Fair wage determination from an employee perspective

(1)

Fair Wage Determination from an Employee Perspective

Master Thesis July 15, 2017

Author: M. Goedman

Program: MSc Business Economics

Track: Managerial Economics and Strategy Institution: Univeristy of Amsterdam Student number: 11371757

Supervisor: Dr. T. Buser ECTS: 15

Abstract

This study makes use of data from a vignette study and from the Research Centre for Education and the Labour Market (ROA). The aim is to shed light on the effect of competition

in the labor market on employees’ perception of what constitutes a fair wage and how does age and gender influence this perception. For companies, it is important to know which factors determine whether a wage seems fair or unfair, because the existing literature shows

that workers reduce their effort when they perceive that they receive less than a fair wage. Companies also want to be attractive to new employees, but do not want to pay too much. This study will demonstrate that competition in the labor market definitely plays a role in the

determination of what constitutes a fair wage. Interesting for further research is to focus on the side of the employers.

(2)

2

Statement of Originality

This document is written by Student Malou Goedman who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

1. Introduction

Wages represent one of the main causes of job dissatisfaction (Klopper, Coetzee, Pretorius & Bester, 2012). Results from a 2015 research study by Loonwijzer in the Netherlands show that 40% of workers are dissatisfied with their jobs. This 40% expressed the greatest

dissatisfaction regarding salary (68%). Also, according to Clark (2001) remuneration, together with job security, are the most important aspects contributing to job satisfaction. However, several companies, such as Semco and Incentro, have shown that when workers were able to choose their own wages they proposed wages that were the same as what the company would initially have offered them (Bueters, 2015). Additionally, Groot and Maassen van den Brink (1999) argue that most empirical studies find that higher wages do not increase job

satisfaction. This results in the interesting question: What drives the perception of a fair wage?

Milkovich, Newman and Milkovich (2002) and many other researchers stress the importance of paying fair wages to obtain workers’ effort. “Effort” can be defined as the amount of time and energy a worker spends on work activities (Campbell & Pritchard, 1976). Fehr, Kirchsteiger and Riedl (1993) and Akerlof (1982) observe reciprocity in the relationship between employer and employee in a free market setting and suggest that fair wage offers are reciprocated with more effort, and vice versa.

This leads the question: what determines “fair wages” for employees, and the perception of fair wages under different circumstances? When students look for a job after finishing school or university, do they take labor market conditions into account when

deciding what constitutes a fair wage? This leads to the following research question: “What is

the effect of competition in the labor market on the perception of a fair wage and how does age and gender influence this perception?” In this paper, a competitive labor market is

regarded as a market in which the number of available potential new employees (unemployed workers) is high compared to the number of required new employees (vacancies).

Prospective employees may be more eager to obtain a job and therefore have lower expectations regarding their prospective wages, because otherwise they could possibly find themselves unemployed. On the other hand, from a company’s point of view, it is interesting and important to know the optimal salary of an employee. Companies do not want to pay too much in a competitive labor market, but in a non-competitive labor market, they need to ensure that they pay enough to be attractive. Surprisingly, Rees (1993) argues that employers never take the supply-and-demand conditions of the labor market into account when

(5)

5

determining wages. If people consider that a fair wage depends on labor market conditions, then it would be better for employers to take this into account.

In this study, satisfaction with wages is compared in two hypothetical job scenarios, one where there is a high demand for labor and the other where there is a lower demand for labor. The objective of this study is to examine whether people are less critical of wage satisfaction in a situation where there is a low demand for labor (a competitive labor market). The outcomes of this study will make an important contribution for both employers and employees. Employers care about the costs arising from wages but also about the productivity of employees. For an optimal balance between the two, it is important to know what

employees take into account when determining what constitutes a fair wage. On the other hand, employees want to earn as much money as possible. The labor market can affect how critical they are in determining a fair wage.

This study shows that competition in the labor market definitely plays a role in the determination of a fair wage. Several factors that are important for determining whether a wage is fair or unfair are discussed. Furthermore, the study shows, both in the literature as well as in the results of the vignette study, that the same wage is rated as less attractive by males than by females, particularly in a non-competitive market. In contrast to the literature, no evidence is found that age determines what is regarded as a fair wage.

This paper is structured as follows. Chapter two provides an overview of the literature relevant to the research question, and the hypotheses are described. Following the literature overview, the methodology is described and discussed. This is followed by the results of the research, which are structured according to the hypotheses. The next section discusses the results, and the final chapter presents an overall conclusion, which answers the research question.

(6)

6

2. Literature Review

This chapter reviews the literature relevant to the research question described in the

introduction. The chapter is divided into sections on neoclassical wage theory, fair wages and productivity, the fairness rule and the unemployment rate, differences in age, differences in gender, other factors important to fairness, the literature related to the hypotheses, and the contribution this study will make to the literature.

2.1 Neoclassical Wage Theory

The neoclassical wage theory emphasises that a worker’s utility is based on his or her own wages and hours of work. According to this theory, the utility function is not related to the wages and hours of others. In fact, wages seem to be very diverse when explained in terms of the neoclassical wage theory. Besides the assumption that labor is homogenous, and that all workers have equal productivity capacity, there is another very important factor for the differences in real-world situations, namely fairness. Fairness, when based on comparison with others, always tends to be positive rather than negative (Rees, 1993).

Rees (1993) argues that “wages are not fully determined by markets”. He finds that employers never argue for taking supply-and-demand conditions in the labor market into account. Labor markets leave space for bargaining, because the labor market is imperfect and leaves an undefined zone. This also creates discretion on the part of wage setters, which is more important than emphasised in the literature on labor economics. The concept of fairness is the key to understanding how discretion is applied (Rees, 1993).

The question is whether the concept of fairness belongs to the supply or the demand side of wage theory. The supply-side of wage theory encompasses the supply of both new and experienced workers. In determining whether wages are fair or unfair, experienced workers take their previous experience into account. When this previous work experience makes a greater contribution than reflected by the offered wage, the wage seems unfair. However, when such experienced workers demand wages that are too high, an employer will move on to the next applicant. Both employers and employees have to make comparisons with the

average level of wages in the market. When a salary is too low and workers think their salary is unfair, they will not give their full effort (Rees, 1993).

According to Rees (1993) labor supply is derived from a utility function that must include the wages of one or more other people, in addition to real wages and leisure. There is a great deal of theoretical literature in behavioural economics on inequality aversion. For example, Hamermesh (1975) writes: “We assume that the utility-maximizing individual

(7)

7

derives pleasure from having a wage above the average at any given level of the average wage of some comparison group that he can readily observe”. Remarkably, in wage comparison, people compare themselves with others close by. The effect of interpersonal comparison on the utility function should not be underestimated (Hamermesh, 1975). Reese (1993) mentions that the effect of the comparative wage function is greater than that of a person’s own wages. The following example makes this clearer. On Tuesday a worker is very happy because his boss tells him that his wages will increase by €20 per month. On Thursday he discovers that the rest of the company’s employees’ wages will increase by €30 per month. Now that he knows this, he is less happy on Thursday than he was on Tuesday. It is also more than likely that he is less happy than he was on the previous Friday before his wages were increased. He prefers that no one gets an increase rather than that he gets a smaller increase than others (Rees, 1990).

Card, Mas, Moretti and Saez (2012)propose a new strategy for evaluating the effect of relative pay comparisons. They argue that the relationship between job satisfaction and

relative pay is nonlinear. “Workers with salaries below the median for their pay unit and occupations report lower pay and job satisfaction, while those earning above the median report no higher satisfaction”.

In the wage setting process, directors refer to surveys of salaries in comparable

companies. It is not always possible for an employer to meet equitable comparisons. When an employee stays in a job but is not satisfied with the salary, he or she can withhold effort, at the risk of being dismissed. It is important for an employer to take interpersonal comparison into account. When an employer hires a new person who asks for a higher wage than what current employees receive, and this higher pay is necessary to recruit this particular person, it might disrupt the structure of the company (Rees, 1993).

Clark and Oswald (1996) and Loewenstein, Thompson and Bazerman (1989) also find strong evidence that wage comparison is an important factor for overall job satisfaction. Clark and Oswald (1996) show from a study of 10,000 British individuals that income comparison has a significantly negative effect on overall job satisfaction.

2.2 Fair Wages and Productivity

Employees use a range of criteria to examine the fairness of their reward. Research shows that it is hard to satisfy all employees in a company with an absolute reward. In practice, everyone wishes to earn more and looks for arguments to justify higher wages as fair (Prins, Brouwers, Mertens & Segers, 2010). “The fundamental assumption of fair-wage models is that workers

(8)

8

may withhold effort if their actual wage falls short of what they perceive to be a fair wage rate” (Chouliarakis & Correa-López 2014). In both the papers of Prins et al. (2010) and Chouliarakis and Correa-López (2014), the authors argue that a fair wage often depends on the wages of other workers, the wages of other firms, past wages and current labor market conditions.

Akerlof and Yellen (1990) conceived “the fair wage-effort hypothesis”, which explains workers’ conception of what constitutes a fair wage. If the actual wage is less than the worker’s view of a fair wage, he or she will supply a corresponding share of normal effort. This can be described in terms of the following formula:

effort supplied = min (𝑎𝑐𝑡𝑢𝑎𝑙 𝑤𝑎𝑔𝑒 (𝑤)_{𝑓𝑎𝑖𝑟 𝑤𝑎𝑔𝑒(𝑤}_∗₎ , 1). The normal effort is denoted as 1.

Akerlof and Yellen (1990) show in their paper that workers reduce effort when they perceive that they are paid less than a fair wage. The reduction in effort is proportional to the gap between a fair wage and the actual wage. They state: “When people do not get what they [think they] deserve, they try to get even”. Blinder and Choi (1990) tested competing theories of wage stickiness by interviewing managers from firms in New Jersey and Pennsylvania. They found that a majority of 79% of responding managers confirmed that workers

considered wages as unfair when the managers took advantage of labor market slack to reduce wages. They also found that a majority of managers (95%) believed that when workers

considered wages to be unfair, this led to a drop in morale and a concomitant decline in work effort.

2.3 The Fairness Rule and the Unemployment Rate

The existence of unemployment can also be explained by the fair wage hypothesis of Akerlof and Yellen (1990). In short, they assume that when a fair wage exceeds the market-clearing wage (supply is equal to demand), there is a fixed supply of labor, which is independent of wages. The relationship between the marginal cost and the marginal product of effective labor determines the labor demand of firms. There are two options.

First, the marginal product of effective labor is lower than the fair wage (α < 𝑤∗_{). The}

marginal cost of effective labor is at least as large as the fair wage. In this case, the marginal cost of effective labor exceeds the marginal product. This implies that it cannot be profitable for a firm, so the demand for labor is zero and there is a great deal of unemployment.

The second option is that the marginal product of effective labor is bigger than the fair wage (α > 𝑤∗_{). In this case, there is unemployment; this means that the aggregate supply of}

(9)

9

According to Akerlof and Yellen (1990), a firm will choose a wage level that minimises the wage per unit of effort, i.e. the marginal cost of effective labor. The marginal cost of effective labor is equal to a fair wage for the worker. When the marginal cost is lower than the marginal product of effective labor, this results in an aggregate demand for labor. Under these

circumstances, there exists competition between firms for workers. This will force firms to pay workers wages higher than fair wages (𝑤∗_{). Firms will set wages between what}

constitutes fair wages (𝑤∗_{) and the marginal product of effective wages (α); between these}

wages the demand for labor will be infinite. At full employment, all firms will pay the

market-clearing wage (w=α) because the demand for labor is infinitely elastic when the wage 𝑤 = 𝛼.

The above equilibrium infers that w* reflects fairness. If workers consider their wages as unfair, the reason for reducing effort is not that they are better off doing so, but that they are angry. Angry people behave in a way that does not maximise their utility (Akerlof & Yellen, 1990).

When the unemployment rate is high, people may be grateful to be employed and set their expectations of a fair wage at a low level. Conversely, when the unemployment rate is low, people are likely to consider being employed as normal and so may set their expectations of a fair wage at a high level (Akerlof & Yellen, 1990).

Akerlof and Yellen (1990) generalize many assumptions in their model. For example, a fair wage may depend on several factors, such as past wages of the workers, the wages of other workers and the profits accruing to the firm’s owners. Besides, certain types or labor might be complementary or substitute labor, and the effort they represent may not multiple the production function. This may lead to a nonlinear production function:

𝐹 = 𝐴₀+ 𝐴₁(𝑒1𝐿1) + 𝐴2(𝑒2𝐿2) − 𝐴11(𝑒1𝐿1)2+ 𝐴12(𝑒1𝐿1)(𝑒2𝐿2) − 𝐴22(𝑒2𝐿2)2

The model consists of four key behavioural assumptions, namely endowments, tastes, technology and fairness. In this thesis, the focus is on the assumption of fairness. Fairness implies three premises. First, there is the fair wage-effort hypothesis for the two labor groups: 𝑒1 = min (𝑤_𝑤1

1∗, 1) and 𝑒2 = min (

𝑤2

𝑤2∗, 1), which has been discussed earlier. Second, fairness entails the determination of a fair wage (w*) by using a reference wage, which means that a fair wage for Group 2 depends on the wages received by Group 1, and vice versa. A fair wage is also determined by market conditions. In contrast to Kahneman, Knetsch, and Thaler

(10)

10

(1986), Akerlof and Yellen (1990) argue that market conditions influence what constitutes a fair wage. Accordingly, the fair-wage formula is as follows:

𝑤₁∗ _{= 𝛽𝑤}

2+ (1 − 𝛽)𝑤1𝑐

𝑤₂∗ _{= 𝛽𝑤}

1+ (1 − 𝛽)𝑤2𝑐

The market-clearing wages for Groups 1 and 2, expressed by 𝑤₁𝑐_{and 𝑤}

2𝑐, are defined

according to Akerlof and Yellen (1990) as “those wages that would clear the market for labor of a given type in a simple neoclassical economy where workers exert full effort regardless of the wage they are paid”. If this holds, which means 𝑒1 = 𝑒2 = 1, labor demand functions take

the following forms:

𝐿₁ = 𝑎₁− 𝑏₁𝑤₁+ 𝑐₁𝑤₂ 𝐿2 = 𝑎2+ 𝑏2𝑤1− 𝑐2𝑤2

Akerlof and Yellen (1990) assume that the effect of the “own” wage is stronger than the effect of the “cross” wages, which means in these formulas that b1>c1 and b2<c2.

Finally, the third premise of fairness is that when the profits of a firm are not affected by paying a fair wage, firms still have a small preference for paying a fair wage.

Kreickemeier and Nelson (2006) agree with Akerlof and Yellen (1990) that a fair wage is determined by either the market wage of the other group, or the “own” wage if workers become separated from their current job. Kreickemeier and Nelson (2006) developed a model that follows the model of Akerlof and Yellen (1990), but this model includes two sectors instead of one. In this case, the economy produces two types of goods, X and Y, and uses the factors of unskilled (L) and skilled (H) labor.

𝑤_𝐿∗ _{= 𝜃𝑤}

𝐿+ (1 − 𝜃)(1 − 𝑈𝐿)𝑤𝐿

𝑤_𝐻∗ _{= 𝜃𝑤}

𝐻+ (1 − 𝜃)(1 − 𝑈𝐻)𝑤𝐻

UL and UH are the factor-specific rates of unemployment and θ is the weight of other factors

which have influenced the determination of a fair wage. In a perfectly competitive market, Kreickemeier and Nelson (2006) assume that the wages for unskilled workers would be lower than the wages for skilled workers. If this holds, the following will be true in equilibrium:

UL > UH = 0

𝑤_𝐻> 𝑤_𝐻∗ _{> 𝑤}

𝐿 = 𝑤𝐿∗

(11)

11

Where 𝜀𝐿 expresses the effort of an unskilled worker and 𝜀𝐻 expresses the effort of a skilled

worker, these formulas describe full employment for skilled workers, but for unskilled workers the rate of unemployment is positive. This means that a fair wage is binding only for unskilled workers, and in terms of the last equation both unskilled and skilled workers provide normal effort. Kreickemeier and Nelson (2006) describe the relationship between the wage differential and the rate of unemployment in terms of the following formula: ω ≡ 𝑤𝐿/𝑤𝐻.

This is called the fair wage constraint, which describes conditions of equilibrium between the rate of unemployment of unskilled workers and the relative gross wages of skilled and

unskilled workers (Kreickemeier & Nelson, 2006). Applying the derivative of the fair wage constraint shows that when the unemployment rate for unskilled workers increases, this leads to relatively lower wages. This can be explained by the fact that the fair wage needed to obtain normal effort from (unskilled) workers is lower when the unemployment rate is higher (Kreickemeier & Nelson, 2006).

2.4 Differences in Age

The perception of the fairness of wages seems to be related to workers’ wishes to equal or better wages achieved in the past. This can explain the differences in perception of what constitutes a fair wage between younger and older employees. Young employees have not yet achieved top wages and want to receive equal pay for equal work. Older employees want to be rewarded commensurate to their seniority level, and care about maintaining acquired rights (Prins et al., 2010).

2.5 Differences in Gender

Clark (1997) uses data from a British Household Panel Survey to show that the female-male pay satisfaction differential largely rises with age. Clark (1997) separates job satisfaction into three categories: job satisfaction with the work itself, job satisfaction with pay, and overall job satisfaction. He found for all three categories significantly lower job satisfaction levels for men than for women. The biggest difference in job satisfaction was in respect of remuneration (Clark, 1997). In a study with a sample of British workers, Groot and Maassen van den Brink (1998) found that wages have a statistically significant positive effect on male job

satisfaction, but that for women this effect was insignificant. Sloane and Williams (2000) also used data from British workers and found for both female and male workers a significantly positive relationship between job satisfaction and wages, although they also found a stronger positive effect for male workers than for female workers.

(12)

12

The labor inspectorate report of 2006 (Erdem & Hoeben, 2008) shows that females earn on average 77% of the gross hourly earnings of males. The difference is higher for skilled workers than for low-skilled workers (26% versus 15%). This is also reflected in high-ranking functions. The difference in earnings between women and men in high-high-ranking functions is higher than the difference in earnings between women and men in low-ranking functions (22% versus 0%). The difference in earnings is also bigger in full-time functions (20%) than in part-time functions (4%). Finally, the report shows differences between working sectors. The biggest differences in earnings are found in corporate services (29%), repair consumer products and trade (28%), and health and social care (25%). The smallest difference is found in the catering sector (9%).

One possible explanation for the differences in earnings between the genders may be differences in respect of negotiation. Bowles, Babcock and Lai (2007) found evidence for differences in initiating negotiations. Women are less inclined than men to negotiate when there are male evaluators, owing to nervousness. They also found that male evaluators punish female candidates more than male candidates for initiaing negotiations. These differences could be explained by perceptions of being “nice” or “demanding”. Female evaluators make no distinction between males and females that could be regarded as punitive, and no gender distinctions were found in negotiations when the evaluator was a female (Bowles, Babcock & Lai, 2007).

Recent research by Babcock, Recalde, Vesterlund and Weingart (2017) found another explanation for the differences between genders. They show that barriers to the advancement of women in organizations and society as a whole are due to differences in the allocation of tasks with low promotability. Relative to men, women progress more slowly in organizations because they are allocated more non-promotable tasks. They found evidence that: “women are more likely to volunteer, more likely to be asked to volunteer, and more likely to accept direct requests”. These differences even exist when there are no gender differences in ability and preferences.

2.6 Other Factors Important to Fairness

Kahneman, Knetsch, and Thaler (1986) show that the perception of fairness is also dependent on the reason for a cut in wages. Just like Blinder and Choi (1990), discussed earlier, they argue that if firms want to cut pay because of declining labor market conditions, this will be seen as unfair, while cutting pay because of the threat of bankruptcy will be seen as more acceptable. These concerns affect the effort choices of workers, because workers will exert

(13)

13

more effort if wages are considered fair. Because of incomplete contracts in employer-employee relations, it is important to pay a fair wage. When workers consider their pay to be fair, they will exert more effort, but this also leads to higher payoffs in response to wage increases for the firm(Akerlof 1982; Akerlof & Yellen 1990).

2.7 Hypotheses

Four hypotheses follow from the discussed literature. The first concerns the main factor that drives the perception of a fair wage. According to Kreickemeier and Nelson (2006) and Akerlof and Yellen (1990), a fair wage is determined by either the relative wage or the wage which a worker could expect outside their current job, taking into account that they might be unemployed with a probability that is equal to the factor-specific rate of unemployment. This second factor is also known as the competitiveness of the labor market. Prins et al. (2010) and Chouliarakis and Correa-López (2014) include past wages as an important factor in

determining what constitutes a fair wage.

Hypothesis 1: The competitiveness of the labor market is one of the factors that drives the perception of a fair wage.

Second, Prins et al. (2010) argue that in practice everyone wishes to earn more and looks for arguments to justify a higher wage as fair. By contrast, Groot and Maassen van den Brink (1999) argue that most empirical studies find that higher wages do not increase job satisfaction. The model of Akerlof and Yellen (1990) assumes that a group with a higher wage rate is fully employed while a group with a lower wage rate experiences some

unemployment. This means that when there is some unemployment, wages are lower. Also, Kreickemeier and Nelson (2006) mention in their paper that the fair wage constraint shows that when the unemployment rate for unskilled workers increases, this leads to relatively lower wages. This can be explained by the fact that the fair wage needed to obtain normal effort by the (unskilled) worker is lower when the unemployment rate is higher.

Hypothesis 2: People in a competitive labor market are more easily satisfied with a low wage.

Third, there are differences in gender. The literature discussed earlier shows that the wages of females are lower than the wages of males. Wages have a significantly positive effect on job satisfaction for both males and females, but this effect is stronger for male workers. Groot and Maassen van den Brink (1998) only found this effect for male workers. The differences in wages may be due to differences in respect of negotiation.

(14)

14

Hypothesis 3: Females in both a competitive and non-competitive labor market are more easily satisfied with a low wage.

Finally, the fourth hypothesis deals with differences in age. Prins et al. (2010) found evidence that the perception of wages is related to past achieved wages. This explains the difference in perception of what constitutes a fair wage between younger and older employees.

Hypothesis 4: Older (30+) participants, when compared to younger participants, value past achieved wages more often as an important factor in the determination of fair wages.

2.8 Contribution to the Literature

As can be seen from the overview of the literature, there is a great deal of theoretical literature on how people achieve a fair wage. What is missing is that in certain situations the perception of what constitutes a fair wage can be adjusted to specific situations. Particularly, following the financial crisis of 2007-2008, unemployment was high and it was hard for students to find a job immediately after graduating. During this time, people were probably more easily satisfied with finding a job and the proposed salary of this job. According to Rees (1993), employers do not take into account supply-and-demand conditions in the labor market when setting wages. Additionally, Kahneman, Knetsch, and Thaler (1986) and Blinder and Choi (1990) conclude that cutting wages because of declining labor market conditions is seen as unfair by employees. These conclusions are in contrast to the general opion among students. It is important, especially for employers, to know what drives a fair wage. Wages are major expenditures for companies. It is important for companies not to pay too much in a non-competitive labor market, but in a non-competitive labor market, companies need to pay enough to be attractive. Besides that, the literature shows that workers reduce effort when they perceive that they are being paid less than a fair wage. Therefore, a vignette study was conducted and data from ROA was used to examine the effect of competition in the labor market on the perception of what constitutes a fair wage.

(15)

15

3. Methodology

This section provides an overview of the methodology used in this study. First, the method of data collection of the vignette study is described. In this part, the advantages and

disadvantages of this method are discussed, and the method used to obtain the results explained. The data of school graduates obtained from ROA is then discussed.

3.1 Collecting Data

To test the hypotheses of this thesis respondents were asked to fill out a vignette

questionnaire. Research methods such as interviews and questionnaires are less suited to determining attitudes and behavior patterns because self-reporting might be unreliable and one-sided. This led to less validity and varied results. According to Alexander and Becker (1978), a vignette comprises “short descriptions of a person or a social situation which contains precise references to what are thought to be the most important factors in the decision-making or judgment-making processes of respondents”.

The vignette questionnaire in this study contains two different labor market scenarios. In both scenarios, the job circumstances are closely related to the needs and wishes of a job seeker. The differences between the scenarios are the labor market conditions and the wage levels. The differences between the sectors are important because these factors determine the unemployment rate and what constitutes fair wages. One of the scenarios is a very

competitive labor market where it is difficult to find a job and the other scenario is a non-competitive labor market where it is easier to find a job. The proposed wages are different in the two situations: in the non-competitive labor market the wage is higher than in the

competitive labor market. The different wages could give extra insight into whether people in a competitive labor market are more easily satisfied with a low wage. The proposed wages in the two situations are based on starting wages in a non-competitive labor market and in a competitive labor market. Table 1 shows the differences in both situations (non-competitive and competitive). Appendix A1 gives a full description of the job choice scenarios and Appendix A2 gives the corresponding questions.

(16)

16

In this study, a within-subjects design is used, which means that all participants took part in both situations. According to Charness, Gneezy and Kuhn (2012), the within-subjects design has three advantages: “First, their internal validity does not depend on random assignment. Second, in many frameworks, they offer a substantial boost in statistical power. Finally, they are more naturally aligned with most theoretical mindsets: a theorist is likely to imagine an agent in a market reacting to a price change, not two agents in separate markets with different prices”. However, it has also a disadvantage, namely the so-called “order effects”. This means that the answers of the participants in the first situation may affect the answers in the second situation. To control for order effects, the two situations were assigned randomly to the participants. A Whitney test will check if the randomization was successful. The Mann-Whitney test compares whether the means differ significantly between the groups. If no significant difference is found, the randomization was successful.

Because all participants took part in both vignettes, different wages were proposed to the participants in each situation. This made the participants aware of looking at the salary

amount in different labor market situations. Because the same proposed wage was proposed in Situation X

Non-competitive situation

Situation Y Competitive situation You know that three other recent graduates

are also applying for the job. You know that there are many unfilled vacancies at other companies with comparable jobs. For a full-time job (40 hours) a monthly gross salary of €3250 is defined in the contract.

You know that twenty other recent graduates are also applying for the job. You know that there are few unfilled vacancies at other companies with comparable jobs. For a full-time job (40 hours) a monthly gross salary of €2400 is defined in the contract.

You know that three other recent graduates are also applying for the job. You know that there are many unfilled vacancies at other companies with comparable jobs. For a full-time job (40 hours) a monthly gross salary of €2400 is defined in the contract.

You know that twenty other recent graduates are also applying for the job. You know that there are few unfilled vacancies at other companies with comparable jobs. For a full-time job (40 hours) a monthly gross salary of €2000 is defined in the contract.

Table 1 - Defining the Different Parts of each Treatment in the Vignette

Note: This table shows the crucial part of the vignette in which each treatment differs. Only this part was altered, leaving the other parts of the vignette the same.

(17)

17

two different situations, hypotheses 2 could be tested:“People in a competitive labor market

are more easily satisfied with a low wage.” An important aspect of the structure in the

vignette study is discussed in the discussion section.

The aggregate survey consists of three parts. The first two parts consist of the vignette questionnaire and the last part consists of some general questions about the respondents’ age, gender, highest qualification and employment status, and some questions on opinions related to labour market conditions.

The vignette approach outlines a hypothetical work scenario, which has certain limitations. The most important limitation of a vignette approach is the loss of external validity. Therefore, Evans, Roberts, Keeley, Blossom, Amaro, Garcia and Reed (2015) emphasize that a “vignette study should produce results that generalize to real-world situations encountered by the participants and others like them”.

Vignettes are not proposed to recreate real-world situations; they are designed to measure key aspects of the decision-making process of individuals used in real-world situations. Much of the weight of a study’s validity is established in one element: the vignettes. This make a vignette unique (Evans et al., 2015).

Schoenberg and Ravdal (2000) mention another benefit of the vignette method. The vignette approach encourages the respondent to think outside their own situations. This allows discussing sensitive subjects. Further, because vignettes supply the relevant information to the respondent, it enables respondents with little knowledge about a subject to give reasoned answers.

3.2 Data of the Vignette Study

Data from 200 respondents was collected via the online survey tool Qualtrics. Questionnaires were distributed through various channels (WhatsApp, email, Facebook). Qualtrics provides a function to ensure that the situations (non-competitive and competitive) were shown randomly to the respondents. This was done to prevent the “order effects” which was discussed earlier in this section. A precise description of the data is given in the next section.

The data obtained from the vignette study consists of ordinal and nominal variables. This means that the data does not follow a normal distribution, which is one of the premises of a linear regression model. Therefore a linear regression model is not an appropriate model, and an ordered probit regression analysis was applied in this study, using the cumulative normal distribution. The model is estimated using maximum likelihood. The variable “attractiveness of the wage” was most often used as the dependent variable. This is a

(18)

18

categorical variable, i.e. it is classified in different (ordered) categories ((1) very unattractive, (2) unattractive, (3) neither unattractive nor attractive, (4) attractive, (5) very attractive). The linear regression assumes that the difference between “very unattractive” and “unattractive” is the same as that between “unattractive” and “neither unattractive nor attractive”. The category used in this study only reflects ordinality. Ordinality means the number only indicates a position in a category. There is no logical reason for expecting these differences between the number to be the same (Daykin & Moffatt, 2002). For probit regressions, the measure of fit for models with a binary dependent variable is the pseudo-R squared measures; this measures the fit of the model using the likelihood function (Stock & Watson, 2003).

The dataset consists of too few “unemployed” respondents and respondents with the working status “other”; hence, these respondents were deleted from the sample. The variable “student” is therefore a dummy (1-student, 0-employed). This also applies to “other” in the variable “highest level of qualification”. The “highest level of qualification” presented the following options: High school, MBO, HBO Bachelor degree, University Bachelor degree and Master degree. These variables are included in the model. The model will not be controlled for nationality, because 98,5% of the respondents were of Dutch nationality.

3.3 Data from the Research Centre for Education and the Labour Market (ROA) Every year, ROA carries out a survey among school graduates in transition from school to a connecting study or to the labor market in the Netherlands. The survey is done among school graduates approximately one and a half years after leaving school. For this study, a survey of graduates in transition from school to the labor market is interesting. The data includes unemployment, job satisfaction, gross hourly wages, gross monthly wages, duration of unemployment until the first job (in months), the demand of labor market, and connection level of the job. All these variables are defined separately for each study major. In this study, ROA data will help to investigate the correlation between unemployment, satisfaction and wage level. Two regression analyses were performed, the first to test if there was a correlation between unemployment, duration of employment until the first job and gross hourly wage, and the second to test if there was a correlation between gross hourly wage, unemployment, duration of employment until the first job and job satisfaction. Table 2 shows a summary of the descriptive statistics of the sample.

(19)

19

To examine the correlations, a linear regression model was used. Only the correlation can be tested because no control variables were available. The data consisted of 140 different HBO studies, in six different study sectors. HBO studies which did not contain all information were dropped from the sample. In Table 3, the number of respondents for each study sector is shown. Appendix A3 shows all HBO studies in the data which were used in this study.

Variable Mean Standard Deviation

Gross hourly income (in euros) 13,794 1,634

Unemployment (number of people) 208,090 188,072

Duration of unemployment until the first job (in months) 1,813 0,945

Job satisfaction (number of people) 1691,482 773,368

Number of Respondents 2013/2014

Agriculture 406

Education 1186

Technology 2044

Economics 4190

Health & Welfare 1915

Behaviour & Society 1816

Language & Culture 1063

Note: This table shows the number of respondents of the data of ROA in each study sector.

Table 2 - Descriptive Statistics of the Data of the Research Centre for Education and the Labour Market

Note: This table shows the four variables of ROA which are used to test the correlations in this study.

(20)

20

4. Results

In this part, the results are provided. This section begins with descriptive statistics, followed by the main results, which are discussed per hypothesis.

4.1 Descriptive Statistics

4.1.1 Sample

In total 314 participants started the vignette survey. Of those 314 participants, only 200 participants completed all questions. The remaining 114 participants who did not complete all questions in the survey were deleted from the sample. All participants were recruited online using the online survey tool Qualtrics. The sample consists of 118 female and 82 male participants. The average age was 27,6, and the majority of respondents (68%) were between 22 and 26 years of age, while 13,5% were older than 30 years. Fifty-four percent of the sample were students, 42% were employed, 2% were unemployed and 4% were “other” (retired, in an interim period). Regarding the highest qualification respondents had obtained, 30% had a master’s degree, 28% had a university bachelor’s degree, 23% had an HBO bachelor’s degree, 4% had completed the MBO, 14% had completed high school and 2% were “other”.

(21)

21

Variable Total sample First situation X First situation Y p-value

Gender (Male) 41% 46% 36% 0,285 Age 27,6 28,7 26,5 0,110 Occupation Studying 53,5% 50% 57% 0,024** Employed 41,5% 44% 39% Unemployed 1,5% 2% 1% Other 3,5% 4% 4% Highest qualification 0,930 High School 14% 17% 11% MBO 4% 3% 5%

HBO Bachelor’s Degree 22,5% 17% 28%

WO Bachelor’s Degree 28% 24% 32% Master’s Degree 29,5% 36% 23% Other 2% 3% 1% Nationality 1,00 Dutch 98,5% 98% 99% Belgian 0,5% 1% 0% Other 1% 1% 1% N 200 100 100

Table 4 shows a summary of the descriptive statistics of the sample. To test the differences between the two groups a chi-square test was used for the variables gender, occupation, highest qualification and nationality. A paired t-test was used for the variable age. In the fourth column of Table 4, the p-values are shown. These p-values show that the participants per group did not differ significantly for the variables gender, age, highest qualification and nationality. The p-value of the variable occupation shows a significant difference between the two groups. Because a within-subject design was used, this has no effect on the results. The internal validity does not depend on random assignment in this case.

More important for the results is the test of order effects. To control for order effects, the non-competitive and competitive situations were presented to the respondents randomly. To determine whether this was successful, the results from the vignette study should be Note: This table shows all descriptive statistics for the vignette sample as a whole and per start position in the vignette. The p-values in the last column indicate that the characteristics gender, age, education and nationality do not differ between the different start positions in the vignette. Only the variable “employment” differs between the two groups. Because of using a within-subject design, this would have no effect on the results. *,**,*** indicate p<0.01, <0.05, <0.01.

(22)

22

checked to see whether the results from “across respondents” are different from the results from “within respondents”.

“Across respondents” means to check whether answers for the same vignette differ between people who start with different vignettes. Thus the attractiveness of the proposed wage in the different situations can be compared. A Wilcoxon-Mann-Whitney test will show if the respondents who started with the non-competitive situation produced different results from the respondents who started with the competitive situation. Table 5 shows the results of the Wilcoxon-Mann-Whitney test in respect of the four proposed wages. The results show that the randomization between the two vignettes was successful, but it also shows a significant difference between the two groups for the valuation of the wage for the follow-up questions within the situations. An ordered probit regression analysis shows that respondents who started the vignette with the competitive situation value the wage of €2400 in the non-competitive situation lower than respondents who started the vignette with the non-competitive situation. This also holds for the wage of €2000 in the competitive situation (Apendix A4). This should be taken into consideration when analysing the results. This will be discussed more fully in Chapter 5.

Situation Proposed wage Situation first Mean z p-value

1. Non-competitive Wage of €3250 1. Non-competitive 4.29

1.054 0.2917 2.Competitive 4.39

Wage of €2400 1. Non-competitive 3.03

1.876 0.0607* 2.Competitive 3.2

2. Competitive Wage of €2400 1.Non-competitive 2.57

1.413 0.1576 2.Competitive 2.8

Wage of €2000 1.Non-competitive 1.92

2.120 0.0340** 2.Competitive 2.15

“Within respondents” means checking whether the same respondent’s answer differs between vignettes and wages. A fixed-effect panel data model is used to check this. In Table 6 the differences “within respondents” are checked for (1) a wage of €2400 in both the

non-competitive and non-competitive situations; (2) a wage of €3250 in the non-non-competitive situation and a wage of €2400 in the competitive situation; (3) the two wages in the non-competitive Note: This table shows the means separately for each group and the p-values from Mann-Whitney test to compare the means of the different proposed wages between the two groups. The means shown in this table are the averages of the answers to the questions related to the attractiveness of the wage for the respondent. The answer possibilities range from 1 (very unattractive) to 5 (very attractive).*,**,*** indicate p<0.01, <0.05, <0.01.

(23)

23

situation (€3250 and €2400); and (4) the two wages in the competitive situation (€2400 and €2000).

Table 6 shows that respondents rate all four wages differently. Column 1 shows that a wage of €2400 in the competitive situation is valued higher than a wage of €2400 in the non-competitive situation1. Column 2 shows that a wage of €2400 in the competitive situation is valued lower than a wage of €3250 in the non-competitive situation. Columns 3 and 4 show the valuation of wages by respondents when only the wage decreases and the labor market conditions remain the same. Respondents value wages significantly differently in different situations. In general, this result is logical because higher wages are often valued as more attractive.

The differences between columns 2 and 3 are interesting. In both situations the wage decreases from €3250 to €2400. The only difference here is the labor market situation. A decrease in wages between the non-competitive situation and the competitive situation is regarded as being less negative than a decrease in wages in the non-competitive situation. This can be due to the different labor market conditions2.

Wage of €2400 (1) Non-competitive: €3250 Competitive:€2400 (2) Non-competitive Situation (€3250-€2400) (3) Competitive Situation (€2400-€2000) (4) Competitive situation 0.439*** (0.053) Lower wage -1.674*** (0.061) -1.101*** (0.051) Competitive situation*lower wage -0.623*** (0.029) Constant 1.813*** (0.135) 4.976*** (0.066) 6.037*** (0.097) 6.433*** (0.181) Number of observations 374 374 374 374 R2 within 0.268 0.703 0.7998 0.714 F 0.000 0.000 0.000 0.000 1

Important note to this result: the proposed wage is not randomized between the respondents and is asked after the proposed wage of €3250 in the non-competitive situation. This could influence the results. Further explanation will be given in the discussion chapter.

2

This conclusion cannot be made with certainty, because the proposed wage in the non-competitive situation is not

randomized between the respondents and is asked after the proposed wage of €3250. This could influence the results. Further explanation will be given in the discussion chapter.

Table 6 – Within Respondents Analysis (PANEL: FIXED EFFECT)

Note: This table shows the “within respondents” analysis with a fixed-effect panel data model.

Dependent variable: attractiveness of the wage (Likert scale: 1-very unattractive;5-very attractive). Independent variable: Competitive situation (1-competitive situation, non-competitive situation) and Lower wage (1-lowest proposed wage, 0-highest proposed wage). Robust standard errors are reported in parenthesis. *,**,*** indicate p<0.01, <0.05, <0.01.

(24)

24 0% 10% 20% 30% 40% 50% 60% 70% 80% Yes No Male Female 4.2 Main results 4.2.1 Hypothesis 1

After the two hypothetical job scenarios were outlined, the respondents were asked a number of questions. One of these questions was: “Which factors determine your wage satisfaction?” Figure 1 shows the answers that were given. The background characteristics and

attractiveness of the job were mentioned most frequently. Comparative wages formed a part of fair wage determination, which is in line with the literature. A slight majority of 52% of respondents said that demand in the labour market determined their wage satisfaction. Other factors which were mentioned included career opportunities and secondary employment conditions. For this question, it was possible to give multiple answers.

Figure 2 shows the answers provided to the question: “Does the demand in the labor market play a role in determining your wage satisfaction?” The majority (75%) of respondents said that they took the labour market into account when determining what constitutes a fair wage. No significant gender differences for this question were found ([χ2 (1, N=200) =1,35

p=0.25]).

Figures 1 and 2 may lead to the conclusion that competition in the labor market is an important factor driving the perception of a fair wage. Figure 1 shows that demand in the labor market is not the most important factor driving the perception of wages. Background Note: This figure shows, by percentage, the selected

answers in the vignette study to the question: “Which factors determine your wage satisfaction?” Multiple answers were possible.

Note: This figure shows, by percentage, the selected answers in the vignette study to the question: “Does the demand in the labor market play a role in determining your wage satisfaction?” 83% 77% 67% 63% 52% 14%

Background (education, experiences) Attractivenss of the job

Comparative wages Current or previous wages Demand of labor market Other factors

Figure 1 –Selected answers by percentage: “Which factors determine your wage satisfaction?”

Figure 2 –Selected answers by percentage: “Do you take the demand of the labor market into account when determining a fair wage?”

(25)

25

characteristics, such as education and experience, and attractiveness of the job, are often more important determinants. Nevertheless, competition in the labor market is clearly one of the factors driving the perception of a fair wage. Hypothesis 1, “Competitiveness in the labor

market is one of the factors that drives the perception of a fair wage”, is supported.

Differences in answers are found between the two situations. Table 7 shows that the factor “competitiveness in the labor market” is more significant in driving perception of a fair wage in a competitive labor market than in a non-competitive market.

(1) (2) Competition 0.409*** (0.127) 0.411*** (0.127) Male -0.039 (0.135) Age -0.002 (0.009) Student 0.004 (0.155) Highest qualification 0.073 (0.051) Pseudo R2 0.019 0.023 Number of observations 400 374

Note: Dependent variable: number of times “competition in the labor market” is selected as a factor in determining what constitutes a fair wage (1-yes, 0-no). Independent variable: Competition (1-competitive situation, 0-non-competitive situation). Robust standard errors are reported in parentheses, *,**,*** indicates p<0.01, <0.05, <0.01.

Table 7 – Differences in answers between the competitive and non-competitive situation (ORDERED PROBIT)

(26)

26

Note: This table shows the answers of respondents as percentages of the whole sample in the different situations of the vignette study.

4.2.2 Hypothesis 2 4.2.2.1 Vignette

One of the main questions in the vignette study was: “Do you think the proposed wage is attractive?” Participants could rate the attractiveness of the salary using a five-point Likert scale, ranging from 1 (very unattractive) to 5 (very attractive). Table 8 shows the mean and standard deviations to this question.

Table 9 shows the answers of respondents as percentages of the total sample. Only the wage of €3250 in a non-competitive situation is viewed by almost all respondents as very

attractive/attractive. A remarkable result is the different views respondents had of a wage of €2400 under different labor market conditions. In a competitive labor market, a wage of €2400 was rated more often as very-attractive/attractive than in a non-competitive labor market. The mean values in Table 8 support this difference too.

As described in the section on methodology, in this case the only difference between the two situations was the number of other graduate students who applied for the job and the number of unfilled vacancies, in other words, the competition in the labor market. The difference in how attractive a wage of €2400 was regarded by men and women is confirmed in Table 10.

Situation Proposed wage Scale Mean Standard

Deviation

Non-competitive

€3250 1 (very unattractive)- 5 (very attractive) _4,340 _0,690

€2400 1 (very unattractive)- 5 (very attractive) _2,685 _0,812

Competitive €2400 1 (very unattractive)- 5 (very attractive) 3,115 0,846 €2000 1 (very unattractive)- 5 (very attractive) _2,035 _0,719

Situation Proposed wage Very attractive (5)/Attractive (4) Neither unattractive nor attractive (3) Unattractive (2)/very unattractive (1) Non-competitive €3250 94% 3,5% 2,5% €2400 15% 44% 41% Competitive €2400 35% 42,5% 22,5% €2000 2,5% 18,5% 79%

Table 8 –The mean and standard deviations to the key questions

Table 9 – Valuation of wages by the respondents in the vignette study

Note: The mean in this table shows the average of the answers to the questions related to the attractiveness of the wage for the participant. The answer possibilities range from 1 (very unattractive) to 5 (very attractive).

(27)

27

The independent variable competition is a dummy variable. The non-competitive situation is interpreted as low competition and the competitive situation is interpreted as high competition between potential employees.

Attractiveness of wage (1) Attractiveness of wage (2) Competition 0.554*** (0.110) 0.512*** (0.156) Male -0.334** (0.147) Male* Competition 0.137 (0.219) Age 0.005 (0.007) Student -0.219 (0.135) Highest qualification -0.021 (0.047) Pseudo R2 0.026 0.039 Number of observations 400 374

Figure 10 shows that there was a significant difference between the ways respondents viewed the same wages in the two situations. In a highly competitive labor market, respondents valued the wages as more attractive than in a labor market where there was less competition. This effect is significant at a 1% significance level. This supports hypothesis 2. Figure 10 also shows that the wage is regarded as less attractive by males than by females. No significant differences were found in the way males and females regarded the attractiveness of a wage of €2400 in a competitive labor market.

The following question supports hypothesis 2: “Are you more satisfied with a lower wage if you know it is hard to find a job?” A large majority of 78% of respondents answered: “Yes, it

Note: Dependent variable: Attractiveness of wage (Likert scale: 1-very unattractive; 5-very

attractive). Competition is a dummy variable (1-competitive situation, 0-non-competitive

situation). Robust standard errors are reported in parentheses, *,**,*** indicates p<0.01, <0.05, <0.01.

Table 10 – Differences in attractiveness of a wage of €2400 between the competitive and non-competitive situation (ORDERED PROBIT)

(28)

28

is more important to have a job”. Only 22% said that this would not influence their

satisfaction. There was a significant gender difference in the answers to this question ([χ2 (1, N=200) = 4,051 p=0.044]). Females were more satisfied than males with lower wages when they knew it was hard to find a job.

Another question of the vignette study also supports hypothesis 2: “When you know a lot of companies are looking for someone like you, what are your wage expectations?” A majority of respondents (57%) were more critical in this case and selected the option that the wage conditions should match their expectations. Thirty-eight percent said that this would not have an influence. They believed if the job and the company felt good, it did not necessarily guarantee a higher salary than other companies with comparable jobs. No significant gender difference was found ([χ2 (1, N=190) =0,512 p=0.472]). Five percent answers this question with “other”. The results of this question show a marginal support in favour of hypothesis 2.

4.2.2.2 The Research Centre for Education and the Labour Market (ROA) Data

ROA data was used to examine if there was a correlation between unemployment, the duration of unemployment until the first job and the gross hourly wage. The dataset is much larger than that of the vignette study and the respondents were asked about their own

experiences. Therefore, this data is a useful addition to this study. Firstly, looking at the effect of the rise in unemployment on the gross hourly wage, Table 11 shows that unemployment and the duration of unemployment before the first job had a significant negative effect on gross hourly wages. This means that when there is greater unemployment after the graduation of a specific HBO study, the gross hourly wage is lower.

Gross hourly wage (1)

Gross hourly wage (2) Unemployment -0.002*** (0.001) -0.002*** (0.001) Duration of unemployment

until first job

-0.435*** (0.133) Constant 13.913*** (0.206) 14.671*** (0.328) R2 0.083 0.184 Number of observations 75 75

Note: The linear regression model shows the correlation of the dependent variable gross hourly wage and the independent variable unemployment. Robust standard errors are reported in parentheses, *,**,*** indicates p<0.01, <0.05, <0.01.

(29)

29

Secondly, looking at how wages affect job satisfaction and whether unemployment moderates this effect, Table 12 shows a linear regression model with job satisfaction as a dependent variable and gross hourly wage as an independent variable. Column 2 controls for

unemployment to see whether that changes the coefficient on gross hourly wages.

Job Satisfaction (1)

Job Satisfaction (2)

Gross hourly wage -27.217

(36.999) 71.721 (69.763) Unemployment 2.932*** (0.693) Duration of unemployment

until first job

-182.005* (106.598) Constant 2066.918 *** (545.171) 555.349* (1100.782) R2 0.003 0.343 Number of observations 87 87

The linear regression in Table 12 shows that wage has no significant effect on job satisfaction. The coefficient on gross hourly wage clearly changes when controlled for unemployment. Without controlling for unemployment, the gross hourly wage has a negative effect and after controlling a positive effect. This means, without controlling for unemployment, that when the gross hourly wage increases, people are less satisfied with their job. With controlling for unemployment, when the gross hourly wage increases, people are more satisfied with their job. This demonstrates a significant effect of unemployment on job satisfaction. It means that when there is greater unemployment, people are more satisfied with their jobs. Because not all control variables are available, these results only show the correlation between the variables and no causal effects.

Note: The linear regression model shows the correlation between the dependent variable job satisfaction and the independent variable gross hourly wage. Robust standard errors are reported in parentheses, *,**,*** indicates p<0.01, <0.05, <0.01.

(30)

30

4.2.3 Hypothesis 3

First, the gender difference is analysed for the question discussed under hypotheses 2: “Do you think this salary is attractive?” Participants were asked this question four times, twice in each situation, in respect of the two different wages. A five-point Likert scale was used for this question, ranging from 1 (very unattractive) to 5 (very attractive). To determine if there was a significant difference in the perception of the fairness of the wages between the genders, an ordered probit regression analysis was performed.

First, the differences in the attractiveness rating of the proposed wage of €3250 in a non-competitive situation are confirmed in Table 13. In this situation, males rated the

attractiveness of the wage significantly lower than females. Column 2 shows the rating of the attractiveness of the proposed wage of €2400 for both situations. The dummy variable

“Competitive Situation” was added to the regression model, which represents the competitive situation. Column 2 also shows significant differences in the rating between males and females. However, significant differences were found between the competitive and non-competitive situation. As shown earlier, respondents rated the wage of €2400 as more attractive in a competitive situation than in a non-competitive situation. The interaction term showed no significant difference in the valuation of attractiveness for males in a competitive situation. The wage of €2000 showed no significant differences in rating between females and males. Even though not all differences are significant, the coefficient effects were negative, which means that males rated the wages less attractive than females did. These findings are in line with the theory of Clark (1997).

(31)

31 Attractiveness €3250 (1) Attractiveness €2400 (2) Attractiveness €2000 (3) Male -0.346 * (0.178) -0.333* (0.147) -0.310 (0. 173) Competitive Situation 0.512*** (0.156) Male*Competitive Situation -0.137 (0.219) Age 0.006 (.0112) 0.005 (0.007) 0.019 (0. 0115) Student -0.218 (0.212) -0.219 (0.135) 0.143 (0. 187) Highest qualification 0.170** (0. 066) -0.021 (0.047) -0 .099 (0.067) Pseudo R2 0.012 0.039 0.022 Number of observation 200 374 187

The supplementary question “Are you more satisfied with a lower wage if you know it is hard to find a job?” has already been discussed. The chi-squared test shows a significant difference ([χ2

(1, N=200) =4,051 p=0.044]). Females were more satisfied than males with lower wages when they knew it was hard to find a job. The coefficients in Table 13 support this, but the difference is not significant.

Significant gender differences were found in respect of negotiating wages. For the non-competitive situation, the chi-square test is χ2 (1, N=200) = 8,96, p=0.003. This means a significant difference in respect of considering renegotiating; males will negotiate more often than females. This also holds for the competitive situation, where the chi-square is χ2 (1, N=200) = 4,02, p=0.045). At 5%, it can be considered a significant difference in respect of negotiating between the genders. Even though there is no male or female evaluator in this scenario, males will initiate negotiations more often than females will.

Note: Dependent variable: Attractiveness of the wage (Likert scale: 1-very unattractive; 5-very attractive). Attractiveness of the wages of €2400 is tested in both situations (competitive and non-competitive), therefore the dummy variable “competitive situation” is created (1-competitive situation, 0-non-competitive situation). Robust standard errors are reported in parentheses, *,**,*** indicates p<0.01, <0.05, <0.01.

Table 13 – Differences in Gender in the Valuation of Attractiveness of Different Wages

(32)

32

Note: Dependent variable: number of respondents taking current and past achieved wages into account when determining what constituted a fair wage (1-yes,0-no). Robust standard errors are reported in parentheses, ***P<0.01 **P<0.05 *P<0.1

4.2.4 Hypothesis 4

Respondents were divided into two age groups: Young participants (18 to 30 years of age) and older participants (30+). Younger participants have fewer experiences with wages and are less likely to take achieved wages into account. Table 13 shows no significant differences between the two groups in terms of rating the attractiveness of the wages. This holds for all situations. This means that older respondents did not rate the attractiveness of the wages differently than younger respondents (18 to 30 years of age).

A probit regression analysis was used to determine whether older (30+) participants valued current and past achieved wages more often as an important factor in the determination of fair wages than younger participants. A majority of 66,5% of the younger respondents, compared to 40,1% of the older respondents, took past achieved wages into account when determining what constituted a fair wage. The probit regression model in Table 14 supports this. The variable “Old (30+)” shows a negative coefficient, which means that young respondents (18 to 30 years of age) take current and past achieved values into account more often than older respondents do. Contrary to expectations, hypothesis 4 is not supported by the data of the vignette study.

(1) (2) Old(30+) -0.6596** (0.263) -0.652** (0.329) Male -0.579 *** (0.199) Student -0.131 (0.221) Highest qualification 0.127 (0.075) Pseudo R2 0.003 0.008 Number of observations 200 187

Table 14 –Current and Past Achieved Wages as an Important Factor in Fair Wage Determination (PROBIT)

Fair wage determination from an employee perspective