The influence of company performance on CEO compensation

(1)

The Influence of Company Performance on

CEO Compensation

Wouter Valk

10200568

Finance and Organization

Organizational Economics

Spring Semester 2015

Supervisor: Pepijn Trietsch

(2)

Statement of Originality

This document is written by Wouter Valk, 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)

1. Introduction

This thesis focuses on the area of executive compensation and how it is affected by company performance in the Netherlands. We know that performance is reflected in CEO

compensation, because this is stated in their contracts. But we want to know to what extent compensation is positively determined by performance, because the higher the effect of performance on compensation, the more the relatively high amounts of CEO compensation can be justified.

Over the last decade, CEO compensation has gained increasing attention due to a rise in criticism on the height of compensation. Especially after the financial crisis of 2007, criticism grew because many bank executives received massive bonuses while at the same time they were asking for government bailouts. Furthermore, the extremely high amounts of salary given to CEOs and the fact that CEOs are increasingly paid more than the average household, received a lot of criticism and questions were raised whether these high pays were

reasonable (Thomsen & Conyon, 2010). Winter (2010) adds to this by saying that in the wake of the financial crisis many felt that the extreme variable pay schemes had led executives and managers to excessive risk taking with short term private benefits for themselves. Also in the Netherlands, we have seen a rise in criticism on compensation policies. For example, the executives at Royal Dutch Shell and Ahold received substantial bonuses in the period of 2003 through 2005, while at the same time the company performances were relatively poor. This led to the fact that people questioned whether these relatively high pays were justified. Recently, the pay for performance issue gained attention in the Netherlands when ABN Amro announced a salary increase for their executives in March

(5)

2015. Unions objected against this raise because the employees had not received a pay raise in years and on top of that the bank fired a lot of employees. Following the criticism, the executives decided to refuse the salary increase.1

Relatively little research on the relationship between performance and compensation has been conducted in the Netherlands and have showed all kinds of results. For example

Duffhues and Kabir (2007) found a negative correlation between performance and pay in the period of 1998 through 2001. On the other hand, van Ees et al. (2007) found a strong

positive correlation in the period of 2002-2006. Furthermore, Cornelisse et al. (2005) found an insignificant positive effect of performance in 2002 and 2003. Research in the US has showed us a very significant relationship between performance and compensation, which is mainly caused by the fact that variable compensation accounts for a big share in total compensation in the US.

This thesis will also present a quantitative research on the effect that company performance has on CEO compensation. As Murphy and Jensen (1990, p.1) stated: ‘It’s not how much you pay, but how.’ The central question of this research is:

‘To what extent is CEO compensation affected by company performance in the Netherlands?’ To investigate to what extent there exists a relationship between company performance and CEO compensation, ordinary least squares regressions will be performed. These regressions will show us whether a statistically significant positive correlation between compensation and various performance measures exists. The few researches that have been conducted in the Netherlands have used data prior to 2007.2_{This thesis intends to add something to these} researches by using data from the period of 2011-2013. The sample includes 72 companies from the AEX-index, the AMX-index and the AScX-index.

The following chapter presents a review of previous literature on this matter and will present the hypotheses that will be tested in this research. The third chapter will elaborate on the dataset and how data was retrieved. It will also show how the research is going to be conducted and includes the model and its variables that are going to be used. The fourth 1_{The news items on the ABN Amro case are to be found on the website of NOS (March 20, 2015) or ‘Het}

Financieele Dagblad’ (March 29, 2015).

2_{Duffhues and Kabir (2007), Cornelisse et al. (2005), Duffhues et al. (2003), Mertens et al. (2007), Van Ees et al.}

(2007)

4

(6)

chapter will present the results of the research and these results will be analysed. The last chapter presents an overall conclusion to this thesis and some suggestions for further research.

2. Literature Review

In this chapter the literature regarding CEO compensation and its relationship with company performance will be elaborated upon. First, the principal-agent problem will be discussed. The second paragraph of this chapter will describe how CEO compensation schemes serve as a solution to the principal-agent problem. It will also show how these compensation

schemes are designed and how performance is measured. After that some other determinants of compensation will be presented and the last paragraph gives us earlier findings on the relationship between performance and compensation.

Figure 1 gives us an overview of the link that exists between performance and

compensation. We see that besides performance, other variables determine compensation as well.

Figure1: Link between Performance and Compensation

CEO compensation: - Total Compensation - Variable Compensation

Control Variables, for example: - Firm Size - Board Size - Sector Performance Measures: - Accounting-based - Capital Market-based 5

(7)

2.1 Principal-Agent Problem

The principal-agent problem is seen in large public companies, where a separation of ownership and control exists (Berle & Means, 1932). A separation of ownership and control means that the decisions agents do not bear a major share of the wealth effects of their decisions. It entails that a principal (the shareholders) hires an agent (CEO) to operate the firm and to ensure high performance of the company. This agent acts on behalf of the principal and the principal delegates the decision-making rights to the agent, whose actions and decisions influence the welfare of the principal (Jensen & Meckling, 1976).

The problem occurs when both principal and agent fall in the category of homo economicus. A homo economicus is a rational, opportunistic and individualistic person that always seeks to maximize its own benefits and personal utility. The fact that their goals differ, means that the CEO may not always act in the best interests of the shareholders and therefore we speak of a conflict of interests (Fama & Jensen, 1983).

This conflict of interests can lead to a lot of difficulties for a company. This poses a threat to the return on shares, which is the main goal of the shareholders. So this problem has to be

solved and this can be done by creating a well-designed compensation scheme3_.

2.2 Compensation Schemes for CEOs

This paragraph will begin with an explanation of how compensation schemes can solve the principal-agent problem. After that the structure of the schemes and the performance measures, according to which compensation is determined, will be discussed.

2.2.1 Compensation to solve the Principal-Agent Problem

Companies should create compensation contracts for CEOs that makes compensation for a part dependent on company performance. This gives CEOs incentives to perform for the company and ensures that CEOs become value-maximizing driven entrepreneurs for their companies. With a positive correlation, the alignment of the interests of CEOs and

shareholders can be achieved and therefore the principal-agent problem can be solved (Hall & Liebman, 1998). If a positive correlation is absent, it is not likely that the resources of a company are being managed on a value-maximizing and efficient way. Jensen and Murphy (1990) underlined the importance of a well-designed CEO compensation contract:

3_{Other solutions are: auditing, formal (internal) control systems, budget restrictions and so forth (Jensen and}

Meckling, 1976)

6

(8)

‘compensation policy is one of the most important factors in an organization’s success. Not only does it shape how top executives behave but it also helps determine what kinds of executives an organization attracts.’

More risk-averse executives are more likely to prefer an increase in fixed salary to an increase in variable compensation. Higher quality managers are likely to want a more incentive based pay relative to fixed pay, because they know they are able to reach the performance threshold needed to activate the variable pay. However, it should be noted that 100% variable compensation is not the answer. Potential CEOs will never accept a contract like this, because his compensation is now too much subject to economic

fluctuations. Besides, as Hall and Liebman (1998) stated, 100% variable compensation is not necessary to induce optimal behaviour.

We can conclude that to reduce the principal-agent problem companies need to create a positive relationship between CEO compensation and performance measures. This will induce the CEO to exert extra effort to increase company performance and leads to the alignment of interests of shareholders and CEOs. So we can assume that there exists a positive relationship between total compensation and performance. The first hypothesis is:

H1: The relationship between company performance and total compensation is positive and

significant in the Netherlands.

2.2.2 Structure of CEO compensation

It is not sufficient to just look at total compensation. In order to do a more thorough

research on CEO compensation, we need to examine how total compensation is structured. According to Murphy (1999), CEO compensation contains four pay elements: annual fixed salary, annual bonus, equity compensation and other benefits in the form of pension pay and other perks, like a car or a phone. The annual fixed salary of a CEO is usually determined according to competitive benchmarking across general industries. Which means that the company gives the CEO a base salary that is comparable to companies in the same industry. This base salary will not show any correlation with the company performance (Mertens et al., 2007). Fernandes et al. (2010) created a table that showed the share of every element in total compensation of CEOs across different countries in 2006. They found that fixed salary accounts for 44% of compensation in the Netherlands, compared to 28% in the US.

(9)

As already elaborated upon, compensation schemes need to be incentive compatible to overcome the principal-agent problem. This incentive mechanism of compensation is provided by the annual bonus and the equity compensation for the CEO, also called the variable component of the compensation (Duffhues & Kabir, 2008).

The annual bonus is regarded as a short-term incentive and is decided upon predetermined performance measures, which vary per company (Murphy, 1999). The annual bonus is paid by a lot of companies and most of the times contributes to a substantial share of total compensation: 23% in the Netherlands in 2006, which is about the same in the US (Fernandes et al., 2010).

The other part of variable compensation, equity compensation, is regarded as a long-term incentive, because they cannot be sold straight away. Equity compensation accounts for 22% out of equity compensation in the Netherlands in 2006, compared to 39% in the US (Fernandes et al. , 2010). Jensen and Murphy (1990) stated that ‘CEOs should own substantial amounts of company stock’, because this is the most powerful link between shareholder wealth and executive wealth and therefore aligns their interests.

Equity compensation comes in the form of stock options and restricted stock. A stock option is a contract that gives the holder the right to purchase the underlying stock at some

predetermined price in the future. Restricted stocks are a type of shares that are restricted because they have a time-vesting period: the receiver doesn’t hold the shares until a specific time of years has elapsed (Conyon & Thomsen, 2012).

So in conclusion, the correlation between company performance and compensation is caused by the variable part. This means that our second hypothesis is:

H2: The relationship between company performance and variable compensation is positive

and significant in the Netherlands.

2.2.3 Performance Measures

Both equity compensation and annual bonuses for CEOs are decided upon performance measures. Hall and Liebman (1998) tell us about an important principle that should be incorporated in CEO compensation and that is that CEOs should be rewarded for outcomes over which they have control. But it is also important that a CEO’s pay isn’t based on too

(10)

many performance measures, as Jensen (2001) says: ‘multiple objectives is no objective’. Too many measures, and therefore objectives, lead to confusion and a lack of purpose. Something that companies need to consider is that the performance threshold needs to be set correctly. The threshold must be achievable, which ensures that the CEO will not be demotivated. However, if the thresholds are set too low, the CEO will not be pushed towards high levels of performance and the CEO will perform way below the level he is capable of. Prior research has showed all sorts of performance measures. For example Jensen and Murphy (1990), who conducted one of the first researches on the relationship between compensation and company performance, and Hall and Liebman (1998) looked at

shareholder return and its effect on compensation. Murphy presented us a comprehensive overview of all the performance measures and different elements of compensation that are used by firms in the US. As ‘earnings measures’ he noted that companies use net income, return on assets, return on equity and return on capital. Other measures are: sales, customer satisfaction, shareholder return, costs and economic value added. Van Ees et. al (2007) conducted a research, which was commissioned by the Dutch Corporate Governance Code Monitoring Committee and gave us an overview of performance measures used in the Netherlands. The report showed us that earnings

measures like profitability, return on assets and return on equity are used as a performance measure for short-term variable compensation. For long-term variable compensation most companies use total shareholder return. This overview is quite similar to the overview by Murphy (1999).

Duffhues et al. (2003) presented one of the first researches on compensation in the

Netherlands. Their regressions included earnings measures, like return on equity and return on assets, to measure performance. Duffhues and Kabir (2007) also included return on assets,together with return on sales, total shareholder return and Tobin’s Q, which we had not seen before. They used Q because it offers a good mix between an accounting-based measure and a capital market-based measure and it is used a performance proxy in almost all management literature.

(11)

2.3 Other Determinants of Compensation

Performance is not the only determinant of compensation. Prior research presented us some other determinants like firm size, board size and the sector of the firm.

2.3.1 Firm Size

One variable that is used in almost all prior research is firm size. For example, Jensen and Murphy (1990) found that firm size is the most important determinant of compensation. According to them this is a bad case because firm size says nothing about performance itself. Murphy (1999) called the fact that CEO pay is higher in larger firms the best-documented stylized fact regarding CEO pay. When comparing research in different countries on the relationship between firm size and compensation, Murphy found that this theory holds in every single country.

The fact that compensation is significantly and positively affected by firm size, was investigated by Tosi et al. (2000). They concluded that 5% of the variance in CEO

compensation in the US is explained by changes in firm size, which is higher than the 4% that gets explained by performance.

Dutch research finds the same results. Van Ees et al. (2007) concluded in their research that compensation gets determined by sales and number of employees, which they regarded as measures for firm size. Knop and Mertens (2010) also found that firm size, measured by market capitalization, is the most important determinant of firm size according to their dataset. Because firm size is such an important determinant of compensation, it has led Duffhues and Kabir (2007) and many more to include firm size as a control variable. So the third hypothesis is:

H3: The relationship between total compensation and firm size is positive and significant in

the Netherlands.

2.3.2 Board Size

Larger boards are likely to lead to higher compensation for CEOs. Bebchuk and Fried (2004) presented this statement in their paper on managerial power view, which entails that CEO compensation can be seen as a cause of the agency problem. According to them, large boards, with dispersed ownership, cannot be expected to discuss indefinitely with the CEO over their compensation, because a large board entails that members feel less responsible

(12)

and are less cohesive in correcting the CEO. The result of this is that executives gain

substantial power and therefore influence over their own compensation contracts. So their conclusion was that larger boards have a positive effect on total compensation.

Knop and Mertens (2010) investigated the relationship between compensation and firm size and found that board size has a statistically significant positive effect on total, variable and fixed compensation. The fourth hypothesis that will be tested is:

H4: The relationship between total compensation and board size is positive and significant in

the Netherlands.

2.3.3 Sector

Murphy (1999) pointed out the fact that compensation contracts of CEOs are also

determined according to industry benchmarks and therefore It is likely that pay practices differ across industries. In his research he used four different industries mining &

manufacturing, financial services, utilities and other industries. He found what was already expected: performance measures and pay practices differ across industries. He concluded that absolute performance measures (absolute dollar value) are mostly used in the

manufacturing industry, return performance measures get mostly used in the financial sector and profitability per share is the most popular measure in the utilities sector. Furthermore, his regressions showed that the coefficient of the effect of performance on pay differed across the industries.

Nearly all research includes industry dummies to pick up industry-wide shocks to ensure the unbiasedness of the coefficients. For example, Duffhues and Kabir (2007) used the same distinction in sectors as Murphy did.

H5: CEO compensation differs across sectors.

2.4 Prior Results

This paragraph presents some of the relationships between performance and pay that earlier research has found.

(13)

Murphy and Jensen (1990) used data from 1974 through 1988 in the US and found that correlation was close to zero. According to their research, CEOs received only $3.25 for every $1000 increase in shareholder wealth. They stated: ‘On average, corporate America pays its most important leaders like bureaucrats.’ Hall and Liebman (1998) wanted to test this statement by using data from 1980 through 1994. They showed a significant positive relationship between performance and compensation and therefore they concluded that, based on their dataset, the statement was false. Murphy (1999) showed the same result using data from 1992-1996. Both wrote that the difference in findings was mainly caused by the fact that stock options increased as a compensation element after the mid-80s, which was mainly neglected by Jensen and Murphy.

When looking at prior research in the Netherlands, we need to keep in mind that it wasn’t mandatory for Dutch listed companies to disclose executive compensation prior to 2002. But since September 2002 the Dutch government has implemented a new law, called WOB (Wet Openbaarmaking Bezoldiging Bestuurders en Commissarissen) (Van Ees et al., 2007). This law has made it mandatory for listed firms to disclose financial remuneration of executives. Duffhues et al. (2003) found a significant positive effect of performance on stock options grants in 1997 and 1998 in the Netherlands. They also looked at the reverse relationship and found that granted stock options in 1997 had a positive effect on company performance in 1998, which is the aim of a well-designed compensation contract. Duffhues and Kabir (2007) examined the relationship between pay and performance in the period of 1998-2001 in the Netherlands. Due to data limitation, the maximum amount of firms ranged from 14 in 1998 to 30 in 2001. Their regressions showed that there was no evidence for a positive correlation between cash or total compensation and firm performance and even more striking, they found a statistically significant negative effect of performance on compensation.

Table 1: Summary of Prior Research

This table presents prior statistical research on the correlation between Compensation and Performance. Statistical correlations are shown and the outcomes are either significant positive, weak (positive but insignificant) and negative.

Name Data Period Data Country Compensation Independent

Variables Correlation

Jensen & Murphy

(1990) 1974-1988 US Compensation Total Shareholder Return Weak

(14)

Hall & Liebman

(1998) 1980-1994 US Compensation Total Shareholder Return Significant Positive Murphy (1999) 1992-1996 US Total; Variable;

Base Compensation

Shareholder

Return Significant Positive Tosi et al.

(2000) 1987-1991 US Compensation Total Firm Size Significant Positive Duffhues et al.

(2003) 1997-1998 Netherlands Compensation Equity ROE; ROA Significant Positive Duffhues & Kabir

(2007) 1998-2001 Netherlands Compensation Total ROA; ROS; TSR; Q Negative Van Ees et al.

(2007) 2002-2006 Netherlands Compensation Total Sales; EPS; TSR Sales/EPS: Positive Significant; TSR: Weak Knop & Mertens

(2010) 2006-2008 Netherlands Total; Variable; Base Compensation

Board Size Significant Positive

3 Data and Methodology

This chapter will explain the data and the research methodology that are used. First, the time period and companies are discussed. After that the regression model and the different included variables will be presented.

3.1 Data

The dataset includes multiple entities and more than one different time period and therefore we call it panel data.

3.1.1 Time Period

Data will be collected from the period of 2011 through 2013 in the Netherlands. This is done so the research will use data that is as recent as possible and therefore more interesting for both reader and writer. Furthermore, there is no need to control for the financial crisis in the regressions, because the time period is after the core of the financial crisis. The year 2014 is excluded as it is unlikely that databases already include the information of every firm for this year. Another advantage is the WOB law of 2002, which has made it easier to collect data. Data is collected from December 31 of each year.

(15)

3.1.2 Companies

The dataset contains 72 companies from the AEX-index, AMX-index and the AScX-index. Where, based on market capitalization, the AEX-index represents the 25 largest companies, the AMX-index represents the 25 middle sized companies and the AScX-index contains the 25 smaller firms. These three indices are chosen for two reasons. First, companies that are included on these indices are required to disclose their financial information, which means that data on CEO compensation and other financial measures can easily be collected. Second, larger firms are more likely to give their CEOs higher compensation and therefore their remuneration reports should include more information and data in order to provide transparency to their shareholders.

Companies that were on the indices in December 2013 were used. In the appendix a list of the used companies can be found. Excluded are Cryo Save Group and Aperam, because there is no information available on their compensation. Furthermore, due to the same reason, Ziggo is excluded in 2011 and OCI in 2011 and 2012. Because ING’s policy says that CEOs will never be rewarded with variable compensation, they are excluded as well to prevent outliers in the regressions.

3.2 Research Method

To answer the research question and in order to test the hypotheses, this thesis presents a quantitative research on the relationship between CEO compensation and firm performance. The empirical research will be conducted by making use of an ordinary least squares (OLS) regression model. This regression model will show the causal effect of performance (X) on compensation (Y). The following model is going to be used:

(Ln CEO Compensation)it = α0 + α1(CompanyPerformance)it + α2 (Ln FirmSize)it + α3 (BoardSize)it + α4(Sector) + εit

The dependent variables will first be reviewed. After that the different explanatory variables, company performance, will be discussed. The last paragraph will elaborate on the different control variables.

3.3 Compensation (Y)

For this research both total compensation and variable compensation are used. Due to the fact that CEO compensation is unlikely to be normally distributed and due to the height of

(16)

some of these compensations, compensation will be expressed in natural logarithms in order to decrease outliers.

3.3.1 Total Compensation

Total compensation includes all the elements of compensation: fixed salary, annual bonus, equity compensation and other benefits. The goal of this research is to examine to what extent performance has a statistically significant positive effect on total compensation and therefore this is the most important dependent variable. Total compensation adds fixed salary and other benefits to variable compensation. These elements are not affected by company performance and for that reason total compensation will show a lower correlation with performance than variable compensation.

3.3.2 Variable Compensation

Variable compensation consists of the annual bonus and equity compensation and is determined according to performance measures. Therefore, it is likely that the different performance measures have a significant positive effect on variable compensation. Since 2002 the website of bestuursvoorzitter.nl has collected information on CEO

compensation of Dutch listed firms. They have done this through the annual reports of the firms. The website makes a distinction between fixed salary, annual bonus, stock options, restricted stock and total compensation, where total compensation includes all

compensation elements. The Black-Scholes formula is used to value the stock options and the value at grant date is the value that is used for restricted stock.4

Any missing values are hand-collected through the annual reports. If values are given in another currency, the exchange rate given by the European Central Bank is used to compute the value of the compensation in Euros5. If another CEO was appointed during the year, both the compensation values are used for the respective time they were CEO to form a

representative compensation for the company. Furthermore, severance payments for CEOs that ended their job at the firm are neglected.

4_{Information on the course of action of the bestuursvoorzitter.nl is to be found at:}

http://www.veb.net/bestuursvoorzitter/Uitleg.aspx

5_{The exchange rate of the dollar/euro was retrieved from:}

https://www.ecb.europa.eu/stats/exchange/eurofxref/html/eurofxref-graph-usd.en.html

15

(17)

3.4 Performance (X)

Company performance, the explanatory variable, can be measured in a lot of different ways. In deciding on which performance measures to use, a distinction was made between

accounting-based measures and capital market-based measures. Data on the performance measures is retrieved from Datastream and missing values were hand-collected from the annual reports or the websites of the respective firms.

3.4.1 Return on Assets

Return on assets (ROA) is used as an accounting-based performance measure. It is a good measurement for companies because it shows the profitability of a firm compared to its total assets and therefore shows the efficiency in which companies handle their assets . The ROA is measured as follows:

ROA = Net income / Total Assets

3.4.2 Total Shareholder Return

The capital market-measure is total shareholder return (TSR). CEO compensation is used as a corporate governance mechanism to align the interests of shareholders and CEO in order to increase shareholders’ wealth. So it is beneficial for shareholders to measure company performance along this TSR. The TSR is computed as follows.

TSR = (share price at year end + dividends paid– share price at year beginning) / share price at year beginning

If a company became listed after the beginning of the year, the share price of the first issue was used.

3.4.3 Tobin’s Q

Tobin’s Q represents a mix of an accounting-based measure and a capital market-based measure. Looking at the way it is computed we can see why:

Q = (Market value of common shares + Book value of debt)/ Book value of total assets It is a ratio that presents a firm’s market value relative to its assets. If Q is below one it means that it costs more to replace its assets than to buy the firm, which means that the firm is undervalued. If Q is above one it means that the firm is overvalued.

(18)

3.5 Control variables

Focusing solely on performance, entails that other determinants of CEO compensation are neglected, which will lead to the fact that the coefficients of the regressions will be biased. This is called omitted variable bias and in order to mitigate this bias, certain control variables are added.

3.5.1 Firm Size

Prior research has pointed out that firm size is a strong determinant of CEO compensation, where larger firms pay their CEOs more salary. This is the reason why nearly all research on CEO compensation has included firm size as a control variable to reduce bias in the

research.6_{Total market capitalization will be used to measure firm size, just as is done in} prior research.

Market Capitalization = Shares outstanding at year end x Share price at year end Because market capitalization is not likely to have a normal distribution and is likely to produce extreme outliers, it is expressed in natural logarithms. Both shares outstanding and share price are retrieved from Datastream.

3.5.2 Board Size

The managerial power view tells us that board size is likely to have a significant effect on compensation and therefore needs to be added as a control variable. The size of the board of directors and the board of supervisors are combined as if there is an one-tier board in the company.

The reports of the female board index were used to collect data on board size7_{. Any missing} values were collected through annual reports or the websites of the companies.

3.5.3 Sector

CEO compensation is likely to differ across sectors and therefore this research includes it as a variable. Besides, it has the effect that the data stays independent of each other because it picks up industry-wide shocks.

This research makes use of the method used by Duffhues and Kabir (2007). Their research included dummy variables for four different sectors: manufacturing sector, the trade, 6_{Dutch examples are: Knop & Mertens (2010), Duffhues and Kabir (2007).}

7_{The Female Board Index is constituted by Prof. Dr. Mijntje Lückerath-Rovers. Its aim is to present an overview}

of the female representation in both supervisory and executive boards in the Netherlands.

17

(19)

services and transportation sector, financial institutions sector and information and communication technology sector. Respective sectors the firms belong to were collected through the database of Datastream.

4. Results

Throughout the research a lot of different results were obtained and retrieved and this chapter will summarize and analyse these results. In the first paragraph the descriptive statistics will be reviewed. The regression results will be analysed in the second paragraph. 4.1 Descriptive statistics

The summary statistics of each separate element of CEO compensation over the whole period will first be discussed. After that the distribution of the elements of total

compensation will be reviewed, followed by summary statistics of the three used indices. The last part presents the statistics of the independent variables.

4.1.1 CEO Compensation 2011-2013

As table two shows, the average total compensation of Dutch CEOs in the period of 2011 through 2013 was €1,820,000, which is 65.47 times higher than the average income in the

(20)

Netherlands in 2013 (€27,800)8_{. The average fixed salary over the period was €580,000 and} accounts for 32% of total compensation.

Since companies don’t always pay an annual bonus and equity compensation, it is not surprising that the minimum amount of these elements is €0. A large spread exists in annual bonus: €3,500,000. The spread with equity compensation is even higher: €7,110,000. As would be expected, the standard deviation of the annual bonus and equity compensation is higher than fixed salary. This is because annual bonus and equity compensation get

determined by performance measures, whilst the fixed pay is more stable. Annual bonus and equity compensation account for 55% of total compensation.

Compared to the figures in 2006, that were presented by Fernandes et al. (2010), we see that the share of fixed salary has decreased (44% becomes 32%). And although we see that the annual bonus stayed the same, equity compensation increased significantly (22% becomes 32%). This is the same shift that occurred in the US, where equity compensation accounts for a substantial share of total compensation.

Table 2: Descriptive Statistics of CEO Compensation over 2011-2013

Component Obs. Average Share of

Total Comp. Std. Dev. Min. Max.

Fixed Salary 213 €580000 32% €310000 €120000 €1640000

Annual Bonus 213 €421000 23% €560000 €0 €3500000

Equity Comp. 213 €590000 32% €1130000 €0 €7110000

Total Comp. 213 €1820000 100% €1990000 €160000 €13200000

4.1.2 Distribution of Compensation

Table 3 shows us that pay mix has changed a little over the period of 2011 through 2013. Both the share of annual bonus and equity compensation decreased in 2012 and while annual bonus decreased again in 2013, we saw an increase in equity compensation. Fixed salary remained at the same level throughout the years.

Table 3: The Distribution of Compensation over 2011-2013

Year Observations Average Total

Compensation Annual Bonus% Equity Compensation% Fixed Salary%

2011 70 €1730000 26% 33% 32%

8_{Average income is computed by ‘Centraal Bureau voor de Statistiek’.}

19

(21)

2012 71 €1820000 23% 30% 32%

2013 72 €1910000 21% 34% 32%

4.1.3 Differences in Compensation across Indices

Table 4 shows that there are huge differences in compensation between the indices.

Average compensation is 4.79 times higher in the AEX-index than in the AScX-index and 2.76 times higher than the AMX-index over the period. These statistics are in line with earlier research, which states that firm size is a significant determinant of compensation.

Table 4: Descriptive Statistics for the Indices

Index Obs. Average Total

Compensation Median Std. Dev. Min. Max.

AEX 69 €3450000 €2820000 €2570000 €422000 €13100000

AMX 72 €1250000 €1030000 €819000 €386000 €4930000

AScX 72 €721000 €545000 €599000 €160000 €3450000

4.1.4 Independent variables

If we look at the descriptive statistics of the independent variables, we see that return on assets and total shareholder return are positive and that means that the companies increased their value. The average of Tobin’s Q amounted to 1.40 which means that the firms’ market values are 1.40 times higher than their actual values and therefore we

conclude that, on average, the firms are overvalued. The high standard deviation of firm size, measured by market capitalization, tells us that it is necessary to transform the market capitalization into natural logarithms to decrease outliers. Average board size is 8.74, which does not say that much as literature does not give us a rule of thumb to decide whether a board can be considered large or small.

Table 5: Descriptive Statistics Independent Variables

Variables Obs. Average Std. Dev. Min. Max.

ROA 216 0.03 0.10 -0.36 0.68

TSR 216 0.08 0.32 -0.86 0.83

Q 216 1.40 0.78 0.38 7.36

FirmSize 216 €7.31e+09 €2.21e+10 €1.57e+07 €1.75e+11

BoardSize 216 8.74 2.97 3 17

These are the statistics of the different variables that are included in the regressions. ROA = return on assets; TSR = Total Shareholder Return; Q = Tobin’s Q. Firm size gets measured by market capitalization. Board size includes both directors board and supervisory board.

(22)

4.2 Regression

The first paragraph elaborates on the assumptions for OLS regression. After that, the

regression results for both total compensation and variable compensation will be discussed.

4.2.1 Assumptions for OLS Regression

OLS regressions can only be performed if certain standards are met. Stock and Watson (2012) presented them: ensure that there is a linear relationship between X and Y, large outliers are unlikely, X and Y are independently and randomly distributed and there is no perfect multicollinearity.

Linearity and Outliers

Scatter plots show us whether a linear relationship between X and Y exists and whether outliers are present. Although all scatterplots show a linear relationship, we do see that there are some outliers in the scatterplots with variable compensation on the y-axis. These outliers are caused by the fact that some CEOs received zero variable compensation in a year, while performance was comparable to other firms. Explanation for this is that these firms use other performance standards and these outliers won’t be excluded from the regressions.

X and Y are independently and randomly distributed

Because the companies are independent of each other, means that X and Y are

independently and randomly distributed. We do need to correct for time series, because it is likely that a company’s performances in one year show correlation with its performances in another year. This problem is mitigated by using year dummies.

Multicollinearity

Perfect multicollinearity arises when one of the independent variables is a perfect linear function of the other independent variables, which is a problem because it makes it

impossible to compute the OLS estimator. It makes sense, because in multiple regression the coefficient is the effect on Y following a change in the independent variable, holding the other independent variables constant. But if you hold that other independent variable constant, it means that the other stays constant as well because they are perfectly correlated (Stock & Watson, 2012).

(23)

Imperfect multicollinearity arises when one of the independent variables is very highly correlated with the other independent variables. This doesn’t prevent estimation of the coefficient, but it does mean that one or more coefficients could be estimated imprecisely and therefore we would like to mitigate this.

Table 6: Correlation Independent Variables

ROA TSR Q LNCAP BOARD Sector1 Sector2 Sector3 Sector4

ROA 1 TSR 0.24 1 Q 0.53 0.36 1 LNCAP 0.28 0.13 0.22 1 BOARD 0.06 0.06 -0.03 0.72 1 Sector1 -0.07 -0.01 -0.28 -0.04 0.04 1 Sector2 -0.06 -0.02 0.01 0.15 0.19 -0.31 1 Sector3 -0.03 0.06 0.12 -0.11 -0.12 -0.18 -0.34 1 Sector4 0.14 -0.02 0.10 -0.04 -0.14 -0.28 -0.53 -0.31 1

The correlation between LNCAP (natural logarithm of market capitalization) and BOARD is marked because its value of 0.72 is quite high, which makes sense because larger firms need larger boards. We need to account for this collinearity when performing regressions and see whether using board size and firm size simultaneously provides problems.

Another thing that needs to be accounted for is the so called dummy variable trap. In these regressions we make use of sector dummies and each company falls in one sector. If all the sector dummies are included, the regression will fail because of perfect multicollinearity, therefore in every regression one of the sector dummies is excluded. The same goes for the year dummies: one out of three year dummies is left out.

4.2.2 Results

Table 7 presents the results of the regressions with CEO compensation and the independent variables. Because we have two dependent variables and three different performance measures, six regressions are performed. The coefficients of the year dummies are left out of the table because they are not the focus of this research.

(24)

Because correlation between board size and firm size is quite high, a comparison is made between models that include both variables and models that only include one of the

variables (tables 8 and 9 in the appendix). We see that regressions with both variables have higher R2_{values, which means that these models are better at explaining the variance in} compensation. The F-statistics are similar and this means that both models include variables that are equally significant. Therefore we can conclude that the correlation between firm size and board size provides no serious problems and so all regressions include both variables.

It should be made clear that when deciding on whether to reject or accept a hypothesis, acceptance doesn’t mean it is regarded as true. Further evidence might still reject the hypothesis and vice versa. The results for total compensation will first be discussed. Total Compensation

The first three regressions use the natural logarithm of total compensation as dependent variable. As was expected, the effect of the three performance measures on total

compensation is positive. We see that the coefficient of ROA is 0.62. This means that an one unit increase in ROA (+1), leads to an increase of 62% in total compensation and the same reasoning goes behind the coefficients of TSR and Q. Although the effect is positive, it is not statistically significant as is shown by the p-value of the coefficient of ROA (0.133). We see the same results with the other performance measures and especially the coefficient of TSR (0.05) is very insignificant, measured by its p-value of 0.751. Therefore, based on our

dataset, we reject the first hypothesis (H1: The relationship between company performance and total compensation is positive and significant in the Netherlands).

In line with earlier research we find that firm size, measured by the natural logarithm of market capitalization, is an important determinant of the level of total compensation. The coefficient in the first regression is 0.26. This means that for every 1% increase in firm size, the compensation increases by 0.26%. The p-value is 0.000 in all three regressions and this entails that the firm size has a statistically significant positive effect on total compensation. Therefore, we accept the third hypothesis, which says that the relationship between total compensation and firm size is positive and significant in the Netherlands.

(25)

The regressions also show us that board size has a significant and positive effect on total compensation. The coefficient of 0.08 in regression one and three is significant at the 1% level and means that if the board increases by one, the level of total compensation increases by 8%. The second regression has a coefficient of 0.07 and is also statistically significant. This supports Bebchuk’s and Fried’s managerial power view (2004), which states that larger boards lead to higher compensations for CEOs. This means that we accept the fourth hypothesis: The relationship between total compensation and board size is positive and significant in the Netherlands.

We see that the three sectors provide quite different coefficients. The first regression provides a coefficient of -0.31 for the financial sector, which means that it has a statistically significant negative effect on total compensation. Manufacturing on the other hand has a coefficient 0.32, which is positive and significant at the 1% level according to the p-value of 0.001. And although the technology sector has a positive coefficient of 0.12, it is not statistically significant, measured by its p-value (0.300). Regression two and three show similar results and therefore we conclude, based on our dataset, that CEO compensation differs across sectors (Hypothesis 5).

The R2_{-value is 0.70 for all three regressions and this means that 70% of the variation in total} compensation is explained by the included variables in the model. Therefore, we can

conclude that this model is good at predicting total compensation. The F-statistic measures the overall significance of the model, which means that it test the hypothesis that all

coefficients are zero. For all three regressions the F-statistic is high and significant at the 1% level.

Variable Compensation

Regression four, five and six include the natural logarithm of variable compensation as the dependent variable. Again, the coefficients of the performance measures show a positive effect, which is in accordance with literature. Although the coefficient of ROA is not statistically significant, we see that the coefficient of TSR is significant at the 5% level (p-value of 0.021) and the coefficient of Q is significant at the 1% level, revealed by its p-(p-value of 0.000. So hypothesis two, which states that the relationship between company

performance and variable compensation is positive and significant in the Netherlands, is 24

(26)

accepted for TSR and Q and rejected for ROA according to our dataset. This could mean that the companies in our sample don’t use this measure to determine variable compensation. The height of fixed salary and other benefits explain the findings that performance measures have a statistically significant on variable compensation, but not on total compensation. As table 2 in the descriptive statistics showed, variable compensation amounts up to 55% of total compensation, which means that another 45% of total compensation is not decided upon performances. Although companies in the Netherlands have increased variable pay compared to 2006, it is still not enough to ensure that performance has a statistically significant effect on total compensation. Most research in the US finds a significant positive effect of performance on total compensation, which shows that the Dutch companies haven’t followed their American counterparts in increasing variable compensation’s share in total compensation.

Just as with total compensation, we see that firm size is a very important determinant of variable compensation. In all three regressions the coefficient of firm size is significant at the 1% level, measured by the p-value of 0.000. Interesting is the fact that board size is not statistically significant in regressions four and five, whilst it was a significant determinant of total compensation. Reasoning behind this can again be found in the managerial power view, which states that the higher the firm size, the more influence a CEO has on his compensation contract. Assumed is that executives are risk averse and therefore like to decrease the share of variable compensation and increase the fixed share. Therefore, a larger board doesn’t have a statistically significant effect on variable compensation.

It is interesting that the financial sector has a positive effect on variable compensation, while it has a statistically significant negative effect on total compensation. So we conclude that companies in the financial sector create incentive compatible compensation contracts for their CEOs. The technology sector shows even more incentive compatible contracts for CEOs, measured by 1% significant p-values of the coefficients. Another interesting thing we see is that the coefficients of the manufacturing sector are insignificant, while they have a

significant effect on total compensation. A possible explanation for this is that fixed elements account for a large share in compensation contracts for manufacturing firms.

(27)

The relatively low R2 _{- and F-values for the regressions can be explained by the fact that} some outliers were not excluded from the regressions and therefore the included independent variables are less able to explain the values of variable compensation.

Table 7: Regression Results

This table presents the ordinary least squares regressions that measure the effect of performance on compensation. Regressions 1,2 and 3 use the natural logarithm of total compensation of the CEO as dependent variable. Regressions 4, 5 and 6 use the natural logarithm of variable compensation as dependent variable.

Performance is measured by return on assets (ROA), total shareholder return (TSR) and Tobin’s Q (Q). Firm size is measured by the natural logarithm of market capitalization and board size includes both the size of supervisory boards and directors boards. *, **, *** indicate significant at 1% level, 5% level and 10% level, respectively. P-value in parentheses. Every regression performed includes year dummies and all regressions are performed with robust standard errors. Only three out of four sectors are included to prevent the dummy variable trap. The hypotheses and their expected signs are included below the variables that are explained.

Total Compensation (Y) Variable Compensation (Y)

Regression (1) (2) (3) (4) (5) (6)

ROA 0.62 1.77

(28)

H1/H2: + (0.133) (0.429) TSR H1/H2: + 0.05 (0.751) 1.75** (0.021) Q H1/H2: + 0.07 (0.234) 1.14* (0.000) FirmSize H3: + 0.26* (0.000) 0.27* (0.000) 0.26* (0.000) 0.67* (0.000) 0.68* (0.000) 0.57* (0.000) BoardSize H4: + 0.08* (0.000) 0.07* (0.000) 0.08* (0.000) 0.07 (0.356) 0.06 (0.424) 0.13*** (0.095) Financial -0.31* (0.002) -0.33* (0.001) -0.30* (0.003) 0.45 (0.28) 0.41 (0.327) 0.85*** (0.073) Manufacturing 0.32* (0.001) 0.31* (0.001) 0.31* (0.001) 0.54 (0.313) 0.51 (0.326) 0.55 (0.289) Technology 0.12 (0.300) 0.11 (0.351) 0.10 (0.372) 1.33* (0.002) 1.21* (0.005) 1.18* (0.004) Intercept 7.99* (0.000) 7.79* (0.000) 7.82* (0.000) -2.71 (0.343) -3.10 (0.232) -2.79 (0.277) Summary Statistics R2 _0.70 _0.70 _0.70 _0.27 _0.29 _0.30 F-statistic 95.84 (0.00) 98.51 (0.00) 94.95 (0.00) 16.01 (0.00) 15.39 (0.00) 15.41 (0.00) Number of observations: 213

5. Conclusion

In this thesis the effect of company performance on CEO compensation is investigated. The following question is central to this thesis: ‘To what extent is CEO compensation affected by company performance in the Netherlands?’ An answer to this question is important because the higher the effect of performance on compensation, the more the high compensations for CEOs can be justified.

According to literature, CEO compensation can be used as a corporate governance mechanism to align the interests of CEO and shareholders. This alignment is realized by creating incentives for the CEO, which means that compensation is made dependent on

(29)

performance. The dependency will induce the CEO to exert extra effort and therefore, it is expected that the height of compensation is positively determined by performance. Previous research confirms this hypothesis and most of them have found a statistically significant positive effect of performance on CEO compensation.

This research analyses data from 72 companies in the Netherlands over the period from 2011 to 2013. Ordinary least squares regressions are performed with total compensation and variable compensation as dependent variables. Return on assets, total shareholder return and Tobin’s Q are used as performance measures. After checking whether certain conditions for OLS regressions were met, six regressions were performed.

These regressions show that all performance measures have a positive effect on both total compensation and variable compensation. However, performance does not have a

statistically significant positive effect on total compensation, so the first hypothesis is rejected. The rejection also provides an answer to our research question: performance has an insignificant positive effect on CEO compensation. Regarding variable compensation, we see that although TSR and Q have a statistically significant positive effect, ROA has an insignificant positive effect. Therefore, the second hypothesis is rejected for the latter performance measure, but accepted for TSR and Q.

The height of fixed salary and other benefits provide an explanation to the findings that performance measures have a statistically significant effect on variable compensation and an insignificant effect on total compensation. The summary statistics showed us that 45% of total compensation is not decided upon performances and so we conclude that Dutch companies have not made the shift to variable compensation like we have seen in the US. In line with earlier research, all regressions showed that firm size is a statistically significant positive determinant of total compensation. Therefore, the third hypothesis was accepted. Hypothesis four, which states that board size has a significant positive effect on total

compensation, was accepted as well and so our findings support the managerial power view. Furthermore, the regressions showed that the ratio of variable and total compensation differed a lot across sectors. So the fifth hypothesis was also accepted, which states that CEO compensation differs across sectors.

(30)

One of the limitations of this research is the fact that the dataset is smaller than datasets used in prior research. The more observations are used, the less biased coefficients tend to get. So for further research it is advised to increase the dataset. This can be done by

extending the time period and extending the number of firms. For example, the period from 2002, when the WOB law was introduced, through 2013 could be used and that would significantly increase the observations. Furthermore, this research has not used all the listed companies in the Netherlands and so further research could increase the number of firms. Another limitation is the fact that companies might use other metrics to measure

performance than the three measures that are used in this research, making it possible that other measures do show a positive and significant effect on compensation. There are numerous accounting and capital market measures used throughout the annual reports, so for further research it is advised to include more measures. Furthermore, this analysis looks at the height of compensation, but it might also be interesting to use pay mix, which is the ratio of elements of compensation to total compensation, as a dependent variable.

This research does not include all determinants of compensation and so it is likely that our coefficients are biased because of omitted variables. For example, CEO-specific variables like age, tenure and quality can be added. Also other financial measures like for example

leverage can be included. Biasedness can also be caused by the fact that certain outliers (where variable compensation was zero) were not excluded, because exclusion of these outliers would undermine the overall research.

Something that needs to be considered is the causal relationship between performance and compensation. This research has looked at the effect of performance on compensation. But CEO compensation serves as a way to align interests of shareholders and executives in order to improve company performance. On the basis of this research we cannot conclude

anything on the area of causality. So for further research it is interesting to look at to what extent a better designed or a more incentive compatible compensation contract has led the CEO to exert more effort and to improve company performance.

(31)

Bibliography

Bebchuk, L., Fried, J., 2003. Executive compensation as an agency problem. Journal of Economic Perspectives 17, Summer, 71-92.

Bebchuk, L. and Fried, J., 2004. Pay without Performance: The Unfulfilled Promise of Executive Compensation. Harvard University Press

Berle, A. A., Jr., and Means G.C. ,1932, “The Modern Corporation and Private Property”, New York: Macmillan

Brown, J.S., Davison, L., Hagel III, J., 2010, March 4. The Best Way to Measure Company Performance. Harvard Business Review

(32)

Cadbury Report, 1992, December 2. Report of the Committee on: the Financial Aspects of Corporate Governance

Conyon, M. and Murpy, K., 2000. The Prince and the Pauper: CEO pay in the US and UK. The Economic Journal. Volume 110, pp.640-671

Conyon, M. and Thomsen, S., 2012. Corporate Governance: Mechanism and Systems. New York, NY: McGraw-Hill Education

Core, J., Holthausen, R., Larcker, D., 1999. Corporate governance, chief executive officer, and firm performance. Journal of Financial Economics 51, 371-406.

Core, J. and Guay, W., 2010. Is there a Case for Regulating Executive Pay in the Financial Services Industry. Journal of Accounting Research 40, 613-630

Core, J., Guay, W., Larcker, D., 2003. Executive equity compensation and incentives: a survey. Federal Reserve Bank of New York Economic Policy Review 9, 27-50.

Duffhues, P. and Kabir, R., 2008. Is the pay-performance relationship always positive?: Evidence from the Netherlands. Journal of Multinational Financial Management, 18(1), 45- 60

Ees, van, H, van der Laan, G., Ensegaeth, E.J.P. and Selker, C.A.S., 2007. De relatie tussen de bezoldiging van bestuurders van beursgenoteerde ondernemingen en de prestaties van deze ondernemingen”, Research report commsisioned by the Dutch Corporate Governance Code Monitoring Committee, Rijksuniversiteit Groningen/Towers Perrin Amsterdam

Ely, K., 1991. Interindustry Differences in the Relation between Compensation and Firm Performance Variables. Journal of Accounting Research, Vol.29, No. 1, pp. 37-58 Fama, E.F. and Jensen, M.C., 1983. Separation of Ownership and Control, Journal of Law and

Economics, Vol.26

Fernandes, N. , M. A. Ferreira, P. Matos, and K. J. Murphy (2010). The Pay Divide: (Why) Are U.S. Top Executives Paid More?, ECGI Finance Working Paper, No. 255.

Frydman, C. and Saks, R.E., 2010. Executive Compensation: A New View from a Long-Term Persepective, 1936-2005. Oxford Journals: The Review of Financial Studies.

(33)

Gregg, P., Jewell, S.L., Tonks, I., 2010. Executive Pay and Performance: Did Bankers’ Bonuses Cause the Crisis. Available at SSRN:

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1815210

Hall, B. and Liebman, J., 1998. Are CEOs really paid like bureaucrats?, Quarterly Journal of Economics 113, 653-691.

Haubrich, J. (1994), “Risk Aversion, Performance Pay, and the Principal-Agent Problem, ”Journal of Political Economy, 102(2), 258-76.

Holmström, B., 1979, “Moral hazard and observability”, The Bell Journal of Economics, Vol. 10, No.1, pp. 74-91

Jensen, M. C., and Murphy, K. J. 1990. Performance pay and top management incentives. Journal of Political Economy, 98: 225-264

Jensen, C M and Meckling, H William (1976), „Theory of the firm: Managerial behavior, agency cost, and ownership structure‟, Journal of Financial Economics, Vol. 3, pp.360-395.

Knop, N. and Mertens, G., 2010. The Impact of Ownership and Board Structure on CEO compensation in the Netherlands. Working paper, RSM Erasmus Universiteit Rotterdam

Lückerhart-Rovers, M., 2011, 2012, 2013. The Dutch Female Board Index. Working Paper,TIAS

Mertens, G., Knop, N., en Strootman, R. (2007). Pay-for-performance in Nederland 2002-2006, Working paper, RSM Erasmus Universiteit Rotterdam

Murphy, K.J., 1999. Executive Compensation, Marshall School of Business, University of Southern California

Ozkan, N. (2010). CEO Compensation and Firm Performance: an Empirical Investigation of UK Panel Data. European Financial Management , 17 (2), 260-285.

Prendergast, C., 2002, “The Tenuous Trade-Off between Risk and Incentives” The Journal of Political Economy, Vol. 110, No.5, pp. 1071-1102

(34)

Stock, J.H. and Watson, M.H., 2012. Introduction to Econometrics. Harlow, England: Pearson Education Limited

Tosi, H., Werner, S., Katz, J., Gomez-Mejia, L., 2000. How much does performance matter? A meta- analysis of CEO pay studies. Journal of Management 26, 301-339.

Winter, J.W., 2010. Corporate Governance Going Astray: Executive Remuneration Built to Fail. Duisenberg School of Finance Policy Paper, 5: 1521-1535

Websites:

Bestuursvoorzitter.nl www.cbs.nl

www.ecb.europa.eu

Appendix

Table 8: Regression Statistics: Total Compensation as Dependent Variable

Dependent Variable: Total Compensation (LN)

Regressor (1) (2) (3) (4) (5) (6) ROA H1: + 1.92* (0.000) 0.292 (0.494) TSR H1: + 0.162 (0.390) 0.046 (0.764) Q H1: + 0.30* (0.000) 0.001 (0.984) MarketCap (LN) H3: + 0.336* (0.000) 0.34* (0.000) 0.34* (0.000) 33

(35)

BoardSize H4: + 0.199* (0.000) 0.203* (0.000) 0.204* (0.000) Financial -0.405 (0.002) -0.245* (0.020) -0.47* (0.000) -0.254** (0.013) -0.34** (0.010) -0.254** (0.021) Manufacturing 0.32* (0.003) 0.37* (0.000) 0.269** (0.019) 0.361* (0.000) 0.288* (0.008) 0.361* (0.000) Technology 0.07 (0.579) 0.131 (0.273) 0.023 (0.863) 0.123 (0.296) 0.008 (0.949) 0.126 (0.284) Intercept 12.21* (0.000) 6.9* (0.000) 12.227* (0.000) 7.838* (0.000) 11.78* (0.000) 8.83* (0.000) Summary Statistics R2 _0.57 _0.68 _0.54 _0.68 _0.56 _0.67 F-Statistic 33.74 (0.000) 98.18 (0.000) 28.21 (0.000) 99.87 (0.000) 33.26 (0.000) 100.75 (0.000) *, **, *** indicate significant at 1% level, 5% level and 10% level, respectively. P-value in

parentheses. Every regression performed includes year dummies and industry dummies and all regressions are performed with robust standard errors. Due to collinearity none of the regressions include both board size and firm size. Number of observations: 216

Table 9: Regression Statistics: Variable Compensation as Dependent Variable

Dependent Variable: Variable Compensation (LN)

Regressor (1) (2) (3) (4) (5) (6) ROA H2: + 5.21** (0.012) 1.47 (0.494) TSR H2: + 2.04** (0.014) 1.75** (0.008) Q H2: + 1.64* (0.000) 1.02* (0.000) MarketCap (LN) 0.75* (0.000) 0.74* (0.000) 0.70* (0.000) BoardSize 0.40* (0.000) 0.40* (0.000) 0.41* (0.000) 34

(36)

Financial 0.20 (0.625) 0.51 (0.227) 0.04 (0.923) 0.47 (0.268) 0.76 (0.112) 0.92** (0.06) Manufacturing 0.54 (0.329) 0.58 (0.285) 0.42 (0.452) 0.55 (0.293) 0.50 (0.349) 0.64 (0.234) Technology 1.21* (0.007) 1.34* (0.002) 0.99** (0.028) 1.23* (0.005) 0.97** (0.022) 1.22* (0.004) Intercept 8.47* (0.000) -3.72 (0.126) 8.24* (0.000) -3.88*** (0.09) 5.99* (0.000) -4.42*** (0.061) Summary Statistics R2 _0.19 _0.27 _0.20 _0.29 _0.24 _0.29 F-Value 11.56 (0.000) 16.94 (0.000) 9.62 (0.000) 16.49 (0.000) 11.67 (0.000) 16.19 (0.000) *, **, *** indicate significant at 1% level, 5% level and 10% level, respectively. P-value in

parentheses. Every regression performed includes year dummies and industry dummies and all regressions are performed with robust standard errors. Due to collinearity none of the regressions include both board size and firm size. Number of observations: 213

Table 10: Regression Statistics of Relationship Total Compensation and ROA

Regressor (1) (2) (3) (4) (5) ROA 2.30* (0.000) 0.12 (0.755) 1.89* (0.000) 0.50 (0.186) 0.62 (0.133) FirmSize(LN) 0.35* (0.000) 0.27* (0.000) 0.25* (0.000) BoardSize 0.207* (0.000) 0.08* (0.000) 0.08* (0.000) Financial -0.31* (0.002) Manufacturing 0.32* (0.001) Technology 0.12 (0.300) Intercept 13.99* (0.000) 7.72* (0.000) 12.30* (0.000) 7.83* (0.000) 7.99* (0.000) Summary Statistics R2 _0.05 _0.62 _0.50 _0.65 _0.71 F-Value 6.99 (0.002) 106.99 (0.000) 48.32 (0.000) 98.13 (0.000) 95.84 (0.00) n 213 213 213 213 213

*, **, *** indicate significant at 1% level, 5% level and 10% level, respectively. P-value in parentheses.

(37)

Table 11: Regression Statistics of Relationship Total Compensation and TSR

Regressor (1) (2) (3) (4) (5) TSR 0.31 (0.171) 0.04 (0.813) 0.16 (0.393) 0.04 (0.781) 0.05 (0.751) FirmSize(LN) 0.339* (0.000) 0.28* (0.000) 0.27* (0.000) BoardSize 0.21* (0.000) 0.07* (0.000) 0.07* (0.000) Financial -0.33* (0.001) Manufacturing 0.31* (0.001) Technology 0.11 (0.351) Intercept 13.972* (0.000) 6.94* (0.000) 12.22* (0.000) 7.63* (0.000) 7.79* (0.000) Summary Statistics R2 0.01 0.6209 0.47 0.65 0.70 F-Value 0.82 (0.48) 106.62 (0.000) 36.17 (0.000) 98.40 (0.000) 98.51 (0.000) n 213 213 213 213 213

Table 12: Regression Statistics of Relationship Total Compensation and Q

Regressor (1) (2) (3) (4) (5) Q 0.34* (0.007) 0.06 (0.367) 0.37* (0.000) 0.14** (0.026) 0.07 (0.234) FirmSize(LN) 0.35* (0.000) 0.26* (0.000) 0.26* (0.000) BoardSize 0.21* (0.000) 0.08* (0.000) 0.08* (0.000) Financial -0.30* (0.003) Manufacturing 0.31* (0.001) Technology 0.10 (0.372) Intercept 13.55* (0.000) 6.67* (0.000) 11.67* (0.000) 7.74* (0.000) 7.81* (0.000) Summary Statistics R2 0.05 0.62 0.51 0.65 0.70 F-Value 2.64 (0.05) 105.72 (0.000) 48.82 (0.000) 96.35 (0.000) 94.95 (0.000) n 213 213 213 213 213

(38)

Table 13: Regression Statistics of Relationship Variable Compensation and ROA

Regressor (1) (2) (3) (4) (5) ROA 5.64* (0.009) 1.03 (0.631) 4.85** (0.022) 1.41 (0.533) 1.77 (0.429) FirmSize(LN) 0.74* (0.000) 0.67* (0.000) 0.67* (0.000) BoardSize 0.40* (0.000) 0.08 (0.341) 0.07 (0.356) Financial 0.45 (0.276) Manufacturing 0.54 (0.313) Technology 1.33* (0.002) Intercept 12.31* (0.000) -3.11 (0.167) 8.90* (0.000) -1.98 (0.47) -2.71 (0.343) Summary Statistics R2 _0.04 _0.25 _0.17 _0.25 _0.27 F-Value 2.84 (0.04) 16.58 (0.000) 14.24 (0.000) 14.94 (0.000) 16.01 (0.000) n 213 213 213 213 213

Table 14: Regression Statistics of Relationship Variable Compensation and TSR

Regressor (1) (2) (3) (4) (5) TSR 2.41* (0.007) 1.83** (0.017) 2.11** (0.012) 1.83** (0.017) 1.75** (0.021) FirmSize(LN) 0.73* (0.000) 0.66* (0.000) 0.68* (0.000) BoardSize 0.39* (0.000) 0.07 (0.371) 0.06 (0.424) Financial 0.41 (0.327) Manufacturing 0.51 (0.326) Technology 1.21* (0.005) Intercept 11.90* (0.000) -3.21 (0.126) 8.58* (0.000) -2.31 (0.342) -3.10 (0.232) Summary Statistics R2 _0.05 _0.27 _0.19 _0.27 _0.29 F-Value 2.85 (0.038) 16.34 (0.000) 11.74 (0.000) 14.33 (0.000) 15.39 (0.000) n 213 213 213 213 213

The influence of company performance on CEO compensation