• No results found

M ASTER T HESIS

N/A
N/A
Protected

Academic year: 2021

Share "M ASTER T HESIS"

Copied!
56
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

M

ASTER

T

HESIS

THE EFFECT OF CEO

-

TO

-

WORKER PAY RATIO ON FIRM PERFORMANCE

:

A TOURNAMENT AND EQUITY THEORY PERSPECTIVE

By: Rob Eekhout Student number: 1684035

Course: Master Thesis IB&M Course code: EBM719A20 Supervisor: Dr. K. van Veen Co-assessor: Mr. P.J. Marques Morgado

(2)

ABSTRACT

This study aimed to investigate the effect of the CEO-to-worker pay ratio on firm performance. There are two main theories that can explain this relationship: Tournament Theory and Equity Theory. Tournament Theory hypothesizes that large differences in wages within a firm motivate employees to work hard in order to win a promotion, which increases firm performance, while Equity Theory proposes that large pay differences leads to feelings of inequity and to

demotivation, which decreases firm performance. While there are many studies on this

theoretical framework, this thesis is the first study that uses the CEO-to-worker pay ratio as the predictor for firm performance. To study this relationship, this thesis integrates Tournament Theory and Equity Theory into one framework and hypothesizes an inverse U-shaped

relationship between the CEO-to-worker pay ratio and firm performance, which is moderated by the number of employees in a firm and the power distance of the home country of the firm. However, our statistical analysis of 56 European firms did not show any relationship between CEO-to-worker pay ratio and firm performance, thus casting doubt on our theoretical framework. In our subsequent discussion of the results, we add value to the existing literature by critically discussing the current state of the debate on the theoretical merit of Tournament Theory and Equity Theory.

(3)

Contents

1 Introduction ... 1

2 Theoretical Framework ... 3

2.1 Intra-firm Wage Dispersion ... 3

2.2 Tournament Theory ... 4

2.3 Equity Theory ... 4

2.4 Previous Research ... 6

2.4.1 Positive Linear Relationship ... 6

2.4.2 Negative Linear Relationship ... 8

2.4.3 Inverted U-shaped Relationship ... 9

2.4.4 U-shaped Relationship ... 12

2.4.5 Conclusion of Literature Review ... 15

2.5 Firm Size ... 17 2.6 Power Distance ... 18 2.7 Hypotheses ... 20 2.8 Conceptual Model ... 21 3 Methodology ... 22 3.1 Independent Variables ... 22 3.2 Dependent Variable ... 24 3.3 Control Variables ... 25

3.4 Data Sources and Collection ... 26

(4)
(5)

1 Introduction

Different companies have different levels of intra-firm wage dispersion; in some there are big gaps in salaries between employees and in others wage dispersion is more flat. In the academic world this topic is generally seen as a question of incentives for employees: on the one hand, big gaps in compensation between employees can be an incentive to work hard to win a promotion, but others argue that large differences in compensation lead to feelings of inequity and to reduced morale and motivation (Pfeffer & Langton, 1993). Because of the supposed effect of wage dispersion on employee motivation, differences in relative pay can also lead to differences in firm performance. These arguments roughly correspond to Tournament Theory and Equity Theory. Tournament Theory predicts that large differences in pay within a company lead to better firm performance (Henderson & Fredrickson, 2001), while Equity Theory predicts that such differences in pay will lead to worse performance (Lallemand, Plasman & Rycx, 2004). The existing literature in this field that test these predictions use varying definitions of intra-firm pay dispersion. Whereas some look at the pay differences between similar workers in a firm (e.g. Winter-Ebmer and Zweimüller, 1999), others look at wage differences between

dissimilar workers (e.g. Connelly et al., 2013). Similarly, some studies look at differences in pay between blue-collar workers (e.g., Lallemand, Plasman & Rycx, 2004), while other papers analyze compensation differences between the CEO and the top management team (Henderson & Fredrickson, 2001). Perhaps because of these differing definitions, the results of these studies are mixed, with some studies finding positive relationships between intra-firm wage dispersion and firm performance and others finding a negative relationship, and still others find different shaped relationships or no relationship at all. Although they all purport to study intra-firm wage dispersion, their widely varying measures really show that they are studying quite different phenomena.

(6)

Himmler & Koenig, 2011). However, in spite of the apparent importance of the differential between CEO and worker compensation, research that looks at the effect of the CEO-to-worker pay ratio on firm performance has not been performed yet.

Therefore, to close this gap, our paper will investigate the effect of CEO-to-worker pay on firm performance using Tournament Theory and Equity Theory. In addition, we propose that this relationship is moderated by two other variables. Firstly, we hypothesize that firms with fewer employees are more likely to experience the supposed effects of CEO-to-worker pay ratio on firm performance, because employees will compare themselves to the CEO more easily than in a large firm where the employees never see the CEO. In a large firm, the effect of the CEO salary relative to the average employee may have less impact, because the CEO is such a far-away figure. Secondly, we theorize that the power distance of a firm’s home country moderates the relationship between CEO-to-worker pay ratio and firm performance; in countries with a high power distance, employees are more likely to accept large differences in salary, and thus those firms are less likely to undergo the negative effects of Equity Theory. To study these effects, a self-compiled data set is used which draws the CEO-to-worker pay ratio from the annual reports of companies in addition to more general data like firm performance and firm size, which were extracted from Orbis.

From an academic point of view this study is interesting because it can contribute to the understanding of the effects of CEO compensation. Studies on CEO pay are very common, but our paper can add an original contribution to this body of research by looking at the CEO-to-worker pay differential’s effect on firm performance. Furthermore, the theoretical debate on Tournament Theory and Equity Theory is contentious, and our paper will attempt to integrate both theories in one framework. From a managerial standpoint this research can also be helpful, as it sheds light on the potentially positive or negative effects of the size of the gap between worker and CEO pay on firm performance. Finally, this study can help to inform societal discussions around CEO pay. News articles of CEOs who earn hundreds of times more money than the average worker within their company are common and frequently lead to angry

(7)

This results in the following research question, which will guide this paper:

What is the effect of CEO-to-worker pay differentials on firm performance?

This paper is structured as follows: first, the theoretical background around the issue of intra-firm wage dispersion will be discussed, and the existing literature on this topic will be explored. This chapter will conclude by providing hypotheses regarding the relationship between CEO-to-worker pay differentials and firm performance. Next, the methodology chapter will map out how these hypotheses will be tested. This chapter will include information about the

variables, the sample, data sources, and the statistical methods that will be used to test the hypotheses. Then, in the results chapter, regression assumptions will be examined and the hypotheses will be tested using regression analysis. After that, the discussion chapter will provide analysis of the results of our paper, and situate it in the larger debate in which this paper participates. The discussion chapter will also consider the limitations of our study. Finally, the conclusion brings the entire paper together and finalizes this thesis.

2 Theoretical Framework

This chapter provides an overview of the theories used in this paper and of the existing literature that concerns itself with these theories. It concludes by positing hypotheses based on the

theoretical model that is used in this chapter.

2.1 Intra-firm Wage Dispersion

(8)

equal salaries, and collaborate? The two main theories that are concerned with these issues are Tournament Theory and Equity Theory, which we will explore in the next sections.

2.2 Tournament Theory

Tournament theory was originally proposed by Edward Lazear and Sherwin Rosen in 1981, and suggests that high intra-firm wage dispersion can have a positive effect on the output of workers, as they are motivated to work hard to get a promotion to a position that pays more. As they argue, “the large salaries of executives may provide incentives for all individuals in the firm who, with hard labor, may win one of the coveted top positions” (Lazear and Rosen, 1981, p. 841). One of the prerequisites of this is that there is a wide range of salaries within the firm, so employees have incentives to work hard and to get promoted to one of the better paying jobs. A special case is the CEO salary, which is often significantly higher than that of his or her direct colleagues in the board (O’Reilly, Main & Crystal, 1988). Tournament Theory explains this by viewing the top position in a company as the ultimate prize, which motivates all others to work hard to ultimately aspire to that position (p. 847). In this situation there is a tournament in which the employees in the company compete for the higher paid positions in the company. Thus, according to this theory, a high ratio of CEO-to-worker pay will have a positive influence on firm performance, as workers have a strong incentive to be productive.

Since Lazear and Rosen’s (1981) original paper, Lazear (1989) has argued that

tournaments with too high levels of intra-firm wage dispersion do not just provide incentives for productivity, but also for negative behavior. As the wage levels between job levels rise, the prize in the tournament increases and employees gain incentives to sabotage their colleagues who are competing for the same prize (Lazear, 1989, p. 562; Heyman, 2005, p. 1316; Hunnes, 2009, p. 793). Lazear (1989) thus suggests that the positive effects of intra-firm wage dispersion on firm performance have a limit.

2.3 Equity Theory

(9)

Comparison Theory, which goes back to 1954, when Leon Festinger published an article titled “A Theory of Social Comparison Processes.” This article by Festinger argues that people compare their “opinions and abilities” with other people to judge their own opinions and

abilities. For example, people judge their score on a test not just by the result that they achieved, but primarily by comparing their result to the results of other people. Whereas Social

Comparison Theory was not directly applied to salaries, J. Stacy Adams (1963) used the logic of it to theorize about what employees think of as a fair wage. This theory, dubbed Equity Theory, proposes that whether employees think their salary is fair or not does not just depend on the absolute size of the salary, but rather on its size relative to others. For example, employees may be perfectly happy with their wage until they hear about the higher salary of a colleague.

In the first instance such comparisons are done with colleagues who do similar work, such as direct colleagues. However, comparisons are also made with people at different levels of the organization. When people compare their salary to other levels of the organization this is not done in an absolute sense, because there is an assumption that it is fair that people higher on the corporate ladder earn more than people lower in the organization. Instead, the perceived inputs (such as effort and skill) and outputs (such as pay) are compared, and if the inputs justify the output there is no sense of inequity and thus no problem (Cowherd & Levine 1992, p. 303). However, when this relation between skill and effort on the one hand and pay on the other is significantly different than for oneself, employees can start to experience inequity. Because of the visibility of the CEO and because that is usually the person with the highest remuneration, comparisons between an employee’s wage and that of the CEO are common, and Equity Theory suggests that such comparisons can lead to feelings of inequity if the difference is greater than expected. Such feelings create “tension” in the person who experiences inequity, and the employee will try to alleviate this tension, for example by working less hard (Adams, 1963, p. 427). Research has indeed confirmed that CEO wages that are perceived to be unfair can lower work morale and can subsequently result in reduced work effort (Cornelißen, Himmler & Koenig, 2011). Consequently, Equity Theory predicts that a perceived feeling of inequity with regards to the relation of one’s own salary and that of the CEO will lead to reduced motivation and a decrease in firm performance.

(10)

firm performance. Equity Theory’s predictions do dovetail with Lazear’s 1989 addition to Tournament Theory, when he proposed that too high levels of wage dispersion can lead to

sabotage and consequentially to decreased firm performance. However, the theoretical arguments for this decrease in firm performance are slightly different: whereas Lazear (1989) suggested that this effect comes from the fact that extremely large differences in pay lead to incentives to sabotage others in order to be promoted themselves, Equity Theory posits that this effect comes from feelings of inequity and demoralization. Therefore, Lazear thinks employees become extremely motivated to gain a promotion to the point where they will sabotage other employees who threaten their position, while Equity Theory argues that employees will become very

demotivated from such large inequities.

2.4 Previous Research

Previous studies on the relationship between wage dispersion and firm performance can broadly be divided into four strands, which all hypothesize different kinds of relationships between intra-firm wage dispersion and intra-firm performance. Within these strands it remains critical to also look at the differences in how these studies measure intra-firm wage dispersion.

2.4.1 Positive Linear Relationship

The first strand of the empirical literature are basically tests of Tournament Theory, which state that intra-firm wage dispersion has a positive linear effect on firm performance. This is

(11)

Figure 1: Positive Linear Relationship between Intra-Firm Wage Dispersion and Firm Performance

Leonard (1990) is one of the earliest papers on this as he, among other things, looks at the effect of the variance of managerial pay on firm performance, which he measures as return on equity. His study with survey-based data of firms from the United States concludes that “there is some evidence consistent with a tournament or lottery view of executive compensation,” but the effects are very slim (Leonard 1990, p. 27). Main, O’Reilly & Wade (1993) provides a similar approach as they also study firms from the United States and use variance in pay among managers as their measure of wage dispersion. They too find evidence that greater wage

dispersion leads to better firm performance, corroborating Tournament Theory. Eriksson (1999) joins the previous studies in testing this linear relationship between intra-firm wage dispersion and firm performance. Like the previous studies he uses a measure of wage dispersion among similar workers, as he looks at the differences in pay among executives. His work supports the previous studies, as he also finds evidence for a positive relationship between wage dispersion and firm performance, using a dataset of 210 Danish companies. Heyman (2005), who looks at wage dispersion among similar workers, also finds a positive relationship between intra-firm wage dispersion and firm performance in his dataset with Swedish firms, in accordance with Tournament Theory. Like the others, Heyman (2005) measure intra-firm wage dispersion as pay differences between managers, a measure of similar workers.

Firm p erf o rm an ce

(12)

One of the interesting commonalities in these studies on Tournament Theory is that they measure wage dispersion as the difference in pay between managers, and that none of them look at pay variance among lower level workers (such as blue-collar workers) or between lower-level employees and higher-level employees. This corroborates that there is a common argument that Tournament Theory only applies to the upper echelons of a company. In our paper we contend that this is incorrect and that Tournament Theory also applies to lower-level employees.

2.4.2 Negative Linear Relationship

The second strand of the literature is concerned with Equity Theory and similar theories, which hypothesize a negative linear relationship between intra-firm wage dispersion and firm

performance. This relationship is visualized in the next figure, and is the opposite of the first strand of literature.

Figure 2: Negative Linear Relationship between Intra-Firm Wage Dispersion and Firm Performance

Firm p erf o rm an ce

(13)

An early empirical study on this is Cowherd & Levine (1992). Using a sample of 102 business units from the United States and Europe they find that less wage dispersion between lower-level employees and higher-level managers leads to better firm performance, which they measure as product quality. One of the main contributions of Cowherd & Levine (1992) in this debate is to extend the discussion from the effects of pay differences between direct colleagues (“intraclass pay differential”) to pay differences between lower-level employees and the top management team of an organization (“interclass pay differential”). They point out that there have been significant theoretical contributions that highlight that lower-level employees do compare their salaries to top management, but that there has not been empirical research to test the effects on employee output. In finding that more pay equity leads to higher firm performance, the results of Cowherd & Levine (1992) support Equity Theory.

An unusual study on this is Bloom (1999), entitled “The Performance Effects of Pay Dispersion on Individuals and Organizations.” Whereas most studies use a sample of companies and analyze their economic performance, such as profits or return on equity, Bloom (1999) studies pay dispersion in baseball teams in the Major League Baseball. His analysis of 1644 players from 29 teams reveals that more pay dispersion within a team leads to worse

performance of the players and teams, showing support for Equity Theory. A more conventional study is the one by Siegel & Hambrick (2005), who look at the pay dispersion within top

management teams in high-technology firms. They too conclude, from a sample of 67 firms from the United States, that high intra-firm wage dispersion is detrimental to firm performance.

These first two strands of the literature, which respectively represent Tournament Theory and Equity Theory, thus diametrically oppose each other yet both find empirical evidence for their hypotheses. The other two strands of the literature, which we will now discuss, attempt to reconcile both theories into one framework.

2.4.3 Inverted U-shaped Relationship

(14)

Figure 3: Inverted U-shaped Relationship between Intra-firm Wage Dispersion and Firm Performance

The theoretical justification for this is that as wage dispersion increases, firm performance may increase up to a certain point as per Tournament Theory, until the wage dispersion becomes so large that the effects of Equity Theory start to dominate and decrease firm performance. Such effects have led researchers like Hunnes (2009) to suggest an inverted U-shaped relationship between wage dispersion and firm performance; a low level of wage dispersion does not give enough incentives for workers (Tournament Theory), but a too high level has negative

consequences from Equity Theory. The optimal level for firm performance is therefore theorized to be somewhere in between, as is visualized in the figure.

From Tournament Theory we would expect that low levels of intra-firm wage dispersion lead to lower firm performance due to little motivation to be productive. As wage dispersion increases, an increase in firm performance is also expected as employees gain incentives to participate in the corporate tournament for higher paying jobs. Equity Theory predicts that while employees will tolerate some level of wage dispersion because of different skills and effort, a disproportionately high wage dispersion can be demoralizing and have negative effects on productivity. As Adams argued, “the tension is proportional to the magnitude of inequity

present,” and at higher levels of inequity, employees will be more demoralized (Adams 1963, p. 427). Therefore, an interplay of Tournament Theory and Equity Theory can be expected, where

Firm p erf o rm an ce

(15)

initially a positive effect of increasing intra-firm wage dispersion can be found, as employees have more reason to be productive so as to aspire to higher paying jobs. As wage dispersion increases to higher levels, the effects of Equity Theory become stronger and can be expected to overtake the effects of Tournament Theory, leading to a lower firm performance. This decreasing effect is reinforced by Lazear’s (1989) argument that too high levels of wage dispersion can lead to sabotage among employees who are competing for the same promotion. Thus, an inverse U-shaped relationship, or a hump-U-shaped relationship, is expected between intra-firm wage dispersion and firm performance.

There are also existing empirical studies that propagate this inverted U-shaped

relationship. The earliest paper on this is Winter-Ebmer and Zweimüller (1999), who study the effects of wage differences within the groups of white-collar workers and blue-collar workers in Austria. Although their data is quite limited as they do not have any data on executives or financial performance, their theoretical and methodological model set the stage for future research, including ours. They find that for both white-collar and blue-collar workers, a rise in pay inequality first increases firm performance (which they measure through standardized wages for different workers, as they do not have data on financial performance of the firms), but at a certain level of pay inequality firm performance stops increasing and starts to decrease. They thus find an inverted-U relationship between intra-firm wage dispersion and firm performance, leading them to conclude that “too little inequality is harmful for productivity due to a lack of incentives” but that there is a “fairness constraint” that leads to “less efficient outcomes when inequality becomes too large” (Winter-Ebmer and Zweimüller 1999, p. 570). Although they do not explicitly recognize this, in reality they have created a model that integrates Tournament Theory and Equity Theory.

(16)

that Tournament Theory is dominant for short-term effects and Equity Theory for long-term effects. They find, based on their longitudinal study of 445 firms over 10 years, that high intra-firm wage dispersion initially increases intra-firm performance, but that this effect turns around after a certain time period. As a measure of wage dispersion they use the differential between top management team compensation and average worker compensation, so they use a measure of dissimilar workers.

2.4.4 U-shaped Relationship

The last strand of the literature is the complete opposite of the previous, as it hypothesizes a U-shaped relationship between intra-firm wage dispersion and firm performance. This is visualized in the next figure.

Figure 4: U-shaped Relationship between Intra-firm Wage Dispersion and Firm Performance

This relationship is theorized by Ridge, Aime & White (2015), who measure wage dispersion as the difference in compensation between the CEO and the four other highest compensated

executives. Whereas the previous strand of the literature that we discussed (e.g. Winter-Ebmer & Zweimüller (1999); Lallemand, Plasman & Rycx (2004)) hypothesizes an inverted U-shaped

Firm p erf o rm an ce

(17)

relationship between intra-firm wage dispersion and firm performance when integrating

Tournament Theory and Equity Theory, Ridge, Aime & White (2015) theorize the opposite: a U-shaped relationship between intra-firm wage dispersion and firm performance. They suggest that at low levels of pay dispersion Equity Theory (which they call “Social comparison”) provides benefits to performance by promoting collaboration and harmony, and at high levels of pay dispersion Tournament Theory results in better performance through large incentives for increased productivity. Because their theoretical model is the complete opposite of these earlier studies by Winter-Ebmer & Zweimüller (1999) and Lallemand, Plasman & Rycx (2004), one would hope they would discuss the reasons why they choose a different theoretical model with such radically different hypotheses. However, they do not reference these papers at all, even though they are widely cited and discussed in other studies on intra-firm wage dispersion. This is disappointing, because it would be interesting to see this theoretical discussion in a field which has such different results.

(18)

This theoretical model by Ridge, Aime & White (2015) is a direct challenge to the model used by earlier authors who attempted to integrate Tournament Theory and Equity Theory, and its theoretical underpinnings thus deserve attention. If we look at part (a) of figure 5, one can see two different possible U-relationships: a regular U-shape and an inverted U-shape. In part (b) of figure 5, the authors choose the regular U-shaped relationship as their hypothesis, rather than the bottom part of figure 5 (a) that reveals a potential inverse U-shaped relationship. The authors justify this with two arguments.

Firstly, they argue that Tournament Theory only provides predictions at very high levels of pay disparity and not at medium to low levels. They state, by quoting Henderson &

Frederickson (2001), that “tournament theory was developed to ‘explain the very large gaps typically observed between the pay of CEOs and the pay of executives directly below them.’” While it is certainly true that the original paper by Lazear and Rosen (1981) highlights this situation as a good example of how Tournament Theory works, it is not true that their theory is aimed only at the differences between CEOs and top management teams, as can be deduced from Lazear and Rosen’s argument that “the large salaries of executives may provide incentives for all individuals in the firm who, with hard labor, may win one of the coveted top positions” (p. 841; emphasis added). Neither do Lazear and Rosen argue that their theory only works at very high levels of pay disparity, and that the firm performance effects do not differ between low and medium pay dispersion. This is supported by much empirical research on Tournament Theory, which does find differences between low and medium pay disparity, even if high pay disparity may have larger effects (e.g., Eriksson, 1999; Main, O’Reilly & Wade, 1993). The theoretical argument made by Ridge, Aime & White (2015) that predictions from Tournament Theory only work at the very high level of pay disparity can thus not be supported.

(19)

much higher salary, for the sole reason that that person earns a lot more. And indeed, the theoretical literature on Equity Theory does not state that people only compare themselves with people who have a similar salary. Rather, people generally look at the ratio of the inputs (such as effort and skill) and the salary that person gets and compare that to their own, which means that people do compare themselves to dissimilar others (Adams, 1964; Akerlof & Yellen, 1990). In addition, Cornelißen, Himmler & Koenig (2010) have shown that excessive CEO pay does have a negative influence on work morale, and Wade O’Reilly & Pollock (2006) have found evidence that “CEOs serve as a key referent for employees in determining whether their own situation is ‘fair’”. Therefore, it has to be concluded that Ridge, Aime & White’s argument that people only compare themselves to those who earn similar wages is implausible.

Because of these reasons their theoretical model, which is the exact opposite of the model used in our paper, can be put to the side for now. In spite of this, they did find empirical support for their model, as they concluded from analysis of their dataset of 227 firms from the Fortune 500 that “high firm performance is found around meaningfully low or meaningfully high levels of pay disparity” (p.630). Nevertheless, there are some differences between Ridge, Aime & White (2015) and our paper which lead us to adopt a different theoretical model. Whereas Ridge, Aime & White use the difference between CEO and top management team compensation as a measure, our paper uses the difference between CEO and median worker compensation as a measure. These measures are used for quite different constructs: pay disparity among similar workers and pay disparity among dissimilar workers. For that reason we can, at least for now, use a different theoretical model.

2.4.5 Conclusion of Literature Review

In concluding the literature review, one can see a few things standing out. Firstly, a lot of different measures are used for intra-firm wage dispersion, which actually measure two very different concepts. To illustrate this, one can say that the difference in pay between direct

colleagues or between a secretary and the CEO are completely different things that should not be treated as two measures of the same concept. Perhaps because of these very different

measurements, the results of the studies are also very mixed. We have distinguished four

(20)

intra-firm wage dispersion exactly influences firm performance, using different measures of wage dispersion.

However, there is one measurement of intra-firm wage dispersion which dovetails very well with Tournament Theory and Equity Theory, but which has not been studied yet: the differential between CEO pay and average worker pay. The CEO is the most public figure of the company and their salary is often public as well, making it an obvious object of comparison for workers throughout the company. On the one hand the CEO’s salary can work as an incentive for the other workers. That is not to say that all workers aspire to become the CEO, but the CEO’s salary can have a symbolic function to signal that successful people in the company are

compensated very well. This is how, through Tournament Theory, one can expect that a high CEO-to-worker pay ratio can lead to increased performance, as workers have incentives to work hard. On the other hand, a relatively high CEO salary can also lead to feelings of inequity, as Equity Theory suggests. Therefore, from Equity Theory we would expect that very high levels of CEO-to-worker pay differences can lead to demotivation and decreased firm performance.

The paper that comes closest to our measure of intra-firm wage dispersion is Connelly et al. (2013), who measure the difference between Top Management Team pay and average worker pay. However, theoretically, the CEO as a public symbol has much more power as an object for comparison than the entire TMT. One can thus question whether Connelly et al.’s (2013) approach is sound, as the TMT does not generally fulfil the public role that the CEO does. This research gap, with a lack of studies on CEO-to-worker pay dispersion, is what our paper

addresses.

(21)

focusing on CEO compensation, also tangentially touches upon the discussion in this more general field on CEO compensation.

Apart from looking at the CEO-to-worker pay, we wish to add to the existing literature by suggesting two variables that possibly moderate the relationship between CEO-to-worker pay and firm performance. These two variables, firm size and power distance, will be discussed in the next two sections.

2.5 Firm Size

Studies that explore the relationship between intra-firm wage dispersion and firm performance have included firm size in their models in several ways. For one, McLaughlin (1988) suggested a popular extension to Tournament Theory, which proposed that as the number of contestants (i.e., the number of employees) increases, the size of the prize (i.e., the CEO’s salary) should also increase. The theoretical reasoning behind this is that as there are more contestants in the tournament, the probability of winning the ultimate prize is decreased, and therefore the size of the prize will increase, just like in a lottery. For this reason, some studies use the number of employees (a measure of firm size) as a predictor variable of CEO pay or of CEO-to-worker pay.

A second reason studies use firm size is for its function as a control variable. Almost all studies that hypothesize a relationship between intra-firm wage dispersion and firm performance use firm size in some way as a control variable. The most common reason for this is because of the proposed influence of firm size on salaries in general, on CEO compensation in specific, and on intra-firm wage dispersion. For example, Winter-Ebmer & Zweimüller (1999) argue that “it is a well established fact that larger firms pay higher wages,” and therefore they control for size by including plant size as a control variable (Winter-Ebmer & Zweimüller, 1999, p. 558-559). Most other studies follow this approach, using a variety of measurements for firm size, such as a 4-year average of the number of employees (Eriksson, 1999), total assets (Chen, Ezzamel & Cai, 2001), and total sales (Fredrickson, Davis-Blake & Sanders, 2010). Some other studies control for firm size for a different reason: its possible influence on firm performance. An example of this is Ridge, Aime & White (2015), who argue that “inertial tendencies associated with firm size could influence firm performance” (p. 626). These are all examples of using firm size as a

(22)

can have an effect on the relationship between intra-firm wage dispersion and firm performance, especially in our model which uses CEO-to-worker pay.

When a firm has fewer employees, the distance to the CEO is smaller from the employee’s point of view. In such a situation, it may be more likely that employees compare themselves to the CEO than in a big firm where the CEO is just a far-away executive who the employees never see. As the perceived distance between the employees and the CEO becomes larger, we hypothesize that the effect of CEO-to-worker pay ratio also becomes smaller. It is thus theorized that the number of employees in a firm moderates the relationship between CEO-to-worker pay and firm performance. In very large firms there is less likelihood that the average employees relate themselves to the CEO due to the perceived distance between them, while in companies with fewer employees the CEO is more visible to the average employee, and is therefore more of a comparison target for regular employees. Therefore, we theorize that firms with more employees will have a smaller effect of the CEO-to-worker pay ratio on firm performance than firms with fewer employees.

2.6 Power Distance

Power distance is one of the four original cultural dimensions that were hypothesized by Geert Hofstede, and is still part of the current framework by Hofstede which includes six dimensions. Power distance concerns “the extent to which the less powerful members of institutions and organizations within a country expect and accept that power is distributed unequally” (Hofstede, Hofstede & Minkov, 1991, p. 61). For example, in France it is quite common that bosses and employees have an unequal relationship, while in Sweden employees will not accept large power imbalances between them and their superiors. In the context of intra-firm wage dispersion, it is commonly argued that power distance has a positive effect on CEO pay and the CEO-to-worker pay ratio (Tosi & Greckhamer, 2004). The idea behind this is that in a country with high power distance, employees will accept larger differences in compensation between them and their superiors, and CEOs and other executives will expect to earn significantly more than their employees.

(23)

tries to go across borders and to see what kind of results and relationships that uncovers. Some authors do speculate about possible differences between countries, such as Chen, Ezzamel & Cai (2011), who study the effects of intra-firm wage dispersion on firm performance in China. They suggest that “tournament theory has cultural support in China as its culture is characterized by ‘high power distance’” (p. 1198). Although they acknowledge the possible influence of power distance, they only study Chinese firms and can thus not incorporate this variable into their model. This is a gap in the literature that we wish to address as well.

We hypothesize that the power distance of a firm’s home country has a moderating influence on the relationship between CEO-to-worker pay and firm performance. This is

especially the case for the Equity Theory part of our theoretical model. Earlier, we hypothesized that large differences in pay between the CEO and the average worker can lead to feelings of inequity in the employees, and to decreased productivity. Therefore, in a country like the

Netherlands with a relatively low power distance, a big gap in pay between CEO and worker can demotivate the employees. However, in countries with a higher power distance, such as Spain or Italy, employees will be more willing to accept compensation differences, which will lead to a lower probability of feelings of inequity and demotivation. Therefore, we hypothesize that the power distance of a firm’s home country moderates the relationship between CEO-to-worker Pay Ratio and firm performance. The specific moderation effect that we expect is that firms from countries with a higher power distance will suffer less from the negative firm performance effects from Equity Theory. This moderating effect is visualized in figure 6, which shows the regular proposed relationship on the left side, and on the right side the relationship moderated by a high level of power distance.

Figure 6: Theoretical effect of the power distance moderator on the relationship between CEO-to-worker pay ratio and firm performance

Firm p erf o rm an ce

CEO-to-worker pay ratio

Firm p erf o rm an ce

(24)

2.7 Hypotheses

The discussion of the theory and the existing literature leads to three hypotheses that test the theoretical model used in our paper.

Out of the four theoretical models we discussed in the literature review, we adopt the third one for our paper. This means that we hypothesize that the ratio of CEO-to-worker pay has an inverse U-shaped (hill-shaped) relationship with firm performance. Tournament Theory and Equity Theory work together to explain the impact of CEO-to-worker pay on firm performance. At low levels of CEO-to-worker pay it is expected that the lack of incentives leads to low performance of employees and consequently also to low firm performance. As wage dispersion grows, it is expected that firm performance also increases due to more incentives for employees to be productive and to aspire a higher position within the firm. These are the effects that would be expected in accordance with Tournament Theory. However, as the CEO-to-worker pay ratio continues to increase it is expected that, at some point, the differences in wages will become so great that they create feelings of inequity and of unfairness. These feelings are thought to lead to demotivation and decreased labor productivity. Therefore, at high levels of wage dispersion, it is expected that firm performance will start to decrease again, as expected per Equity Theory.

We choose to use this model, as opposed to one of the other three, because its reasoning is the most theoretically sound and because it fits the CEO-to-worker pay ratio the best. The literature review showed that the opposite model (U-shaped relationship) had weak theoretical foundations and there was only one study that supported this model. Moreover, for CEO-to-worker pay the inverse U-shaped relationship makes the most theoretical sense by integrating both the positive effects of Tournament Theory and the negative effects of Equity Theory.

Therefore, the following hypothesis is proposed:

H1: The ratio of CEO-to-worker pay has an inverse U-shaped relationship with firm performance

(25)

hypothesize that in a smaller firm the effect of CEO-to-worker pay on firm performance will thus be stronger than in firms with more employees. This leads to the following hypothesis:

H2: The number of employees in a firm moderates the relationship between CEO-to-worker pay and firm performance

Another moderating effect on the relationship between CEO-to-worker pay and firm performance that we hypothesize are country differences. Whereas in one country employees may accept certain levels of pay difference between the CEO and the average employee, in other countries the same difference may create feelings of inequity and tension. We hypothesize that this can be explained through Geert Hofstede’s Power Distance, which measures to what degree people in a country expect and accept power differences in society. We hypothesize that this power distance moderates the relationship between intra-firm wage dispersion and firm performance. This leads to the following hypothesis:

H3: The Power Distance of a firm’s home country moderates the relationship between CEO-to-worker pay and firm performance

2.8 Conceptual Model

In this section we present the conceptual model that follows from the hypotheses presented in the previous section.

Figure 6: Conceptual Model

IV: CEO-to-Worker Pay Ratio

DV: Firm Performance

MV2: Power Distance of Firm’s Home Country

H1

H2

MV1: Number of Employees

(26)

In this model one can see the main relationship between the independent and dependent variable as postulated by hypothesis 1. It is hypothesized that this relationship is curvilinear: as CEO-to-worker pay increases, so does firm performance, up to a certain level, when firm performance starts to decline as intra-firm wage dispersion continues to increase. One can also see the two moderator variables in this conceptual model. Hypothesis 2 represents how the number of employees moderates the relationship between CEO-to-Worker Pay and hypothesis 3 does the same for the power distance of the firm’s home country. We now continue to the next chapter, where the methodology of our paper is presented.

3 Methodology

To test the hypotheses developed in the previous chapter, a regression analysis will be performed. This chapter will map out the relevant variables, data sources and collection, and regression model.

3.1 Independent Variables

The main independent variable in our study is CEO-to-worker pay ratio. Almost all the other literature that tests hypotheses based on Tournament Theory and Equity Theory measure wage dispersion by constructing a variable using the wages of relevant groups in a company that are available and then computing a measure of the difference between them (e.g. Eriksson, 1999; Lallemand, Plasman & Rycx, 2004). One downside of this is that it requires a large amount of data on the wages of individual employees at firm level, which may be an explanation for why research on this topic is relatively scarce, as such specific employee-employer linked data sets are hard to obtain and to analyze. Another drawback is that different studies use different methods to compute a measure of difference between wages.

However, recent developments have provided the opportunity to measure intra-firm wage dispersion in a different way: through the ratio of the remuneration of the CEO and the

(27)

provides voluntary guidelines for reporting Corporate Social Responsibility information that companies can use to include in their own social reports, has also started to encourage companies to report the ratio between the CEO’s compensation and the median employee compensation. Their G4 guidelines, published in 2013, include an indicator (G4-54) that requires companies who voluntarily subscribe to the GRI’s “Comprehensive Guidelines” to report this ratio.

In our study this CEO-to-median-worker pay ratio will be used to measure intra-firm wage dispersion. This is a suitable measure for multiple reasons. For one, this measure reflects the important role of CEO compensation in Tournament Theory and Equity Theory. As argued in the previous chapter, employees do not just compare their salaries to workers who are similar to themselves, but also to workers who are dissimilar. The CEO has a special position in this, as this is the most public figure of the company and the representative of top management. As such, the CEO-to-median-worker pay ratio adequately reflects this importance of CEO remuneration, whereas a more general measure that measures differences between all kinds of employees does not do this.

In addition, this methodology will contribute to understanding this CEO-to-median-worker pay ratio. As this ratio becomes more public through the introduction of the Dodd-Frank Act in the United States and through the GRI guidelines, it is important that researchers get a grasp of what this ratio means and for what purposes it can be used. This is especially the case considering the provocative nature of this ratio; a high ratio, indicating a large gap between CEO salary and worker salary, may cause the public to assume greediness on the part of top

management. Using this variable in this study can bring some more nuance into this discussion, by showing that such a wage structure can have a positive impact on a company, or by dispelling this idea.

The second independent variable is power distance, which acts as a moderator variable. This variable is measured by taking the Power Distance value from Hofstede’s cultural

(28)

variables for which a moderating effect is hypothesized, CEO-to-worker Pay Ratio and Power Distance. Then, we compute the moderator variable by taking the product of these newly centered variables of Pay Ratio and Power Distance.

The third independent variable is the number of employees, which also acts as a

moderator between CEO-to-worker pay ratio and firm performance. To compute the moderator variable, we follow the same process as for power distance. Therefore, we centralize number of employees and CEO-to-worker Pay Ratio, and then compute the new moderator variable by taking the product of these two centered variables. The result is a variable that represents the moderating effect.

3.2 Dependent Variable

The dependent variable in our study is firm performance. There is no consensus on the most appropriate measure for firm performance in the previous literature on the effects of wage dispersion on firm performance. Measures that have been used previously include standardized wages (Winter-Ebmer & Zweimüller, 1999), return on assets (Connelly et al., 2013), profits divided by sales (Eriksson, 1999), profits per employee (Heyman, 2005), gross production value per employee (Hunnes, 2009), and real value added at the plant level (Lallemand, Plasman & Rycx 2004). For our study, we choose return on assets (ROA) as our firm performance indicator, for several reasons. Firstly, return on assets is widely recognized as an apt measure of firm performance. Hagel, Brown & Davison (2010) have defended return on assets as a better measure than, for example, return on equity or return on sales, because it “may foster a better view of fundamentals of the business” with regards to return on equity, and because it

“determines whether the company is able to generate an adequate return on these assets rather than simply showing robust return on sales.”

A second reason to choose return on assets is because of the availability of data on this measure. Not all of the firms in our sample are public firms, which means that return on equity is not available for all firms. Data on assets and net income is much more accessible for a wide variety of firms, which makes ROA an appealing measure. Lastly, return on assets as a measure of firm performance is “widely used in other studies of executive compensation” (Balkin,

(29)

3.3 Control Variables

To isolate the effects of the independent variables, it is important to control for other variables that can have an effect on the dependent variable, return on assets. Previous studies have used many different control variables, such as R&D expenditures (Fredrickson, Davis-Blake & Sanders, 2010), level of diversification (Ridge, Aime & White, 2015), and the average pay of managers in the firm (Eriksson, 1999). In general, the most common control variables can be put in two groups: firm characteristics and workforce characteristics. One of the firm characteristics is industry, as many of the existing literature argues that the industry of a firm has an influence on its performance and that it should therefore be controlled (e.g. Connelly et al., 2013; Eriksson, 1999; Henderson & Fredrickson, 2001). However, our sample is too small to add dummy

variables for industry as that would result in very small groups. Therefore we are not able to control for industry. A second example of a firm characteristic that is commonly controlled for is firm size (Eriksson, 1999; Winter-Ebmer & Zweimüller, 1999; Ridge, Aime & White, 2015). For this reason we include total assets and the number of employees as a control variable. This means that the number of employees has two different functions in our model: once as a moderator variable between CEO-to-worker pay ratio and firm performance, and once as a control variable on firm performance.

The second group of control variables are for characteristics of the workforce. Examples of this include the share of workers with more than 12 years of education, the share of workers with than 10 years of tenure, the share of workers who are younger than 25 (Hunnes, 2005), the share of workers who are female (Jirjahn & Kraft, 2007), and the educational background of the workforce (Lallemand, Plasman & Rycx, 2004). The advantage that these studies have is that they have databases that include specific characteristics on all the employees in a firm, which we unfortunately do not have. Therefore, it is not possible for us to control for the characteristics of a firm’s employees as this data is not publicly available for our sample.

(30)

3.4 Data Sources and Collection

In finding the relevant data for our paper, the most important and difficult variable is CEO-to-median-worker pay ratio. This measure is not reported in any of the major databases like Orbis, and companies tend to want to keep this ratio private for fear of controversy over large

differences in pay between the CEO and the average worker. Listed firms in the United States will be forced to report this ratio starting in 2017, but our paper precedes that date and it is therefore not possible to use that data. The data collection in our paper therefore focuses on the firms who adhere to the “Comprehensive Guidelines” of the Global Reporting Initiative. Firms who subscribe to these guidelines are required to report “the ratio of the annual total

compensation for the organization’s highest-paid individual in each country of significant

operations to the median annual total compensation for all employees (excluding the highest-paid individual) in the same country” in their annual reporting. This “highest-paid individual” will usually be the CEO. The Global Reporting Initative provides an Implementation Manual for firms which clarifies how a firm should calculate this ratio. The relevant part of this

implementation manual for the ratio has been added to our paper as appendix 1. Because this ratio has only recently become a requirement, the only year for which enough data is currently available is 2014. Furthermore, we limit our sample to European firms, for multiple reasons. For one, we wish to limit institutional differences in our sample. Another reason is that the CEO-to-worker pay ratio is very scarcely available in countries outside Europe, and that accessing annual reports from these countries to check for the CEO-to-worker pay ratio is difficult, also due to language differences.

These limitations (European firms who adhere to the “Comprehensive Guidelines” of the Global Reporting Initiative) lead to an initial list of 143 firms that are registered as such by the Global Reporting Initiative (GRI). After going through the annual reports and social

(31)

unavailability of data, the argument that this ratio is irrelevant, and some companies simply ignore the ratio altogether without providing justification.

Of the remaining list of 62 companies, some only provide the ratio for their home

country, and others also provide this for other countries in which they operate. For the purpose of this study, if figures for multiple countries were reported only the home country ratio was used. This is done because some firms are only active in one country or only report for their home country. Furthermore, some companies are active in third-world countries, and including the data from those countries can skew the data. For example, a firm can have a large work force in Indonesia, and because the wage levels in Indonesia are relatively low, the CEO-to-worker pay ratio is skewed by this. For those reasons we chose to only include the home country CEO-to-worker pay ratio, so that the ratio reflects the same concept for every company in our sample.

After collecting data for the CEO-to-median-worker pay ratio from all the different annual reports of the companies, the data for firm performance (dependent variable), firm size (control variable), and previous firm performance (control variable) were collected through Bureau Van Dijk’s Orbis database. Of the 62 companies, full data was available for 47

companies in this database, with some companies lacking data for only one variable, and some for multiple variables. For an additional 11 companies the data could be completed with help from other sources, mainly annual reports. For a list of all the manual data entries and their sources, please see appendix 2.

The degree of power distance in the home countries of the firms is collected directly through the website of the Hofstede Centre and added to the database. For example, a company from France will have the general Power Distance value of France to represent the level of acceptance of inequality in that country.

Finally, because Orbis does not provide a separate figure for return on assets, instead net income and total assets were exported from Orbis, and return on assets was calculated by the calculation: (net income in 2014) / (total assets in 2014).

(32)

but the fact is that companies are very reluctant to share information on the CEO-to-worker pay ratio, and that such data is thus very rare. Of course this relatively small sample size does need to be taken into account in interpreting the results of this analysis.

3.5 Regression Model

The regression model we will be testing in our paper can be written as a statistical equation. In this section we develop this equation, step by step. As a first step, we include the control variables.

controls

Y

In this formula, Yrepresents the dependent variable (return on assets),  represents the constant, and  represents the error term. Now, we add in a simple linear effect of CEO-to-Worker Pay Ratio.

X

controls

Y

1 ratio

Here,1Xratio represents the independent variable CEO-to-worker pay ratio. Now, to this

equation we add the hypothesized curvilinear relationship by including the squared term of the CEO-to-worker pay ratio. We then come to the following regression equation:

X

X

controls

Y

1 ratio 2 ratio

²

(33)

X

X

X

X

controls

Y

1 ratio 2 ratio

²

3

(

ratio

*

employees

)

Subsequently, we add the proposed moderating effect of the power distance of the home country of a firm (pd) on the relationship between the dependent and independent variable.

X

X

X

X

X

X

controls

Y

1 ratio 2 ratio

²

3

(

ratio

*

pd

)

4

(

ratio

*

employees

)

This final equation represents the full regression model that is used in our paper, which will be carried out in the next chapter, in which the results of the regression analysis will be presented.

4 Results

To test our hypotheses a hierarchical multiple regression analysis will be performed in order to simulate an inverted U-shaped model. To do so we will first check the dataset for outliers and for the assumptions of multiple regression. Then, we will look at the descriptive statistics of the dataset, and finally we will perform the actual regression analyses that test our hypotheses. 4.1 Outliers

Firstly, it is important to check our main independent variable, CEO-to-worker pay ratio, for outliers. Therefore, we perform a boxplot and find that cases 1 (Abengoa) and 39 (Nestlé) are outliers. After checking these companies’ reporting on this ratio it becomes clear that they both report worldwide ratios rather than a national ratio, indicating a coding mistake. This heavily skews the ratio as they include wages from countries with a much lower standard of living, thus resulting in a very high ratio that does not accurately represent the home country. Consequently, it is justified to delete these cases from our dataset, which results in a final sample of 56 cases. 4.2 Normality

(34)

this reading of the histogram, showing an acceptable skewness (-0,773, std. error 0,319) but a relatively high level of kurtosis (8,346, std. error 0,628). Transformations, such as the natural logarithm or a squared term, do not help this, so we have to accept this level of kurtosis. The other variables in our model are highly non-normal, but because normality is only assumed for the dependent variable this does not harm the validity of our analysis.

4.3 Homoscedasticity

A second assumption is that the dependent variable is homoscedastic. This means that “the variance of residuals when plotted against the predicted values of your dependent variable is relatively the same across all those predicted values” (Pearson, 2010). If this assumption is violated we speak of heteroscedasticity, which can make non-significant relationships seem statistically significant because of an underestimation of the standard errors of the regression coefficients (Pearson, 2010, p.290). We test this assumption of homoscedasticity using a scatterplot that plots the standardized residual against the standardized predicted values. Subsequently, we add a linear fit line and we observe that this line is horizontal, indicating that the variable is indeed homoscedastic. We can thus conclude that the assumption of

homoscedasticity is satisfied, indicating no problems for our analysis.

4.4 Multicollinearity

(35)

Table 1: Multicollinearity Statistics

Variable VIF

CEO-to-worker Pay Ratio 1,926

Number of Employees 1,869

Total Assets (x1000) 1,115 Return on Assets (1 year prior) 1,109 Moderator 1 (power distance) 1,943 Moderator 2 (firm size) 2,724

Multicollinearity problems are usually considered to start at a VIF value higher than 5, with values higher than 10 seen as being definitely problematic (Hoerl & Snee, 2012, p. 263). As the table indicates, the VIF values in our model are very low and do not indicate multicollinearity. Additionally, we perform Pearson correlation analyses on the variables to see if they correlate with each other. The resulting correlation table (in section 4.6) shows some correlations, but most of the significant correlations are only between variables that are computations of each other (such as the moderator variables) and thus do not form a problem. The only significant relationship that is not between two computed variables is between Return on Assets in the focal year (2014) and in the year before that (2013), which is to be expected and is not a risk for multicollinearity as return on assets is a dependent variable, and multicollinearity is only a problem if it is between independent variables. Therefore, on the basis of these two tests we can conclude that there are no multicollinearity problems.

4.5 Curvilinear Relationship

(36)

fit a line along the data in the scatterplot a very slight quadratic relationship is revealed, which shows a small inverted U-shaped relationship. Although the shape of this relationship is as hypothesized, the fact that this relationship appears so small does not bode well for our hypotheses. Nevertheless, because the shape is as hypothesized, we can continue with our analysis.

Figure 7: Scatterplot of the main dependent and independent variable with a fit line

4.6 Descriptive Statistics

Table 2 and 3 on the next page show the correlation matrix and the descriptive statistics of the variables used in this study.

-0,3 -0,25 -0,2 -0,15 -0,1 -0,05 0 0,05 0,1 0,15 0,2 0,25 0 10 20 30 40 50 60 70 RE TU RN ON ASSE TS

(37)

Table 2: Correlation Matrix Return on Assets CEO-to-worker Pay Ratio Total Assets (x1000) Number of Employees Return on Assets (1-year prior) Moderator 1 (Power Distance) Return on Assets Pearson Correlation 1

Sig. (2-tailed)

CEO-to-worker Pay Ratio Pearson Correlation ,028 1 Sig. (2-tailed) ,835

Total Assets (x1000) Pearson Correlation -,054 ,121 1 Sig. (2-tailed) ,694 ,374

Number of Employees Pearson Correlation ,063 ,173 ,213 1 Sig. (2-tailed) ,647 ,203 ,115

Return on Assets (1-year prior) Pearson Correlation ,422** ,067 -,128 ,246 1

Sig. (2-tailed) ,001 ,629 ,351 ,070

Moderator 1 (Power Distance) Pearson Correlation -,044 ,547** -,086 -,112 ,039 1

Sig. (2-tailed) ,749 ,000 ,529 ,413 ,777

Moderator 2 (Firm size) Pearson Correlation ,048 -,475** ,105 ,468** ,061 -,627**

Sig. (2-tailed) ,725 ,000 ,441 ,000 ,658 ,000

Table 3: Descriptive Statistics

(38)

4.7 Testing of Hypotheses

In order to test for an inverted U-shaped relationship between CEO-to-worker pay ratio and firm performance, we use a hierarchical regression analysis with return on assets of the focal year (2014) as dependent variable. This means that we incrementally add variables in our regression model to look at the changes in the different models. In our first model we do a baseline

regression with only the control variables. Regression analysis shows that this model is

significant at the 0,05 significance level as p=0,016. However, a look at the coefficients in table 4 shows that only prior firm performance (the ROA of 2013) has a significant effect.

Table 4: Regression coefficient results of model 1

Standardized Beta P value

(Constant) 0,564

Number of Employees -0,062 0,647

Total Assets (x1000) 0,004 0,975

Prior firm performance 0,438 0,002

In our second model we include our main independent variable, CEO-to-worker pay. This second model represents a linear relationship between CEO-to-worker pay and firm

performance. In the third model we include the squared function of CEO-to-worker pay ratio as an independent variable to simulate the proposed curvilinear relationship between CEO-to-worker pay ratio and return on assets. The results can be seen in table 5.

Table 5: Regression results of models 2 and 3

Model R R Square

Adjusted R Square

R Square

Change F Change Sig. F Change

2 ,426 ,182 ,116 ,182 2,772 ,958

3 ,426 ,182 ,098 ,000 ,017 ,898

Model 2 Predictors: (Constant), Control variables, CEO-to-worker Pay Ratio

(39)

In these regression results, model 2 represents the linear relationship between CEO-to-worker Pay Ratio and Return on Assets. The associated statistics on this model do not reveal a

statistically significant relationship, as the Sig. F Change value is very high (,958) and is not significant at any reasonable significance level. Model 3, which represents the curvilinear relationship between CEO-to-worker Pay Ratio and Return on Assets, has slightly better statistical values but still does not even come close to displaying a significant relationship. In spite of this, the model in its entirety is statistically significant, but this is only because of the control variable on prior firm performance. This analysis, which includes only the relationship between CEO-to-worker Pay Ratio and return on assets, thus gives no evidence for a statistically significant relationship between CEO-to-worker Pay Ratio and firm performance. We thus have to reject hypothesis 1, which proposed an inverted U-shaped relationship between CEO-to-worker Pay Ratio and firm performance.

In our final model we add in the proposed moderating effects of firm size (measured by number of employees) and of power distance on the relationship between CEO-to-worker pay and return on assets. This final model includes all the variables, including the moderating effects of firm size power distance. For this model full regression results are shown in the tables below.

Table 6: Regression coefficient results of model 4

Standardized Beta P value

(Constant) ,780

CEO-to-Worker Pay Ratio -,116 ,849

CEO-to-Worker Pay Ratio (Squared) ,238 ,717

Number of Employees -,084 ,672

Total Assets (x1000) ,000 ,999

Prior Firm Performance ,451 ,002

Moderator 1 (Firm size) ,025 ,916

(40)

Table 7: Regression results of model 4 Statistic Value R Squared 0,194 Adjusted R Squared 0,074 F-value 1,612 P-value 0,155

The second hypothesis stated that “A firm’s number of employees moderates the relationship

between intra-firm wage dispersion and firm performance.” The results in table 6 show that the

moderator of firm size has a p-value of 0,916, which is not statistically significant. We therefore have to reject hypothesis 2.

The third hypothesis stated that “The Power Distance of a firm’s home country moderates

the relationship between intra-firm wage dispersion and firm performance”. The results of

model 4 show that no such moderating effect was found in our data as the p value of the moderator is 0,532. Therefore, we do not find evidence that supports hypothesis 3.

To check the robustness of our results we also performed the same analyses with the net profit ratio (profits / revenue) as the dependent variable to represent firm performance. However, all results stayed the same, and no significant relationships were found in this situation either. We also checked for other relationships within the data, such as an ordinary linear relationship, but this did not provide any results either. Furthermore, we performed an additional check by running our regression analyses without the control variable prior firm performance (return on assets in the previous year) because this variable was highly correlated to the dependent variable (return on assets in 2014), but this also did not reveal any differences in the results. We therefore have to conclude that our dataset shows no evidence at all that supports our hypotheses.

5 Discussion

(41)

we have found no evidence at all for our hypotheses, we are forced to consider what went wrong in our theoretical or methodological work that has concluded in these unexpected results. Firstly, we will discuss possible theoretical consequences, and in the limitations section we will look at the methodological limitations that constrained this thesis.

Theoretically speaking, our results are a setback for those who propose that wage

dispersion between dissimilar workers has an important place within discussions of Tournament and Equity Theory. This is a debate that has been going on since the earliest studies on this matter, and it basically concerns whether Tournament Theory and Equity Theory function when there is wage dispersion among similar workers (e.g. direct colleagues) and/or between

dissimilar workers (e.g. between a blue-collar employee and the CEO). Some scholars have expressed doubt over whether one can apply these theories to dissimilar workers, as we have done in our paper. For example, Hunnes has argued that “the wage spread incentive is most effective among homogenous workers with the same job design” (2009, p. 778). Similarly, Gupta, Conroy & Delery state that “the relevance of similar others is higher than the relevance of dissimilar others for assessing pay equity in work situations” (2012, p. 108). Furthermore, Major & Forcey (1985) concluded that people prefer to compare their wages to people who are most similar to them, such as people who have the same job and even people who are of the same sex. The difficult issue is that it is possible to cite many other sources who state that comparisons between dissimilar workers are in fact relevant, as we did in our theory section. However, we have to conclude that our results are more in line with the view that the comparison between CEO and average worker are not relevant for firm performance, although further studies will have to confirm this.

Referenties

GERELATEERDE DOCUMENTEN

Where Zuse explores information processing theory as a way to give new insights in physics and where Schmidhuber (mostly) hypothetically explores what it would mean for a universe to

The current study contributes to alliance network theory by answering the question whether the performance of firms, who participate in alliance networks, is influenced by the

Additionally, the thesis verifies whether this process of relying on vicarious learning from nearby located organisations affects small and large firms equally.. This is an important

Since management practices have a strong effect on firm performance and differ across countries (Bloom & van Reenen, 2010; Bloom et al., 2016), foreign-owned firms could

Based on the analyses within this study it can be concluded that innovation activities of companies in the food-manufacturing industry indeed generate higher sales

Q2: CEO power affects the amount of M&A activity done by a company: companies with high levels of CEO power tend to have a more intense M&A activity, expressed in

Comparing the percentages of country-level, industry-level, country-industry interaction effects and firm- level heterogeneity, we notice that country factors account for

The calculation below will show how the calculation of the maintenance performance works, furthermore the relation between effectiveness, efficiency and performance will