The relationship between CEO compensation and firms
performance, with a distinction between million-dollar firms
and non-million-dollar firms.
analysis of American companies.
Klaassen, Anne Puck 10252630
February 2, 2016
Supervisor: J. Ligterink
2
Hierbij verklaar ik, Anne Puck Klaassen dat ik deze scriptie zelf geschreven heb en dat ik de volledige verantwoordelijkheid op me neem voor de inhoud ervan.
Ik bevestig dat de tekst en het werk dat in deze scriptie gepresenteerd wordt origineel is en dat ik geen gebruik heb gemaakt van andere bronnen dan die welke in de tekst en in de referenties worden genoemd.
De Faculteit Economie en Bedrijfskunde is alleen verantwoordelijk voor de begeleiding tot het inleveren van de scriptie, niet voor de inhoud.
3
Abstract
The purpose of the International Revenue Code Section 162(m) mandated by the SEC was to increase pay for performance in million-dollar companies. This study first examines the relation between firm performance and executive compensation. Second, this study aims to provide further evidence to earlier articles by observing the recent impact of the section 162 (m) regulation on the relationship between executive compensation and firm performance in million-dollar firms. In contrast to previous work, this study also examines the relationship between executive compensation and firm performance non-million-dollar firms. This study performed a multiple regression analysis on the total yearly compensation of a CEO, the Return On Assets of the related firm, and the million-dollar firm variable. The sample consists of 8262 observations from 2372 different North American companies in the period 2007-2013. The results show statistical evidence to state that the vast majority of North American companies do have a positive relationship between CEO pay and firm performance, that million-dollar firms have a positive relationship between CEO pay and firm performance, and that non-million-dollar firms have a negative relationship between executive compensation and firm performance.
4 Table of contents 1. Introduction ... 5 2. Literature ... 7 2.1 Agency theory ... 7
2.2 Previous literature on pay for performance ... 9
2.2.1 Results from previous literature on pay with performance ... 9
2.2.2 The contribution of inefficient contracts on pay without performance. ... 10
2.3 Million-dollar firms ... 12
3. Research design ... 14
3.1 The hypotheses... 14
3.2 Research Method ... 15
3.3 Data ... 18
3.3.1 Data collection and adjustments for the dependent variable ... 18
3.3.2 Data collection and adjustment for the independent variables ... 19
3.3.3 Merging the two datasets ... 20
4. Results and analysis ... 21
4.1 descriptive statistics ... 21
4.2 Correlation matrix and Multicollinearity ... 23
4.3 Results ... 25
4.4 Discussion ... 27
5. Conclusion ... 30
Appendix ... 31
5
1. Introduction
There has been a stream of negative publicity about excessively high and irrational salaries drawn by corporate executives. The debate about the relation between executive pay and firm performance has been an ongoing discussion. The reason for this public perception is that a large part of CEO compensation is not based on increasing firm value for their shareholders, however, these same shareholders are charged for the associated costs coming from the actions taken by their CEOs. This more than performance-related compensation is in part illustrated by the fact that companies in competitive markets want to pay more in order to retain their CEOS. For example, in 1998 Michael Eisner, chief executive at that time of The Walt Disney Co. and one of the US’s highest compensated CEOs, made almost $575 million. On top of his $750.000 salary, Eisner claimed a $ 9.9 million bonus and cashed in on $565 million in stock options that year. Although Disney had a downturn from a 28% drop in profits in 2000, Eisner still managed to bag a $11.5 million bonus. The question at that time was why should executives like Eisner earn so much money? In part, this excessive increase in compensation was due to a large rise in the stock market and not because of the improved performance of Disney. Bertrand and Mullainathan (2001, p. 901) stated that CEO pay should be fair and should not be tied to observable shocks to performance beyond the CEO’s control.
The persistent criticism has led to a stream of implemented national regulations. For example, on august 5 of last year, The SEC adopted a final rule to the Dodd-Frank Executive Compensation reform Act, the Pay Ratio Disclosure Requirement. This requires a public company to disclose the ratio of the compensation of its chief executive officer to the median compensation of its employees (SEC, 2015).
This ongoing discussion and constantly new implemented regulations are two of the reasons why there have been numerous studies about fair executive compensation. Most studies focus on this concern to see whether there exists a pay for performance relationship in publicly traded companies. Studies have examined this subject by testing different ways of compensation against various variables representing firm performance. Bebchuk and Fried (2004) are one of the better-known studies which represent a pay without performance resulting from a lack of governance structure, a prevalent managerial power and a significant influence of firm size. In contrast to this view, the majority of studies find a positive relation between firm performance and executive compensation due to the ongoing criticism, independent board members, and new remuneration structures (Mehran, 1995, and Hall and Liebman, 1998, among others).
Another part of prior research focuses on the impact of implemented regulations on the pay for performance relation. For example, the study of Perry and Zenner (2001) focused on the International Revenue Code Section 162(m) mandated by the U.S. Securities and Exchange Commission (SEC), which entered in 1992-1993 and required the limitation of tax deductibility in firms that pay non-performance-based CEO compensation above $1 million, i.e., million-dollar firms. These rules had led to a decrease in compensation, a decline in salary growth rates, and an increase in
6
pay-performance sensitivity in million-dollar firms with CEOs near or above a million dollar compensation level (Perry and Zenner, 2001). However, their study only focused on the impact of the regulation on the pay-performance relation in million-dollar firms and did not show the effects on the pay-performance relation in non-million-dollar companies. The same applies to the study of Rose and Wolfram (2002), who only found evidence on the decrease in executive compensation in million-dollar firms.
There have not been many studies examining the impact of the section 162 (m) regulation and, in particular, the impact of this regulation on the pay for performance relationship in non-million-dollar companies. Despite the fact that section 162 (m) is implemented with the purpose to reduce the non-performance-based compensation in million-dollar firms, this regulation might also affect the pay for performance in non-million dollar companies. That is why this study aims to provide an addition to the previously described articles by observing the recent impact of the section 162 (m) regulation on the relationship between executive compensation and firm performance in million-dollar firms and non-million-dollar firms.
The research question for this paper can be formulated as follows: Do total
compensation of CEOs in North American companies have a positive relation with firm performance? Will there be a distinction in this relation between million-dollar firms with CEO compensations above a million dollars and non-million-dollar firms with CEO compensation less than a million dollars?
This study performed a regression analysis on the total yearly compensation of a CEO, the Return On Assets of the related firm, and the million-dollar firm variable, to investigate whether there exist a relationship between firm performance and CEO compensation and if this relationship contains a difference between million-dollar firms and non-million-dollar firms. The dependent variable represents the total compensation of the executive of the given company, which consists of the yearly bonus, salary and stock and options awarded per CEO year. The variable Return On Assets is used to represent the independent variable, firm performance. The financial performance measures: firm size, sector, and net income, are used as control variables, to avoid potential omitted variable bias. The dummy variable million-dollar firm is realized to distinguish firms where executives receive compensation above a million dollars and vice versa. Subsequently the interaction term “ROAMillion-dollar firms” is introduced in order to make a distinction in the pay-performance relationship between million- and non-million-dollar firms. Finally, year-dummies are added. The choice of variables is drawn from previous literature, which is detailed in section two and three. The sample consists of 8262 observations from 2372 different North American companies in the period 2007-2013.
The structure of this paper is outlined as follows. Section 2 presents a discussion of the relevant previous literature, the underlying agency theory, and the million-dollar concept. Section 3 describes the hypothesis, the research model and method used for the empirical analysis, and the data set. In section 4, the results are discussed and presented. In the 6th section, the conclusion will be drawn.
7
2. Literature
This chapter first discusses the underlying agency theory, which is applied to the shareholder-manager relation and leads to executive compensations as an answer to the agency problem. Afterward, the previous literature on pay for performance is discussed. At last, the relevance of the million-dollar firm is briefly explained.
2.1 Agency theory
The literature on the pay for performance studies goes back to a fundamental agency problem. The primary predecessors of this theory were Berle and Means (1932) enhanced by Michael C. Jensen and William H. Meckling (1976) with their theory of the ownership structure of the firm (containing both debt and outside equity claims). The agency model is defined by Jensen and Meckling (1976).as a contractual relationship under which the principal engages the agent to perform some work on their behalf, which implicates delegating some rights, responsibilities and decision-making authority to the agent. In this paper, we describe the shareholder-manager relationship. In this relation, there exist a separation between owners of largely publicly traded companies, and the control of their hired executives. It should be no surprise to discover that this severance can cause some issues, referred to as the “agency problem”.
Murphy (2012) identified, at least, three versions of this problem: the Agency Cost of Debt reflecting a potential conflict of interest that exist between a company’s shareholders and its debtholders; the Agency Costs of Free Cash Flow; and The Agency Cost of Equity. The latter one describes the problem in this paper. As noted by Murphy (2012): “ Executives who own less than 100% of the shares of an
all-equity firm will not make the same decisions they would if they owned 100% of the share. Executives want to be paid more and to take actions that increase their own utility, while shareholders are primarily concerned with providing executives with incentives to take actions that increase value of the shares.”
The agency theory knows different conflicts. First of all, principal and the agent have inconsistent interests. For example, on the one hand, we have the executive who is eager to receive maximum compensations out of short-term delivered performances. He may pursue his self-interest and, therefore, choose to postpone short-term costs, which could have led to future profitable investments. In contrast to the managers view, we have on the other hand the shareholder who is interested in increased firm value. He is concerned with the long-term view of the company and hence would have implemented such future profitable investments (Tosi et al., 2000). The second problem is the fact that the agent can make decisions which are too risk-averse causing the principal to lose opportunities which he would have wanted to achieve (Eisenhardt, 1989, p. 58). Finally, the agent-principal relationship possesses an asymmetry of information, the agents has often more and better information than the principal (Tosi et al., 2000, p. 304).
8
All the described issues cause the principal to suffer from some losses. These losses are called the agency costs and arise by letting the agent do the job instead of carrying it out himself. These agency costs contain three different type of costs, namely: supervision costs it is difficult and costly for the principal to monitor the agent; binding costs caused by the agent to show his principal that he has the best interest in the company; and residual loss arising by engaging the relationship between the agent and the principal (Jensen and Meckling, 1976, p. 308).
Jensen (1993) emphasized four forces that could lighten agency conflicts between managers and shareholders: The concept is the capital markets. Shareholder activists and large institutional shareholders mitigate agency problems by pushing corporations to increase the relation between executive wealth and increased firm value (Murphy, 2012, pp.132-133). The second force devised by Murphy (2012), which has dissolving influences on the agency problems, is the political, legal, and regulatory system. Over the past years, several disclosure standards have been designed to inviolable shareholders. For example, the Securities Act of 1934 protected shareholders by requiring full disclosure. All information that was necessary to make a good investment choice should be made public. In addition, several other regulations were imposed to protect shareholders and other financial claimants, which consisted of business larceny, tax policies, accounting rules, disclosure requirements, security rules, and embezzlement. The third force, product markets, had a lightened effect on agency problems by providing discipline to value-destroying managers. In other words, companies that cannot compete in the product market cannot survive (2012, pp. 134-137). The last concept is the board of directors, which is elected by the shareholders. In the beginning, boards were stocked with former executives and CEOs of the company itself, however, after the shareholder movement in the 1980a (Murphy, 2012, p.69) a strong desire for external board members came into existence (Horstmeyer, 2011). The board ensures executives are hired and monitored and determines their salary.
However, the most direct approach used in prior literature to solve the agency problem is to pay for performance by setting up an incentive contract in order to induce the agent to act on the behalf of the principal. The construction of an executive contract is frequently based on the efficient contracting hypothesis. Structuring the CEO compensation so that it provides the appropriate incentives. The incentive contract binds compensation to shareholders value creation (performance-based contract), to which they respond in predictable ways, and at the same time pays enough so that the executives will take the job. Rewards are regularly distributed in the form of equity-based compensation such as stock options or restricted stock (Murphy, 2012, pp. 132-133).
Concluding, CEO pay solves the agency problem by rewarding executives with the right incentives. However, section 2.2 also described studies who aim to show that these same contracts do not always serve their purpose and are sometimes a problem itself.
9
2.2 Previous literature on pay for performance
The contracting view of executive compensation explains that compensation is used by investors to solve the agency problem. Over the years, several studies have been performed to explain the excessive increase in CEO pay and to determine a link between rewards received by firm’s directors and the results of a corporation. Former literature and studies have shown different results regarding this relationship.
In general, there are two distinct views on the executive compensation literature. The majority of studies such as Perry and Zenner (2001), and Hall and Liebman (1998) among others, support the effective contracting view and find positive relations between firm performance and executive compensation. Contrary, some studies describe that complication in these contracts could generate pay for luck instead of pay for performance (Betrand and Mullinathan, 2001). Bebchuk and Fried (2004), and Jensen and Murphy (1990b), conclude that the compensation of top executives is independent of performance.
Section 2.2.1 describes the literature based on pay for performance. Afterward, in section 2.2.2 different concepts on pay without performance are discussed following by results on pay without performance from previous literature.
2.2.1 Results from previous literature on pay with performance
That before the 80s, several studies did not find a significant relation in pay-performance was not an exceptional situation. Murphy(2012) stated that the executive compensation practices provided few incentives to create long-run value and performance –based terminations were almost non-existent. After these years, the growing pressure for the link between firm performance and manager pay increased dramatically. Murphy (2012) described the years 1983-1992 as he Emerging Market for Corporate Control. This had led to the adding of section 280(g) to the tax code as part of the Deficit Reduction Act of 1984, which was meant to reduce excessive CEO payments. The new rules and the enormous growth in stock markets have led to The Stock Options explosion of 1992-2001 provided a tremendous pressure for equity-based pay. This occurrence increased the proportion of stock options in executive compensation schemes. As the years followed, different regulations were designed that would ensure a more positive outcome in executive contracts to the pay-performance relation.
As described in section 2.1, there were different forces that contributed to the pay with performance view: the ongoing criticism, the pressure from shareholder activists in capital markets, the regulatory systems, the board of directors, but most of all the incentive contracts. They have generally ensured that despite the increase in the level of CEO compensation the sensitivity for pay-performance (i.e., equity-based pay) has risen dramatically (Murphy, 2012).
The study of Mehran (1995) suggests that when executives receive an incentive to increase shareholders value, the form rather than the level of compensation is what motivates managers to increase firm performance. In this study, a compensation dataset of 153 randomly selected manufacturing firms during the
10
period 1979 through 1980 is collected. The hypotheses are tested cross-sectional using ordinary least squares analysis (OLS). Their study presents empirical evidence which shows that firm performance is positively related to the percentage of their compensation that is equity-based.
The study of Clementi and Cooley (2009) showed a significantly strong relationship between the executive payments and the created shareholder value. This study focused on the managers who had 1% or less in the company shares. This way they could eliminate executives with much managerial power from their research. Namely, as will be discussed in the following section, powerful CEOs (with a significant amount of company shares and working through influencing captive board members) can affect the payment structures and compensation levels and subsequently causes for contracts to be inefficient (Murphy, 2012, p. 128). Furthermore, results show that with the used compensation variable (change in the level of remuneration) a conclusion can be drawn with respect to a positive increasing pay-performance relation.
Brian J. Hall and Jeffrey B. Liebman (1998) examined the link between firm performance and CEO compensation. They used a fifteen-year panel data set of CEOs in the largest, publicly traded U.S. companies. They found an existing strong relationship between firm performance and CEO compensation, an increase in CEO pay, and a massive increase in sensitivity for pay-performance. Their conclusion was in favor of the effective contracting view, which stated that this correlation was almost nearly fully attributable to the value of managers rewarded stock and stock options. By concluding this, they subsequently do not claim for all CEO contracts to be efficient, in some contracts there are certainly flaws presented.
In general, these studies described a significant link: executive pay is increasingly dependent on firm performance. However, as explained before, some studies describe the problems arising from the incentive contracts instead of solving problems, which might lead to pay without performance. This concept will be further explained in the following section.
2.2.2 The contribution of inefficient contracts on pay without performance. As described in section 2.1, the efficient contracting hypothesis assumes that pay is used to alleviate agency problems (Bertrand and Mullainathan, 2001, p. 901). The structure of the CEO payments also reflects a competitive equilibrium in the market for managerial talent. The contract provides incentives to managers to implement actions that maximize shareholder wealth (murphy, 2012, p. 128). However, as stated by Bebchuk and Fried (2003) these executive compensation arrangements are partly a product of this same agency problem. The study of Betrand and Mullainathan (2010) shows that contracts which seem efficient are sometimes inefficient and cause companies to perform even worse. Jensen And Murphy (1990) support this statement: in remuneration structures in which the incentive compensation was expected the play a significant role the pay-performance relation was found to be almost insignificant. The study of Core, Holthausen, and Larcker (1999) is in line with this view. Their
11
results show that companies with poorer performing governance structure have greater agency problems, which results in higher levels of CEO pay and worse firm performance.
As discussed in section 2.1, the board of directors and the legal/political/regulatory system, played a mitigating role in reducing the agency problems, however, their control mechanism also entailed new conflicts by negatively affecting incentive contracts, as follows.
Various scientists had thought that the introductions of the outside board of directors would lead to a decline in executive pay and that they would give better incentives to CEOs to increase firm value, however, a study by Fernandes. Et al. (2012) showed that an increase in the number of independent board members is accompanied by an increase in executive compensation. Murphy (2012) follows these lines by denominating the independent board as imperfect agents in determining CEO compensation. For example, the elected representatives tend to pay too much compensation to average performing executives. This is also a proof of a theorem accepted by many researchers which stated that directors assign compensations with the money of shareholders and not their own. Similarly, as documented by Baker. et al. (1988), the members of the board receive no benefits from punishing underperforming CEOs, instead, they personally bear a part of the additional cost and are therefore reluctant in dispensing these punishments. The final argument for a director as an imperfect agent is the power of an executive which could affect the selection of the board. The social connection between the two plays an important role in enabling the CEOs to determine the level and structure of their own compensation (Fracassi and Tate, 2012).
These problems between the shareholders and its board members are laid down in the “managerial power hypothesis” (Murphy, 2012). The study of Bebchuk and Fried (2004) found evidence on the relationship between managerial power and managerial pay. They stated that structural weaknesses in corporate governance enabled managers to influence their own pay. The managerial power and influence, have shaped the failed incentive contracts on executive compensation in publicly traded U.S. companies. Altogether, managerial power creates incentive schemes, which leads to managers who take opposite actions from creating long-term firm value, i.e. pay without performance (Bebchuk and Fried, 2004, p. 648).
As discussed in section 2.1, the legal, political, and regulatory systems have led to an increase in pay-performance sensitivity, however, the regulations were generally inefficient and have in some cases even led to a deterioration of the agency problems (Murphy, 2012). A part of this inefficiency is based on the believe that corporate governance regulations have a “One-size-fits-all” nature though not every company is the same and some regulations have different (sometimes opposing) effects. For example, the prohibition of CEOs to take part in their own compensation committees can be expressed as positively in many cases, but in the case of family businesses and CEO company founders, it will harm shareholders. There are even results that show a substantially lower compensation of CEOs who are part of their compensation committees than other CEOs who are not (Bizjak and Anderson, 2003).
12
In addition, regulations focus often on small issues within the remuneration structures leaving enough scope for circumventing these rules. Finally, the corporate decrees are primarily written by politicians who rarely have experiences in the field of pay-performance and owning stock, and also do not see creating firm value as their major objective (Murphy, 2012, pp. 134-137). Jensen and Murphy (1990) share this opinion by mentioning the decline in the pay-performance ratio due to the political forces who introduced restrictions that led to an opposite effect: the reduction of pay-performance sensitivity.
In addition to the board of directors, managerial power, and regulatory systems, there is another driving factor that might affect incentive contracts in an inefficient way and subsequently causes for compensation without performance namely, firms size. The determination of the CEO pay is largely driven by increases in firm size. Tosi, et al. (2000) show that 40% of the variance of executive pay is motivated by company size. This is among other things caused by the increased managerial power in larger firms (Bebchuk and Fried, 2004). However, Gabaix and Landier (2008) stated that the relationship between firm size and CEO compensation was based on increased complexity in managing large companies. What the demand for better qualified Chief Executive Officers should receive is associate with higher compensation levels. This significant relation between firm size and executive compensation gave managers incentives to pursue expansion of their firm (Murphy, 2012, p.62). However, one should keep in mind managers can follow these incentives even when they are decreasing firm value (Bebchuk and Fried, 2004, p.670). Jensen and Murphy (1990b) even found that in large firms CEOs usually have less stock and less compensation-based incentives than CEOs in smaller firms. According to the results of a paper of Schaefer (1998), managers’ pay-performance sensitivity is also decreasing with firm size. Where pay-performance sensitivity is defined by Jensen and Murphy (1990b) as the change in executive wealth due to the one dollar change in shareholder wealth. When firm size grows the variance of shareholder wealth increases, the risk associated with pay-performance sensitivity increases, the marginal return to the managerial effort is less certain, which lead us to the negative relationship between incentives for managerial effort and company size. On the other hand executive pay-performance sensitivity may decrease due to the decreasing fraction of CEO influence in larger team size (Schaefer, 1998, p. 436).
Concluding, the board of directors, managerial power, regulatory systems and firm size all have contributed in some way to the inefficient contract concept and caused new agency problems. Subsequently, several studies have found a CEO pay without performance (Murthy, et al., 1975, Jensen and murphy, 1990b, and Albrecht and Jhin, 1978).
2.3 Million-dollar firms
Murphy (2012) stated that during the 1992 U.S. presidential campaigns, different candidates made several warnings that corporate boards shouldn’t pay CEOs additional compensation that was unrelated to productivity or performance. The
13
controversy over CEO compensation became a political issue. The most decisive statement came from Clinton who promised to end the infinite tax deductions for executive compensation. All compensation for employees above $1 million were defined by Clinton as unreasonable and therefore disallowed from a tax deduction. However, this statement caused major criticism: options were exercised in large numbers, large investment banks increased their bonuses and publicly traded companies threatened to go private. A modified regulation was composed. Omnibus Budget Reconciliation Act of 1993, section 162(m) ensures all non-performance based CEO compensation above $1 million as not tax deductible. The rule is only applicable to publicly traded firms and only applies remuneration to CEOs and the four highest paid executives as disclosed in annual proxy statements (Murphy, 2012).
The purpose of this regulation was to reduce excessive compensation (other than performance-based), however, the opposite occurred, an increase in the overall top executive compensation. According to a study of Perry and Zenner (2001), there is evidence that the 162(m) regulation has led to a decrease in compensation, an impact on the compensation structure and to a decline in salary growth rates, all in million-dollar firms with CEOs near or above a million dollar compensation level. Meanwhile, non-million dollar companies increased their compensation expenses to exactly $1 million because the cap may have created an acceptable CEO compensation standard (Rose and wolfram, 2002). Perry and Zenner (2001) also documented that compensations are increasing sensitive to stock return and that the sensitivity of CEO wealth relative to shareholder wealth has increased. Overall, their study found that the 162 (m) regulation have had a real economic impact on top executive pay and the link between pay and performance, and caused for an increase in pay-performance sensitivity in million-dollar firms. Rose and Wolfram (2002), also found evidence on the decrease in executive compensation in million-dollar firms.
This decrease in compensation in million-dollar firms implies the fact that companies are not likely to pay non-performance-based compensation above the million dollars. The new regulation ensures an additional negative impact on all performance-based CEO compensation above one million due to the non-detectability, which is a situation that many do not want to pursue (Murphy, 2012, p.75). However, both studies focus only on the impact of the regulation on the pay-performance relation in million-dollar firms and did not show the effects on this relation in non-million-dollar companies. Despite the fact that section 162 (m) is implemented with the purpose to reduce the non-performance-based compensation in million-dollar firms, this regulation might also affect the pay for performance in non-million dollar companies. Non-non-million dollar firms seem to take this non-million dollar limit as an acceptable standard for executive compensation even when the top executive compensation were not based on firm performance (Murphy, 2012 and Rose and Wolfram, 2002).
14
3. Research design
First of all, two hypothesis are derived from the literature described in previous sections and connects to the research question. In the following section, the research method is discussed and the choice of the variables are derived from the previous literature. Subsequently, the multiple linear regression models are implemented that will test the hypotheses. Finally, different sources for collecting the required data are indicated and the process of data collection is described. The research focuses on the total compensation of CEOs and performance in million-dollar firms and non-million-dollar firms of North American companies during a period from 2007 to 2013.
3.1 The hypotheses
The hypotheses are based on the above-described empirical studies and underlying literature in section 2. In order to answer my research question, the following hypothesis will be tested with a multiple linear regression:
= The relationship between firm performance and CEO’s total compensation is positive.
As described in section 2, there were four forces described by Jensen (1993): board of directors, capital markets, the legal/political/regulatory system, and product markets, that mitigated the agency problems by providing an increased pay-performance sensitivity. This increase is partly explained by the increase in stocks and stock options as compensation method (Mehran, 1995, p. 163). And yet, caused by the growing criticisms of shareholders, politicians, journalists and scientist played an important role. However, the most direct approach used in prior literature to solve the agency problem is by setting up an incentive contract. These issues generally ensured a positive relation between firm performance and top executive compensation (Murphy, 2012).
Different studies show a positive relation:
- Jensen and Meckling (1976) Variable compensation provides most of the incentives to do better for greater firm performance and solving agency problems. - Perry and Zenner (2001) showed an increase in the pay for performance
sensitivity.
- Murphy (2012) stated the ongoing criticism, new remuneration structures, and the many implemented regulations have generally ensured the rise in sensitivity for pay-performance.
- Mehran (1995) showed firm performance is positively related to CEO compensation, This is partly explained by the increase in stocks and stock options as compensation method.
- Clementi and Cooley (2009) concluded a significantly strong relationship between the executive payments and the created shareholder value.
15
- Hall and Liebman (1998) found an existing strong relationship between firm performance and CEO compensation and a massive increase in sensitivity for pay-performance.
In general, these studies described a significant link; executive pay is increasingly dependent on firm performance. I think my study will not differ from these results.
= The relation between firm’s performance and CEO compensation is positive in million-dollar firms and not positive in non-million-dollar firms.
In this study, the million-dollar hypothesis is based on The Omnibus Budget Reconciliation Act of 1993, section 162(m). Despite the insignificant results of Perry and Zenner (2001) on the relation between increased pay for performance and the million-dollar firm definition, they also found a decrease in compensation, a decline in salary growth rates and an increase in the sensitivity of CEO wealth relative to shareholder wealth (i.e. pay-performance sensitivity), all in million-dollar firms. Rose and Wolfram (2002), also found evidence on the decrease in executive compensation in million-dollar firms. This decrease in compensation in million-dollar firms implies the fact that companies are not likely to pay non-performance-based compensation above the million dollars. The new regulation ensures an additional negative impact on all performance-based CEO compensation above one million due to the non-detectability, which is a situation that many do not want to pursue (Murphy, 2012, p.75). For this reason and in section 2.3 described literature, I think my results will differ and I think this regulation is expected to lead to the significant pay-performance relation in companies with CEO compensation over $1 million.
These studies do not provide results on the impact of the regulation on the pay-performance relation in non-million-dollar companies. Therefore, it is difficult to adopt an expectation on this relation. However, Rose and wolfram (2002), and Murphy (2012) stated that non-million seem to take this million dollar limit as an acceptable standard for executive compensation. They often increased their compensation expenses to exactly $1 million even when this compensation was not based on firm performance. Despite the fact that these studies have not showed any results they do show that something (based on non-performance pay) is happening in these non-million dollar firms. Therefore, I think my results will show no positive relation between pay and performance in non-million-dollar firms.
3.2 Research Method
In this section, the research method is discussed and choice of variables based on previous literature is developed and explained. Subsequently, the regression models are implemented.
The dependent variable or the regressand is synonymous to the total compensation of the CEO. The total compensation of a CEO is measured by the sum of the following four relevant variables; salary, bonus, and other long-term-incentive bonus namely stocks and options awarded. These variables are the most commonly
16
used measures for CEO compensation (Frydman and Saks, 2006, & Bebchuk and Grinstein, 2005). Also, the study of Mehran (1995) shows 74% of all CEO’s total compensation is in salary and bonus. Salary is assumed to be the fixed pay of compensation and the bonus pay is assumed to be a part of the variable pay. Moreover, the other two variable compensations namely long-term incentive plans, i.e. the stock and options awarded, are included in the model because they, as the name suggest, give incentives to CEOs to perform well and ensure better business results. Equity awards can be valued in two ways namely, “realized” pay when stocks and options awards are actually exercised or “Grand-date” pay (i.e., grand-date stock price) when the awards are granted. I choose to adopt the grant-date approach because, in general, most studies since the 1980s have measured the equity awards using this valuation (Murphy, 2012, p. 7). Jensen and Murphy (1990b) among others, used the outstanding stock options to estimate CEO wealth. As describes in section 2 the compensation in the form of stock and option rewards has increased in recent years and makes an increasingly larger part of the variable compensation. The dependent variable and its database are given in table 1.
The independent variable is a proxy for firm performance. This is an important measure of the reasonableness of compensation. The choice of this performance variable supported by the agency theory and described in section 2. In previous literature, an ongoing discussion exists about the applied different measures for the firm performance variable. Some measures, such as stock return, applied by Murthy and Salter (1975) as for Albert and Jhin (1978), are discussed to be a good measure only for all-equity firms. Perry and Zenner (2001) find evidence the return on assets has a large impact on top executive compensation. They used another approach concerning the improved compensation disclosure that offers new insight into the pay and performance relation. They made a table of most applied performance measures by different compensation committees. In this table, 60% of the firms applied Net Income and was therefore the most frequently used variable. 20% of the firms used ROA, i.e. return on Assets. Also the study of Mehran (1995) ROA to represent firm performance. Their result’s found positive and significant coefficients of the variable on all CEO compensation. Another argument for applying this variable is that when compensation is determined accounting outcomes play an important role because they are reliable figures which give directors intelligence about the value added by the executive (Jensen and Murphy, 1990b). Hence, this research chose as well as many other studies to represent Return on Assets as firm performance, measured by the ratio of net income to the book value of the firm’s total assets. Finally, it is important to study the annual cash flow because; the independent variable of ROA is given per year.
In the regression, three different control variables are imposed because they are likely to affect compensation levels: two firm-specific characteristics (e.g., firms size and economic sector) and net income.
Smith and Watts (1992), Coughlan and Schmidt (1985), Albrecht & Jihn (1978) and Perry & Zenner (2001), among others, emphasize the concern of firm size towards compensation. The natural logarithm of total assets is used as a proxy for
17
firms size, which is also used by these researchers. See section 2.2.1 for a discussion of this relationship between firm performance and firms size.
Another additional control variable, that may also affect compensation and return on assets, is the industry specific variable called sector. Despite the fact that Baum, Sever, and Strickland (2004) indicated that a variable sector had no significant influence on the pay-performance relation, we consider sector as a valuable control variable, because Perry and Zenner (2001) stated that industry specific variables may also affect compensation. This sector is based on the S&P economic sector codes that are used to identify companies within any of the eleven broad economic industry groups as defined by Standard & Poor’s and shown in table 1. To control for this fact, different dummy variables for each sector are created. Sector-dummies are equal to 1 if the observation is for the said sector and to zero otherwise.
Moreover, net income is implied as the last control variable. It was by many studies considered as an important variable in determining the executive compensation (Perry and Zenner, 2001).
Finally, different year dummies are implied to separate the results of each year in our panel dataset. Year-dummies are equal to 1 if the observation is for the said year and to zero otherwise.
The dummy variable in our regression model is the million-dollar firm. The dummy variable equals one when a CEO earned more than one million dollars in total compensation at least once over the sample period and equals zero otherwise. “ROAMillion-dollar firm” is an interaction variable added to the model to estimate the relationship between return on assets and total CEO-compensation when an executive is currently working in a million-dollar firm as compared to a non-million dollar firm. The choice of the variables is described in section 2.3
This leads to the following empirical model to be tested by a multiple regression:
β ∗ β ∗ β ∗
β ∗ β ∗ β ∗
(4)
= natural logarithm of firm i’s sum of CEO’s salary, bonus, Stocks Awarded Grand Date Fair Value of Stock Awarded, and Options Awarded Grand Date Fair Value of Options Granted ($) in year t
α = constant
18
= natural logarithm of firm i’s total assets in year t to control for firm size
= net income or loss of firm i in year t
= dummy variable that equals 1 if firm i’s CEO’s total compensation in year t was greater than one million dollars and 0 otherwise
= dummy variables that equal one for a given S&P sector and 0 otherwise
= return on assets in a million-dollar firm
= dummy variables that equal one for given year and 0 otherwise
= error term
3.3 Data
Two different databases are used to collect the required data in this research. They’re both to be found in Wharton Data Services. The dependent variable, Total Compensation of the CEO, is extracted from the Executive Compensation database. As where the dependent variable and control variables are chosen from the Compustat database, which is specialized in annual accounting fundamentals. Both databases contain more than thousands of different North American firms. The date range has a point-in-time coverage from 1986 until the recent year. It includes S&P 500, Midcap, and Smallcap index firms. Together they cover the S&P 1500. Both databases are used by renowned researchers in high ranked articles (Perry and Zenner, 2001 & Mehran, 1995).
3.3.1 Data collection and adjustments for the dependent variable
As discussed above, this study selected the Total Compensation to represent the compensation variable of the CEO. For specific information about this variable a database namely Execomp was used. It presents all different executive compensation in American Dollars (USD). Herein one can select various variables out of the compensation categories. Subsequently, only the relevant compensation variables for this research model were selected; salary, bonus, and long-term-incentive plans namely stocks and options awarded to CEOs. The total compensation is the representative of the four components together (Perry and Zenner, 2001, pp. 459-461). In addition, all variables are yearly presented. Incorporated all of the above brings this research to a number of 17.635 observations.
The total compensation data is selected for the years 2007 up to and including 2013. We decided to leave 2014 out of the set because not all observations were shown. In addition, we decided to the start up from 2007 because the increasing presence of long-term incentive pay observations. Unfortunately, in the years
19
previous to 2007 almost all of this information was omitted and, therefore, irrelevant for our compensation data.
After selecting the relevant compensation variables, a new column is created to display of the total compensations of a CEO, which equals the sum of salary, bonus and stock and options awarded. Although there exist some disclosure regulations that must be followed by all publicly traded companies, there are still fiscal years for which there is no value of total compensation. 87 CEO years have been omitted since they covered all CEOs who did not receive any compensation for a given fiscal year. For a statistic efficient reliable test, it is important to remove outliers. These 14 observations of CEOs who earn less than 100.000 US are seen as outliers and have therefore been removed. Finally, this Total Compensation variable was modified with a natural logarithm transformation and the dependent variable dataset has now decreased to 17.264 CEO years.
3.3.2 Data collection and adjustment for the independent variables
In this study, the Return On Assets is selected to represent a performance measure of the firm. As for the regressors, we took a different dataset from a different database namely Compustat. Herein they present many different accounting variables distributed over the income statement, balance sheet, and cash flows. As described in section 3.3.3 only two variables are relevant four the representation of our independent variable namely ACT (Total Assets) and NI (net Income (loss)). By taking the ratio of net income to the book value of the firm’s total assets we receive the variable Return on Assets (ROA) in US dollars, per firm, per fiscal year. The ticker codes, which represent a company, were used to collect the same companies as we took from Compustat. An amount of 15.905 observations occurred. Subsequently, all negative, blank and zero values have been omitted from the dataset. so that an effective statistical test could be done.
Different dummy variables are created, which are called the million dollar firm or yearly dummies. The million-dollar firm dummy equals one when a CEO earned more than one million dollars in total compensation at least once over the sample period and equals zero if not. The yearly dummies are constructed as follows. For example, the 2007 yearly dummy equals one when the observation occurred in the year 2007 and equals zero if not, and so on. The interaction variable is created by multiplying the ROA with the dollar firm variable so the relation of million-dollar firms on ROA and Total compensation can be shown. Moreover, I took three control variables from the Compustat database. The variable firm size is based on the natural logarithm of the total assets (ACT) of the firm. There were no negative values presented. Net income was also selected from this database as well as for the economic sector variable. The same construction as for the year dummies applies for all sector dummies. The eleven broad economic industry groups of Standard & Poor’s, are described in table 1 below.
20
Table 1. S&P economic sector codes.
code Economic sector
970 Basic materials 925 Capital Goods 974 Communication Services 976 Consumer Cyclicals 978 Consumer Staples 935 Energy 800 Financials 905 Health Care 940 Technology 600 Transportation 700 Utilities
An amount 6181 observations have removed from the dataset. All these variable sector values as for all assets and net income values were left blank. A total of 9326 observations remained. I present the economic sector codes and their frequencies in Appendix A.
3.3.3 Merging the two datasets
The fiscal year and ticker codes have been selected to link both datasets simultaneously into one dataset. When the datasets were coupled, a number of 9002 fiscal CEO years from Execomp dataset did not correspond to the dataset from Compustat. Hence, after adjustments, a significantly lower number of observations will be used. However, there is still a large number of observations left namely 8262. This completed merged dataset will be used to analyze our regression model that is described in section 3.2.3. Stata used for this analysis. Table 3 briefly shows the applied adjustments.
Table 2. Removed observations from datasets.
S&P 500 Data N=CEO years
Available data Dataset ExeComp 17,365
Removed Total Compensation equal to zero -87
Removed Outliers -14
Available data Total Adjusted Dataset ExeComp 17,264
Available data Dataset Compustat 15,905
Removed Blank Observation -6,181
Available data Total Adjusted Dataset Compustat 9,724
Removed No matches with Compustat -9,002
21
4. Results and analysis
In this section, first the descriptive statistics of all different variables are presented and discussed. Supplemented by Table 4 which identifies a clear overview. Thereafter, a correlation matrix is deducted to test the research variables on correlation and the independent variables are tested on dependency on the basis of multi collionearity. Finally, regression analyzes are performed to test our hypotheses along with the summary and description of the regression results. The sample contains 8262 observation of CEO compensation in different firms for the given period from 2007 to 2013.
4.1 descriptive statistics
Table 4 provides descriptive statistics over 8262 observations for the CEO compensation and for the independent variables for period 2007-2013. Statistics of the sector and year dummies are given in Appendix A and B The mean total compensation is $4.652 the minimum is $123 and the maximum is $128.452 (all in thousands). This maximum total compensation was earned by Eugene M. Isenberg in the fiscal year 2008 who worked at Nabors Industries Ltd. His total compensation was almost equal to this variable compensation namely $127.727.000, his fixed income was $725.000. The highest return on Assets was 40.0338 at Azzurra Holding Corp in 2007 with a total compensation of $130.638 (far below the mean and almost equal to the minimum total compensation) earned by Daniel W. Rumsey. The minimum return on Assets occurred in Gulport Energy Corp, however, the CEO still received a total compensation of $518.668. This is remarkable. When comparing the two different observations, we can state that the highest return on assets observation, and thus “best-performed firm” in the relevant fiscal year, had total earnings of the CEO lower than the total compensation of the executive with the lowest return on assets observation.
Table 3. Descriptive Statistics
Part 1: Descriptive Statistics for CEO compensation (N=8262)
(in thousands of $) Mean Std. Dev Min Max
Salary 828.32 411.09 0 8100.00 Bonus 254.07 1749.63 0 76951.00 Stocks 2235.54 3490.30 0 66549.90 Options 1334.37 3135.65 0 90693.40 Total compensation 4652.30 5631.64 123.05 128452.00 Ln(Total compensation) 8.0371 .8963 4.8126 11.7633
Part 2: Descriptive Statistics for independent variables (N=8262)
Mean Std. Dev Min Max
ROA .1278 .5795 -8.9421 40.0338
LnAssets 6.7069 1.4791 -1.9105 11.5275
NetIncome 480.5324 1998.28 -29580 45220
Million-dollar firm .8922 .3102 0 1
Note: ROA is the return on assets calculated by dividing Net Income by Total Asset.
22
The average compensation components for the fiscal CEO years from 2007 to 2013 are presented in figure 1. Total compensation and variable compensation move simultaneously over time. Salary made a small increase from $766.626 to $884.317. As where the variable bonus experienced a small decline from $296.671 to $ 186.926. Where stocks increase over the years, the variable options made a small decrease. As described by Murphy (2012, pp. 97-98), the stock market crash of the early 2000s was accompanied by the internet bubble in 2000 and the terrorist attacks of 2001. The expectation of future stock prices declined, options were canceled and replaced with restricted stock. This continued until 2011. In 2009, all components, except for salary, had a downturn. This is supported by Murphy (2012, pp. 97-98), as the result of the economic crisis which occurred in 2007-2008. In addition, from the graph, we can conclude that the long-term variable intensive pays, namely stocks and options awarded, are important and weight more heavily in the total variable earnings compared to the bonus variable.
Figure 1. Graphically average compensation components
The average dependent (natural logarithm of total compensation) and independent (return on assets) variable components for the fiscal CEO years from 2007 to 2013 are presented in figure 2. Total compensation made a small increase over the years from 4502 to 5296 (in thousand $). As where the overall dependent variable return on assets experienced an overall decline from 0.1978 to 0.1562 over the years. The ROA was volatile with a standard deviation of 0.5795. It had experienced a significant decline in 2008 as where the total compensation made a decline in 2009. Thereafter, both ratios represent an upward trend. The explanation for these effects is in line with the statement of Murphy (2012) given above namely, an occurring economic crisis in 2007-2008. From the graph, we can conclude that the ratios do not move
0 1000 2000 3000 4000 5000 6000 2007 2008 2009 2010 2011 2012 2013 CEO compensation (i n thousands of $) Executive Compensation total compensation Salary Bonus Options awarded Stocks awarded variable compensation
23
simultaneously over time. This effect is shown due to the fact that a large part of total compensation is mostly based on the performance of the previous year. When the return on assets decreased in 2008 it resulted in a decrease of the total pay in the beginning of 2009. Murphy (2012, p.97) shows this same movement in CEO compensation. The average components of compensation and return on assets per fiscal year are presented in Appendix B.
Figure 2. Graphically average dependent and independent variables
4.2 Correlation matrix and Multicollinearity
In this section, a correlation matrix is deducted to test the research variables on correlation. The results are shown in table 4. This table proves a significance correlation between the return on assets and the total compensation which is in line with our expectations. The control variables firm size and Net Income have a significant correlation with the independent variable return on assets (i.e., firm performance) as for the dependent variable (total compensation). This is in line with the literature of Perry & Zenner(2001) who emphasized the concern of firm size and net income towards compensation. A new correlation matrix for the control variable
0 0.05 0.1 0.15 0.2 0.25 2007 2008 2009 2010 2011 2012 2013
ROA
ROA 7.7 7.8 7.9 8 8.1 8.2 8.3 2007 2008 2009 2010 2011 2012 2013ln (Total Compensation)
ln (TC)24
sector is provided in Appendix C. From this table we can state that only Utilities Consumer Technology and transport have a significant correlation with the return on assets. We can also conclude that every sector except for Utilities and Basic Materials and consumer have a significant relation with the dependent variable. The dummy variable million-dollar firm in table 5 shows a significance correlation with total compensation and ROA. This is an obvious result since this dummy variable was based on the total compensation of a CEO
Table 4. Correlation Statistics
Ln(TC) ROA LnAssets NetIncome Million$ F
Ln(Total Compensation) 1
ROA 0.0604*** 1
LnAssets 0.7091*** 0.0322*** 1
NetIncome 0.2858*** 0.1608*** 0.4259*** 1
Million-dollar firm 0.5858*** 0.0328*** 0.3960 *** 0.0801*** 1
Note: * indicates significance at the 10% level. ** indicates significance at the 5% level. *** indicates significance
at the 1% level Ln(TC) is the natural Logarithm of executives Total Compensation ROA is the return on assets calculated by dividing Net Income by Total Asset. LnAssets is the natural Logarithm of Total Asset, that is used as a proxy of firm size
Million-dollar firm is a dummy which is 1 for Executive Total Compensation greater than one million
In addition, we used our variance inflation factors (VIF) findings in table 6 to see if there is any multicollinearity, i.e., a perfect linear relationship between the independent variables. When the VIF amount increases above 10 the estimators of the coefficients will need to be further investigated. Removal of an independent variable is not of interest since all values are below 10.
Table 5. Variance Inflation factors
Variable VIF ROA 1.03 LnAssets 1.47 NetIncome 1.27 Million-dollar firm 1.20 Mean VIF 1.20
25
4.3 Results
In this section, the results of the regression analysis and, therefore, OLS estimators are described along with the evaluation of the hypotheses. The results are presented in table 6. The sample contains 8262 observations.
As seen in table 8, there is no evidence that the coefficient of the return on assets, which is the proxy for financial performance measure namely firm performance, is significantly different from zero when considering the simple regression on executives compensation in model 1. However, after controlling for firms specific factors such as firm size, net income, and sector, and year-dummies the ROA shows a positive effect on CEO compensation with a 5% significance level (refer to model 2). Subsequently, when adding the dummy million-dollar firm (model 3), the coefficient even becomes significant at a 1% level. Yet, the last model shows an insignificant coefficient of the ROA when the interaction term is implemented.
The natural logarithm of total assets is a proxy for firm size and is suggested in the different literature as an important firm-specific factor in determining top executive compensation. All three models show a positive relation between firm size and executive compensation. The obtained results are significant at the 99% confidence level. Net income and sector are also suggested as important firms specific factors. The coefficient of net income is negative and significant at the 1% level in model 2. However, after controlling for the million-dollar firm and interaction term (refer to model 3 and 4), the variable is negative but insignificant All sectors are positive and significant except for financials which is only significant at a 5% level in model 2, capital goods which is only significant at a 10% level in model 2, and utilities which in any case insignificant.
In model 3 the million-dollar firm dummy variable is introduced. Both models (3 and 4) show a significant and positive outcome, which is an obvious result since the dummy is based on one million dollar compensation.
In model 4, the interaction term “ROAMillion-dollar firm”, is added. It is positive and significant,
Robust regression results are shown in Appendix D. It offers an alternative to my OLS regression that is less sensitive to outliers and still defines a linear relationship. The new outcomes obtained by this regression are not significantly different with an exception of the coefficient of the ROA in model 1 and 4. This coefficient now shows a significant (at 1% level) positive effect in model 1 and a significant (at 10% level) negative effect in model 4
26
Table 6. OLS Estimates
Dependent variable – Ln(Total Compensation)
Independent Variable (1) (2) (3) (4) Constant 8.025151*** 4.821337*** 4.604619*** 4.595155*** [.0154697] [.0556623] [.0446697 ] [.0446037 ] ROA .0934641 .0677369** .0490285*** .0073629 [.0999462] [.0319767] [.0199611 ] [.0103593 ] LnAssets .4410187*** .346567*** .3469974*** [.005267 ] [.0050546] [.0051198 ]
NetIncome -.0000213*** -2.21e-06 -6.68e-06
[5.37e-06 ] [4.89e-06 ] [5.06e-06 ]
Sector-dummies included No Yes Yes Yes
Million-dollar firm 1.030203*** 1.016783***
[.0172761 ] [.0175167 ]
ROAMillion-dollar firm .1414897***
[.0289866]
Year-dummies included No Yes Yes Yes
Regression Results (1) (2) (3) (4) R-squared 0.0037 0.5281 0.6332 0.6349 F-Value 0.3497 485.50 711.38 682.36 Significance 0.87 0.0000 0.0000 0.0000 Number of observation 8262 8262 8262 8262
Note: Robust standard errors are reported between the brackets. * indicates significance at the 10% level. **
indicates significance at the 5% level. *** indicates significance at the 1% level. ROA is the return on assets calculated by dividing Net Income by Total Asset. LnAssets is the natural Logarithm of Total Asset, that is used as a proxy for firm size. Sector-dummies are equal to 1 if the observation is for the said sector and to zero otherwise. All observations are compared to the sector transportation, which was omitted from the model in order to avoid a dummy trap. The Million-dollar firm is a dummy variable which is 1 for Executive Total Compensation greater than one million and 0 otherwise. The ROAMillion-dollar firm is an interaction term added to the regression to estimate the relationship between return on assets and executive compensation when an executive is currently working in a million-dollar firm as compared to a non-million dollar firm. Year-dummies are equal to 1 if the observation is for the said year and to zero otherwise. All observations are compared to the year 2007, which was omitted from the model in order to avoid a dummy trap.
27
4.4 Discussion
According to the results obtained, the coefficient of the return on assets is positive and significant when all control variables are included, excluding model 4 where the interaction term was introduced. The outcomes in model 2 and 3 indicates that ROA has a positive effect on CEO compensation. This is in line with our expectations and the findings in prior research (Jensen and Meckling (1976); Perry and Zenner (2001); Mehran (1995), among others). However, the outcome in model 4 has an insignificant value of ROA. The result of this effect is caused by the implementation of the interaction variable which includes a distinction in the pay-performance relation between million-dollar firms and non-million dollar firms. The insignificant coefficient of the ROA indicates no relationship between pay and performance in non-million-dollar firms. However, when performing a robust regression, outcomes differ and return on assets now shows a significant (at 10% level) negative effect in model 4. Subsequently, when performing a regression only on non-million dollar firms, we also find a significant negative relationship (Appendix D, model 5). The outcome indicates that a decrease in return on assets in non-million-dollar firms will lead to an increase in executive compensation.
As described in section 2.3, the studies of Rose and wolfram (2002) and Perry and Zenner (2001) only provide results on the insignificant results of the regulation on the pay-performance relation in million-dollar companies and not for non-million-dollar firms. However, my results are in line with my expectations described in section 3.3, there is no positive relation between pay and performance in the non-million-dollar firm. A reason for this negative effect could be “One-size-fits-all” believe on corporate governance regulation as described in section 2.2.2. Not every company is the same and some regulations have different (sometimes opposing) effects. This might be the case in the section 162(m) regulation, which created an increase pay for performance sensitivity in million-dollar firms but created an unwritten million-dollar compensation ceiling in non-million-dollar firms. As Rose and Wolfram (2002) and Murphy (2012) mention, because non-performance-based pay under $1 million has no effect on tax deduction, non-million dollar firms seem to take this million dollar limit as an acceptable executive compensation standard and increase their CEO compensation up to $1 million even when firm performance is negative.
However, we cannot automatically claim that all of this negative effect is attributable to the million-dollar regulation. Another reason might be: when testing for firms size, non-million dollar firms tend to be smaller firms. This is confirmed by the results of Perry and Zenner (2001) who also state that smaller firms have higher CEO ownership levels. Corporations with high CEO ownership are companies were managerial power can be assumed to play an important role. As described in section 2.2.2 greater managerial power leads to managers who take opposite actions from creating (destroying) long-term firm value (Bebchuk and Fried, 2004, p. 648). Greater managerial power can also lead to companies with poorer performing governance structure, in which executive could affect the selection of the board, which results in
28
higher levels of CEO pay and worse firm performance (Core, et al., 1999). And last, this significant relation between firm size and executive compensation gave managers incentives to pursue expansion of their firm (Murphy, 2012, p.62). Non-million-dollar firm managers might want to follow these incentives even when they are decreasing firm value (Bebchuk and Fried, 2004, p.670).
Considering the model 2 and 3, why do these results show a different (significant positive) outcome on return on assets? A reason for this effect could be the much smaller sample of million-dollar firms with 891 observations, allowing the million-dollar firms dataset with 7371 observations to prevail (Appendix E).
The negative and significant coefficient of the control variable net income in model 2 is remarkable since it suggests that an increase in net income results in a decrease in CEO compensation. The outcomes in model 3 and 4 indicates no relationship between net income and executive compensation. This is in contrast to the expectations based on the results by Perry and Zenner (2001) who stated that 60 percent of all compensation committees use net income as a financial performance measure to determine CEO compensation. However, in the end, Perry and Zenner (2001) stated that they did not use basic net income because of its expected correlation with firms size. But, when testing the regression for multicollinearity in section 4.2, we did not find any troubling results. Perry and Zenner (2001) used another form of net income namely earnings per share, these results showed an insignificant effect as well. This can be explained by the fact that salaries are set before the contemporaneous performance measures are known and, therefore, have no influence on CEOs total compensation (Perry and Zenner, 2001, p.469).
Furthermore, almost all sectors seem to have a significant effect on CEO compensation except compared to the sector transportation. However, Perry and Zenner (2001), Rose and Wolfram (2002), among others, did not introduce the sector variables to their models, because they are not of interest in determining the executive compensation when other firm-specific control variables are added (i.e., firms size and net income).
The natural logarithm of total assets shows a positive relation between firm size and executive compensation in every model. This implies that executives of larger companies receive higher compensations. This is in line with the results of Bebchuk and Fired (2004), Tosi, et al. (2000), Perry and Zenner (2001), among others, who also find significant positive relations. Firm size appears to be the most important control variable in our model, however, one might consider implementing other by the literature know determinants that may affect the CEO compensation. For example CEO tenure, implemented by Perry and Zenner (2001) to describe the quality and power of the CEO by implementing the number of years of which the CEO has held his position.
The implemented interaction term “ROAMillion-dollar firm” in model 4 shows a positive and significant coefficient which is consistent with my expectations: CEO compensation has a positive relationship with firm performance in million-dollar firms. However, this does not support the prior research done by Perry and Zenner (2001) who found no empirical evidence in million-dollar firms on the fact
29
that the CEO compensation above the one million dollars is considered to be based on firm performance. A possible reason for this difference in results could be the difference in dataset periods. Murphy (2012, pp. 97-98) showed a significant increase in stock grants (i.e., restricted stock and performance shares based on performance criteria), between 2002-2011, the results of this paper involved CEO compensation 2007-2013. However, Murphy (2012) showed no increase in stock grants between 1992-2001, Perry and Zenner (2001) had a dataset based on 1992-1997.
Although we found a positive and significant coefficient for the interaction variable “ROAMillion-dollar firm” which tells a lot about the relationship between CEO pay and firm performance in million-dollar firms, we cannot automatically claim that all of this positive effect is attributable to the million-dollar cap regulation. There might be several other factors affecting this positive effect. First of all, as mentioned earlier in this discussion, million-dollar firms are tend to be large companies where CEOs have little ownership. On the one hand, this could mean that due to a lack of shareholder equity they do not receive the appropriate incentives and are subsequently cause agency problem. However little ownership means they have less control in determining their own non-performance-based pay and are more likely to receive pressure from shareholders to increase firm value. Second, large firms have governance characteristics which cause for in increased sensitivity in pay for performance based on shareholders’ interest (Perry and Zenner, 2001). Furthermore, external board members, which in large firms are elected by the shareholders appear to play an important role in determining this pay for performance (Horstmeyer, 2011). Finally, large publicly traded companies are often the target of shareholder activists and exert pressure by pushing corporations to increase the relation between executive wealth and increased firm value (Murphy, 2012, pp.132-133).
