1
Thesis
Name: Annelies Slik
Student number: 11018429
Specialisation: Economics and Finance
Field: Organizational Economics
Number of credits: 12 EC
Title: Agency theory and moral hazard: the influence of gender on
remuneration
2
Statement of originality
This document is written by Student Annelies Slik who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 Table of contents Abstract 4 1 Introduction 4 2 Literature review 6 3 Methodology 9
3.1 Data collection of the sample 11
3.2 Tests and hypotheses 12
4 Results 12 4.1 Fixed remuneration 13 4.2 Variable remuneration 14 4.3 Total remuneration 17 4.4 Multicollinearity tests 18 5 Discussion 21 6 Conclusion 23 Reference list 25 Appendix 27
4
Abstract
This thesis examines the influence of gender on the remuneration policies of the directors of the largest companies in the Netherlands. Fixed, variable and total remuneration are regressed on gender and control variables. Total and fixed remuneration are dependent on gender, variable remuneration is not. This implies that a gender wage gap exists for total and fixed remuneration. Moral hazard is defined by excessive risk-taking in this thesis. The moral hazard problem can be solved by motivating a director through variable remuneration. The examination of the results of the regression of variable remuneration indicates that men and women act similarly in a situation that allows for moral hazard.
1 Introduction
In 1966 the iconic James Brown released the song ‘‘It’s a Man’s Man’s Man’s world’’, which was probably the truth in that period of time. However, a lot has changed since then. Nowadays in the Netherlands the number of employed women keeps rising and the universities are filled with more female students than male students (Ministerie van Onderwijs, Cultuur en Wetenschap, 2018). Moreover, on January 15th 2016 the Dutch government decided to extend the implementation of a regulation that stimulates the number of females in executive positions in large companies (Rijksoverheid, 2016). Keeping these facts in mind it will not be a surprise that the number of females in the boards of the largest firms in the Netherlands has been rising (Top machtigste vrouwen, 2017). Due to the alterations the composition of the Dutch executive boards is facing, other factors might be changing as a result.
It is essential to keep the people in executive positions as motivated as possible, no matter their gender. This may sound easier than it actually is. The relationship between the owner of the company and its director can be explained by agency theory (Eisenhardt, 1989). One of the problems that might occur when an owner and a director work together is that the director starts to show signs of moral hazard (Eisenhardt, 1989).
Moral hazard is one of the agency problems in the principal agent relationship that is described by agency theory (Eisenhardt, 1989). This problem occurs in a situation where the owner and the director face a conflict of interests and when the actions of the director cannot be perfectly observed by the owner (Eisenhardt, 1989). The outcome of this unfortunate issue is that the director does not exert all of his/her potential effort at the expense of the owner (Eisenhardt, 1989). In this paper, the moral hazard problem implies excessive risk-taking by
5
the director. The solution to this complication is finding the appropriate remuneration policy to keep the director motivated (Koch & Peyrache, 2008).
Combining the information on moral hazard with the increase of the number of female directors in the largest Dutch firms gives rise to a natural question: is the appropriate remuneration policy influenced by gender? Therefore, this thesis investigates if gender diversity among the directors of the 50 largest Dutch firms influences the remuneration policies. The results of this analysis are important to understand what motivates the group of people that run the largest Dutch firms.
Albanesi, Olivetti and Prados (2015) did research on gender differences in the remuneration of executives using data from the S&P 500, an American stock market index. They found proof that the structure of the remuneration policy is different for males and females (Albanesi et al., 2015). Their research is focused on the comparison of men and women and uses the United States of America as a target region (Albanesi et al., 2015). Benkraiem, Hamrouni, Lakhal and Toumi (2017) also indicated that the number of independent female executives has a negative effect on the level of CEO remuneration. The focus of this analysis is on the market of France and on board independence (Benkraiem et al., 2017). The influence of gender on the structure of remuneration of the directors of the largest firms in the Netherlands has not been analysed before. This paper aims to fill this gap in the existing literature. The research that is conducted by this thesis builds on the groundwork of previous economists by investigating a different region and focusing on the dependence of the remuneration policy on gender.
This paper examines the influence of gender on the structure of remuneration by use of a regression. The total remuneration policy, the fixed salary and the variable salary are regressed on gender and control variables. The main finding of this paper is that gender influences the fixed and total remuneration. Nevertheless, variable remuneration is not affected by gender. This implies that there exists a gender wage gap in the base salary and total salary of the directors of the largest Dutch firms. Women receive less of these types of remuneration than men. Moreover, since there is no difference in variable pay among male and female directors, they seem to be motivated in a similar manner.
First, a literature review is conducted, examining the research done about this topic. Second, the research method of this paper is explained along with the process of data collection and tests for this examination. After this the results of the investigation are presented. Finally, the discussion and conclusions of this thesis are stated.
6
2 Literature review
Agency theory has been studied extensively by multiple economists since the 1970’s (Eisenhardt, 1989). Eisenhardt reviews this theory in her article ‘‘Agency Theory: An Assessment and Review’’ (1989). According to her, agency theory examines the relationship between an employer and an employee, referred to as principal and agent respectively (1989). A contract is used to define this particular relationship (Eisenhardt, 1989). This thesis focuses on the specific setting of the interrelation between the owner and the directors of the largest firms in the Netherlands.
One of the issues that can occur when a principal and an agent interact is the agency problem (Eisenhardt, 1989). Eisenhardt (1989) states that this problem is contingent on two conditions. First, there must be a conflict of interests between the two parties involved in the relationship (Eisenhardt, 1989). The principal and the agent only care about maximizing their own interests (Eisenhardt, 1989). Therefore, if the two do not see eye to eye, a conflict arises. Second, it must be impossible to observe and verify all the actions of the agent (Eisenhardt, 1989). This implies that the principal deals with information asymmetry regarding the effort of the agent (Hölmstrom, 1979). According to Eisenhardt (1989) there are two types of agency problems, moral hazard and adverse selection. The focus of this thesis is on the issue of moral hazard.
Moral hazard materializes when the agent takes advantage of the information asymmetry the principal is facing (Hölmstrom, 1979). Moreover, the interests of the agent and the principal do not align, which gives the agent an incentive to exhibit undesirable behaviour (Eisenhardt, 1989). The agent exerts less effort than (s)he should and shirks (Eisenhardt, 1989). The moral hazard problem implies that the agent start to take excessive risk.
One of the possible solutions to moral hazard is the construction of the appropriate remuneration policy (Eisenhardt, 1989). According to Eisenhardt (1989), it is best to make this policy based on the outcome delivered by the agent. If the salary of the agent is based on the preferred results of the principal, the contract aligns the interests of both parties (Eisenhardt, 1989). Hölmstrom also studied moral hazard and its solution and confirmed that a fitting contract can prohibit the agent from lacking effort (1979). Moreover, Stevens and Thevaranjan (2010) concluded that a flat salary will not provide any incentives to the agent in favour of the principal. This implies that according to these economists, moral hazard can be prevented by the suitable remuneration policy (Eisenhardt, 1989; Hölmstrom, 1979; Stevens & Thevaranjan, 2010). It is crucial that a contract is adjusted to the particular agent it is made for to ensure that the motivation of this agent is maximized (Koch & Peyrache, 2008). A remuneration policy
7
can consist out of various components. This thesis considers that a remuneration policy consists out of a base salary, bonuses and stocks options. The bonuses and stock options make up the variable pay. This part of the remuneration is based on outcome and keeps the agent motivated (Eisenhardt, 1989).
An individual’s remuneration policy is dependent on his/her own characteristics and several other factors. This thesis focuses on gender and control variables as the determinants of a compensation package. The objective of this thesis is to examine the influence of gender on the remuneration of the directors of the largest firms in the Netherlands, which implies that the determinant of interest is gender. There is no previous research about this topic in this particular setting, so this thesis fills a gap in the existing literature. As stated before, a remuneration policy is designed to prevent moral hazard from occurring (Eisenhardt, 1989). If gender influences the design of such a policy, males and females might act differently in the case of moral hazard. This paper investigates this issue thoroughly.
Albanesi et al. (2015) did research on all the top executives of the companies that are included in the S&P 500. They stated that the variable compensation of women is less based on performance than the variable compensation of men (Albanesi et al., 2015). Most differences between the remuneration policies of men and women originate from the variable pay (Albanesi et al., 2015). Next, Perryman, Fernando and Tripathy (2016) examined the relationship between gender diversity and the risk level of a firm. One of their findings was that female directors receive a lower compensation package then male directors (Perryman et al., 2016). This difference starts to disappear once the gender diversity among the board of executives increases (Perryman et al., 2016). Benkraiem et al. (2017) also investigated this topic in France and stated that an increase in the number of independent women on a board of executives causes a decrease in the base salary of the CEO. Independent women tend to increase the level of supervision and control which affects the fixed part of the CEO remuneration policy (Benkraiem et al., 2017). The increase in control reduces the agency costs in a firm (Benkraiem et al., 2017).
Contradictory to the previous finding of Benkraiem et al. (2017) mentioned, they also confirmed that the number of female executives on the board and the complete remuneration policy have a positive relationship (Benkraiem et al., 2017). This positive relation also holds for the number of female executives on the board and the variable part of the remuneration policy (Benkraiem et al., 2017). Next, Bugeja, Matolcsy, and Spiropoulos (2016) analysed the influence of the level of gender diversity on the board of executives on the level of CEO compensation among U.S.-listed firms. This influence turned out the be insignificant (Bugeja
8
et al., 2016). In a different article Bugeja, Matolcsy and Spiropoulos (2012) did research on a similar topic and again concluded that there is no gender wage gap. Finally, Geiler and Renneboog (2015) examined the gender wage gap among all executive positions in the UK and used all the UK-listed firms in their dataset. They stated that a gender wage gap exists for all executive positions except for the CEO position (Geiler et al., 2015).
The following papers examined how gender diversity influences the excessive risk-taking and the overall level of risk in a company. This thesis defines moral hazard as excessive risk-taking. Ali, Liu, and Su (2018) stated that the degree of default risk in firms is depressed when the boards are more gender-diverse. In addition, Faccio, Marchica and Mura (2016) concluded that women have a less risk-taking mindset than men. Finally, Baixauli-Soler, Belda-Ruiz and Sanchez-Marin (2015) confirmed that women are more risk-averse then men. As mentioned before, Benkraiem et al. (2017) extensively examined the influence of gender diversity on the monitoring of the remuneration policies of executives in France. In 2011 France enforced a regulation to stimulate the number of women in executive positions of the largest French firms (Benkraiem et al., 2017). This is very similar to the situation in the Netherlands, which is analysed in this thesis. One of their main findings is that independent female executives reduce the agency costs in a company by increasing the level of supervision (Benkraiem et al., 2017). According to Benkraiem et al. (2017) the French regulation for the women quota benefits the decision-making process in firms. Benkraiem et al. (2017) concluded that even though there exists a positive relation between the number of women on board and the remuneration packages, overall female executives receive less compensation then male executives.
Another group of economists that were mentioned earlier on in this thesis are Albanesi et al. (2015). They did research on all the U.S. companies that are listed on the S&P 500 and acknowledged that a gender wage gap between male and female executives exists (2015). The difference in pay comes from the parts of the remuneration policy that are designed to motivate the executives, in particular the stock options (Albanesi et al., 2015). Albanesi et al. (2015) concluded that it is important for firms to include more women in their boards of executives.
Finally, Perryman et al. (2016) also did important research on the relationship between gender and the remuneration policy of an executive. Their findings are very interesting (Perryman et al., 2016). Again, the gender wage gap is acknowledged between males and females on the boards (Perryman et al., 2016). However, this gap decreases as the number on women on the boards climbs (Perryman et al., 2016). This implies that gender diversity among
9
the boards of executives has a positive effect on the level of CEO compensation, and in particular the remuneration of female executives (Perryman et al., 2016).
This thesis aims to expand the existing literature by focusing on the influence of gender on the structure of remuneration of the directorsof the largest firms in the Netherlands, which has not been analysed before. The total remuneration policy, the fixed pay and the variable pay are regressed on gender and several control variables. The control variables include age, the possession of a college degree, tenure and the size of the company based on turnover. The beta of gender is expected to be significant, especially for the regression of the variable remuneration. The data required for this research entails the information on the directors of the 50 largest companies in the Netherlands.
3 Methodology
This study aims to investigate the influence of gender on the remuneration of the directorsof the 50 largest Dutch firms. The model created by Benkraiem et al. (2017) is used as a baseline model for the regressions conducted in this thesis. They conducted their research in France which is facing a similar situation as the Netherlands (Benkraiem et al., 2017). The aim of their study is to define the influence of the number of females in the executive boards on the level of supervision on the CEO compensation (Benkraiem et al., 2017). Since the research of Benkraiem et al. (2017) is in some parts similar to the investigation of this thesis, their model is used as a basis. Benkraiem et al. (2017) regresses various forms of compensation on the gender diversity among boards, size of the board, experience and ownership. The different types of pay are defined as total, fixed and variable pay (Benkraiem et al., 2017). The main finding of their paper is that the number of independent females is positively correlated to several parts of the CEO remuneration (Benkraiem et al., 2017).
The model of this paper is performing three regressions. Fixed remuneration, variable remuneration and total remuneration are regressed on gender and several control variables. Age, the possession of a college degree, tenure and the size of the company make up the control variables. The regressions use the estimation method of ordinary least squares. This method has the most optimal statistical properties. The regressions are presented as the following.
𝐹𝑖𝑥𝑒𝑑 𝑟𝑒𝑚𝑢𝑛𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑖
= 𝛽0+ 𝛽𝑔𝑒𝑛𝑑𝑒𝑟𝑔𝑒𝑛𝑑𝑒𝑟𝑖+ 𝛽𝑎𝑔𝑒𝑎𝑔𝑒𝑖+ 𝛽𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝑑𝑒𝑔𝑟𝑒𝑒𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝑑𝑒𝑔𝑟𝑒𝑒𝑖 + 𝛽𝑡𝑒𝑛𝑢𝑟𝑒𝑡𝑒𝑛𝑢𝑟𝑒𝑖+ 𝛽𝑠𝑖𝑧𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑐𝑜𝑚𝑝𝑎𝑛𝑦𝑠𝑖𝑧𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑐𝑜𝑚𝑝𝑎𝑛𝑦𝑖+ 𝑒𝑖
10 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒 𝑟𝑒𝑚𝑢𝑛𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑖 = 𝛽0+ 𝛽𝑔𝑒𝑛𝑑𝑒𝑟𝑔𝑒𝑛𝑑𝑒𝑟𝑖+ 𝛽𝑎𝑔𝑒𝑎𝑔𝑒𝑖+ 𝛽𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝑑𝑒𝑔𝑟𝑒𝑒𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝑑𝑒𝑔𝑟𝑒𝑒𝑖 + 𝛽𝑡𝑒𝑛𝑢𝑟𝑒𝑡𝑒𝑛𝑢𝑟𝑒𝑖+ 𝛽𝑠𝑖𝑧𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑐𝑜𝑚𝑝𝑎𝑛𝑦𝑠𝑖𝑧𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑐𝑜𝑚𝑝𝑎𝑛𝑦𝑖+ 𝑒𝑖 𝑇𝑜𝑡𝑎𝑙 𝑟𝑒𝑚𝑢𝑛𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑖 = 𝛽0+ 𝛽𝑔𝑒𝑛𝑑𝑒𝑟𝑔𝑒𝑛𝑑𝑒𝑟𝑖+ 𝛽𝑎𝑔𝑒𝑎𝑔𝑒𝑖+ 𝛽𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝑑𝑒𝑔𝑟𝑒𝑒𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝑑𝑒𝑔𝑟𝑒𝑒𝑖 + 𝛽𝑡𝑒𝑛𝑢𝑟𝑒𝑡𝑒𝑛𝑢𝑟𝑒𝑖+ 𝛽𝑠𝑖𝑧𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑐𝑜𝑚𝑝𝑎𝑛𝑦𝑠𝑖𝑧𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑐𝑜𝑚𝑝𝑎𝑛𝑦𝑖+ 𝑒𝑖
To fully assess the influence of gender on the remuneration policies of the directors of the largest Dutch companies, three regressions are required. A compensation package often consists out of multiple components. The fixed remuneration is a constant factor that is not affected by the performance of the director. This part of the remuneration policy entails a base salary, privileges (car allowance, transportation costs, medical insurance, etc.) and retirement benefits. The variable compensation is dependent on performance and can take multiple forms. This paper assumes that the variable remuneration consists out of an annual bonus and stock options granted. Since the variable part of the compensation is based on outcome and the fixed part is not, they both play different roles in motivating the director (Eisenhardt, 1989). The variable remuneration is designed to maximize motivation and prevent moral hazard (Eisenhardt, 1989). If gender has an influence on this particular part of the compensation, the level of excessive risk-taking could differ among men and women. Therefore, it is important to separately regress fixed remuneration, variable remuneration and total remuneration on gender and the other control variables.
Gender is the variable of interest in this model. The coefficient of this variable shows the influence of gender on the dependent variable, which is a particular type of remuneration. Gender is represented by a dummy variable in all of the three regressions. When a observation in the sample depicts a male, the dummy is equal to 1. When a female is observed in the sample, the dummy is equal to 0.
The control variables in this model control for the factors that also affect the types of remuneration, but the influence of these variables is not examined. The first control variable is age. The older a person is, the more likely (s)he is to be educated and to be employed for a longer period. This influences the level of remuneration an individual receives. Therefore, age is included as a control variable. Next, the possession of a college degree is also a control variable. This factor is represented by a dummy variable. This implies that if an observation in the sample portrays an individual that possesses a college degree, the dummy variable is equal
11
to 1. If an individual does not possess a college degree, the dummy is equal to 0. Education is a determinant of the personal skills of an individual which influence his/her job performance. Therefore, the remuneration policy is also affected by the possession of a college degree. Another control variable is tenure, which is the time of employment. The longer a person is employed at the same company, the more chances (s)he has had for promotion. This also influences the compensation of an individual. The final control variable is the size of the company the director works for. The size of the firm is determined by the level of operating revenue (turnover) in this thesis. Companies with more operating revenue have the opportunity to award their directors with a higher level of compensation. Therefore, the types of remuneration are also affected by the size of the company.
3.1 Data collection of the sample
The data collected originates from the 50 largest companies located in the Netherlands. Among these 50 companies, the information of 356 directors was gathered. This sample size was chosen because within this range the firms are of similar size and structure. This makes the remuneration policies of their directors comparable. Increasing the sample size could cause certain p-values to become significant, while in reality they are not. To keep the regressions relevant, the sample size of 50 firms was chosen. The database ‘Orbis’ was used to select these companies. The size of the companies is determined by the operating revenue attained in 2017. A list of the largest Dutch companies with their directors and the data on gender, age, the possession of a college degree, tenure and the size of the company is collected by use of the database ‘Orbis’. As mentioned before, the data of 356 directors was gathered. In this sample there are 91 women and 331 directors with a college degree. This implies that the proportion of women in this sample is a little over 25%. ‘Orbis’ provided all the appointment dates of the directors in the sample and these dates were transformed into the time of employment. Tenure is defined in years and the size of the company is denoted in the amount of operating revenue in euros.
The information on the remuneration was more difficult to obtain. Unfortunately, a database with all the information on the remuneration packages of Dutch firms does not exist. Since a list of companies with their accompanying directors was already obtained through Orbis, the annual reports of these firms were used to find the information of the compensation packages of the directors. All the companies in the sample have published their annual reports up to 2017. In these annual reports there is a remuneration report which contains the information on the individual compensation packages of all the people in the board of directors. Using these remuneration reports, all the data about total, fixed and variable remuneration was
12
collected. As mentioned before, the fixed remuneration consist out of a base salary, privileges (car allowance, transportation costs, medical insurance, etc.) and retirement benefits. The variable compensation contains the annual bonuses and the stock options granted. Naturally, the total remuneration package is the sum of these two.
3.2 Tests and hypotheses
After running the three regressions, the results of each individual regression are examined with the use of the same set of hypotheses. The purpose of these regressions is to estimate the influence of gender on the dependent variable, which is one of the types of remuneration. To look at the impact of gender, its coefficient (beta) is examined. The three gender coefficients, one for each type of remuneration, are tested using a 95%-confidence interval. The p-value is testing the impact of gender. This value defines the probability of finding an equal or more extreme value than the one observed, given that the null hypothesis is true. Combining this with the confidence interval implies that if the p-value of the coefficient of gender is smaller than 0.05, the coefficient is significant and gender influences the dependent variable, a type of compensation. When the p-value is larger than 0.05, gender does not affect the type of remuneration.
The null hypothesis refers to the state where the beta of gender is not significant. The coefficient does not influence the dependent variable and can be considered to be equal to 0. When the p-value of the coefficient is larger than 0.05 there is not enough statistical evidence to reject this hypothesis. The null hypothesis is represented as the following.
H0: 𝛽𝑔𝑒𝑛𝑑𝑒𝑟= 0
If it is found that the p-value of the beta of gender is smaller than 0.05, the null hypothesis is rejected. In this case, there is enough statistical evidence to state that the gender of an individual impacts the level of remuneration that is received. This implies that the beta is significant and can be considered to be unequal to 0. The accompanying hypothesis is presented as below.
H1: 𝛽𝑔𝑒𝑛𝑑𝑒𝑟≠ 0
4 Results
The results are reviewed for each regression individually. First the results for the fixed remuneration regression are discussed. This is followed by the results of the variable remuneration regression. Finally, the total remuneration regression is examined.
13
4.1 Fixed remuneration
The linear prediction of the ordinary least squares regression and the main results are presented below. The complete regression table can be found in appendix C.
Fixed remuneration
𝜷𝒈𝒆𝒏𝒅𝒆𝒓 310444.1
p-value 0.038
The p-value of the coefficient of gender for this particular regression is smaller than 0.05, which implies that the null hypothesis is rejected. Fixed remuneration is influenced by gender. To be more specific, if a director is a male and the dummy variable is equal to 1, he earns €310,444.10 more in fixed remuneration than his female counterpart. This result shows that there is a gender wage gap in fixed remuneration.
Albanesi et al. (2015) also find a gender wage gap in their research. They discuss a few possible explanations for the difference in remuneration (Albanesi et al., 2015). First, it is more difficult for women to enter the business world, especially in a company where there are a lot of male directors (Albanesi et al., 2015). It is more troublesome for women to get in touch with the right contacts and thus receive useful advise from a mentor (Albanesi et al., 2015). This
3 0 0 0 0 0 4 0 0 0 0 0 5 0 0 0 0 0 6 0 0 0 0 0 7 0 0 0 0 0 F it te d v a lu e s 0 1 Gender
Fixed remuneration
14
prevents female directors from obtaining certain positions as easily as male directors (Albanesi et al., 2015). Secondly, among the married executives, women spend a lot more time with their families than men (Albanesi et al., 2015). According to Albanesi et al. (2015), this results in females receiving a lower level of fixed remuneration. Finally, women are less likely to begin a discussion about the level of their compensation (Albanesi et al., 2015). All these arguments explain the gender wage gap in fixed remuneration.
Another possible explanation is provided by Perryman et al. (2016). They state that the gender wage gap of the fixed remuneration is caused by the level of gender diversity (Perryman et al., 2016). The difference in compensation is negatively correlated to the number of women in the boards (Perryman et al., 2016). The proportion of females in the sample that is used for this research is 25%. The base salary of women might be lower because the number of female directors among the boards of the largest Dutch companies is not very substantial(Perryman et al., 2016).
The final explanation for the result of this regression stems from the remuneration committee. The remuneration committee designs the compensation packages for all the employees of the firm, so also for the directors. Bugeja et al. (2016) investigate the influence of a gender-diverse remuneration committee on the level of fixed remuneration of executives. Their main finding is that the level of fixed compensation decreases when there are more independent women on the committee and the board (Bugeja et al., 2016). Female directors tend to increase the level of surveillance on the design of the remuneration policies made by the committee (Bugeja et al., 2016). This leads to a overall lower level of fixed remuneration (Bugeja et al., 2016). As mentioned before, the sample only has a proportion of 25% of women. Therefore, the remuneration policies might be less supervised by the directors. This could cause the gender wage gap that is found in this research.
Overall it seems like the result of this regression stems from the difference in the opportunities that men and women receive and from the underrepresentation of females directors among the boards of the largest Dutch companies. This paper thus concludes that the fixed remuneration is influenced by gender.
4.2 Variable remuneration
The linear prediction of the ordinary least squares regression and the main results are presented below. The complete regression table can be found in appendix D.
15
Variable remuneration
𝜷𝒈𝒆𝒏𝒅𝒆𝒓 212176.6
p-value 0.199
Referring to the table, it is clear that the null hypothesis cannot be rejected. The p-value for the regression which has variable remuneration as a dependent variable is 0.199. Therefore, using a 95% confidence interval there is not sufficient statistical evidence to reject the null hypothesis. This implies that gender does not influence the level of variable remuneration among the largest Dutch firms. The result is quite unexpected and the possible explanations are mentioned below.
Again, Albanesi et al. (2015) provides an explanation for the result of this particular regression. They state that the risk aversion of both men and women is equal in a financial setting (2015). A person’s level of risk aversion indicates how opposed someone is to risk. The variable part of remuneration is relatively risky because it is based on a certain outcome and could potentially be low if the objectives are not met. Therefore, people with different levels of risk aversion accept different levels of variable pay. An individual with a very high degree of risk aversion does not accept a lot of variable pay and vice versa. Since the level of risk
3 0 0 0 0 0 3 5 0 0 0 0 4 0 0 0 0 0 4 5 0 0 0 0 5 0 0 0 0 0 F it te d v a lu e s 0 1 Gender
Variable remuneration
16
aversion between males and females is equal according to Albanesi et al. (2015), the variable remuneration should not be influenced by gender. This was proven by the regression with variable remuneration as dependent variable.
The article written by Perryman et al. (2016) is also able to shed some light on this issue. They show that according to agency theory, if men and women are equally productive, they should get paid the same amount of variable remuneration (Perryman et al., 2016). An increase in gender diversity among the boards of firms has proven to reduce the level of overall risk and enhance the performance of the company (Perryman et al., 2016). Perryman et al. (2016) find that female directors are just as capable (if not more) to run large firms as their male counterparts. Therefore, it makes sense that gender does not have an impact on the variable remuneration.
Benkraiem et al. (2017) came to similar conclusions as Perryman et al. (2016). Women are just as efficient as men and because they often have different ways of approaching problems, women can improve the performance of a company significantly (Benkraiem et al., 2017). Benkraiem et al. (2017) confirm that a gender-diverse board is proven to be more effective. Female directors reduce the agency costs of a firm by improving the quality of supervision, which causes a lower level of compensation and a better design of the remuneration policies (Benkraiem et al., 2017). If both men and women perform the same, there should not be a difference in their compensation (Benkraiem et al., 2017). This explains the result of this regression, which shows that gender does not influence variable remuneration.
There is no difference in variable compensation between male and female directors according to Bugeja et al. (2012). Once a person reaches the top in the business world, (s)he is going to be paid as such (Bugeja et al., 2012). This confirms that there is no influence of gender on the variable remuneration, which was found by the ordinary least squares regression. However, there is a large gap between the number of men and women who are capable of entering the highest ranks in this industry (Bugeja et al., 2012). Unfortunately, the proportion of female directors is a lot smaller than the proportion of their male counterparts (Bugeja et al., 2012). This is also represented by the sample used for the regression of the variable remuneration. There are only 91 females out of the 356 people included in the sample.
Overall it seems like the result of this regression can be explained by the similar qualities between men and women. They perform equally and since variable remuneration is based on performance, they receive the same amount of variable compensation.
17
4.3 Total remuneration
The linear prediction of the ordinary least squares regression and the main results are presented below. The complete regression table can be found in appendix E.
Total remuneration
𝜷𝒈𝒆𝒏𝒅𝒆𝒓 522753.6
p-value 0.029
From the table presented above it is clear that the p-value is smaller than 0.05. This implies that the null hypothesis is rejected for this regression. Gender does influence total remuneration, which makes sense because total remuneration is just the sum of fixed and variable remuneration. If the dummy variable of gender is equal to 1, the total remuneration increases with 522753.6. This means that a male director earns € 522,753.60 more in total than a female director.
Since the total remuneration is the sum of the fixed and variable remuneration, the result can also be explained by the same argument. The gender wage gap in the total compensation
6 0 0 0 0 0 8 0 0 0 0 0 1 0 0 0 0 0 0 1 2 0 0 0 0 0 F it te d v a lu e s 0 1 Gender
Total remuneration
18
exists for the same reasons as the gap in the fixed compensation. Therefore, the causes are not extensively described in this section again.
In short, the result of this regression most likely stems from the difference in the opportunities that men and women receive and from the underrepresentation of females directors among the boards of the largest Dutch companies. This paper thus concludes that the total remuneration is influenced by gender.
4.4 Multicollinearity tests
To provide some additional explanations for the results that are found through the regressions, two multicollinearity tests are performed. The tables with the main results and the graphs including the linear approximations are presented below. The complete regression tables can be found in appendix F and G.
Multicollinearity arises when one of the independent variables in the model can be estimated by one or more of the other independent variables in the model. This implies that a relationship between the independent variables of the model might be able to explain the results of the three regressions that are performed in this thesis.
The first multicollinearity test is analysing the relationship between the dummy variable of the possession of a college degree and the dummy variable of gender. Gender could possibly influence the level of education a person has obtained. Sometimes men and women do not have the same opportunities when it comes to schooling, especially in developing countries. In order to check if gender has any impact on whether someone possesses a college degree, a regression is conducted. The dummy variable of the possession of a college degree is regressed on the dummy variable of gender. The regression that is performed is presented below.
𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝑑𝑒𝑔𝑟𝑒𝑒𝑖 = 𝛽0+ 𝛽𝑔𝑒𝑛𝑑𝑒𝑟𝑔𝑒𝑛𝑑𝑒𝑟𝑖+ 𝑒𝑖
Again, a 95% confidence interval is used to assess the influence of gender. The null hypothesis states that there is no influence of gender on whether someone has obtained a college degree. If the p-value of the coefficient of gender is larger than 0.05, the null hypothesis is not rejected and is considered to hold. In this case there is not enough statistical evidence to believe that gender affects the dependent variable. The null hypothesis is represented as the following.
19
If the p-value is smaller than 0.05, the null hypothesis is rejected. In this case the dummy variable of gender does influence the possession of a college degree of an individual. This implies that there is enough statistical evidence to believe that gender influences the dependent variable. The accompanying hypothesis is presented below.
H1: 𝛽𝑔𝑒𝑛𝑑𝑒𝑟≠ 0
The results of the regression are presented below.
The possession of a college degree
𝜷𝒈𝒆𝒏𝒅𝒆𝒓 0.0385237
p-value 0.216
As is shown in the table, the p-value of the coefficient of gender is larger than 0.05. This means that gender does not impact the level of education of an individual. This result can be explained by the job title and the country of interest of this analysis. All the people in the sample of this investigation are referred to as director, which implies that they all have an executive position. This position involves a lot of responsibility and difficult tasks, which requires a certain level of knowledge. Therefore, most directors, whether they are male or
.9 .9 1 .9 2 .9 3 .9 4 F it te d v a lu e s 0 1 Gender
20
female, are highly educated and possess a college degree. Moreover, the Netherlands, where education is very accessible, is the country of interest in this examination. It is mandatory for all Dutch citizens to attend school until they are 18 years old. The Dutch school system has many different levels education, so there is a place for every student. Furthermore, students who do not have the means to afford education can apply for a student loan. The interest on these loans is minimal. Since schooling is very accessible in the Netherlands and most directors are highly educated, it is not surprising that gender does not affect whether a person possesses a college degree. Unfortunately, the relationship between these two independent variables cannot help explain the results of the previous three regressions.
The second multicollinearity test examines the relationship between tenure and whether someone possesses a college degree. Education is a determinant of the personal skills someone has acquired, and these skills might affect the length of the period someone is employed in the same company. A person who performs bad will most likely not employed for a very long time. This particular relationship can possibly give some new insights on the remuneration regressions. Therefore, tenure is regressed on the dummy variable of the possession of a college degree. The regression that is performed is presented below.
𝑡𝑒𝑛𝑢𝑟𝑒𝑖 = 𝛽0+ 𝛽𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝑑𝑒𝑔𝑟𝑒𝑒𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝑑𝑒𝑔𝑟𝑒𝑒𝑖 + 𝑒𝑖
Again, a 95% confidence interval is used determine the impact of the level of education on the dependent variable. The null hypothesis states that the possession of a college degree does not affect an individual’s time of employment. If the p-value of the coefficient of the dummy variable is larger than 0.05, the null hypothesis is not rejected and is considered to hold. In this case there is not enough statistical evidence to believe that the level of education affects tenure. The null hypothesis is represented as the following.
H0: 𝛽𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝑑𝑒𝑔𝑟𝑒𝑒 = 0
If the p-value is smaller than 0.05, the null hypothesis is rejected. In this case there is statistical proof that whether someone has a college degree determines his/her period of employment. The hypothesis is presented below.
H1: 𝛽𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝑑𝑒𝑔𝑟𝑒𝑒 ≠ 0
21
Tenure
𝜷𝒄𝒐𝒍𝒍𝒆𝒈𝒆 𝒅𝒆𝒈𝒓𝒆𝒆 0.7897583
p-value 0.494
Also for the second multicollinearity test the p-value of the coefficient is larger than 0.05. The influence of the level of education on tenure is therefore insignificant. The relationship between these to independent variables is also not able to explain any of the results obtained from the remuneration regressions.
To conclude, the multicollinearity tests did not provide any additional explanations of for the influence of gender on the several types of remuneration. Even though these tests did not give an very useful answer, it is still very important to check whether possible relationships between the independent variables could potentially explain the main results of the regressions.
5 Discussion
This section examines the results and explanations of the influence of gender on variable remuneration. First, the role of gender when committing moral hazard is discussed. The examination of difference in behaviour between men and women involving moral hazard
4 .6 4 .8 5 5 .2 5 .4 F it te d v a lu e s 0 1
The possesion of a college degree
22
is one of the issues of this paper. Second, the results of the regression of variable remuneration are questioned.
Moral hazard materializes when the interests of the owner and the director of the company do not align and the actions of the director are not completely observable (Eisenhardt, 1989). This thesis defines moral hazard as taking excessive risk. As mentioned before, the moral hazard problem can be solved by finding the appropriate remuneration policy for each individual (Eisenhardt, 1989). Specifically, if the compensation of the director is based on the outcome desired by the owner, the conflict of interests is solved and the moral hazard problem is prevented (Eisenhardt, 1989). The part of the remuneration policy that is based on outcome is the variable remuneration. This part of an individual’s compensation package is made up by bonuses and stock options granted. By use of this variable remuneration, the owners of the largest Dutch companies try to solve moral hazard and keep the directors motivated (Eisenhardt, 1989).
The influence of gender on variable remuneration is one of the topics this paper is investigating. If gender affects the variable remuneration, men and women potentially need a different structure of variable pay to prevent them from committing moral hazard. This implies that men or women could be more likely to commit moral hazard than their counterparts. To see if male and female directors act differently in the case of moral hazard and thus have different variable remuneration policies, variable remuneration was regressed on gender and control variables.
The results of this particular regression, which are discussed in section 4.2, show that gender does not have a significant impact on variable remuneration. This implies that men and women are motivated in the same manner. According to the result of the regression of variable compensation, men and women act in the same manner in a situation which allows for moral hazard. They take the same amount of excessive risk.
So, according to the regression results there exists no gender wage gap in variable pay. The article written by Benkraiem et al. (2017) creates some controversy around that fact. Benkraiem et al. (2017) state that women have acquired more skills and perform better on certain areas than men. Female directors often have had more education and have a better understanding of the expertise needed in the business world (Benkraiem et al., 2017). Moreover, women have a different approach at work and a unique world-perspective (Benkraiem et al., 2017). Female directors tend to improve the overall decision-making process because they are team players and are very good at problem-solving (Benkraiem et al., 2017).
23
As a result, women enhance the performance of a company and reduce the overall level of risk (Benkraiem et al., 2017).
Bugeja et al. (2012) states that once a person reaches the top in the business world, (s)he is going to be paid as such. However, if women perform better then men and variable remuneration is based on performance, why do women not receive more variable pay than men? Male and female directors are paid equal amounts of variable pay according to the results of the regression. However, this does not mean that there exists no gender wage gap. If one were to account for the director’s performance and reduced level of risk, women might still get paid less variable remuneration than men. Does this imply that male directors can be more likely to commit moral hazard? Since variable remuneration is supposed to keep the directors in line and men receive more of it, this can be necessary to prevent male directors from taking excessive risk.
To conclude, more research is needed to fully assess how gender plays a role in the problem of moral hazard.
6 Conclusion
Is the solution to moral hazard, an individual’s appropriate remuneration policy, influenced by gender? This is the question that this thesis attempted to find an answer to. Fixed, variable and total remuneration are regressed on gender and several control variables. The examination of the multiple types of remuneration showed that fixed and total remuneration are both influenced by gender. Male directors receive more of these types of compensation then female directors, thus a gender wage gap exists in these areas. Variable remuneration, which is supposed to motivate the director who receives the pay, is not affected by gender. This thesis concludes that men and women are motivated in the same manner, and thus treat moral hazard similarly.
The analysis done in this paper is bounded by some limitations. The initial idea was to use only CEO’s for the data sample of the regression. Unfortunately, there are not many female CEO’s among the largest Dutch firms and the job title ‘Director’ was used to select the people in the sample. Next, there does not exist a common database with all the information needed for this research. Therefore, the information of the remuneration policies was looked up for every company individually. Since this is cumbersome, the sample size is limited to a few observations. Finally, due to the lack of a relevant dataset, some variables were omitted from the regressions. The initial idea was to also include variables like the number of promotions a person has had, if the individual is married or not, the number of children, if a female director
24
went on maternity leave, the number of days worked, etc. However, all this data had to be looked up for every observation in the sample individually. Due to time limitations, this idea was dropped.
Future research could attempt to solve the limitations of this thesis. Using a larger and more versatile dataset might give different results. Moreover, the level of performance of the director should be measured. This allows for a more profound examination of the relationship between performance and variable pay. Variable remuneration between male and female directors can be compared more accurately if performance measures are used in the research. To conclude, more research is needed to fully determine the impact of gender on the remuneration policies of the directors of the largest Dutch companies.
25
Reference list
Albanesi, S., Olivetti, C., & Prados, M. J. (2015). Gender and Dynamic Agency: Theory and Evidence on the Compensation of Top Executives. Federal Reserve Bank of New
York Staff Reports, 718, 1-44.
Ali, S., Liu, B., & Su, J. J. (2018). Women on board: Does the gender diversity reduce default risk? Griffith Business School, 1-55.
Baixauli-Soler, J. S., Belda-Ruiz, M., & Sanchez-Marin, G. (2015). Executive stock options, gender diversity in the top management team, and firm risk taking. Journal of
Business Research, 68(2), 451-463.
Benkraiem, R., Hamrouni, A., Lakhal, F., & Toumi, N. (2017). Board independence, gender diversity and CEO compensation. Corporate Governance: The International Journal
of Business in Society, 17(5), 845-860.
Bugeja, M., Matolcsy, Z., & Spiropoulos H. (2012). Is there a gender gap in CEO compensation? Journal of Corporate Finance, 18(4), 849-859.
Bugeja, M., Matolcsy, Z., & Spiropoulos, H. (2016). The Association Between
Gender-Diverse Compensation Committees and CEO Compensation. Journal of
Business Ethics, 139(2), 375-390.
CBS (2018). Arbeidsdeelname, kerncijfers. Statline. Retrieved from
https://opendata.cbs.nl/statline/#/CBS/nl/dataset/82309ned/table?ts=1522846191779
Consuelo Pucheta-Martìnez, M., Bel-Oms, I., & Olcina-Sempere, G. (2016). Corporate governance, female directors and quality of financial information. Business Ethics; A
European Review, 25(4), 363-385.
Eisenhardt, K. M. (1989). Agency Theory: An Assessment and Review. The Academy of
Management Review, 14(1), 57-74.
Faccio, M., Marchica, M. T., & Mura, R. (2016). CEO gender, corporate risk-taking, and the efficiency of capital allocation. Journal of Corporate Finance, 39, 193-209.
Geiler, P., & Renneboog, L.(2015). Are female top managers really paid less? Journal of
Corporate Finance, 35, 345-369.
Hölmstrom, B. (1979). Moral Hazard and Observability. The Bell Journal of Economics,
10(1), 74-91.
Koch, A. K., & Peyrache, E. (2008). Moral hazard contracts: Does one size fit all? Economics
26
Ministerie van Onderwijs, Cultuur en Wetenschap (2018). Deelnemers wetenschappelijk onderwijs. Onderwijs in cijfers. Retrieved from
https://www.onderwijsincijfers.nl/kengetallen/wetenschappelijk-onderwijs/deelnemerswo
Perryman, A. A., Fernando, G. D., & Tripathy, A. (2016). Do gender differences persist? An examination of gender diversity on firm performance, risk, and executive
compensation. Journal of Business Research, 69(2), 579-586.
Rijksoverheid (2016). Opnieuw wettelijk streefcijfer voor vrouwen en mannen in bestuur ondernemingen. Retrieved from
https://www.rijksoverheid.nl/onderwerpen/vrouwenemancipatie/nieuws/2016/01/15/o pnieuw-wettelijk-streefcijfer-voor-vrouwen-en-mannen-in-bestuur-ondernemingen Stevens, D. E., & Thevaranjan, A. (2010). A moral solution to the moral hazard problem.
Accounting, Organizations and Society, 35(1), 125-139.
Top-100 Corporate Vrouwen (2017). Management Scope. Retrieved from
27
Appendix A – Summary of all the variables
Variable Obs Mean Std. Dev. Min Max
Turnover 356 3.28e^07 3.55e^07 1.02e^07 1.48e^08 NumEmpl 356 61611.78 91043.27 25 369000 Tenure 356 5.374298 5.554198 0.25 50 DumGender 356 0.7744382 0.436822 0 1 Age 356 56.40449 8.837142 32 75 DumCollege 356 0.9297753 0.255885 0 1 BaseSalary 356 606278.2 1223361 0 1.35e^07 IncenPay 356 302872.3 113390 0 1.38e^07 Stocks 356 133362.5 588927.3 0 9100045 TotVarComp 356 436234.7 1341277 0 1.38e^07 TotComp 356 1042471 1947244 0 1.51e^07
Appendix B – Information of the dummy variable for gender
DumGender Freq. Percent Cum.
0 91 25.56 25.56
1 265 74.44 100.00
Total 356 100.00
Appendix C – Regression of fixed remuneration on gender, age, the possession of a college degree, tenure and the size of the company
Source SS df MS Number of
obs =
356 Model 1.2528e^13 5 2.5056e^12 F(5, 350) = 1.69 Residual 5.1877e^14 350 1.4822e^12 Prob > F = 0.1361 Total 5.3130e^14 355 1.4966e^12 R-squared = 0.0236
Adj. R-squared = 0.0096 Root MSE = 1.2e^06
Base salary Coef. Std. Err. t P> ǀ t ǀ 95 % Conf. Interval
Turnover 0.0010213 0.0018479 0.55 0.581 -0.0026121 0.0046557 Tenure -7075.387 12065.08 -0.59 0.558 -30804.57 16653.79 DumGender 310444.1 149148.9 2.08 0.038 17103.2 603785 Age -663.7311 7566.936 -0.09 0.930 -15546.12 14218.65 DumCollege 447683.9 253969.7 1.76 0.079 -51814.82 947182.6 _cons 952.4499 476127.8 0.00 0.998 -935478.9 937383.8
28
Appendix D – Regression of variable remuneration on gender, age, the possession of a college degree, tenure and the size of the company
Source SS df MS Number of
obs =
356 Model 5.1203e^12 5 1.0241e^12 F(5, 350) = 0.57 Residual 6.3353e^14 350 1.8101e^12 Prob > F = 0.7263 Total 6.3865e^14 355 1.7990e^12 R-squared = 0.0080
Adj. R-squared = -0.0062 Root MSE = 1.3e^06
TotVarComp Coef. Std. Err. t P> ǀ t ǀ 95 % Conf. Interval
Turnover -0.0008729 0.0020421 -0.43 0.669 -0.0048892 0.0031434 Tenure -314.6839 13333 -0.02 0.981 -26537.56 25908.19 DumGender 212176.6 164823 1.29 0.199 -111991.4 536344.7 Age -8983.239 8362.143 -1.07 0.283 -25429.61 7463.132 DumCollege -34895.48 280659.3 -0.12 0.901 -586886.4 517095.4 _cons 847719.1 526163.9 1.61 0.108 -187121.7 1882560
Appendix E – Regression of total remuneration on gender, age, the possession of a college degree, tenure and the size of the company
Source SS df MS Number of
obs =
356 Model 2.5018e^13 5 5.0037e^12 F(5, 350) = 1.33 Residual 1.3211e^15 350 3.7744e^12 Prob > F = 0.2526 Total 1.3461e^15 355 3.7918e^12 R-squared = 0.0186
Adj. R-squared = 0.0046 Root MSE = 1.9e^06
TotComp Coef. Std. Err. t P> ǀ t ǀ 95 % Conf. Interval
Turnover 0.0001447 0.0029488 0.05 0.961 -0.005655 0.005944 Tenure -7389.096 19253.24 -0.38 0.701 -45255.69 30477.5 DumGender 522753.6 238009.1 2.20 0.029 54645.55 990861.7 Age -9650.594 12075.18 -0.80 0.425 -33399.63 14098.44 DumCollege 413370.1 405280.2 1.02 0.308 -383720.8 1210461 _cons 848310 759796 1.12 0.265 -646030.1 2342650
29
Appendix F – Multicollinearity test: Regression of the possession of a college degree on gender Source SS df MS Number of obs = 356 Model 0.100529648 1 0.100529648 F(5, 350) = 1.54 Residual 23.1438524 354 0.065378114 Prob > F = 0.2158 Total 23.244382 355 0.065477132 R-squared = 0.0043 Adj. R-squared = 0.0015 Root MSE = 0.25569
DumCollege Coef. Std. Err. t P> ǀ t ǀ 95 % Conf. Interval
DumGender 0.0385237 0.0310669 1.24 0.216
-0.0225751
0.0996226 _cons 0.9010989 0.0268037 33.62 0.000 0.8483843 0.9538135
Appendix G – Multicollinearity test: Regression of tenure on the possession of a college degree Source SS df MS Number of obs = 356 Model 14.4979438 1 14.4979438 F(5, 350) = 0.47 Residual 10936.9394 354 30.895309 Prob > F = 0.4938 Total 10951.4373 355 30.8491192 R-squared = 0.0013 Adj. R-squared = -0.0015 Root MSE = 5.5584
Tenure Coef. Std. Err. t P> ǀ t ǀ 95 % Conf. Interval
DumCollege 0.7897583 1.152888 0.69 0.494 -1.477613 3.05713 _cons 4.64 1.111671 4.17 0.000 2.45369 6.82631
