Perceptions Towards Immigrants: Does the Current
Level of Household Income Matter?
Victor Koster – s2371073
University of Groningen
Date: 11-01-19
Contact: v.r.koster@student.rug.nl
Supervisor: Dr. R.M. Jong-A-Pin
Abstract
1
1. Introduction
Despite the upward business cycle the economy is in today, populist parties are gaining support throughout Europe and the US. In many European countries, populism gathers around the pillar of the right wing. Rodrik (2018) defines right wing populism as follows: “right wing populism mobilizes along ethno-national and cultural cleavages, as immigration and refugees are held responsible for economic downturn”. The literature explains this mobilization of right wing populist politicians and their growing support based on, amongst other socio-political reasons (population size preferences, safety issues, integration/cultural issues), two economic arguments. The first explanation comes from the self-interest theory, which focuses on “competition over limited resources, such as jobs, social welfare benefits, and education, as the source of opposition to immigration” (Citrin, Green, Muste, and Wong (1997)). A second explanation for attitudes towards immigration focuses on the country’s financial success. If the economic situation of a country deteriorates, its inhabitants find themselves increasingly opposed to immigrants, as they perceive immigrants as a threat to their financial wellbeing. On the contrary, inhabitants of a country in a stable financial situation find themselves less opposed towards immigrants (Miller (2012), Coenders and Scheepers (1998), Quillian (1995)).
In this thesis, I focus on attitudes towards immigrants in the Netherlands. As the economy in the Netherlands (and in other Western European countries) is improving, I argue that the main economic argument people hold against immigrants is based on the self-interest theory.
Scholars have looked into the self-interest theory extensively in order to explain attitudes toward immigrants. Empirical studies have examined the impact of immigration on competition over resources by natives and immigrants. Most of them focus on the effect of immigration on the labor market and consequently on wages. The majority of studies performed in this field conclude that there is little impact of immigration on the labor market in terms of employment, unemployment and wages (Friedberg and Hunt (1995), Dustmann, Fabbri, Preston and Wadsworth (2003), Card (2005)), while others do find a significant impact on labor market competition and wages (Borjas (2003))1.
Another line of study did not focus on the direct effect of immigration on the local economy, but rather the economic fears imposed and enhanced by the inflow of immigrants. Scheve and Slaughter (2001) find that low-skilled native workers will lose when forced to compete for jobs with low-skilled immigrants, as immigrant workers price themselves in the market at lower wages. They conclude that this perceived threat lies at the heart of anti-immigration feelings. A majority of scholars, however, conclude that the most important driver for this perceived economic threat is the educational attainment of individuals, where anti-immigrant sentiments decline with the level and years of education
1 Borjas (2003) proves that immigration lowers wages of competing workers, where a 10% increase in supply
2 (Espenshade and Calhoun (1993), Hainmueller and Hiscox (2007) (2010), Hainmueller, Hiscox and Margalit (2015), Citrin et al. (1997), Card, Dustmann, and Preston (2012)).
However, as the Netherlands has a highly educated population, still anti-immigrant statements are popular tools for politicians to gain support. This implies that other factors than just education are important in explaining immigration attitudes. Some scholars have proven that this is true. Amongst them Gang, Rivera-Batiz, and Yun (2002) who find that the strength of educational attainment declines in explaining negative attitudes towards foreigners during the period between 1988 and 1997 in the US. Burns and Gimpel (2000) state that prejudices regarding immigrants and thus negative attitudes towards them are, at least partly, caused by economic insecurity (rather than education only).
In line with Burns et al. (2000) I will draw attention away from the role of educational attainment and examine which other economic variables help explain the formation of anti-immigration sentiments in society. I will focus on the role of individual economic factors, and in particular on the role of the current level of household income in explaining immigration attitudes.
An important extension to this thesis, one that makes this article different from the other articles explaining immigration sentiments, is the interaction of the current level of household income with proximity to immigrants. As Burns et al. (2000), and Gang et al. (2002) point out, an important role is played by the concentration of immigrants in one’s local neighborhood in the formation of attitudes. Markaki and Longhi (2013) show that discriminating between regions is important in studies of perceptions. Perceptions are influenced by one’s local environment, and hence different factors might be relevant for people who live in close proximity to foreigners compared to people who live in far proximity to foreigners. The question that I will focus on is whether proximity influences the strength of the current level of household income, in explaining immigration attitudes.
The dataset I use, is from a survey named ‘Political Ideology’, performed in March 2016, and made available by the Rijksuniversiteit Groningen. It constitutes cross-sectional data on 2.460 households in the Netherlands. Amongst them, there are households of every social class and income group. Questions in the survey regard, amongst other things, the impact of immigrants on perceived labor market competition, and perceived financial threat due to immigrant inflows. I combine this with CBS data on the ratio of natives to immigrants in different regions in the Netherlands.
3
2. Theoretical framework
There are two extremes in the spectrum of perceptions towards immigrants (Hainmueller et al. (2007)). On the one hand, immigration can be seen by natives as a potent source of labor to supply the key industries within a country (Freeman (2006)). Immigration can alleviate the threat of an ageing population on the countries’ pension system, by injecting the local population with young immigrant workers (Freeman (2006)). From a cultural perspective, immigrants can contribute to society by enriching the world of arts, philosophy, cuisine and other intellectual property within a country (O’Rourke and Sinnot (2006)). The other side of the spectrum, however, might perceive immigrants as ‘job stealers’ at the expense of natives (Chomsky (2018)). Immigrants might be associated with fortune seekers who subtract from government budgets by taking advantage of a countries’ generous social welfare policies, and hence are costly for taxpayers (Chomsky (2018)). On top of that, they are a source of polarization within society, as they are perceived to undermine traditional cultures (Miller (2008)) , cluster within their own groups (Dustmann and Preston (2001)) and increase the number of crimes committed (Merton (1938), Peterson and Krivo (2005)). But who are the people at both ends of the spectrum? Who are the people that feel economically most threatened by immigrant inflows, and who are the people actually in favor of liberal immigration policies? In order to answer that question, we first look at sophisticated economic theories developed on this matter.
Economic theories explaining the impact of immigration
The economic models considered in this section all discuss the income effects of immigration. The effect of immigration on peoples’ income depends on the productive factors people own (Borjas (1997), (1999)). If immigrants compete for the same productive factors, they impose a threat to natives’ financial wealth. Most studies performed in this field have assumed that the inflow of immigrant workers is an inflow of low-skilled labor into the local economy, meaning that natives’ who own this factor of production are the ones competing with the immigrants. Consequently, economic thinking teaches us that it is plausible that supporters of immigration are amongst those who expect to gain, and opponents amongst those who expect to lose.
4 low-skilled workers, since the increase in low-skilled labor enhances the productivity of high-skilled workers (due to the complementarity between factors of production), which might increase their wages (Chiswick, 2005). Scheve et al. (2001) build on the Factor Proportions Analysis and conclude that there is a link between attitudes to immigration and the distributional consequences of immigration. Low-skilled workers expect to lose from inflows of low-Low-skilled immigrants, and hence favor restrictive policies. Also Mayda (2006) finds that in countries where most people are high-skilled compared to immigrants, support is quite prominent as they expect to have higher wages due to (low-skilled) immigrant inflows (O’Rourke et al. (2006) find similar results). Malhora, Margalit, and Mo (2013) weaken the results of Scheve et al. (2001) and Mayda (2006) by concluding that the threat for native workers is only pronounced for those who actually compete with immigrants, and not low-skilled or high-skilled workers in general.
The second economic model that can be used for explaining the impact of immigrants on the national economy is the Heckscher-Ohlin model. It assumes an open economy and comparative advantage between factors of production (Williamson (2002)). According to this theory, the impact of immigration is offset by world trade. An increase in skilled labor simply means that the factors produced by low-skilled labor are imported less from other countries, as it can be produced locally at lower cost. Hence, countries change their output mix. In a small open economy, this change will have no impact on world prices, and therefore wages will not change at all if the local economy is small enough (this is called factor price insensitivity, as discussed by Learner and Levinsohn (1995)).
If we allow for different industry specific skill levels of immigrants in an open economy model, as described in the article of Scheve et al. (2001), the picture again becomes quite like the Factor Proportions Analysis. When we assume that all goods are traded so that prices are fixed in the world market, the inflow of low-skilled workers indeed puts pressure on the wages of native low-skilled workers (and the same holds for inflows of skilled workers and the effect on the wages of high-skilled natives, ceteris paribus). However, if we assume non-traded goods are included in the model, and the inflow of immigrants is such that they have skill levels specific to those non-traded goods, it can be the case that those non-traded goods become either cheaper or more expensive. This is dependent on the consumption preferences within the country after immigration. Hence, the effect on wages is ambiguous.
Empirical evidence for the effects of immigration on labor market outcomes
5 which was used to compute local labor marker outcomes, and data of the Current Population Survey, which contains data on wages of individual households. The most important finding of Card (2005) is that immigration does not reduce the labor market opportunities of the less-skilled natives. “Recent evidence on the response of the local industry structure to immigration induced supply shocks shows that the absorption of unskilled immigrants takes place within industries in high-immigrant cities, rather than between industries, as implied by simple trade models” (Card (2005)). So, the findings of Card (2005) suggest that an inflow of skilled immigrants does not deteriorate the opportunities for low-skilled natives in a country. The labor market changes only within certain geographical regions where immigrants tend to cluster.2
Dustmann et al. (2003) researched the labor market in the UK based on data from the Labor Force Survey and the 1971, 1981 and 1991 UK Census. They conclude, like Card (2005) did for the US case, that there is no discernable effect of immigration on the employment or wages of existing workers, and hence that labor market fears based on immigration are not justified.
Borjas (2003) on the other hand does find empirical evidence for the labor market impact of immigration, based on the 1960-1990 US Decennial Census and the 1998-2001 Current Population Surveys. Assuming a labor market where people with the same educational level but different levels of experience are not perfect substitutes, Borjas finds that an inflow of immigrants into the national labor market puts pressure on the wages of competing workers, by as much as 3%-4% when the increase in labor supply is as large as 10%.
As empirical research points out, the effect of immigration on employment and the wages of natives fluctuates, although most of them seem to cluster around zero. However, still for policy makers equilibrium outcomes are not yet fixed, and hence it is difficult to implement effective immigration policies.
Economic self-interest: perceptions of the impact of immigration on the economic perspective of natives As the real equilibrium effect of immigration continued to be ambiguous, scholars shifted their attention to perceptions of the effects of immigration, rather than the actual effects. Negativity and fears regarding immigration constitute a potent source of polarization within a lot of Western societies. Gaining insight in which individual factors contribute to the formation of negative believes about immigrants might even be of more value than the actual equilibrium effects, as policymakers can use these insights to take away fears and gain more public support regarding immigration policies.
2 Friedberg and Hunt (1995) conclude similar things for the US. They find no evidence for a deterioration of
6 An interesting finding is presented by Hainmueller et al. (2007), based on the European Social Survey which was performed over 22 countries in 2003. They conclude that educational attainment is the most important driver of the formation of perceptions, rather than economic self-interest.3 This finding is
reinforced by later research of Hainmueller et al. (2010), where they conclude that the correlation between education and immigration stems not from economic self-interest4, but from cultural values and
beliefs.5
However, another line of scholarship continues to argue that besides education, there is a significant importance of individual economic factors in the formation of perceptions towards immigrants. Burns et al. (2000) are the first ones to prove that downward economic trajectories enhance stereotypical thinking (i.e. generalize the characteristics of outside groups and assess them negatively) within society, as people’s financial situation deteriorates during recessionary periods.6 Citrin et al. (1997) conclude
that amongst other things, beliefs about the state of the national economy and anxiety over taxes7 play a
significant role. Miller (2012), who uses the 2007 Global Attitudes Project (PGAP) which entails data on 47 countries, finds that self-assessed poverty enhances anti-immigrant sentiments significantly, independent of the effects of objective measures such as education and employment.
Reassessing the role of individual economic factors
One thing that can be concluded from previous research done by numerous scholars is that education is an important driver for explaining attitudes towards immigrants. However, some scholars tend to disregard other economic factors, and in particular one’s personal financial situation, as being an unimportant explanatory variable (Hainmueller et al. (2007) (2010), and Hainmueller et al. (2014)). My analysis will try to prove that this is ungrounded. I base this hypothesis on two theoretical arguments.
3 Hainmueller et al. (2007) were the first ones to prove that the Factor Proportions Analysis is not true, as
high-skilled workers are more supportive of all kinds of immigrant inflows (both low-high-skilled and high-high-skilled). Hence, other things than economic fears and self-interest must lie at the heart of anti-immigrant sentiments. This finding is reinforced by the article of Hainmueller et al. (2015), as they find that different workers in very different industries share similar thoughts on immigration, which again nullifies the Factor Proportions Analysis and hence the importance of one’s labor market position in explaining anti-immigrant sentiments.
4 I.e. the better economic position associated with a higher education does not explain the formation of their
attitudes toward immigrants.
5 Other scholars also emphasized the importance of education in explaining attitudes toward immigration, for
example Espenshade and Calhoun (1993), and Card et al. (2012).
6 Dancygier and Donnely (2013) also find that national economic conditions play a significant role in the
formation of attitudes, as opposition to immigration increased during recessionary periods due to enhanced fears by people for their individual financial situation. Kehrberg (2007) also emphasized the importance of the state of the national economy.
7 Immigration policies receive a lot of resentment due to the tax burden accompanied with it. Immigration
7 First of all, Hainmueller et al. (2010) prove that not economic self-interest, but rather cultural values and beliefs is the source of correlation between education and immigration. People with a higher educational attainment are expected to postulate less anti-immigrant sentiments, as education is associated with the learning of tolerance (Jackman (1978)). However, as Burns et al. (2000) point out, education is not synonymous for the learning of tolerance. Well educated people may harbor the same stereotypical prejudices as low-educated people, but it might not surface due to economic well-being. Secondly, as Markaki et al. (2013) point out, not discriminating between regions but just assessing country level effects, might show different results than when accounting for regional differences. Research up to now focused extensively on country level effects, not discriminating between regions within a country. As immigrants tend to cluster within a few regions (Dustmann et al. (2001)), this might impose that different variables might be relevant within the regions in which immigrants cluster and the regions in which few immigrants live.
Hence, I will reassess the effect of a variable often disregarded by scholars, the current level of household income8, on the formation of attitudes towards immigrants. I will let proximity to immigrants
(which can function as a proxy for the learning of tolerance) act as a moderator on the effect of the current level of household income, something that to my knowledge has never been done before. This way, regional differences can be accounted for. I expect that the current level of household income is important in explaining attitudes towards immigrants, at least for some levels of proximity.
Proximity as a moderator
The direction in which proximity to immigrants influences perceptions of natives is not straightforward. Does it enhance economic fears amongst natives, or does it reduce those fears? Some scholars find evidence for the effect of proximity on attitudes towards immigrants to be negative, in the sense that it creates hostility. Others find that proximity creates mutual understanding, leading to a decline in anti-immigrant sentiments amongst natives.
For example, Citrin et al. (1997) investigate the relationship between proximity and immigration sentiments in the US. They conclude that an increased inflow of immigrants does have significant impact on anti-immigrant sentiments and the upswing of restrictive movements. They find an increase in the number of hate crimes, and an increase in people starting initiatives aimed at eliminating social benefits for illegal immigrants in states where immigrant inflows were large9. Olzak (1992), in his turn, finds
8 Like Burns and Gimpel (2000), Forbes (1997) and Cummings (1997).
9 For similar results, see also Fetzer (2000), Gang and Rivera-Batiz (1994), Dustman and Preston (2001), Bauer,
8 that inflows of immigrants in communities intensified the competition for economic factors between natives and newcomers, and thus anti-immigrant sentiments flourished.
Therefore, I hypothesize that as interaction with foreigners increases (i.e. higher proximity), hostility towards them increases as well. The moderating effect of proximity is such that as proximity increases, the effect of income increases.
On the other hand, Burn et al. (2000) find that people living in proximity to foreigners are less negative of foreigners than people living in rural areas where generally less foreigners settle. However, they note that it is dependent on the level of integration of foreigners, as more integration leads to less hostility amongst natives and less integration means more hostility. If immigrants cluster in their own ethnic groups and do not integrate, contact between natives and foreigners actually decreases despite the fact that they live in proximity to each other. If assimilation is sufficient, an increased interaction with foreigners decreases the probability that negative perceptions are formed towards them (Bauer et al. (2000))10. Glaeser (2005) finds that “hatred relies on people accepting, rather than investigating,
hate-creating stories”. According to his theory, people living far away from immigrants have incentive to accept hate creating stories surrounding immigrants, as the cost of investigating whether it is actually true is high (i.e. they have to move to more immigrant invested areas to see whether it is actually true). People already living in close proximity, do not have to make this investment, and most often the hate creating stories do not influence their perceptions as they experience the opposite in real life.
Hence, I formulate a second hypothesis regarding the effect of proximity. I hypothesize that proximity enhances tolerance towards immigrants. I expect the effect of proximity to moderate the effect of income in such a way, that as proximity increases, the effect of income fades away due to the learning of tolerance associated with proximity.
3. Data:
The dataset used to assess the predictions comes from the survey titled ‘Political Ideology’, which was conducted in 2016 and funded by the Rijksuniversiteit Groningen. It constitutes cross-sectional data on household in the Netherlands, and in total 2.460 respondents of different social classes and income groups completed the survey. The survey gives insight in, amongst other things, the income position of households. The different households entail breadwinners with a high educational attainment and relatively low income, but also breadwinners with a low educational attainment and relatively high income11. This way, the effect of income can be crystallized, despite peoples’ level of educational
attainment. On top of that, the survey gathered respondents’ views on immigration both on economic
10 See also McClaren (2003) or Ellison, Shin, and Leal (2011) who find that natives living in proximity to
foreign-born people are less likely to favor expulsion sentiments toward legal immigrants.
9 grounds and cultural grounds. The measure of proximity was obtained by combining data from the ‘Political Ideology’ survey, where respondents were placed in COROP areas, and data from the Central Bureau of Statistics (CBS) on the number of non-western immigrants in those COROP areas in 2016. Dependent variables:
As my dependent variable is attitudes towards immigrants on economic grounds, first I focus on the question regarding perceived labor market competition from immigrants:
“Someday, it will get to the point where Dutch natives will get fired so ethnic minorities can be hired”12
Respondents could answer this question on a scale ranging from 1 to 5 (totally disagree, disagree, neutral, agree, totally agree). I recoded this variable a bit for presentation purposes later on, where totally disagree and disagree where merged, and totally agree and agree were merged as well, making it into an ordinal variable of 3 categories called Et2Cat (1 = totally disagree, disagree, 2 = neutral, 3 = agree, totally agree)13.
The second question under consideration regards the perceived financial threat to one’s individual financial situation due to immigrant inflows:
“I am afraid that my financial perspectives will deteriorate due to the presence of ethnic minorities” Again, respondents could answer this question on a scale ranging from 1 to 5 (totally disagree, disagree, neutral, agree, totally agree), and it was transformed in the same way as presented above, making it into an ordinal variable of 3 categories called Et5Cat (1 = totally disagree, disagree, 2 = neutral, 3 = agree, totally agree).
Independent variables:
The first independent variable in my model is the level of net monthly income of the household. Respondents were asked to make known the net monthly income of their respective household. Some missing values were reported, due to people not willing to tell their household’s net monthly income, or due to people not knowing their net monthly income. Those respondents were excluded from the sample. For presentation purposes, the variable Inccat was created14. Four categories of household income are
distinguished. The first category, which is labeled ‘1’ in the model, is the household that has a net monthly income of €1.150,- or less. The second category, labeled as ‘2’ in the model, is the household that has a net monthly income between €1.151,- and €1.800,-. The third category, labeled as ‘3’ in the model, is the household that has a net monthly income between €1.801,- and €2.600,-. The final
12 I took the liberty to translate the questions from the questionnaire from Dutch to English.
13 Doing this does not fundamentally change the outcomes of my analysis, as is shown by the Robustness
Analysis in section 6.
10 category, labeled as ‘4’ in the model, is the household that has a net monthly income of €2.600,- or more.
The second independent variable used in the model is proximity, which I define as the ratio of non-western immigrants to natives. In order to get the ratio, I first looked at the ‘Political Ideology’ survey, which contains data on the postal codes of the respondents. The postal codes of the households were combined with the regions in which they fell. The Netherlands distinguishes 40 COROP-regions. However, this research distinguishes 39 COROP-regions, as COROP-region Oost-Groningen and Delfzijl e.o. are bundled in one category called Oost-Groningen. Also, COROP-regions IJmond and Zaanstreek are bundled in one category. Finally, the city of Almere is unbundled from COROP-region Flevoland and forms a different category.
I combined the COROP-regions codes in the dataset with data of the CBS on the number of non-western immigrants per COROP region in 2016. Non-western immigrants are in the dataset defined as first- and second generation non-western immigrants. Later generations of non-western immigrants (third, fourth, etc.) are considered natives. The ratio of non-western immigrants to natives was calculated per COROP-region15, and the code given to the different COROP-regions was replaced by this ratio. This way, the
COROP-regions are defined by the percentage of non-western immigrants to natives. In my model, this variable is called Proximity16.
Proximity: The proximity to immigrants differs per COROP-region, where the smallest ratio of non-western immigrants to natives in the sample is 3,78% and the highest ratio of non-non-western immigrants to natives is 29,69%. In the analysis, for presentation purposes, I categorized these regions by margins of 3%. The first category ranges between proximity levels of 2,97% - 5,97% (far proximity). The category for the highest ratio of non-western immigrants to natives (close proximity) ranges from 26,98% - 29,97%. This distribution is shown in Table 1.
Table 1: Distribution of the COROP-regions in categories of proximity
Category COROP-regions Proximity range
1 Zuid Friesland Noord Friesland Zuidoost Drenthe Zuidwest Drenthe Noord Drenthe Oost Groningen Achterhoek Zeeuwsch Vlaanderen Noord Overijssel Overig Zeeland 2,97% - 5,97%
15 Regions in the Netherlands are divided in so called COROP-regions as of research purposes. The appendix
provides an overview of these regions including a proximity ratio describing the number of immigrants to natives.
16 See Table A1, provided in the appendix, for a detailed overview of the variable Proximity. Table A1 presents
11 2 Zuid Limburg Midden Limburg Noord Limburg Veluwe Zuidwest Gelderland Kop van Noord-Holland Overig Groningen
Noordoost Noord-Brabant Agglomeratie Leiden & Bollenstreek 5,98% - 8,97% 3 Zuidwest Overijssel Twente Alkmaar e.o. West Noord-Brabant Zuidoost Noord-Brabant Midden Noord-Brabant Oost Zuid-Holland Zuidoost Zuid-Holland Arhnem/Nijmegen Het Gooi/Vechtstreek Delft en Westland Flevoland 8,98% - 11,97% 4 Agglomeratie Haarlem 11,98% - 14,97% IJmond en Zaanstreek Utrecht 5 14,98% – 17,97% 6 17,98% - 20,97% 7 Groot Rijnmond 20,98% - 23,97% 8 23,98% - 26,97% 9 Groot Amsterdam Agglomeratie ’s Gravenhage Almere 26,98% - 29,97%
Note: For some categories there is no COROP-region under consideration. Still, the marginal effects of these values can be calculated.
Control variables:
12 in which they could not buy foods or clothing. However, the comparison with others who were worse off made them assess their own wealth as good. For developed countries, I argue that this might be an important factor as well. One’s contentment with its level of income is partly determined by the comparison with its peers. Someone who is in a high income class, may therefore assess his wealth as not good due to the fact that he assesses his own wealth as worse than that of its pears. To control for this effect, I include the variable self-assessed poverty (1 = the respondent is unhappy about his relative level of current income, 0 = the respondent is happy about his relative level of current income).
Other control variables included in the model that might influence attitudes towards immigrants, come from the article of Gang, et al. (2002). The number of members in the household might contribute to the attitudes towards immigrants. Gang, et al. (2002) find that as the number of household members increases, the attitudes towards foreigners become more positive. This may be due to the fact that large households have children that mix with the children of immigrants, and hence increase understanding and reduce the prejudices of native adults.
To control for this effect, the variable memhh (members of the household) is included in the analysis (1 = the household consist of 1 member, ranging up to 9, where 9 = the household consists of 9 members or more).
To control for the effect of current employment status (i.e. a person in general might perceive a higher level of job market competition and might fear more for its financial wellbeing, when unemployed), I include the dummy variable unemploydum. This dummy describes the people who could participate in the labor market (so not the elderly, the incapacitated and children), but do not (1 = looking for work after job loss, looking for work for the first time, caretaker of the housekeeping of the household, works an unpaid job for retention of benefits, and 0 = otherwise).
Markaki et al. (2013) emphasize the effect of regional characteristics on the formation of attitudes. However, there might be a self-selection bias in my results, as it is possible that people who are more likely to view immigrants as a threat move further away from immigrants to more rural areas, and vice versa. If this is true, the effect of proximity on income might be biased. Dustmann et al. (2001) argue that this bias is unlikely to happen in larger regions. To account for this bias, I control for regional effects. I distinguish the control variable region (1 = the most urban area in the Netherlands, consisting Amsterdam, Rotterdam and The Hague, 2 = the West of the Netherlands, excluding the three big cities just mentioned, 3 = the North of the Netherlands, 4 = the East of the Netherlands, 5 = the South of the Netherlands).
13 negatively affected by an inflow of low-educated immigrants as there is no direct job competition and no direct threat to their financial wellbeing.
The survey asked the respondents to select the highest level of education they had attained, from now on called educat (1 = primary education, 2 = VMBO, 3 = HAVO/VWO, 4 = MBO, 5 = HBO, 6 = WO).17
The direction of the effect proximity and income
In order to ease the reader into the analysis that follows, I first show some preliminary graphs. Figure 1 and 2 display graphs of the effects of the independent variables on the two dependent variables. It displays the separate effects of proximity and income, without the inclusion of the interaction term.18
Figure 1A shows that the probability of perceiving labor market competition decreases as proximity increases. Interestingly, figure 1B shows that the effect of income in non-linear, i.e. the probability of perceiving labor market competition does not decrease when income increases. This suggests that people in different income categories perceive immigrants differently. Figure 1C shows that proximity does not seem to have a large impact on the probability of perceiving a financial setback. Figure 1D shows a similar pattern as figure 1B, whereas the probability of a perceived financial setback due to immigrant inflows does not linearly decrease as income increases.
Figure 1: Perceived Labor Market Competition and Perceived Financial Setback
17 See Table B5 of the Appendix for descriptive statistics.
18 In order to obtain the Figure 1 and 2, I estimate a probit model with dummies for my dependent variables.
.2 5 .3 .3 5 .4 .4 5 P ro b a b ili ty o f p e rc e iv in g l a b o r m a rk e t c o m p e ti ti o n 1 2 3 4
Net monthly household income in categories
Perceived Labor Market Competition with 95% Confidence Intervals
.2 .2 5 .3 .3 5 .4 P ro b a b ili ty o f p e rc e iv in g l a b o r m a rk e t c o m p e ti ti o n .0296 .0596 .0896 .1196 .1496 .1796 .2096 .2396 .2696 .2996 Proximity ratio
Perceived Labor Market Competition with 95% Confidence Intervals
14 Figure 2A shows a reversed effect of proximity as compared to Figure 1A, whereas the probability of not perceiving labor market competition increases as proximity increases. Figure 2B also shows a reversed effects as compared to figure 1B. The effect of income is non-linear, as the probability of perceiving no labor market competition first decreases before it starts to increase again as income increases. Figure 2C shows no large impact of proximity on the probability of perceiving no financial setback. Figure 2D again shows a reversed effect as compared to figure 1D, where the probability of perceiving no financial setback does not increase linearly as income increases.
Figure 2: No Perceived Labor Market Competition and No Perceived Financial Setback
.1 5 .2 .2 5 .3 P ro b a b ili ty o f p e rc e iv in g a f in a n c ia l s e tb a c k .0296 .0596 .0896 .1196 .1496 .1796 .2096 .2396 .2696 .2996 Proximity ratio
Perceived Financial Setback with 95% Confidence Intervals
Figure 5B Figure 2A .3 5 .4 .4 5 .5 P ro b a b ili ty o f n o t p e rc e iv in g l a b o r m a rk e t c o m p e ti ti o n .0296 .0596 .0896 .1196 .1496 .1796 .2096 .2396 .2696 .2996 Proximity ratio
No Perceived Labor Market Competition with 95% Confidence Intervals
Figure 2B .1 5 .2 .2 5 .3 .3 5 P ro b a b ili ty o f p e rc e iv in g a f in a n c ia l s e tb a c k 1 2 3 4
Net monthly household income in categories
Perceived Financial Setback with 95% Confidence Intervals
.3 .3 5 .4 .4 5 .5 P ro b a b ili ty o f p e rc e iv in g n o l a b o r m a rk e t c o m p e ti ti o n 1 2 3 4
Net monthly household income in categories
No Perceived Labor Market Competition with 95% Confidence Intervals
15 Figure 1 and 2 show us how attitudes change as the current level of household income and proximity change. However, I am interested in the way in which income influences perceptions at different levels of proximity. In the next section, I provide the specification of my model.
4. Model specification:
To estimate the relevant coefficients an order probit model is used, as the dependent variable is an ordinal categorical variable. If a linear probability model was used the estimations would suffer from probabilities outside the [0,1] interval, and the error term would have a highly non-normal distribution and would suffer from heteroskedasticity. To test the effect of income and proximity on attitudes toward immigration on economic grounds, the following interaction model is constructed:
𝐸[𝑌] = 𝛷[𝛽0+ 𝛽1(𝐼𝑛𝑐𝑐𝑎𝑡) + 𝛽2(𝑃𝑟𝑜𝑥𝑖𝑚𝑖𝑡𝑦) + 𝛽3(𝐼𝑛𝑐𝑐𝑎𝑡 ∗ 𝑃𝑟𝑜𝑥𝑖𝑚𝑖𝑡𝑦) + 𝛽4𝑋 + 𝜀]
Y in this case can be either two of the dependent variables under consideration, that together constitute attitudes toward immigration on economic grounds. The first one is perceived labor market competition induced by immigration, the second one is perceived financial threat induced by immigration. Inccat captures the effect of the category of household income, Proximity is the variable that describes the density of immigrants to natives, X is a vector of control variables, and 𝜀 constitutes the error term which has a standard normal distribution.
The interaction term included in this model might enhance multicollinearity, increasing the size of the standard errors and mitigating the explanatory power of the coefficient of the interaction term as its significance decreases. However, in tackling this problem we follow the article of Brambor, Clark and Golder (2006). Although their article mainly focusses on models constituting interaction terms and continuous dependent variables, they note that all key lessons learned also apply for situations in which there is a limited dependent variable, as is the case with the ordered probit model used in this paper. Brambor et al. (2006) state that the problem of multicollinearity is not that big of an issue, because due
.4 .4 5 .5 .5 5 P ro b a b ili ty o f n o t p e rc e iv in g a f in a n c ia l s e tb a c k .0296 .0596 .0896 .1196 .1496 .1796 .2096 .2396 .2696 .2996 Proximity ratio
No Perceived Financial Setback with 95% Confidence Intervals
.3 5 .4 .4 5 .5 .5 5 P ro b a b ili ty o f p e rc e iv in g n o f in a n c ia l s e tb a c k 1 2 3 4
Net monthly household income in categories
No Perceived Financial Setback with 95% Confidence Intervals
16 to the admission of interaction terms in the model the interpretation of the coefficients change from that of a linear model anyway. The coefficients of the constitutive terms do not represent the average effect of that variable, as they did in a linear additive model. Hence, Brambor et al. (2006) teach us that for this research, the coefficients on the constitutive terms (in this paper income category (Inccat) and the proximity to foreigners (Proximity)) cannot be interpreted as the unconditional or average marginal effects, nor can the significance and magnitude of the interaction term say anything about the conditional effect. It can even be possible that the marginal effect of income category on attitudes towards immigrants is statistically significant for relevant values of proximity, even when the coefficient of the interaction term is insignificant.
5. Results:
Table 2 presents the ordered probit estimation for perceived labor market competition due to immigrant inflows. The coefficients of the independent variables cannot be interpreted as direct marginal effects. However, I can say something about the direction in which the effect works.
Not surprisingly, the coefficient for education (Educat) shows a high level of significance. Being in a higher educational category would increase the likelihood of being in a lower category of perceived labor market competition (Et2Cat). The unemployment dummy (Unemploydum), also shows a significant effect. People who are unemployed but could participate in the labor force are more likely to be in a higher category of perceived labor market competition. Self-assessed poverty (Selfasspov) also has some explanatory power. People who assess their wealth negatively with regard to others, are more likely to be in the higher categories of perceived labor market competition. Age (Age) shows a small but significant effect, whereas age increases, the likelihood of being in lower categories of perceived labor market competition increases. It does not care whether one is female or male (Gender) in experiencing labor market competition, as the effect is insignificant. Being married or not (Married) also has some explanatory power, as can be concluded form the high level of significance. When a person is married the probability of being in a higher category of perceived labor market competition increases. The number of household member (Memhh) also seems to have some explanatory power, as when this number increases, the likelihood of being in higher perceived labor market categories increases. Finally, the control variable region does not seem to have any significant impact.
Looking at the current level of household income (Inccat) we can say that being in income category 2 increases the probability that one is in the higher categories of perceived labor market competition (Et2Cat) as compared to a person in income category 119. The same thing holds for income category 3.
When being in income category 4, however, there is an increased likelihood of being in the lower
19 Income category 1 is taken as a benchmark and the other categories (2,3, and 4) are compared with the
17 categories of perceived labor market competition, as compared to income category 1. Examining proximity (Proximity), we observe that the likelihood of being in a lower category of perceived labor market competition increases as proximity increases.
For the interaction term (Inccat*Proximity) the interpretation of the coefficients is as follows. Considering income category 2, the effect of Inccat 2 would be 0.042 + 0.827*Proximity. As proximity increases, the likelihood of being in a higher category of perceived labor market competition increases. The effects of proximity on income category 3 and 4 can be obtained in a similar manner. For income category 3, the effect of proximity is such that as proximity increases the likelihood of being in lower categories of perceived labor market competition increases. Focusing on income category 4, we observe that proximity increases the likelihood of being in higher categories of perceived labor market competition.
However, considering the interaction term and its constitutive terms, we observe that the coefficients are insignificant. As Brambor et al. (2006) teach us, it might still be the case that income has a significant effect on whether or not one perceives immigrants as a source of labor market competition, despite the insignificance of the estimates of Inccat, Proximity, and Inccat*Proximity. Besides, nor the significance nor the sign or magnitude of the interaction term says anything about the conditional effect. Hence, we turn to table 3 and 4 presenting the marginal effects.
18 (0.075) Selfasspov 0.115* (0.065) Age -0.004** (0.002) Gender -0.058 (0.052) Married 0.197** (0.078) Memhh 0.056** (0.028) Region 0.020 (0.022) Cut-off point 1 -0.964 Cut-off point 2 -0.368
Note: Standard errors are reported in brackets, with *** p < 0.01, ** p < 0.05, * p < 0.10; The results are estimated using an
ordered probit model.
Table 3 presents the marginal effect of income at different levels of proximity when people disagreed with the statement that more immigrants increases labor market competition. A person in income category 1 living in far proximity to foreigners is 39,68% likely to not perceive labor market competition from immigrants. A person in income category 1 living in close proximity to foreigner is 47,28% likely to not perceive labor market competition form immigrants. The probability that a person in the lowest income category disagrees with the statement that immigrants lead to more labor market thus increases with 7,60% as we move from far proximity to close proximity. For people in income category 3, we see a similar pattern, where the probability of not perceiving immigrants as a source of labor market competition increases with 9,14% as we move from far proximity to close proximity. As for income category 2 and 4, proximity does not seem to have a moderating effect on the likelihood of not experiencing labor market competition.
What is interesting to notice is that people in income category 1 and 3 are initially less likely to disagree with this statement compared to people in the highest income category (respectively 39,68% and 38,17% versus 47,80%). However, as proximity increases, tolerance increases for both income groups (i.e. the likelihood of disagreeing with this statement increases). Ultimately, being in either income group 1, 3 or 4 does not matter much for the level of tolerance when living in close proximity, and thus the effect of being in income category 1 or 3 becomes obsolete as proximity increases.
19
Table 3: Predicted probabilities of perceiving no labor market competition per income group at different values of proximity
Inccat*Proximity Margins Inccat = 1, Proximity = 1 Inccat = 1, Proximity = 4 0.3968 0.4250 Inccat = 1, Proximity = 9 0.4728 Inccat = 2, Proximity = 1 0.3729 Inccat = 2, Proximity = 4 0.3740 Inccat = 2, Proximity = 9 0.3757 Inccat = 3, Proximity = 1 0.3817 Inccat = 3, Proximity = 4 0.4155 Inccat = 3, Proximity = 9 0.4731 Inccat = 4, Proximity = 1 Inccat = 4, Proximity = 4 0.4780 0.4759 Inccat = 4, Proximity = 9 0.4722
Note: The dependent variable in this analysis is Et2Cat, and we show the results of Et2Cat = 1 (i.e. people disagree with the statement that more immigration leads to more labor market competition). For presentation purposes, we used the categorized variables for the level of current net household income (Inccat) and we categorized the level of proximity by margins of 3%, i.e. Inccat (1= < €1150, 2= €1151 - €1800, 3= €1801 - €2600, 4 = > €2600). Proximity (1 = the percentage of non-western immigrants to natives ranges between 2,97% and 5,97% (far proximity), 4 = the percentage of non-western immigrants to natives ranges between 11,97% and 14,97%, 9 = the percentage of non-western immigrants to natives ranges between 26,97% and 29,97% (close proximity)).
Table 4 displays the marginal effect of income at different levels of proximity when people agree with the statement that an increased inflow of immigrants enhances labor market competition. A person in income category 1 living in far proximity to foreigners is 38,29% likely to perceive labor market competition from immigrants. A person in income category 1 living in close proximity to foreigners is 31,16% likely to perceive labor market competition from immigrants. The probability that a person in income category 1 agrees with the statement that more immigrants lead to more labor market competition decreases with 7,13% as we move from far proximity to close proximity. Again, for a person in income category 3, we see a similar pattern. The probability that a person in income category 3 perceives labor market competition decreases with 8,67% when we move from far proximity to close proximity. Looking at income category 2 and 4, we see that proximity does not influence the probability of agreeing with the statement.
20 When taking a glance at income category 2, we see that proximity does not have a moderating effect on the likelihood of experiencing labor market competition. There is quite a significant difference between the eventual (i.e. when living in close proximity) likelihood of perceiving immigrants as a source of labor market competition compared to income category 1, 3, and 4.
Considering income category 4, we can conclude that the moderator proximity seems to have little impact. The level of tolerance for people in income category 4 is, compared to the other income categories, high over all values of proximity. Apparently, when being in the highest income category, it does not matter whether a person lives in far or close proximity to foreigners.
These results show that it is not said that as income increases, perceived threats decrease. Contrary, it shows that it is extremely important in which income group a person is as to how immigration is perceived initially (at far proximity), but also ultimately (close proximity).
Table 4: Predicted probabilities of perceiving labor market competition per income group at different values of proximity Inccat*Proximity Margins Inccat = 1, Proximity = 1 Inccat = 1, Proximity = 4 0.3829 0.3555 Inccat = 1, Proximity = 9 0.3116 Inccat = 2, Proximity = 1 0.4069 Inccat = 2, Proximity = 4 0.4059 Inccat = 2, Proximity = 9 0.4041 Inccat = 3, Proximity = 1 0.3980 Inccat = 3, Proximity = 4 0.3646 Inccat = 3, Proximity = 9 0.3113 Inccat = 4, Proximity = 1 Inccat = 4, Proximity = 4 0.3070 0.3089 Inccat = 4, Proximity = 9 0.3121
Note: The dependent variable in this analysis is Et2Cat, and we show the results of Et2Cat = 3 (i.e. people agree with the statement that more immigration leads to more labor market competition). For presentation purposes, we used the categorized variables for the level of current net household income (Inccat) and we categorized the level of proximity by margins of 3%, i.e. Inccat (1= < €1150, 2= €1151 - €1800, 3= €1801 - €2600, 4 = > €2600). Proximity (1 = the percentage of non-western immigrants to natives ranges between 2,97% and 5,97% (far proximity), 4 = the percentage of non-western immigrants to natives ranges between 11,97% and 14,97% (medium proximity), 9 = the percentage of non-western immigrants to natives ranges between 26,97% and 29,97% (close proximity)).
Perceived financial setback
21 increases the likelihood of being in a lower category of perceived financial threat due to immigrant inflows. The other exception is the control variable married, which becomes insignificant.
For the interaction term and its constitutive terms, the interpretation is similar to the interpretation as discussed for table 2. As we can observe, the effect of proximity (Proximity) is reversed as compared to table 2. Being in closer proximity now increases the likelihood of being in a higher category of perceived financial threat. Another difference worth mentioning considers our interaction term (Inccat*Proximity). Proximity decreases the level of perceived threat for all income categories. For example, the effect of Inccat 2 would be 0.149 - 0.091*Proximity. As proximity increases, the likelihood of being in a lower category of perceived financial threat increases. The interaction effects of income category 3 and 4 can be interpreted in a similar way. However, as mentioned before, nor the interaction term or its constitutive terms can say anything about the true conditional effect (Brambor et al. (2006)). Hence, we move to table 6 and 7 presenting the marginal effects.
22 Married 0.130 (0.080) Memhh 0.079*** (0.028) Region 0.023 (0.022) Cut-off point 1 -0.229 Cut-off point 2 0.462
Note: Standard errors are reported in brackets, with *** p < 0.01, ** p < 0.05, * p < 0.10; The results are estimated using an ordered probit model.
Table 6 shows us the marginal effect of income at different levels of proximity when people disagree with the statement that immigrants lead to a deterioration of their individual financial situation. What directly catches the eye when looking at the table, is that effect of proximity is reversed for people in income category 1 and 2 as compared to people in income category 3.
A person in income category 1 living in far proximity is 52,93% likely to not perceive immigrants as a potential financial threat, while a person in income category 1 living in close proximity is 45,83% likely to not perceive immigrants as a potential financial threat. So, the probability of not perceiving immigrants as a financial threat decreases by 7,1% when we move from far proximity to close proximity. A similar pattern can be observed for people in income category 2.
However, a person in income category 3 is 45,26% likely to not perceive immigrants as a financial threat when living in far proximity, while a person in income category 3 living in close proximity is 51,40% likely to not perceive immigrants as a financial threat. For people in income category 3, the moderator proximity works in the opposite direction, as the probability of not perceiving immigrants as a financial threat increases by 6,14% as proximity increases.
Taking a glance at income category 4, we see that the moderator proximity seems to have little impact on the effect of income. For income category 4, the likelihood of not perceiving immigrants as a financial threat is roughly the same over all levels of proximity.
Table 6: Predicted probabilities of perceiving no perceived financial setback per income group at different values of proximity
23 Inccat = 4, Proximity = 1 Inccat = 4, Proximity = 4 0.5498 0.5435 Inccat = 4, Proximity = 9 0.5330
Note: The dependent variable in this analysis is Et5Cat, and we show the results of Et5Cat = 1 (i.e. people disagree with the statement that more immigration leads to a financial setback w.r.t. their own financial situation). For presentation purposes, we used the categorized variables for the level of current net household income (Inccat) and we categorized the level of proximity by margins of 3%, i.e. Inccat (1= < €1150, 2= €1151 - €1800, 3= €1801 - €2600, 4 = > €2600). Proximity (1 = the percentage of non-western immigrants to natives ranges between 2,97% and 5,97% (far proximity), 4 = the percentage of non-western immigrants to natives ranges between 11,97% and 14,97%, 9 = the percentage of non-western immigrants to natives ranges between 26,97% and 29,97% (close proximity)).
Table 7 shows the marginal effect of income on perceiving immigrants as a potential financial threat, again at different values of proximity. Similar to table 6, we observe that the effect of proximity is reversed for income category 1 and 2 as compared to income category 3. A person in income category 1 is 5,86% more likely to perceive immigrants as a financial threat when we move from far to close proximity. A person in income category 2 is 5,58% more likely to perceive immigrants as a financial threat when moving from far to close proximity. A person in income category 3, on the other hand, is 5,20% less likely to perceiving immigrants as a financial threat when we move from far to close proximity. For a person in income category 4, the moderator proximity seems to have little impact on how immigrants are perceived.
The effect of income is dependent on the moderator proximity, as we see that people in income category 3 seem to build op tolerance as proximity increases, while people in income category 1 and 2 seem to build op fear as proximity increases. Apparently, the direction in which income influences perceptions of people is not a linear one. It differs per income group how immigrants are perceived, and moreover, it differs per income group in which direction the moderator proximity works. More income and more proximity are not necessarily factors contributing to the acceptance of immigrants, at least for specific levels of income.
Table 7: Predicted probabilities of perceiving perceived financial setback per income group at different values of proximity
24
Inccat = 4, Proximity = 9 0.2358
Note: The dependent variable in this analysis is Et5Cat, and we show the results of Et5Cat = 3 (i.e. people agree with the statement that more immigration leads to a financial setback w.r.t. their own financial situation). For presentation purposes, we used the categorized variables for the level of current net household income (Inccat) and we categorized the level of proximity by margins of 3%, i.e. Inccat (1= < €1150, 2= €1151 - €1800, 3= €1801 - €2600, 4 = > €2600). Proximity (1 = the percentage of non-western immigrants to natives ranges between 2,97% and 5,97% (far proximity), 4 = the percentage of non-western immigrants to natives ranges between 11,97% and 14,97%, 9 = the percentage of non-western immigrants to natives ranges between 26,97% and 29,97% (close proximity)).
6. Robustness analysis:
Table 8 displays my robustness analysis20, as to see whether my results stay roughly the same when
conditions change.
The first two columns test the appropriateness of the form of the dependent variable. The first column displays my dependent variable as constructed in the survey. There are five categories at which people can rank the statement of perceived labor market competition, ranging from totally disagree to totally agree. Although the four cutoff points from the ordered probit estimation with five categories of the dependent variables can be distinguished (i.e. there are significant differences between the steps)21, one
can argue that the size of the steps between the categories differs. For example, the step from agree to totally agree (disagree to totally disagree) might be smaller or larger than the step from neutral to agree (disagree). Also, one can argue whether there is additional explanatory power if a person ‘totally agrees’ with a statements versus a person who ‘agrees’ with a statement. Hence, column 2 shows my estimation with three categories (agree, neutral, and disagree).
Column 3 and 4 displays a sample split for the employed. People who do not participate in the labor market are economically more vulnerable. Therefore, they may postulate more anti-immigrant sentiments as they perceive immigrants as more of a threat to their own situation. Hence, I exclude the unwillingly unemployed from the sample, and check whether my results will still hold.
Column 5 and 6 displays a sample split for age. Karlins, Coffman, and Walters (1969) show that stereotypes/prejudice are formed due to preadult socialization. Therefore, the period in which a person comes of age is likely to be of importance for the formation of their attitudes towards immigrants. Hence, I distinguish between people who are retired (65 years and older) and people still participating in the labor force. A clear distinction between the current working population and the retired is that a lot of guest workers were present in the time the retired were still at work (guest workers from Turkey and Morocco came to the Netherlands in the sixties/seventies22 in order to accommodate the economic
20 I only performed a robustness analysis examining the dependent variable ‘labor market competition’ (Et2Cat),
as I consider it sufficient to test the overall robustness of my model.
21 This is not shown in table 8.
22 Source:
25 growth of that time). Current immigrant inflows consist mostly of refugees from the Syrian war (inflow started in 2014)23.24 The discrepancy in reasons for the inflow of ethnic minorities into the Netherlands
might be a source for diverging beliefs between generations.
23 Source: https://www.cbs.nl/nl-nl/nieuws/2018/42/recordaantal-immigranten-en-emigranten-in-2017 24 Also, the number of eastern European workers has increased drastically. They are a direct source of labor
market competition. However, the focus of my research is only on non-western immigrants (ethnic minorities).
Table 8: Robustness Analysis
Dependent variable: Perceived Labor Market Competition
Et2 Et2Cat Employed Unemployed Age ≤ 65 Age ≥ 66
Inccat Inccat 2 0.022 0.042 0.059 -0.039 -0.117 0.470 (0.194) (0.213) (0.253) (0.414) (0.260) (0.420) Inccat 3 0.062 0.047 0.121 -0.289 0.030 0.247 (0.183) (0.201) (0.237) (0.416) (0.242) (0.410) Inccat 4 -0.239 -0.247 -0.234 -0.415 -0.208 -0.253 (0.180) (0.198) (0.233) (0.429) (0.233) (0.412) Proximity -1.142 -0.859 -1.73 -0.606 -1.370 2.241 (1.109) (1.203) (1.439) (2.284) (1.335) (3.055) Inccat*Proximity Inccat 2 0.919 0.827 1.598 -0.693 2.092 -2.774 (1.360) (1.489) (1.776) (2.795) (1.842) (3.281) Inccat 3 0.150 -0.178 -0.319 1.934 0.179 -2.474 (1.228) (1.343) (1.579) (2.774) (1.549) (3.180) Inccat 4 1.154 0.924 1.236 1.767 0.980 -0.510 (1.165) (1.271) (1.497) (2.894) (1.411) (3.173) Educat -0.205*** -0.219*** -0.215*** -0.234*** -0.260*** -0.144*** (0.017) (0.018) (0.020) (0.050) (0.023) (0.032) Unemploydum 0.140** 0.152** 0.154* 0.091 (0.068) (0.075) (0.092) (0.139) Selfasspov 0.137** 0.115* 0.130* 0.016 0.181** -0.010 (0.059) (0.065) (0.072) (0.167) (0.080) (0.115) Age -0.003* -0.004** -0.004** -0.006 -0.004 -0.010 (0.002) (0.002) (0.002) (0.006) (0.002) (0.008) Gender -0.062 -0.058 -0.088 0.201 -0.027 -0.125 (0.047) (0.052) (0.055) (1.760) (0.062) (0.100) Married 0.189*** 0.197** 0.238*** -0.013 0.266*** 0.109 (0.070) (0.078) (0.085) (0.215) (0.098) (0.206) Memhh 0.041* 0.056** 0.050* 0.059 0.041 0.258 (0.025) (0.028) (0.030) (0.073) (0.029) (0.173) Region 0.017 0.020 0.029 -0.039 0.006 0.058 (0.020) (0.022) (0.024) (0.061) (0.027) (0.039) Observations 2.298 2.298 1.978 320 1.594 704 Pseudo R2 0.0398 0.0523 0.0517 0.0525 0.0643 0.0392 Prob > chi2 0.000 0.000 0.000 0.0011 0.000 0.000
Note: In order to get the coefficients I estimated all columns using an ordered probit estimation. Column Et2 presents the estimates for 5 categories of the dependent variable (i.e. 1=totally disagree, 2=disagree,3= neutral, 4=agree, 5=totally agree). Column Et2Cat presents the estimates for 3
26 Taking a glance at the first two columns, which concern the form of the dependent variable, we observe that the estimates are quite robust. However, when taking a look at the interaction term (Inccat*Proximity), it shows that for Inccat 3 the effect becomes reversed. Despite the notion that the coefficients of the interaction term do not say anything about the true conditional effect (Brambor et al. (2006)), it does compromise the generalizability of my results a bit.
Column 3 shows that the results of my analysis stay roughly the same, when I exclude the unemployed from the sample. Column 4 on its turn shows quite dissimilar results compared to the column of reference (column 2). As my results are generalizable for the employed, the same cannot be said for the unemployed. Hence, it is a good thing that my model accounts for whether a person is unemployed or not (i.e. included as a control variable), as this increases the robustness of my analysis.
Column 5 and 6 show that a sample split for age does change the outcomes of the analysis. It is difficult to say anything about our coefficients of interest (i.e. the current level of household income), as a marginal effects table should be derived. However, we can see by looking at the control variables (for example Educat), that there are differences between people still participating in the labor force and those who are pensioners. For people still participating in the labor force, the results of my analysis seem quite robust. For pensioners, my results seem less robust.
So, the results of my estimation are not extremely robust. As Guiliano and Spilimbergo (2009) point out, especially for studies on belief formation, one easily omits relevant variables. In order to further increase the robustness of my analysis, future research should focus on which variables to include as to make my results more generalizable.
7. Conclusion:
In this paper I investigated the role of the current level of household income in explaining attitudes towards immigrants based on economic grounds, for residents of the Netherlands. I accounted for regional differences following Markaki et al. (2013) by including the moderator proximity to immigrants into the analysis, and assessed which impact income had at different values of proximity. In order to perform this analysis, I used data from the survey ‘Political Ideology’ (2016) made available by the University of Groningen which contained information on the perceptions of people towards immigrants. I combined this survey data with data of the CBS on the ratio of first and second generation immigrants to natives in COROP-regions in the Netherlands (i.e. my measure of proximity). The results were estimated using an ordered probit model.
27 of immigrants), and for some levels conditional on proximity. Moreover, I find an adverse effect of the moderator proximity. There is no outspoken direction in which this moderator works, as it depends on the dependent variable under consideration.
An important challenge in this thesis was to prove that economic self-interest is still an important driver in the formation of perceptions. Hainmueller et al. (2010) state that education is the most important driver of attitudes, and disregard the effect of economic factors such as the current level of household income. I prove that this is ungrounded. Following the article of Markaki et al. (2013), I show that when regional differences are accounted for, the current level of household income is important in explaining attitudes towards immigrants.
The insights provided by this thesis can be of societal importance, as policy makers can use this information in order to come up with immigration policy that has public support. Policy makers should account for regional differences, and hence an important role can be played by local politics in the success of immigration policies.
Although my findings impose a role for the current level of household income, further research is needed. My results show that the effect of income is non-linear. This might suggest that income becomes irrelevant after it passes a certain threshold. At which thresholds income is significant is beyond the scope of this thesis and is something further research should focus on.
A caveat is in order. As Hainmueller et al. (2010), Hainmueller et al. (2014), and Card et al. (2012) find no effect for the current level of household income, my findings suggest otherwise. An explanation (besides the fact that my research accounts for regional differences) for the fact that my results differ from scholars emphasizing education, comes from Segovia and Defever (2010). They state that the results of immigration research depend on the specific questions being asked. Differences in the questionnaires might be part of the explanation for the mixed results.
28
References:
Bauer, T.K., Lofstrom, M., and Zimmerman, K.F. (2000). Immigration Policy, Assimilation of Immigrants and Natives’ Sentiments Toward Immigrants: Evidence From 12 OECD Countries. Swedish Economic Policy Review, 7, 11 – 53.
Borjas, G.J. (1995). The Economic Benefits From Immigration. The Journal of Economic Perspectives, 9(2), 3 – 22.
Borjas, G.J. (1999). The Economic Analysis of Immigration. Handbook of Labor Economics, 3(A), 1697 – 1760.
Borjas, G.J. (2003). The Labor Demand Curve is Downward Sloping: Reexamining the Impact of Immigration on the Labor Market. The Quarterly Journal of Economics, 118, 1135 – 1174.
Borjas, G.J., Freeman, R.B., Katz, L.F., DiNardo, J., and Abowd, J.M. (1997). How Much Do Immigration and Trade Affect Labor Market Outcomes? Brookings Papers on Economic Activity, 1997(1), 1 – 90.
Brambor, T., Clark, W.R., and Golder, M. (2006). Understanding Interaction Models: Improving Empirical Analyses. Political Analysis, 14(1), 63 – 82.
Burns, P., and Gimpel, J.G. (2000). Economic Insecurity, Prejudicial Stereotypes, and Public Opinion on Immigration Policy. Political Science Quarterly, 115(2), 201 – 225.
Card, D. (2005). Is The New Immigration Really So Bad? The Economic Journal, 115(507), 300 – 323. Card, D., Dustmann, C., and Preston, I. (2012). Immigration, Wages, and Compositional Amenities. Journal of European Economic Association, 10, 78 – 119.
Chomsky, A. (2018). “They Take Our Jobs!”: and 20 Other Myths About Immigration. Beacon Press, Boston, Massachusetts.
Chiswick, B.R. (2005). The Economics of Immigration, Edward Elgar Publishing, Cheltenham, U.K. and Northampton.
Citrin, J., Green, D.P., Muste, C., and Wong, C. (1997). Public Opinion Toward Immigration Reform: The Role of Economic Motivations. The journal of Politics, 59(3), 858 – 881.
Coenders, M., and Scheepers, P. (1998). Support for Ethnic Discrimination in the Netherlands 1979 – 1993: Effects of Period, Cohort, and Individual Characteristics. European Sociological Review, 14(4), 405 -422.
