Pay and performance benchmarking in executive compensation : the importance of competitive benchmarking for Chief Executive Officer compensation level and its adjustments

(1)

AMSTERDAM BUSINESS SCHOOL

MSc Business Economics

Master Specialisation Finance

Pay and Performance Benchmarking in Executive Compensation

The importance of competitive benchmarking for Chief Executive Officer

compensation level and its adjustments

Author:

ZV Badzheva

Student number:

10398589

Thesis supervisor: dr. JJ Lemmen

Finish date:

July 2016

(2)

PREFACE AND ACKNOWLEDGEMENTS

I would like to express my deep gratitude to my supervisor, dr. Jan Lemmen, for his continuous guidance and support, insightful comments and valuable recommendations. His meritable and timely advice was highly appreciated throughout the entire research and writing process, and made this final work a graceful closure of my degree.

I would also like to thank my fellow students for the warm support during the time of completing the Master Thesis, and for contributing to make the whole Master Program a memorable and pleasant experience.

Statement of Originality

This document is written by Student Zlatina V Badzheva, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

ABSTRACT

The research provides empirical evidence on the way Chief Executive Officer (CEO) compensation is affected by benchmarking, focusing on the period after the SEC changes in disclosure regulation of 2006, using US S&P500 firms and simulated peer groups. The findings indicate that the level of CEO pay and its subsequent increases are a function of the relative CEO remuneration, but not the relative firm performance. The less an executive is paid in the prior year in comparison to similar companies, the higher his contemporaneous increase in rewards, regardless of the firm’s position in the performance distribution of the peer group. The results support the notion that benchmarking might indeed be used simply for boosting CEO remuneration and could be part of the reason for the executive pay upward ratcheting in recent years.

Keywords: Executive compensation, Benchmarking, Peer groups, CEO pay JEL Classification: G34, J31, J33

(4)

LIST OF TABLES

Table 1 Independent Variables Summary [11]

Table 2 Descriptive Statistics [17]

Table 3 OLS Regressions with the Peer Group Pay and Performance Medians [21] Table 4 OLS Regressions with the Firm’s Relative Pay and Performance Positions [25] Table 5 OLS Regressions with the Firm’s Relative Pay and Performance Positions (CDFs) [27] Table 6 OLS Regressions with the Firm’s Relative Pay [34]

Table 7 Variables descriptions [35]

Table 8 Correlation Matrix (Covariates Table 3) [38]

(6)

CHAPTER 1 Introduction

CEO compensation is no longer a matter strictly contained within the board room but a concern for academia, regulators and the large public. Frequently making the media headlines, it has struck heated debates over the need and appropriateness of the seemingly outrageous director pay checks. Attempting to credibly and precisely disentangle the reasons for the growth and composition of compensation, a large part of the corporate governance literature has repeatedly examined manager rewards over the years. The focus, however, has largely been on the effect of the relationships between executives, the board of directors, and shareholders, with very little attention on the impact of firm-to-firm interaction. Nevertheless, the broader context involving other companies is not to be overlooked, as it is factored in the CEO compensation through methods such as competitive benchmarking. Often cited as one of the leading causes of the executive pay upward ratcheting in recent years, the process of using a peer group pay-level information for firm’s own remuneration decision is as controversial as it is common in practice. It accounts for numerous unsettled discussions due to the ambiguity of both its incentives and implementation.

On the one side, it can be seen as a cover for simply inflating managerial pay, if firms look only at the compensation levels of the comparison universe when deciding on remuneration. On the other side, it might be used as an effective mechanism for determining pay, which will not only retain but also motivate the CEO to exert effort, as long as other group indicators are used for comparison, such as performance levels. Crucial for the efficient use of benchmarking and its assessment is the selection of comparables, which is also the main topic of most academic work on the subject.

This research focuses on examining the effects of relative CEO pay and performance on the annual chief executive compensation and its adjustments, using simulated peer groups. The firm’s overall performance is a standard determinant of remuneration but understanding whether its level in relation to a benchmark is also material will shed new light on the current debate. Learning in further detail how pay is set in practice will be beneficial for policy makers as well as company shareholders and academics.

The research question is “How important are pay and performance benchmarking for CEO compensation level and its adjustments?”. Building on previous empirical work and various economic theory predictions, the following two hypotheses will be examined:

1. Pay relative to the peer group median has a higher effect on CEO compensation than firm performance relative to the median.

(7)

2. Firms with pay below the median see higher compensation adjustments in the following year than firms above the median.

The paper aims at making a three-fold contribution to the existing research. First and foremost, through testing the simultaneous effects of the peer pay and peer firm performance on CEO remuneration and its changes, rather than studying the relations separately. Second, it uses a more comprehensive set of controls for the governance structure and director characteristics, thus attempting to estimate the impact of the comparison group with increased precision. A large part of the benchmarking literature so far has focused on a single set of features (firm, board or CEO specifications) rather than a combination of the three, which gives scope for further investigation. Lastly, this study will make use of post-2008 data, thus taking into account the 2006 change of the SEC disclosure regulation.

The rest of the paper is structured as follows: Section 2 discusses the current benchmarking practice and debate, and the fundamental theories behind it. Section 3 describes the data and the empirical approach. Following is the presentation and discussion of findings in section 4. Section 5 concludes and discusses limitations and recommendations for future research.

(8)

CHAPTER 2 Literature Review

2.1 Executive Compensation and the Benchmarking Practice

“Advertising company WPP chief’s £60m pay set to spark anger: including the 2015 pay scheme, Sr Martin Sorrell’s remuneration in the last 5 years will add up to £150m.” (Vandeveld & Oakley, 2016).

Executive1 compensation determination, albeit not a central task in a company’s operations, has managed to spark discussions more extensive than the majority of other corporate practices. The debates diverge in two main directions – what are the reasons for the current colossal pay checks and how much the managers should actually be receiving. Being at the top of the corporate ladder, the CEO’s labour is naturally rewarded higher than the average employee within the firm. The differentials are usually justified as a compensation for the costs of managing a complex organization, having higher responsibilities, effort levels, turnover risk and others (Core et al. 1999). They have, however, largely risen in the last few decades together with the soaring up of overall executive pay levels. As Goergen and Renneboog (2011) illustrate, it is a worldwide phenomenon, not discriminating between industries and sectors. Some attribute it to the forces governing the market for managerial talent and the increased return to CEO ability in recent years. The organizations, ever growing in size and complexity, require an ever rising level of CEO skill for overseeing them, and this instinctively has a boosting effect on compensation (Cremers & Grinstein, 2014). Supporting this view, Gabaix & Landier (2006) document a six fold growth in CEO pay between 1980 and 2003, which they mainly attribute to the preceding six fold increase in the average firm size and the low dispersion and scarcity of CEO talent. On the opposite end, critics blame the raises largely on corporate governance failures, such as managerial power, asymmetric information or transaction costs of executive turnover and the effect they have on wage determination (Bebchuk & Fried, 2003).

Nevertheless, CEO remuneration setting is usually a standard procedure, and theoretically very transparent for the US publicly listed companies. Generally, it begins with the HR department making a pay proposition, later to be reviewed and adjusted by the board of directors’ compensation committee and externally hired consultants (Murphy & Sandino, 2010). Additional practices, increasing in popularity, however, significantly broaden the scope for discretion. Competitive benchmarking is amongst the most prevalent, using a selection of companies for information gathering on pay levels and structures. It is the process of choosing similar firms that organization measure themselves against whenever they are determining compensation. The majority of the entities

1

(9)

employing it anchor all components of the CEO remuneration to this peer group in some manner. The target total reward is typically the median (50th percentile) of the comparison universe or higher, and the bonuses, stock plus option pay and other rewards are directly or indirectly tied to the metric. The adoption of benchmarking is seen either as enhancing the power of a CEO’s reward package for attracting, incentivizing and retaining employees, or as contributing to furthering the misconduct of pay setting. Understanding how firms choose their peer group is informative of the role benchmarking plays in determining the CEO compensation. Hypothetically, the goal is to make a selection, which resembles the firm as closely as possible in order for the compensation be effective in its role. A large number of companies exhibit an analogous trend when selecting a benchmarking universe, most frequently considering industry and size criteria. An entity’s own industry is an intuitive place to search for comparables because the firms providing similar products or services and competing for customers are likely to be the main competitors for executive talent as well. As size gauges the complexity of the organization and its scope, it is also a logical choice for a yardstick (Bizjak et al., 2011). Nevertheless, it is possible that firms select peers on the basis of other considerations, such as geographical location or accounting standards (Bizjak et al., 2008).

Bijzak et al. (2008) report that 96 out of 100 randomly selected companies indicated that they base executive’s wage on the analysis of comparable firms. Despite the fact that the practice is widely adopted and its impact largely recognized, a significant portion of the corporate governance literature does not account for it when examining a CEO pay package. It was only in the last decade that academia focused on competitive benchmarking, which also coincides with the change in SEC’s 2006 disclosure regulation. According to the latest US requirements as of August 2006, public firms are obliged to report their peer group members if they impact compensation setting. A comparison of pre- and post-amendment groups (between 2005 and 2006), shows that approximately 25% of them saw a considerable alternation in the selection. The rule itself sparks a separate discussion concerning its overall effect on the benchmarking practice and consequently on remuneration. Evidence is mixed – some argue that the increased disclosure leads to less bias in the peer group selection (Bizjak et al., 2008), while others see it as an easier way to justify higher levels of CEO pay (Faulkender & Yang, 2012).

2.2 The Benchmarking Debate

Despite being widely used, benchmarking is just as controversial as almost any other practice related to executive pay setting. Critics and proponents argue on ideologically different grounds - on the one side, the current pay methods are being justified as a reflection of the equilibrium of the managerial

(10)

labour market and on the other they are seen as the result of entrenched CEOs (Faulkender & Yang, 2010).

The opponents attest that the practice creates perverse remuneration incentives, and usually present concerns related to corporate governance failures and the abuse of managerial power for self-servicing. As Bebchuk and Fried (2003) find, a combination of a powerful CEO and a co-opted board can result in an opportunistically chosen peer group, such to inflate CEO pay. In this case, benchmarking to firms with highly paid executives acts as a justification of a company’s choice of an outrageous pay package. Albuquerque et al. (2013) also find that the selection of the CEOs in the peer group often consists of managers with a high salary, in order to justify a higher salary for their own CEO. The SEC requirements of 2006 for a higher transparency of peer groups are aimed at limiting exactly this kind of behavior, but how successful they are in achieving it is still undetermined. The rationalization of the selection is that these firms are the primary competition for managerial talent. Besides inflating a single firm’s CEO pay, benchmarking is also blamed for the overall upward ratcheting of compensation levels over time (Faulkender & Yang, 2012). The ‘‘Lake Wobegon Effect’’ is cited as one of the reasons for the exponential increase – firms are not willing to admit having a below-average CEO, hence they are not willing to pay below the median of comparable companies. This way, the wage is used as a signalling device to affect the value of the firm, as well as boost pay, and meanwhile leads to the negative externality of an upward spiral (DiPrete et al., 2010). As former DuPont CEO Edward S. Woolard Jr. summarizes in his 2002 Harvard talk: “The main reason compensation increases every year is that most boards want their CEO to be in the top half of the CEO peer group because they think it makes the company look strong. So when Tom, Dick, and Harry receive compensation increases in 2002 I get one too, even if I had a bad year. “ (Elson, 2003). This relates to another issue of using peer groups for reward determination – it can potentially lead to increases in target-based components not justified by the actual performance exhibited. Examining proxy statements, Bizjak et al. (2008) report that the majority of performance-based rewards, such as bonus and stock and option grants, were awarded without justifying their values by meeting or exceeding the peer group performance indicators. An example is Amoco Inc., which bases its pay increases on the need to reach the median for the industry, despite the fact that it has underperformed its peers. Albuquerque et al. (2013) also support the findings that the compensation is not tied to the firm performance as it should be. On a related note, Core et al. (1999) show that paying a CEO too much will hurt firm performance – the authors find a negative relation between an executive’s remuneration and the subsequent company results.

According to the opposite stance, benchmarking is economically vital for the compensation determination process, as efficiently structured peer groups are indicative of the level of competitive pay necessary for retention and incentivizing of the CEO. Thus, pay packages are simply a reaction to

(11)

market forces such as supply and demand for executive talent. Operating in a competitive labour market, information on the CEO’s outside options (the prevailing market wage), is crucial for constructing rewards consistent with the goal of preserving valuable human capital (Albuquerque et al., 2012). As Holmstrom (2004) argues, wage is a price for labour and as any other economic price it cannot be objectively determined by solely focusing on intrinsic value and disregarding the price of equivalent products. Holmstrom and Kaplan’s (2003) simple conclusion is that ‘‘we need more effective benchmarking not less of it.”

The practice is controversial and criticized on all levels from its primary objectives to the final execution. As the peer group selection is fundamental for the outcomes, it has been the major focus of most of the academic work on competitive benchmarking. Findings and conclusions have, however, been only partially informative and largely ambiguous. Faulkender & Yang (2010) provide support for the wide spread belief that comparison groups contain an upward compensation bias. Examining newly disclosed peers after the 2006 change in regulation, they show that the selection made was largely due to the need to justify boosts in remuneration. Others hypothesize that even if the comparison universe features wages above the firm’s, the board of directors could exercise discretion and mitigate the bias effects. Taking this stance, Bizjak et al. (2011) argue that pay is only partially adjusted in response to the differential with the peer group and it is mainly a result of the standard accounting and performance indicators. The authors also find very limited evidence for the actual existence of the bias in constructing benchmarking groups, and conclude that even if there is any opportunistic behaviour in choosing peers, the benefits to the CEO pay are relatively small.

Taking a different perspective on the peer group bias, part of the research is dedicated to disentangling the motivation behind comparing to higher compensated CEOs. Despite the blame that benchmarking is primarily used for personal benefits, Albuquerque et al. (2013) present evidence that it is largely a compensation for managerial talent. They calculate a “peer pay effect” as the difference between the compensation of the actual and the “efficiently constructed” peer groups, and split the effect into a self-servicing component and a compensation for talent. Results indicate that even though actual peers are subjectively chosen and have a different mean compensation used for benchmarking, the difference is a reward for CEO ability. Complementing Bizjak et al. (2011)’s finding, the authors suggest that the benchmarking practice is indeed driven by the company’s objective to attract and retain an executive and not a result of a weak corporate governance. Proxying talent by abnormal stock and accounting results, their results suggest that the peer group performance is also factored in when determining compensation (Albuquerque et al., 2013). Examining S&P1500 firms after the 2006 change in SEC disclosure rules, Gong et al. (2011) confirm that benchmarking in order to gauge a CEO reward package is often accompanied by either an implicit or explicit relative performance comparison. Theoretically, the agent (CEO) and the principal (board of directors) share the benefits -

(12)

as peer performance includes the common exogenous shocks, it filters out the effect of systemic risk on the agent’s pay, while providing the principal with additional information on ability (Gong et al., 2011). In practice, however, the advantages prove to be limited because this provides scope for further misconduct of remuneration determination. Even if there is no apparent upward pay bias in the comparison universe, there might be a strategic choice of underperforming peers, which would artificially boost the firm’s relative performance and raise the executive compensation (Aggarwal & Samwick, 1999). An organization’s choice of comparables might be biased towards either higher paying companies or underperforming ones, depending on whether the effect of benchmarking the pay or the performance to the median is stronger (Gong et al., 2011).

2.3 Theories

Competitive benchmarking’s primary aim is to facilitate the designing of a CEO compensation package which would ensure attracting, retaining and incentivizing of executives (Aggarwal & Samwick, 1999). The need for a peer group comparison of wages and performance is condoned by a number of behavioural and labour economic theories as well as sociological and psychological arguments.

Support for the necessity of a uniform wage level in order to retain managers is found in relative deprivation theory, as discussed in Cowherd and Levine (1992). Agents experience deprivation if they are underpaid in comparison with a reference group, which might result in them leaving the company. Reference groups could include employees or managers with the same or different job specification within the firm. The findings are closely related to the predictions of the equity theory of human behaviour in social exchange, which builds on the idea of “fairness” of compensation. It attests that individuals believe everyone should receive a pay reflecting his own contribution. Stated differently, reward distribution should be such that the ratio of effort to wage is similar to the one of a suitable comparable universe. If there is pay inequity, a person would attempt to reduce it by altering effort levels or ultimately leaving the firm. One of the theory’s predictions is that the need for justice is weaker when the executive is relatively overpaid rather than underpaid (Shin, 2013). Testing the proposition, Ezzamel and Watson (1998) show an asymmetric response to peer information – overpaid managers face downward adjustments of much smaller magnitude than the magnitude of the upward adjustments of underpaid CEOs. Building up on equity theory, Akerlof and Yellen (1990) present a “fair-wage” hypothesis, according to which a worker’s performance would deteriorate when their wage falls below the perceived “fair” one. A proposition is that a worker’s perception of a fair-wage system is one where the pay differences are more compressed than the productivity differences. The model illustrates how smaller variation in compensation can benefit a company and facilitate the retention and motivation of executives. Inequity causes distress to the employees which they would

(13)

attempt to reduce by ultimately seeking new employment (Cowherd & Levine, 1992). The tendency of employees to evaluate their position in comparison to a representative group is also explained by Festinger’s social comparison theory (Festinger, 1954). It poses that agents define their own utility relative to a peer universe, as it is impossible to determine it objectively in absolute terms. Thus, the participants in the market for executive talent will set their “market value” and reward expectations in relation to the other employment opportunities available (Göx & Heller, 2008).

With respect to incentivizing, the tournament theory, as discussed by Lazear and Rosen (1981), provides a rationale for using relative performance in determining a CEO’s remuneration. One of the propositions is that relative performance, rather than absolute, should be guiding an employee’s reward for achieving an efficient outcome. The authors argue that this manner of setting wages leads to an increase in executive motivation, since higher productivity is also rewarded higher, and thus to additional benefits for the company. This, however, creates a trade-off between incentivizing and retaining the executive, because the wage differentials that would result might force the CEO to look for other options. A second prediction is that that being below the highest pay level gives you the possibility to win further prizes. Therefore, a manager in the lower half of the compensation distribution has room for larger reward raises in the future (Heyman, 2005).

The argument that benchmarking is a response to the need to compete for executive talent is built on both theoretical and empirical grounds. The standard agency theory and the competitive sorting models both acknowledge the important role the differences in executive ability play in determining compensation. Furthermore, empirical studies building on the models’ predictions show that the variations in CEO human capital affect corporate operations and results, hence the more skilled the executive, the more attractive he is (Albuquerque et al., 2013). Firms are using remuneration as a mean to obtain this desirable set of qualities and as Falato et al. (2011) show, its level reflects the board’s perception of the CEO’s talent and reputation.

Finally, when firms defend their choice of a peer group in proxy statements, they explicitly discuss the competition for CEO talent as one of the leading reasons for the specific selection (Albuquerque et al., 2013).

The aforementioned discussion leads to the development of the following two hypotheses to be empirically examined:

1. Pay relative to the peer group median has a higher effect on CEO compensation than firm performance relative to the median.

2. Firms with pay below the median see higher compensation adjustments in the following year than firms above the median.

(14)

CHAPTER 3 Methodology and Data

To empirically test the effect of benchmarking on CEO rewards and their yearly changes, different specifications of a compensation model are estimated for each hypothesis. Initially, the amount of executive pay is examined solely as a function of each company’s own characteristics, without taking into account external factors. The internal features considered include firm, board and CEO characteristics for each year. Following the creation of a peer group per firm per year, the absolute values of the last year median pay and median performance of these comparison universes are used to augment the original model, in order to test and compare their effects on rewards. After examining the relationship between benchmarking and the compensation level, the effect of a peer group use on the remuneration adjustments from year to year is investigated. The dependent variable is altered from the annual amount of the reward, to the change in CEO pay from year t-1 to t. Computing the differences between a CEO’s pay and the median, and firm performance and the median, the relative positions of each company within its peer group are established. These positions are included in the following regressions as binary indicators, reflecting whether the CEO’s remuneration last year was below or above the median and whether the firm’s last year stock returns were below or above the median. To further assert the results regarding the effect of the relative “underpaying” of an executive on the change in his compensation, the same specification are tested using the dollar value differences in positions (or in other words, the distance from the medians). To manage skewness, the natural logarithm of the variables was taken in all regressions.

Finally, to attest whether relative performance2 or relative pay is more important for setting rewards, the cumulative distribution functions of performance and remuneration are constructed and included as determinants.

The main covariates of interest in all regressions are the peer group related ones: absolute levels of median pay and median performance, the position of the firm and the CEO relative to the medians, and the cumulative distribution functions (CDF).

A detailed description of the estimation process of each model specification, including the creation of the variables and peer groups, is presented below.

2

From here on, the term “(relative) performance” refers to the (relative) stock return values of a company for a given year. The other measure of performance used in the paper (ROA), is referred to as “accounting performance” or “accounting returns”.

(15)

3.1 The Effect of Peer Group Pay and Performance Medians on CEO Compensation

To understand the effect of pay and performance benchmarking on the CEO pay more clearly, the initial estimations aim at explaining compensation without taking into account peer group characteristics. The determinants of executive rewards have been examined extensively over the years and the literature presents a wide spectrum of potential variables 3. Thus, this study relies on a combination of covariates specific to the paper’s framework in question, and on the ones asserted as material for the director’s pay regardless of the setting. The base regression equation is:

Ln(Compensationi,t)= α+ β1 Firm Characteristicsi,t+ β2 Board Characteristicsi,t+ β3 CEO Characteristicsi,t+ εi,t

The primary dependent variable is total annual compensation of an executive measured in US dollars, but annual cash payments (salary plus bonus) and stock and option grants are also examined as robustness checks. The natural logarithm of these variables is taken in all regressions.

Table 1 Independent Variables Summary

Summary of the independent variables used in the estimations of CEO compensation, grouped by characteristics type.

Firm Characteristics Board Characteristics CEO Characteristics

Sales Entrenchment index CEO gender

Market to book Board size CEO is above 60

Leverage % Outsiders CEO age

ROA CEO is chairman CEO tenure

Stock return

Investigating parts of the CEO remuneration independently allows for a better understanding of the impact on the separate components as well. The total cash is used, instead of solely salary or bonus, in an attempt to limit the biasing effect of outside influences, such as varying company and industry rules on the mix of fixed and variable components of short term incentives. The salary and bonus are also usually determined in a complimentary manner and the relative weight put on either often times

3

Tosi et al. (2000) provide a meta-analysis of CEO pay studies, focusing especially on the diversity of reward determinants. Murphy’s (1999) study on executive compensation is frequently used as a foundation for theselection of covariates, butmost papers rely on their own combination and specification of variables, guided by data availability and research direction.

(16)

reflects the company’s strategy. If the absolute performance of the peer group and the relative performance of the firm are significant determinants of executive pay, then they should have the largest impact through the target-based components of compensation, which are the bonus and the stock and option grants. Nevertheless, the total annual reward is the main focus, as most of the times it is the one whose value is directly targeted at the median (or above). It is not necessary for all components of pay to be tied to the peer group characteristics - some might be determined only as a fraction of the overall compensation, hence only indirectly influenced. In this case, the most pronounced effect will be on the total annual remuneration.

The firm characteristics considered are representative of the company’s size (last year sales revenues (Ln(Sales t-1))), investment opportunities (market-to-book ratio (Market to book t-1)) , leverage (total

debt scaled by total assets (Leverage t-1)), and performance (current and prior year return on assets

(ROA and ROA t-1) and current and prior year stock returns (Stock return and Stock return t-1).

From these indicators, the sales are shown to be the most important for compensation, as in Albuquerque et al. (2013)’s investigation. Aligned with this view are also the findings of Gabaix & Landier (2006), who attest that the six-fold increase of CEO pay between 1980 and 2003 is fully attributable to the six-fold growth in firm size. In a later study, focusing on the 2004 - 2011 period, the authors find the same one-to-one trend in both crisis and pre- and post-crisis times (Gabaix et.al., 2014). This is worth noting as it may have implications for the results of the present study, since it also includes the years of the global financial crisis.

For the sales, as well as the different pay measures, the natural logarithm (Ln) values were used in all regressions, instead of their dollar-amounts. The board characteristics are the number of people on the firm’s board of directors (Board size), the percent of outside directors on the board (% Outside directors), whether the CEO is also a chairman of the board and an entrenchment index. The latter variable is constructed similar to Gompers, Ishii and Metrick’s (2003) GIM index of shareholder rights (which is only available until 2006). As previously discussed, director entrenchment is often cited as one of the potential problems with CEO compensation in general and with benchmarking in particular. Borokhovich, Brunarski and Parrino (1997) present evidence in favor of the hypothesis that CEOs’ compensation increases in firms which adopt takeover defenses, thus it is important to examine their and other provisions’ effects on an executive’s rewards. The index constructed in the present study combines six different indicators of director entrenchment, concerning delay, voting and other provisions (golden parachute, poison pill, majority vote required, classified board, blank check preferred and supermajority) and attaches a weight of one to each. The outcome is an ordinal variable

(17)

with value between zero (high shareholder rights) and six (low shareholder rights). The CEO covariates are the binary indicators CEO gender, whether the CEO is above 60 years of age, CEO age, and a continuous measure for CEO tenure.

To test for the effect of including a peer group in the compensation determination, the model is augmented by the values of the last year median pay and median performance of the company’s respective comparison universe. In other words, the specification investigates the importance of the absolute levels of the peer group compensation and the peer group stock returns for the absolute level of CEO pay. Visually:

Ln(Compensationi,t)= α + β1 Median Payi,t-1+ β2 Median Performancei,t-1+ β3 Firm Characteristicsi,t+ β4 Board Characteristicsi,t+ β5 CEO Characteristicsi,t+ εi,t

The lagged values of the medians were used, because the remuneration’s structure and target levels are usually determined a year in advance. Therefore, if the peer group features are material for the CEO pay, then these should be the prior year ones. The results are presented in Table 3.

3.2 The Effect of a Firm’s Relative Pay and Performance Positions on CEO Compensation

The following estimations regard the impact benchmarking has on compensation adjustments from year to year. The main dependent variable is the dollar value change in the total CEO compensation from year t-1 to year t, consequently replaced by the changes in cash payments and in stock and option grants.

Instead of the absolute levels of the peer group independent variables, relative measures are constructed to represent the positions of each executive’s pay in relation to the median and every firm’s performance in relation to the median, for a given year. These pay and performance status indicators take three different forms – as binary variables, as continuous variables and as cumulative distribution functions (CDFs), used in separate regressions. The model in question is:

∆Ln(Compensationi,t)= α+ β1 BelowMedianPay i,t-1 + β2 BelowMedianPerformance i,t-1+ β3 Firm Characteristics i,t+ β4 Board Characteristics i,t+ β5 CEO Characteristics i,t + εi,t

Below Median Pay is a dummy equal to 1 if the CEO was relatively “underpaid” in the previous year (compensation was below the median at t-1), and 0 otherwise. Similarly, Below Median Performance is equal to 1 if the firm underperformed its peers last year (firm stock returns were below the median at t-1), and 0 otherwise. This specification allows to test whether being below the median in a peer

(18)

group has consequences for the value of the changes in a director’s compensation. To mitigate the information-loss effect of using a dichotomous measure, the two dummies were substituted for the continuous variables Difference Pay and Difference Performance, calculated as the dollar-value of the CEO remuneration net of the median reward and the company’s stock returns net of the median stock returns, respectively. The pay distance points out how much more or less the director was awarded last period compared to his industry peers. The regression coefficient will show how the company adjusts the current CEO’s reward as a function of his relative compensation in the prior year. Positive signs on these determinants indicate firm results and director awards above the midpoints of the comparable companies. This will provide more refined information about the effect of relative positions on remuneration and is used as a robustness check.

Finally, based on these measures, the CDFs of relative pay and relative performance are constructed, which allows for a direct comparison of the economic impact of the main covariates of interest, as they would have equivalent standard deviations, means and medians (of approximately 0.5) (Aggarwal & Samwick,1999). The lagged values of the CDFs are used and the higher they are, the better the company performance and the larger the pay in relation to the peer group.

Some of the other determinants in the model are also altered, in order to better fit the fact that the dependent variable in this part is the adjustment of rewards, rather than their total yearly worth. While the absolute amount of sales is material for the absolute value of the executive compensation, the change in remuneration is theoretically better explained by the change in sales. This also holds for using the change in leverage, Market to book ratio and ROAfrom t-2 to t-1 (Shin, 2013). Regression results are presented in Table 4 and 5.

Additional tests were performed using only the variables related to peer pay (Below Median Pay, Difference Pay and the CDF of the relative pay). Results are reported in Table 6, Appendix A.

3.3 Peer Group Creation

For analyzing the effect of benchmarking on compensation, the way in which peer companies are selected and the characteristics they carry are fundamental. Proxies for these comparison clusters are created in a manner mimicking the actual procedures firms use, as described in compensation committee reports. Usually, the peers are chosen from the same industry and with similar sales revenues. In most of the current literature 4, the companies are separated into industries by their

4

See Bizjak, J., Lemmon, M., & Nguyen, T. (2011); DiPrete, T. A., Eirich, G., (2010) ; Gong, G., Li, L. Y., & Shin, J. Y. (2011) amongst others.

(19)

digit standard industrial classification (SIC) codes. Moreover, Aggarwal and Samwick (1999) and Gibbons and Murphy (1990) argue that the exact length of the SIC code is immaterial, and show that using the two-digit, four-digit or the 49 Fama and French industries produce very similar compensation regression results. Relying on their findings, the firms in this paper are also split according to the two-digit codes, in order to construct peer groups of a representative size. For each year, all firms are ranked in comparison to the others in their industry, according to last year sales. Further, the companies are split into two sales categories: the high group, if their previous year sales were above the median previous year sales for the industry; and the low group, if their sales were below this median. Within each sales category in each industry, the executives are ranked based on last year total remuneration and similarly separated in two compensation categories: high, if executive pay was above the prior year median, and low, if it was below it. This procedure provides the median peer group reward and performance, as well as the CEO’s position relative to the pay median and his firm’s position relative to the performance median, which will be used in the estimations. Last year information is used, as determining groups for compensation setting is typically done far before the current year results are known.

This partitioning is recognizably imperfect, as some firms might rely on other characteristics for constructing their peer group (or not use a peer group at all). There is also the possibility that the executives handpick their comparison universe in order to inflate their own pay. In either case, the ability of the models used in this paper to assess the effect of benchmarking on compensation, would be weakened.

3.4. Data Sources and Summary Statistics

As this study aims at providing empirical evidence on the theoretical relationship between pay and performance benchmarking and executive rewards, data collection is fundamental. Four different databases were used in order to obtain a comprehensive and complete set of the firm, director and board characteristics needed. The sample covers US S&P500 firms over the time period 2008-2014, and the focus is exclusively on executives classified as CEOs. This selection allows for examining the relation in question post-SIC changes in disclosure requirements of 2006. Clearly, this time span includes the potential effects of the global financial crisis on CEO remuneration.

CEO compensation information is obtained from ExecuComp and is matched with director characteristics from the ISS (previously Risk Metrics) Director database. ISS Government is used for the construction of the entrenchment index, which reduces the initial sample (extending until 2015), as

(20)

the source keeps record of the variables only up to 2014. The CEO and board features are further supplemented by firm-data from Compustat.

Additional details on the variables’ definitions and sources can be found in Appendix A, Table 6.

The information sets contained a large amount of gaps in the data, which led to a considerable reduction in the number of observations. In order to obtain a representative sample, the missing values were manually filled in with hand-collected data, where possible, before merging the sets. Further manipulations include removing the executives with less than two years in the current CEO position because they might be receiving only partial pay and cannot be used to estimate the effects on changes in remuneration. The compensation indicators are winsorized at the 1% level to manage any skewness in the data, and the directors with reported total pay of 0 are excluded.

To prevent inconsistencies, the firms with non-consecutive year observations were also eliminated as well as the ones with missing values for any of the explanatory variables. As CEOs are the main focus of the paper, only the directors classified as CEO are kept (and from here onwards referred to also as executives or directors). Following this procedure, the final sample consists of 3,751 CEO-year observations covering 709 firms. Note that in Table 4, as a result of additional variables requirements, the number of observations in the regressions declines further.

Summary statistics of the sample are presented in Table 2. The main compensation measure used is total annual pay, consisting of salary+ annual bonus + other annual payouts + restricted stock grants + other long term incentives. Other reward indicators examined are cash (salary + annual bonus payouts) and stock and option grants (referred to as long-term incentives (LTI) in $, valued by the Black-Scholes model). The mean total pay, mean change in total pay, mean cash and mean stock plus option value are $5.75 million, $0.26 million, $0.984 million and $3.35 million. The average number of peers for each firm is 13 (untabulated), slightly lower than the typical count in the benchmarking literature (usually around 16). The average size of the companies in the sample, measured as sales revenue, is $6.92 million but there is a great variance between the smallest and largest firm. The Market to book ratio also exhibits a large difference between its minimum and maximum values. Looking at the board and CEO features, on average 80% of the directors on a company’s board are outsiders. The mean board size consists of 9 directors and the CEO is also a chairman of the board in slightly over 50% of the cases. The average age of the CEO is 55 years, with 25% being above 60, and only 3% being women.

(21)

Examining the correlations, presented in Tables 8 and 9, most of the independent variables have weak associations between each other. Only covariates with correlations lower than 0.5 are included in the regressions. The only determinants with higher associations are the relative pay ones, as they are different manipulations of the same data, and represent similar ideas. However, as they are not simultaneously included in the models, this does not pose estimation issues.

Table 2 Descriptive Statistics

Summary statistics on compensation, firm, board and CEO characteristics. The total sample consists of 3,751 observations, over the time period 2008-2014. The CEO compensation data is obtained from ExecuComp; Firm Characteristics: from Compustat; Board and CEO Characteristics: from ISS (previously Risk Metrics). Total compensation is the annual dollar value of salary+ bonus + other annual payouts + restricted stock grants + other long term incentives. Cash compensation is the annual salary + annual bonus payouts. The LTI: stock and option awards are the Black-Scholes model values of annual stock and option grants (referred to as long-term incentives (LTI) in $). ∆Total compensation is calculated as total compensation t -total compensation t-1. Similarly defined are ∆LTI and ∆Cash compensation. Sales is the total annual sales

revenue. Leverage is the ratio of total debt value to the market value of assets; ROA is the ratio of EBITDA to total assets;

Market to book is the ratio of market value of assets to book value of assets; Stock return is the firm’s annual stock return. Board size is the number of individuals on the firm’s board of directors; % Outside directors is the fraction of directors

classified as outsiders; Entrenchment index is a 0 to 6 ordinal indicator of shareholder rights (0= high shareholder rights, 6=low) and combines six different indicators of director entrenchment, concerning delay, voting and other provisions, and attaches a weight of one to each. CEO is chairman is a binary variable equal to one if the CEO is also a chairman of the board of directors. CEO age is the age of the executive; CEO above 60 is a binary variable equal to 1 if the executive is older than 60 years; CEO tenure is the number of years the executive has served on the board until the year in question; Gender is a binary variable equal to 1 if the CEO is a female.

Panel A: CEO Compensation

Variables Mean Median Minimum Maximum Number of observations

Total compensation (in thousands $) 5,750 4,103 0.001 156,078 3,751

LTI: stock and option awards (in thousands $)

3,349 2,097 0 145,060 3,751

Cash compensation (in thousands $) 984 825 0.001 71,634 3,751 ∆Total compensation (in thousands $) 259 96.6 -106,145 148,208 3,128 ∆LTI: stock and option awards (in

thousands $)

234 104 -38,054 122,521 2,972 ∆Cash compensation (in thousands $) 23.8 0 -50,601 19,884 3,128

Panel B: Firm Characteristics

Sales (in thousands $) 6,923 1,606 3.18 433,526 3,751 Leverage (scaled by assets) 0.208 0.192 0 2.93 3,751

ROA 0.102 0.0951 -1.81 1.18 3,751

Market to book 3.43 2.26 -164 1,374 3,751 Stock return 0.149 0.101 -0.941 26.9 3,506

(22)

Table 2 (Continued)

Panel C: Board Characteristics

Board size 8.94 9 3 20 3,751

% Outside directors 0.793 0.818 0 0.947 3,751 Entrenchment index 3.5 4 0 6 3,751

CEO is chairman 0.559 1 0 1 3,751

Panel D: CEO Characteristics

CEO age (in years) 55.5 55 29 88 3,751

CEO is above 60 0.251 0 0 1 3751

CEO tenure (in years) 9.46 7.5 0.08 52.6 3,751

(23)

CHAPTER 4 Results and Discussion

4.1 Analysis of the Effect of Peer Group Pay and Performance Medians on CEO Compensation

Table 3 presents the outcomes of the estimations of the compensation models regarding the level of CEO remuneration. In columns (1) and (2) the dependent variable is the main measure of executive pay, the log of total annual compensation, in columns (3) and (4) it is the log of the dollar value of stock and option grants and in (5) and (6) it is the log of cash payouts (salary plus bonus). Columns (1), (3) and (5) show the results for the base specification, not including any peer group characteristics; in (2), (4) and (6) the log of the prior year median pay and the median performance of the comparison firms are added to the regressions.

Understanding the determinants of executive compensation solely in terms of the company’s internal characteristics will be beneficial for interpreting the main results of the study further on. Focusing first on the base model, a number of the predictors of total awards are evidently significant for the CEO pay and the direction of their effect is consistent with the expectations. From the firm features, only the lagged value of the accounting performance (ROA t-1) is negatively related to the overall

compensation and it is also the only one not statistically significant. However, with respect to the LTI, it does account for sizeable increases of 44%, which is consistent with the fact that stock and option grants are traditionally related to long term projects. The log of sales, the past stock returns, contemporaneous ROA and lagged Market to book are all material for the total executive remuneration at the 1% level, with a 1% higher sales revenue leading to approximately 0.35% raise in pay. Following conventional logic, larger companies and the ones with numerous investment opportunities usually have the resources to provide more substantial remuneration packages. An increase in the Market to book ratio has no consequences for the cash compensation but corresponds to a 0.08% and 0.1% raise in total rewards and stock and option grants respectively, which is consistent with the practice of firms with more investment opportunities trying to stimulate a long run focus.

The modest economic effect could be attributed to the fact that some of the companies with significant growth potential are actually small young firms, which do not have enough capital for extremely generous awards. The firm leverage and current stock performance also have a boosting effect on the main compensation measure, albeit only at the 10% level. The stock returns are not statistically significant for the cash payments, which is unexpected, as the bonus part is often tied to changes in performance. The firm size Ln(Sales t-1), accounting returns (ROA) and leverage are the only firm

(24)

All of the board characteristics are positively related to the total level of executive compensation and, besides the CEO is chairman, are statistically meaningful. The signs on the coefficients of the board size, entrenchment index and CEO duality, which measure the strength of corporate governance, are compatible with the notion that weak governance allows the executive to secure additional rewards. The fact that a higher percentage of outside directors leads to increases in remuneration adds to the mixed evidence on the value of outsiders as monitors. Several studies are re-evaluating the extent to which the traditional distinction between outside and inside managers represents board independence in practice (Johnson et al., 1996). It would be beneficial to explore the connection more thoroughly by separating the variable in number of outside members appointed after (or by) the CEO and the ones already in place when the CEO took office.

Weaker are the results related to the CEO characteristics, where none of the coefficients is statistically significant for compensation. The findings for women executives are ambiguous but females appear to receive 0.5% more than males in overall rewards, which is in contrast with the on-going debate on the gender pay gap. Not surprisingly, the director’s tenure effect is also uncertain, as it could represent unwillingness of the firm to invest in human capital as the time progresses but simultaneously proxy for the CEO’s skill and effort or political and social influences (Shin, 2013).

The covariates explain between 44-51% of the variation in stock and option awards and total CEO pay but only 14% when it comes to the salary plus bonus. This suggests that the determination of the short term component of pay might be governed by a different set of characteristics. The lower number of statistically significant results for the cash payouts could also be due to the fact that the values of the fixed (salary) and the variable (bonus) part might be affected in opposite directions by the company’s indicators. In this case, the total overall impact would be ambiguous and would depend on which effect is stronger.

The inclusion of the prior year values of the log of median pay and of the median performance of the peer group (columns (2), (4) and (6)), does not lead to radical changes in either the coefficients of the control variables or their significance, compared to the base specification results. The R2 for all three dependent variables improves only slightly by 0.2 to 0.4%. Nevertheless, an increase in the median level of pay of the comparable companies leads to a raise in all of the components of CEO compensation, which confirms that benchmarking is material for reward determination. A 1% higher median in the previous year translates into 0.003% additional total earnings in the current year. For the cash payments the economic effect is greater, amounting to 0.2%. The other variable of interest, the median stock returns, exhibits a positive (negative) link to the total pay and stock plus option grants (cash) but is not significant in any of the specifications. This suggests a limited to non-existing role of peer group performance for director compensation determination. However, no conclusion can be

(25)

drawn yet, because using the absolute measures of the benchmarking variables, without relating them to the specific firm, is quite a simplified approach. Still, the preliminary findings indicate that peer groups are non-trivial for the setting up a CEO’s remuneration package.

The following model augmentations, directly relating the pay and performance indicators of the company to the ones of its comparable universe, should provide more refined results, to be used for evaluating the hypotheses.

Table 3 OLS Regressions with the Peer Group Pay and Performance Medians

OLS regressions of the effect of firm, board and CEO characteristics and peer group median pay and performance on executive compensation. The dependent variable in columns (1) and (2) is the log of total annual compensation, in columns (3) and (4) it is the log of the Black-Scholes dollar-value of stock and option grants and in (5) and (6) it is the log of cash payouts (salary plus bonus). Columns (1), (3) and (5) show the results for the base specification, not including any peer group characteristics; in (2), (4) and (6) the log of the prior year median pay and the median performance of the comparison firms are added to the regressions. Independent variables: Ln(Sales t-1) is the log of the firm’s last year’s total annual sales revenue; Leverage t-1is the lag of the ratio of total debt value to the book value of assets; ROA is the ratio of EBITDA to total assets;

ROAt-1 is the prior year’s firm ROA; Market to book t-1 is the lagged value of the ratio of market value of assets to book

value of assets; Stock return is the firm’s annual stock return. Stock returnt-1 is the lag of the stock return.Board size is the

number of individuals on the firm’s board of directors; % Outside directors is the fraction of directors classified as outsiders;

Entrenchment index is a 0 to 6 ordinal indicator of shareholder rights (0= high shareholder rights, 6=low) and combines six

different indicators of director entrenchment, concerning delay, voting and other provisions, and attaches a weight of one to each. CEO is chairman is a binary variable equal to one if the CEO is also a chairman of the board of directors. CEO age is the age of the executive; CEO above 60 is a binary variable equal to 1 if the executive is older than 60 years; CEO tenure is the number of years the executive has served on the board until the year in question; Gender is a binary variable equal to 1 if the CEO is a female. Median performance is the median stock return of the peer group in the prior year; Ln(Median pay) is the dollar-value of the median compensation of the peer group in the prior year. The total sample period is 2008-2014. Significance levels are standardly denoted as * p < 0.1, ** p < 0.05, *** p < 0.01. Heteroskedasticity robust t statistics with firm-clustered standard errors are in parentheses. Time and industry fixed effects included.

Independent variables

Dependent variable: CEO compensation

(1) Ln(total compensation) (2) Ln(total compensation) (3) Ln(LTI) (4) Ln(LTI) (5) Ln(cash) (6) Ln(cash) Ln(Sales t-1) 0.348*** 0.310*** 0.421*** 0.382*** 0.153*** 0.125*** (15.02) (10.01) (16.18) (13.56) (7.75) (4.81) Leverage t-1 0.000* 0.000** 0.000** 0.000** 0.000* 0.000** (1.86) (2.34) (2.01) (2.31) (1.87) (2.06) ROA 0.962*** 0.954*** 0.674* 0.649* 0.286* 0.291* (2.61) (2.60) (1.83) (1.80) (1.62) (1.68) ROA t-1 -0.199 -0.174 0.439* 0.489** -0.199 -0.195 (-0.66) (-0.59) (1.74) (2.11) (-1.09) (-1.06) Stock return 0.0517* 0.051* 0.036 0.039 0.0164 0.013 (1.71) (1.64) (0.80) (0.88) (1.45) (1.02) Stock return t-1 0.078*** 0.076*** 0.133** 0.138*** 0.005 0.007 (2.80) (2.79) (3.00) (3.09) (0.67) (0.88) Market to book t-1 0.001*** 0.001*** 0.001*** 0.001*** 0.000 0.000 (3.85) (3.60) (3.20) (3.23) (0.43) (0.51) Board size 0.021* 0.019 -0.008 -0.0092 0.010 0.008 (1.70) (1.58) (-0.46) (-0.56) (0.69) (0.60)

(26)

Table 3 (Continued) % Outsiders 0.863*** 0.856*** 1.458*** 1.461*** 0.494* 0.497* (3.36) (3.37) (3.94) (3.97) (1.78) (1.81) Entrenchment index 0.086*** (4.15) 0.084*** (4.11) 0.124*** (3.99) 0.122*** (3.98) 0.063** (2.50) 0.061** (2.47) CEO is chairman 0.056 0.051 0.032 0.0253 0.031 0.031 (1.15) (1.08) (0.49) (0.39) (0.68) (0.68) CEO age -0.001 -0.001 -0.000 -0.000 0.001 0.000 (-0.63) (-0.23) (-0.08) (-0.12) (0.27) (0.31) CEO is above 60 0.039 0.037 -0.118* -0.116* 0.128*** 0.126*** (0.01) (0.78) (-0.29) (-1.79) (2.62) (2.59) CEO gender 0.005 0.000 -0.098 -0.101 0.111*** _0.108** (0.04) (0.00) (-0.53) (-0.56) (2.61) (2.51) CEO tenure -0.004 -0.004 -0.001 -0.000 0.000 0.001 (-1.04) (-0.96) (-0.15) (-0.14) (0.15) (0.19) Median performance 0.115 (1.25) 0.303 (2.34) -0.010 (-1.17) Ln(Median pay) 0.003** 0.005*** 0.028* (2.39) (2.63) (1.56) _cons 4.555*** 4.753*** 3.259*** 3.543*** 4.797*** 4.801*** (15.87) (15.59) (9.39) (9.99) (21.45) (20.99) Industry fixed effects

Yes Yes Yes Yes Yes Yes

Year fixed effects Yes Yes Yes Yes Yes Yes

N 3061 3061 2813 2813 3061 3061

adj. R2 _0.510 _0.512 _0.440 _0.442 _0.144 _0.148

4.2 Analysis of the Effect of a Firm’s Relative Pay and Performance Positions on CEO Compensation

In the following set of regressions the effect of benchmarking is examined in further detail by focusing on relative, rather than absolute measures of pay and performance. These are the aforementioned binary Below Median Pay and Below Median Performance, indicating whether the CEO was “underpaid” last year and whether the firm underperformed, with respect to its peers; and the continuous Difference Pay and Difference Performance, equal to the prior year CEO’s dollar-value compensation and firm’s stock return, net of the respective peer group medians. The dependent variable in regressions (1) and (2) is the log of the change in total annual compensation from year t-1 to t; in (3) and (4) it is the change in the dollar-value of stock and option grants; and in (5) and (6) it is the change in cash payouts (salary plus bonus). Columns (1), (3) and (5) present the results of the estimations including the binary variables, while in (2), (4) and (6) the focus is on the continuous differences.

The findings for Below Median Pay indicate that, after controlling for company’s size and absolute performance, an executive paid less than the median at t-1 is granted a compensation increase 79% higher than the compensation increase of a counterparty, paid at or above the median in the previous

(27)

year. The effect is statistically significant and consistent across the various remuneration measures, and is in line with Hypothesis 2. From the separate components, the biggest impact is on LTI, about twice the size for the salary and bonus (52% - 21%). These larger stock and option awards to low paid executives, resulting from the benchmark process, could be seen as part of the explanation for the dramatic increase in the use of LTI documented by Hall and Liebman (1998). As expected, a company’s stock underperforming the comparable companies is negatively related to the executive wage adjustments. The effect, however, is economically small – the CEO of a firm with below median performance, faces a decrease in total rewards of approximately 4.9% more than the CEO of similar entities. Moreover, this link is not significant at any conventional level.

With respect to the relative measures in (2), (4) and (6), the coefficients exhibit the same trends, both in direction and significance. The further a CEO pay is below the peers’ mid-value in the previous year, the higher the current increase in rewards. A 1% lower CEO reward last year, leads to 0.009% increase in the contemporaneous remuneration package. Translated into dollar-value terms, this is a change of $22,621.5 On the contrary, the compensation would decrease between the two years by 72% for an executive in a firm with a stock return lower by 1 than the median of the peers. In line with executive compensation theory, the effect is greatest for the change in cash payments (84%) and smallest for the change in stock and option grants (8,7%). The salary and bonus, as short-term components, are more dependent on the absolute annual stock returns than LTI, as the latter is traditionally determined based on several consecutive years of performance. The inclusion of the continuous measures for the firm’s net pay and performance drives out most of the explanatory power of the current and past stock returns. Nevertheless, the effect of the Difference Performance variable is also insignificant in all specifications. This is potentially due to the correlations of the lagged performance covariate and the measures of the returns’ distance, which are relatively high compared to the others.

To examine this possibility in more detail, the regressions from part 4.2 were replicated only using the three different relative pay measures. Results are presented in Appendix A Table 6. Almost all coefficients on past and contemporaneous returns are positive and have a highly significant relation to CEO rewards in both economic and statistical terms. This is in contrast with the mixed and weaker findings of the models including the relative performance indicators, and in line with the idea of correlations driving the results. The other control variables do not exhibit large changes in their impact on compensation .There is also no considerable difference in the estimated effects of the relative pay measures and they still pay an important role for the increase of the executive

5

The dollar-value $22,621 is calculated following Albuquerque et al. (2013); [exp(7.45+0.00895*1.46)- exp(7.45)], where 7.45 is the mean of the log of the difference in pay, 1.46 is the standard deviation of the difference in pay (untabulated) and 0.00895 is the coefficient on the variable (Table 4, regression (2)). This means CEO compensation increases from year t-1 to t by $22,621 for each 1% by which the executive was underpaid last year.

(28)

compensation. Omitting the net performance measures leads to a decrease of the adj.R2 for all specifications. The findings are consistent with research on the peer pay effect on compensation for the pre-2006 period.

Focusing on the regressions in Table 4, the change in sales, the absolute measures of the firm’s stock returns and the increase in growth opportunities are material for a bigger raise in executive compensation, while the change in leverage has an ambiguous effect. From the absolute performance indicators, the return on assets has higher economic implications for executive rewards, however the coefficients are insignificant at any conventional level and the direction of the effect is undetermined. As previously, an increase in the prior year absolute stock returns leads to positive changes in this year’s compensation. These trends are consistent across the various estimation specifications. The adj. R2 improves slightly when using the continuous pay difference measure because of its higher information content, compared to the dichotomous Below Median Pay. It should be noted that due to the change in some of the independent variables in Table 4, the number of observations in the regressions falls by approximately one half as compared to Table 3.

The estimations yield consistent results, regardless of the binary or continuous variables being used. In either case, being a director with pay below the median leads to a higher compensation increase in the following year, but a firm’s relative underperformance (lower stock returns net of the peer median) would be detrimental for the pay. For the economic impact of the peer group performance and pay to be compared, however, the CDFs of the relative measures are included in the last set of regressions, presented in Table 5. The dependent variables in columns (1), (2) and (3) are respectively the log of the dollar-value change in total annual compensation from year t-1 to t; the change in stock and option grants; and change in cash payouts (salary plus bonus).

The findings are consistent with Hypothesis 1.The impact of the relative rewards is about 3 times higher than the one of performance – moving to the top of the pay distribution is associated with an increase of rewards more than 100% higher than the increase following a movement to the top of the performance distribution. Regarding the relative compensation measure, getting from the 1st to the 100th percentile results in an average raise of overall awards by 170%. The difference in the effects of CDF pay and performance is ostensibly smaller for the change in LTI and cash: for the salary and bonus it is approximately 39%. While the CDF pay is statistically significant at the 1% level in all three specifications, the performance is once again not material for the changes of pay.

The fact that performance relative to the comparable companies appears to have no role for the change of compensation has two possible implications. First, it could mean that typical performance-based indicators, such as bonuses and LTIs, are actually solely determined in relation to the peer pay instead