Executive Compensation, Firm Performance and Anglo-American Influences: Evidence from Europe

(1)

Executive Compensation, Firm Performance

and Anglo-American Influences:

Evidence from Europe

University of Groningen

Faculty of Economics and Business Msc Business Administration, Finance

Author: Annemarie Schulenburg Student number: 1400932

Date: August 2008

1st_supervisor: _{Prof. dr. C.L.M. Hermes} 2nd_supervisor: _{Dr. T.J.B.M. Postma}

(2)

PREFACE II

PREFACE

In recent years, executive compensation has attracted considerable media, political and academic attention. In particular, the debate has centered on the structure of compensation packages and the level of compen-sation, which is often considered to be excessive. A recent article in the Dutch popular press criticized the remuneration packages received by Dutch CEO’s1_{. During the past 6 years, total compensation of Dutch} CEO’s rose on average 62% during the period, while wages paid to employees increased approximately 3% a year. Although the executives’ base salary during that period rose with only 2.8% a year, the notice-able increase in compensation can be attributed to increases in annual bonuses and other long-term incen-tive plans. These high levels of compensation might be justified by outstanding company performance. However, Gerard Kleisterlee, CEO of Philips saw his total remuneration rise 82% while the group’s turn-over fell sharply and shareholders suffered an average loss of around 2%. As I found it difficult to com-prehend that top executives received exorbitant salaries even when their companies performed poorly, the idea of studying executive remuneration was born.

This thesis could not have been completed without the support and guidance of several persons. First of all, I would like to express my gratitude to Niels Hermes for his valuable suggestions and feedback. Fur-thermore, I would like to thank my family and friends who supported me during my years at the univer-sity, and especially during the process of writing this thesis. With their assistance and advice, they encour-aged me to overcome all obstacles to complete this thesis.

Rijssen, August 2008

(3)

ABSTRACT III

ABSTRACT

This study examines the relationship between firm performance, Anglo-American influences, and execu-tive compensation within several European countries. To be more specific, a variety of accounting-based and market-to-book performance measures are used to examine the link between pay and performance. Several variables are analyzed to capture Anglo-American influences: listing on an Anglo-American stock exchange, being an Anglo-American executive serving on a non-Anglo-American management board, and having at least one Anglo-American executive director. The analysis also encompasses some firm and individual executive characteristics to control for confounding effects. A hand-collected data set is com-piled of cash compensation paid to individual executive team members. The sample consists of 906 execu-tive directors, representing 228 listed companies from the United Kingdom, Ireland, France, Germany, and The Netherlands.

The results show that firm performance, using a market-to-book measure, is significantly positively related to total cash compensation and these results were robust to alternative model specifications. A significant positive relationship is also found between Anglo-American listing and executive compensation and this is consistent with a prior study. In addition, the results reveal a positive relationship between being an An-glo-American executive serving on a non-AnAn-glo-American board and compensation. However, no signifi-cant relationship is found between having at least one Anglo-American on the board and executive com-pensation for non-Anglo-Americans who sit on the same board. Finally, a significant positive relationship was observed between executive compensation and the control variables tenure, board size, and firm size proxied by sales.

(4)

TABLE OF CONTENTS IV

TABLE OF CONTENTS

1. Introduction ...1

2. Theoretical Background...3

2.1. Agency Theory... 3

2.2. Anglo-American governance model versus Continental European governance model ... 5

2.2.1. Listing on an Anglo-American stock exchange ... 6

2.2.2. Anglo-American executives... 7

3. Literature Review ...9

4. Data and Methodology... 14

4.1 Data... 14 4.1.1. Dependent variable ... 15 4.1.2. Explanatory variables ... 15 4.1.3. Control variables ... 16 4.1.4. Sample characteristics... 18 4.2. Methodology ... 20 4.2.1. Regression model... 20

4.2.2. The assumptions for Ordinary Least Squares estimation ... 21

(5)

CHAPTER 1:INTRODUCTION 1 1. INTRODUCTION

Executive compensation has been extensively studied by many authors. However, the existing literature that examines executive compensation has focused mainly on companies in the Anglo-Saxon economies of the U.S., and to a lesser extent the U.K., primarily due to a lack of international data on the topic. Nowa-days, more and more countries, especially in Continental Europe, require publicly listed firms to disclose compensation information, providing new research opportunities. Moreover, few papers investigate the determinants of compensation received by CEOs, as well as other executive board members, and this issue has only recently begun to garner the attention in the academic literature (for example Main et al., 1996; Buck et al. 2003; Crespí-Cladera & Gispert 2003; and Duffhues & Kabir 2008).

The objective of this thesis is to fill this gap in the literature by examining executive compensation issues in five European economies, namely the United Kingdom, Ireland, Germany, France and The Netherlands. This thesis attempts to determine how company performance and Anglo-American influences affect com-pensation of the top management team in the aforementioned countries. The link between company per-formance and executive compensation has been widely studied. What distinguishes this study from some previous work, however, is that this study comprises firms from both common and civil law countries. Also, this study includes firms from Ireland and France and, to the best of my knowledge, there are no empirical studies examining executive compensation practices in these countries. There is only one study that examines he effects of Anglo-American influences on executive compensation. Therefore, this study can be a valuable extension of the existing compensation literature.

This thesis makes several contributions to the present literature on executive compensation. First, this study focuses on compensation of CEOs, as well as compensation received by other executive board members, whereas prior research was often restricted to studying CEO compensation. Second, this study extends the existing literature as this is, to the best of my knowledge, the first empirical study on executive team com-pensation that covers several European countries. Third, this study analyses corporate governance variables that have only recently garnered the attention of academics and therefore their relationship with compensa-tion need to be further investigated. Finally, this study draws upon data that was hand-collected from the annual reports of the selected firms, which yields results potentially more reliable than those inferred from survey data (Muslu, 2003).

(6)

(7)

CHAPTER 2:THEORETICAL BACKGROUND 3

2. THEORETICAL BACKGROUND

2.1. Agency theory

(8)

granting shares. In the literature, designing contracts aimed at providing optimal incentives for executives is often referred to as the ‘optimal contracting approach’ (Bebchuk et al., 2002). From an agency theory perspective, the best practice is to link performance to the firm’s stock performance measures through equity based instruments such as stock options, inducing the executives to maximize stock returns and thereby maximizing shareholder’s wealth. Moreover, this could also alleviate the asymmetric information problem as the firm’s stock performance is determined outside of the executive’s personal control, while executives might attempt to window-dress accounting performance measures. Additionally, accounting performance measures reflect the firm’s past performance, whereas equity based instruments also capture the firm’s future firm performance (assuming efficient stock markets). However, prior research shows that in the short run executive compensation is more closely tied to accounting performance measures than to the firm’s stock performance (Lambert & Larcker 1987; Sloan 1993). Over the long run, Boschen et al. (2003) find that chief executive compensation responds more strongly to stock price performance than to accounting performance measures. Following the logic of agency theory the first hypothesis to be tested will be:

H1: There is a positive relationship between company performance and executive remuneration.

As will be discussed in the next section, the literature that examines executive compensation finds only limited support for the optimal contracting approach. Therefore, it is useful to consider an alternative approach known as the “managerial power approach”, presented by Bebchuk & Fried (2003). Unlike the optimal contracting approach, which considers compensation as a potential instrument to address the agency problem, the managerial power approach views executive compensation as part of the agency problem itself. The main premise of the managerial power approach is that executives have power over the board of directors and thus over the compensation committee and are therefore able to influence both the level and structure of the compensation package. Bebchuk & Fried (2003) argue that because of the executives’ influence over the board of directors, the board of directors cannot bargain at arm’s length with executive directors, and consequently could not act in the best interest of the shareholders. The ne-gotiated contracts thus enable the executives to extract ‘rents’, which are benefits the executive obtains in excess of what he would receive under optimal contracting. According to Bebchuk & Fried, the board of directors will be loyal to the executives rather than the shareholders for various reasons. The directors might collaborate with the executives because they wish to be renominated to the board because a direc-torship yields several benefits. Besides the membership to prestigious social circles, directors are paid sala-ries and other benefits that are not to be sneezed at.2_{Another reason to favor the executive directors is} that they can affect directors’ compensation (Bebchuk & Fried 2003, pp. 4). Although the managerial power approach will not be tested separately, I will include some control variables that represent the ex-ecutives’ power as well as some variables that represent the board’s independence.

(9)

As prior research failed to observe a significant relationship between pay and performance, and therefore did not provide strong support for the agency theory, researchers concentrated on a richer set of social, political and environmental explanations. One of these explanations that will also be examined in this thesis is whether Anglo-American influences affect executive compensation in non-Anglo-American com-panies. To be more specific, this thesis investigates whether listing on an Anglo-American stock exchange and having Anglo-American executives on the management board influences compensation. However, before considering these specific items, I will briefly discuss the most important characteristics of the two main corporate governance models: the Anglo-American model, also known as the shareholder model, and the Continental European model, also known as the stakeholder model.

2.2 Anglo-American governance model versus Continental European governance model When it comes to corporate governance, two main models are distinguished based on a country’s legal tradition: the Anglo-American or Anglo-Saxon model3_{, which is dominant in countries with a common} law legal system and the Continental-European model, which is dominant in countries with a civil law tradition (Aguilera & Cuervo-Cazurra, 2004). The first difference between the two models stems from the level of ownership concentration. The Anglo-Saxon model is characterized by highly dispersed share-holder ownership, whereas in the Continental-European model shareshare-holder ownership is more concen-trated and as a result shareholders hold larger percentages of total share capital. Not only does the level of ownership concentration differ between the two governance models, the nature of ownership differs as well. In common-law countries, shares are primarily held by individual investors and institutional inves-tors, whereas in civil-law countries families, other firms and banks are the main shareholders (La Porta et al., 1999; Mayer & Sussman, 2001). Another distinction between the two models is that common-law countries, and hence the Anglo-Saxon model, provide a higher level of protection to shareholders than do the civil-law countries. This higher level of shareholder protection leads to differences in capital structure and liquidity. Specifically, in common-law countries, more companies are publicly listed and these compa-nies are characterized by higher market capitalization facilitated by highly liquid stock markets. In civil-law countries on the other hand, companies have higher levels of leverage as debt is more widely used to fi-nance companies (La Porta et al., 1999). Moreover, the Anglo-Saxon goverfi-nance system is associated with a more stringent disclosure environment. Overall, the general conclusion is that …“the Anglo-Saxon system is commonly regarded as the most demanding corporate governance system” (Oxelheim & Randøy, 2003).

Non-Anglo-Saxon companies that list on an Anglo-American stock exchange, for example The New York Stock Exchange, need to comply with the more stringent disclosure requirements. The next paragraph discusses why this stricter disclosure environment might affect executive pay.

(10)

2.2.1. Listing on an Anglo-American stock exchange

Over the last two decades an increasing number of firms have listed their shares abroad. Pagano et al. (2001 and 2002), who researched the cross-listing of European firms, document that some exchanges, especially those in the United States, attracted a larger number of these cross-listings by European compa-nies than other exchanges.

Companies may list their shares on a foreign stock exchange for a variety of reasons. For example, firms may list abroad to attract capital, either because they can raise the funds on better terms abroad or because the company faces financial constraints in the country of incorporation. When companies list on a large stock exchange, they may enhance their stock’s liquidity because they have access to a larger pool of inves-tors. Another motive for listing abroad, proposed by Cheung & Lee (1995), stems from the fact that com-panies may signal their credibility by listing on certain exchanges with stringent disclosure requirements. Cheung & Lee state that “…a higher level of disclosure conveys to investors the management’s confidence in its future earnings…[which]… will result in better pricing for the share of high quality firms” .

(11)

executive directors in the US and UK companies generally contain more equity-based compensation com-ponents and hence is more incentive-oriented (Cheffins, 2003). However, studies reveal that the level of cash compensation in the US and UK is also higher than in Continental Europe (Conyon & Schwalbach, 2000).

Besides the spillover effects, the executives of the cross-listed firms are exposed to greater risks because they will be monitored by international investors that may put higher demands on the top executive team. Since Anglo-American investors are less lenient toward underperforming executives, there is a greater risk of being dismissed and, therefore the executives will ask higher pay by means of a premium. All these factors make the job of top executive more demanding. Therefore, the pool of capable managers is lim-ited, which induces a higher level of compensation for the executives that have the competences necessary to accomplish such complex tasks. Following Oxelheim & Randøy (2005), the next hypothesis is:

H2: There is a positive relationship between a listing on an Anglo-American stock exchange and executive com-pensation in non Anglo-American firms.

2.2.2. Anglo-American executive directors

Similar to listing on an Anglo-American stock exchange, hiring an Anglo-American executive might also influence the compensation packages. According to Cheffins (2003), companies may hire an American executive because the pool of talented people in the US is deeper than the domestic pool. Moreover, Cheffins (2003) suggests that by hiring a CEO from a country that considers shareholders’ interests of paramount importance, a company can signal that maximizing shareholder value is a high priority. Oxel-heim & Randøy (2005) suggest that by recruiting an outsider Anglo-American board member, a non-Anglo-American firm can signal its compliance with a more demanding corporate governance system. Oxelheim & Randøy argue that “independent outsider Anglo-American board members bring with them the corporate governance culture of their home country which could boost the incentive structure of top management”. In an earlier paper, Oxelheim & Randøy (2003) found, among others, that the inclusion of an Anglo-American board member to the board of directors increases the firm’s value. This increase in firm value may be partially captured by the top executives and consequently pay levels for the executives are higher. However, another and probably more plausible rationalization for the expected higher level of compensation when an Anglo-American is a member of the management team is proposed by Cheffins (2003). According to this author, in order to attract Anglo-American executives, a foreign company needs to offer a compensation package commensurate to what he would obtain in his country of origin. Since pay levels in Anglo-American coun-tries are generally higher, the following hypothesis is posited:

(12)

Having an Anglo-American executive on the management board might cause a spillover effect. To be more specific, the Anglo-American board member’s fellow executives - encouraged by the high pay pack-ages of the Anglo-American executive - will also demand higher pay levels. Hence, the following hypothe-sis is developed:

H4: There is a positive relationship between the inclusion of an Anglo-American member to the executive board and fellow executive compensation in non Anglo-American firms.

In summary, the hypotheses regarding the Anglo-American influence on executive compensation are based on the theories suggested by Oxelheim & Randøy (2005) and Cheffins (2003). I explicitly test the theory proposed by Oxelheim & Randøy (2005) that an Anglo-American stock exchange listing is posi-tively associated with executive compensation in hypothesis 2.

Hypothesis 3 relies on the assumption made by Cheffins (2003) that a non-Anglo-American company needs to pay an Anglo-American a compensation package comparable to what he would receive in his own country.

(13)

CHAPTER 3:LITERATURE REVIEW 9

3. LITERATURE REVIEW

During the last decades, the academic literature on compensation has grown at a pace that is only rivaled by executive compensation itself. General overviews of the literature are provided by Gomez-Mejia & Wiseman (1997), Barkema & Gomez-Mejia (1998), Murphy (1999), Tosi et al. (2000) and Core et al. (2003), among others.

One of the most examined relationships in the executive compensation literature is the relationship be-tween pay and performance. However, the empirical evidence on this relationship is mixed. Murphy (1985), in his analysis of 500 executives from 73 of the largest U.S. manufacturing companies over the period 1964 – 1981, finds a significant positive relationship between executive compensation and share-holder return. The results indicate that for every 10% increase in shareshare-holder wealth, total executive com-pensation (base salary, bonuses and deferred comcom-pensation) raises with 2.1%.

Gibbons & Murphy (1990) also analyze the relationship between CEO compensation and the firm’s stock market performance. In contrast to Murphy (1985), who studied total executive compensation, this study only analyzes base salaries and bonuses. Using a sample of over 1,600 CEO’s from 1,049 U.S. companies spanning the period 1975 - 1986, the authors conclude that base salaries and bonuses change approxi-mately 1.6% for each 10% return on common stock, which confirms the hypothesis that relationship be-tween pay and performance is positively and highly significant. Although the results are statistically highly significant, the economic significance is very small, as every $ 1,000 increase in shareholder wealth corre-sponds to a pay raise of only $ 0.09, on the basis of a median market value of $ 812 million and a median CEO pay of $ 465 thousand.

In their widely cited paper, Jensen & Murphy (1990) estimate the pay-performance relation for a sample of 1,688 executives from 1,049 companies between 1974 and 1986. The results demonstrate that CEO weal-th changes $ 3.25 for every $ 1,000 change in shareholder wealweal-th. Even weal-though weal-the results are statistically significant, the authors suggest they are too weak to provide evidence for a relationship between pay and performance. Many other authors have reproduced the study of Jensen & Murphy with other datasets and methodologies and reported higher pay-performance sensitivities. For example, Hall & Liebman (1998) study 478 large U.S. companies over the period 1980 - 1994 and document that a $ 1,000 increase in firm value leads to a $ 5.29 increase in CEO wealth at the median. The authors argue that the pay-performance sensitivity increased dramatically throughout the investigated period, and this was mainly the result of much higher issues of stock options.

Core et al. (1999) hypothesize a reversed cause-and-effect relationship with respect to pay and perform-ance. They conclude that CEO excess compensation, due to weak corporate governance structure, has a significant negative impact on subsequent accounting performance. Hence, excess compensation does not induce subsequent higher firm performance.

(14)

Empirical studies of executive compensation outside the U.S. have also yielded inconsistent results. For a sample of 755 Canadian firms over the period 1991-1995, Zhou (2000) finds positive relationships be-tween pay and performance. When the applied measure for CEO wealth includes both cash compensation and options and stockholdings, the estimated pay-performance sensitivity equal 0.00562; for every $ 1,000 increase in shareholder wealth, CEO wealth increases with $ 5.62. However, when the measure for CEO wealth only comprises cash compensation, the estimated pay-performance sensitivity decreases to 0.0001. Another variable investigated by Zhou is size, proxied by both the firm’s sales and its total assets. The estimated size elasticity based upon sales is 0.25, which implies that for every 10% increase in the firm’s sales, CEO compensation increases by 2.5%

Analyzing CEO compensation in Australia, Izan et al. (1998), use a sample of 99 firms for the period 1987 – 1992 and found no evidence of a relation between CEO compensation and performance. Contrary to Izan et al., Merhebi et al. (2006) find a positive and statistically significant association between CEO pay and (lagged) performance. Using data for 722 companies over a ten-year period from 1990-1999, the au-thors find that for every 10% increase in shareholder wealth CEO pay increases with 1.16%. Merhebi et al. also examined the pay-performance sensitivity and documented that a $A 1,000 increase in shareholder wealth leads to a $A 1.82 increase in executive wealth. Besides the pay-performance relationship, the au-thors investigated the linkage between pay and firm size and observed highly significant results. Using total revenue as a proxy for size, the estimated size elasticity is 0.274, implying that a 10% increase in revenue corresponds to a 2.74% increase in CEO compensation.

Kato & Kubo (2006), studying individual CEO cash compensation among 51 listed and non-listed Japa-nese companies, find a significant positive pay-for-performance relationship. For the sub-sample of listed firms, the authors documented that a 10% increase in shareholder value leads to an increase of approxi-mately 0.9% in CEO cash compensation. However, the authors argue that using accounting measures of performance provides the most robust results.

Recently, compensation in the emerging Chinese market economy has been scrutinized by Firth et al. (2006). They examine the compensation of Chinese CEO’s with a sample of 549 listed companies over a three-year period from 1998 – 2000. They find positive but very small pay-performance sensitivities for firms with private blockholders.

As this thesis focuses on executive compensation in European countries, some European studies will be discussed now, starting with some empirical evidence from the UK.

(15)

results indicate that a 10% increase in shareholder wealth leads to an increase of total compensation re-ceived by the highest paid director, the CEO and the entire board of 8.94%, 7.20%, and 7.03% respec-tively. The authors argue that these highly statistically significant results are mainly due to executive stock options, a compensation component that has become increasingly important during the period of investi-gation. Buck et al. (2003) examine executive director pay for a sample of 287 large UK companies over a two-year period from 1996 – 1997. Similar to Main et al. (1996), this study also includes long-term re-wards. The authors document that for the model that included all executives, the presence of long-term incentive plans (LTIP) was associated with statistically significant higher average total rewards. For chief executive officers, the presence of LTIP’s was also associated with higher rewards, though this result was statistically insignificant. Furthermore, Buck et al. (2003) reported the estimated pay-performance elasticity and sensitivity. For the model that includes all executives, a 10 percentage point increase in total share-holder return corresponds to a 15.5 % increase in total compensation, LTIP’s included. However, in the absence of LTIP schemes, the elasticity increases to 19.9%. The reported sensitivity reveals that a £ 1,000 increase in shareholder wealth leads to a £ 1.06 increase in total executive compensation when LTIP’s are included, versus a £ 1.36 increase in total executive compensation in the absence of LTIP’s. Therefore, LTIP’s increases the absolute total rewards executives receive, but the presence of LTIP schemes reduces the pay-performance elasticity and sensitivity.

In contrast to the US and UK, there is a dearth of literature concerning compensation issues in continen-tal Europe. A lack of disclosure on executive compensation has been a major impediment to analyses among continental European countries. However, an increasing number of countries have adopted corpo-rate governance regulations that require publicly listed firms to disclose compensation information, facili-tating research into European executives’ compensation packages.

Conyon & Schwalbach (2000) examine executive pay in the UK and Germany using 15-year panel data of 102 and 48 companies respectively. The results reveal that UK CEO’s are paid vastly more than their German counterparts. However, both countries exhibit a positive association between pay and perform-ance. The authors have also estimated the elasticity between cash compensation and firm performance for both the UK and Germany. The results suggest that a 10 percentage point increase in shareholder return yields increases of 0.67% for UK CEO’s. For German executives, the pay-performance elasticity is 0.071, which implies that compensation for German executives increases by 0.71% for every 10 percentage point increase in shareholder return. Regarding German executive compensation, Elston & Goldberg (2003) and Haid & Yurtoglu (2005) confirm Conyon & Schwalbach’s (1999) results. For a sample of 160 large listed German companies over the period 1987 – 2003, Haid & Yurtoglu find a small but significant positive relationship between executive compensation and firm performance. The linkage between executive com-pensation and firm size appeared to be slightly stronger.

(16)

was positively and significantly related to executive compensation. Several robustness checks were per-formed and confirmed the above results. The significant negative pay-performance relationship remained even after including data on stock option grants in the pay measure for a small sub-sample. The authors explain these interesting findings by the fact that ...“managers are able to influence their own compensation, and managerial entrenchment is more likely to take place when there is less corporate governance pressure”…, as was the case in the Netherlands during the period under review. Van Ees et al. (2007) studied executive pay over the period 2002 – 2006 using a sample of 107 Dutch listed companies. In contrast to the results of Duffhues & Kabir (2008), these authors find a significant positive relationship between some components of execu-tive compensation and company performance. Specifically, no relationship exists between base salary and performance, while bonuses and performance are highly correlated. It should be noted that this study uses, among other measures, revenues as a proxy for performance, a commonly used measure for size. Besides cash components, the authors also investigated long term incentive plans in the form of stock option grants and share grants. They find strong evidence of a positive relationship between long-term compensation and relative shareholder return. Mertens et al. (2007) also examined compensation data of the entire top management for a sample of 90 Dutch listed firms during the period 2002 – 2006. The key finding is that company size is the most important determinant of executive pay. Furthermore, the results show that there is a slightly positive relationship between top management pay and company perform-ance. The authors report that the positive relation is stronger for CEO’s than for other members of the top management team.

Crespí-Cladera & Gispert (2003) also study compensation received by the entire management board, using a sample of 113 Spanish listed companies during the period 1990 – 1995. They found a positive relation-ship between pay and performance, especially for accounting measures of company performance. More-over, the authors find supportive evidence for a positive linkage between firm size and executive compen-sation. Finally, the results show that ownership concentration explains part of the pay-performance rela-tionship.

A study conducted by Brunello et al. (2001) among Italian executives reveals a significant positive relation-ship between cash compensation and firm size. The linkage between pay and performance was also posi-tive, though the sensitivity coefficient was very small. The authors further report that pay-performance relationship is non-linear, which implies that …”a manager is penalized more when profits fall, even in they are positive, than he is rewarded when profits increase, even if they are negative.”

(17)

compen-CHAPTER 3:LITERATURE REVIEW 13

sation and company accounting performance. In addition to the pay-performance relationship, the study also examines the linkage between several corporate governance measures and CEO compensation. The results reveal a significant positive relationship between compensation and board size, foreign board membership and market capitalization. To the best of my knowledge, no empirical research has investi-gated the determinants of executive pay for Ireland and France.

Little empirical evidence exists on the relationship between compensation and the specific governance issues studied in this thesis: foreign board membership and foreign listing.

(18)

CHAPTER 4:DATA AND METHODOLOGY 14

4. DATA AND METHODOLOGY

4.1. Data

A large cross-sectional data set has been constructed by former students4_{, which contains data on 363} companies and their top management team for fifteen EU member states. Initially, it was my intention to complement this database with compensation data for the individual executive directors. However, since not all EU member states require publicly listed firms to disclose compensation information, constructing a sample based upon the existing database appeared to be unfeasible. Hence, a new data set was compiled and the selection criterion used for composing this sample was the availability of compensation informa-tion. As publication of executive compensation of individual executive board members is mandatory in the UK, Ireland, Germany, France, and the Netherlands, I gathered data on publicly listed companies of the aforementioned countries.5_{The sample is drawn from all firms listed on the London Stock Exchange,} Irish Stock Exchange, Frankfurt Stock Exchange, Euronext Paris and Euronext Amsterdam for the year 2005 (for UK companies the fiscal year 2005-2006). All members of the management board (companies with a two-tier board) or all executive directors (companies with a one-tier board) for whom individual compensation data is available will be included in the sample. Information on the executive’s compensa-tion, as well as company informacompensa-tion, were hand-collected from the annual reports or otherwise extracted from Amadeus. Companies that were not headquartered in the aforementioned countries were excluded, as were dual-listed companies with headquarters in both common and civil law countries (e.g. Reed El-sevier, Unilever, and Royal Dutch Shell). Finally, companies that had insufficient compensation data were excluded. The final sample consists of 906 observations, representing 228 companies (see Appendix table A1). Table 1 presents the distribution of the number of observations per country.

Table 1

Sample summary

Country Number of companies Number of CEO's Number of other executives Total number of observations

United Kingdom 54 56 149 204 Ireland 32 33 88 121 Germany 42 43 178 221 France 47 53 124 177 The Netherlands 53 53 129 183 Total 228 238 668 906

4_{Ilse Marsman and Daan van der Mee}

(19)

4.1.1. Dependent variable Compensation

Generally, the total compensation package encompasses four components: base salary, annual cash bo-nuses, long-term incentive schemes (e.g. stock option grants, share appreciation rights), and benefits. The level of the base salary, as well as the other compensation components, is usually assessed against a peer group. In general, the emphasis of the annual bonus is on the company’s annual performance and is there-fore considered to be a short-term incentive component. The objective of long-term incentive schemes is to encourage long-term sustainable (shareholder) performance and to align the interests of the executives with those of the shareholders. Finally, benefits include all other emoluments, such as pension provisions, employer’s contributions to social security plans, representation allowances, living expenses and other perquisites. In this study, the dependent variable executive compensation comprises only the cash compo-nents base salary and the annual bonus. Although the vast majority of companies announced whether they operate long-term incentive schemes, they did not provide sufficient information to estimate the value of these incentives using the Black-Scholes model. Therefore, a dummy variable is included that equals one if the executive received any long-term incentive payments.

4.1.2. Explanatory variables Firm performance

Consistent with the literature, the explanatory variable firm performance is measured in several ways. First, following Zhou (2000) the standard accounting measures of firm performance return on equity (ROE) and return on assets (ROA) will be used6_{. Both contemporaneous and lagged accounting} perform-ance measures are included, as previous studies report a significant influence of the firm’s one year lagged performance on the current level of compensation7_{. Although capital market-based performance measures} like total shareholder return are commonly used proxies for performance, this study does not include these capital market-based measures as, given the limited time, obtaining this data appeared to be too laborious. However, it has been suggested that some accounting performance measures are highly corre-lated with stock returns (e.g. Landsman & Shapiro, 1995). Another measure of firm performance em-ployed in this study is Tobin’s Q, a hybrid measure of accounting and market-based performance (Duff-hues & Kabir, 2008). Following previous studies, Tobin’s Q is measured as the market value of common stock plus the book value of total debt divided by the book value of total assets at the end of the fiscal year. The market value of common stock, or market capitalization, is calculated by multiplying the number of shares outstanding by the fiscal year-end market price of a share.

6_{ROE =} 2 / ) equity s r' shareholde equity s er' (sharehold holders equity to le attributab income tax after 1 t t + − − ROA = 2 / ) assets total assets (total holders equity to le attributab income tax after 1 t t + − −

(20)

Anglo-American exchange listing

Following Oxelheim & Randøy (2005), a dummy variable is used that equal 1 if the firm is listed on an Anglo-American stock exchange (New York Stock Exchange, NASDAQ, London Stock Exchange), and 0 otherwise. Needless to say, this variable will only be included for the sub sample of Continental Euro-pean companies. Of the sample of 142 Continental EuroEuro-pean companies, 49 companies were listed on an Anglo-American stock exchange, or 34.5% of the sub sample firms are listed on an Anglo-American stock exchange.

Anglo-American board membership

The variable Anglo-American board membership is assigned a value of 1 if the executive is born in a country with a common law legal system (e.g. United States, United Kingdom, Ireland, South Africa, Aus-tralia, New-Zealand, Hong-Kong, Singapore, among others)8_{. Like the explanatory variable Anglo-American} exchange listing, the variable Anglo-American board membership is only applicable to Continental European companies. Of all the 581 individuals who are executive of a Continental European company, 34 persons, or approximately 5.9% of the sample, conform to the definition of an Anglo-American board member. I will include another dummy variable that equals 1 if one or more executive members are born in a country with a common law legal system, as was suggested by Oxelheim & Randøy (2005). Using this latter defini-tion, 27 of the 142 firms, or 19% have at least one Anglo-American member on the executive board.

4.1.3. Control variables

To control for firm characteristics, governance characteristics, and individual executive characteristics, several variables will be included in the regression model.

Firm size

As mentioned earlier in the literature review, the extant empirical studies found that firm size is signifi-cantly and positively correlated with executive compensation. Several theories try to explain this result. For example, it is theorized that larger firms are more complex and therefore require specific managerial skills. Because the pool of capable managers that have these skills is limited, the managers should be paid higher levels of compensation. Moreover, managing a larger company generally means a manager has greater responsibility and needs to complete more complex tasks and decisions (Doucouliagos et al., 2007). It has also been argued that large companies have a greater number of executive levels, and each level has a salary differential, which leads to a higher pay level for senior executives (Simon, 1957). Consis-tent with Duffhues & Kabir (2008), Merhebi et al. (2006), and Zhou (2000), among others, the natural logarithm of sales is used as a proxy for firm size. Brunello et al. (2001) use the number of employees as a proxy for size. The authors argue that the number of employees is less likely to suffer from collinearity

8

(21)

problems than sales. Therefore, I will also include the natural logarithm of the number of employees as a proxy for size.

Leverage

Given the widely accepted proposition that monitoring by debt-holders reduces the agency problem, I expect higher debt levels to be negatively related to compensation. However, Duffhues & Kabir (2008) suggest that this relationship may be positive, given the fact that higher debt levels lead to increases in the firm’s risk, which induces higher payments to the executives. Leverage is operationalized as the ratio of total debt to total assets at fiscal year-end, for both contemporaneous and lagged terms.

Board structure

Scant attention has been paid to the effects of board structure on executive compensation. One could expect a negative relationship between board size and executive compensation, as a larger board can di-vide tasks. Hence, each director can focus on a specific monitoring task and this facilitates a high level of monitoring. However, it is often asserted that at a certain point, a large board of directors is less effective in monitoring the executive board and these large boards allow higher levels of compensation. This as-sumption is confirmed by several studies (Core et al., 1999; Randøy & Nielsen, 2002; Ozkan, 2007) who all find that larger boards pay their executives higher salaries. Core et al. (1999) found that an increase in board size of one member corresponds to a $ 30,601 increase in total CEO compensation. To allow for this non-linear relationship, I will also include a squared board size variable.

Compensation committee

Another control variable considered in this study is the compensation committee, as this committee is responsible for determining the amounts and types of compensation paid to the executive directors. Ac-cording to Conyon & Peck (1998) argue that in the absence of a compensation committee, there exists an opportunity for executives to award themselves exorbitant compensation packages. Therefore, the pres-ence of a compensation committee is expected to be negatively associated with compensation. Although prior studies often use a dummy to indicate whether a company has established a compensation commit-tee, I will use the ratio of compensation committee members to total board members as a proxy for com-pensation committee, since almost all companies of the sample do have a comcom-pensation committee.

Blockholders

(22)

Age

Several studies investigate the link between the executive’s age and the level of compensation (e.g. McKnight & Tomkins, 2004 and Doucouliagos, 2007). According to McKnight & Tomkins (2004), ex-perience and knowledge increases with the executives’ age. It is assumed that the managerial labor market will value the increased capabilities, inducing higher levels of compensation. However, previous studies found mixed results on the relationship between compensation and age. An explanation for these mixed results is that the relationship is not a linear one, as most researchers assume, but the association is non-linear, as suggested by Doucouliagos (2007). This author assumes that older executives might be less ef-fective and consequently pay may be negatively related to age. Therefore, I will also include a squared age variable to allow for a non-linear relationship between age and compensation.

Tenure

In a similar vein, tenure is expected to be positively related to compensation. Over the length of their tenure, the executive develops skills and becomes more experienced which results in higher productivity, justifying higher levels of compensation. Moreover, Hill and Phan (1991) posit that over the course of their tenure, chief executives become more entrenched and accumulate more power over the supervisory board or board of directors. Since new board members in general are appointed by the executives, these new board members owe their position to the executives and therefore …“their loyalty may be to the executives rather than to the stockholders”. Therefore, through their power over the board, the executives are able to influence their compensation package. Prior research has yielded inconsistent results with respect to the linkage between tenure and the level of compensation. For example, Hambrick & Finkelstein (1995), who studied CEO compensation, found no association between the length of tenure and the change in pay. On the other hand, Johnston (2002) documented a positive relationship between tenure and compensation. McKnight & Tomkins (2004) found no direct relation between tenure and pay, though tenure appeared to increase the relationship between compensation and performance. Tenure is measured as the number of years the executive is in its current positions as of the end of the fiscal year.

4.1.4. Sample characteristics

Table 2 presents an overview of the descriptive statistics of the variables used in this study. Panel A pre-sents the descriptive statistics of the dependent variables on compensation data, panel B prepre-sents some firm characteristics and panel C and D show the explanatory and control variables respectively. The com-pensation figures in panel A and company figures in panel B are expressed in Euros. Total cash compen-sation for the sample of 906 executives is on average € 1,184,448 (median € 990,338). As can be expected, CEO’s receive higher cash compensation than the average executive director, as the average total cash compensation for a CEO is € 1,597,461 (median € 1,357,316). Panel B shows that the firm characteristics

(23)

are skewed to the right, as the average figures differ substantially from the median figures. This implies that a few firms present very high values for the expressed firm characteristics.

Table 2

Descriptive statistics

Variables N Mean Median Min Max St. deviation

Panel A: compensation variables

Total compensation in € (all executives) 906 1.184.448 990.338 36.600 8.430.944 901.173

Base salary in € (all executives) 906 562.466 495.850 0 4.453.144 315.550

Annual bonus in € (all executives) 906 621.982 426.900 0 7.233.750 689.285

CEO total compensation in € 238 1.597.461 1.357.316 100.000 8.383.750 1.098.776

CEO base salary in € 238 749.858 709.500 0 2.121.347 360.313

CEO annual bonus in € 238 847.603 634.000 0 7.233.750 869.017

Panel B: firm characteristics

Total assets (mln €) 228 81.675 6.201 24 1.351.401 236.625

Sales (mln €) 228 13.556 5.988 2 200.535 21.825

Market capitalization (mln €) 227 14.837 5.756 74 186.940 24.957

Employees 228 49.695 20.415 5 502.545 77.644

Panel C: explanatory variables

ROA 228 0,058 0,047 -0,177 0,516 0,067 ROA t-1 227 0,023 0,040 -4,823 0,628 0,332 ROE 228 0,242 0,176 -0,231 4,770 0,439 ROE t-1 227 0,202 0,137 -6,775 15,179 1,179 Tobin's Q 228 1,646 1,340 0,833 6,860 0,933 Tobin's Q t-1 223 1,495 1,260 0,789 5,865 0,766

Anglo-American listing (dummy) 142 0,345 - - - 0,477

Anglo-American board member (dummy) 581 0,060 - - - 0,238

At least 1 Anglo-American board member (dummy) 142 0,190 - - - 0,394

Panel D: control variables

Leverage 228 0,668 0,679 0,136 1,094 0,193

Leverage t-1 228 0,677 0,694 0,183 1,188 0,195

Board size 228 11,6 11,0 3,0 21,0 5,1

Compensation committee 213 0,365 0,333 0,105 1,000 0,169

Blockholders (dummy) 225 0,764 - - - 0,425

Long-term incentive plan (dummy) 888 0,780 - - - 0,414

Age 874 52,1 52,0 33,0 71,0 7,0

Tenure 888 5,4 4,0 0,2 32,0 4,9

Industry

(24)

4.2. Methodology

Following the literature (for example Oxelheim & Randøy, 2005; Duffhues & Kabir, 2008; Ozkan 2007) the ordinary least-squares regression method will be used to test the hypotheses developed in chapter 2. The regressions will be run using the EViews econometrics package.

4.2.1. Regression model

To test the hypothesis that firm executive compensation is positively related to firm performance (H1), the following equation will be estimated:

(1)

The dependent variable Ln(Comp)it is the amount of total cash compensation paid to executive directors of firm i in period t. The explanatory variable will be proxied by return on assets, return on equity and Tobin’s Q, as was stated in subsection 4.1.2. In addition to the explanatory variable (Perf), the following control variables are added. Firm size (Size) proxied by the natural logarithm of sales and the number of employees, leverage (Lev), proxied by the ratio of total debt to total assets, the number of board members (Board), a non-linear measure for the number of board members (Board2_{), compensation committee (CC),} measured as the ratio of compensation committee members to total board members, a dummy for the presence of blockholders (Blockholder), the age of the executive (Age), a non-linear measure for the ex-ecutive’s age (Age2_{), number of years the executive is in his current position (Tenure), a dummy for the} presence of a long term incentive scheme (LTIS), and a dummy to indicate whether the firm is incorpo-rated in a common law country (ComLaw).

To test whether there is a positive relationship between a listing on an Anglo-American stock exchange and executive compensation in non-Anglo-American firms (H2), the following regression will be esti-mated:

(2)

The explanatory variable (Listing) is a dummy that indicates whether a firm has a listing on an Anglo-American stock exchange. Since the hypothesis can only be tested for firms incorporated in non-Anglo-American countries, the dummy ComLaw is omitted.

To examine hypotheses 3 and 4, an Anglo-American executive receives a higher salary than other execu-tives and having at least one Anglo-American on the executive board positively influences the level of cash compensation of the other executive team members, the following equations will be estimated:

(25)

(3)

The explanatory variable (AngloNat) is a dummy variable equal to one if the board member is of Anglo-American origin.

(4)

The explanatory variable (OneAnglo) is a dummy variable that equals one if an executive has at least one fellow executive team member of Anglo-American origin.

4.2.2. The assumptions for Ordinary Least Squares estimation

Before testing the equations stated above, it is important to consider the assumptions of the OLS regres-sion method, as violating these assumption might produce poor results. The first assumption is that the mean value of the errors is zero, and this can be accomplished by including a constant term in the regres-sion equation, as is done in equation 1 – 3 (Brooks, 2002). Another assumption of the OLS regresregres-sion model is that there is no heteroskedasticity. Heteroskedasticity is present when the variance of the errors is not constant. Heteroskedasticity does not affect the parameter estimates and therefore the estimated coef-ficients should be unbiased when heteroskedasticity is present. However, heteroskedasticity biases the variance of the estimated parameters and hence the t-statistics. Consequently, this causes difficulties in determining whether or not the estimated coefficients are statistically significant. I tested for and found heteroskedasticity using White’s heteroskedasticity test. Therefore, to avoid the aforementioned problems, heteroskedasticity-consistent standard error estimators will be used in the OLS regressions (Brooks, 2002). Furthermore, the OLS regression method assumes that there is no multicollinearity, that is, the explana-tory variables should not be correlated with one another. Whenever multicollinearity is present and hence variables are highly correlated, the variables convey the same information and the p-values can be mislead-ing. Moreover, the regression becomes very sensitive to small changes in the specification and confidence intervals can be very wide resulting in inappropriate conclusions9_{. To assess the possibility of} multicollin-earity, a Pearson correlation test is performed and the results are presented in table 3. Since some variables had a non-normal distribution, a Spearman’s rho correlation test was performed as well. However, testing the non-normally distributed variables with Spearman’s rho yielded similar results. For the sake of brevity, the results for Spearman’s rho test are not reported. Although no specific cutoff point is defined, multicol-linearity is typically not problematic until correlations exceed 0.75.

9_{Brooks, Introductory econometrics for finance, page 191-192.}

(26)

C H A P T E R 4: D A T A A N D M E T H O D O L O G Y 22 Table 3

Pearson Correlation Matrix

Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

1. Ln (Total compensation)

2. Ln (Base salary) 0,863 ** 3. Ln (Annual bonus) 0,929 ** 0,660 ** 4. Long-term incentive plan (dummy) 0,308 ** 0,304 ** 0,276 ** 5. ROA -0,126 ** -0,145 ** -0,116 ** -0,049 6. ROE 0,059 0,065 0,058 0,055 0,364 ** 7. Tobin's Q -0,085 * -0,160 ** -0,006 0,090 ** 0,514 ** 0,407 ** 8. ROA t-1 0,039 0,032 0,049 0,058 0,293 ** 0,103 ** 0,180 ** 9. ROE t-1 0,083 * 0,077 * 0,081 * 0,059 0,206 ** 0,676 ** 0,289 ** 0,355 ** 10. Tobin's Q t-1 0,040 -0,046 0,048 0,141 ** 0,531 ** 0,458 ** 0,886 ** 0,212 ** 0,349 **

11. Anglo-American listing (dummy) 0,380 ** 0,277 ** 0,369 ** 0,327 ** -0,083 * -0,060 -0,020 0,035 0,074 0,056 12. Anglo-American board member (dummy) 0,069 0,098 * 0,042 0,046 0,026 0,132 ** 0,070 0,062 -0,002 0,092 * 0,081 13. At least 1 Anglo-American board member (dummy) 0,090 * 0,089 * 0,080 0,152 ** -0,016 0,169 ** 0,053 0,063 0,089 * 0,064 0,200 **

(27)

It can be observed from table 3 that there is a high correlation between the natural logarithm of total compensation and the natural logarithm of base salary and the annual bonus. This is not surprising, as the variable total compensation is the sum of base salary and the annual bonus. Moreover, a high correlation is detected between the variables Tobin’s Q and Tobin’s Qt – 1, ln (sales) and ln (employees), and leverage and leveraget – 1. Therefore, these variables should not be included in the same regression estimation. Test-ing for multicollinearity usTest-ing VIF (Variance Inflator Factor) in SPSS, the non-linear variables age² and board size² were highly correlated with age and board size, respectively. However, as they should be in-cluded in the same regression equation, these variables will first be regressed on each other.10_The residu-als of this equation represent the change in the dependent variable not caused by the independent variable. These residuals are then utilized in the initial equation to replace one of the variables and remove the cor-relation between the two correlated variables11 ₁₂_.

Furthermore, the OLS estimation technique requires that there is no correlation between the error terms, which implies that the covariance between errors is zero. The Durbin-Watson test is performed to assess whether autocorrelation is present. The results fall well within the acceptable range and hence the data fit the assumption of no autocorrelation. Finally, the assumption is made that residuals are normally distributed. The Jarque-Bera normality test reveals that this assumption poses no problems for the database used for this thesis.

10_{I would like to thank my fellow student A.K. Mensink for suggesting this alternative specification.} 11_{The following initial equation will be tested:}_y _α _α _x _α _z₊_ε

2 + 1 + 0

= . However, explanatory variables x and z are characterized by a high correlation. To remove this correlation, the following equation must be estimated: z=β₀+β₁x+ν. The residual v represents the change in z not caused by x. This residual is then used to replace z in the initial equation. This equation then becomes: y=α₀+α₁x+α₂v+ε.

12_{Regressions were also run with the unmodified variebles Age, Age², Board size, and Board size². This yielded results at similar significance levels}

(28)

CHAPTER 5:RESULTS 24

5. RESULTS

Table 4 presents OLS estimation results for total cash compensation. The models 1 – 4 present the results for hypothesis 1 – 4 respectively. These models examine whether the explanatory variables individually can explain the executive compensation. Model 5 presents the results for the regression estimation that includes all explanatory variables, except the variable ‘At least 1 Anglo-American executive. Including this variable in model 5 would cause a near singular matrix error. Due to the fact that variables capturing the Anglo-American influences are included, model 5 only comprises data for non-Anglo-American firms.

5.1. Regression analysis

The regression estimate for model 1 indicates that, as was hypothesized, total compensation is signifi-cantly positively related to firm performance, measured by Tobin’s Q, at the 5% significance level. The collective model – model 5 – also finds a highly significant and positive relationship between compensa-tion and Tobin’s Q, though this result is drawn from a smaller sample. I only report the results for Tobin’s Q as proxy for firm performance, as this performance measure yields the most significant results. The performance variables return on assets and return on equity were positively though insignificantly related to compensation, as is presented in the Appendix, table A2. Similarly, I included the natural logarithm of sales as a proxy for firm size and did not include the natural logarithm of the number of employees. As suggested in hypothesis 2, model 2 shows that there is a positive and significant association between Anglo-American stock exchange listing and executive cash compensation. This result also applies to the collective model for the same significance level, namely the 5% level.

Table 4

Regression results for total cash compensation

Model 1 2 3 4 5

Explanatory variables

Tobin's Q 0.097** (3.159) 0.179*** (5.419)

Anglo-American listing 0.161** (2.875) 0.135** (2.491)

Anglo-American executive 0.319*** (3.308) 0.283*** (2.891)

At least 1 Anglo-American executive 0.027 (0.431)

Control variables Ln(sales) 0.223***(15.162) 0.183*** (8.112) 0.196*** (9.119) 0.191*** (8.546) 0.198*** (9.041) Leverage 0.073 (0.613) -0.048 (-0.259) -0.134 (-0.714) -0.116 (-0.571) 0.214 (1.192) Board size 0.035*** (6.460) 0.031*** (5.380) 0.034*** (6.002) 0.035*** (5.746) 0.035*** (6.175) Board size2 _-0.002_** _(-2.074) _-0.001 _(-1.127) _-0.002 _(-1.561) _-0.001 _(-1.446) _-0.001 _(-0.536) Compensation committee 0.272* (1.752) 0.052 (0.295) 0.098 (0.557) 0.093 (0.541) 0.149 (0.809) Blockholders -0.109** (-2.082) 0.001 (0.020) -0.061 (-0.833) -0.027 (-0.351) -0.033 (-0.451) Age 0.002 (0.553) -0.004 (-1.000) -0.004 (-0.819) -0.004 (-0.901) -0.004 (-0.916) Age2 0.000 (-0.328) 0.000 (0.267) 0.000 (0.225) 0.000 (-0.027) 0.000 (0.266) Tenure 0.017*** (4.265) 0.025*** (5.121) 0.025*** (5.004) 0.024*** (4.844) 0.024*** (5.052) Long term incentive scheme -0.031 (-0.494) 0.008 (0.109) 0.018 (0.226) 0.037 (0.449) -0.054 (-0.710) Common law country 0.040 (0.907)

Intercept 9.492***(32.752) 10.551***(29.301) 10.397***(29.305) 10.440***(28.186) 9.842***(26.564)

Number of observations 787 488 488 457 488

Adjusted R2 0.467 0.423 0.427 0.424 0.460

F-statistic 58.31 33.51 33.93 31.54 32.91

(29)

According to the results for model 3, being an executive of Anglo-American origin working for a non-Anglo-American firm positively influences cash compensation. This expected positive relationship also holds for the collective model at a significance level of 1%.

The results for model 4 indicate that having an Anglo-American executive on the management board does not significantly influence the compensation level of the other individual non-Anglo-American manage-ment board members. Though not reported in table 4, adding the other explanatory variables Tobin’s Q and Anglo-American listing yields similar results for the variable ‘At least 1 Anglo-American executive’. Hence, I find no support for hypothesis 4. That is, the results do not confirm the presence of the hy-pothesized “spillover effects”. Therefore, although non-American firms pay executives of Anglo-American origin higher compensation levels, they are not willing to pay their fellow continental European executives higher salaries.

The results in table 4 also show that there is a positive association between total cash compensation and the control variables firm size, board size, and tenure. Moreover, these control variables were significant at the 1% level and these results hold for all of the five models. The relationship between firm size and compensation is even stronger than the link between performance and compensation, a result that is con-sistent with prior research (Tosi et al., 2000) Although the significant results for firm size and tenure are straightforward and consistent with prior research, the relationship between board size and compensation was supposed to be more complicated. That is, it was suggested that monitoring by the board would ini-tially improve as the board comprises more members, though at a certain size, monitoring activities would diminish. Therefore, it was expected that the non-linear variable Board2_{would exhibit a significant} rela-tionship but this is only the case for model 1. This can be explained by the fact that contrary to model 2 – 5, model 1 includes British and Irish firms, and these firms have one-tier boards that comprise generally a larger number of members. Nevertheless, the result for the variable ‘Board’ is consistent with prior re-search (for example Ozkan, 2007).

As was expected, the coefficient of the dummy variable for blockholders is negatively and significant for model 1 at the 5% level. For models 3-5, the presence of blockholders had a negative though insignificant influence on executive compensation. The regression results for model 2 even exhibit a positive relation-ship, although the magnitude of the coefficient is low.

Contrary to the expectations, having a pay package that contains a long term component does not have a negative impact on the executive’s cash compensation. The signs of the coefficient of the LTIS dummy differed across the models, but none of the models provides significant results. It could therefore be ar-gued that executives obtain long term incentive packages without giving up corresponding amount of cash compensation, as is suggested by Bebchuk & Fried (2003).

(30)

5.2 Robustness checks

In order to test the robustness of the results, I will replicate the regressions using alternative specifica-tions. First, the sample is divided into a subsample of chief executive officers and a subsample of execu-tive directors who are not a CEO. Table 5 shows the results for the CEOs and other directors in panel A and panel B, respectively. It can be observed from table 5 that, overall, the relationship between the ex-planatory variables and compensation is somewhat stronger for the subsample of other directors than it is for the subsample of CEOs.

Table 5

Regression results for subsamples Chief Executive Officers and Other Executive Directors

Model 1 2 3 4 5

Panel A: Chief executive officers Explanatory variables

Tobin's Q 0,133** (2.591) 0,167*** (3.159)

Anglo-American listing 0,146* (1.697) 0,135 (1.578)

Anglo-American executive 0,411** (3.033) 0,393*** (2.653)

At least 1 Anglo-American executive 0,143 (1.349)

Control variables Ln(sales) 0,253***(10.233) 0,193*** (4.950) 0,194*** (5.061) 0,196*** (5.065) 0,199*** (5.416) Leverage 0,041 (0.189) -0,101 (-0.287) -0,131 (-0.368) -0,240 (-0.636) 0,180 (0.619) Board size 0,031*** (3.326) 0,025*** (2.632) 0,030*** (3.403) 0,032*** (3.541) 0,031*** (3.350) Board size2 -0,003 (-1.601) -0,001 (-0.629) -0,002 (-1.013) -0,001 (-0.769) -0,001 (-0.507) Compensation committee 0,023 (0.105) -0,248 (-1.004) -0,185 (-0.778) -0,141 (-0.570) -0,126 (-0.549) Blockholders -0,153* (-1.864) -0,173 (-1.507) -0,233** (-1.987) -0,213* (-1.803) -0,214* (-1.833) Age 0,003 (0.513) -0,006 (-0.787) -0,004 (-0.623) -0,005 (-0.666) -0,005 (-0.819) Age2 0,000 (-0.074) 0,001 (1.532) 0,001 (1.444) 0,001 (1.391) 0,001 (1.138) Tenure 0,012** (2.589) 0,012* (1.963) 0,012* (1.892) 0,013** (1.991) 0,013** (2.106) Long term incentive scheme 0,035 (0.323) 0,076 (0.547) 0,099 (0.700) 0,071 (0.505) 0,029 (0.237) Common law country 0,027 (0.382)

Intercept 9,465***(18.257) 11,215***(17.824) 11,148***(18.301) 11,147***(17.749) 10,599***(16.780)

Adjusted R2 0,619 0,521 0,527 0,522 0,554

F-statistic 26,94 13,66 13,96 13,23 13,24

Panel B: Other executive directors Explanatory variables

Tobin's Q 0,085** (2.468) 0,178*** (4.977)

Anglo-American listing 0,209*** (3.289) 0,174*** (2.853)

Anglo-American executive 0,380*** (3.594) 0,339*** (3.180)

Control variables Ln(sales) 0,215***(13.827) 0,181*** (7.036) 0,203*** (8.383) 0,197*** (7.818) 0,202*** (8.140) Leverage 0,095 (0.753) -0,005 (-0.025) -0,140 (-0.663) -0,092 (-0.403) 0,195 (0.998) Board size 0,045*** (7.457) 0,040*** (6.357) 0,043*** (6.795) 0,042*** (6.263) 0,044 (6.957) Board size2 _-0,002_* _(-1.837) _-0,001 _(-0.901) _-0,001 _(-1.314) _-0,001 _(-0.994) _0,000_***_(-0.167) Compensation committee 0,361* (1.923) 0,188 (0.833) 0,231 (1.038) 0,193 (0.844) 0,258 (1.107) Blockholders -0,098* (-1.753) 0,070 (0.889) -0,010 (-0.128) 0,031 (0.388) 0,043 (0.567) Age -0,004 (-1.054) -0,009* (-1.804) -0,008 (-1.646) -0,008 (-1.581) -0,008* (-1.743) Age2 0,000 (-0.225) 0,000 (0.407) 0,000 (0.465) 0,000 (0.173) 0,000 (0.722) Tenure 0,012*** (2.660) 0,024*** (4.197) 0,024*** (4.031) 0,022*** (3.760) 0,022*** (4.139) Long term incentive scheme -0,034 (-0.465) 0,008 (0.096) 0,011 (0.132) 0,046 (0.520) -0,071 (-0.841) Common law country 0,016 (0.335)

Intercept 9,638***(31.044) 10,436***(27.085) 10,225***(26.900) 10,241***(25.870) 9,668 (23.958)

Adjusted R2 0,513 0,476 0,481 0,479 0,520

F-statistic 51,39 30,59 31,11 28,8 30,88

(31)

For all models, the signs of the coefficients are similar to those found for the full sample presented in table 4. Furthermore, the magnitudes of the coefficients agree well with those obtained for the full sample, and this especially applies to the subsample of other executives presented in panel B.

To further check whether results are robust, industry dummies are added. Two-digit SIC industry codes are retrieved from Amadeus. The SIC (Standard Industrial Classification) classification is well accepted and widely applied system for categorizing companies into industry groups (Porac et al. 1999). Initially, the sample was divided into eight groups. However, some subsamples included only a few firms, which could lead to severe errors using regression analysis. Therefore, the sample is divided into three industry groups, namely manufacturing firms, financial institutions, and other firms. This distinction is made as a large part of the sample consists of manufacturers and therefore could be well classified. Financial companies are classified as a separate group since researchers often exclude these firms because their financial character-istics differ from other firms and because these firms are subject to more stringent government regulation which might affect the structure of executive compensation (Firth et al., 2007).

The results presented in table 6 reveal that controlling for different industries yields results that are consis-tent with those obtained earlier. With respect to the explanatory, as well as the control variables, the re-gressions produce signs and magnitudes of coefficients similar to those presented in table 4.

Table 6

Regression results for total cash compensation with industry dummies

Model 1 2 3 4 5

Explanatory variables

Tobin's Q 0,116** (3.862) 0,187*** (5.426)

Anglo-American listing 0,158*** (2.699) 0,113** (1.993)

Anglo-American executive 0,321*** (3.298) 0,287*** (2.911)

Control variables Ln(sales) 0,239***(15.272) 0,187*** (7.261) 0,205*** (8.475) 0,201*** (7.808) 0,210*** (8.243) Leverage -0,141 (-1.167) -0,101 (-0.493) -0,228 (-1.098) -0,212 (-0.970) 0,120 (0.637) Board size 0,033*** (5.884) 0,030*** (5.184) 0,032*** (5.614) 0,033*** (5.242) 0,033*** (5.808) Board size2 -0,002** (-1.781) -0,001 (-1.111) -0,001 (-1.399) -0,001 (-1.305) 0,000 (-0.417) Compensation committee 0,269* (1.732) 0,054 (0.310) 0,082 (0.480) 0,073 (0.415) 0,137 (0.753) Blockholders -0,087** (-1.645) 0,001 (0.017) -0,054 (-0.734) -0,016 (-0.216) -0,033 (-0.453) Age 0,002 (0.680) -0,005 (-1.017) -0,004 (-0.860) -0,004 (-0.918) -0,004 (-0.902) Age2 _0,000 _(0.016) _0,000 _(0.310) _0,000 _(0.348) _0,000 _(0.120) _0,000 _(0.401) Tenure 0,017*** (4.513) 0,025*** (5.126) 0,025*** (5.068) 0,025*** (4.927) 0,024*** (5.101) Long term incentive scheme -0,027 (-0.430) 0,009 (0.118) 0,016 (0.200) 0,038 (0.462) -0,056 (-0.732) Common law country 0,045 (1.042)

Intercept 9,320***(31.412) 10,543***(27.667) 10,353***(27.907) 10,356***(26.610) 9,707***(23.868) Control variables Manufacturers -0,020 (-0.501) -0,016 (-0.286) 0,009 (0.160) 0,011 (0.184) 0,025 (0.472) Financial institutions 0,211*** (3.637) 0,028 (0.374) 0,082 (1.115) 0,090 (1.165) 0,103 (1.324) Number of observations 787 488 488 457 488 Adjusted R2 _0,476 _0,421 _0,423 _0,423 _0,460 F-statistic 52,01 28,28 28,75 26,75 28,63