Neighbourhood Ethnic Composition and Voting for the Radical Right in The Netherlands. The Role of Perceived Neighbourhood Threat and Interethnic Neighbourhood Contact

Abstract

Support for radical right parties has grown rapidly in many Western countries over the past few decades. In recent years, many studies have addressed the relationship between the presence of ethnic minorities in people’s living environment and their support for a radical right party, but consensus is hard to find as to how ethnic minority density is related to support for the radical right, let alone why. In this contribution, we demonstrate that in The Netherlands, ethnic minority density is positively related to the likelihood to vote for the Party for Freedom. This is particularly the case when the size of the minority group exceeds 15 per cent of the total neighbourhood population. We could establish this relationship by using the Dutch 1Vandaag Opinion Panel data set, a unique large-scale, individual-level data set comprising 21,200 native Dutch respondents living in 3,068 different neighbourhoods. We enriched this data set with contextual information derived from Statistics Netherlands. The reason why ethnic minority density is linked to support for the radical right is that these residents see non-Western migrants as a threat for their neighbourhood. This is particularly true for residents who do not mingle with their non-coethnic neighbours.

Introduction

Support for radical right parties has grown rapidly in many Western countries over the past few decades. Although defining the radical right is not easy, radical right parties share a core of ethno-nationalist xenophobia and anti-establishment populism (Rydgren, 2007). In the wake of their growing popularity, scholarly attempts to explain their success have likewise increased in number (Kitschelt, 2007; Rydgren, 2007). Many studies have addressed the relationship between ethnic minority density and the likelihood of voting for a radical right party (Lubbers and Scheepers, 2000; Coffé et al., 2007; Rink et al., 2009; Rydgren and Ruth, 2011), but there is still no consensus as to how the ethnic composition of people’s living environment is related to support for the radical right, let alone why.

In The Netherlands, the most important radical right party is the Partij voor de Vrijheid (Party for Freedom; PVV). The PVV was founded in 2006 by Geert Wilders, and in a recent expert survey, it has been indicated as a radical right party, similar to, e.g., Vlaams Belang in Belgium or Front National in France (Immerzeel et al., 2011). The PVV ‘[…] has been particularly vocal in […] resorting to racist and xenophobic discourse, targeting above all Muslim communities’ (European Commission against Racism and Intolerance [ECRI], 2008: p. 35). The PVV attracted 15 and 10 per cent of the votes in the past two general elections (2010 and 2012, respectively).

Previous studies investigating the impact of the ethnic composition of people’s living environment on support for the radical right can be grouped into macro- and micro-level studies. Macro-level studies relate the ethnic composition of, for instance, voting districts, municipalities, or countries to the mean support for the radical right within this context (Knigge, 1998; Golder, 2003; Coffé et al., 2007; Bowyer, 2008; Biggs and Knauss, 2012; Rydgren and Ruth 2013; Van der Waal et al., 2013; Valdez, 2014; Van Gent et al., 2014). A major advantage of macro-level studies is that they use actual election results. The downside of this approach is the risk of ecological fallacies (see also Rydgren, 2007). When using election results, it is difficult to determine the percentage of radical right voters among the native population in each macro-level unit. This is because both party preferences and turnout rates differ between ethnic groups to an unknown degree, and these differences may even be district-specific. Micro-level studies explain individual voting intentions or self-reported past voting behaviour and enrich survey data with characteristics of often relatively large geographic areas, such as countries, regions/districts, or municipalities (Lubbers and Scheepers, 2002; Lubbers et al., 2002; Norris, 2005; Arzheimer and Carter, 2006; Rink et al., 2009; Ford and Goodwin, 2010; Werts et al., 2012; Green et al., 2015; Stockemer, 2015). The impact of the ethnic composition of small-scale neighbourhoods has received only scarce attention in micro-level studies [but see for an exception, Dinas and Van Spanje (2011)].

Given the variety of research designs (i.e. macro-level versus micro-level), the geographical areas covered (e.g. voting districts, municipalities, or countries), and measures of radical right voting (i.e. intentions or behaviour), it might not come as a surprise that findings with respect to the impact of an environment’s ethnic composition are mixed. Some studies have found a positive relationship between ethnic minority density and radical right voting (Lubbers and Scheepers, 2002; Coffé et al., 2007; Valdez, 2014); others have found no significant relationship (Lubbers and Scheepers, 2000; Stockemer, 2015) or even a negative relationship (Bowyer, 2008; Rydgren and Ruth, 2013). Further, although some studies have found evidence for the idea that it is an increase in the percentage of ethnic minorities in particular that drives people to vote for radical right parties (Lubbers and Scheepers, 2000; Kessler and Freeman, 2005), others have not (Lubbers et al., 2000; Arzheimer and Carter, 2006). Finally, some scholars have pointed to a non-linear nature of the relationship between minority group size and radical right voting (Rink et al., 2009; Biggs and Knauss, 2012).

Success has also been limited when it comes to explaining the observed relationship between the presence of ethnic minorities and support for the radical right. The most important drawback of studies that rely solely on macro-level data—besides the absence of a valid dependent variable—is that they are not able to test the underlying micro-level, demand-side, explanations. Micro-level studies have focused on explanations derived from conflict theory (Coser, 1956; Blalock, 1967; Olzak, 1992). The idea is that a higher, or increasing, percentage of ethnic minorities in people’s living environment fuels perceptions of ethnic threat. Because grievances arising from immigration and feelings of ethnic threat are considered to be the central attitudinal drivers of the success of radical right parties (Rydgren, 2007; Ivarsflaten, 2008), this would explain a presumed positive relation between ethnic minority density and support for the radical right.

However, based on Blau’s (1977) opportunity structure theory, combined with contact theory (Allport, 1954/1979), an opposite expectation can be derived. According to this line of reasoning, an increase in contact opportunity leads to an increase in interethnic contact (Blau, 1977; Martinovic, 2013), which, in turn, fosters tolerance (Allport, 1954/1979; Pettigrew and Tropp, 2011). In theory, this would render people less likely to vote for a radical right party. This would explain a negative relation between ethnic minority density and support for the radical right, or at the least suppress a positive relationship between the two. Remarkably, the indirect effect of minority group size on radical right voting via interethnic contact has hardly been addressed empirically. Only very recently, Green et al. (2015) considered the influence of interethnic contact on voting for the radical right in Switzerland. Although living in Swiss regions with a larger minority group did not induce interethnic contact, they showed that interethnic contact is negatively related with ethnic threat perceptions and the likelihood to vote for the Swiss People’s Party.

Interethnic contact may not only affect radical right support by reducing perceptions of threat but also moderate the processes linking ethnic minority density with right-wing voting. For people who have contact with ethnic minority members, ethnic minority density may not as readily translate itself into feelings of ethnic threat (cf. McLaren, 2003) and consequently into a vote for the radical right. Although there is evidence for such a moderating influence for perceived ethnic threat, how contact moderates the relationship between ethnic minority density and support for the radical right has not been tested.¹

In this micro-level study, we will answer the questions as to how (increases in) minority group size within neighbourhoods is related to support for the PVV among native Dutch, and to what extent the observed relationship can be explained by perceptions of ethnic neighbourhood threat and interethnic neighbourhood contact. We will use recent and unique individual-level data from the Dutch 1Vandaag Opinion Panel (1VOP, 2015) enriched with neighbourhood characteristics provided by Statistics Netherlands. 1VOP is a very large sample; it consists of 21,200 native Dutch living in 3,068 different neighbourhoods. Whereas smaller representative samples face the problem that ethnically diverse localities are sparse, and samples focusing on individuals living in big cities contain relatively few neighbourhoods without a substantial minority group size, our sample includes an abundance of residents living in all types of neighbourhoods.

We will build on earlier research in several ways. First, we will consider the impact of the presence of ethnic minorities within neighbourhoods, in contrast to almost all previous micro-level studies that focused on larger-scale geographic areas, such as districts (Green et al., 2015) or countries (Werts et al., 2012). We assume that residents are more aware of the ethnic composition of their immediate living environments, meaning that local contexts are more important than larger contexts when it comes to the influence of ethnic minority density (see also Dinesen and Sønderskov, 2015). Moreover, focusing on neighbourhoods gives the advantage of more variation in ethnic minority density across macro-level units. This will make it possible for rigorously investigating a possible non-linear relationship between ethnic minority density and support for the radical right. Second, we will not only assess the impact of the existing size of the minority group within neighbourhoods but also simultaneously investigate whether recent increases in ethnic minority density are related to more support for the PVV. Finally, building on the recent work of Green et al. (2015), we will investigate the interplay between the contact and threat mechanism taking place within the neighbourhood when explaining the relationship between (increases in) minority group size and support for the radical right. Focusing on The Netherlands, we will show that for people with more interethnic contact, the presence of ethnic minorities is experienced less as a threat to the neighbourhood and thereby less likely to lead to a vote for the PVV.

Theories and Hypotheses

According to contact theory, positive interethnic contact reduces outgroup derogation (Allport, 1954/1979; Pettigrew and Tropp, 2011). Allport (1954/1979) stressed that this would only hold if contact takes place under ‘optimal’ conditions of which equal group status; common objectives; intergroup cooperation; and the support of authorities, law, or custom are the best known. The meta-study of Pettigrew and Tropp (2011) convincingly showed that these four conditions are not necessary for interethnic contact to reduce prejudice, but that contact under these conditions will reduce prejudice more strongly. The influence of interethnic contact can be explained in terms of increased levels of knowledge, empathy, and perspective taking and decreased levels of intergroup anxiety.

A higher percentage of ethnic minorities in a given environment increases the likelihood that its residents will have a positive interethnic contact, which, in turn, reduces negative attitudes towards ethnic minorities (Wagner et al., 2006; Schlueter and Scheepers, 2010). Scholars have repeatedly proposed that higher levels of interethnic contact would also reduce people’s likelihood to vote for a radical right party (Biggs and Knauss, 2012; Van der Waal et al., 2013), but only two studies have put this relationship to an empirical test. Rydgren (2008) solely addressed the direct influence of interethnic contact on radical right voting in six West European countries, finding only support for a negative influence of contact in two countries. Green et al. (2015) tested the mediating role of interethnic contact between ethnic minority density and radical right voting in Switzerland. Although their findings revealed a negative influence of interethnic contact on radical right voting, minority group size was unrelated with interethnic contact.

In this contribution, we address the role of neighbourhood contact with stigmatized ethnic minority groups in The Netherlands. Although Geert Wilders’ rhetoric includes the presumed negative consequences of immigration from (Eastern) European countries such as Poland, his main concern lies with Muslims (ECRI, 2008). In this study, we assess how contact with non-Western minorities—the vast majority of which is Muslim—in the neighbourhood affects support for the radical right. Should the contact mechanism be the sole, or most important, mechanism behind the impact of ethnic minority density on radical right support, we would expect that:

(H1a) The percentage of non-Western ethnic minorities in the neighbourhood is negatively related to support for the PVV, because (H1b) the percentage of non-Western ethnic minorities in the neighbourhood is positively related to interethnic neighbourhood contact, and (H1c) interethnic neighbourhood contact is negatively related to support for the PVV.

While the presence of minorities fosters interethnic contact, it may, at the same time, lead to feelings of ethnic threat. According to conflict theory (Coser, 1956; Blalock, 1967; Olzak, 1992), a higher percentage of ethnic minorities within a given environment increases competition between natives and ethnic minorities for economic resources and cultural issues. Ethnic competition, actual or perceived, fuels feelings of ethnic threat. Feelings of ethnic threat are, together with the anti-immigrant attitudes that result from experiencing threat, the main determinants of support for the radical right (Lucassen and Lubbers, 2012; Werts et al., 2012). That said, evidence for a (positive) relationship between ethnic minority density and ethnic threat is not convincing at the local level (Semyonov et al., 2004; Savelkoul et al., 2014; but see Green et al., 2015; Newman et al., 2015). Olzak (1992) was one of the first authors to stress that ethnic threat is more likely to be triggered by recent substantial increases in the size of the minority group than by stable levels of ethnic minority density, as the latter influence might flatten out as people become accustomed to the presence of ethnic minorities, or move out. We argue that the absence of more corroborative evidence might also be owing to the fact that in previous studies, the local level of ethnic density has been related to measures of threat at the national level, rather than to threats specific to the locality under consideration. Following the aforementioned line of reasoning, we propose that:

(H2a) The percentage of non-Western ethnic minorities in the neighbourhood—and increases thereof—are positively related to support for the PVV, because (H2b) the percentage of non-Western ethnic minorities in the neighbourhood—and increases thereof—are positively related to ethnic neighbourhood threat, and (H2c) ethnic neighbourhood threat is positively related to support for the PVV.

Interethnic contact and feelings of threat are negatively related, although the causal order can be debated (Savelkoul et al., 2011). Previous research has also shown that negative consequences of ethnic minority density on ethnic threat perceptions or trust are weaker for people with interethnic contact experiences (McLaren, 2003; Schneider, 2008; Stolle et al., 2008). A higher percentage of ethnic minorities might foster perceptions of ethnic threat, but if residents of ethnically dense environments have interethnic contact, the negative impact of a larger minority group size may be mitigated. This would then also lead to a weaker relationship between ethnic minority density and radical right voting. While we will take into account the negative relationship between ethnic threat and interethnic contact, our focus is on testing the following hypothesis:

(H3a) The positive relationship between the percentage of non-Western ethnic minorities—and increases thereof—and support for the PVV, is weaker for people who have interethnic neighbourhood contact, because (H3b) the positive relationship between the percentage of non-Western ethnic minorities—and increases thereof—and ethnic neighbourhood threat is weaker for people who have interethnic neighbourhood contact.

Using macro-level data, Biggs and Knauss (2012) found that the predicted probability for being a member of the British National Party is higher in neighbourhoods in which the relative minority group size reaches 25 per cent than in neighbourhoods in which there are no ethnic minorities. However, in neighbourhoods where ethnic minorities form three-quarters of the population, the probability falls strongly. Rink et al. (2009) reached similar conclusions, using individual-level data on voting behaviour during general elections in Belgium. After an initial steep increase in the likelihood of voting for the Vlaams Blok in Belgian municipalities with higher percentages of ethnic minority members, with the further growth of the relative minority group size (more than 4.8 per cent), the likelihood of radical right voting substantially decreases. In this study, we will rigorously explore possible non-linear relationships between (changes in) ethnic minority density and radical right voting.

Data

This study uses individual-level data from the 1VOP. Respondents sign up for this opinion panel voluntarily, after which they are invited to participate in small web surveys on a regular basis. Results of these opinion polls are regularly used in national news items. All 1VOP members were invited to participate in the web survey designed for the current study by email in February 2015, and 25,774 respondents—of whom 24,790 (96.2 per cent) were native Dutch—took up the invitation.² In our 1VOP sample, we found that men, older individuals, and those with higher education were somewhat overrepresented. Because of the uniquely large sample size, however, we could survey a wide range of people within the Dutch population: we conducted our analyses on specific subsamples (e.g. men/women, young/old, or lower/higher educated) to assess if, and to what extent, the non-representativity of our sample created a bias. This turned out not to be the case. This and other additional analyses are available upon request.

The ‘neighbourhood’ identifier included in the 1VOP is the four-digit part of the postal code. In The Netherlands, complete postal codes are combinations of four digits and two letters (e.g. 1011AB), resembling small parts of a specific street. The median surface area of our four-digit neighbourhoods is 5.3 km² and they are, on average, inhabited by 4,000 people. The respondents included in the final sample reside in 76 per cent of all the four-digit postal codes in The Netherlands (3,068/4,044). As Statistics Netherlands does not offer contextual information for these neighbourhoods directly, we constructed neighbourhood characteristics on the basis of grid data (0.01 km² grid cells; Statistics Netherlands, 2014a).

Support for the PVV

To examine populist radical right voting in The Netherlands, we measured respondents’ intended voting behaviour: ‘Which party would you vote for if parliamentary elections were held today?’ The answer categories to this question consisted of the 11 largest political parties represented in the Dutch Parliament and the option ‘Another party’. We removed from the analysis the 3,396 respondents (13.7 per cent) who answered ‘I don’t know’ (11.5 per cent), ‘blank vote’ (0.6 per cent)’, ‘I’m not allowed to vote’ (0.004 per cent), ‘I’m allowed to vote, but I wouldn’t’ (1.1 per cent), and ‘No answer’ (0.5 per cent). We constructed a dichotomous variable measuring the intention to vote for the PVV.

In total, 18.5 per cent of our respondents intend to vote for the PVV if parliamentary elections would be held. Compared with the last parliamentary election held in 2012 in which 10.1 per cent voted for the PVV, this percentage (18.5) is considerably high, but this will be, at least in part, the result of our selection of native Dutch respondents, the decision to exclude the ‘I don’t know’ category from our analyses, and the increased popularity of the PVV. Additional analyses in which we analysed (self-reported) voting behaviour during the parliamentary elections of 2012 led to substantially similar findings.

The Percentage of Non-Western Minorities

Following the definition of Statistics Netherlands, people are considered to be non-Western immigrants when at least one of their parents was born in a non-Western country. About 12 per cent of the Dutch population are considered to be of non-Western descent in 2014. The largest non-Western minority groups in The Netherlands are Moroccan (2.2 per cent), Turkish (2.4 per cent), and Surinamese and Antillean (2.9 per cent; Statistics Netherlands, 2014b). The percentage of non-Western minorities for the neighbourhoods in which our respondents live ranges from 0 to 67.6 per cent; the mean value is 9.8 per cent.

To create a thorough test of the extent to which minority group sizes, and changes to them, affect support for the PVV, we included both the average percentage between 2014 and 2011 and the difference in the percentage of non-Western minorities between these two years in our explanatory model. To scrutinize a possible curvilinear effect of the presence of non-Western minorities, we used categorized versions of these group size variables. We used the following categories for the percentage of non-Western minorities: [0,5]; ⟨5,10]; ⟨10,12.5]; ⟨12.5,15]; ⟨15,20]; ⟨20,30]; ⟨30, HI].³ In addition, we used the following categories for the difference measure: [LO,0]; ⟨0,0.5]; ⟨0.5,1]; ⟨1,2]; ⟨2,HI]. Although these changes might not seem substantial, the period in which either the influx or outflow of minorities took place spanned only three years.

Interethnic Neighbourhood Contact and Ethnic Neighbourhood Threat

To reduce variations in individual perceptions of neighbourhood contact and neighbourhood threat, we provided our respondents with the following definition: ‘The neighbourhood is understood to mean the living environment that is reachable on foot in fifteen minutes from your own house.’ This corresponds roughly to a neighbourhood with a surface area of 4.5 km² (i.e. an area with a radius of 1.2 km).

We measured interethnic neighbourhood contact with the following question: ‘How often do you have personal contact in your neighbourhood with people of non-Western descent? By personal contact, we mean that you know the name of this person and occasionally have a conversation with this person.’ Our contact measure thus referred to a positive contact experience. The answer categories for this item included: ‘Never’ (0); ‘About once a year’ (1); ‘Several times a year’ (2); ‘About once a month’ (3); ‘Several times a month’ (4); ‘Once or several times a week’ (5); ‘(Almost) every day’ (6); and ‘Not applicable (in my neighbourhood, there are no people of non-Western descent)’ (7), which we labelled as ‘No opportunity’. Ethnic neighbourhood threat was measured with the item: ‘I sometimes worry that my neighbourhood is deteriorating because of the arrival of ethnic minorities.’ The answer categories were: ‘Totally disagree’ (0); ‘Somewhat disagree’ (1); ‘Neither disagree nor agree’ (2); ‘Somewhat agree’ (3); ‘Totally agree’ (4); and ‘I don’t know/no opinion’ (5). In our main analyses, which explain voting for the PVV, the categories of these variables were entered as separate dummies, to be able to detect possible non-linear relationships.

Control Variables

We included respondents’ gender, age, and level of education. Moreover, we controlled for household composition, differentiating the following five categories: ‘Single, no kids’; ‘Single, kids’; ‘Married, no kids’; ‘Married, kids’; and ‘Student or living communally/Living with parents/Other’. We included main daily activity in the following categories: ‘Employee/Self-employed’; ‘Looking for work’; ‘Unable to work’; ‘Student’; ‘Housewife/house-husband’; ‘Pensioner’; and ‘Other’. For all categorical control variables, we used deviation coding, meaning that all coefficients show deviations from the overall mean. At the neighbourhood level, we controlled for affluence and neighbourhood decline, using information on the average house price in each neighbourhood (in €1,000), available for 2011 and 2012. We included the natural logarithm of the average and that of the absolute difference between both years.

Working Sample and Methods

We removed 194 respondents (less than 1 per cent) from our sample for whom we could not match contextual information. This left us with 21,200 respondents in 3,068 neighbourhoods. The descriptive statistics for our main variables can be found in Table 1.⁴ Because our respondents are nested in neighbourhoods, we used multilevel logistic regression analyses.

Table 1

Descriptive statistics (N_individual = 21,200; N_{neighbourhood} = 3,068)

	Mean/Prop	SD	Minimum	Maximum
% Non-Western minorities ([0–5])	0.42		0	1
% Non-Western minorities (⟨5,10])	0.27		0	1
% Non-Western minorities (⟨10,12.5])	0.06		0	1
% Non-Western minorities (⟨12.5,15])	0.06		0	1
% Non-Western minorities (⟨15,20])	0.08		0	1
% Non-Western minorities (⟨20,30])	0.06		0	1
% Non-Western minorities (⟨30, HI])	0.05		0	1
Δ % Non-Western minorities ([LO,0])	0.26		0	1
Δ % Non-Western minorities (⟨0,0.5])	0.36		0	1
Δ % Non-Western minorities (⟨0.5,1])	0.22		0	1
Δ % Non-Western minorities (⟨1,2])	0.12		0	1
Δ % Non-Western minorities (⟨2,HI])	0.04		0	1
Economic affluence (housing value × 1,000)	232.50	71.94	94.95	1,167
Δ Economic affluence (housing value × 1,000)	−5.69	5.44	−74.80	69.75
Contact non-Western minorities (‘Never’)	0.24		0	1
Contact non-Western minorities (‘Once a year’)	0.07		0	1
Contact non-Western minorities (‘Several times a year’)	0.16		0	1
Contact non-Western minorities (‘Once a month’)	0.09		0	1
Contact non-Western minorities (‘Several times a month’)	0.15		0	1
Contact non-Western minorities (‘Several times a week’)	0.15		0	1
Contact non-Western minorities (‘Almost every day’)	0.09		0	1
Contact non-Western minorities (‘No opportunity’)	0.06		0	1
Ethnic threat (‘Totally disagree’)	0.20		0	1
Ethnic threat (‘Somewhat disagree’)	0.25		0	1
Ethnic threat (‘Neither disagree nor agree’)	0.20		0	1
Ethnic threat (‘Somewhat agree’)	0.14		0	1
Ethnic threat (‘Totally agree’)	0.18		0	1
Ethnic threat (‘I don’t know/no opinion’)	0.03		0	1

Notes: Sample comprises native Dutch only. ‘% non-Western minorities’ refers to the mean score between 2011 and 2014. ‘Δ % non-Western minorities’ refers to the difference between 2011 and 2014. ‘Economic affluence’ refers to the mean score between 2011 and 2012. ‘Δ Economic affluence’ refers to the difference between 2011 and 2012.

Sources: 1VOP (2015); Statistics Netherlands (2011, 2014).

Results

Our individual-level control variables are similarly related with radical right voting as found in previous research (Lubbers et al., 2002; Rink et al., 2009; Lucassen and Lubbers, 2012); for example, women, older individuals, and those with higher education are less likely to vote for the PVV (Table 2). The economic status of the neighbourhood is not significantly related to support for the PVV.⁵

Table 2

Cross-sectional models explaining voting for the PVV

	Model 1		Model 2		Model 3
(Intercept)	−1.62	(0.06)***	−1.20	(0.07)***	−3.68	(0.14)***
Contextual-level variables
% Non-Western minorities (REF: [0–5])
% Non-Western minorities (⟨5,10])	0.05	(0.05)	0.06	(0.05)	−0.12	(0.06)*
% Non-Western minorities (⟨10,12.5])	0.07	(0.08)	0.09	(0.08)	−0.18	(0.10)
% Non-Western minorities (⟨12.5,15])	0.12	(0.09)	0.12	(0.09)	−0.22	(0.10)*
% Non-Western minorities (⟨15,20])	0.21	(0.08)**	0.22	(0.08)**	−0.26	(0.09)**
% Non-Western minorities (⟨20,30])	0.29	(0.09)***	0.31	(0.09)***	−0.24	(0.10)*
% Non-Western minorities (⟨30, HI])	0.33	(0.10)***	0.32	(0.10)**	−0.29	(0.11)*
Δ % Non-Western minorities (REF: ⟨0,0.5])
Δ % Non-Western minorities (⟨0.5,1])	−0.05	(0.05)	−0.04	(0.05)	−0.07	(0.06)
Δ % Non-Western minorities (⟨1,2])	0.11	(0.06)	0.10	(0.07)	0.10	(0.08)
Δ % Non-Western minorities (⟨2,HI])	0.04	(0.10)	0.02	(0.11)	0.00	(0.12)
Δ % Non-Western minorities ([LO,0])	−0.14	(0.05)**	−0.14	(0.05)**	−0.07	(0.06)
Log(Economic affluence)	−0.05	(0.08)	−0.06	(0.08)	−0.09	(0.09)
Log(Δ Economic affluence)	0.01	(0.02)	0.00	(0.02)	0.01	(0.02)
Individual-level variables
Contact non-Western minorities (REF: ‘Never’)
Contact non-Western minorities (‘Once a year’)			−0.50	(0.08)***	−0.30	(0.09)**
Contact non-Western minorities (‘Several times a year’)			−0.66	(0.06)***	−0.45	(0.07)***
Contact non-Western minorities (‘Once a month’)			−0.73	(0.08)***	−0.61	(0.09)***
Contact non-Western minorities (‘Several times a month’)			−0.59	(0.06)***	−0.35	(0.07)***
Contact non-Western minorities (‘Several times a week’)			−0.52	(0.06)***	−0.36	(0.07)***
Contact non-Western minorities (‘Almost every day’)			−0.24	(0.07)***	−0.16	(0.08)*
Contact non-Western minorities (‘No opportunity’)			−0.92	(0.10)***	−0.29	(0.11)**
Ethnic threat (REF: ‘Totally disagree’)
Ethnic threat (‘Somewhat disagree’)					0.91	(0.14)***
Ethnic threat (‘Neither disagree nor agree’)					2.00	(0.14)***
Ethnic threat (‘Somewhat agree’)					3.05	(0.13)***
Ethnic threat (‘Totally agree’)					4.40	(0.13)***
Ethnic threat (‘I don’t know/no opinion’)					2.49	(0.16)***
Gender (female)	−0.32	(0.02)***	−0.31	(0.02)***	−0.13	(0.03)***
Gender (male)	0.32	(0.02)***	0.31	(0.02)***	0.13	(0.03)***
Education (in years/10)	−1.64	(0.06)***	−1.60	(0.06)***	−0.94	(0.07)***
Age (in years/10)	−0.09	(0.02)***	−0.09	(0.02)***	−0.08	(0.03)***
Household (single, no kids)	−0.24	(0.04)***	−0.25	(0.04)***	−0.22	(0.05)***
Household (single, kids)	0.28	(0.08)***	0.29	(0.08)***	0.36	(0.10)***
Household (married, no kids)	−0.18	(0.04)***	−0.18	(0.04)***	−0.17	(0.05)***
Household (married, kids)	0.08	(0.05)	0.09	(0.05)	0.05	(0.06)
Household (other)	0.05	(0.09)	0.05	(0.09)	−0.02	(0.11)
Daily activity (working)	−0.06	(0.05)	−0.06	(0.05)	0.03	(0.06)
Daily activity (looking for work)	0.12	(0.09)	0.12	(0.09)	0.22	(0.10)*
Daily activity (unable to work)	−0.03	(0.07)	−0.04	(0.07)	−0.03	(0.08)
Daily activity (student)	−0.55	(0.21)**	−0.56	(0.21)**	−0.64	(0.24)**
Daily activity (housewife/husband)	0.35	(0.10)***	0.36	(0.10)***	0.33	(0.11)**
Daily activity (retired)	−0.09	(0.06)	−0.09	(0.06)	−0.05	(0.07)
Daily activity (other)	0.27	(0.09)**	0.28	(0.09)**	0.14	(0.10)
Variance component
Neighbourhoods	0.05		0.05		0.05
Model fit
Pseudo R2	0.11		0.13		0.45
AIC	19,146		18,912		14,270
BIC	19,361		19,183		14,580

Notes: Regression coefficients with standard errors in parentheses.

*

P < 0.05; **P < 0.01; and ***P < 0.001 (two-tailed test). N_individual = 21,200; N_{neighbourhood} = 3,068. The sample comprises native Dutch only. The control variables ‘gender’, ‘household composition’, and ‘daily activity’ are included as deviation contrasts. The control variables ‘age’, ‘education’, ‘economic affluence’, and ‘Δ economic affluence’ are grand mean centred. The variance_{neighbourhoods} of the empty model is 0.10. The individual-level variance is fixed (π²/3) because we use the logit link function.

Sources: 1VOP (2015); Statistics Netherlands (2011, 2014).

People living in neighbourhoods with a higher percentage of non-Western minorities are more likely to vote for the PVV (Table 2, Model 1); but only in neighbourhoods where the relative minority group size exceeds 15 per cent do parameter estimates reach significance. In neighbourhoods with between 15 and 20 per cent non-Western ethnic minorities, the odds of voting for the PVV are 1.23 higher than in neighbourhoods with 5 per cent or fewer non-Western minorities (b = 0.21, se = 0.08; exp(b) = 1.23). The likelihood of voting for the PVV seems to flatten when neighbourhoods become ethnically more dense; in neighbourhoods in which non-Western minorities comprise more than 30 per cent of the total population, the odds of voting for the PVV are 1.39 (b = 0.33, se = 0.10; exp(b) = 1.39) higher than in neighbourhoods with 5 per cent or fewer non-Western minorities. Respondents living in neighbourhoods facing either a stable or a decreasing minority group size are somewhat less likely to vote for the PVV (b = −0.14, se = 0.05; exp(b) = 0.87) compared with those living in neighbourhoods with a small increase (<0.5). So far, our results seem to be in line with conflict theory (H2), while being at odds with expectations derived from contact theory (H1).

In Model 2 of Table 2, we introduce our interethnic neighbourhood contact variable. For residents with opportunities for interethnic contact, contact is negatively related with the probability of voting for the PVV, as we expected based on the contact theory (H1c). When people meet non-Western minorities once a month, the likelihood of voting for the PVV is lowest (b = −0.73, se = 0.08; exp(b) = 0.48).⁶ When contact is more frequent, the chance that residents will vote for the PVV increases again. This might indicate that people with more positive interethnic neighbourhood contact might likewise have more negative interethnic contact experiences. Respondents who claim to have no opportunity for interethnic contact are least likely to vote for the PVV (b = −0.92, se = 0.10; exp(b) = 0.40). Including the contact variable does not substantially alter the estimates of (increases in) minority group size and radical right voting, and interethnic neighbourhood contact does not suppress an even stronger positive relationship between minority group size and support for the PVV. Thus, even though residents of neighbourhoods with a higher percentage of non-Western minorities have more interethnic neighbourhood contact (Appendix 1, column 1), we refute H1.

Natives who experience higher levels of ethnic neighbourhood threat are more likely to vote for the PVV (Table 2, Model 3): compared with residents who do not experience any ethnic threat (i.e. ‘Totally disagree’), the odds of residents who do experience ethnic threat are 81 times higher (b = 4.40, se = 0.13; exp(b) = 81). Perceptions of ethnic neighbourhood threat explain the positive influence of the percentage of non-Western minorities and changes to it on radical right voting voting: in Model 3, the impact of minority group size on voting for the PVV is no longer positive; the impact is even negative and significant. Increasing minority group size is no longer significantly related with voting for the PVV once we take into account ethnic neighbourhood threat. Combined with the finding that living in neighbourhoods with a larger non-Western minority population is related to greater ethnic neighbourhood threat and the level of threat is lower in neighbourhoods with a decreasing or stable minority population, as compared with neighbourhoods with a small increase (<0.5; Appendix 1, column 2), we thereby find clear support for H2 derived from the conflict theory.

In line with previous studies (Schlueter and Scheepers, 2010; Savelkoul et al., 2011), we observe a negative relation between contact and threat (Appendix 1). However, we expected that interethnic contact would also mitigate the positive influence of ethnic minority density (and changes thereof) on radical right voting, because the (increasing) minority group size would be weaker related to feelings of ethnic threat for residents who have interethnic contact on a regular basis than for residents with no or irregular interethnic contact. The positive relationship between minority group size and interethnic neighbourhood threat is significantly less pronounced for people who regularly mingle with non-Western ethnic minorities as expected (Appendix 2). Further support for H3 can be found in Figure 1 (see also Appendix 3). In the left panel of Figure 1, we show that for residents who have no interethnic neighbourhood contact (or only on an irregular basis), ethnic minority density is more strongly related to support for the PVV than for residents with regular interethnic contact; the slope of the redline (with dots) is steeper than that of the blue line (with triangles). Once we control for levels of neighbourhood threat (right panel, Figure 1), the impact of ethnic minority density is very similar across residents with and without contact. Given the dichotomous nature of our dependent variable and the resulting interpretation problems with interactions (Ai and Norton, 2003), we refrain from formally testing the significance of the interaction effect of ethnic minority density with contact. Although we should thus be cautious in drawing bold conclusions, our results are in line with both H3a and H3b.

Figure 1

Probability to vote for the PVV for residents with different levels of interethnic contact. Notes: Predicted probabilities (and the uncertainties therein; 90% CI) are based on estimates of fixed effects only. X-axis labels refer to the mean value within each category of the ‘% non-Western ethnic minorities’ variable. ‘No or irregular contact’ is defined as at most ‘several times a year’ (respondents with no opportunity for contact are not included). ‘Regular contact’ is defined as at least ‘about once a month’. We controlled for same set of variables as reported in Table 2. The sample comprises native Dutch only.

In contrast to our expectation, interethnic neighbourhood contact moderates the impact of changes in minority group size on neither threat nor radical right voting (Appendix 2 and 3). Possibly because effects of changes in ethnic minority density were not very substantial to begin with.

Discussion

Anti-immigrant sentiments are the most important determinants of support for the radical right. These feelings are triggered by continuous migration flows and by the subsequent (perceived) threat of migrants to ‘our way of life’. Several authors have shown a positive relation between (increases in) ethnic minority density at the country or regional level and support for the radical right (Knigge, 1998; Lubbers and Scheepers, 2000, 2002; Lubbers et al., 2002; Kessler and Freeman, 2005). More recently, attention has shifted to the impact of ethnic minority group sizes at the very local level (e.g. neighbourhoods or voting districts) on support for the radical right. However, so far, empirical evidence has been based largely on macro-level studies (Biggs and Knauss, 2012; Rydgren and Ruth, 2013; Valdez, 2014; but see also Dinas and Van Spanje, 2011), probably owing to the lack of large-scale, individual-level data sets. This has consequences not only for the interpretation of the findings (as the votes of ethnic minorities are also included in election results) but also for the possibility to unravel underlying explanations.

In this contribution, we have shown that in The Netherlands, both the existing proportion of non-Western ethnic minorities in neighbourhoods and, to a lesser extent, recent increases thereof are positively related to the likelihood of voting for the PVV. The relationship between ethnic minority density and the odds to vote for the PVV proved to be non-linear; the odds increase rapidly once the group size exceeds 15 per cent and seems to flatten in ethnically highly dense neighbourhoods (with more than 30 per cent minorities). We could establish this non-linear relationship because we relied on a unique large-scale, individual-level data set enriched with relevant information about neighbourhoods that varied widely in ethnic composition. Not only could we establish a positive link between ethnic minority density and support for the radical right, we showed that this relation can be explained by the fact that in particular, residents of ethnically more dense neighbourhoods are of the opinion that the presence of ethnic minorities leads to neighbourhood deterioration.

In neighbourhoods with a larger proportion of non-Western minorities, native Dutch have more positive contact experiences with this ethnic group. We expected that interethnic contact would make residents less likely to vote for the PVV, but we found only weak support for a direct impact of interethnic contact (cf. Rydgren, 2008). Although interethnic contact is negatively and significantly related with the odds of voting for the PVV, its impact is not substantial. This is not to say that interethnic neighbourhood contact is irrelevant. Unlike commonly used indicators of interethnic contact, our contact measure enabled us to differentiate between respondents lacking the possibility to interact with non-Western minorities in their neighbourhood and those without interethnic neighbourhood contact, while having the opportunity to interact. Predominantly, the latter group perceives ethnic minorities as a threat to their neighbourhood and is more likely to vote for the PVV. Our findings also illustrate that interethnic contact moderates the process by which ethnic minority density leads to support for the radical right: only for residents who seldom interact with their non-Western neighbours, the size of the minority group is related to more support for the PVV, because for these residents in particular, a larger minority group in the neighbourhood induces feelings of ethnic neighbourhood threat.

Together with Green et al. (2015), we were the first to simultaneously test the mediating role of ethnic threat perceptions and interethnic contact, explaining influences of ethnic minority density on radical right voting. Surprisingly, once we take into account the role of neighbourhood contact and threat, residents of neighbourhoods with more than 5 per cent ethnic minorities are less likely to support the PVV than those of neighbourhoods with up to 5 per cent minorities. This is a puzzling finding that warrants further research.

Given that we had to rely on cross-sectional data, we had to be cautious in making very strong causal interpretations. Selective residential mobility may be part of the reason why we observe a (positive) relationship between ethnic minority density and interethnic neighbourhood contact. At the same time, it is likely to suppress the positive relationship between ethnic minority density and perceived ethnic threat. Future research could further test whether the influences of (increasing) ethnic minority density and both mediators can be generalized to other Western countries and preferably test these influences using longitudinal data.

Neither Rydgren (2008) nor we found very convincing evidence for a direct influence of interethnic contact on radical right voting. However, we should not dismiss the contact mechanism too hastily. Our measure of interethnic contact referred to positive contact with stigmatized minority groups taking place in the neighbourhood. We neither captured interethnic friendships within the neighbourhood nor assessed the impact of positive and negative interethnic contact experiences that take place outside the direct residential neighbourhood. However, contact taking place outside the neighbourhood will not explain the impact of ethnic minority density of the neighbourhood on support for the radical right, which was the focus of this contribution. Future research could further investigate how interethnic contact in different domains and with different minority groups (e.g. with immigrants from Eastern Europe) affect support for the radical right and how these types of contacts moderate the relationship between ethnic minority density and support for radical right parties.

Summarizing, we found that residents of ethnically dense neighbourhoods are more likely to vote for the PVV. Using a unique large-scale, individual-level data set enriched with information about people’s direct living environments, we could detect that only when the proportion of non-Western minorities reaches a substantial size (in our study, a tipping point of 15 per cent) does this relationship become apparent. The reason why residents of ethnically dense neighbourhoods are more attracted to the rhetoric of the radical right is because they see non-Western migrants as a threat to their neighbourhood. This is particularly true for residents who do not mingle with their non-coethnic neighbours.

Michael Savelkoul is Assistant Professor at the Department of Sociology, Radboud University Nijmegen, The Netherlands. His main research interests include ethnic diversity, social cohesion, and radical right voting. His work has appeared, amongst others, in European Sociological Review, Journal of Ethnic and Migration Studies, Non-profit and Voluntary Sector Quarterly, and Social Science Research.

Joran Laméris is a PhD-student at the Department of Sociology, Radboud University Nijmegen, The Netherlands. Her main research interests are neighbourhood effects, ethnic diversity, and social cohesion.

Jochem Tolsma is Associate Professor at the Department of Sociology, Radboud University Nijmegen, The Netherlands. His research interests include neighbourhood effects and social cohesion. He has published in, among other journals, Acta Politica, Annual Review of Sociology, British Journal of Sociology, and European Sociological Review. In 2012, he received a VENI fund from The Netherlands Organisation for Scientific Research (NWO) to investigate the impact of the ethnic composition of communities for social cohesion.

Acknowledgements

The authors would like to thank anonymous reviewers of the European Sociological Review for their constructive comments and suggestions.

Funding

This work was supported by the Netherlands Organisation for Scientific Research (NWO; grant numbers 406-12-068 to J. Laméris and 016.135.161 to J. Tolsma).

1

However, see Biggs and Knauss’ (2012) macro-level study in which segregation is used as a proxy for interethnic contact.

2

We only selected respondents whose parents were Dutch or those who identified with The Netherlands in case one parent was non-Dutch.

3

In preliminary analyses, we also used quartiles and quintiles, which led to substantially similar results, but the non-linear relationship between minority group size and support for the radical right is best demonstrated with the presented categorization.

4

The descriptives of our individual-level control variables will not be publicly disclosed at the request of the owners of the 1VOP panel but have been sent to the editor and reviewers before publication.

5

The economic status of the neighbourhood is negatively and significantly related to support for the radical right as long as we do not control for neighbourhood composition effects.

6

The category ‘once a month’ deviates significantly from ‘never’, ‘once a year’, ‘several times a week’, and ‘almost every day’.

Appendix

Table A1

Cross-sectional models explaining interethnic contact and ethnic threat

	Interethnic contact	Ethnic threat
(Intercept)	2.27 (0.05)***	1.88 (0.03)***
Contextual-level variables
% Non-Western minorities (REF: [0–5])
% Non-Western minorities (⟨5,10])	0.60 (0.04)***	0.18 (0.02)***
% Non-Western minorities (⟨10,12.5])	0.77 (0.07)***	0.28 (0.04)***
% Non-Western minorities (⟨12.5,15])	0.96 (0.07)***	0.33 (0.04)***
% Non-Western minorities (⟨15,20])	0.99 (0.06)***	0.46 (0.04)***
% Non-Western minorities (⟨20,30])	1.43 (0.07)***	0.54 (0.05)***
% Non-Western minorities (⟨30, HI])	1.78 (0.08)***	0.62 (0.05)***
Δ % Non-Western minorities (REF: ⟨0,0.5])
Δ % Non-Western minorities (⟨0.5,1])	0.05 (0.04)	0.00 (0.03)
Δ % Non-Western minorities (⟨1,2])	−0.08 (0.05)	0.03 (0.03)
Δ % Non-Western minorities (⟨2,HI])	0.15 (0.09)	0.04 (0.05)
Δ % Non-Western minorities ([LO,0])	−0.04 (0.04)	−0.08 (0.03)***
Log(Economic affluence)	0.02 (0.06)	−0.02 (0.04)
Log(Δ Economic affluence)	0.03 (0.01)*	0.00 (0.01)
Individual-level variables
Ethnic threat (REF: ‘Totally disagree’)
Ethnic threat (‘Somewhat disagree’)	−0.15 (0.04)***
Ethnic threat (‘Neither disagree nor agree’)	−0.22 (0.04)***
Ethnic threat (‘Somewhat agree’)	−0.08 (0.05)
Ethnic threat (‘Totally agree’)	−0.29 (0.05)***
Ethnic threat (‘I don’t know/no opinion’)	−1.09 (0.08)***
Contact non-Western minorities (REF: ‘Never’)
Contact non-Western minorities (‘Once a year’)		−0.21 (0.04)***
Contact non-Western minorities (‘Several times a year’)		−0.26 (0.03)***
Contact non-Western minorities (‘Once a month’)		−0.22 (0.04)***
Contact non-Western minorities (‘Several times a month’)		−0.29 (0.03)***
Contact non-Western minorities (‘Several times a week’)		−0.23 (0.03)***
Contact non-Western minorities (‘Almost every day’)		−0.22 (0.04)***
Contact non-Western minorities (‘No opportunity’)		−0.55 (0.04)***
Gender (female)	−0.08 (0.02)***	−0.20 (0.01)***
Gender (male)	0.08 (0.02)***	0.20 (0.01)***
Education (in years/10)	0.17 (0.04)***	−1.03 (0.03)***
Age (in years/10)	−0.02 (0.02)	0.00 (0.01)
Household (single, no kids)	−0.02 (0.03)	−0.08 (0.02)***
Household (single, kids)	0.14 (0.07)*	0.04 (0.04)
Household (married, no kids)	−0.15 (0.03)***	−0.06 (0.02)**
Household (married, kids)	0.16 (0.04)***	0.07 (0.03)**
Household (other)	−0.13 (0.07)	0.03 (0.04)
Daily activity (working)	0.08 (0.04)*	−0.09 (0.02)***
Daily activity (looking for work)	0.09 (0.07)	−0.04 (0.04)
Daily activity (unable to work)	0.12 (0.06)*	−0.02 (0.04)
Daily activity (student)	0.02 (0.14)	−0.10 (0.09)
Daily activity (house wife/husband)	−0.18 (0.08)*	0.11 (0.05)*
Daily activity (retired)	−0.15 (0.04)***	−0.02 (0.03)
Daily activity (other)	0.03 (0.07)	0.16 (0.05)
Variance component
Neighbourhoods	0.09	0.03
Individuals	3.99	1.64
Model fit
Pseudo R2	0.09	0.11

Notes: Regression coefficients with standard errors in parentheses.

*

P < 0.05; **P < 0.01; and ***P < 0.001 (two-tailed test). N_individual = 21,200; N_{neighbourhood} = 3,068. The sample comprises native Dutch only. When analysing the relationship between minority group size and interethnic contact, we recoded the ‘No opportunity’ category as ‘Never’ (0) and treated the resulting contact variable as continuous. When analysing the relationship between minority group size and ethnic threat, we recoded the ‘I don’t know/no opinion’ category as ‘Neither disagree nor agree’ (2) and treated the resulting threat variable as continuous. Excluding these categories yields substantially similar findings. The control variables ‘gender’, ‘household composition’, and ‘daily activity’ are included as deviation contrasts. The control variables ‘age’, ‘education’, ‘economic affluence’, and ‘Δ economic affluence’ are grand mean centred. For interethnic contact, the variance components of the empty model are: variance_{neighbourhoods} = 0.38, variance_individuals = 4.09. For ethnic threat, the variance components of the empty model are: variance_{neighbourhoods} = 0.09 and variance_individuals = 1.79.

Sources: 1VOP (2015); Statistics Netherlands (2011, 2014).

Table A2

Cross-sectional models explaining relation between minority group size (and changes thereof) and ethnic threat for different levels of interethnic contact

	Ethnic threat (subsample: ‘No or irregular contact’)	Ethnic threat (subsample: ‘Regular contact’)	Significance of difference
(Intercept)	1.72 (0.04)***	1.70 (0.04)***
Contextual-level variables
% Non-Western minorities (REF: [0–5])
% Non-Western minorities (⟨5,10])	0.28 (0.03)***	0.06 (0.04)	***
% Non-Western minorities (⟨10,12.5])	0.35 (0.06)***	0.20 (0.06)***
% Non-Western minorities (⟨12.5,15])	0.41 (0.07)***	0.24 (0.06)***	*
% Non-Western minorities (⟨15,20])	0.59 (0.06)***	0.34 (0.05)***	**
% Non-Western minorities (⟨20,30])	0.59 (0.07)***	0.45 (0.06)***
% Non-Western minorities (⟨30, HI])	0.89 (0.09)***	0.46 (0.06)***	***
Δ % Non-Western minorities (REF: ⟨0,0.5])
Δ % Non-Western minorities (⟨0.5,1])	−0.02 (0.04)	0.01 (0.04)
Δ % Non-Western minorities (⟨1,2])	0.00 (0.05)	0.05 (0.05)
Δ % Non-Western minorities (⟨2,HI])	0.06 (0.09)	0.04 (0.07)
Δ % Non-Western minorities ([LO,0])	−0.04 (0.04)	−0.12 (0.04)**
Log(Economic affluence)	−0.03 (0.06)	−0.04 (0.06)
Log(Δ Economic affluence)	0.00 (0.01)	0.00 (0.01)
Individual-level variables
Gender (female)	−0.22 (0.02)***	−0.20 (0.02)***
Gender (male)	0.22 (0.02)***	0.20 (0.02)***
Education (in years/10)	−1.05 (0.04)***	−1.10 (0.04)***
Age (in years/10)	−0.01 (0.02)	0.00 (0.02)
Household (single, no kids)	−0.07 (0.03)*	−0.11 (0.03)**
Household (single, kids)	0.08 (0.07)	0.03 (0.06)
Household (married, no kids)	−0.07 (0.03)*	−0.07 (0.03)*
Household (married, kids)	0.11 (0.04)**	0.03 (0.04)
Household (other)	−0.05 (0.07)	0.11 (0.07)
Daily activity (working)	−0.14 (0.04)***	−0.02 (0.03)	*
Daily activity (looking for work)	0.01 (0.07)	−0.07 (0.06)
Daily activity (unable to work)	0.00 (0.05)	−0.02 (0.05)
Daily activity (student)	−0.17 (0.14)	−0.06 (0.13)
Daily activity (house wife/husband)	0.07 (0.07)	0.10 (0.08)
Daily activity (retired)	−0.05 (0.04)	0.00 (0.04)
Daily activity (other)	0.26 (0.07)***	0.08 (0.07)
Variance component
Neighbourhoods	0.04	0.03
Individuals	1.63	1.71
Model fit
Pseudo R2	0.11	0.10

Notes: Regression coefficients with standard errors in parentheses.

*

P < 0.05; **P < 0.01; and ***P < 0.001 (two-tailed test). The sample comprises native Dutch only. ‘No or irregular contact’ is defined as at most ‘several times a year’ (respondents with no opportunity for contact are not included). ‘Regular contact’ is defined as at least ‘about once a month’. For subsample ‘No or irregular contact’: N_individuals = 9,722; N_{neighbourhoods} = 2,648. For subsample ‘Regular contact’: N_individuals = 10,160; N_{neighbourhoods} = 2401. When analysing the relationship between minority group size and ethnic threat, we recoded the ‘I don’t know/no opinion’ category as ‘Neither disagree nor agree’ (2) and treated the resulting ethnic threat variable as continuous. Excluding this category yields substantially similar findings. The control variables ‘gender’, ‘household composition’, and ‘daily activity’ are included as deviation contrasts. The control variables ‘age’, ‘education’, ‘economic affluence’, and ‘Δ economic affluence’ are grand mean centred. For ethnic threat (subsample: ‘No or irregular contact’), the variance components of the empty model are: variance_{neighbourhoods} = 0.11, variance_individuals = 1.77. For ethnic threat (subsample: ‘Regular contact’), the variance components of the empty model are: variance_{neighbourhoods} = 0.09, variance_individuals = 1.83.

Sources: 1VOP (2015); Statistics Netherlands (2011, 2014).

Table A3

Cross-sectional models explaining relation between minority group size (and changes thereof) and voting for the PVV for different levels of interethnic contact

	PVV voting (subsample: ‘No or irregular contact’)		PVV voting (subsample: ‘Regular contact’)
	Model 1	Model 2	Model 1	Model 2
(Intercept)	−1.43 (0.08)***	−3.52 (0.19)***	−1.78 (0.09)***	−4.64 (0.25)***
Contextual-level variables
% Non-Western minorities (REF: [0–5])
% Non-Western minorities (⟨5,10])	0.11 (0.07)	−0.20 (0.08)**	0.01 (0.08)	0.00 (0.09)
% Non-Western minorities (⟨10,12.5])	0.20 (0.12)	−0.14 (0.13)	−0.02 (0.12)	−0.18 (0.14)
% Non-Western minorities (⟨12.5,15])	0.01 (0.13)	−0.46 (0.14)**	0.22 (0.12)	0.05 (0.14)
% Non-Western minorities (⟨15,20])	0.28 (0.11)*	−0.32 (0.12)*	0.18 (0.11)	−0.15 (0.13)
% Non-Western minorities (⟨20,30])	0.41 (0.13)**	−0.19 (0.15)	0.25 (0.12)*	−0.21 (0.14)
% Non-Western minorities (⟨30, HI])	0.59 (0.16)***	−0.28 (0.18)	0.28 (0.13)*	−0.17 (0.15)
Δ % Non-Western minorities (REF: ⟨0,0.5])
Δ % Non-Western minorities (⟨0.5,1])	−0.03 (0.07)	−0.03 (0.08)	−0.04 (0.08)	−0.07 (0.09)
Δ % Non-Western minorities (⟨1,2])	0.05 (0.09)	0.09 (0.11)	0.16 (0.09)	0.14 (0.11)
Δ % Non-Western minorities (⟨2,HI])	0.17 (0.16)	0.09 (0.18)	−0.01 (0.14)	−0.02 (0.16)
Δ % Non-Western minorities ([LO,0])	−0.08 (0.07)	−0.04 (0.08)	−0.17 (0.08)*	−0.05 (0.09)
Log(Economic affluence)	−0.05 (0.11)	−0.11 (0.13)	−0.09 (0.12)	−0.09 (0.14)
Log(Δ Economic affluence)	0.04 (0.02)	0.06 (0.03)*	−0.02 (0.02)	−0.04 (0.03)
Individual-level variables
Ethnic threat (REF: ‘Totally disagree’)
Ethnic threat (‘Somewhat disagree’)		0.77 (0.19)***		1.15 (0.25)***
Ethnic threat (‘Neither disagree nor agree’)		1.85 (0.18)***		2.30 (0.24)***
Ethnic threat (‘Somewhat agree’)		2.77 (0.18)***		3.47 (0.23)***
Ethnic threat (‘Totally agree’)		4.14 (0.18)***		4.88 (0.23)***
Ethnic threat (‘I don’t know/no opinion’)		2.29 (0.22)***		3.17 (0.31)***
Gender (female)	−0.27 (0.03)***	−0.07 (0.04)	−0.37 (0.04)***	−0.21 (0.04)***
Gender (male)	0.27 (0.03)***	0.07 (0.04)	0.37 (0.04)***	0.21 (0.04)***
Education (in years/10)	−1.67 (0.09)***	−1.03 (0.10)***	−1.62 (0.09)***	−0.86 (0.11)***
Age (in years/10)	−0.08 (0.03)*	−0.05 (0.04)	−0.11 (0.03)**	−0.11 (0.04)**
Household (single, no kids)	−0.26 (0.07)***	−0.26 (0.08)***	−0.26 (0.06)***	−0.20 (0.08)**
Household (single, kids)	0.43 (0.12)***	0.50 (0.15)***	0.19 (0.12)	0.24 (0.14)
Household (married, no kids)	−0.24 (0.06)***	−0.24 (0.07)***	−0.16 (0.06)**	−0.13 (0.07)
Household (married, kids)	0.06 (0.07)	−0.03 (0.09)	0.08 (0.07)	0.10 (0.09)
Household (other)	0.02 (0.13)	0.04 (0.15)	0.16 (0.13)	−0.01 (0.16)
Daily activity (working)	−0.08 (0.07)	0.07 (0.09)	−0.02 (0.07)	0.00 (0.09)
Daily activity (looking for work)	0.28 (0.12)*	0.34 (0.15)*	0.04 (0.12)	0.20 (0.15)
Daily activity (unable to work)	−0.03 (0.10)	−0.02 (0.12)	−0.03 (0.10)	−0.04 (0.12)
Daily activity (student)	−0.71 (0.31)*	−0.73 (0.36)*	−0.50 (0.29)	−0.71 (0.34)*
Daily activity (housewife/husband)	0.28 (0.14)*	0.28 (0.16)	0.39 (0.15)*	0.37 (0.18)*
Daily activity (retired)	−0.14 (0.09)	−0.12 (0.10)	−0.01 (0.09)	0.06 (0.10)
Daily activity (other)	0.41 (0.13)**	0.18 (0.15)	0.13 (0.14)	0.11 (0.16)
Variance component
Neighbourhoods	0.042	0.007	0.075	0.093
Model fit
Pseudo R2	0.11	0.43	0.11	0.49
AIC	9,310	7,135	8,896	6,493
BIC	9,503	7,365	9,091	6,724

Notes: Regression coefficients with standard errors in parentheses.

*

P < 0.05; **P < 0.01; and ***P < 0.001 (two-tailed test). The sample comprises native Dutch only. ‘No or irregular contact’ is defined as at most ‘several times a year’ (respondents with no opportunity for contact are not included). ‘Regular contact’ is defined as at least ‘about once a month’. For subsample ‘No or irregular contact’: N_individuals = 9,722; N_{neighbourhoods} = 2,648. For subsample ‘Regular contact’: N_individuals = 10,160; N_{neighbourhoods} = 2,401. The control variables ‘gender’, ‘household composition’, and ‘daily activity’ are included as deviation contrasts. The control variables ‘age’, ‘education’, ‘economic affluence’, and ‘Δ economic affluence’ are grand mean centred. For subsample ‘no or irregular contact’: the variance_{neighbourhoods} of the empty model is 0.04. For subsample ‘regular contact’: the variance_{neighbourhoods} of the empty model is 0.08. The individual-level variance is fixed (π²/3) because we use the logit link function.

Sources: 1VOP (2015); Statistics Netherlands (2011, 2014).