Research question thesis:
Performance of the S&P 500 listed firms in relation to CEO payments since the financial crisis of 2008.
Student: Bas Weinberg
Student number: 10089195 Thesis supervisor: Aaron Kamm
Field: Organizational Economics Faculty: Faculty Economics and Business Study: BSc Finance and Organization
Table of content
1 Introduction 2
2 Literature
2.1 The start of the financial crisis of 2008 3
2.2 Agency Theory 4
2.3 Corporate governance in U.S. 5
2.4 Prior Research 6
3 Hypotheses 8
4 Research Methodology
4.1 Sample, Data Source and Time 10
4.2 Variables of Interest 4.2.1 Dependent Variables 11 4.2.2 Independent Variables 11 4.2.3 Control Variables 12 4.3 Research Method 13 4.4 Robustness Checks 15 5 Results 5.1 Descriptive Statistics 16 5.2 Hypothesis 1 21 5.3 Hypothesis 2 22 5.4 Robustness Checks 5.4.1 Log-Transformed Model 24 5.4.2 Reversed Relationship 25
6 Conclusion and summary 26
References 30
Table 1-3: Descriptive Statistics 16
Figure 1-3: Descriptive Statistics 16
Table 4: Results Hypothesis 1 32
Table 5: Results Hypothesis 2 34
Table 6: Results Log-Transformed Model 36
Table 7: Results Reversed Relationship 38
Appendix 1: List of S&P 500 companies included 40
1 Introduction
In November 2008 Richard Fuld, CEO of Lehman Brothers from 1994 until 2008 when it filed for bankruptcy protection, was called to testify before a U.S. Congressional committee examining the unexpected collapse of Lehman Brothers. The legislators wanted to know from Fuld what he had achieved to take home $500 million over the last 9 years (Nordberg, 2012, pp. 2-3).
Currently, following the Google news page, many shareholders and also increasingly stakeholders still question the executive remuneration package of their executive. It is important to know what effect executive payment has on firm performance. Central questions in the discussions found on the internet are: Do executives really need to be paid what they are paid? And is firm performance directly related to the payments made to the top executives? Before judging the excessive payments we should ask ourselves if there is scientific prove for paying those amounts.
There exist many empirical studies around the world which has studied the relation between CEO payments and firm performance. Examples are Murphy and Jensen (1990), Hall and Liebman (1998) and Zhou (2000). These studies are done years before the financial crisis of 2008. Most of other studies which examined the relation are done before the financial crisis as well. The financial crisis and taking in consideration all changes and requirements which firms had to adapt in their businesses due to the crisis could differentiate this study to the already existing ones. The main research question of this thesis is if performance of the S&P 500 listed firms are in relation to CEO payments since the financial crisis of 2008.
The three hypotheses tested for every year in the research period are: S&P 500 listed firms’ performance is significant positively related to total CEO payments since the financial crisis of 2008; S&P 500 listed firms’ performance is significant positively related to CEO cash payments since the financial crisis of 2008 and the reversed pay-performance relation is positive and significant. The relation between total CEO payments and firm performance is examined with two different models, a linear and a log-transformed model. The first model found evidence for a positive significant relation in the years 2009 and 2012. The second model found evidence for a positive significant relation in 2008, 2009 and 2012. The CEO cash pay-performance relationship show positive significant results for 2009 and 2012. Finally the reversed pay-performance model show positive significant results for 2009 and 2
2012. Different proxies are used for the independent and control variable. It is shown that the choice of proxies for firm performance variables is very important for showing significant relationships. Next to this, evidence is found that not all years after the financial crisis of 2008 show positive significant relationships, where former literature mostly showed a clear positive relationship over periods of time.
Primarily this thesis will give a theoretical background in chapter 2. Thereafter hypotheses of interest for this research will be discussed in chapter 3. Chapter 4 focusses on the sample and research methodology. Chapter 5 discusses and analyzes results obtained out of the regressions. Finally chapter 6 comes up with a short summary, conclusion and suggestions for further research.
2 Literature
In this chapter a theoretical background is given. In the first subchapter the start of the financial crisis is discussed. In 2.2 the Agency Theory is described. Thereafter the corporate governance structure in the U.S. is discussed. Finally some prior studies related to the thesis topic are described.
2.1 The start of the financial crisis of 2008
This subchapter is about the start of the financial crisis. To put things in perspective for this thesis, it is important to know what is considered as the start of the financial crisis.
There is not one clear starting point of the recent financial crisis. Since it hasn’t happened long ago many researchers are still investigating the crisis. In this thesis the collapse of the Lehman Brothers is seen as the start of the financial crisis. Lehman Brothers, an investment bank, was deeply involved in the asset-backed security market. It was the subprime mortgages and derivatives contracts which caused the collapse of the investment bank (Nordberg, 2011). Lehman Brothers borrowed large amounts to create leverage for their investments. A significant large part of the borrowed amounts were invested on the mortgage market. When the subprime mortgage crisis started, and Lehman Brothers had leveraged their investments done in the mortgage market, they were quite vulnerable. The asset-backed securities were traded between banks and institutions all over the world. This resulted in the vulnerability of a wider spread financial system. In 2008 Lehman Brothers was confronted with enormous losses due to ongoing subprime mortgage crisis. 3
Finally when Lehman Brothers filed for bankruptcy on September 15 in 2008 and the U.S. governments refused to bail out, wider financial markets were affected.
The collapse of Lehman Brothers was the first clear signal of a financial crisis and had a huge impact on the wider financial markets and the economy. Since the collapse almost brought down the world’s financial system and gave a first clear signal that the collapse would affect wider spread systems, it is seen in this thesis as the start of the financial crisis (The Economist, 2013).
2.2 Agency Theory
This subchapter gives a better understanding in the problems arising by a separation between ownership and control. This is important because of the ownership structure of the S&P 500 listed firms.
Central to The Modern Corporation and Private Property, a book published in 1932 by Berle and Means, was the divorce of ownership from the control of the modern corporation. In 1976 Jensen and Meckling identified this as the agency problem. The foundation of the agency theory was established. Jensen and Meckling described the divorce between ownership and control as the agency relationship in which the principal, the owner of the firm, contracts the agent, the manager of the firm, to perform services on behalf of him. They viewed the firm as the nexus of a set of contracting relationships among individuals.
Jensen and Meckling (1976) argue that if the agent and the principal are both utility maximizers, there is a good reason to believe that the agent will not always act in the best interests of the principal. They further argue that the principal can limit divergences by aligning incentives for the agent and by monitoring the agent to limit the divergent activities. Next to this the manager can expend to reduce agency divergence and conflict.
Maskin and Tirole (1992) analyze the principal-agent relationship when the principal has private information as a three-stage non cooperative game: proposal of the contract; acceptance or rejection and finally contract execution. A crucial difference to the classical theories is that now the principal has private information. In 2003 Bénabou and Tirole argued that performance incentives, offered by the informed principal, can undermine the intrinsic motivation of the agent. They showed by their model that extrinsic motivation or incentives can crowd out intrinsic motivation.
Hendrikse (2003) argues that an interesting principal-agent model has three ingredients: available surplus; conflict of interest and asymmetric information. He assumes that the agent knows more than the principal and addresses three solutions for this. The first is that the principal has to structure the payoffs in the contract such that the interests are better aligned (only with complete information). Secondly the principal and the agent have to change the information structure. Or finally change the principal or the agent.
The view of the board is built around the separation of ownership and control and thereby to act in the interests of the principals. The board has to direct the efforts of their agents to advance shareholder value (Nordberg, 2011, p. 38). Getting back to the research questions, giving agents contracts in which they are paid on performance should align firm performance with CEO payments.
2.3 Corporate governance in U.S.
This subchapter is about corporate governance in the U.S. It gives a definition and describes a potential corporate governance structure in the U.S. Understanding the corporate governance structure in the U.S. allows to better analyze the results and draw conclusions.
There is not one clear definition for corporate governance. Corporate governance is seen from multiple perspectives. Grazell (2006, p. 293), explains it as “a system of procedures and structures that is used to govern and control the firm, within a field of forces of involved stakeholders, with the goal to contribute to the utility of these stakeholders”. In the U.S., most of the companies are able to choose in which state they want their legal seat without having the main part of their operations in the same state. States are competing to offer the best terms for corporations. About half of the U.S. stock market listed companies have their legal seat in Delaware. The business judgment rule, a provision in Delaware, describes that decisions by the board in the normal line of business may not be challenged by shareholders. This provision results in the ignorance of recommendation done by shareholders (Nordberg, 2011).
Nordberg also argues that the provision in Delaware gives managers a large amount of power. They can choose directors and influence the policies, since shareholders can only vote in favor of board candidates and not against them. Actually he argues that shareholders have to choose between directors nominated by directors which are influenced by managers. Delaware law allows a powerful manager to become CEO and Chairman. Auditors
with the task of checking the CEO performance will be selected by the manager. And his salary, set by the board, will be based on the performance, while performance measures are set by the board. Nordberg argues that many other states have the same rules concerning the appointment of the board.
Nordberg (2011) shows cases like WorldCom and Enron that gives the potential drawbacks of this form of corporate governance. While other cases like Microsoft and General Electric, give the potentials benefits of this system.
2.4 Prior Research
This subchapter gives an overview of prior research that has investigated the relation between firm performance and CEO payments and it gives an overview of other important literature related to the topic. Many empirical studies around the world have studied this relationship. Only a selection of those empirical studies will be mentioned in this subchapter. The described studies are useful to learn how to manage and steer this study in the best possible way.
Primarily the study of Murphy and Jensen (1990) is discussed. Murphy and Jensen studied the pay-performance relation with change in shareholder wealth as independent variable and change in CEO wealth as dependent variable. They were the first to use total CEO compensation to examine the relation between performance and pay. Their final sample consisted out of 1,049 large U.S. corporations, studied over the period 1974-1986. They found that CEO wealth changes $3.25 for every $1,000 change in shareholder wealth.
Hall and Liebman (1998) studied the pay-performance relationship with absolute measures as well with relative measures. The latter is known as the pay-performance elasticity. Change in total CEO wealth is used as dependent variable and firm performance as independent variable. Their sample consisted out of the 478 largest publicly traded U.S. companies over the period 1980-1994. The authors found that the elasticity tripled from 1.2 in 1980 to 3.9 in 1994. This shows that both the level of CEO compensation and the sensitivity of compensation to firm performance have risen dramatically since 1980, largely because of increases in stock option grants. Hall and Liebman also say that this relationship is almost entirely generated by changes in the value of CEO holdings of stock and stock options. They found that stock and stock option revaluations increase median CEO wealth by about 1.25 million dollars for a 10% increase in firm performance. This increase in median
for salary and bonus is $23,400 for the same 10% increase in firm performance.
Zhou (2000) studied the relationships of CEO pay, firm size and corporate performance. He did this for 755 Canadian firms over the period 1991-1995. The evidence found is in line with previous studies: CEO pay rises with firm size and compensation is tied to company performance. For every $1,000 change in shareholder wealth, Canadian CEOs’ earnings change by $5.62. Zhou used change in CEO wealth as dependent variable and change in shareholder wealth as independent variable.
Further studies that investigated the relation between CEO pay and firm performance are discussed in short. Kato and Kubo (2006) studied Japanese firms over the period 1986-1995. They found that CEO cash compensation increased about 1.3% for every 1% increase in ROA. Conyon and Schwalbach (2000) studied the relation between shareholder return and CEO cash payments in the U.K. and Germany over the period 1968-1995. They reported a positive significant relation for firms in both countries with an elasticity of 0.067 (U.K.) and 0.071 (Germany). Duffhues and Kabir (2008) studied the pay-performance relationship for Dutch listed companies over the period 1998-2001. Their robust empirical analyses haven’t found evidence for a positive relationship. This result doesn’t underpin the idea that managers’ interest can be aligned with shareholders’ interest. But it is in line with the idea that managers can influence their own pay if they are powerful.
Next to the already discussed studies which are in line with the research question, some other interesting studies related to the topic are described. Tosi et al. (2000) found, through meta-analytic reviews, that firm size accounts for 40% of the variance in total CEO compensation, while firm performance accounts for less than 5% of the variance. Cichello (2005) studied the impact of size on pay-performance sensitivities over the period 1993-2000 in the U.S. He provides evidence that the negative effect of variance in stock return on estimated pay-performance sensitivity could mostly be diminished by controlling for firm size. Aggarwal and Samwick (1998) say that the existing empirical evidence, which support that executive compensation is of central importance in the principle-agent model, is quite weak. Using a comprehensive sample of executives at large corporations, they found strong empirical evidence on the prediction that the executive’s pay-performance sensitivity is decreasing in the variance of the firm’s performance. Besides, Core et al. (1999) found that the larger the board, the easier for the CEO to manipulate the board and increase his remuneration. Finally Brick et al. (2006) found that excess compensation for both director 7
and CEO is associated with firm underperformance. They found the reason for this in cronyism. The different described studies in this paragraph show that compensation of the CEO is not completely determined by firm performance. This is of importance when making conclusions.
Most of the discussed studies show a positive relationship between firm performance and CEO payments. This thesis is going to conduct a research somewhat in line with the first three mentioned studies. Comparing is hard since all studies have their own identity. Neither of the discussed studies above have done their research after the financial crisis of 2008. This thesis could build on the existing literature by taking the financial crisis in consideration and testing hypothesis not over a period of multiple years, but for every year in the research period.
3 Hypotheses
This chapter will give information about the hypotheses with respect to the research question formulated as the performance of the S&P 500 listed firms in relation to CEO payments since the financial crisis of 2008.
In 2.2 Agency Theory it is argued that the principal can limit divergences by aligning incentives for the agent. From this argumentation it follows that firm performance should be related to CEO payments. Subchapter 2.4 Prior Research showed that almost all former studies found evidence to conclude there is a positive relationship between firm performance and CEO payments. Most former studies ratified the argumentation by Jensen and Meckling (1976).
This study expects, in line with former research and theory, that CEO payments are positively related with S&P 500 listed firms’ performance since the financial crisis of 2008. The mentioned studies are all done before the financial crisis. However it is expected that the agency theory holds during the financial crisis of 2008. This thesis will test the following hypothesis:
H1: S&P 500 listed firms’ performance is significant and positively related to total CEO payments since the financial crisis of 2008.
Based on financial intuition this study believes that in case of an economic downturn and imperfect incentives alignments, the strength of the relationship between CEO payments and firm performance could weaken. Argued in a different way it will mean that if firm performance is worse during a crisis and incentives aren’t perfectly aligned, the economic downturn could enlarge imperfection of the alignments and this could finally result in a weaker relationship close to the years of the start of the financial crisis. If evidence is found for a significant positive relationship as hypothesized under H1 for all years in the research period, it will be estimated if there is a difference in strength in relation between firm performance and total CEO payments over the years researched.
Many former studies tested if cash compensation (salary and bonus) is related to firm performance. Hall and Liebman stated that many researchers have limited their study only looking to the relation between firm performance and cash compensation and that their findings with respect to the relation were quite small (1998, p. 668). It is interesting to look for the relation between firm performance and cash compensation, because is leaves out equity based compensation. It can say something about CEOs paid on relative performance measures rather than performance measures in which the wealth of shareholders are taken into account (Gibbons and Murphy 1991; Hall and Liebman 1998). This thesis will test the following hypothesis:
H2: S&P 500 listed firms’ performance is significant and positively related to CEO cash
payments since the financial crisis of 2008.
A third hypothesis is formulated for the reversed pay-performance relationship in subchapter 4.4 Robustness Checks. After the hypotheses are tested, this thesis will come with conclusions about relevant results and suggestions for further research.
4 Research Methodology
This chapter gives information about the research design and methodology. Subchapter 4.1 will discuss the sample, source of the data and research period. Next it will describe all variables of interest. Thereafter 4.3 will give the methodology and regression models used to test the hypotheses. Finally robustness checks are described.
4.1 Sample, Data Source and Time
This subchapter discusses the sample, data source and research period used to conduct this study. It gives a better understanding of the sample and time frame of this study.
The study focuses on the S&P 500 listed firms in order to find a relation between CEO payments and firm performance. The S&P 500 index includes 500 leading, based on market capitalization, firms listed on NYSE or NASDAQ in the U.S. and captures approximately 80% coverage of available market capitalization (Standard & Poor’s, 2013). The S&P 500 gives a good representation of all U.S. listed firms. The focus is on the U.S. since the financial crisis started in the U.S. and data about the variables are easily accessible for the S&P 500 listed firms. Data is obtained from Wharton Research Data Services. WRDS is a web-based business data research service (2013a). WRDS contains Compustat which contains ExecuComp, a database that offers information about annual executive compensations and company financials. The research period is between the 1th of January 2008 and the 31th of December 2012. Hypotheses are tested for every year in the research period. Former studies like Murphy and Jensen (1990) and Zhou (2000) used year to year changes in their dependent and independent variables. This thesis stays in line with former studies so datasets are needed for every year between 2007 and 2012. The year 2008 is taken as starting point of the research since this study is looking for a relation between firm performance and CEO payments since the financial crisis of 2008. The year 2013 is left out of this research due to the unavailability or incompleteness of information.
Not all S&P 500 firms are included in this research. Only firms listed for the whole period between 2007 and 2012 are included (leaves out 15 firms). Next to this all needed data to run regressions should be available for the period 2007-2012 (leaves out 33 firms). Besides, the CEO’s total compensation should be more than $50,000 in all years between 2007 and 2012 (leaves out 10 firms). The reason for setting the last criterion is to ensure compensation packages are set or at least a part of it, for the extrinsic motivation of CEOs. It is in line with the argumentation of Jensen and Meckling (1976) that the principal can limit divergences by aligning incentives for the agent. After removing firms a final sample of 442 U.S. S&P 500 listed firms are examined over the period 2008-2012. Appendix 1 gives a review of all included firms by full name. An additional note to the data is that the ROE for the Lockheed Martin Corporation is changed to its correct value of 527.9 in the fiscal year 2012 (iStock Research, 2013). WRDS showed a ROE of 7038.462.
4.2 Variables of Interest
In this subchapter the variables of interest for this study are discussed. The dependent, independent and control variables are explained and given. All monetary variables are represented in dollars.
4.2.1 Dependent Variables
Two different dependent variables are used in order to test the first two hypotheses formulated in chapter 3 Hypotheses. This section describes the dependent variables used in this study. The dependent variables are expressed in thousands of dollars.
To test H1 the dependent variable total CEO payments (TOTALp) is used. Murphy and Jensen (1990), Hall and Liebman (1998) and Zhou (2000) all used CEO wealth as dependent variable, where total CEO wealth can be explained as total CEO payments. Following the agency theory CEOs should be compensated on the base of firm performance. Argued in a different way it means that CEO compensation is a result of firm performance. Total CEO payments, which is found in ExecuComp as TDC1, consist of: salary, bonus, other annual, total value of restricted stock granted, total value of stock options granted (using Black-Scholes), long-term incentive payouts (LTIP), and all other total (WRDS, 2013b).
Cash payments (CASHp), comprised out of salary and bonus is used as dependent variable in order to test H2. Gibbons and Murphy (1991), Hall and Liebman (1998), Murphy and Jensen (1990) and Zhou (2000) used cash compensation as dependent variable in order to test if there is a relation between salary and bonus and firm performance. Using cash payments and leaving out equity compensation enables this study to look at the change in cash payments by a change in firm performance without taking equity based payments into account. Salary and bonus can be found in ExecuComp respectively as SALARY and BONUS. 4.2.2 Independent Variables
Multiple measurements exist in order to estimate firm performance (Perf). Former studies used different proxies. Tosi et al. (2000, p. 310) found 30 different proxies for firm performance. Firm performance measures can affect the estimated pay for performance relationship (Joskow and Rose, 1994). This section describes different proxies of firm performance used in this study.
Primarily two accounting-based measurements, return on equity (ROE) and return on assets (ROA), are considered. Many researchers have used these performance
measurements in their study. Examples are: Zhou (2000), Finkelstein and Boyd (1998), Belliveau et al. (1996) and Kato and Kubo (2006). Tosi et al. (2000, p. 330) concluded in their study that only the unadjusted relationship between CEO pay and ROE and CEO pay and ROA appear not to be affected by moderators. ROE and ROA are calculated in ExecuComp respectively by dividing net income by equity and dividing net income by total assets. ROE and ROA are given in percentages and can be found in ExecuComp as respectively ROEPER and ROA.
Next to ROE and ROA return to shareholder (TRS) is discussed. TRS is the sum of the share price gain or loss and dividends, divided by the share price at the beginning of the year. TRS is given in percentages. Murphy and Jensen (1990), Hall and Liebman (1998) and Zhou (2000) all looked for change in shareholder wealth. TRS is a good approximation for shareholder wealth. Where change in shareholder wealth is explained by Murphy and Jensen (1990, p. 5) as the inflation adjusted rate of return on common stock realized in the fiscal year multiplied by the firm value at the end of the previous year. Return to shareholder is found in ExecuComp as TRS1YR, where the period is set for 1 year.
4.2.3 Control Variables
Control variables could control for conditions that have an effect on CEO payments and cash payments and are therefore of importance in this study. This section describes the control variable and proxies used in this thesis.
As discussed earlier Tosi et al. (2000) found, through meta-analytic reviews, that firm size accounts for 40% of the variance in total CEO compensation, while firm performance accounts for less than 5% of the variance. Schaefer (1998) found evidence for the dependence of pay-performance sensitivity as defined by Murphy and Jensen (1990) on the size of the firm. Zhou (2000) confirmed existing studies that showed a positive relationship between firm size and CEO payments. Tosi et al. found 16 different measures of firm size in their meta-analytic reviews (2000, p. 307). Sales, total assets and market value are the three most used proxies for firm size. This study uses sales and market value as proxies for firm size due to the relevance in former literature and the ability to indicate the size of a firm. Examples of studies which used sales as proxy for firm size are: Zhou (2000), Belliveau (1996) and Ciscel (1974). Hall and Liebman (1998), Gibbons and Murphy (1991) and Lewellen (1968) used market value (MKTVAL) as proxy. Net sales and market value can be found in
ExecuComp respectively as SALES and MKTVAL and are expressed in millions of dollars. Firm size is the only used control variable in this study. Tosi et al. (2000) argue that there are only two widely used economic theories regarding the determinants op CEO payments. The first theory believes that CEOs are paid primarily in relation to firm size. The second theory believes that the main determinant of CEO pay is firm performance, like explained by the agency theory. They argue that other theories aren’t widely used and show very diverse results. For that reason this study limits itself to the control variable firm size. 4.3 Research Method
This subchapter gives information about the research method and regression models used to test the hypotheses formulated.
In order to examine the relation between firm performance and CEO payments for every year in the research period this thesis makes use of the ordinary least squares (OLS) method. The method of OLS is used among others by Murphy and Jensen (1990), Hall and Liebman (1998) and Zhou (2000). This study won’t deviate from their used method since it has shown to be a method which allows estimating the relation between firm performance and CEO payments. It is noticed that all data are viewed in subsequent equal periods and are thus dependent. A problem of ignoring serial correlation in regressions over multiple years is that standard errors are underestimated which could lead to incorrect conclusions about the significance of coefficients. To overcome the problem of serial correlation hypotheses are tested for every year in the research period. A disadvantage of testing hypotheses every year is the loss in statistical power. Losing statistical power decreases the probability that it will correctly lead to the rejection of a false hypothesis (Greene, 2000). However the sample size of this study is equal for every year, which could reduce the loss in power.
Many studies like Conyon and Schwalbach (2000), Hall and Liebman (1998), Zhou (2000) and Kato and Kubo (2006), based their models on the study of Murphy and Jensen in 1990. The regressions used in this study are based on the models used by Murphy and Jensen (1990) as well. Murphy and Jensen (1990) used the following models in their research to measure the total pay-performance sensitivity:
Δ(CEO wealth)t = a + b Δ(shareholder wealth)t (1)
And for the variable cash compensation:
Δ(CEO salary + bonus)t = a + b Δ(shareholder wealth)t (2)
The dependent variables Δ(CEO wealth)t and Δ(CEO salary + bonus)t are changes of the compensation of a given CEO for company i in year t compared to one year before (t-1). The independent variable Δ(shareholder wealth)t is the change in firm performance of company i in year t compared to one year before (t-1). Looking at change, the difference year to year, allows estimating the margin of change in CEO payments driven by the change in firm performance.
By adding the control variable SIZE after Schaefer (1998), Tosi et al. (2000) and Zhou (2000) showed a relation between size and CEO payments the model of Murphy and Jensen (1990) is extended to the following OLS model used in this thesis:
To test H1:
Δ(TOTALp)i,t = β0 + β1*Δ(Perf)i,t + β2*(SIZE)i,t + εi,t (3)
The dependent variable Δ(TOTALp)i,t is a change in total CEO payments for a CEO of company i in year t compared to year t-1. The independent variable Δ(Perf)i,t is a change in ROE, ROA or TRS of company i in year t compared to year t-1. If one of the proxies for firm performance changes by 1, Δ(TOTALp) is expected to change by β1*1, in thousands of dollars. The control variable SIZE is measured by sales and MKTVAL. If one of the proxies for firm size changes by 1, Δ(TOTALp) is expected to change by β2*1, in thousands of dollars. εi,t is the error term. Different proxies for firm performance and firm size are used to see if the estimations of β1 changes.
The same model is used to test H2. Only Δ(TOTALp)i,t is changed for Δ(CASHp)i,t. Where the dependent variable Δ(CASHp)i,t is a change in total cash payments for a CEO of company i in year t compared to year t-1.
To test H2:
Δ(CASHp)i,t = β0 + β1*Δ(Perf)i,t + β2*(SIZE)i,t + εi,t (4)
4.4 Robustness Checks
This subchapter describes small changes in the given specification in order to asses if those changes affect the estimations. First a log-transformed model is discussed. Thereafter the reversed pay-performance relationship is addressed. Finally the different proxies are mentioned.
Where this thesis is focused on linear terms to estimate the pay-performance relation, former researchers also looked for the pay-performance relationship by using log in their variables. Hall and Liebman (1998), Conyon and Murphy (2000) and others studied the pay-performance relationship by relative measures. Hall and Liebman (1998, p. 668) argued that the pay-performance sensitivity as used by Murphy and Jensen (1990) is more sensitive to outliers than regressions based on elasticity. The model used by former studies as Hall and Liebman (1998), Conyon and Murphy (2000) and this study is:
Δln(TOTALp)i,t = β0 + β1*Δ(Perf)i,t + β2*ln(SIZE)i,t + εi,t (3’)
The dependent variable Δln(TOTALp)i,t is the difference between the natural logarithm of total CEO payments of company i at year t and the natural logarithm of total CEO payments of company i at the previous year (t-1). The independent variable Δ(Perf)i,t is a change in ROE, ROA or TRS of company i in year t compared to year t-1. ROE, ROA and TRS are in percentages and proxies for firm performance. The performance variables aren’t log-transformed since Δln(Perf) is the same as ΔPerf/Perf which is the same as ROE, ROA or TRS. The changes in the proxies for firm performance are semi-elasticity. For example if ROE, ROA or TRS changes by 1 (1%), total CEO payments changes by β1*1%. The control variable SIZE is ln transformed and measured by the different proxies of sales and MKTVAL. If SIZE changes by 1%, the change in total CEO payments is expected to increase or decrease by β2*1%.εi,t is the error term.
Next to the log-transformed model the reversed pay-performance relationship is examined in order to see if a change in CEO payments affect firm performance. It doesn’t address the compensation structure like piece rates or hourly wage. This study is interested to see if higher CEO payments increase firm performance. The hypothesis used to test the reversed relationship is deducted from the agency theory where Jensen and Meckling (1976) argue that managers are paid on behalf of the principal. If a CEO receives higher payments,
this should be reflected in the interests for the principal. The following hypothesis is used to test the relation:
H3: S&P 500 listed total CEO’s payment is significant and positively related to firm performance since the financial crisis of 2008.
The control variable firm size is still valid since Hall and Weiss (1967) found evidence that large firms perform better. They based their research on the theory of Baumol (1959) that higher money capital increases the total profits of a firm and finally increases earnings per dollar investment. The next model is used to test H3:
Δ(Perf)i,t = β0 + β1*Δ(TOTALp)i,t + β2*(SIZE)i,t + εi,t (5)
The same interpretation, for the variables, is used as in regressions (3) and (4), only Δ(TOTALp) has switched with Δ(Perf). The different proxies described in the earlier regressions are used for this model as well.
Besides, this study uses different proxies, described earlier, for performance variables and size variables. All regressions are run on the different proxies to see if estimates change.
5 Results
This chapter discusses and analyzes the results of the regressions run. In the first subchapter means and standard deviations of the most important variables are given. Next the results and analyses from the interested regressions to test the first hypothesis are discussed. Thereafter results and analyses are given for the second tested hypothesis. Finally the results of the robustness checks are given and analyzed. It is important to know that p-values for the independent variables and control variables are one-tailed values. One tailed values are used since hypotheses and theories predict a positive relation between performance, size and CEO payments. For the sake of clarity the tables 4-7 can be found below in the thesis. All the tables have added notes.
5.1 Descriptive Statistics
This subchapter gives means and standard deviations (SD) of the most important variables used in this thesis for the years in the research period, 2008-2012, plus the year 2007 since some variables are interested in change. For every year and variable 442 observations are 16
available. It totals 2,652 observations for the complete period 2007-2012.
Table 1 gives the mean and SD in thousands of dollars of the pay variables salary, bonus, CASHp and TOTALp. The last part of the table shows the mean and SD for the period 2007-2012. Remarkable is the high SD in the year 2008 for bonus CASHp and TOTALp. That year the CEOs of the companies Nabors Industries and Chesapeake received a bonus of over 70 million dollars. It is noticed that the mean of TOTALp dropped in 2009, one year after the start of the financial crisis, from 10.14889 million dollar in 2008 to 8.557641 million dollar in 2009 while the predetermined salary slightly increased. Bonus and salary are only a small part of a CEO’s total compensation. For the period 2007-2012 bonus and salary (CASHp) accounted for 15.45% (CASHp/TOTALp) of TOTALp. This is in line with the argument of Hall and Liebman (1998, p. 656) that one of the most direct solutions to the agency problem is to: “align the incentives of executives with the interests of shareholders by granting (or selling) stock and stock options to the CEOs”. Summarized, the mean of salary, bonus, CASHp and TOTALp of all observations over the years 2007-2012 are respectively: 1.081859, 0.486889, 1.568747 and 10.15089 million dollars. Figure 1 gives a graphical impression for the means of TOTALp and CASHp.
Figure 1:
Table 1: Pay Variables1; Mean and Standard Deviation
Variables Mean Standard Deviation
2007, 442 obs
Salary 1014.101 536.516
Bonus 626.570 3178.112
CASHp (Bonus + Salary) 1640.672 3277.67
TOTALp 10026.11 9674.988
2008, 442 obs
Salary 1051.445 513.955
Bonus 630.869 5061.997
CASHp (Bonus + Salary) 1682.314 5093.275
TOTALp 10148.89 11541.26
2009, 442 obs
Salary 1069.758 571.886
Bonus 348.482 1479.495
CASHp (Bonus + Salary) 1418.24 1636.167
TOTALp 8557.641 6611.546
2010, 442 obs
Salary 1110.914 563.999
Bonus 425.972 1707.981
CASHp (Bonus + Salary) 1536.886 1883.155
TOTALp 10360.79 8103.959
2011, 442 obs
Salary 1124.876 544.925
Bonus 422.142 1862.693
CASHp (Bonus + Salary) 1547.018 2137.169
TOTALp 11077.86 10389.76
2012, 442 obs
Salary 1120.06 554.504
Bonus 467.295 2269.316
CASHp (Bonus + Salary) 1587.355 2474.995
TOTALp 10734.03 7315.909
2007-2012, 2,652 obs
Salary 1081.859 548.937
Bonus 486.889 2870.081
CASHp (Bonus + Salary) 1568.747 2987.264
TOTALp 10150.89 9133.478
Notes: 1 Observations are in thousands of dollars and rounded on 3 decimals.
Table 2 gives the mean and SD in percentages of the firm performance variables ROE, ROA and TRS. The last part of the table shows the mean and SD for the period 2007-2012. Remarkable is the negative mean of TRS in the year 2008. TRS dropped from about 17 in 2007 to about minus 32 in 2008, while TOTALp slightly increased from 2007 to 2008. TRS fluctuates over the years while ROE and ROA stay relatively stable. The SD of ROE in the year 2008 is more than eight times the mean of ROE. This is caused by some very large negative ROEs and some large positive ROEs. Summarized, the mean of ROE, ROA and TRS of all
observations over the years 2007-2012 are respectively: 13.725, 5.705, and 11.399 per cent. Figure 2 gives a graphical impression for the means of ROE, ROA and TRS.
Table 2: Performance Variables1; Mean and Standard Deviation
Variables Mean Standard Deviation
2007, 442 obs ROE 16.574 19.349 ROA 6.723 7.804 TRS 16.743 55.652 2008, 442 obs ROE 8.032 66.673 ROA 4.593 11.338 TRS -32.413 24.486 2009, 442 obs ROE 10.745 22.138 ROA 4.525 7.677 TRS 37.803 52.396 2010, 442 obs ROE 14.934 16.468 ROA 6.255 6.072 TRS 23.818 26.969 2011, 442 obs ROE 16.074 21.010 ROA 6.330 5.756 TRS 4.302 25.913 2012, 442 obs ROE 15.991 30.002 ROA 5.807 6.116 TRS 18.138 26.187 2007-2012, 2652 obs ROE 13.725 34.086 ROA 5.705 7.741 TRS 11.399 43.602
Notes: 1 ROE, ROA and TRS are given in percentages and rounded on 3 decimals.
Table 3 gives the mean and SD in millions of dollars of the firm size variables sales and market value of firm (MKTVAL). Again the last part of the table shows the mean and SD for the period 2007-2012. The numbers are relatively stable with the exception of the decline in MKTVAL from the year 2007 to 2008. Summarized, the mean of sales and MKTVAL of all observations over the years 2007-2012 are respectively: 18.41855 and 22.95586 billion dollars. Figure 3 gives a graphical impression for the means of sales and MKTVAL.
Figure 2:
Table 3: Control Variables1; Mean and Standard Deviation
Variables Mean Standard Deviation
2007, 442 obs Sales 17349.6 34155.69 MKTVAL 26641.57 47002.02 2008, 442 obs Sales 18204.09 37771.31 MKTVAL 17593.45 34769.55 2009, 442 obs Sales 16652.41 32425.65 MKTVAL 20545.07 34425.53 2010, 442 obs Sales 18192.39 35283.13 MKTVAL 23479.33 37634.93 2011, 442 obs Sales 19903.04 39862.04 MKTVAL 23463.19 38769.26 2012, 442 obs Sales 20209.78 39429.11 MKTVAL 26012.55 41504.55 2007-2012, 2652 obs Sales 18418.55 36578.37 MKTVAL 22955.86 39340.21
Notes: 1 Observations are in millions of dollars and not rounded.
Figure 3:
5.2 Hypothesis 1
This subchapter discusses and analyzes the results obtained out of the regressions in order to test H1. Table 4 shows the results of the regression (3) Δ(TOTALp)i,t = β0 + β1*Δ(Perf)i,t + β2*(SIZE)i,t + εi,t. For every year there are 442 observations. Only significant and remarkable results are discussed. H1 is tested for every year in the research period with the dependent variable ΔTOTALp. Different proxies, as described earlier, are used for firm performance and firm size. In total 6 different regressions are run for every year indicated in table 4 under {1}-{6}.
The year 2008 doesn’t show any significant results for the proxies of firm performance and firm size. The observed R2 is between 0.001 and 0.012, which indicates that the fraction of the variance of the dependent variable that is explained by the independent variable and control variable is low.
The year 2009 show that ROA is positively related to total CEO payments for a significance level of 1% for both the control variables sales and MKTVAL. This result means that if ROA changes positively by 1%, total CEO payments increase by about {3} $99,000 and {4} $96,000. In the regressions {1}-{4} the constant reports a negative amount for a
significance level of 1% or 5%. The negative amount means that when firm performance and firm size remains unchanged total CEO payments decrease by about, for example in {4}, $1.3 million or 15% relative to the mean for total CEO payments. Murphy and Jensen (1990, p. 12) reported constants of about $0.8 million. These constants were positive and measured between 1969 and 1983. It should be taken into consideration that the value of money changes over time. R2 is between 0.001 and 0.015. The highest R2 is measured in regression {4} where total CEO payments is regressed on ROA and MKTVAL.
In 2010 constants of about $1.5 and $1.8 million are measured for a significance level of 1%. The constants are positive and mean that keeping firm performance and firm size constant, total CEO payments increases by the observed amount. Regression {2} shows that MKTVAL is positively related to total CEO payments for a 5% significance level. It means that for every million increase in market value, total CEO payments increase by 0.009 units of total CEO payments or $9.00. R2 is between 0.001 and 0.012.
The year 2011 doesn’t show any significant results for a 5% or 1% significance level. R2 gives values between 0.000 and 0.001.
In 2012 all firm performance measures are positively related to total CEO payments for a significance level of 5% (ROE and TRS) or 1% (ROA). A 1% change in the performance measure ROE, ROA or TRS increases total CEO payments for about, $32,000, $189,000 or $22,000 respectively. The table reports R2 values between 0.007 and 0.016. The highest R2 is measured in regression {4}, like in 2009 where total CEO payments is regressed on ROA and MKTVAL.
Only 2009 and 2012 give significant (1%) positive relationships between firm performance variables and total CEO payments. It doesn’t allow this study to estimate if there is a difference in strength in β1 over the years. None of the control variables are significant positively related for 1%.
5.3 Hypothesis 2
This subchapter discusses and analyzes the results obtained out of the regressions in order to test H2. Table 5 shows the results of the regression (4) Δ(CASHp)i,t = β0 + β1*Δ(Perf)i,t + β2*(SIZE)i,t + εi,t. For every year there are 442 observations. Only significant and remarkable results are discussed. H2 is tested for every year in the research period with the dependent variable ΔCASHp. Different proxies, as described earlier, are used for firm performance and
firm size. In total 6 different regressions are run for every year indicated in table 5 under {1}-{6}.
In 2008 no significant results are observed. Standard errors are in most cases higher than the absolute value of the coefficients. R2 is between 0.003 and 0.005.
In the year 2009, ROA is positively related to cash payments for a significance level of 1% for both the control variables sales and MKTVAL. This result means that if ROA changes positively by 1%, cash payments increase by about $51,000 dollars. In the regression with ΔTOTALp as dependent variable in 2009, ROA was significant positively related to ROA with changes in ΔTOTALp for every unit change in ROA of between $96,000 and $99,000. This indicates that the relation between ROA and total CEO payments is driven for an important part by salaries and bonuses. R2 is between 0.001 and 0.016. The highest two R2 are measured in regressions {3} and {4} where total cash payments are regressed on ROA and sales or ROA and MKTVAL.
The year 2010 shows no significant results for the constant or the firm performance variables. Sales is significant positively related to cash payments, with a coefficient of 0.002, for a significance level of 10% in {1}, {3} and {5}. It means that for every million increase in sales, cash payments change positively by $2.00. MKTVAL is positively related to cash payments for a significance level of 5% in the regressions {2}, {4} and {6}. Coefficients are respectively 0.003, 0.003 and 0.002. This means for example in regression {2} that for every million increase in market value cash payments change positively by $3.00. R2 is between 0.005 and 0.010.
In 2011 ROA is positively related to total cash payments with coefficients of 15.027 {3} and 14.920 {4} for a 5% and 10% significance level respectively. TRS is positively related to cash payments for a significance level of 5% with a coefficient of 1.832 in {5} and 1.896 in {6}. The control variables sales and MKTVAL show in all regression a positively relation, between 0.001 and 0.002, for a significance level of 10% or lower. R2 is between 0.008 and 0.018.
The year 2012 show that ROE and ROA are positively related to CEO cash payments. ROE is positively related to cash payments with a coefficient of 6.502 in {1} and 6.550 in {2} for a significance level of 5%. ROA is positively related to cash payments with a coefficient of 43.205 in {3} and 43.418 in {2} for a significance level of 1%. R2 is between 0.004 and 0.027. The highest two R2, like in 2009, are measured in regressions {3} and {4} where total cash 23
payments are regressed on ROA and sales or ROA and MKTVAL.
Again only 2009 and 2012 give significant positive relationships between firm performance variables and total CEO payments for a significance level of 1%. Regressions in which ROA was included and significant related showed generally the highest R2. None of the control variables are positively related for a significance level of 1%.
5.4 Robustness Checks
In this subchapter the results of the robustness checks are discussed and analyzed. First the results for the log-transformed model, in order to test H1, are presented. Thereafter the results for the reversed pay-performance relationship model, to test H3, are given. Different proxies, as described earlier, are used for firm performance and firm size. In total 6 different regressions are run for every year, indicated under {1}-{6}, in table 6 (log-transformed model) and table 7 (reversed relationship model). For every year there are 442 observations. Only significant and remarkable results will be discussed.
The robustness checks with different proxies are mentioned within the subchapters or sections for every used model in this thesis so not mentioned apart.
5.4.1 Log-Transformed Model
This section discusses and analyzes the results of regression (3’) Δln(TOTALp)i,t = β0 + β1*Δ(Perf)i,t + β2*ln(SIZE)i,t + εi,t in order to test H1. The results can be found in table 6. The dependent variable Δln(TOTALp)i,t is the log-transformed dependent variable already used to test H1.
In 2008 ROE is positively related to total CEO payments for a significance level of 1% for both the control variables lnsales and lnMKTVAL. The coefficients of ROE in {1} and {2} are 0.001. It means that if ROE increases by for example 10%, total CEO payments increases by 0.01%. ROA has a coefficient of 0.004 and is positively related to total CEO payments for a significance level of 10% with the control variable lnsales. In contrast to regression (3) with ΔTOTALp as dependent variable, ROE now show significant results. R2 is between 0.002 and 0.022, which is higher than in regression (3) with Δ(TOTALp) as dependent variable.
ROE and ROA are positively related to total CEO payments, for a significance level of 1% in the year 2009. Like in 2008 ROE has a coefficient of 0.001 in {1} and {2}. ROA amounts to 0.011. R2 is between 0.002 and 0.040 and is higher than in regression (3).
In the year 2010, ROE is positively related to total CEO payments for a significance
level of 5%. ROE has coefficients of 0.002 {1} and 0.003 {2}. ROA is positively significant (10%) related to total CEO payments with coefficients of 0.06 in {3} and {4}. R2 is between 0.004 and 0.012.
In 2011 results don’t show any significance for performance variables and control variables. This is the same for 2011 in regression (3) with the dependent variable ΔTOTALp. R2 is between 0.000 and 0.008.
The year 2012 show positively significant results for all performance variables. ROE has coefficients of 0.002 in {1} and {2} for a significance level of 5%. ROA has coefficients of 0.014 in {3} and {4} for a significance level of 1%. TRS has coefficients of 0.002 in {5} and {6} for a significance level of 1%. R2 is between 0.006 and 0.020.
In all years none of the control variables show significant results. Significant intercepts are rare and only for 10%. The years 2008, 2009, and 2012 show significant (1%) positive relationships between firm performance and total CEO payments. In regression (3) with ΔTOTALp as dependent variable only the years 2009 and 2012 showed (1%) significant results. Regression (3) had a R2 of between 0.000 and 0.016. The R2 in this regression (3’) is between 0.000 and 0.040. A possible explanation of the higher R2 measured is that natural logarithms reduce the effect of outliers. Results of Hall and Liebman (1998, p. 686) show relationships that are much stronger than the significant relationships found in this research. As described earlier they found an elasticity of between 1.2 in 1980 and 3.9 in 1994. It should be taken into consideration that variables and the precise model differ from Hall and Liebman (1998).
5.4.2 Reversed Relationship
This section discusses and analyzes the results of regression (5) Δ(Perf)i,t = β0 + β1*Δ(TOTALp)i,t + β2*(SIZE)i,t + εi,t in order to test H3. The results can be found in table 7.
The year 2008 doesn’t show any significant (5% or less) results for the pay or size variables. The constant however is significant (1%) for every regression. The negative amount, for example in {4} means that when total CEO payments and firm size remains unchanged ROA decreases by about 2.26%. This is almost half of ROA’s mean. It could indicate that much of the dependent variable is explained by explanatory variables not in the regression. The R2 is between 0.001 and 0.009.
In 2009 total CEO payments is positively significant (1%) related to ROA for both
control variables. The coefficient in {3} of 0.000133 means that if pay increased by 1000 units ($1 million) ROA changes by 0.133%. The R2 is between 0.000 and 0.027.
In 2010 constants are again significant (1%) for every regression. In {5} sales is positively significant (5%) related to TRS. The coefficient of sales amounts to 0.00013, which means that if sales increases by 1000 units ($1 billion) TRS changes by 0.13%. The R2 is between 0.01 and 0.013.
The year 2011 show significant (1%) constants for all regressions. The control variables sales (5%) and MKTVAL (1%) are positively related to ΔTRS in regressions {5} and {6}. The coefficients of sales and MKTVAL are respectively 0.000083 and 0.0001306. The R2 is between 0.000 and 0.018.
In 2012 total CEO pay is positively related (5%) to ΔROE in {1} and {2}; positively related (1%) to ΔROA in {3} and {4} and positively related (5%) to ΔTRS in {5} and {6}. The coefficients of ΔTOTALp are between 0.0000758 and 0.0005092. If pay increases by $1 million this indicates a change in the dependent variables of between 0.0758% and 0.5092%. The observed R2 is between 0.007 and 0.015.
Only 2009 and 2012 give significant positive relationships between total CEO payments and firm performance for a significance level of 1%. Only when ΔROA is used as the dependent variable, ΔTOTALp is positively related for a significance level of 1%. The lowest observed R2 is 0.000 and the highest 0.027.
6 Conclusion and summary
Primarily a short summary of the study is presented. Thereafter conclusions are given. Finally some implications of the study are discussed and suggestions for further research are described.
This study examines the relationship between performance of the S&P 500 listed firms and CEO payments since the financial crisis of 2008. In total 442 companies and their payments to CEOs have been examined by using OLS regressions for four different models. The agency theory and former literature assumes a relation between firm performance and CEO payments. Therefore the two hypotheses tested are formulated as: S&P 500 listed firms’ performance is significant positively related to total CEO payments and S&P 500 listed firms’ performance is significant positively related to CEO cash payments since the financial crisis of 2008. These hypotheses are tested for every year in the research period using two 26
different models. Several different proxies for performance and control variables have been used to see if estimates differ. Besides, a robustness check with a nonlinear model, by using natural logarithms, is performed and the reversed pay-performance relationship is examined.
For the regression used to test H1, 2009 and 2012 show for a 1% significance level, positive relations between firm performance and total CEO payments. Only the firm performance proxy ROA shows a positive 1% significant relation in both years. A 1% change in ROA changes total CEO payments between $96,417 and $98,507 in 2009 and between $188,527 and $189,111 in 2012. This indicates a strong positive relationship for 2009 and 2012. For the 1% significant results R2 is between 0.014 and 0.016.
The results of the regression to test H2 show that again only 2009 and 2012 give 1% significant results. ROA was 1% significant in both 2009 and 2012. For a 1% change in ROA cash payments changed between $51,243 and $51,306 in 2009 and between $43,205 and $43,418 in 2012. This indicates a strong positive relationship for 2009 and 2012. In comparison to the results of 2009 and 2012 with ΔTOTALp as dependent variable it is shown that the relation of ROA to total CEO payments is driven for an important part by predetermined salaries and bonuses. For the 1% significant results R2 is between 0.016 and 0.027.
H1 has been tested for a second time with a log-transformed model. It gives 1% positive significance results for 2008 (ROE), 2009 (ROE and ROA), 2012 (ROA and TRS). ROE has coefficients of 0.001 in 2008 and 2009. ROA amounts to 0.011 in 2009 and to 0.014 in 2012. TRS has coefficients of 0.002 in 2012. A coefficient of 0.001 means that for every 10% change in one of the performance variables, total CEO payments change by 0.01%. R2 has values for the 1% significant results between 0.014 and 0.040. The founded relation by this model is weaker than Hall and Liebman (1998, p. 686) found in their research. But it should be taken in consideration that the used performance variables differ.
The reversed pay-performance model showed positive 1% significant results for 2009 and 2012, but only for the dependent variable ΔROA.
For all models the years 2009 and 2012 show positively 1% significant results when ROA is used as independent variable. It can be concluded that in 2009 and 2012 total CEO payments and CEO cash payments are positively and significant (1%) related to ROA since the financial crisis of 2008. It is impossible to draw conclusions why other years didn’t show 27
1% positive significant results. However possible explanations for the insignificance of results are: managers have great influence over their pay; the financial crisis misaligned the incentives or CEO payments are dependent on many other variables not examined in this study. The most important conclusion that can be drawn is that the relation and the significance of the relationship are highly dependent on the choice of firm performance measures and the choice of the regression model. This is in line with the argumentation of Joskow and Rose (1994) that performance measures can affect the estimated pay for
performance relationship.
In the nonlinear model R2 values increased. This is probably due to log transforming variables and limiting the effect of large outliers. However values of the R2 are small for all models. The low R2 values mean that the fraction of variance in the dependent variable explained by the regressors is low. A possible explanation for low R2 values is that only objective measures for firm performance proxies are used in the regressions. By excluding subjective measures R2 values could stay low. However a low R2 doesn’t necessarily mean that the models are useless. If results of the performance variables are significant it still allows this study to draw conclusions.
To examine multiple years and to deal with time series it should be interesting to see the results of a panel regression. The panel regression could be obtained by adding firm specific variables like: CEO tenure or age; firm leverage and firm industry. This thesis is already using a balanced data set, since the amount of observations are exactly the same for every year. For further research it is advised to include more control variables to reduce the possibility of omitted variable bias. Variables included in the regression might be affected by excluded variables which could cause the estimators to be biased. Biased estimates are misleading and could result in wrong conclusions.
Since CEO payments are constructed out of different kind of payments like: stock, options and more. regressions could be run on all kind of the different pay variables to see which pay variables are more related to firm performance. Next to this it would be good to further investigate what kind of measures firms exactly use to base their pay for CEOs. It is hard to draw clear conclusions when using all kind of performance proxies. But by choosing one proxy, the possibility of making wrong conclusions increase. As discussed earlier, including subjective proxies could possible increase the explanation power of the models. Finally it would be interesting to see if further research can connect CEO payments to not 28
only incentives aligned contracts as explained by the agency theory but with a more society focused perspective.
References
Aggarwal, R., & Samwick, A. A. (1998). The other side of the tradeoff: The impact of risk on executive
compensation (No. w6634). National Bureau of Economic Research.
Baumol, W. J., Business Behavior, Value, and Growth. New York: MacMillan, 1959.
Belliveau, M. A., O'Reilly, C. A., & Wade, J. B. (1996). Social capital at the top: Effects of social similarity and status on CEO compensation. Academy of Management Journal, 39(6), 1568-1593.
Benabou, R., & Tirole, J. (2003). Intrinsic and extrinsic motivation. The Review of Economic
Studies, 70(3), 489-520.
Berle, A. A., & Means, G. G. C. (1932). The modern corporation and private property. Transaction Books.
Brick, I. E., Palmon, O., & Wald, J. K. (2006). CEO compensation, director compensation, and firm performance: evidence of cronyism? Journal of Corporate Finance, 12(3), 403-423.
Cichello, M. S. (2005). The impact of firm size on pay–performance sensitivities. Journal of Corporate
Finance, 11(4), 609-627.
Ciscel, D. H. (1974). Determinants of executive compensation. Southern Economic Journal, 613-617. Conyon, M. J., & Murphy, K. J. (2000). The prince and the pauper? CEO pay in the United States and
United Kingdom. The Economic Journal, 110(467), 640-671.
Conyon, M. J., & Schwalbach, J. (2000). Executive compensation: evidence from the UK and Germany. Long Range Planning, 33(4), 504-526.
Core, J. E., Holthausen, R. W., & Larcker, D. F. (1999). Corporate governance, chief executive officer compensation, and firm performance. Journal of financial economics, 51(3), 371-406. Duffhues, P., & Kabir, R. (2008). Is the pay–performance relationship always positive?: Evidence from
the Netherlands. Journal of multinational financial management, 18(1), 45-60. Finkelstein, S., & Boyd, B. K. (1998). How much does the CEO matter? The role of managerial
discretion in the setting of CEO compensation. Academy of Management Journal, 41(2), 179-199.
Gibbons, R., & Murphy, K. J. (1991). Relative performance evaluation for chief executive officers (No. w2944). National Bureau of Economic Research.
Grazell, J. (2006). Institutions, corporate governance and firm performance. Advances in Corporate
Finance And Asset Pricing, 293.
Greene, W. H. (2000). Econometric analysis 4th edition.
Hall, B. J., & Liebman, J. B. (1998). Are CEOs really paid like bureaucrats? The Quarterly Journal of
Economics, 113(3), 653-691.
Hall, M., & Weiss, L. (1967). Firm size and profitability. The Review of Economics and Statistics, 49(3), 319-331.
Hendrikse, G. (2003). Economics and Management of Organizations. New York: The McGraw-Hill Companies, ISBN 9780077099923.
iStock Research (2013). Intrinsic Value of Lockheed Martin Corporation (LMT). (http://www.istockresearch.com/value/LMT), 20 December 2013.
Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of financial economics, 3(4), 305-360.
Joskow, P. L., & Rose, N. L. (1994). CEO pay and firm performance: dynamics, asymmetries, and
alternative performance measures (No. w4976). National Bureau of Economic Research.
Kato, T., & Kubo, K. (2006). CEO compensation and firm performance in Japan: Evidence from new panel data on individual CEO pay. Journal of the Japanese and International
Economies, 20(1), 1-19.
Lewellen, W. G. (1968). Introduction to" Executive Compensation in Large Industrial Corporations". In Executive Compensation in Large Industrial Corporations (pp. 1-9). NBER.
Maskin, E., & Tirole, J. (1992). The principal-agent relationship with an informed principal, II: Common values. Econometrica: Journal of the Econometric Society, 1-42.
Murphy, K., & Jensen, M. (1990). Performance pay and top management incentives. Available at SSRN: http://ssrn.com/abstract=94009 or http://dx.doi.org/10.2139/ssrn.94009 Nordberg, D. (2011). Corporate governance: principles and issues. Sage Publications.
Schaefer, S. (1998). The Dependence of pay—Performance Sensitivity on the Size of the Firm. Review
of Economics and Statistics, 80(3), 436-443.
Standard & Poor’s (2013). S&P 500.
(http://www.spindices.com/indices/equity/sp-500), 12 November 2013. The Economist (2013). The origins of the financial crisis.
(http://www.economist.com/news/schoolsbrief/21584534-effects-financial-crisis-are-still-being-felt-five-years-article), 13 Januari 2014.
Tosi, H. L., Werner, S., Katz, J. P., & Gomez-Mejia, L. R. (2000). How much does performance matter? A meta-analysis of CEO pay studies. Journal of Management, 26(2), 301-339.
Wharton Research Data Services (2013a). WRDS.
(https://wrds-web.wharton.upenn.edu/wrds/), 12 December 2013. Wharton Research Data Services (2013b). WRDS.
(https://wrds-web.wharton.upenn.edu/wrds/ds/documentation/comp/legacy/tdc1.cfm), 12 December 2013.
Zhou, X. (2000). CEO pay, firm size, and corporate performance: evidence from Canada. Canadian
Journal of Economics/Revue canadienne d'économique,33(1), 213-251.
Table 4: Hypothesis 1; Dependent Variable ΔTOTALp
For notes see next page.
N: 442 {1} {2} {3} {4} {5} {6} B B B B B B 2008 Constant 564.618 (461.305) 322.231 (468.396) 617.624 (467.223)* 382.828 (475.723) 367.846 (563.142) 106.808 (571.805) ΔROE 6.932 (6.380) 6.874 (6.413) ΔROA 52.846 (43.051) 53.320 (43.317) ΔTRS -2.635 (6.539) -2.818 (6.569) Sales -.021 (.011) -.021 (.011) -.021 (.011) MKTVAL -.008 (.012) -.008 (.012) -.007 (.012) R2 0.011 0.003 0.012 0.004 0.009 0.001 2009 Constant -1699.074 (500.876)*** -1313.492 (518.630)** -1700.698 (497.105)*** -1333.737 (514.694)*** -985.734 (700.615) -478.170 (718.768) ΔROE .495 (6.768) .188 (6.765) ΔROA 98.507 (40.808)*** 96.417 (40.811)*** ΔTRS -9.513 (6.570) -10.966 (6.571) Sales .006 (.014) .007 (.014) .004 (.014) MKTVAL -.014 (.013) -.012 (.013) -.017 (.013) R2 0.001 0.003 0.014 0.015 0.005 0.009 2010 Constant 1762.24 (270.529)*** 1557.017 (283.437)*** 1739.618 (274.714)*** 1532.284 (287.307)*** 1670.561 (273.948)*** 1470.95 (284.919)*** ΔROE 6.590 (12.108) 7.198 (12.085) ΔROA 27.273 (37.990) 30.005 (37.929) ΔTRS -7.266 (4.221) -7.523 (4.202) Sales .001 (.007) .001 (.007) .002 (.007) MKTVAL .009 (.006)** .009 (.006)* .010 (.006)* R2 0.001 0.006 0.001 0.006 0.007 0.012 32
Table 4 continued: Hypothesis 1; Dependent Variable ΔTOTALp
Notes: OLS regressions are run in Stata version 12 for every year apart in the research period. N=442 for every year. Standard errors are in parentheses. *,** and *** indicates significant levels of respectively 10%, 5% and 1%. P-values for the independent variables and control variables are one-tailed values. All results are rounded on 3 decimals. All deltas are defined as the observed variable value of the research year minus the observed variable value of the previous year. ΔTOTALp is the year to year change of salary, bonus, other annual, total value of restricted stock granted, total value of stock options granted (using Black-Scholes), long-term incentive payouts (LTIP), and all other total. Sales and MKTVAL are observed in millions, pay variables in thousands and performance variables in percentages.
N: 442 {1} {2} {3} {4} {5} {6} B B B B B B 2011 Constant 725.345 (428.316)* 584.188 (447.624) 720.665 (427.806)* 579.543 (447.215) 805.857 (478.458)* 653.114 (502.832) ΔROE -7.764 (22.410) -8.081 (22.403) ΔROA -33.005 (86.210) -32.904 (86.173) ΔTRS 4.164 (10.129) 3.373 (10.178) Sales .000 (.010) -.000 (.010) -.000 (.010) MKTVAL .006 (.010) .006 (.010) .006 (.010) R2 0.000 0.001 0.000 0.001 0.000 0.001 2012 Constant -406.202 (423.137) -113.852 (444.020) -311.482 (423.097) -16.933 (444.022) -742.426 (447.928)* -445.728 (468.084) ΔROE 32.278 (18.413)** 32.054 (18.394)** ΔROA 189.111 (74.791)*** 188.527 (74.712)*** ΔTRS 22.733 (10.026)** 22.036 (10.016)** Sales .003 (.010) .003 (.010) .004 (.010) MKTVAL -.009 (.009) -.009 (.009) -.008 (.009) R2 0.007 0.009 0.015 0.016 0.012 0.013 33
