What are the effects of individual characteristics on the
attitude people have towards immigrants and their
preference for social welfare spending?
Evidence from Dutch households
Abstract: This paper uses individual-level survey data from Laméris, Jong-A-Pin and Garretsen (2019) to analyze the characteristics that have a direct effect on a person’s attitude towards social welfare state spending. It also tries to test if some characteristics might have an indirect effect, through a person’s attitude towards immigrants. For this research, a factor analysis will be used, as well as multivariate linear regressions. Based on the models, the attitude towards immigrants is found to be exogenous and has an insignificant effect on a person’s attitude towards social welfare. A person’s age, perspective of his/her own financial situation and having a nationalistic view are highly significant in explaining the attitude towards social welfare.
Keywords: immigrants, social welfare, attitudes, preferences, individual-characteristics
Sanne van der Laan (S2734508)
Master thesis: Economics
University of Groningen
Supervisor: dr. R.M. Jong-A-Pin
11 – 01 – 2019
2
1. Introduction
Nowadays, there live around 4 million immigrants in the Netherlands (CBS, 2017). The inflow of more immigrants has led to economic, social and political changes and to some degree of discomfort among Dutch citizens. People disagree about the degree to which immigration should be reduced, expanded or if it should be kept at the same level. The rights of these immigrants, together with the Dutch immigration policy are also issues that are debated strongly within the Netherlands. Due to the increased immigration in Europe, and especially the Netherlands, there is a more heterogeneous society which leads to new pressures. Ethnicity has led to social division and started a political debate in the Netherlands. More (poor) people lead to more pressure on the welfare state, which leads to challenges and could also lead to a change in opinion about welfare spending. There is an increasing concern about the trade-off between a generous immigration policy and a generous welfare state. In Europe, there is a strong debate whether or not to offer the same welfare benefits and services to immigrants as to non-immigrants.
Immigration is a topic that is often debated in society and in academics. In the Netherlands, there are many different opinions about immigration and the rights of immigrants. What are the consequences of the attitude people have towards immigrants for the belief about the Dutch social welfare state? Knowing what and how people think is important for politicians since it might lead to different immigration and social welfare policies.
Throughout this research, I study the effect of the attitude towards immigrants on the support for social welfare. This relationship comes from the fact that many people have the idea that the immigrants excessively use the assistance payments and that they do not economically contribute to the welfare state. The percentage of people with a non-western ethnicity depending on assistance allowance is higher compared to people with a western ethnicity (CBS, Jaarrapportage Integratie 2016)1, which might make an inflow of more immigrants potentially harmful for the welfare state.
It is often argued that the debate is distorted by misperceptions and the media. People might have the idea that immigrants, when entering the Netherlands, have direct access to all the social services. However, many services depend on the amount build up during your working life. The media always portrays immigrants as being poor or low-educated, which makes people scared that a large inflow of immigrants increases the costs to the welfare state2. This concept is linked to the self-interest
1 In 2015: 13,8 percent of the people with a non-western ethnicity were dependent on assistance allowance, while this percentage was only 4,4 for people with a western ethnicity.
3 argument that states that native residents could be negative about an inflow of immigrants who pay less in taxes than they receive in benefits (Card, Dustmann and Preston, 2005).
There exists a literature gap about the role of attitudes in explaining the relationship between immigrants and social welfare. There are however some concepts found by previous researchers working in this field. The study of Dustmann and Preston (2007) finds that welfare concerns are most relevant when people think about the effect of immigration on the economy. This finding makes this thesis more relevant, since it studies how the attitude people have towards immigrants influences their support for social welfare. The study of Senik, Stichnoth and van der Straeten (2009) find a strong positive relationship between the attitude towards immigrants and their support for the welfare state. While other research is often related to politics and policies, this thesis is more focussed on economics.
This thesis contributes to the existing literature by using a large representative sample of Dutch households that contains an extensive number of characteristics to provide a broad picture of the drivers of the attitude towards social welfare. Due to the use of survey data, attitudinal heterogeneity within a population can be portrayed and examined3. The use of individual data concerning the attitudes of people is interesting, since the democratic theory of politics states that changes in public opinion will eventually lead to changes in politics and policy (Eger, 2010; Downs, 1957). People’s preferences and attitudes shape the demand side of immigration-policy decisions and are therefore critical. Alterations in immigration policies are not just due to economic considerations but can also be driven by the attitude of the public towards immigration. Knowing what drives this attitude might therefore be important for policy makers. For this research latent variables and factor analysis will be used, which is not often done by previous researchers.
The effect of immigrants on the government depends on the demographic and economic characteristics of the immigrants and the nature of the tax and benefit system in a country. This will be explained using the fiscal burden theory (Hainmueller and Hiscox, 2010) and the welfare magnet theory (Borjas, 1999). Both are used in this thesis and will be discussed in further detail in the literature section.
The main research question of this research is: What is the effect of the attitude people have towards immigrants on the attitude towards social welfare? Besides this research question, there are several sub-questions that will also be discussed in this article:
4
Sub-question (1): Do demographic and or socioeconomical variables have impact on the attitude people have towards immigrants?
Sub-question (2): Do demographic and or socioeconomical variables have impact on the attitude people have towards the social welfare system?
Sub-question (3): Is there evidence for the fiscal burden theory for the Netherlands, based on the survey data?
The first part of this thesis consists of a literature survey of earlier research about the link between immigration, social welfare and the importance of attitudes. The second part is about the data that will be used for this research. The third part describes the start of the research, where the focus is on the factor analysis. In the fourth part of this thesis the model and the estimated regressions are explained and estimated, together with some robustness checks. In the last part, the shortcomings and the conclusion will be stated that is derived from the results of the regressions.
2. Literature overview
This literature overview starts with the discussion of the interaction between immigration and social welfare. Besides looking at previous research for the relationship between immigration and social welfare, previous research is used to study which variables have an impact on this relationship. It will be shown that the attitude people have towards immigrants is dependent on demographic, economic and other individual characteristics4. Besides having an impact on the attituded people have towards immigrants; these characteristics might also have an impact on the attitude towards social welfare. To identify which individual-level variables could have an impact on which attitude (either towards immigrants or towards social welfare), the studies of several researchers will be used. At the end of each literature review section, the relevant hypotheses are stated.
Immigration and social welfare:
First of all, the link between immigration and social welfare is explained. Stichnoth and van der Straeten (2013) have done a literature review about the relation/trade-off between a generous immigration policy and a generous welfare state. They concluded that the evidence is mixed. Most
5 researchers in this field, found a significant relationship between the level of ethnic diversity and public spending (Mau and Burkhardt, 2009; Alesina, Baqir and Easterly, 1999). However, this negative relationship is only found to be weak. While there is much research about the level of ethnic diversity on public spending, there is not much research about the effect of people’s preferences about ethnic diversity on the support for public spending. This thesis, studies exactly this interaction between the attitude towards immigrants (and therefore ethnic diversity) and the support for public spending.
There are many different arguments used in debates about the relationship between immigration and redistribution. The fiscal burden theory (Hainmueller and Hiscox, 2010) is one of the two important economic theories for explaining the anti-immigrant attitude people might have. The fiscal burden model focuses on the effect of immigrants on the government budget (including social welfare). The model assumes that low-skilled immigrants are a burden for public services, while high-skilled immigrants are beneficial due to the fact that the government receives more tax revenues. In a country with a progressive tax system (as is the case for the Netherlands), the high-skilled people have more to lose than the low-skilled people if the tax rate increases due to an inflow of immigrants. The different types of public services that are part of the welfare state are often only accessible to poor citizens. According to the fiscal burden model, rich natives favor highly skilled immigration more than the poor natives do. However, this part of the fiscal burden model is rejected by Hainmueller and Hiscox (2010). The results of their research confirm the part of the fiscal burden model that states that low-skilled natives are concerned about an inflow of low-skilled immigrants since their own benefits might decrease. This would indicate that people with a higher educational level are more positive about immigration. This is one of the effects that is tested in this thesis.
A group of researchers (Garand, Xu and Davis, 2015; Soroka et al.,2015) study this specific field of research and find that immigrants in America are more likely to be low-skilled, jobless or poor. These characteristics make them more likely to be recipients of welfare benefits. A large inflow of immigrants will therefore likely drive up the costs of the welfare state (especially in generous welfare states). Due to the lower socioeconomic status and the higher social welfare-program participation rates for immigrants, the attitude towards the welfare state is likely to be at least partly influenced by the attitude towards immigrants. Garand, Xu and Davis (2015) find that the attitude towards immigration and immigrants, are important components of how the citizens care about welfare5. While both studies6 agree on this possible explanation of the relationship between immigrants and social welfare, they differ in other explanations given for this relationship. Soroka et al. (2015) mention the ‘welfare magnet theory’ (Borjas, 1999), which means that low-skilled migrants will cluster in generous welfare
5 Based on their study of America
6 states. Migrants participate in welfare shopping, since the cost of migrating has to be offset by the lower wage dispersion and the more generous welfare state. Garand, Xu and Davis (2015) mention two other explanations to get an understanding of this relationship. The first one mentioned is the effect of stereotypes about immigrants. Americans might wrongly think that the majority of the welfare recipients are (poor) immigrants. These stereotypes affect their evaluation abilities and alter their perceptions and attitudes towards immigrants. As van Oorschot (2006) stated, immigrants are perceived to be the least deserving group in the American society to benefit from welfare policies. The last possible explanation given by Garand, Xu and Davis (2015) is related to higher racial and ethnic diversity. Immigrants could be seen as not equally deserving the help from the government, since they have different values and a different culture compared to the natives. As the racial diversity increases, more of the social welfare benefits go to people who are not similar to them7. The support for social welfare therefore decreases once the ethnic heterogeneity increases.
The influence of racial and ethnic diversity on the support for the welfare state is researched by several other researchers as well (Banting and Kymlicka, 2006; Alesina, Baqir and Easterly, 1999; Habyarimana et al., 2007; Luttmer, 2001; Mau and Burkhardt, 2009). This group of studies all conclude that ethnic diversity hampers the support for social welfare. In the study of Banting and Kymlicka (2006), two important arguments are mentioned that involve the role of diversity in the relationship between immigration and redistribution. One of them is the heterogeneity/redistribution tension, which states that ethical and racial diversity is likely to weaken redistributive social policies. This is due to the fact that people do not feel trust or national solidarity with people from other ethnicities. The other argument mentioned is called the recognition/redistribution tension (Banting and Kymlicka, 2006). This means that multiculturalism policies often exacerbate tensions between diversity and social solidarity. This eventually further weakens the support for redistributive policies. Luttmer (2001) states that people become more positive about welfare spending if the number of recipients from their own racial group rises (or if the society becomes more homogeneous). People from the same ethnicity cooperate more often, as they adhere to the in-group norms. Mau and Burkhardt (2009) confirm this by stating that an increasing ethnic heterogeneity8 negatively relates with the support for welfare state redistribution. The ethnic diversity is also found to be negatively linked to people’s support for the inclusion of foreigners. Immigrants might be seen as people that violate the standard culture and behavior and therefore don’t deserve the same rights as natives. The feeling of having in-group and out-group norms and people, is something that is studied by many researchers. Garand, Xu and Davis (2015) state that increased immigration, increases the racial and ethnic heterogeneity within a political
7 system. This can impact the national identity and the solidarity among people, which reduces the support for the welfare state.
The fact that ethnic heterogeneity is negatively related with the support for social welfare is clearly visible in the research of Eger (2010), who does research about the effect of immigration on the support for welfare state spending. It is shown that the most recent immigration led to a negative attitude towards welfare spending. In countries with a high proportion of recent immigrants, people are less likely to be positive about welfare spending. For the case of Sweden, an increase in the proportion of the country residents who receive social assistance, leads to a more positive attitude towards welfare spending. This reinforces the theory that a growing amount of in-group people a positive effect has on the attitude towards social welfare. The recent immigration in Sweden together with the proportion of foreign-born citizens makes it more likely that people reduce their support for welfare spending. Immigration was found to be the only country-level variable that negatively influences the attitudes. The more foreign-born residents are in Sweden, the more people react negatively towards welfare state spending.
The results9 found for America, are likely not all applicable for Europe, due to the fact that the institutions, politics and level of trust differ10. According to Luttmer (2001), America is relatively more racially and ethical heterogeneous compared to most western European countries. This higher level of ethnic heterogeneity makes people less positive about social welfare11. The effect of immigration on the welfare state might therefore be more limited for the Netherlands compared to America (Luttmer, 2001). While Soroka et al. (2015) and Garand, Xu and Davis (2015), focus on America, Senik, Stichnoth and van der Straeten (2009) focus on Europe and found a weak negative relationship between the perceived presence of immigrants and citizens’ support for the welfare state. This conclusion is based on the use of the perceived presence of immigrants and differs substantially across areas. When Senik, Stichnoth and van der Straeten (2009) focus on the direct link between the attitude towards immigrants and the amount of support, they do find a strong positive relation that does not depend on the perceived presence of immigrants. In this thesis a similar research will be performed but specified for the Netherlands, rather than Europe.
9 Especially the magnitude of the effects
10 America has a less generous welfare state and a racial division black and white (Garand, 2015). Van Oorschot (2006) mentions that racial stereotyping is common in America and that there is a strong racial element in the American welfare support. The conditions under which diversity unfolds is different in Europe compared to America (Mau and Burkhardt (2009). The different political institutions between America and Europe (proportional representation) might make it more difficult to transfer money to the poor (Alesia, Glaeser, 2004).
8 One has to be careful when interpreting these results as people often have a misperception about the actual number of immigrants within their country. These misperceptions make them more averse towards redistribution. Once people are confronted with the true numbers, people often alter their attitude. Alesina, Miano and Stantcheva (2018) state that the support for welfare spending is more related with the perceived composition of immigrants than with the perceived share of immigrants. However, Naumann and Stoetzer (2018), concluded that peoples’ awareness of migration has no clear effect on the support for redistribution.
Besides looking at the effect on the welfare state using quantitative data about the amount of welfare spending or the number of immigrants, using personal data about the attitudes people have towards welfare spending and immigrants might also be interesting. In the articles mentioned above, it is already shown that the amount of welfare spending might depend on immigration. Research has shown that either the presence of immigrants or individual’s attitudes towards immigrants can influence support for welfare.
The study of Xu and Davis (2015) conclude that people who have negative attitudes towards immigration are more likely to support decreases in welfare spending (compared to people with a more positive attitude towards immigrants). The view people have towards poor people is very important in explaining the attitude people have towards welfare spending. It is stated that there exists a positive relationship between the attitudes people have towards (illegal) immigrants and their attitude towards welfare recipients and to social welfare spending.
Hypothesis 1: The attitude people have towards immigrants is positively related to the attitude towards social welfare.
Individual characteristics that influence the attitude towards immigrants:
9 variables relate to each other12. Mau and Burkhart (2009) looked both at macroeconomic variables and individual characteristics for their research and concluded that age, education and political preference have a considerable impact on the attitude people have towards immigrants. Below the variables that might have an impact on the attitude towards immigrants will be discussed.
Location:
The effect of the degree of urbanization of a person’s residential area is expected to be negatively related to the attitude a person has towards immigrants. This assumption is based on the finding of Espenshade and Hempstead (1996), who find that the place of residence has a strong effect on the attitude people have towards immigrants. People living in rural areas have more negative feelings towards immigrants compared to people living in the city. People do feel more negative towards immigrants if they live in low population/urbanized localities. This might be due to the fact that people living in big cities, are more adapted to living with immigrants (Van Dalen and Henkens, 2005). Not only the type of area the people live in, but also the location is of great importance. Espenshade and Hempstead (1996) found that the northeast part of America is more negatively focussed on immigrants compared to other parts of America. Although America is larger and more diverse, it might be the case that people in the north of the Netherlands have a different, more negative, attitude towards immigrants than the rest of the Netherlands.
Gender:
Dustmann and Preston (2001) found that gender is a significantly important variable that influences the attitude people have towards minorities. It is often stated that women are less prejudiced and have a more positive attitude towards immigrants (Kessler, 2001; Dustmann and Preston, 2001). On the other hand, Mayda (2006) found that males are more likely to be pro-immigration. Scheve and Slaughter (2001) found no evidence that gender had a significant effect on the attitude people had towards immigrants. The effect of gender on the attitude towards immigrants is therefore not as clear-cut. For this research, an insignificant effect of gender on the attitude towards immigrants is expected. This assumption is based on the finding that there is no clear evidence whether or not being a male or a woman has an impact on the attitude towards immigrants. Since the evidence found is mixed, an insignificant effect is expected.
Hypothesis 2: Gender is likely to have no effect on the attitude towards immigrants.
10 State of the person/country:
The feelings people have about their own economic situation or about the current state of the country, also have an impact on how people see immigrants. There is a strong relationship between optimism about the national economy ant the attitudes towards immigrants. People who have a positive attitude towards immigrants, are often those that see their country as having a strong economy (Espenshade and Hempstead, 1996; Citrin et al., 1997). People who have a global perspective about international economic matters, are often those people with a more positive attitude towards immigration compared to people with a more isolationistic view. Mayda (2006) states that people who are more nationalistic and have national pride, are more negative about immigration compared to others. According to Espenshade and Hempstead (1996), two of the most important determinants of the attitude towards immigrants are the sense of alienation and an isolationist mentality. It is also stated that people who feel dissatisfied about how their life is going, are often more negative about immigrants and prefer lower immigrational levels. In this thesis, the effect of a person’s own economic situation is expected to be positively related to their support for immigrants. This is based on the finding that feeling positively about a country is already shown to be positively related to the attitude towards immigrants. Since this is true for a person’s view of a country, it might also be true for a person’s perceptions of his/her own financial situation.
Political:
An interesting contribution to this field of research, is looking at the effect of political preference on the degree to which the respondent is pro-immigration. Scheve and Slaughter (2001) found no evidence that party identification had a significant effect on the attitude towards immigrants, but the effect of ideology was significantly positive. It is found that the level of affiliation with the more “conservative” parties is negatively correlated with the attitude towards immigrants (Mayda, 2006; Scheve and Slaughter, 2001; Espenshade and Hempstead, 1996; Kessler, 2001). This indicates that the more conservative people wanted a more restrictive immigration policy. Being left-wing or right-wing oriented is therefore an important contributor to the attitude people have towards immigrants. Alesina, Miano and Stantcheva (2018) concluded that right-wing oriented voters often have larger misperception about immigration and also have a more negative attitude towards immigration. Meanwhile, left-wing voters are more likely to have a positive attitude towards immigrants (Mau and Burkhart, 2009). These findings confirm the assumption that being left oriented is positively related to a person’s attitude towards immigrants, while being right oriented is negatively related to a person’s attitude towards immigrants.
11 egalitarian views, show a more positive attitude towards immigrants. Due to the fact that people with an egalitarian view want equality for all people, having a strong egalitarian view is assumed to be positively related to the attitude towards immigrants.
Education:
Besides the political preference of people, the educational level of the respondents might also be related to the attitude towards immigrants. While the study of van Oorschot and Uunk (2007) concludes that there is a higher concern about immigrants among higher educated people, the studies of Espenshade and Hempstead (1996), Mau and Burkhardt (2009) and Dustmann and Preston (2007) concluded that people with high education are more likely to have a positive attitude towards immigrants. This conclusion is confirmed by Alesina, Miano and Stantcheva (2018), who state that low-educated people have a larger misperception about immigration and a more negative attitude towards immigration. Dustmann and Preston (2001) find a similar conclusion that states that if a person has been educated beyond the age of 18, it is found to be more positive toward minorities. People with a low education are likely to be more negative towards immigrants. The reason why higher educated people might be more positive might be due to the fact that high educated people often occupy more skilled positions which makes them feel less threatened by labor market competition from unskilled immigrants. According to the labor market competition theory, education and income are expected to be positively related with the attitude towards immigrants. Mayda (2006) concluded that people with high education are more positive abut immigrants in high per capita GDP countries, but are more negative about immigrants in low per capita GDP countries. Although the overall effect of education on the attitude towards immigrants seems to be positive, it might not be clear-cut as Garand, Xu and Davis (2015) found some mixed effects for the variable “education”. For this research, a positive effect of education on the attitude towards immigrants is expected, as people with a higher educational level might be less threatened by an inflow of (poor/low-skilled) immigrants.
Hypothesis 3: Low educated people are likely to have a more negative attitude towards immigrants compared to higher educated people.
Income:
12 towards immigrants (van Oorschot and Uunk, 2007; Card, Dustmann and Preston, 2005). People who are unemployed due to retirement or being disables showed an even more negative attitude towards immigrants (compared to other groups of unemployed people). Since people with a high income are often those people with a higher education, the effect of income on the attitude towards immigrants will be positive13. People who are unemployed might be more vulnerable and scared for an inflow of immigrants, as it narrows their chances of finding a job. The effect of unemployment on the attitude towards immigrants is therefore expected to be negative14.
Age:
Besides income or place of residence, Espenshade and Hempstead (1996) also investigates the influence of age on the attitude towards immigrants. While they find that people aged 18-24 and 45-54 are more tolerable towards immigrants, Scheve and Slaughter (2001) found no evidence that the age had a significant effect on the attitude people had towards immigrants. Some studies find age to have a significant positive effect on the attitude towards immigrants (Kessler, 2001), but other studies find that older people have a strong anti-immigrant attitude (Card, Dustmann and Preston, 2005). It is assumed that the effect of age on the attitude towards immigrants is insignificant, as previous research finds mixed effects.
Religion:
Not many researchers have done research about the effect of religion/Christianity on the attitude people might have towards immigrants, while Christianity is known for charity. Dustmann and Preston (2001) looked at different types of religion and concluded that Catholicism is the least prejudiced religious group in their study. Christianity15 is assumed to have a positive effect on the attitude towards immigrants, as Christianity involves charity and caring for others.
Individual characteristics that influence the attitude towards social welfare:
Besides the possibility that there are characteristics that have an impact on the way people feel about immigrants, these (or other) characteristics might also influence the way people think about social welfare. In this section some of the most important studies will be discussed to state whether or not these variables do impact the attitude toward social welfare. In the beginning it is already stated
13 As is the case for education.
14 Being unemployed is likely to make a person less positive about immigrants.
13 that the feelings about immigrants and immigration might be an important contributor to this, but there may be more variables that impact the attitude towards social welfare.
Gillens (2004) did research about the different interest the rich and poor have towards social security but found no evidence that these two groups had different support levels for social security reforms. Other researchers (Burgoon, Koster, and van Egmond, 2010; Hjerm and Schnabel, 2012; Sumino, 2014) found evidence that people with a higher education, more income or who are employed are less supportive of redistribution. For this thesis, the effects of education and income on the attitude towards social welfare is assumed to be negative. People with a low educational level and poor people, are likely to depend more on social assistance16 and are therefore more positive about social welfare.
Hypothesis 4: People with a higher income or education level are likely to be negative about social welfare.
As is the case for the attitude towards immigrants, gender and having a positive image of the state of the economy have an impact on the attitude towards social welfare (Burgoon, Koster, and van Egmond, 2010). Eger (2010) found that being female or being part of a union lead to a more positive preference towards welfare spending. Other researchers (Hjerm and Schnabel, 2012; Burgoon, Koster, and van Egmond, 2010; Sumino 2014) also find evidence that women and older people have a more positive attitude towards redistribution compared to men and younger people. The conclusion that women are more positive about redistribution is also found by Garand, Xu and Davis (2015). This indicates that gender is an important variable in explaining the attitude people have towards welfare recipients. Gender and age are therefore expected to be positively related to a person’s attitude towards immigrants.
Hypothesis 5: Women and older people are likely to be more positive than man when it comes to the support for social welfare
Since this thesis also looks at the influence of several political aspects on the attitude towards social welfare, studies doing research in this field are also important. A group of studies (Hjerm and Schnabel, 2012; Sumino, 2014; Eger, 2010) find that people that consider themselves to be more left-wing, tend to have more support for redistribution and are more positive about social welfare. Left-wing parties are known for being supporters of social welfare and its voters are therefore likely to be positive about social welfare.
14
Hypothesis 6: Left-wing oriented people are likely to be more positive about social welfare, compared to right-wing oriented people.
Unlike many other researchers, Sumino (2014) also looked at the effect of religion on the support of welfare state policies. It is found that protestants are less likely to support egalitarian schemes. While Sumino (2014) did not find a positive effect, it will be assumed that Christianity has a positive effect on the support for social welfare, as Christianity believes in doing good for others and helping them17.
3. Data:
For this research, survey data from Laméris et al. (2019) will be used, which is constructed in March 2016. 3083 people were approached for the survey, of which 2460 households have completed the survey. This gives a response rate of 80%, which is considered to be sufficient and trustworthy. The data of this survey is constructed by the CentERpanel, which is an organization that sends surveys every week to about 1900 and is believed to be a good representation of the Dutch society. The survey used in this thesis contains several questions about the socioeconomic and political characteristics of the person. Within this survey, several statements were presented to the households. For each statement, they had to indicate to which degree they agreed with these statements. The scale of these statements is coded from 1 to 5 as such that a 1 means that the person totally disagrees, while a 5 means that the person totally agrees. For this research, the answers to some of these statements will be used to give a better insight in the attitude of people. This dataset allows to identify both the attitude people have towards immigrants, as well as their preference for the social welfare state.
Using the data of Laméris, Jong-A-Pin and Garretsen (2019), map 1 and map 2 are constructed for the Netherlands18. As can be seen in map 1, the attitude people have towards social welfare greatly differs among the provinces within the Netherlands. Some provinces have a positive attitude towards social welfare (on average), while others are more negative (less positive). This can be seen by the lighter color of these provinces in the map of the Netherlands. What are the determinants of this attitudinal difference? Could it be related to the difference in the attitude towards immigrants? These are some of the questions that will be addressed in this thesis. As can be seen in map 2, the attitude
17 For example, through social assistance
15 towards immigrants also greatly differs between provinces within the Netherlands. The center of the Netherlands is darker than the other areas, which indicates that the people in these provinces are (on average) more positive about immigrants compared to the other provinces. The determinants of this attitudinal difference will also be shortly discussed as this attitude could be endogenous and relevant in the explanation of the attitude towards social welfare. The attitude towards social welfare are modeled using factor analysis and multivariable linear regressions and are discussed in the analysis part of this paper.
Map 1. Attitude towards social welfare Map 2. Attitude towards immigrants
(using: Laméris, Jong-A-Pin, Garretsen (2019))19 (using: Laméris, Jong-A-Pin, Garretsen (2019))20
Descriptive statistics:
To get an insight in the relationship between the attitude towards immigrants and the attitude towards social welfare, and to get an understanding of the impact of different characteristics on the attitude people have, some descriptive statistics will be shown. This research focusses on the attitude towards social welfare and it is therefore insightful to get a view of the attitudinal distribution. As can
19 The black lines in the map indicate the border between the provinces in the Netherlands. For the survey each respondent had to fill in their province of residence, from which this map is made. The different colors indicate the mean of the answers of the people in a province about the statement about social welfare. The darker the color, the more positive the people in that province are about social welfare (on average). A light color indicates that the people in this province are on average not very supportive of social welfare.
20 The black lines in the map indicate the border between the provinces in the Netherlands. The different colors in this map indicate the average answer of the people in a province to those statements that indicate a
16 be seen in graph 1, most people are relatively more positive about social welfare as the distribution is right skewed.
Graph 1 Distribution attitude social welfare (using: Laméris, Jong A Pin, Garretsen, 2019)21
Besides the attitude towards social welfare, the attitude towards immigrants is another critical factor where this thesis focusses on. To get a good insight in the distribution of the attitudinal difference, the following boxplot is presented (graph 2). In this boxplot, it is clearly visible that the people are relatively positive, as was the case for the attitude towards social welfare.
Graph 2 Boxplot attitude towards immigrants (using: Laméris, Jong-A-Pin, Garretsen (2019))22
21 This graph shows the distribution of the attitude towards social welfare. On the y-axis the different levels of positivity towards social welfare are listed. The higher this value, the more positive a person is towards social welfare. The x-axis shows the frequency of the number of votes for each of the values about the attitude towards social welfare. The numbers besides the bars show the exact number of votes for each level of positivity towards social welfare.
17 In the graph below (graph 3), the distribution of the attitude towards immigrants is show per level of the attitude towards social welfare. As can be seen, the more positive a person is about social welfare, the more positive it scores on the attitude towards immigrants. There are more people who score above 20 for their attitude towards immigrants, the higher their attitude towards social welfare23.
Graph 3 Distribution attitude towards immigrants per attitude towards social welfare (using: Laméris, Jong-A-Pin, Garretsen (2019)24)
To get an insight in the difference in characteristics between immigrants and Dutch citizens within the sample, a comparison is reported in table 1 on the next page.
18
Table 1. Comparison natives and non-natives (using: Laméris, Jong-A-Pin, Garretsen (2019))
Variable Born in NL Born abroad
Left-right (mean) 5,297 4,865
Gender (mean) 0,483 0,554
Age(mean) 54,071 51,366
Gross individual income (mean) 2268,691 1954,232
Gross household income (mean) 3892,096 3702,08
Social status (mean)25 2,423 2,291
Primary school (percentage) 0,039 0,053
VMBO (percentage) 0,243 0,125
HAVO/VWO (percentage) 0,106 0,098
MBO (percentage) 0,225 0,259
HBO (percentage) 0,252 0,268
WO (percentage) 0,134 0,196
Occupation: not employed, with social welfare (amount) 4 1 Occupation: not employed, with social welfare (percentage) 0,2% 0,9%
Number of respondents 2198 112
Number of unemployed people26: 58 5
People between 15 -75: 1965 101
Unemployed (percentage)27: 3,0% 5,0%
Based on the table above (table 1), there are no great differences between the Dutch and non-Dutch people in the sample. The belief that immigrants are low-educated and/or poor is not confirmed based on the table. Another detailed division of the sample can be found in table B.2. in appendix B. This table shows some of the characteristics of people who are more left, center or right oriented. Based on the table, there are no large differences between the different political ideological groups. One thing to note from this table is that most highly educated people (HBO and WO) are more left-wing oriented.
The fact that the data provides information of about 2000 households, the sources of attitudinal differences can be modeled precisely. Before the data will be studied in more detail, the data is cleaned in order to be able to work with it.
25 The variable social status has the following values: 1 (high social-class), 2 (high/middle social-class), 3 (middle social-class), 4 (middle/mow social-class), 5 (low social-class). An average of 2,423 means the average social status is between high/middle and middle social-class.
26 An unemployed person belongs to the workforce (everyone between the ages 15 and 75) but is without a job. These people are looking/searching for a job (CBS)
19
Data – cleaning:
After collecting the data from the survey, the data needs to be cleaned and organized. Since outliers can lead to inconsistent conclusions and results, the outliers are removed from the data set. After omitting the outliers (income levels28 that are far beyond the confidence interval as can be seen in figure B.2 in appendix B), the sample is reduced. The consequence of omitting these observations for the results, is explained at the end of this thesis29.
Besides looking for outliers, observations with missing data also need to be inspected. Since many different variables are used for this research, it is important that the respondent has answered all the statements, questions and their background information. The respondents that completely filled in the survey will be considered for this research. The subjects that did not completely filled in the survey will be eliminated from this research (listwise deletion). After keeping only those respondents that completed the entire survey and filled in every question, the sample is reduced to 1974 people. Deleting observations with missing values can be argued that this would lead to a bias, since the people that skipped these questions might be more likely to leave some questions open than others. However, after omitting these observations, the sample is still relatively large.
In order to regress the effect of several individual characteristics on the attitude towards the social welfare and how they might be indirectly affecting this attitude through the attitude towards immigrants, the people that are born outside the Netherlands are removed from the sample since the questions concerning their agreement to several statements about immigrants were not asked to them. These questions are critical in determining the attitude one has about immigrants and cannot be forgotten. This research will therefore be solely based on people who are born in the Netherlands. After omitting the respondents that did not fill in these statements and the respondents that did not fill in their social class, 1856 respondents were left.
Creating relevant variables:
After cleaning the data-set, the variables needed for this research needed to be created from the answers/statements in the survey.
As mentioned in the literature part of this paper, there are several other variables that might also contribute to the attitude people might have towards immigrants or towards social welfare. Based on these articles, methods and data availability, the following variables needed to be constructed:
20 attitude towards immigrants, attitude towards social welfare, nationalistic view, egalitarian view, and a person’s perception of his/her own financial situation.
The dataset contains numerous statements and questions which might be related to the same factor. This factor is called an unobserved or latent variable and explains the correlation between variables. This variable is not observed, but rather inferred from other observed variables30. This latent variable drives the responses of people to these observed variables. Using latent variables instead of individual indicators reduces the number of variables under consideration. Having many individual indicators that are correlated to each other in a model, can lead to (multi)collinearity31. Having fewer dimensions makes it easier to interpret the relevant effect. This dimensionality reduction is being done using factor analysis. Factor analysis is a statistical method to describe the variability among the observed, correlated variables in dataset. Having less variables that are correlated makes the problem of (multi)collinearity less likely to occur.
Selecting those statements that would be optimal for that specific factor is both difficult and might lead to different conclusions. Some statements within the survey are very detailed, which makes this statement too narrow to state that a person who agrees with it as being positive about immigrants. Other statements could be interpreted differently, and therefore assigned to other characteristics. This means that each statement needs to be studied in-depth. In order to know which statement, belong to which factor, arguments based on theoretical knowledge are provided for each of these factors. Some statements for this research needed to be rewritten and recorded in order for them to indicate a positive attitude, the higher the person agrees with them. The manner in which all the statements are rewritten can be found in table B.1. in appendix B. To analyze whether or not these statements are indeed appropriate variables to describe a certain characteristic, a factor analysis is performed. The approach in this research is weighting each single relevant question and/or statement when combining them into a single factor. The factors will therefore not only be an average of the answers but uses appropriate weighs in order to distinct important statements from less important questions. In the next section, the statements belonging to each factor are described. To see if these statements indeed belong to this specific factor, a factor analysis is performed, which can be found in appendix A of this thesis.
Nationalistic view:
The first factor that will be discussed in this thesis is the nationalistic view people have. As stated by Mayda (2006), people with a more nationalistic view are more likely to be negative towards
30 These are the statements and questions from the survey.
21 immigrants. Using a factor analysis, the followings statements are confirmed to be good indicators to model the nationalistic view the respondents have32.
“Nationalistic sovereignty is more important than international cooperation”
“The Netherlands should exit the European Union (EU)”
Egalitarian view:
Besides the effect of being more nationalistic, the effect of a more egalitarian view is also important to model the way people feel about immigrants. As discussed in the literature review, people with a strong egalitarian view show a more positive attitude towards immigrants. Having an egalitarian outlook on the world, means that you believe that everyone is equal and that everyone deserves the same rights and opportunities. For this the following statement is used:
“Everyone in the Netherlands should be treated equally, no matter their religion, race or ethnicity33.”
Attitude towards immigrants:
One of the key factors in this thesis is the attitude towards immigrants. As described in the literature part of this paper, the attitude towards immigrants depend on several characteristics, but also has a substantial impact on the attitude towards social welfare. The attitude people have towards immigrants is based on their answers to a group of questions and statements in the survey. Some of these statements/questions, directly show how a person feels about immigrants, while others are more detailed/indirect. The answers of these statements will be used to give an indexation of how positive or negative a person feels towards immigrants. The following statements will be used to portray the “attitude towards immigrants”34:
“Immigrants are entitled to social welfare assistance” “The borders need to be closed for asylum seekers”
“I am concerned that my residential area is deteriorating due to the arrival of ethnic minorities” “One day, the Dutch are fired to accept ethnic minorities”
“The arrival of ethnic minorities in the Netherlands is a threat for the Dutch culture”
32 The reasoning behind these two statements can also be found in appendix A. 33 The reasoning behind this statement can be found in appendix A.
22
“Education for children of ethnic minorities is at the expense of education for Dutch children” “I am afraid that my financial situation will deteriorate due to the presence of ethnic minorities”
“Till which degree would you allow people from other ethnic groups to come live here in the Netherlands?”
Throughout this thesis, the assumption is made that respondents think about low-skilled/low-educated immigrants when answering the immigration questions. This assumption is based on the fact that the media and in public debates people often focus on poor immigrants if they are talking about immigrants35.
The reasoning behind the choice of these statements is clearly described in appendix A. Since not all statements are written in the same direction, some have to be rewritten as can be found in appendix A and in table B.1 in appendix B.
Attitude towards social welfare:
Besides peoples’ attitudes towards immigrants, the attitude people have towards the social welfare state is just as important for this research. This attitude is dependent on several characteristics and also on the attitude towards immigrants (as is being stated by several researchers in the literature part of this paper). To model the attitude towards social welfare, the following statement will be used.
“The government spends too much money on social welfare”
Perception of a person’s own financial situation:
One of the characteristics that might influence a person’s attitude toward immigrants is the feeling of the financial state in which they currently are. As mentioned before, people who see their current financial situation as something negative (or less than others), might be more negative about immigrants. To see whether a person is more positive or negative about their own financial situation, the variable that describes the way the respondents would describe their current income position as well as the way how they would position themselves in the income distribution, are used.
35 It is important to keep in mind that immigration and racial diversity are 2 different things. Not every immigrant will become member of a racial minority in the new country and not all racial minorities came from immigration. But if people answer questions about immigrants, they often relate this question to racial minorities. Although immigration and racial diversity are not the same, the members of the panel might have equated these two concepts when answering the questions. As mentioned by the migration observatory at the university of Oxford35, it is often unclear who is considered a migrant if people talk about immigrants in public
23 Characteristics:
Besides these latent variables mentioned above, there are other socioeconomic characteristics that might contribute to the attitude towards immigrants and the attitude towards social welfare. To model a person’s socioeconomic characteristics, the person’s gender, age, occupation (which is divided into categories), degree of urbanization of their residential area, household’s monthly net income, social class (dummies), political affiliation with right or left (as is done by Mayda, 2006), being a member of a political party, a person’s political preference, voting for a Christian party and their educational level (CBS – categories) are used. As stated by Scheve and Slaughter (2001), people form policy opinions according to their interests as labor-force participants. Empirical labor economists often measure skills using the educational attainment or occupation of the respondents. However, in this thesis (as is done by Scheve and Slaughter, 2001), both of these will be used. Adding a variable that indicates whether or not a person would vote for a Christian party is added due to the assumption that attitudes could be influenced by religion. This reasoning is based on the fact that Christianity put much focus on tolerance.
To model the influence of net income, the household net monthly income will be used. Economists see households as single decision-makers and therefore the household income rather than the individual income is used. The choices in a household are made as a group and need to be made on the total income of that household. This is called the unitary approach, which is also used by Donni and Ponthieux (2011). Although people make decisions as a group, their attitudes might differ on a personal level. As is often done by researchers, the logarithm of the income will be used to deal with the skewness. The degree of urbanization is used, since it has been shown that people in rural areas differ in their attitude from people living in other areas.
The dataset contains 13 employment categories that are clustered in several groups: employed, student, volunteering (since these people might be more concerned with caring for others and might be more positive about immigrants since they want them to succeed), looking for a job, unemployment benefit (unpaid work), retirement, incapacitated, and other36. For each of these occupations, a dummy is created to measure the individual effects.
Data inspection:
After generating the relevant variables based on the answers of the statements and questions from the survey, the summary statistics are reported in the table 2. In the data set used in this thesis,
24 most people are male, relatively rich and have an HBO diploma. The ages and occupations of the people are quite diverse, which gives a broad view of society. To get a better view of the composition of the sample, a diagram (figure B.1.) is added to appendix B, that shows the composition of the sample based on their occupation.
Table 2. Summary statistics
37 1 = only completed primary school; 2 = VMBO, 3 = HAVO/VWO; 4 = MBO; 5 = HBO; 6 = WO.
38 1 = work on payroll; 2 = works in family company; 3 = freelance, own boss; 4 = searching for a job after job loss; 5 = searching for a job for the first time; 6 = goes to school/college; 7 = does homecare; 8 = retired; 9 = (partially) incapacitated; 10 = provides unpaid services and is dependent on social assistance; 11 = volunteer; 12 = does something different.
39 1 = high social class; 2 = middle-high social class; 3 = middle social class; 4 = middle-low social class; 5 = low social class.
Variable Mean Standard deviation Min Max
Attitude towards immigrants 18,7579 5,5532 6,139 30,013
Attitude towards social welfare 3,3492 0,8583 1 5
Net income households 2831,386 1217,844 0 8250
Age 53,639 17,0206 16 93
Perspective of own financial situation
4,664 1,1066 1,430 7,151
Having a nationalistic view 3,310 1,1992 1,286 6,429
Having an egalitarian view 4,071 0,8969 1 5
Gender 0,4728 0,4994 0 1
Urbanization 2,978 1,3194 1 5
Education37 3,843 1,4838 1 6
Being left oriented 0,377 0,4848 0 1
Being right oriented 0,286 0,4522 0 1
Employed 0,507 0,5001 0 1
Occupation38 4,459 3,4860 1 12
Member of a political party 0,075 0,2641 0 1
Vote for Christian party 0,180 0,3841 0 1
25
4. Analysis:
Model:
An important variable used to model a person’s attitude towards immigrants, is a person’s opinion about how much people from other ethnicities are allowed to live and work in the Netherlands. As can be noted from table 3, most people in the Netherlands are hostile about people from other countries to come live in this country. In this section, the determinants of this hostility will be discussed using several models.
A multivariate analysis is needed to model the effect of all relevant characteristics that could explain the attitudinal differences. This involves the following variables: percentage immigrants in a province, percentage people depending on social welfare in a province, percentage people employed in a province, the logarithm of the population in a province, age, educational dummies40, left oriented, right oriented41, net household income, occupational dummies42, attitude towards immigrants, province-dummies43, perception of a person’s own financial situation, egalitarian view, degree of urbanization and dummies for the different classes44. Not all these variables are used in the same regression, as this would lead to several problems (for example multicollinearity).
If variables are correlated with each other, inserting both of them could lead to exaggerated and wrong results. Many variables are used for this research, which makes it most likely that there would be some evidence of (multi-) collinearity in the data. Looking at the correlation matrix in table B.3. in appendix B, it can be seen that some coefficients are slightly correlated with each other. However,
40 VMBO, HAVO/VWO, MBO, HBO and WO (elementary school is the reference category) 41 Center oriented is used as reference group since these people have less strong preferences.
42 Employed, student, retired, volunteer, dependent on social welfare, incapacitated, searching for a job, homecare (different occupation is used as reference group)
43 For which South Holland is used as reference group
44 High-class, high/middle-class, middle-class and low/middle-class (low class is used as the reference group) Table 3. Welcoming other ethnicities in the Netherlands (using: Laméris, Jong-A-Pin, Garretsen (2019))
Degree to which you would allow people from
other ethnic groups to come live here: Number of people who agree:
Allow almost nobody 201
Allow a minority 1105
Allow the majority 380
26 there were no significantly high correlation coefficients45 that could lead to a multicollinearity problem. Since the correlation coefficients are not significantly high, all these variables will still be used within this research.
Regression:
For the regression of the attitude towards social welfare, the attitude towards immigrants will first be treated as an exogenous variable. The Ordinary Least Squares model that belongs to this looks like the following:
(1) 𝑎𝑡𝑡𝑖𝑡𝑢𝑑𝑒 𝑡𝑜𝑤𝑎𝑟𝑑𝑠 𝑠𝑜𝑐𝑖𝑎𝑙 𝑤𝑒𝑙𝑓𝑎𝑟𝑒
= 𝛽1+ 𝛽2%𝑖𝑚𝑚𝑖𝑔𝑟𝑎𝑛𝑡𝑠 + 𝛽3%𝑝𝑒𝑜𝑝𝑙𝑒 𝑤𝑖𝑡ℎ 𝑠𝑜𝑐𝑖𝑎𝑙 𝑏𝑒𝑛𝑒𝑓𝑖𝑡𝑠 + 𝛽4%𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑑 + 𝛽5log(𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛) + 𝛽6log (𝑛𝑒𝑡 ℎ𝑜𝑢𝑠𝑒ℎ𝑜𝑙𝑑 𝑖𝑛𝑐𝑜𝑚𝑒)
+ 𝛽7𝑎𝑡𝑡𝑖𝑡𝑢𝑑𝑒 𝑎𝑏𝑜𝑢𝑡 𝑖𝑚𝑚𝑖𝑔𝑟𝑎𝑛𝑡𝑠 + 𝛽8𝑍𝑖+ 𝜀
Z is a matrix that captures the control variables as well as the different gender, educational and occupational dummies46. The first four variables in this model are numbers at province level47, while the other variables are individual characteristics. Since this model includes province characteristics, adding province dummies would lead to the problem of multicollinearity. For the attitude towards social welfare only redistribution statement 4 of the survey is used, this statement states that the government spends too much money on social welfare. This question directly asks about the attitude a person has towards social welfare. As was mentioned in the factor analysis part of this paper, the statements belonging to the latent variable “attitude towards social welfare” were similar to those belonging to the latent variable “egalitarian view”. Since these two variables are closely connected with each other, only one clear statement is used for “attitude towards social welfare”. By using only one statement, and other (multiple) statements for the “egalitarian view”, both of these variables can be used for this research. The reference groups for the educational dummies is elementary school, the reference group of political preference are the center oriented voters and the reference group for the different occupation one might have is the group that chose a different occupation.
The focus of this research will be on 𝛽7, which is expected to be positive. This beta shows how the attitude a person has towards immigrants influences their attitude towards social welfare, considering both province and individual characteristics. The more a person cares about immigrants,
45 Above 0.7
46 VMBO, HAVO/VWO, MBO, HBO, WO, female, left oriented, right oriented, employed, student, retired, volunteer, dependent on social welfare, incapacitated, searching for a job and homecare.
27 the more caring this person is and the more positive it will be about social welfare. If 𝛽7 equals 1, this would indicate that the way people feel about immigrant does not affect their feelings about the welfare state. If 𝛽7 is negative, this would indicate that the attitude towards immigrants is negatively correlated with the attitude towards social welfare. A person that is very positive about immigrants, would then mean that this person is negative about social welfare. This is something that is not expected for this research.
Besides 𝛽7, other coefficients are also noteworthy for this research. Since numerous researches concluded that poor people are more likely to be positive about social welfare, 𝛽6 is assumed to be negative. People with a higher income might be less supportive of an increase in social welfare spending. As was mentioned prior in this thesis, higher educated people are also more likely to have a negative attitude towards social welfare, which makes the expected sign of HBO and WO to be negative.
The results of this ordinary OLS can be seen in the first model in table B.4. in appendix B. One can see that the province characteristics are insignificant and that only age, the logarithm of net household income, being left oriented, being right oriented, and the attitude towards immigrants are significant individual characteristics48.
As is often the case in various models, there might still be the problem of heteroskedasticity in this research. Heteroskedasticity means that the errors do not have the same variance. To test for heteroskedasticity, the Breusch-Pagan test is used. The null hypothesis of this test states that the errors are homoscedastic49. For this test, a chi-squared statistic of 96.6450 and a p-value of 0 are found. These findings result in the rejection of the null hypothesis and provide evidence of heteroskedasticity in the model. White robust standard errors will be used in order to control for this heteroskedasticity.
Another issue that often arises, is the omitted variable bias. This problem arises if one or more relevant variables are not included in the model. Omitting these significant variables might lead to inconsistent estimates and unreliable results. To check for this problem, the Ramsey RESET test is performed, which states that the model has no omitted variables under the null hypothesis. Since the
48 The second model in table B.4. is the same model as model 1, but without the province level-based variables. The third model includes province dummies48 instead of the province level-based characteristics. Area-level variables (size immigrant inflow, income per capita) cannot be included in the individually based regressions together with the area fixed effects. Doing this would lead to the problem of perfect multicollinearity, which means that the two independent variables are intercorrelated. Province-dummies are added to measure unobserved province-specific effects. Would province-dummies not be added to the regression, the coefficients of the individual variables might become insignificant.
49 Have the same variance
28 p-value is found to be high (above 0.3), we fail to reject this hypothesis, which provides evidence that this model might not have the problem of omitted variables.
Is the attitude towards immigrants endogenous?
The focus of this research is to model the effect of the attitude toward immigrants on the attitude towards social welfare. As stated in the literature part of this thesis, the attitude towards immigrants might depend on several characteristics. For example, a person’s age or income might influence the way they feel about immigrants (Card, Dustmann and Preston, 2005). For this research, the following instruments will be used: having an egalitarian view, having a nationalistic view, thoughts about his/her own financial situation, the different occupation dummies, the social class dummies, log (net household income), gender, political orientation, and the person’s age. When the attitude towards immigrants depends on several characteristics and is endogenous, the model becomes a simultaneous equation model, which can be represented in the following way51:
𝑌 = 𝛽1+ 𝛽2𝑥 + 𝛽3𝑊 + 𝛽4𝑉 + 𝜖𝑖 where 𝑥 = 𝛾1+ 𝛾2𝑍 + 𝛾3𝑉 + 𝑣𝑖
For the research within this thesis, x indicates the attitude towards immigrants, W are the variables that only affect the attitude towards social welfare (based on previous research)52, V are the variables that affect both the attitude towards immigrants as well as the attitude towards social welfare53 and Z are the variables that only affect the attitude towards immigrants54.
To model this endogeneity, the 2SLS method is used. In the first stage of this method, the predicted value of the attitude towards immigrants is obtained. In this stage the endogenous variable (the attitude towards immigrants) is regressed on the instruments and the other exogeneous variables, which gives the results in table 4.
It can be seen that the following variables are significant in explaining the attitude people have towards immigrants: HBO, WO, female, left oriented, right oriented, egalitarian view, nationalistic view, being member of a political party and voting for a Christian party.
51 It implies that 𝑥
𝑖 and 𝜖𝑖 are correlated with each other
52 Immigrants in a province, percentage people depending on social welfare in a province, percentage people employed in a province, the logarithm of the population in a province.
53 different occupation dummies, the social class dummies, net household income, gender, political orientation and age
29
Table 4. First stage regression
Attitude towards immigrants
Age -0.0006 (-0.07) Employed -0.0290 (-0.02) High class 1.2554 (1.31) Vmbo 0.1631 (0.31) Student 0.0613 (0.04) High/middle class 0.8121 (0.88) Havo 1.1235 (1.96) Retired -0.3879 (-0.26) Middle class 0.9082 (0.99) Mbo 0.1728 (0.32) Volunteer 0.3639 (0.23) Low/middle class 0.3983 (0.44) Hbo 1.4158* (2.57) Egalitarian view 1.4739*** (12.23) Urbanization -0.0734 (-0.95) Wo 1.9297** (3.04) Nationalistic view -1.8531*** (-18.70) Perspective of own financial situation 0.0870 (0.80) Female 0.6013** (3.11) Member political party 0.8675* (2.48) Dependent on social benefits -1.3193 (-0.69) Left oriented 1.9329*** (8.40) Voting for a Christian party 0.7180** (2.80) Incapacitated -0.3770 (-0.24) Right oriented -1.4393*** (-5.86) Searching -0.2307 (-0.24)
Log (net household income) -0.2421 (-1.18) Homecare -0.7569 (-0.50) *P < 0,05 **P < 0,01 ***P < 0,001; The t-statistics are reported in the brackets;
Adjusted 𝑅2= 0.55; F-statistic = 6555.
Once the predicted value of the attitude towards immigrants is obtained, the attitude towards social welfare is regressed on this predicted value (together with the other exogenous variables) to get the estimators. The difficulty dealing with endogenous variables lies in finding the right instruments. For an instrument to be applicable to use, it has to be relevant and exogenous. A relevant instrument is an instrument that is correlated with the regressor that it is supposed to be instrumenting. Whether an instrument is indeed relevant can be easily concluded from the first stage, as the instruments should be significantly important in explaining the attitude towards immigrants. An exogenous instrument is a variable that only has an impact on the attitude towards social welfare through the attitude a person has towards immigrants. The results of the 2SLS method can be found in table 556.
55The 𝑅2 indicates that the regression does explain most of the variance in the attitude people have towards
immigrants. The value of the F-statistic is higher than threshold of 10, which means that the instruments are in fact relevant.
