 0.039** (0.018) 0.047* (0.026) 0.038** (0.016) 0.066*** (0.018) Peer performance -0.010 (0.009) -0.028** (0.013) -0.005 (0.006) -0.006 (0.006) Industry concentration × peer performance 0.068
(0.066)
Book-to-market ratio × peer performance 0.050*** (0.017)
Firm size × peer performance 0.003**
(0.002)
Controlled × peer performance 0.066***
(0.021) Firm size 0.047*** (0.004) 0.045*** (0.006) 0.053*** (0.005) 0.048*** (0.004) Book-to-market ratio -0.221*** (0.031) -0.043* (0.030) -0.093*** (0.015) -0.078*** (0.018) Industry concentration -0.557** (0.240) -0.590* (0.313) -0.743*** (0.217) -0.631*** (0.215) CEO tenure 0.007*** (0.002) -0.002 (0.002) 0.046*** (0.013) 0.051*** (0.012) CEO age 0.005*** (0.002) 0.006** (0.003) 0.105 (0.100) 0.001 (0.002) Exec/Dir 0.084 (0.065) 0.072 (0.091) 0.089 (0.067) 0.070 (0.065) Controlled -0.086 (0.227) Number of observations 6968 6963 6791 6969 R2 0.18 0.24 0.24 0.24
Year dummies Yes Yes Yes Yes
Industry dummies Yes Yes Yes Yes
that the coefficient of own firm performance is positive and significant at the 5% level, and the coefficient of peer performance is negative but not significant. The negative coefficient of peer performance is lower in absolute value than the coefficient of own firm performance, which gives evidence for weak-form RPE. The coefficient of interest in equation (2) is the coefficient of the interaction term between the industry concentration (measured by the HHI) and the performance of peers. For hypothesis I to be true, this coefficient should be positive. The coefficient is indeed positive but not significant. The positive coefficient indicates that the higher the HHI and thus a higher industry concentration and less industry competition, the less companies use RPE. It demonstrates that it may be optimal to soften RPE incentives in concentrated industries. In a highly concentrated industry, a lower emphasis on RPE can be favorable to avoid excessive competition between rival managers. In this way, companies in a concentrated industry can maximize joint (industry-wide) returns. However, the coefficient of the interaction term is not significant. Hence, the results reveal no systematic pattern for the effect of industry competition on the weight given to peer performance relative to own firm performance in setting CEO compensation. Furthermore, the results of testing equation (2) show that total compensation is positively affected by firm size, CEO tenure and CEO age since all these coefficients are positive and significant at the 1% level. For example, if the age of the CEO increases by one year, CEO total compensation increases by 0.5%.
Column (3) of Table IV tests the impact of growth opportunities on the use of RPE. Hypothesis II states that companies with high growth opportunities are more likely to use RPE. Said differently, for companies with high growth opportunities, the weight given to peer performance should be lower than for companies with fewer growth opportunities. This hypothesis would assume a positive coefficient of the interaction term between the book-to-market ratio (BTM) and the peer performance, since the book-to-book-to-market ratio is an inverse proxy for growth opportunities. Table IV shows a positive significant (10% significance level) coefficient for the effect of firm performance on total compensation and a negative significant (5% significance level) coefficient for the effect of peer performance on total compensation. Since peer performance and total compensation are both measured in logarithms, a 1% increase in peer performance causes a decrease of 0.028% in total compensation. The coefficient of peer performance is lower in absolute value than the coefficient of own firm performance (0.028 < 0.047), indicating weak-form RPE. The
coefficient of the interaction term between the book-to-market value and peer performance is positive and significant at the 1% level. This states that the lower the growth opportunities of a firm (the higher the book-to-market ratio), the less companies use RPE. Vice versa,
companies with high growth opportunities are more likely to use RPE. As explained in the methodology, the effect of a change in peer performance on compensation in equation (3) is 𝛼!+ 𝛼! × 𝐵𝑇𝑀. In this case, that is −0.028 + 0.050 × 𝐵𝑇𝑀. Because the coefficient of the interaction term (𝛼!) is positive, the effect on log total compensation of an additional unit increase in log peer performance is more positive, by the amount of the coefficient (0.050), for each additional unit increase in the book-to-market ratio.
Results of the regression of equation (4) are presented in column (4) of table IV. Recall that Hypothesis III is: Large firms use less RPE. The coefficient of the interaction term of firm size (measured by market capitalization) with peer performance is positive and
significant at the 5% level. The effect of a change in peer performance on total compensation is −0.005 + 0.003 × 𝐹𝑖𝑟𝑚 𝑠𝑖𝑧𝑒 according to the results. The effect on log total
compensation of an additional unit increase in log peer performance is more positive by 0.003 for every 100 million increase in the market capitalization of a company. This implies that larger firms indeed give a less negative weight to peer performance in setting CEO
compensation and thus are less likely to use relative performance evaluation. An explanation for this is the following. Since the supply of talented CEOs is relatively inelastic, it may be optimal to reward these talented CEOs for common shocks in the market if these shocks increase the market value of the firm and the outside employment opportunities of the CEO. Rajgopal et al. (2006) found that firms are less likely to use RPE for more talented CEO’s. Because Himmelberg and Hubbard (2000) argue that more talented CEOs manage larger firms, the results of this research are consistent with the finding that companies with more talented CEOs and thus larger companies are less likely to use RPE.
Column (5) of table IV shows the results of regressing total compensation on firm performance, peer performance, an interaction term of a dummy variable for controlled firms with peer performance, control variables and industry and time fixed effects. Recall that the hypothesis for this regression is that controlled firms use less RPE since they have weaker corporate governance and thus allow more pay-for-luck. Besides this, well-governed firms limit the CEO’s ability to capture the pay process The coefficient of own firm performance is positive and significant at the 1%. This coefficient tells that if the buy-and-hold return
increases by one unit, the total CEO compensation (measured in logarithms) increases with 6.6%. The effect of peer performance on total compensation is negative. However, this
coefficient is not significant. The interaction term between the controlled dummy variable and the performance of peers causes that the coefficient of peer performance is not the only effect of peer performance on total compensation if a company is controlled. Recall from the
methodology that the effect of peer performance on total compensation is 𝛼!+ 𝛼! × 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑙𝑒𝑑. If a company is controlled, then the dummy variable equals one, causing that the effect of log peer performance on log total compensation is −0.006 + 0.066 × 1. If a company is non-controlled, then a 1% increase in peer performance will cause a decrease in total compensation of -0.006%. The coefficient of the interaction term is positive and
significant at the 1% level. The positive coefficient tells us that if the dummy variable equals one (if the company is controlled), the effect of peer performance on total compensation will become more positive. This indicates that controlled firms are less likely to use RPE, as was expected. Furthermore, the results in column (5) show that firm size, CEO tenure, CEO age and the dummy variable for CEO’s who served as a director have a positive effect on total compensation, while the book-to-market ratio and industry concentration have a negative effect on total CEO compensation. For example, if the book-to-market ratio increases by one unit, implying fewer growth opportunities for the company, total compensation decreases by 7.8%.
Overall, the results support all four hypotheses, but not all the coefficients are significant. Results are strongest for hypothesis II and hypothesis IV, indicating that firms with high growth opportunities are more likely to use RPE and controlled firms use less RPE compared to non-controlled firms.
6. Robustness checks
This section checks for robustness of the results. To test for robustness, the five equations tested will be re-estimated for two alternative industry definitions: one-digit and three-digit SIC industry (Gibbons & Murphy, 1990). Table V shows the outcome of testing equation (1) for one-digit SIC and three-digit SIC industries.
Table V – OLS regressions for testing RPE for one- and three-digit industries
The table shows the results from regressions of total compensation on performance measures, firm characteristics and CEO characteristics for a sample that consists of CEOs of all S&P 1500 firms during the period 2010 to 2015. Firm performances are 12-month buy-and-hold stock returns and peer performances are 12-month buy-and-hold stock returns based on companies in the same size quartile and the same industry (one-digit SIC or three-digit SIC). The firm characteristics include the market capitalization and the book-to-market ratio.
Industry concentration (measured by the Herfindahl-Hirschman Index) measures the degree of competition the firm faces. The CEO characteristics include CEO tenure, CEO age and a dummy variable that equals one if the executive served as a director during the fiscal year. Total compensation and peer performance are in logarithms. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, ** and *, respectively.
Dependent Variable: Total Compensation Independent Variables
Predicted
sign one-digit SIC three-digit SIC
Firm performance + 0.028* (0.017) 0.062*** (0.021) Peer performance - 0.019*** (0.007) -0.001 (0.007) Control variables Firm size 0.051*** (0.004) 0.043*** (0.004) Book-to-market ratio -0.218*** (0.029) -0.199*** (0.036) Industry concentration -0.469 (0.613) -0.443*** (0.169) CEO tenure -0.001 (0.002) 0.002 (0.002) CEO age 0.004** (0.002) 0.001 (0.002) Exec/Dir 0.097 (0.065) 0.230*** (0.075) Number of observations 7573 5243 R2 0.14 0.27
Year dummies Yes Yes
