The relationship between equity-based compensation and risk-taking behavior

(1)

The relationship between equity-based

compensation and risk-taking behavior

Abstract

This thesis aims to test the relationship between equity-based compensation and firm risk. A sample of 384 firms is used to test this relationship. Data is collected from WRDS for the period 2009-2013. A positive relationship between these variables is expected. However, in this thesis a negative significant relation is found between equity-based compensation and risk-taking. Equity-equity-based compensation can be divided into two components: stock and option compensation. This

research found a negative significant relationship between stock compensation and risk-taking. Besides that, this research found a positive relationship between option compensation and risk-taking

behavior.

Student Jacqueline van Velzen

Student number 10410872

Supervisor Dr. S. Dominguez Martinez

Date February 2015

Study BSc Finance and Organization

(2)

2 Statement of Originality

This document is written by Jacqueline van Velzen who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Table of content Page

1. Introduction 4

2. Literature review 6

2.1. Principal agent theory 6

2.2. Risk behavior shareholders and CEO 7

2.3. Compensation package 7

2.4. Equity-based compensation 8

2.4.1. (Restricted) Stock compensation 9

2.4.2. Option compensation 10

2.5. Other research 11

3. Hypothesis 13

4. Research Methodology 14

4.1. Sample, period and database 14

4.2. Method 14

4.3. Variable description 17

4.3.1. Dependent variables 17

4.3.2. Independent variable 17

4.3.3. Control variables 18

5. Results and Discussion 19

5.1. Descriptive statistics 19 5.2. Regression results 21 5.3. Limitations 25 6. Conclusion 27 References 28 Appendix 29

(4)

4 1. Introduction

The conflict of interest is a big issue in today’s large corporations. There are two sorts of conflict of interests between shareholders and managers. First of all, the objectives differ. The main objective of shareholders is to increase firm value. But the main objective of CEOs is to maximize his own utility. Second, they have different risk preferences. Shareholders are risk-neutral because they can diversify their investment portfolio, but managers are risk averse. Moreover, managers cannot diversify their portfolio, which can stimulate them to take risk-reducing actions (Mehran 1995). This research focuses on the second conflict of interest.

Managers are risk averse and are likely to minimize firm risk. As a result, managers reduce their risk-taking behavior. By taking actions with lower cash flow volatility, managers can decrease the firm risk. On the other side, shareholders want to maximize firm value. This can only be maximized when a manager takes risky decisions. Therefore it is essential to provide incentives to increase managers’ risk taking behavior. A lot of literature has argued that equity-based compensation is a mechanism that helps managers to overcome their risk aversion and induce optimal risk-taking behavior (Jensen and Murphy, 1990; Low, 2009; Mehran, 1995). With equity-based compensation the managers’ claim on firm outcomes increase, which makes them more willing to accept risky, positive NPV projects.

Although, executive compensation has always been a point of discussion. Amihud and Lev (1981); Canella and Gray (1997) conclude that equity-based compensation has a negative effect on manager’s risk behavior because they want to compensate for the risk associated with their firm-specific human capital. Hence, it is still questionable whether equity-based compensation has a positive effect on a manager’s risk-taking behavior. Therefore, the main objective of this thesis will be to investigate whether there exists a positive significant relationship between equity-based compensation and the managers’ risk-taking behavior. Hence, the main research question in this thesis is:

“Does equity-based compensation increase risk-taking behaviour of managers?” In this thesis, two proxies (Beta and R&D) are used for risk-taking behavior. I will focus on the S&P 500 for the period 2009 through 2013. Coles, Daniel, and Naveen (2006) found that the relation between vega and firm risk runs both ways. Vega is the sensitivity of CEO wealth to stock volatility, which increases with equity-based compensation. Therefore the

(5)

5 I want to expand the existing literature by examining the possible relationship between

equity-based compensation and risk-taking behavior in the most recent years. Most of the previous studies are based on old data, but the financial crisis could have changed the relation between risk-taking behavior and the level of equity-based compensation of the manager. Through the financial crisis, managers could have become more careful with risky decisions.

This thesis is organized as follows. The second chapter will discuss the principal-agent theory and will give an overview of previous studies. The third chapter will introduce the hypotheses. The research method will be explained and discussed in chapter four. The fifth chapter will analyze and discuss the results of this thesis. Finally, the last chapter will give a conclusion.

(6)

6 2. Literature review

In this chapter the principal-agent theory will be discussed. The first paragraph will describe the general principal-agent theory and the conflict of interest between CEOs and shareholders. After that, the risk behavior of CEOs and shareholders will be compared. Paragraph three will explain the importance of a compensation package. Paragraph four will discuss (restricted) stock and option compensation. And finally, in paragraph five other previous research will be discussed.

2.1 Principal-agent theory

In many relations between parties, there exists an agency relationship. Jensen and Meckling (1976) define the agency relationship as a contract under which one party (the principal) engage another party (the agent) to perform some service on their behalf, which involves delegating some decision-making authority to the agent.

When the principal and the agent are both utility-maximizers, there is a high probability of a conflict of interest between the principal and agent (Jensen and Meckling, 1976). It is

important for the principal to try to incentivize the agent to act in the best interests of the principal. For example, the agency relationship between an employer and a worker. The employer (principal) wants to pay as little as possible, though the worker (agent) wants to earn as much as possible. These two preferences are in conflict, but the principal can solve this problem for example by using compensation contracts based on piece-rate or profit sharing.

Another classic example of the principal-agent problem is the conflict of interest between shareholders of a publicly owned corporation and the corporation's CEO (Jensen and Murphy, 1990). Within these firms, shareholders are the owners (principals) of the firm and hire a CEO (agent) to act on their behalf. There are two sources of a conflict of interest in this

relationship. First, the desired actions of the shareholders are not always in line with the actions of the CEO. Shareholders want CEOs to take particular actions whenever it has a positive NPV, so the expected return on the action exceeds the expected costs. But the CEO only compares his private benefits and cost from taking a particular activity (Jensen and Murphy, 1990). Second, shareholders and managers have different risk preferences. Risk-averse managers prefer to undertake less risky projects. On the contrary, shareholders prefer

(7)

7 all positive NPV projects regardless of the associated risk (Low, 2009). Positive NPV projects are important to increase firm value. Risk behavior will be further discussed in paragraph 2.2.

Due to the misalignment of interests, the CEO will not always act in the best interests of the shareholders. Costs associated with the misalignment of interests are called agency costs. With monitoring or the right compensation package, shareholders try to limit these costs. A well-designed compensation policy ties the CEO's welfare to shareholder wealth and provides incentives for CEOs to take more value-maximizing decisions (Jensen and Murphy 1990).

2.2 Risk behavior shareholders and CEO

There exists a problem of risk sharing when the principal and agent have different risk preferences (Eisenhardt, 1989). Shareholders would like CEOs to undertake all projects that are firm value increasing, regardless of their risk (Rajgopal and Shevlin, 2002). In addition, shareholders can diversify their investment portfolio, so they do not mind risk associated with positive NPV projects. By diversifying their portfolio, they can create their desired level of risk (Mehran, 1995).

CEOs, like most individuals, are considered risk averse. Because they hold significant stakes in their own firms, they are undiversified and more exposed to firm risk. Therefore, executives prefer their compensation structured so that they bear less personal risk (Harris and Raviv, 1979; Lewellen, 2006). In addition, executives know that bad organizational outcomes are likely to be attributed to them, this is called ‘employment risk’. When performance is poor, this can induce risk of losing job or bad professional reputation. CEOs will take less risky actions to reduce the variability of the firm’s outcomes. With lower cash flow volatility, his undiversifiable employment risk will be reduced (Canella and Gray, 1997; Amihud and Lev, 1981). Furthermore, Mehran (1995) suggests that incentive-compensation plans motivate managers to take on more risk. This will be discussed in the next paragraph.

2.3 Compensation package

Shareholders structure CEO compensation to provide appropriate incentives. Mehran (1995) suggest that the level of compensation is less important than the composition of

(8)

8 incentives such as stock ownership, performance-based bonuses, stock options, and

performance-based dismissal decisions (Jensen and Murphy, 1990).

The fixed components of executive compensation protect executives from factors beyond their control. Risk borne by CEOs will increase when the compensation is based on firm outcomes. A CEO cannot fully control firm outcomes such as stock prices, because the market also influences fluctuations in stock prices. Risk borne by CEOs will increase because of these fluctuations (Canella and Gray, 1997). This enlarged amount of risk can result in decisions that are not fully value maximizing.

Although, Jensen and Murphy (1990) suggest that setting a lower variable part to transfer risk from managers to shareholders will result in poor executive incentives. When designing a contract, it is important to put fixed and variable compensation on the balance. High variable compensation creates incentives and high fixed compensation reduces

managers’ risk. Hence, designing optimal compensation contracts reflect a trade-off between the goals of providing efficient risk sharing and providing the CEO with incentives to take appropriate decisions.

2.4 Equity-based compensation

Jensen and Murphy (1990) conclude that equity-based rather than cash compensation gives managers the correct incentive to maximize firm value. With equity-based compensation the managers’ claim on firm outcomes increase, which makes them more willing to accept risky, positive NPV projects. Because of these risk-taking incentives, shareholders should prefer that managers’ compensation contain more equity-based forms of compensation.

Although, it may not be optimal to tie all of the components of compensation packages to the firms stock price. Mangers may prefer fixed cash compensation over equity-based compensation. This has several reasons. First of all, equity-based compensation is tied to the firm’s stock return. Stock prices are partly influenced by factors beyond managers’ control such as market fluctuations. These implications in the compensation package can lead to risks in the compensation structure for the CEO (Amihud and Lev, 1981). Second, CEOs human capital is already related to the firm’s stock performance. This increases the managers’ employment risk (risk of losing job, professional reputation etc.). Finally, a managers’ wealth is limited to a single firm, therefore his wealth is undiversified (Canella and Gray, 1997).

In reaction to these increases in risk borne by managers, managers may ignore positive NPV projects to avoid additional risk. For example, because of his undiversified portfolio the

(9)

9 risk associated with risky projects cannot be diversified. Besides, managers may even engage in activities that reduce the firm risk (e.g. conglomerate mergers). This reduced risk-taking behavior will result in less variable stock price and thus less variability in managers’ human capital and compensation. This behavior in turn can adversely affect shareholders wealth, because shareholders prefer actions with positive net present value, independent of the associated risk (Amihud & Lev, 1981).

In short, equity-based compensation has two effects on managers. First, managers with more security holdings will have more incentives to engage in risky projects when it has a positive NPV. And second, the risk borne by managers will increase, which will lead to managers avoiding risky projects. Equity-based compensation includes (restricted) stock compensation and option compensation. First, stock compensation will be discussed. After that, option compensation will be discussed.

2.4.1 (Restricted) stock compensation

When a manager receives (restricted) stock as part of his total compensation, his ownership increases. The amount of equity held by the manager is an important factor influencing CEO behavior. When a firm is managed by the owner, the manager will make decisions which will maximize firm value. When his amount of equity falls, his claim on the firm outcomes falls and he will no longer bear the full cost of any non-pecuniary benefits that he consumes. This will encourage him to larger amounts of the corporate resources in the form of perquisites.

Furthermore, his incentive to devote significant effort to creative activities such as searching for new profitable ventures falls. He may even avoid creative activities, because it requires too much effort to manage or to learn about new technologies. Besides, creative activities can create risks for the CEO, such as risk of failure or reputation (Jensen and Murphy, 1990). Hence, ownership is important for a manager to take actions that are value-maximizing, regardless of their risk or amount of effort.

Besides, Mehran (1995) found a strong positive relationship between equity-based compensation and firm performance for 153 manufacturing firms between 1979 and 1980. They conclude that firm performance is positively related to the percentage of equity held by the managers. In addition, Low (2009) found that CEOs of 2,399 Delaware firms decrease firm risk by 6% in response to the greater protection brought by the takeover regime shift in the mid 1990s. The risk reduction is concentrated among executives with low equity-based compensation. This risk reduction is calculated using variance of daily stock returns.

(10)

10 Finally, Amihud and Lev (1981) investigate the difference between strong owner-, weak owner- and management-controlled firms. They found a positive significant relationship between management-controlled firms and conglomerate mergers for 309 of the largest US industrial firms. Hence, they conclude that managers who have less equity holdings engage more often in risk-reducing projects (e.g. conglomerate mergers) than managers with more equity holdings. Besides, the operations of manager-controlled firms were found to be more diversified than those of owner-controlled firms. This means that the managers with small ownership try to diversify their wealth.

2.4.2 Option compensation

A (restricted) stock option gives a manager the right to buy a share at a pre-determined exercise price on or before a pre-determined date. When the share price is above the exercise price, it is profitable for the manager to buy the share against the exercise price. The option is then called in-the-money. Out-of-the money means that the option has an exercise price above the market price. At that point, it is not profitable to exercise the option. An important

principle of options is that the value of an option increases with the volatility of the stock. When the volatility of the stock increases, the probability of very high and very low returns for the stock increases. An option gives the right, not the obligation, to exercise. Therefore options provide protection against price declines. When a stock price is below the exercise price, the manager decides not to exercise the option. But when the stock price is high, the manager will exercise the option.

Lewellen (2006) suggests that out-of-the-money options tend to encourage risk-taking behavior. With risky behavior, the stock price volatility will increase and hence the expected return on the option will increase. She tries to explore how the firm’s mix of stock and option compensation affects managerial incentives to raise or lower risk-taking behavior. She uses a large sample of 1,587 US firms for the period 1993 through 2001. In this research, leverage is used as a proxy for risk-taking behavior. This research has several main findings. First, stock-based compensation can make debt financing costly to managers. Stock-stock-based compensation exposes managers to firm-specific risk, which gives them an incentive to lower debt. Besides, it is costly to perfectly align managerial incentives with those of shareholders, because

managers have better information than shareholders about the costs and benefits of debt. Second, she finds that the probability of debt issue is positively related to option ownership and negatively related to share ownership. This means that option compensation encourages risk-taking behavior and stock compensation does not. Hence, the main conclusion of this

(11)

11 research is that executive compensation has a significant impact on managers’ financing incentives.

Besides, Rajgopal and Shevlin (2002) investigate whether executive stock options (ESOs) provide managers with incentives to invest in risky projects. They found a positive significant relation between risk-taking and the ESO values for 117 oil and gas CEOs for the period 1992 through 1997. Overall the results are consistent with managers taking riskier action in response to ESOs. When stock return volatility increases, managers have greater incentives to undertake actions that increase firm risk. This is because option values increase with stock return volatility.

2.5 Other research

Previous research has been done to test the relationship between executive compensation and the risk-taking behavior of CEOs. This paragraph will describe a couple of papers that investigated this relationship.

Canella and Gray (1997) examine total compensation, compensation risk, and compensation time horizon. They argue that compensation may be used to encourage risk-taking behavior. The final dataset includes information on 100 firms listed on the New York Stock Exchange, the American Stock Exchange or the NASDAQ Exchange. Regulated industries, such as banking and public utilities, were excluded because regulation restricts executive discretion. They used generalized least squares regression (GLS) and used Beta, Sigma, and Income Stream Risk as a proxy for firm risk. One of the most important conclusions of this research is the negative significant relation between total compensation and firm risk.

Agrawal and Mandelker (1987) examine the relationship between a manager's holdings of common stock and options and the characteristics of the execued investment decisions. They examine three types of investment decisions: acquisitions by mergers, acquisitions by tender offers, and divestitures by sell-offs. An initial sample of all firms that were acquired during 1974 to 1982 is taken from CRSP. This resulted in a sample of 153 firms. They first investigated whether there was a relationship between the stock and option holdings of managers and their incentives to make variance-increasing investments. The investments resulting in increased variance (risk) are significantly more common among managers with higher stock and option holdings. Hence, they conclude that stock and options holdings induce managers to take more risky investments.

(12)

12 Coles, Daniel, and Naveen (2006) find that the relation between vega and firm risk runs both ways. They use the logarithm of the variance of daily stock returns as a proxy for firm risk. Vega is the sensitivity of CEO wealth to stock volatility, which increases with equity-based compensation. They use the period 1992-2002 for their sample period. Data requirements limit the final sample size to a maximum of 10,687 observations. They found that riskier firms are more likely to increase CEO portfolio vega, and increased vega also lead to riskier firm policies and higher firm risk.

(13)

13 3. Hypotheses

This part gives an explanation for the main hypotheses of this research. The main objective is to test whether equity-based compensation and risk-taking behavior have a positive

significant relation. Jensen and Murphy (1990); Mehran (1995) conclude that equity-based compensation provides the best incentives to managers to increase firm value. Agrawal and Mandelker (1987) find a positive relationship between stock and option holdings and taking more risky investments. Besides, Low (2009) concludes that equity-based compensation leads to optimal risk-taking behavior of managers. Hence, the main hypothesis is:

H1: Equity-based compensation has a positive effect on risk-taking behavior.

Equity-based compensation can be divided into stock- and option-based compensation. Stock-based compensation increases the managers’ ownership. CEO ownership is important for aligning the managers’ interests with the interests of the shareholders. When the

managers’ ownership increases, his interests for firm-value maximization increase (Canella and Gray, 1997). Such maximization requires the right, probably risky, investments.

Therefore, hypothesis two states:

H2: (Restricted) stock-based compensation has a positive effect on risk-taking behavior.

Finally, several researches have investigated the impact of option-based compensation on managers’ risk-taking behavior. Additionally, Rajgopal and Shevlin (2002); Lewellen (2006) conclude that stock options encourage managers to invest in risky projects. Options provide protection against price declines, so it is reasonable to expect that this will increase the risky-behavior of managers. Hence, the third hypothesis is:

(14)

14 4. Research methodology

This chapter explains the set up of this thesis. The first paragraph will discuss the sample, period and database that will be used. Paragraph two will explain the research methodology used to test the three hypotheses. In the last paragraph, the (control) variables will be discussed.

4.1 Sample, period and database

The main objective of this thesis is to investigate the impact of equity-based compensation on the risk-taking behavior of executives. A combination of risky and non-risky industries has to be used to avoid a biased selection. Hence, the sample contains Standard & Poor’s 500 biggest U.S. firms. A lot of research has been done on this subject in the mid 1990’s, but his thesis investigates the five most recent years. The fiscal year 2009 till 2013 will be investigated to give a contemporary representation of the impact of equity-based compensation. This period is chosen to investigate equity-based compensation after the financial crisis of 2008. With those 500 companies, the sample would contain 2500 observations. Because of missing values in Execucomp, the data set represents a panel of observations representing 489 firms. Further data requirements limit the final sample size to a maximum of 384 firms for the beta sample and a maximum of 274 for the R&D sample.

All data is obtained from Wharton Research Data Services (WRDS). The data for CEO compensation was collected from S&P’s ExecuComp database. This database contains top executive’s bonus, salary and characteristics since 1992 for the S&P 1500 Index. CEO ownership is also found in this database. Furthermore, CRSP is used to find the beta per company. Finally, Compustat database is used to find the company’s R&D expenses and net sales.

4.2 Method

The dataset represents a panel of observations representing 394 firms for the sample with beta as a proxy. For the R&D sample 274 firms are used. Both of the samples are observed

between 2009 and 2013. This presents problems in analysis. The observations of these panel data are likely to be dependent, which can result in divergent variances across cross-sectional units (Canella and Gray, 1997).

(15)

15 First it has to be tested whether to use fixed or random effects. This can be tested with the “Wu-Hausman test”. This tests whether the unique errors (εi,) are correlated with the regressors. With the random effects model, the errors are not correlated. The outcome of the Hausman test is presented in Table 7.1 The outcome is insignificant, which means that equity-based compensation is not correlated with the error term. Then, “Breusch-Pagan Lagrange multiplier test” was done. This tests whether to use a simple OLS regression or the random effects model.The null hypothesis in this test is that the variances across the entities are zero, which means that there is no significant difference across the entities. The output showed that there is a significant difference, so the random effects model is preferred with this data.2 Therefore, we have to use a GLS random effects model in this research. Then, the presence of heteroskedasticity was tested and confirmed. Because heteroskedasticity was present, the regressions are performed with robust standard errors. The following GLS random effects regression model is used in this thesis:

Model 1: Risk takingit = α + ß1* (Equity-based compensation)it + ß2* ln(Firm Size)it +

ß3* (CEO Ownership)it + ß4* (Consumer Discretion)i+ ß5* (Consumer

Staples)i + ß6* (Energy)i +ß7* (Financials)i + ß8* (Health Care)i + ß9*

(Industrials)i +ß10* (Information Technology)i + ß11* (Materials)i+ ß12*

(Telecommunication services)i + ß13* (Utilities)i + εit

In this regression, i indicates firm and t indicates the year. The dependent variable risk taking will have Beta or R&D as a proxy. Beta is used because this indicates firm risk caused by executive risky actions. R&D expenses used as a proxy because these expenses are risky and can be controlled by the CEO. Consumer Discretion, Consumer Staples, Energy, Financials, Health Care, Industrials, Information Technology, Materials, Telecommunication services and Utilities are the company industries. These are dummy variables and are used as control variables. This model will estimate the relation between equity-based compensation and risk-taking. In this research, a positive significant ß1 is expected and tested. Also, hypotheses 2

and 3 have to be tested to indentify the individual effects of stock and option compensation on risk-taking. The following model is used:

1_{See Appendix}

2_{Chibar2(01) = 1542.14}

(16)

16 Model 2: Risk takingit, = α + ß1* (Stock compensation)it + ß2* (Option compensation)it

+ ß3* ln(Firm Size)it + ß4* (CEO Ownership)it + ß5* (Consumer Discretion)i +

ß6* (Consumer Staples)i + ß7* (Energy)i +ß8* (Financials)i + ß9* (Health

Care)i + ß10* (Industrials)i +ß11* (Information Technology)i +ß12*

(Materials)i + ß13* (Telecommunication services)i + ß14* (Utilities) + εit

In addition, Mehran (1995) suggests that form rather than level of compensation is what drives managers. A manager could get a lot of equity-based compensation compared to other managers but it may be the case that this manager also gets a lot of other compensation, which may also influence firm risk. Therefore, the variable (equity-based compensation/total compensation) is used for the next model to control for level of compensation and focus on form. The following model is used:

Model 3: Risk takingit, = α + ß1* (Equity-based compensation/Total compensation)i,t+

ß2* ln(Firm Size)it + ß3* (CEO Ownership)it + ß4* (Consumer Discretion)i +

ß5* (Consumer Staples)_i + ß6* (Energy)_i +ß7* (Financials)_i + ß8*

(Health Care)i + ß9* (Industrials)i +ß10* (Information Technology)i +ß11*

(Materials)i + ß12* (Telecommunication services)i + ß13* (Utilities)i + εit

Furthermore, Coles et al. (2006) found that the relation between vega and firm risk runs both ways. In this sample, simultaneous causality is also possible. Beta measures the volatility of a firm’s stock compared to the market volatility. When a company’s beta is high, the firm risk is high compared to the market. Executives in risky companies usually receive relatively more Base Salary to compensate for this high company risk. These executives therefore receive less variable (e.g. equity-based) compensation. Therefore, the following model is tested:

Model 4: (Base salary/Total compensation)it = α + ß1* (Beta)it+ ß2* ln(Firm Size)it +

ß3* (CEO Ownership)it + ß4* (Consumer Discretion)i + ß5* (Consumer

Staples)i + ß6* (Energy)i + ß7* (Financials)i + ß8* (Health Care)i + ß9*

(Industrials)i + ß10* (Information Technology)i + ß11* (Materials)i + ß12*

(17)

17 After this, R&D expenses are used as a proxy for risk-taking to test robustness. R&D expenses are obtained from Compustat, but not all the data was available. This resulted in a sample of 171 companies, but the distribution of these firms´ industries was good compared to the original sample.3 Hence, it was possible to use this sample for the robustness check. 4.3 Variable description

This paragraph gives an explanation and description of the variables used in this research. First, the independent variables will be discussed. Second, dependent variables will be explained. And finally, the control variables will be discussed.

4.3.1 Dependent variables

Beta is a commonly used measurement for firm risk (Canella and Gray, 1997; Low, 2009). Beta measures the volatility of a firm’s stock compared to the volatility of the market. Beta would be a good proxy because it indicates the firm specific risk that the investor would diversify within his investment portfolio. This firm risk consists of the net effects of all executive risk-taking activities (Low, 2009). Furthermore, a high beta could be associated with high firm return, which is advantageous for shareholders. Beta is the year-end beta value and NYSE is used as the market type.

R&D expenses, leverage, acquisitions or high risk/high return projects are used in several previous studies to measure the level of executive risk taking (Argawal and

Mandelker, 1987; Lewellen, 2006). R&D expenses represent all costs incurred during the year that relate to the development of new products or services and is measured in millions. R&D expenses are risky, because managers do not know in advance if this investment will work. This makes R&D a good proxy for risk-taking behaviour.

4.3.2 Independent variable

Equity-based compensation consists of (restricted) stock-based compensation and option-based compensation (Canella and Gray, 1997; Low, 2009). Because stock and option compensation are both in monetary terms, they are summed.

(18)

18 4.3.3 Control variables

Log(net sales) is used as a proxy for Firm Size. Net sales represent gross sales reduced by cash discounts, trade discounts, and returned sales and allowances for which credit is given to customers. The sales are log transformed to reduce outliers. Rajgopal and Shevlin (2002); Canella and Gray (1997) state that larger firms are likely to pay more to their executives. In this research size is controlled, because equity-based compensation is an important

component of total compensation.

Executives in risky companies usually receive relatively more Base Salary to

compensate for this high company risk. These executives therefore receive less variable (e.g. equity-based) compensation. Base salary (cash and non-cash) is calculated in thousands of dollars during the fiscal year.

In addition, we control for different industries. Coles et al. (2006) say that optimal investment and financial policies depend on industry. Risky industries are expected to give more base salary and less equity-based compensation.

Finally, CEO Ownership is controlled. This is to separate the stock compensation effects from prior CEO ownership (Miller, Wiseman and Gomez-Mejia, 2002). When a CEO with high ownership and high stock-based compensation takes risky actions, it could be caused by his high ownership instead of the compensation. In addition, executives with large equity holdings may need little or no additional incentives from the compensation contract because equity holdings could be a substitute for compensation incentives (Canella and Gray, 1997). CEO ownership is measured as the percentage of the company's shares owned by the CEO in units.

(19)

19 5. Results and Discussion

This chapter gives an explanation of several results. The first paragraph will show and explain the descriptive statistics. The second paragraph will show and discuss the regression results. Paragraph three will show some limitations.

5.1 Descriptive statistics

This paragraph gives an explanation for the used sample. First, an overall description of all variables will be shown. After that, a more detailed description of the compensation methods will be given.

Data about CEO compensation are retrieved from 489 firms. We are dealing with missing values, so not every company had available data of all the five years. The dependent variables Beta and R&D were obtained from different databases and had fewer observations. The beta sample includes 394 firms with a maximum of 1,771 observations and the R&D sample includes 274 firms with a maximum of 1,239 observations. Furthermore, the dataset contains outliers, but it consists of a large panel data so this should not give any problems with the output.

Because both beta and R&D expenses are used as a proxy for firm risk, both of the descriptive statistics will be compared to see if there are large differences between the

samples. For the beta sample the median of total compensation over the whole sample period was $8,433,000. Total compensation includes stock- and option compensation and a base salary. Stock compensation had an overall mean of $4,056,000, which is higher than the average option compensation of this period. The median of base salary was $1,000,000 and is the lowest of the three components. CEO Ownership had an average of 0.598%, which means that the average CEO had 0.598% of the total shares in his possession. But there were also CEO’s who had no shares and the maximum ownership was 33.68%. Net sales, which

represent the size of the company, had a large spread and a mean of $21,397,000. Some firms do not spend on Research and Development, therefore the minimum of this variable was zero. The average R&D expenses among this sample were $631,540.

For the R&D sample, there is no big difference between the median of total

compensation. For base salary, the median is exactly the same. Furthermore, ownership and net sales are also approximately the same. For stock and option compensation, the difference

(20)

20 is bigger. These averages are both significantly higher in the R&D sample than in the beta sample.

The differences in descriptive statistics can be caused by the fact that the composition of the sample is different. These differences are shown in table 15 and 16.4 For example, the beta sample consists for 10.91% of energy companies, compared to 4.01% for the R&D sample. Also, the beta sample included 30 utility companies and the R&D sample had no utility firms.

Table 1: Descriptive statistics beta sample

Variable Mean Median Std. Dev. Min Max Obs. Firms

Stock ($ thousands) 4,056 3,201 4,745 0 131,940 1,771 394 Option ($ thousands) 1,954 1,319 2,632 0 27,663 1,771 394 Total comp ($ thousands) 9,911 8,433 7,585 0.001 137,207 1,763 394 Base salary ($ thousands) 1,069 1,000 442.93 0 5,600 1,771 394 Ownership (%) 0.598 0.095 2.464 0 33.68 1,730 392 Net sales ($ thousands) 21,397 9,095 40,613 425.10 474,259 1,873 394 R&D ($ thousands) 631.54 137.15 1,419 0 10,991 950 207 Beta 1.100 1.070 0.484 0 3.941 1,873 394

Table 2: Descriptive statistics R&D sample

Variable Mean Median Std. Dev. Min Max Obs. Firms

Stock ($ thousands) 4,297 2,858 11,962 0 376,180 1,239 273 Option ($ thousands) 2,496 1,551 5,160 0 90,693 1,239 273 Total comp ($ thousands) 10,267 8,409 13,395 0.001 377,996 1,230 273 Base salary ($ thousands) 1,033 1,000 450.673 0 4,000 1,239 273 Ownership (%) 0.82 0.09 3.034 0 24.42 1,182 272 Net sales ($ thousands) 22,788 7,829 46,245 101.89 474,259 1,335 274 R&D ($ thousands) 747.84 184 1,565 0 10,991 1,335 274 Beta 1.048 1.029 0.443 0 2.690 950 207 4_{See Appendix}

(21)

21 5.2 Regression Results

This research investigates whether equity-based compensation has a positive effect on firm risk. To test the hypothesis, Model 1 is used. First, regressions are performed without any control variables. Equity-based compensation is the only independent variable because we are interested in the relationship between equity-based compensation and firm risk. Having one independent variable ensures precision, but could also lead to omitted variable bias. Omitted variable bias happens when omitting an important variable which determines the dependent variable and is correlated with at least one independent variable. With omitted variable bias, the regression could be biased (Stock and Watson, 2012). Therefore, after the first regression, control variables are added.

With GLS random effects regression, R squared is incorrect. Though, in this research R squared is not important because this research focuses on the relationship between equity-based compensation rather than the explanatory value of equity-equity-based compensation. Besides, a low R squared is obvious because there are a lot of other variables influencing firm risk. Therefore, Chi square is used to assess the models. Chi square measures the models´ goodness of fit. It tests whether all the coefficients in the model are different than zero. When the p-value is smaller than 0.05, the model is good.

In regression (1) of table 3, only equity-based compensation is included. The

coefficient of equity-based compensation (EB comp) is -0.536 and significant. In regression (2) all the control variables are added and the coefficient changes to -0.50. This means that an increase in equity-based compensation of $100.000 will result in a reduction of the beta of 0.50. The coefficient of equity-based compensation remains almost constant. This means that the relationship between equity-based compensation and beta is robust in this model.

When using R&D as a proxy in model 1, without control variables the coefficient on equity-based compensation is negative. This coefficient becomes stronger and significant when adding the control variables. The control variable ln(net sales) has a positive significant effect on R&D expenses, which means that larger firms expend more on R&D. Furthermore the industries consumer discretionary, financials, health care, industrials, information

technologies and materials all have a positive significant effect on the R&D expenses. The p-value of the Wald chi square changes from 0.933 to 0.000 after adding control variables. This means that the control variables contribute to making the model a good predictor. The beta sample and R&D sample both show a negative relationship between equity-based

(22)

22 compensation and risk-taking. Hence, using R&D as a proxy for risk taking behavior leads to the same conclusion.

These findings are inconsistent with the results from Low (2009). She found a positive relationship between high CEO vega and beta for the period 1990 till 2004. This sample is before the financial crisis. The financial crisis could have changed executives’ perspective on risky behavior. Additionally, high equity-based compensation increases managers’ incentives, but on the other hand it increases managers’ risk (Jensen and Murphy, 1990). Equity-based compensation increases the managers’ claim on firm outcomes. This will incentivize managers to increase their risk-taking behavior. But the risk borne by managers will also increase with equity-based compensation. With these results, it seems that the increase in risk borne by managers has a greater influence on the risk-taking behavior of managers than the increase in incentives.

Table 3: regression model 1

EB comp is divided by 100,000 for the Beta sample. EB comp is divided by 1,000 for the R&D sample. Telecommunication service and utility companies are omitted in the R&D sample because of collinearity. Utility companies are omitted in the beta because of collinearity.

Beta (1) (2) RD (3) (4) Constant 1.131*** (0.000) 0.689*** (0.000) 743.80*** (0.000) -6,744*** (0.000) EB comp -0.536*** (0.001) -0.500*** (0.001) -0.082 (0.933) -2.306*** (0.000)

Control No Yes No Yes

Wald chi2 10.54 0.0012 447.67 (0.000) 0.01 (0.933) 518.65 (0.000) Firm years 1,771 1,730 1,239 1,182 Firms 394 392 273 272

*, **, *** = significant at respectively 10%, 5% and 1% P-value in parenthesis

Furthermore, hypotheses 2 and 3 have to be tested. For these hypotheses we use model 2 and the results are presented in table 4. The expectations of the relationship between stock and option compensation and risk taking were both positive. It is remarkable that only the coefficient on option compensation is positive. Without any control variables, the coefficient of stock compensation is -0.735 and significant at a 1% level. The coefficient of option-based compensation without control variables is 0.458 but not significant. For the regression with R&D, only with including the control variables, the coefficient of stock compensation is

(23)

23 negative and significant. Again for R&D, the control variables induce a great increase in goodness of fit.

These results are consistent with the results of Lewellen (2006). She also found a negative relationship between stock compensation and risk taking and a positive relationship between option compensation and risk taking. Option compensation drives managers to increase stock price volatility. But with stock compensation, managers want a stable and high stock price. On average, managers get more than twice as much stock compensation

compared to option compensation. This could explain the negative coefficient of equity-based compensation of model 1.

Stock and option compensation are divided by 100,000 for the Beta sample. Stock and option compensation are divided by 1,000 for the R&D sample. Telecommunication service and utility companies are omitted in the R&D sample because of collinearity. Utility companies are omitted in the beta sample because of collinearity.

Beta (1) (2) (3) RD (1) (2) (3) Constant 1.128*** (0.000) 1.089*** (0.000) 0.706*** (0.000) 744.0953** (0.000) 737.156*** (0.000) -6,838*** (0.000) Stock -0.735*** (0.003) -0.699*** (0.002) -0.199 (0.826) -2.044*** (0.000) Option 0.458 (0.306) 0.494 (0.241) 2.410 (0.702) -7.498 (0.279)

Control No No Yes No No Yes

Wald chi2 8.62 (0.003) 1.05 (0.307) 453.28 (0.000) 0.05 (0.826) 0.15 (0.702) 482.97 (0.000) Firm years 1,771 1,771 1,730 1,239 1,239 1,182 Firms 394 394 392 273 273 272

The results of model 3 are shown in table 5. The same regressions are tested as in table 3 with relative equity-based compensation instead of the level of equity-based compensation. The industries consumer discretionary, energy, financials, health care, industrials, information technology and materials all have a positive significant coefficient at the 1% level.5 So these industries all lead to a higher beta. Hence, when we focus on the form instead of the level of equity-based compensation, the coefficient becomes less significant but stays negative.

Agrawal and Mandelker (1987) also used the ratio (equity-based compensation/total compensation) to test their hypotheses. However, they found a significant relation which showed that increased relative equity-based compensation resulted in variance increasing decisions. Although, they used an old sample (1974-1982) compared to this sample. Besides,

(24)

24 they only used 152 firms in their sample.

Telecommunication service and utility companies are omitted in the R&D sample because of collinearity. Utility companies are omitted in the beta sample because of collinearity.

Beta (1) (2) RD (1) (2) Constant 1.134*** (0.000) 0.765*** (0.000) 750.132*** (0.000) -6,713*** (0.000) Relative EB comp -0.061* (0.068) -0.053* (0.074) -2.600 (0.646) -6.804 (0.501)

Wald chi2 3.33 (0.0682) 452.34 (0.000) 0.21 (0.6461) 459.83 (0.000) Firm years 1,763 1,725 1,230 1,177 Firms 394 392 273 272

After that, we use model 4 to test simultaneous causality. The results from table 6 show that higher beta results in higher relative base salary. The coefficient of beta is 0.020 without control variables and 0.015 with control variables. Both are significant at a 5% level. This indicates that riskier firms will give higher base salary to compensate managers for the risk. This result is in line with the expectation.

In addition, the relationship between total compensation and firm risk is tested. This resulted in a negative significant effect of beta on total compensation. This means that CEOs of riskier firms get less total compensation. This result is consistent with the findings of Canella and Gray (1997), because they also found a negative significant relation between total compensation and firm risk.

Table 6: regression model 4 Utility companies are omitted in the sample because of collinearity.

Rela base (1) (2) Total compensation (3) (4) Constant 0.124*** (0.000) 0.310*** (0.999) 11,243*** (0.000) 14,746 *** (0.000) Beta 0.020** (0.026) 0.015** (0.049) -1,163*** (0.008) -1,348*** (0.007)

Wald chi2 4.96 (0.026) 96.59 (0.000) 6.98 (0.008) 198.16 (0.000)

(25)

25

Firm years 1,763 1,725 1,763 1725

Firms 394 392 394 392

There are researches that used mergers, high risk/high return projects or Research & Development expenses as a proxy for risk-taking. R&D expenses are used as robustness check, because these expenses are easy to measure. Because different samples for beta and R&D are used in the previous regressions, we now test model 1 with the same sample. There was only complete data available for 171 firms. With this smaller sample, the coefficient of equity-based compensation becomes -0.364.6 This is smaller than the coefficient when using the bigger sample. This smaller negative coefficient could be due to omitting firms that had a more negative effect relationship between equity-based compensation and beta. The

coefficient of equity-based compensation in the model using R&D as a proxy is slightly positive without control variables and negative when adding all the control variables. Both regression (3) and (4) have coefficients which are not significant. This means that we cannot conclude that R&D leads to the same conclusion.

To contribute to previous research, a recent sample is chosen for this thesis. Although, 2009 and 2010 may be too close to the financial crisis of 2008. Therefore, a sample of 2011 through 2013 is used to see whether these more recent years will lead to the same outcomes. The results are shown in table 26 in the appendix. Coefficients on equity-based compensation still remain negative for both the beta and the R&D sample. Only for the R&D sample the coefficients are strongly significant on a 1% level. Hence, the results remain the same if we look at the three most recent years.

5.3 Limitations

There are several limitations that can be improved in future research. First of all, there are a lot of missing values in the dataset. When you have a complete dataset, this would lead to stronger evidence. In addition, several control variables are included in the models, but still several important variables could be forgotten in the models. For example, controlling for investment opportunities, this can be measured by the market-to-book ratio. Firms with more investment opportunities will automatically invest more in risky investments. However, to

(26)

26 find the market-to-book ratio was too time-consuming because this cannot be found in

databases. Another example is CEO tenure, the fact that a CEO has high tenure this implies that the shareholders are content with the CEOs risk-taking behavior. Leaving possible important variables could lead to omitted variable bias, which can result in inconsistent and biased estimates of the coefficients.

Furthermore, it is difficult to define firm risk. In this thesis Beta is used as a proxy for firm risk. But Gray and Canella (1997) say that in high beta firms, stock price movements are strongly associated with market influences such as economic cycles, interest rates, and

government policies. Management exerts little or no control on these macroeconomic variables, so this proxy for firm risk is not totally controlled by the executive. Finally, the sample the sample contains Standard & Poor’s 500 biggest U.S. firms. It might be interesting to include firms from other countries as well. Besides only the biggest firms are used in the sample, but adding smaller firms (e.g. S&P MidCap 400 and S&P SmallCap 600) could also lead to stronger results.

(27)

27 7. Conclusion

Executive compensation has always been a point of discussion. Due to all the criticism, investigating whether equity-based compensation has a positive effect on the shareholders risk preferences is necessary. To test this relationship, this thesis used beta as a proxy for firm risk. Based on previous studies, a positive relation between equity-based compensation and risk taking was expected. This thesis discovers the opposite: a negative significant relation between equity-based compensation and managerial risk taking behavior. Hence, we can conclude that for this sample, equity-based compensation does not increase managerial risk taking behavior.

Equity-based compensation consists of two components, stock and option

compensation. When testing these two components separately, a negative significant relation between stock compensation and firm risk was found, which was against the expectations. Besides, a positive relation between stock and firm risk was found. These findings are comparable to the results Lewellen (2006) found in her research. Lewellen (2006) used the level of leverage as a proxy for firm risk and found a negative relationship between stock compensation and leverage. The opposite was found for the relation between option

compensation and leverage, which was also in line with the results of Rajgopal and Shevlin (2002). They found a positive relationship between ESOs and managers’ risky behavior. Several studies (Agrawal and Mandelker, 1987; Amihud and Lev, 1981; Low, 2009) found that equity-based compensation had a positive effect on the firm risk, but these researches may be dated.

High equity-based compensation increases managers’ incentives, but it also increases the risk borne by managers. Incentives and increased risk have an opposite effect on

managerial risk-taking behavior. With the results of this thesis, it seems that the increase in risk borne by managers has a greater influence on managerial risk-taking behavior than the increase in incentives. Further research can expand this research by investigating these effects of equity-based compensation separately.

(28)

28 References

Agrawal, A. and Mandelker, G. N. (1987) Managerial Incentives and Corporate Investment and Financing Decisions. The Journal of Finance, 42 (4), 823-837

Amihud, Y. & Lev, B. (1981) Risk Reduction as a Managerial Motive for Conglomerate Mergers. The Bell Journal of Economics, 12 (2), 605-617

Boulton, T. J., Braga‐Alves, M. V. & Schlingemann, F. P. (2014) Does equity-based compensation make CEOs more acquisitive? The Journal of Financial Research, 37 (3), 267-293

Canella, A. A. and Samuel R. Gray, S. R. (1997) The role of risk in executive compensation. Journal of management, 23 (4), 517-540

Coles, J. L., Daniel N. D. and Naveen, L. (2006) Managerial incentives and risk-taking. Journal of Financial Economics, 79 (2), 431-468

Eisenhardt, M. (1989) Agency Theory: An Essessment and Review. The Academy of Management Review, 14 (1), 57-74

Jensen, M. C. and Meckling, W. (1976) Theory of the firm: managerial behavior, agency costs, and ownership structure. Journal of Financial Economics, 3 (4), 305-360 Jensen, M. C. and Murphy, K. J. (1990) Performance Pay and Top-Management Incentives.

Journal of Political Economy, 98 (2), 225-264

Low, A. (2009) Managerial risk-taking behavior and equity-based compensation. Journal of Financial Economics, 92 (3), 470-490

Lewellen, K. (2006) Financing decisions when managers are risk averse. Journal of Financial Economics, 82 (3), 551-589

Mehran, H. (1995) Executive compensation structure, ownership and firm performance. Journal of Financial Economics, 38 (2), 163-184

Miller, J. S., Wiseman, R. M. and Gomez-Mejia, L. R. (2002) The Fit between CEO

Compensation Design and Firm Risk. The Academy of Management Journal, 45 (4), 745-756

Rajgopal, S. and Shevlin, T. (2002) Empirical evidence on the relation between

stock option compensation and risk taking. Journal of Accounting and

Economics, 33 (4), 145-171

Stock, J.H. & Watson, M.M. (2013) Introduction to Econometrics (pp. 396-397) (Third Edition). Global Edition. England: Pearson Education Limited

(29)

29 I. Appendix

Table 7: Hausman Test

Outcome is >0.05, so insignificant. Therefore use Random Effects Model

Fixed Effects Random Effects Difference

Equity-based comp -6.14e-06 -5.36e-06 -7.87e-07 Chi2(1)= 2.33 Prob>chi2 = 0.1265 Table 8: Total comp 2009-2013 ($ thousands)

Year Mean St. Dev Min. Max. Obs.

2009 7,118.70 6,285.05 .001 70,143.07 423 2010 9,252.41 8,525.97 .001 84,469.51 434 2011 10,972.41 19,935.14 .001 377,996.50 460 2012 10,464.18 7,498.96 .001 78,440.66 478 2013 11,664.26 8,132.61 .001 95,246.19 471

Table 9: Net Sales 2009-2013 ($ thousands)

2009 16,708.14 32,344.98 101.89 406,103 480 2010 18,330.82 35,244.21 143.37 420,016 487 2011 20,183.99 39,946.72 445.82 444,948 488 2012 20,494.91 39,427.65 599.00 467,231 489 2013 20,805.65 39,175.25 681.20 474,259 489 Table 10: Ownership 2009-2013 (%)

2009 0.881 3.079 0 24.556 375 2010 0.884 3.292 0 30.642 429 2011 0.901 3.375 0 33.678 452 2012 0.847 3.058 0 24.419 474 2013 0.699 2.525 0 23.577 471 Table 11: Beta 2009-2013

2009 1.1967 0.6641 0 3.9405 368

2010 1.0673 0.4117 0.0117 2.0889 370

2011 1.0780 0.4336 0.2188 2.2471 375

2012 1.0860 0.4882 0 2.7885 377

2013 1.0782 0.3582 0 2.4764 383

Table 12: R&D 2009-2013 ($ thousands)

2009 644.02 1,359.05 0 9,010 259

2010 696.78 1,515.72 0 10,991 268

2011 752.80 1,573.60 0 9,112 269

2012 802.46 1,647.47 0 10,148 270

(30)

30 Table 13: Correlation

Stock Option Base Salary Total Comp Net Sales Ownership

Stock 1.00 Option 0.0015 1.000 Base Salary 0.1206 0.0880 1.000 Total Comp 0.8376 0.4144 0.3070 1.000 Net Sales 0.1522 0.0362 0.2419 0.1897 1.000 Ownership -0.0556 0.2011 -0.0076 0.0514 -0.0446 1.000

Table 14: All firms

Industry Frequency Percent

Consumer Discretionary 80 16.36 Consumer Staples 38 7.77 Energy 44 9.00 Financials 83 16.97 Health Care 54 11.04 Industrials 65 13.29 Information Technology 60 12.27 Materials 29 5.93 Telecommunication services 6 1.23 Utilities 30 6.13 Total 489 100

Table 15: Firms beta sample

Consumer Discretionary 61 15.48 Consumer Staples 32 8.12 Energy 43 10.91 Financials 72 18.27 Health Care 40 10.15 Industrials 57 14.47 Information Technology 25 6.35 Materials 28 7.11 Telecommunication services 6 1.52 Utilities 30 7.61 Total 394 100

Table 16: Firms R&D sample

Consumer Discretionary 49 17.88 Consumer Staples 31 11.31 Energy 11 4.01 Financials 19 6.93 Health Care 49 17.88 Industrials 41 14.96 Information Technology 50 18.25 Materials 23 8.39 Telecommunication services 1 0.36 Utilities 0 0 Total 274 100

(31)

31 Table 17: regression model 1

EB comp is divided by 100,000 for the Beta sample. EB comp is divided by 1,000 for the R&D sample. Telecommunication service and utility companies are omitted in the R&D sample because of collinearity. Utility companies are omitted in the beta sample because of collinearity.

Beta (1) (2) RD (1) (2) Constant 1.131*** (0.000) 0.689*** (0.000) 743.80*** (0.000) -6,744*** (0.000) EB comp -0.536*** (0.001) -0.500*** (0.001) -0.082 (0.933) -2.306*** (0.000) Ownership 0.003 (0.696) -21.163 (0.510) Ln(Net Sales) -0.002 (0.870) 683.992*** (0.000) Consumer Discretionary 0.593*** (0.000) 833.614** (0.030) Consumer Staples -0.030 (0.574) 271.321 (0.323) Energy 0.740*** (0.000) 155.624 (0.658) Financials 0.604*** (0.000) 1,610*** (0.001) Health Care 0.174*** (0.001) 1,918*** (0.000) Industrials 0.589*** (0.000) 1,056*** (0.000) Information Technology 0.464*** (0.000) 2,343*** (0.000) Materials 0.624*** (0.000) 871.626*** (0.008) Telecommunication services 0.136 (0.387) Wald chi2 10.54 0.0012 447.67 (0.000) 0.01 (0.933) 518.65 (0.000) Firm years 1,771 1,730 1,239 1,182 Firms 394 392 273 272

Table 18: regression model 2, beta

EB comp is divided by 100,000 for the Beta sample. Utility companies are omitted in the beta sample because of collinearity. (1) (2) (3) (4) (5) Constant 1.128*** (0.000) 1.089*** (0.000) 0.690*** (0.000) 0.756*** (0.000) 0.706*** (0.000) Stock -0.735*** (0.003) -0.697*** (0.002) -0.699*** (0.002) Option 0.458 (0.306) 0.480 (0.260) 0.494 (0.241)

(32)

32 Ownership 0.003 (0.704) 0.003 (0.673) 0.003 (0.704) Ln(Net Sales) -0.002 (0.890) -0.012 (0.390) -0.004 (0.788) Consumer Discretionary 0.585*** (0.000) 0.570*** (0.000) 0.575*** (0.000) Consumer Staples -0.036 (0.500) -0.042 (0.424) -0.042 (0.431) Energy 0.736*** (0.000) 0.722*** (0.000) 0.730*** (0.000) Financials 0.602*** (0.000) 0.584*** (0.000) 0.596*** (0.000) Health Care 0.164*** (0.001) 0.146*** (0.005) 0.152*** (0.003) Industrials 0.580*** (0.000) 0.571*** (0.000) 0.571*** (0.000) Information Technology 0.450*** (0.000) 0.425*** (0.000) 0.433 (0.000) Materials 0.615 (0.000) 0.607*** (0.000) 0.607 (0.000) Telecommunication services 0.142 (0.356) 0.111 (0.443) 0.140 (0.348) Wald chi2 8.62 (0.0033) 1.05 (0.3065) 448.85 (0.000) 453.85 (0.000) 453.28 (0.000) Firm years 1,771 1,771 1,730 1,730 1,730 Firms 394 394 392 392 392

Table 19: regression model 2, R&D

EB comp is divided by 1,000 for the R&D sample. Telecommunication service and utility companies are omitted in the R&D sample because of collinearity.

(1) (2) (3) (4) (5) Constant 744.095*** (0.000) 737.156*** (0.000) -6,701*** (0.000) -6,837*** (0.000) -6,838*** (0.000) Stock -0.199 (0.826) -2.085*** (0.000) -2.044*** (0.000) Option 2.410 (0.702) -7.737 (0.260) -7.498 (0.279) Ownership -21.752 (0.490) -19.299 (0.556) -19.874 (0.545) Ln(Net Sales) 680.059*** (0.000) 689.472*** (0.000) 691.780*** (0.000) Consumer Discretionary 822.564** (0.033) 874.619** (0.027) 862.719** (0.029) Consumer Staples 262.166 (0.339) 309.188 (0.275) 296.253 (0.295) Energy 146.576 (0.675) 192.363 (0.594) 180.1 (0.617) Financials 1,593*** (0.001) 1,655*** (0.001) 1,650*** (0.001) Health Care 1,902*** (0.000) 1,969*** (0.000) 1,957*** (0.000) Industrials 1,043*** 1,104*** 1,090***

(33)

33 (0.001) (0.000) (0.000) Information Technology 2,325*** (0.000) 2,390*** (0.000) 2,384*** (0.000) Materials 858.149*** (0.009) 918.342*** (0.007) 905.899*** (0.008) Telecommunication services Wald chi2 0.05 (0.8263) 0.15 (0.7018) 478.99 (0.000) 454.44 (0.000) 482.97 (0.000) Firm years 1,239 1,239 1,182 1,182 1,182 Firms 273 273 272 272 272

Telecommunication service and utility companies are omitted in the R&D sample because of collinearity. Utility companies are omitted in the beta sample because of collinearity.

Beta (1) (2) RD (1) (2) Constant 1.134*** (0.000) 0.765*** (0.000) 750.132*** (0.000) -6,713*** (0.000) Relative EB -0.061* (0.068) -0.053* (0.074) -2.600 (0.646) -6.804 (0.501) Ownership 0.003 (0.747) -23.443 (0.468) Ln(Net Sales) -0.009 (0.484) 679.248*** (0.000) Consumer Discretionary 0.580*** (0.000) 841.077** (0.031) Consumer Staples -0.034 (0.512) 278.082 (0.315) Energy 0.731*** (0.000) 161.335 (0.647) Financials 0.592*** (0.000) 1,606*** (0.001) Health Care 0.162*** (0.001) 1,919*** (0.000) Industrials 0.584*** (0.000) 1,061*** (0.000) Information Technology 0.451*** (0.000) 2,350*** (0.000) Materials 0.617*** (0.000) 875.720*** (0.008) Telecommunication services 0.116 (0.433) Wald chi2 3.33 (0.0682) 452.34 (0.000) 0.21 (0.6461) 459.83 (0.000) Firm years 1,763 1,725 1,230 1,177 Firms 394 392 273 272

(34)

34

Table 21: regression model 4 Utility companies are omitted in the beta sample because of collinearity.

Rela base (1) (2) Total compensation (3) (4) Constant 0.124*** (0.000) 0.310*** (0.999) 11,243*** (0.000) 14,746 *** (0.000) Beta 0.020** (0.026) 0.015** (0.049) -1,163*** (0.008) -1,348*** (0.007) Ownership 0.008 (0.239) 51.326 (0.809) Ln(Net Sales) -0.015*** (0.000) 2,402*** (0.000) Consumer Discretionary -0.067*** (0.000) 6351.901*** (0.000) Consumer Staples -0.038*** (0.000) 1824.019** (0.023) Energy -0.050*** (0.002) 4476.776*** (0.000) Financials -0.043*** (0.000) 4775.972*** (0.000) Health Care -0.050*** (0.002) 4422.343*** (0.000) Industrials -0.047*** (0.000) 3,526*** (0.000) Information Technology -0.064*** (0.003) 5,597*** (0.000) Materials -0.040*** (0.002) 3,712*** (0.000) Telecommunication services -0.064*** (0.000) 5,670** (0.026) Wald chi2 4.96 (0.026) 96.59 (0.000) 6.98 (0.008) 198.16 (0.000) Firm years 1,763 1,725 1,763 1725 Firms 394 392 394 392

Table 22: 171 firms robustness test

A Health Care EW Health Care MMM Industrials

AA Materials F Consumer Discretionary MO Consumer Staples

ABC Health Care FDO Consumer Discretionary MON Materials

ABT Health Care FLS Industrials MRK Health Care

ACN Information Technology

FMC Materials MWV Materials

(35)

35

AGN Health Care GD Industrials NWL Consumer Discretionary

AIV Financials GE Industrials PBI Industrials

AME Industrials GIS Consumer Staples PEP Consumer Staples

APD Materials GLW Industrials PFE Health Care

APH Industrials GME Consumer Discretionary PG Consumer Staples

ATI Materials GPC Consumer Discretionary PH Industrials

AVB Financials GPS Consumer Discretionary PKI Health Care

AVGO Information Technology

GT Consumer Discretionary PLD Financials

AVP Consumer Staples GWW Industrials PLL Industrials

AVY Materials HAL Energy PM Consumer Staples

AZO Consumer Discretionary

HAR Consumer Discretionary PPG Materials

BA Industrials HCN Financials PX Materials

BAX Health Care HCP Financials RAI Consumer Staples

BCR Health Care HD Consumer Discretionary RHT Information Technology

BDX Health Care HOG Consumer Discretionary RL Consumer Discretionary

BHI Energy HON Industrials ROK Industrials

BLL Materials HOT Consumer Discretionary ROP Industrials

BMY Health Care HP Energy RTN Industrials

BSX Health Care HPQ Information Technology SEE Materials

BWA Consumer Discretionary

HRL Consumer Staples SHW Materials

BXP Financials HRS Information Technology SJM Consumer Staples

CAG Consumer Staples HSP Health Care SLB Energy

CAM Energy HST Financials SNA Consumer Discretionary

CAT Industrials HSY Consumer Staples SPG Financials

CI Health Care HUM Health Care STJ Health Care

CL Consumer Staples IBM Information Technology SWK Consumer Discretionary

CLX Consumer Staples IFF Materials SYK Health Care

CMI Industrials IP Materials T Telecommunications

Services

COL Industrials IR Industrials TDC Information Technology

COP Energy ITW Industrials TEL Information Technology

COV Health Care JCI Consumer Discretionary TGT Consumer Discretionary

CPB Consumer Staples JNJ Health Care THC Health Care

CRM Information Technology

JWN Consumer Discretionary TJX Consumer Discretionary

CVS Consumer Staples K Consumer Staples TMO Health Care

CVX Energy KIM Financials TXT Industrials

DD Materials KMB Consumer Staples TYC Industrials

DE Industrials KMX Consumer Discretionary UHS Health Care

DHR Industrials KR Consumer Staples UNH Health Care

DOV Industrials KSS Consumer Discretionary UTX Industrials

DOW Materials LLL Industrials VAR Health Care

DPS Consumer Staples LLY Health Care VMC Materials

DRI Consumer Discretionary

LMT Industrials VNO Financials

DVA Health Care LO Consumer Staples VTR Financials

ECL Materials LOW Consumer Discretionary WAT Health Care

EL Consumer Staples M Consumer Discretionary WHR Consumer Discretionary