Perceived income positions in the Netherlands

(1)

Perceived income positions in the

Netherlands

Master’s thesis

Koos Johan Beumer (1676881)

1st reader: Marike Knoef

2nd reader: Eduard Suari

January 2021

Abstract

Laypeople’s perceived income position is biased. This thesis aims to investigate that biased perception with data from Dutch inhabitants. In addition, I link this biased perception to the preferences for redistribution. This contributes to an ongoing debate in the literature by adding Dutch data. In addition, this thesis investigates the influence of the measurement methods on the outcome when measuring the perceived income position. The descriptive evidence suggests that the poorest 20% do overestimate their income position substantially by around 25 percentage points, while the richest 20% underestimate by nearly the same amount. On average, the Dutch underestimate their income position around 1 percentage point. Moreover, 37% of the respondents underestimated their income position by more than 10 percentage points, while 30% underestimated by the same amount. In addition, the respondents who were asked what percentage of the population has a lower disposable income, have on average a more negative biased perception of 9 percentage points than those who were asked about the percentage with a higher disposable income. Therefore, the measurement methods do matter when measuring the perceived income position. The outcomes do not prove that a perceived higher income is associated with less preferences for redistribution. However, people who belief that economic success is achieved in a fair manner, do have significant less preferences for redistribution. This thesis contributes to the ongoing debate concerning the perceived income position, the measurement methods and the effect on preferences for redistribution. With this, the foundation has been laid for further research on this topic with Dutch data.

(2)

1. Introduction

One of the most pressing debates in the 21st_{century is without a doubt the growing (income)}

inequality. It has increasingly become the subject of substantial public concerns (Roex et al., 2019; Seery, 2014; World Economic Forum, 2014; Piketty, 2015). Western countries suffer inequality both in wealth as in income. According to NRC Handelsblad (Dutch Newspaper), the Netherlands is no exception: the rich became increasingly richer over the past recent years (Kalse 2019). However, not every study agrees with the statement that inequality is growing in the Netherlands. Caminada et al. (2017) claim that the income inequality barely changed since the 1990’s. Indeed, Statistics Netherlands (CBS, Centraal Bureau voor de Statistiek) give a more or less stable Gini coefficient1_{for more than 20 years now. But the Gini coefficient is}

just an amount that tries to indicate the complex concept inequality. In the end, this does not say anything about the way laypeople experience (income) inequality, or how individuals perceive their own income position.

According to Hauser and Norton (2017), it is not only the actual level of inequality which is critical for shaping the debate concerning inequality. The perceived income position and the understanding of the levels of inequality is of crucial importance (p. 21). Governments attempt to shape distributional policies to flatten the inequality within their scope. ‘Who gets what’ is a frequently asked question in debates concerning economic distributions. Since the debate concerning inequality in democracies is (indirectly) shaped by the preferences of voters, it is important to understand how these preferences come about.

Cruces, Perez-Truglia and Tetaz (2013) stated the following: “If agents have biased perceptions of their own rank in the income distribution, their evaluations of how these costs and benefits will affect them are likely to be inaccurate” (p. 102). In the end, both perceived income position and perceived income inequality shape the debate, not the actual income inequality. Therefore, it is interesting and relevant to look at the perceived inequality by inhabitants of a country, and how this affects the preferences for redistribution of these inhabitants.

1.1 Scope

There exist two types of earlier research on this topic. At first, studies that investigate the inequality in a country, and what estimation people have on the size of this inequality. Hauser and Norton (2017) give a summary of the studies on this topic. For example, in the United

(5)

States and United Kingdom people generally underestimate the inequality within their country (p.1-3). In another study, Niehues (2014) used the Social Inequality module of the International Social Survey Program (ISSP) and found that similar levels of inequality may be perceived differently in Germany, France and Switzerland (p. 17). According to this research, redistributive preferences are influenced less by actual inequality than by perceived inequality. On the other hand, there is a tremendous amount of studies about the question how people evaluate their own income position compared to others in a country. Subsequently, this relative income position is linked to preferences for redistribution. According to Karadja, Mollerstrom and Seim (2017), a relatively richer person benefits less from redistribution, and therefore prefers less distribution. They invigorate this argument by conducting research with a Swedish sample. The first step in their research was investigating to what extent the average Swede over- or underestimates himself concerning income. They conclude that almost 70% of Swedish individuals believe that they are, relative to their nationals, poorer than they actually are. Thereafter, they conclude in theory that this could affect political outcomes concerning distributional measures. In addition to the Swedish sample of Karadja et al. (2017), Cruces, Perez-Truglia and Tetaz (2013) used an Argentinian dataset in their research. They suggest, based on their results: “those who had overestimated their relative position and thought that they were relatively richer than they were tend to demand higher levels of redistribution when informed of their true ranking.” This suggests that a change in perceived income position, rather than actual income position, affects the preferences for redistribution.

This research aims to contribute to the debate by adding Dutch data. Since the Netherlands has never been a case before on this topic, there is a gap in the literature. This study tries to fill that gap. With this, the literature becomes stronger on the topic, and it contributes to the ongoing debate. Furthermore, it can provide a fundament for further research with the Dutch dataset. With the knowledge whether Dutch people over- or underestimate themselves, further research can be done about the implications of this phenomenon.

In addition, this thesis contributes to the understanding of the influence of measurement methods when measuring the perceived income position. The influence whether a respondent is asked about the percentage of households that has a lower disposable income or the percentage of households that has a higher disposable income to measure the income position, is subject of this thesis. This is never done before, and therefore the major contribution of this thesis.

(6)

A sidenote must be made in front of this thesis. The data used is representative for Dutch inhabitants aged 50 or older. This is due to the fact that only respondents with the age of 50 or older filled in the survey used.

1.2 Research question

The first step in this research is to investigate whether the Dutch inhabitants over- or underestimate their own income position on average. In addition, I look at the differences in over- or underestimation between different measurement methods for measuring the perceived income position. The next step is to look at the biased perception for those on the bottom of the income distribution, and those in the top. This stands out in the literature (Engelhardt & Wagener, 2018; Cruces et al., 2013). The link between the perceived income position and the preferences for redistribution is the next step in the research. The literature tells us that there is a link between income and preferences for redistribution. This research investigates this link with Dutch data. To conclude, Ilook at the relation between the belief that economic success is achieved in a fair manner and the preferences for redistribution. The research question is as follows:

Do Dutch people have a biased perception of their income position, and to what extend does this affect their preferences for redistribution?

The main research question is built up from four sub questions. These four sub question both provide an answer to the main question, as well as structure the research.

1. Do the Dutch over- or underestimate their income position?

2. What is the influence of the formulation of a question when measuring the perceived income position?

3. To what extent is the actual income position a determinant for misperceptions of the own income position?

4. To what extent can preferences for redistribution be explained by an individual’s perceived income position and what extent can be explained by other variables?

1.3 Main findings

Based on the analyses in this thesis, we can conclude that Dutch individuals underestimate on average their income position. In addition, this thesis provides for evidence that the

(7)

measurement methods affect the outcome when measuring the perceived income position of an individual. As a result of this effect, the biased perception also differs. People who are asked about the percentage that has a lower disposable income are on average 9 percentage points more negative about their income position. Also, we see that rich people hold a large negative incorrect belief on their income position compared to their actual income position, while poor people overestimate about the same size. This thesis gives not enough evidence to conclude that richer individuals do have less preferences for redistribution, and vice versa. However, it did find that those who belief that economic success is achieved in a fair manner have less preferences for redistribution.

With these results, this thesis makes two major contributions to the literature. First, the debate is extended with Dutch data. But more importantly, there is evidence for the effects of measurement methods on the outcome when measuring the perceived income position.

1.4 Structure of the thesis

To start, chapter 2 provides for an overview of the relevant literature. This theoretical framework examines the main concepts underlying the research scope. Chapter 2 concludes by discussing the hypotheses (2.8).

Thereafter, chapter 3 justifies the research methods used in this study. Section 3.1 discusses the measurement methods, and section 3.2 the biased perception. Section 3.3 discusses the third sub question, followed by the link between the actual income position and biased perception (3.4). Section 3.5 discusses the achievement of economic success. Chapter 4 provides for the descriptive evidence, starting with the survey data (4.1) and the descriptive statistics of the samples used (4.2). Section 4.3 till 4.7 consist of the descriptive evidence for the variables of interest.

Chapter 5 provides for the outcomes as well as justifications of the statistical analyses done. The conclusion provides for a summing up of the main results and recommendations for further research (6).

(8)

2. Theoretical framework

This chapter elaborates on the existing literature of the relevant topics, starting with the definition of disposable income (2.1). Followed by the actual income inequality in the Netherlands. The following paragraph briefly summarizes the existing literature on the perceived income inequality (2.3) which describes to what extent individuals estimate the income inequality in a country. Then, 2.4 describes the perceived income position, which aims to describe a household’s income compared to others, and how this is measured (2.5). Followed by the factors that could cause these (mis)perceptions of income inequality (2.6) and the median voter model (2.7). The hypotheses based on the findings in the literature close this chapter (2.8)

2.1 Disposable income

There is a tremendous amount of definitions for the concept ‘income’. Disposable income (besteedbaar inkomen) is leading in this thesis. Fisher, Johnson and Smeeding (2013) give the following definition for disposable income: “Disposable income is money income from employment, investment, government transfers, and inter-household transfers of money, plus the value of food stamps and federal tax credits less the cost of federal and state income taxes, FICA taxes, and property taxes” (p.3).

Statistics Netherlands published an essay whit the Dutch definitions of income (Bos, Brakel en Otten, 2018). The definition used by Statistics Netherlands is the following: “the government redistributes a part of the primary income by levying contributions and taxes on the one hand and providing benefits and allowances on the other. Disposable income results from the process of acquisition and redistribution” (p. 8, translated from Dutch).

So disposable income is the income of a household after redistribution. This disposable income separates itself from net income as follows: net income does not account for possible bonusses, holiday pays, etcetera. Self-evident, the disposable income of a household contains the sum of multiple earners in that household. In this research, the definition of disposable income is important due to the use of this definition in the survey used. The respondents were asked about the yearly disposable income. The next part elaborates on the actual income inequality in the Netherlands.

(9)

2.2 Income inequality in the Netherlands

The disposable income of a household differ widely. One way to measure this inequality is the Gini-coefficient. Appendix A provides for a brief explanation how the Gini-coefficient is measured. Statistics Netherlands measures the Gini-coefficient every year. In a report launched in 2019, Statistics Netherlands documented several measurements concerning income (in)equality. According to van den Brakel and Pouwels-Urlings (2019, p. 6) the inequality in disposable income is stable in the Netherlands. From 2011 until 2017 the Gini-coefficient was 0,29. In fact, due to redistribution measures this coefficient has remained stable for over 20 years. This is contradicting with different warnings on growing inequality mentioned in the introduction. However, as mentioned before, there is disagreement on (the measurement of) inequality. Salverda (2014) claimed that the Gini-coefficient only looks at the median of the spectrum, and therefore is an incomplete indication (p. 40).

Also Thomas Piketty claimed the contrary of Statistics Netherlands; he said that inequality is increasing in most of the world. A shrinking group of people owns a growing concentration of capital in the world (Piketty, 2013). The urgency to reduce this inequality is clear when looking at the consequences. According to the OECD (2014) the growing inequality leads to a slower development of economic activities in western countries.

This is not the place to repeat this discussion, but be aware of the fact that inequality can be measured with several methods, and that these methods are determinative for the outcome. Now the context of income inequality is clear, it is time to elaborate on studies that aim to measure the perceived income inequality. The following paragraph provides for the distinction between perceived income inequality and perceived income position.

2.3 Perceived income inequality

Perceived income inequality and perceived income position are closely related concepts, yet different. The first concept, perceived income inequality, measures the question: ‘how does a person estimates the income inequality in a country/ in the world?’ This concept does not examine the own income position of the individual.

Perceived income position, on the other hand, asks the question: ‘how does an individual estimates his/her own income compared to others in a country/the world?’ This concept includes the own income of a household or individual, contrary to perceived income inequality. Perceived income position, also referred to as relative income position, is the concept used in this thesis.

(10)

However, perceived income inequality does give us some information about the appraisal of respondents. Since this is an estimation related to income as well, it is relevant information. Moreover, the perceived income inequality is often linked to preferences for redistribution. This is the same link as intended to make in this thesis, what makes it relevant. Therefore, studies concerning perceived income inequality are discussed below. Note that in earlier studies the names of the concepts (perceived income inequality and perceived income position) are often used interchangeably. Section 2.4 provides for an elaboration of the studies that measure perceived income position.

The first research discussed is executed by Hauser and Norton in 2017: they gave an overview of earlier research on perceived inequality. In a summary they stated the following: “Overall, the bulk of the current evidence suggests that people around the world hold incorrect perceptions of inequality in their country- but with variation. In the U.S. and United Kingdom, for example, underestimation of inequality is relatively common (Norton & Ariely 2011; Kiatpongsan & Norton 2014), while overestimation occurs in other countries, such as France and Germany (Niehues 2014). Moreover, there are a few exceptions of high accuracy: respondents in Norway, for instance, were relatively accurate in estimating their country’s income inequality (Niehues 2014). (Note that different methods can produce differences in estimated levels of inequality)” (p. 21-23). So, the question whether people over- or underestimate the inequality is asked in different countries.

Norton and Ariely (2011) proved that Americans are far too optimistic about inequality. In other words, there is more inequality than people perceive. The second step in this study was to ask respondents what their ideal amount of redistribution would be. They found that “respondents constructed ideal wealth distributions that were far more equitable than even their erroneously low estimates of the actual distribution” (Norton & Ariely 2011, p. 1).

This type of research is also conducted in Germany and France (Niehues 2014). In these countries, contrary to the United States, overestimation of inequality occurs. Niehues proves that “redistributive preferences are strongly correlated with the level of perceived inequality” (p. 17).

According to above mentioned studies, there is a correlation between perceived income inequality and preferences for redistribution. The next step is to look at studies that have aimed to link perceived income position to preferences for redistribution.

(11)

2.4 Perceived income position

This section provides for an elaboration on studies that investigated the perceived income position of individuals. The first study discussed was conducted with a Swedish sample by Karadja, Mollerstrom & Seim (2017). Their research investigated how people’s relative income position (or perceived income position) affects preferences for redistribution. In the first stage of their research, they sent 4,500 Swedish inhabitants the following question: “How many percent of the Swedish population (18 years or older) do you think have a total annual income which is lower than yours?” (p. 202-203) This question generates a scale, which, if people hold correct beliefs and the sample is representative, must be from 0 to 100. Furthermore, respondents were asked about their political party preferences and views on how distortive income taxes are.

When comparing these survey data to the actual income distribution in Sweden, they documented that nearly 70% of the respondents underestimates their relative income position by more than 10 percentiles. In other words, they believe they are poorer than they actually are. On the contrary, only 6% overestimate their relative position by 10 percentiles (the same amount).

Three months after this first survey, a randomly selected group received correct information about their true relative income position. They found that those who were informed that they had relatively a higher income than they thought, makes the demand for redistribution fall by 28%. In addition, Karadja et al. (2017) investigated whether the misperceptions differ between subgroups of their dataset. They found that people with above-average cognitive skills as well as younger respondents do have more precise beliefs (p. 206). They also looked at individuals who recently experienced shifts in social mobility and found a less negative bias than others. The study conducted by Karadja et al. corresponds in many ways with the methods used by Cruces, Perez-Truglia & Tetaz (2013). Cruces et al. used an Argentinian survey, including 1100 representative households. First, they found that systemic biases are present in perceptions of own income position: “a significant portion of poorer individuals place themselves in higher positions than they actually occupy, while a significant proportion of richer individuals underestimate their rank” (p. 101). Hereby, a distinction is made between people with higher and lower incomes, which was not done in the Swedish study. On page 104 they give some more detailed evidence. According to Cruces et al., 30% of the sample holds a positive biased perception, while 55% underestimates their income position. In addition, they conclude that the poorest 20% holds an incorrect belief of 2.98 deciles, which corresponds with

(12)

almost 30 percentage points overestimation. The richest group underestimates -2.88 deciles (p. 104).

Cruces et al. (2013) gave two rough explanations for the misperception of income position: limitations in information and bounded rationality (p. 102). They demonstrate this claim by illustrating the situation where a rich individual compares his or her income to a reference group. When everyone in this reference group is relatively rich, and he or she will take this as a representative group, an underestimation of their own income position is evident. This also holds for an individual with less income, who will overestimate evidently.

The second step in the study was to randomly assign a treatment group, who was provided with feedback and were actually confronted with accurate information. The conclusion is the same: correct information does have a significant effect on the preferences for redistribution: “those who overestimated their relative position and who were provided with accurate information demanded more redistribution than those in the control group” (p. 101). These findings are in line with the idea that a higher income leads to less demand for redistribution.

A third comparable study used a German dataset, consisting of 1,100 households. The study was conducted by Engelhardt & Wagener (2018). They found that knowledge about income inequality in Germany is little. Their first observation was the following: “Relatively poor respondents tend to overestimate their own rank while relatively rich respondents tend to underestimate their relative income” (p. 1). This is in line with the findings of Cruces et al. (2012) and suggests that the income distribution is far less equalized than perceived. Engelhardt & Wagener divided the sample in ten deciles. The poorest decile holds an average bias of 2.106, so an overestimation of 21 percentage points, while the richest decile holds a bias of -3.760. The second observation of this research was, according to the authors, quite unexpected: “Respondents across all income groups asked for more redistribution” (p. 1). This could be explained by the fact that the rich individuals (with income deciles between 8 and 10) located themselves far more down in the income distribution (between 5 and 6). Therefore, they might think they are profiting from the distribution, rather than net contributing.

Fernández-Albertos & Kuo (2018) conducted another research by asking whether individuals are accurately informed about their location in their national income distribution. They used a survey from a national representative population in Spain and asked people about their perceived placement in the income distribution. The next step in this research was to inform a randomly assigned group of their true position and assess the impact of such information on tax progressivity preferences.

(13)

Their findings are in line with earlier mentioned studies; “individuals are not necessarily well-informed about their placement in the income distribution” (p.2). This comes down to the same thing as Engelhardt & Wagner (2018) and Cruces, Perez-Truglia & Tetaz (2013) found: poorer people overestimate their position, while richer people underestimate their position.

These findings are certainly relevant for this thesis, since all these previous conducted studies did have more or less the same aim as this thesis. The next part provides for an elaboration on the measurement methods when measuring the perceived income position of respondents.

2.5 Measuring perceived income position

Since this thesis aims to investigate the influence of the measurement methods on the outcome of the perceived income position of a household, it is interesting to look at the measurement methods used in earlier research. Particularly, the formulation of the question when asking about the perceived income position of a household.

Karadja et al. (2017) asked in the survey they used about the annual income of a household. The question related is: “How many percent of the Swedish population (18 years or older) do you think have a total annual income which is lower than yours?” This question is comparable with the question asked in the survey Cruces et al. (2013) used: “There are 10 million households in Argentina. Of those 10 million, how many do you think have an income lower than yours?” Both ask about an amount or percentage that is lower than the respondent’s. Engelhardt & Wagner (2018) used the following question: “What do you think, how many households in Germany have an equal or lower standard of living than yours?” There are two major differences. First, Engelhardt & Wagner (2018) asked about the standard of living instead of an income (whether it is monthly or annual), and they ask about the same or lower, instead of lower.

The last measurement method discussed is those of Fernández-Albertos & Kuo (2018). They asked their respondents what percentage of households in Spain earns less than them, and what percentage earns more. Both answers had to sum to 100. This measurement method still asks the ‘lower’ question, however is supplemented with a ‘higher’ question.

We can conclude that previous studies measured the perceived income position by asking a ‘lower’ question, with minor differences. According to Cruces et al. (2013) this is the income-rank evaluation question (p. 103). The next part provides for an elaboration on the possible factors that shape misperceptions of income position.

(14)

2.6 Shaping perceptions

We see differences in the intensity of misperception in estimating the own income position. Therefore, it is interesting to look at factors that contribute to over- or underestimation of perceived income position.

Although the empirical evidence is thin, Hauser and Norton (2017)came up with three possible factors. The first factor that can be distinguished is the influence of the direct environment of individuals. There is strong evidence that people project local perceptions onto their estimates of social inequality. So one might ask himself; ‘What is it like in my environment?’ However, this environment is self-evident shaped by study-friends, neighbors or coworkers; people with more or less the same financial situation. Therefore, this creates a bubble wherein inequality is little.

Secondly, media coverage plays an important role in shaping the perceptions of inequality. According to Diermeier et al. (2017) the greater exposure to inequality-related stories, the greater concerns about income inequality and unfairness in society. Which means that a large media coverage of inequality-related stories leads to overestimation of the perceived income inequality.

Finally, the acceptance of hierarchy and beliefs in the role of personal choice on outcomes affects the perceptions of inequality. Hauser and Norton (2017) have put this third factor as follows: “Kteily et al. (2017) have shown that individuals who generally endorse hierarchies are also less likely to perceive inequality between groups.”

Moreover, Hauser and Norton (2017) stated that there are several experiments that proved the belief that choice and merit have considerable effects of approving inequality (p. 23). These aspects let us believe that when individuals believe that economic success is due to merit, the preferences for redistribution will decrease. More specifically, when people belief that hard work is the main determinant for economic success, the acceptance for inequality grows. Therefore, the demand for redistribution decreases. Note that these factors lead to over- or underestimation of inequality, and not of the own income position.

The last concept in this theoretical framework is the median voter model. This model forms the link between perceived income position and preferences for redistribution (sub question 3).

2.7 Median voter model

Different earlier mentioned studies linked perceived income position to preferences for redistribution (Karadja et al. 2017; Cruces et al. 2013; Engelhardt & Wagener 2018;

(15)

Fernández-Albertos & Kuo 2018). The assumption is that a relatively higher perceived income leads to less preferences for redistribution, and that a lower perceived income results in more demand for redistribution. This theorem stems from a principle introduced by Romer (1975) and Meltzer and Richard (1981). Both argued that richer people benefit less, pure economically, from redistribution. Therefore, it makes sense that a relatively higher income leads to less demand for redistribution.

This principle is based on the ‘Median voter model’. According to Congleton (2002), this is the most simple and straightforward explanation for political outcomes in a democracy (p. 1). The median voter model says in principle that voters can express their political preferences on a one-dimensional scale. In addition, the theorem says that a representative democracy exists, and every preference of voters is represented. This means that eventually, the median voter is most represented in politics. It is clear that this model is an oversimplification of reality, but it can help us to understand that same reality (Congleton 2002). However, we have to be careful with oversimplifying. According to Karadja et al. (2017) there are some underlying variables that cause a correlation between income and political preferences. In general, individuals with a high-IQ do have a higher salary. Mollerstrom and Seim (2014) have found that high-IQ individuals do have a preference for less distribution, so this can be a biased correlation. This study aims to investigate the biased perception on income position (difference between actual- and perceived income position) in the Netherlands. In addition, this thesis aims to link the perceived income position to preferences for redistribution. The next section provides for the hypotheses of this thesis.

2.8 Hypotheses

Following the concepts explained in the theoretical framework, this paragraph reveals the hypotheses underlying the data analytics in this study. The first hypothesis concerns the over- or underestimation of the perceived income position. The expectation is that, mainly based on Karadja et al. (2017) with Swedish data, the Dutch underestimate on average their perceived income position. This hypothesis is supported by Cruces et al. (2013):

Hypothesis 1: The Dutch underestimate on average their income position.

The second hypothesis concerns the formulation of the questioning in the survey used. When asking about the perceived income position of an individual, two different questions are used

(16)

(higher/lower). The theorem here is that this should not influence the outcome of the question, since the subject of the question stays exactly the same. Therefore, the second hypothesis:

Hypothesis 2: The measurement methods does not affect the outcome when measuring the perceived income position.

To proceed, the third hypothesis makes a distinction in the misperceptions of the relative income position of respondents. As stated, Engelhardt & Wagner (2018), Cruces, Perez-Truglia & Tetaz (2013) and Fernández-Albertos & Kuo (2018) all found that relatively richer people underestimate, and relatively poor people overestimate their income position. There is no reason to assume this is not the case in the Netherlands, hence the third hypothesis:

Hypothesis 3: Relatively poor people overestimate, and relatively rich people underestimate their income position.

The next hypothesis is based on the idea that the richer an individual is, the less he or she prefers redistribution (Alesina & Giuliano 2009, p.1). This, in combination with the median voter model, brings us to the fourth hypothesis:

Hypothesis 4: People who perceive themselves to be relatively rich, have less demand for redistribution.

A fifth hypothesis can be deduced. This hypothesis concerns another determinant for explaining the preferences for redistribution. Hauser and Norton (2017) summarized three possibilities for shaping the preferences for redistribution. The third determinant, beliefs on the achievement of economic success, is tested in this thesis. The hypothesis is based on the findings of Hauser and Norton (2017):

Hypothesis 5: Individuals who belief that economic success is due to hard work, have less preferences for redistribution.

The aim of this research is to test these hypotheses. The next chapter has the goal to explain the methods that lie underneath the analyses.

(17)

3. Research Methods

This chapter contains a description of the methods used in this thesis to answer the sub questions.

3.1 Measuring perceived income position

To start, this thesis investigates the perceived income position of Dutch households. This perceived income position is measured by asking respondents the question what percentage of the Dutch households has a lower disposable income, answered on a scale from 0 to 100. However, not every respondent gets the lower question: half of the respondents answered what percentage of the Dutch households has a higher income. This thesis aims to determine the effect of the way of asking about the perceived income position. To test this, I first look at the mean of both groups. Also, the distribution of both variables is given, as well as some graphical evidence.

3.2 Biased perception

The first sub question concerns the over- or underestimation of Dutch households regarding their income position. This sub question can be answered using two variables included in the dataset. The first variable measures the perceived income position of a household2_.

The second variable arises from the net income of the household filled in by respondents. Note that this is the estimation of the respondent, and that there is no evidence that this corresponds with the actual net income. Further research may check this with administrative data. I made a scale from 0 to 100, where 100 is the highest net income filled in, and 0 is the lowest. With this, both variables are on a scale from 0 till 100, which makes them comparable with each other3_.

The difference between the perceived- and actual income is the biased perception4_{. When the}

outcome of this biased perception is positive, the respondent estimates the own income position

2_{When a respondent is asked what percentage has a ‘higher’ disposable income, the outcome is adapted as}

follows: 100-perceived income when asked higher= perceived income when asked ‘lower’

3_{For example, when someone scores 80 on the scale of their actual income position and fills in 70 on the scale}

of the perceived income position, he or she estimates that 70% of the household has a lower disposable income, while in fact 80% has a lower income. This means that he or she underestimates in this case the own position by 10 percentage points.

(18)

higher than it actually is, which indicates overestimation. To answer the second sub question5_,

I computed both the biased perception of the whole sample6_{, as well as for the ‘higher’ and}

‘lower’ group.

3.3 Actual income position and biased perception

After determination of the biased perception, it is interesting to look at differences in the biased perception for different income groups. This is to answer the third sub question7_{. Both the}

whole dataset and the ‘lower’ and ‘higher’ group are examined. There are two phases to achieve a correct view.

The first phase is descriptive evidence. To start, I look at the distribution of the biased perception of the whole dataset, to continue with the biased perception per income group for both the ‘lower’ group and the ‘higher’ group.

I created three income groups: poor (poorest 20%), middle and rich (richest 20%). For each group (poor, middle, rich), I computed the mean. The descriptive evidence is supplemented with multiple regressions to answer the third sub question8_{, both for the whole sample as for}

the ‘lower’ and ‘higher’ group.

The outcome variables of these regressions are both the biased perception of the income position as well as the size of the biased perception. The size of the biased perception does not say anything about the ‘direction’ (positive/negative) of the bias, but measures whether rich- or poor individuals do have a larger or smaller biased perception in comparison to the reference group. The main explanatory variable is the actual income position of Dutch households. I control for the gender of an individual, as well as the age. Besides, the living environment is controlled for (urban or not), as well as the fact whether a respondent is high-educated or not. In addition, I use the dummy variable for the measurement methods as an explanatory variable. With this, I test for the difference in in outcome when asking about a lower or higher disposable income.

5_{What is the influence of the formulation of a question when measuring the perceived income position?} 6_{When a respondent is asked what percentage a higher disposable income has, the outcome is adapted as}

follows: 100-perceived income when asked higher = perceived income when asked lower.

7_{Sub question 3: To what extent is the actual income position a determinant for misperceptions of perceived} income position?

(19)

3.4 Perceived income position and preferences for redistribution

The fourth sub question of this thesis asks whether a perceived richer individual has less demand for redistribution. Therefore, the outcome variable are the preferences for redistribution, and the main explanatory variable is the perceived income position of an individual.

The outcome variable, preferences for redistribution, is measured in different variants. I focus on the role of government to reduce income inequality in the Netherlands. Therefore, I first determine whether, according to respondents, the income inequality in the Netherlands is too high. In addition, I use an outcome variable that measures whether respondents belief that it is the responsibility of the government to reduce the inequality.

The main explanatory variable is the perceived income position of respondents. Also, I control for the fact whether an individual overestimates the own income position (concluded when measuring sub question 1). To conclude, I control for gender, urbanity, high-educated and age.

3.5 Economic success

The fifth hypothesis claims that people who belief that economic success is achieved in a fair manner, do have less preferences for redistribution. To determine whether this hypothesis is correct, I use the same outcome variables as in 3.4 to answer sub question 4, namely whether the income inequality in the Netherlands is too high and whether it is the responsibility of the government to reduce the income inequality.

The explanatory variables for this regression are two dummies. The first dummy measures whether people belief that economic success is achieved in a fair manner. The second dummy variable measures whether the achievement of this economic success is due to hard work and merit, rather than luck. Each regression has both dummy variables as explanatory variables.

(20)

4. Data

This chapter contains the descriptive statistics (4.1), followed by the graphical evidence of the variables, beginning with the perceived income position (4.2), biased perception (4.3), the determinant income (4.4), preferences for redistribution (4.5) and economic success (4.6).

4.1 Survey data

The data underlying this research is the LISS (Longitudinal Internet studies for the Social Sciences) panel survey. The LISS panel consists of nearly 5,000 Dutch households. The participants of the LISS panel are selected by Statistics Netherlands and CentERdata, so self-selection is no concern. The survey is held for a tremendous number of topics. Monthly surveys provide for an enormous database.

The representativeness of the LISS panel is elaborately investigated by De Vos (2010). Despite the goal of the LISS panel to draw a representative sample of the Dutch population, there are some discrepancies. There is no examination of the representativeness for the year used in this thesis (2018). The most notable difference, according to De Vos (2010) is that the elderly are underrepresented, mainly in the age category 80+. In 2009, Knoef & de Vos conducted a research concerning the difference between filling in an online survey and a survey on a more traditional manner. The LISS panel took several measures to reduce the coverage of error in samples (for example providing a free PC for those who did not have access to the internet). Despite these measures, elderly are still underrepresented. Note that there are some discrepancies in generalizing the results of this paper to the whole population. However, this is not enough reason to abandon representativeness completely.

This thesis uses two specific datasets from the LISS panel. First, a survey held in June 2018, called ‘Pensioenambitie’, is used (dataset 1). This survey contains questions on the perceived income position of respondents. Moreover, the dataset provides for background variables, such as gender, age, income, etcetera. The variables under investigation for this thesis have only been filled in by respondents older than 49 years. Therefore, the focus in this thesis is on respondents 50 years or older. A total of 2,688 respondents filled in the relevant questions. The second dataset used is called ‘Perceptions of the determinants of economic success and demand for redistribution of income’. This dataset can be linked to dataset 19_{. In this survey,}

(21)

questions about the role of government in the redistribution process are posed (dataset 2). The survey stems from June 2013. There are 808 respondents who filled in all the relevant questions (so both from dataset 1 as dataset 2).

Combining two datasets involves both practical issues as well as some methodological concerns. The first dataset stems from June 2018, and the dataset for the questions concerning redistribution from June 2013. This 5-year gap results in a smaller number of overlapping households. However, more than 800 respondents answered all the relevant questions. This amount is still considered a sufficient sample.

Another concern could be the development of opinions concerning perceived income position and preferences for redistribution. However, the estimation is that this effect is negligible. Therefore, the combination of these two datasets is considered valid.

4.2 Descriptive statistics

This section provides for the descriptive statistics of the two datasets used10_{. I compare the}

datasets used with a sample of the population to ensure validity. For that, the original sample of the LISS panel dated June 2018 is used (the same month as the first dataset used in this thesis). Since the LISS panel is considered to be reasonably representative for the Dutch population, it is safe to draw conclusions if the datasets used are comparable with the original data. I filtered out the respondents with an age of 49 or lower, since I use only 50+ in this thesis. The table below summarizes the relevant data. Table 1 provides for the frequencies among different categories. Appendix B gives the mean and standard deviations of each dataset for gender, age, net income, education and urbanity.

Table 1: tabulated descriptives per category for the population, dataset 1 and dataset 2.

Population Dataset 1 Dataset2

N Percent N Percent N Percent

Gender 4,532 2,688 808

Men 2,249 49,62% 1,358 50.52% 421 52.10%

Woman 2,283 50,38% 1,330 49.48% 387 47.90%

Age 4,532 2,688 808

10_{Dataset 1: ‘Pensioenambitite’ (June 2018) and dataset 2: ‘Perceptions of the determinants of economic}

(22)

50-54year 736 16.24% 319 11.87% 86 10.64% 55-64year 1,592 35.13% 869 32.33% 236 29.21% 65+ year 2,204 48.63% 1,500 55.80% 486 60.15% Net income 4,513 2,688 808 No income 291 6.45% 145 5.42% 53 6.58% 500 or less 130 2.88% 77 2.88% 20 2.48% 501-1000 753 16.69% 438 16.36% 156 19.38% 1001-1500 840 18.61% 562 20.99% 178 22.11% 1501-2000 828 18.35% 548 20.47% 157 19.50% 2001-2500 653 14.47% 453 16.92% 127 15.78% 2501-3000 352 7.80% 234 8.74% 64 7.95% 3001-3500 171 3.79% 107 4.00% 23 2.86% 3501-4000 93 2.06% 55 2.05% 15 1.86% 4001-4500 38 0.84% 28 1.05% 6 0.75% 4501-5000 18 0.40% 10 0.37% 2 0.25% 5001-7500 27 0.60% 14 0.52% 3 0.37% 7500 or more 16 0.35% 6 0.22% 1 0.12% Don’t know 303 6.71% Education 4,521 2,685 807 Elementary 373 8.25% 219 8.16% 79 9.79% Vmbo 1,307 28.91% 750 27.93% 237 29.37% Havo/vwo 389 8.60% 245 9.12% 63 7.81% Mbo 958 21.19% 565 21.04% 179 22.18% Hbo 1,109 24.53% 675 25.14% 193 23.92% University 385 8.52% 231 8.60% 56 6.94% Urbanity 4,509 2,676 805 ++ 580 12.86% 359 13.42% 82 10.19% + 1,141 25.30% 665 24.85% 194 24.10% +- 1,045 23.18% 611 22.83% 194 24.10% - 1,033 22.91% 622 23.24% 200 24.84% -- 710 8.52% 419 15.66% 135 16.77%

(23)

It is interesting to look at the differences between the population and the datasets used (1 & 2). To do this, I conducted a Chi2 test for the three samples11_{. Table 2 presents the results of this}

Chi2 test.

Table 2: p-values for the Chi2-test comparing the datasets used (1 and 2) to the population. ***significant on a 0.01 level.

Chi2-test p-value dataset 1 p-value dataset 2

Gender 0,462 0,194

Age 0,000*** 0,000***

Net income 0,000*** 0,000***

Education 0,926 0,422

Urbanity 0,951 0,190

Note that there exist significant differences in age and net income between the datasets. Gender, education and urbanity do not differ significantly. However, the fact that there are significant differences between the datasets does not mean that the datasets cannot be used. When looking at the descriptive statistics in table 1, there is no substantial difference between tabulated statistics. This indicates that the small differences are significant, but not substantial. Therefore, the datasets are considered fairly representative to the Dutch population.

4.3 Perceived income position

The following paragraphs provide for an elaboration on the variables of interest in further detail. The first variable discussed is the ‘perceived income position’. I took the information of a question asked in dataset 1. The relevant question is responded by 2,688 individuals. It is made clear to the respondents that the question alludes to yearly disposable income. Individuals can answer on a scale between 0 and 100 to the following question: “What percentage of Dutch households (excluding students) has a yearly disposable income that is lower/higher than yours?”

From now on, the interpretation of the value is as follows: when an individual answer for example ‘40’, it means he or she estimates that 40% of the Dutch households has a lower

11_{When the p-value of the Chi2 is significant, it means that there is a significant difference between the}

(24)

disposable income, so 60% must have a higher disposable income. This holds for the whole sample12_.

Whether a respondents got the lower or higher option in the question is randomized. According to the second hypothesis, this should not affect the outcome of the variable13_{. Figure 1 provides}

for the distribution of both variables after making the interpretation comparable14_.

Figure 1: This figure shows the histograms of the question what percentage of the Dutch households a Lower/Higher disposable income has than the respondent. The orange histogram represents lower, the blue represents higher. For this graph, the outcomes of the ‘higher’ question are reversed. So, these numbers also represents the percentage of Dutch household with a lower disposable income.

This graph tells us that there are differences between the measurement methods when measuring the perceived income position. The comparison of the means confirms this. When the question is formulated with ‘lower’, the mean of the sample is 44.181 (.573)15_{. While}

formulated with ‘higher’ the mean is 53.161 (.561). The mean of the whole sample is 48.781 (.404). Figure 2 presents the outcome of the question after merging both question options16_{. In}

12_{Lower=100-higher. Therefore, the variable represents the percentage of households that has a lower}

disposable income

13_{Hypothesis 2: The measurement methods does not affect the outcome when measuring the perceived income} position.

14_{The percentage for ‘lower’} 15_{Standard error in parentheses.}

16_{Lower=100-higher. Therefore, the variable represents the percentage of households that has a lower}

(25)

the middle area, there is a larger concentration than expected17_{. On both sides, however, the}

concentration is remarkably low.

For a more detailed view on the distribution of the perceived income position of individuals, I divided the outcome into three categories: high-, middle- and low perceived income position. The middle group reaches from the outcomes 21 till 79, high and low include the extremes. Table 3 includes the tabulation. Remarkable is the large concentration in the middle group. This contains 76.79% of the respondents, while in comparison with the actual income position this must be 50%.

Table 3: the tabulation of the variable perceived income position

Frequency Percentage

Low perceived (<21) 362 13.47%

Middle perceived 2,064 76.79%

High perceived (>79) 262 9.75%

Total 2,688 100%

17_{When there is absolutely no bias, the distribution would be a flat line at the 10% (y-axis) mark.}

Figure 2: This figure shows the histogram of the question what percentage of the Dutch households has a lower yearly disposable income than the respondent for the whole sample. No bias would mean that every group has 10% (the poorest 10% estimates that 10% has a lower disposable income etc.).

(26)

4.4 Biased perception

The perceived income position must be compared to the actual income position to detect whether individuals over- or underestimate their income position18_{. Note that the actual income}

variable measures the net income of the household. The first hypothesis of this thesis claims that the Dutch underestimate on average their income position, so a negative biased perception.19_{Section 4.4.1 provides for an elaboration on the biased perception of the whole}

sample (both the lower and higher question are merged). Section 4.4.2 elaborates on the influence of the different measurement methods on the biased perception.

4.4.1 Whole sample

The mean of the biased perception is -1.332 for the whole sample. The Standard Error (SE) is .522, and therefore the t-value of this variable is -2.552. This tells us that the mean of this variable is significant different from zero. Also, the mean is negative, what indicates that the average individual underestimates their income position.

In other words, Dutch individuals think on average that they earn less than they actually do. This confirms the first hypothesis20_{. However, the underestimation is little, since an outcome}

of zero would mean that there is on average no under- or overestimation (and on the scale of the population) no biased perception.

In addition, the data tells that 37% of the Dutch underestimates the own income position by more than ten percentage points, contrary to 33% who holds a positive bias with that same amount (overestimation). In comparison with Karadja et al. (2017), less people underestimate their income position in the Netherlands: in Sweden, 70% of the population underestimates the own income position. However, the Dutch overestimate far more often their income position in comparison with the Swedish (in Sweden only 6%). In Argentina, 30% holds a negative bias (underestimation) and 55% overestimates (Cruces et al., 2013).

4.4.2 Influence measurement methods

This paragraph provides for an elaboration on the differences in biased perception for the higher- and lower questions. Therefore, this reveals the influence of the measurement methods when measuring the perceived income position on the biased perception.

18_{The actual income is also a scale from 0 till 100, whereby 0 is the lowest income and 100 the highest. The}

distribution of this variable is therefore a flat line

19_{Biased perception = Perceived income position – actual income position} 20_{Hypothesis 1: The Dutch underestimate on average their income position.}

(27)

The mean of the biased perception for the group who got the ‘lower’ question is -5.894 with a standard error of 0.73521_{. The group with the ‘higher’ question has an average biased}

perception of 3.124 (0.722)22_{. For the ‘higher’ group, the answers are reversed}23_{. Contrary to}

the mean of the biased perception for the ‘lower’ group, this mean is a positive number, and therefore indicates overestimation24_.

So when respondents are asked what percentage of the Dutch households has a ‘lower’ disposable income, there occurs underestimation. However, when asked what percentage has a ‘higher’ disposable income, there occurs overestimation. The difference in the mean is around 9 percentage points.

4.5 Actual income position and biased perception

The distribution of the the perceived income position along the actual income scale gives a more detailed view. The assumption here is that relatively poorer people overestimate-, and relatively richer people underestimate their income position. This corresponds with the third hypothesis. For testing this hypothesis, I dive deeper into the data. First, 4.5.1 provides for descriptive evidence for the whole sample. Then, 4.5.2 makes a sidestep to the differences between the ‘lower’ and ‘higher’ questions and reveals the influence of the measurement methods.

4.5.1 Whole sample

Figure 3 shows descriptive evidence for the third hypothesis25_{. This figure compares the actual}

income position (x-axis) with the perceived income position (y-axis). The green reference line indicates no bias. In the lower part of the income distribution, the relation between the actual income position and the perceived income position (blue line), is above the green reference line. This means that there is overestimation. In the higher part of the income distribution, there is underestimation. Looking at figure 3 tells us that individuals that are considered the poorest, do overestimate on average the most, as this overestimation decreases towards the middle of

21_{T-value: -8.019} 22_{T-value: 4.326} 23_{Lower=100-higher.}

24_{Note that for the calculation of the biased perception per group every individual received a new number for the}

actual income position. Still, the lowest income of the sample has 0 and the highest 100. However, the sample is twice as small since half of the individuals got another question. Therefore, there may be slightly differences in the computation of the biased perception for each individual. This is, evidently, of influence on the mean.

25_{Hypothesis 3: Relatively poor people overestimate, and relatively rich people underestimate their income} position.

(28)

the income positions. Appendix C includes two comparable figures for the ‘lower’ and ‘higher’ groups.

The next step is to divide the sample into three groups: poor, middle and rich. The poor group consists of the poorest 20% of the sample, and the rich group of the richest 20%. The means of the biased perceptions in these three groups demonstrate large differences between the groups. According to these results does the poorest 20% of the sample overestimate their income position (24.008). This means that the average individual in the poor group overestimates the own income position with 24 percentage points. On the other hand, the richest 20% underestimates: -26.823. The middle group holds a small negative biased perception: -1.262. In line with Karadja et al. (2017) and Cruces et al. (2013), table 4 shows information on the differences between ten income groups for the whole population. Appendix D provides for two comparable tables for the ‘lower’ and ‘higher’ groups. Remarkable in column 1 is the fact that the average perceived income position covers the area between 32 and 65. So, the average

Figure 3: the y-axis shows the outcome of the question what percentage of the Dutch households has a yearly disposable income that is lower than the disposable income of the respondent. The x-axis represents the income position of that same respondent. The green line refers to the situation with no bias. The mean of the biased perception is -1.332.

(29)

perceived income position covers an area around 30 percentage points, which is remarkably small. Column 2 shows the mean of the bias per income group. The poorest half overestimates, but towards the middle this overestimation decreases. From income group 60 and higher, there occurs underestimation which increases towards the richest individuals. Columns 3 and 5 show the proportion of a positive/negative bias bigger than 10 percentage points/ smaller than -10 percentage points. The large amount of overestimation in actual income group 10 (78%) and underestimation in income group 100 (84%) are remarkably high, but explainable26_.

Comparing this table to Engelhardt & Wagener (2016), there are only minor differences. The tendence is the same: extreme over-/underestimation in the poor/rich groups, and less biased perception in the middle groups.

On the basis of these data, we can conclude that the actual income position is a clear determinant for misperceiving the income position. The richer, the more underestimation, the poorer, the more overestimation. The more towards the middle of the actual income distribution, the less biased perceptions. The results chapter provides statistical evidence for this assumption.

Table 4: the summarizing statistics per actual income group.

(1) (2) (3) (4) (5) (6) Actual income group Average perceived income position

Mean bias Proportion with positive bias27 Average positive bias Proportion with negative bias28 Average negative bias 10 32.914 27.910 0.784 34.725 0.000 0.000 20 35.992 21.000 0.695 29.699 0.029 -13.950 30 40.758 15.758 0.579 30.071 0.100 -16.876 40 46.765 11.758 0.579 23.879 0.115 -17.919 50 46.701 1.705 0.312 22.437 0.256 -21.854 60 50.011 -4.973 0.189 21.711 0.379 -23.859 70 51.453 -13.53 0.081 18.763 0.546 -26.922 80 56.044 -18.95 0.059 13.833 0.669 -28.914 90 61.052 -23.95 0.014 16.332 0.776 -30.419 100 65.302 -29.69 0.000 0.000 0.847 -34.009

26_{We take actual income group 10 as an example. When an individual finds himself in the poorest 10% of the}

income distribution, underestimation is not possible since one cannot estimate lower than their actual income position, which is between 0 and 10. Therefore, no one holds a negative biased perception in the lowest income group, and no one holds a positive biased perception in the highest income group for the same reason.

27_{more than 10 percentage points overestimation} 28_{more than 10 percentage points underestimation}

(30)

4.5.2 Influence measurement methods

This section describes the differences between the higher- and lower question on the distribution of the biased perception per income group. In line with 4.5.1, the sample is divided in three groups: poor, middle and rich. Those who got the ‘lower’ question and are located in the poorest 20% of that sample, hold an incorrect belief of 19.846. This means an overestimation of about 20 percentage points. The richest 20% of the group with the ‘lower’ question has an average biased perception of -30.992; underestimation.

For the group who got the ‘higher’ question, the poorest 20% overestimates with almost 28 percentage points, while the richest 20% underestimates almost 23 percentage points. Table 5 summarizes these values, in comparison with the whole dataset values. Appendix C provides for two graphs displaying the relationship between the actual income position and the perceived income position.

Table 5: over- and underestimation for the richest and poorest 20%, divided by the ‘lower’- and ‘higher’ question and the whole dataset.

(1) (2) (3)

Biased perception ‘Lower’ question ‘Higher’ question Whole dataset

Poorest 20% 19.846 27.915 24.008

Richest 20% -30.992 -22.880 -26.823

Table 5 tells us that those who are asked the ‘higher’ question, hold on average a more positive biased perception in comparison with those who are asked the ‘lower’ question. The poorest 20% overestimates in all three columns, but the largest overestimation is in column 2 (those with the ‘higher’ question). On the other hand, the largest underestimation is in column 1 for the richest 20% (column 1 belongs to those who were asked the ‘lower’ question). The difference in overestimation for the poorest 20% between the ‘lower’ and the ‘higher’ group is around 8 percentage points. The same difference applies for the richest 20%. This is a remarkable large difference.

This table tells us that, in the rich- and poor group, those who are asked the ‘lower’ question, are on average more negative. Therefore, they hold on average a more negative biased perception.

Therefore, we can conclude that the measurement methods do have an influence on the outcome when measuring the biased perception of the own income position.

(31)

Figure 4 displays the slope of all three samples (‘Lower’- and ‘higher’ question and the whole sample). This figure demonstrates that those with the ‘higher’ question holds the most positive biased perception, and those with the ‘lower’ question the most negative. The whole sample is located, self-evidently, in the middle. In the actual income groups around 20 till 40, the largest difference between the ‘lower’ and ‘higher’ group is observable. From the actual income group 10 till 100, the difference in average biased perception remains more or less the same.

All these results taken together point in the direction that the measurement methods do matter (contrary to hypothesis 2). The results chapter provides for some statistical evidence of this presumption.

Figure 4: This figure shows the biased perception per actual income group for: those who were asked what percentage of the Dutch households has a lower disposable income, those who were asked what percentage of the Dutch households has a higher disposable income and the whole dataset, which combines both questions.

(32)

4.6 Preferences for redistribution

Preferences for redistribution is the dependent variable for the fourth and fifth hypothesis in this thesis. The data underlying this variable are derived from dataset 2. The number of respondents that filled in both these questions as well as the questions concerning the perceived income position is 808. Three questions built up the preferences for redistribution, namely: “Income differences in the Netherlands are too large”, “It is the responsibility of the government to reduce the differences in income between low incomes and high income” and “It is the responsibility of the government to reduce the income inequality, even if this means that the purchasing power of rich people decreases.” The last two questions are combined in a dummy variable ‘government’s responsibility’. Figure 5 display the frequency of the answers to these questions. The dummy variables for these questions are divided by those who have more preferences for redistribution (options 4 and 5) and those not.

Figure 5: This figure shows the frequency of the answers of the following questions: 1. “income differences in the Netherlands are too large.” 2. “It is the responsibility of the government to reduce the differences in income between low incomes and high incomes.” 3. “It is the responsibility of the government to reduce the income inequality, even if this means that the purchasing power of rich people decreases.”

Perceived income positions in the Netherlands