Support for the political community in a community that doesn’t support you: immigrant youths’ exposure to anti-immigrant attitudes

ABSTRACT

Expressions of hostility in immigrants’ surrounding environment have been shown to negatively influence their relation to the political society. The impact of such contextual hostility on immigrant youths and political community support however remains unexplored. Filling these gaps, I test the hypothesis that contextual hostility decreases immigrant youths’ political community support, captured with a sense of national pride. In addition, I test two rivalling hypotheses regarding the influence of contextual hostility on the so-called second generation of immigrants, expecting the impact to be either weaker or stronger among the second generation of immigrants in comparison to the first generation. Analysing survey data on the political beliefs of youths in secondary school classes in Sweden, my main finding is that immigrant youths in school classes with high levels of anti-immigrant attitudes are less supportive of the political community, entailing a potential legitimacy problem for the democratic system. The impact of anti-immigrant attitudes among the second generation of immigrants is in turn weak or non-existing.

KEYWORDS:

• Anti-immigrant attitudes
• contextual effects
• political community support
• immigrant youths
• Sweden
• national pride

1. Introduction

Anti-immigrant sentiments are one of many societal consequences following international migration. Scholars have traditionally attempted to explain the development of hostile attitudes towards immigrants, but a growing literature is shifting the attention, focusing instead on the impact of hostility on immigrant residents (Maxwell 2009; Connor 2010; Heath and Demireva 2014; Just and Anderson 2014; Just 2017; Pérez 2015). My paper engages with this literature, studying the hostility immigrants face in their surrounding environment and how such contextual hostility influences immigrant youths’ attitudes towards the host society.

Contextual hostility ranges from anti-immigrant attitudes in the fine-grained context of neighbourhoods to national state-level political debates, taking different forms of direct or indirect expressions. An example of a direct form of hostile expression is outspoken racism specifically aimed towards the individual, whereas indirect expressions of hostility include negative statements about immigration or immigrants as a group in general. Discrimination is also a specific form of hostile expressions, often the main focus in previous studies (Waisman 2008; Heath and Demireva 2014; Pilati 2018). Exposure to contextual hostility is thus a broader set of experiences that includes but is not exclusive to perceptions of discrimination.

Exposure to contextual hostility is likely to influence the political beliefs among the targeted individuals, threating social status or material interests. The impact of contextual hostility on immigrant youths and political community support as a potential outcome however remains unexplored. Addressing these gaps, my study makes three main contributions. First, by focusing on immigrant youths as the targeted group. Second, by testing the impact of contextual hostility on political community support as a previously untested outcome. In addition, I make an empirical contribution by measuring contextual hostility using anti-immigrant attitudes in the micro context of school classes.

Political community support is a diffuse form of support that typically arises during early adolescence and is captured by basic national attachment including national pride and national identification (Easton 1975; Norris 1999). Political community support is essential for the legitimacy of a democratic society. The development of such support is however likely to be challenged by the anti-immigrant rhetoric, which generally questions immigrants’ national belonging. The potential impact of contextual hostility among immigrant youths is especially problematic, considering early adolescence is an important transition time where experiences can shape identities and influence adult political beliefs (Beck and Jennings 1982; Krosnick and Alwin 1989; Pacheco 2008). Exposure to hostile contexts during the teenage years can therefore lead to long-term effects, causing immigrant residents to feel excluded from the political society and eroding future support for the political system.

I test the influence of contextual hostility in youths’ immediate surrounding environment, making use of anti-immigrant attitudes in school classes. This is a novel approach in the research field of contextual hostility and political beliefs, where studies mainly apply spatial approaches with national or regional boundaries (e.g. Jones-Correa 2001; Pantoja, Ramirez, and Segura 2001; Just and Anderson 2014) where the contextual impact is difficult to capture. The use of attitudes in school classes addresses the problem with capturing experiences of contextual hostility, since school classes are close environments where youths spend most of their day. If anti-immigrant attitudes exist in the class room, students are likely to encounter them. Centring on the hostile context created by anti-immigrant attitudes, I thereby focus on the ‘everyday’ form of hostility that immigrant youths face to various degrees in their fine-grained surrounding, where exposure to contextual hostility, to some extent, is inevitable.

Building on previous empirical findings, my first hypothesis suggests contextual hostility to decrease support for the political community among immigrant youths. The political community support is thus expected to be lower among immigrant youths in school classes with higher levels of anti-immigrant attitudes. Secondly, I raise two rivalling hypotheses regarding the so-called second generation of immigrants. While the assimilation theory expects subsequent generations of immigrants to converge with natives in terms of identity and socio-political attitudes (Alba and Nee 1997), the ‘paradox of social integration’ theory in contrast posits that the second generation of immigrants will be more aware of discrimination, and therefore react stronger to unjust treatment and prejudice than the first generation (Heath and Demireva 2014; Platt 2014). Applied to the setting of this paper, the assimilation theory would suggest the impact of contextual hostility to be weaker or non-existing among the second generation where the immigrant identity is less salient, whereas the social integration paradox posits the negative influence of contextual hostility to be stronger among the second generation in comparison to the first generation.

I set the study in Sweden; a country characterized by a high share of immigrant reception during the last decades. The attitudes towards immigrants in Sweden are in general positive when compared to other European countries (Hainmueller and Hiscox 2007), and Sweden was for long an exception to the electoral success of anti-immigrant parties (Rydgren and van der Meiden 2019). There are however differences in anti-immigrant attitudes at the subnational level (Demker 2013). This is further illustrated by the support for the Sweden Democrats – the biggest anti-immigrant party in Sweden – that gained parliamentary representation at the subnational level in 1991 before they reached the national level in 2010. I use data gathered in 2009 by the International Civic and Citizen Education Study (ICCS), which means my study is set in a time period where the salience of the immigration issue was lower on the national level with variation on the subnational level. The high level of immigration, limited political salience for the immigration issue on the national level and subnational variation in anti-immigrant attitudes makes Sweden at the time a unique case on which to test the influence of hostility, where hostility in the micro context to some extent can be isolated from the national level opinion climate.

My study yields three main results. First, as expected, I find lower political community support among immigrant youths in school classes with high levels of anti-immigrant attitudes. This result is robust to a number of model specifications. Secondly, I find weaker or at times non-existing correlations between contextual hostility and political community support among the second generation of immigrants, in line with the assimilation theory. Consequently, I find no signs of a social integration paradox, where the second generation was expected to react stronger to contextual hostility than the first generation.

The remainder of this paper proceeds as follows. First, I introduce the theoretical framework and review previous empirical findings, leading up to presentations of the proposed hypotheses. The research strategy is then presented, followed by the results from the statistical analyses. The paper ends with a concluding discussion, where I summarize the main findings and implications for future studies.

2. Theory and previous research

Exposure to contextual hostility is one of many forms of personal experiences that can shape an individual’s political beliefs. Two theoretical perspectives focusing on material and symbolic threats respectively generate insights as to how contextual hostility may influence political beliefs. Starting with the material threat, contextual hostility can provide distinct threats towards individual or group material interests, for instance by decreasing access to public welfare. Based on the assumption that political attitudes reflect material interest (Buchanan 1984), contextual hostility can thereby influence political beliefs by changing individuals expected personal or group utility. From this self-interest perspective, experiences of hostile contexts may lead individuals to conclude that the political community has less to offer you or members of your group, reducing political community support among individuals with immigrant background.

The second theoretical perspective is psychological and relates to social identity, where individuals are assumed to strive for a positive social identity (Tajfel and Turner 1979). Contextual hostility can here provide a symbolic threat, since it jeopardizes the status of one’s social identity. While the term immigrant refers to a broad category of individuals rather than a social identity, and identification with an ‘immigrant group’ more likely is something that others ascribe to you than something that comes from within, the level of group identification can vary depending on circumstances (Hylland Eriksen 2007). Experiences of anti-immigrant contexts can create heightened group boundaries between the majority and minority, increasing the perception of a shared minority identity (Alba 2005; Okamoto and Ebert 2010). Exposure to hostility may thus raise the salience of a minority immigrant background regardless of the individual’s own level of commitment to an immigrant identity. The heightened group boundaries between the majority and minority, i.e. natives and immigrants, may consequently impact immigrants’ political beliefs about the majority society.

Previous studies testing the impact of contextual hostility have found negative effects on immigrant’s relation to the political society. Anti-immigrant party parliamentary representation and restrictive immigration policies have for instance been shown to lower satisfaction with democracy among non-citizens (Just 2017) and decrease applications for citizenship (Jones-Correa 2001; Van Hook, Brown, and Bean 2006). Contextual hostility has also been found to mobilize political participation (Jones-Correa 2001; Ramakrishnan and Espenshade 2001; Pantoja, Ramirez, and Segura 2001; Okamoto and Ebert 2010; but see Just and Anderson 2014 for demobilizing findings), which at first may appear as an exception to the negative impact. The causal mechanisms used to explain this relationship however relates to the mentioned aspects of threats to personal or group interests (Ramakrishnan and Espenshade 2001; Pantoja, Ramirez, and Segura 2001), or heightened group boundaries (Okamoto and Ebert 2010), indicating an initial negative impact on the targeted individual. Overall, empirical findings thereby suggest that hostile surroundings risks eroding support for the political system.

2.1. Immigrant youths and political community support

Political community support is a diffuse form of system support that is distinct and independent from other more specific evaluations of political actors or formal institutions, and in general tends to be more durable than specific support, meaning it can be difficult to strengthen when weak, and vice versa. Political community support is therefore needed in a democratic system since it can function as a reservoir in periods with public dissatisfaction (Easton 1975; Norris 1999). Captured by basic national attachment including national pride and national identification (Easton 1975; Dalton 1999; Norris 1999), political community support is likely to be the form of support that is most threatened in a hostile context where immigrants’ belonging to the community is questioned.

The impact of contextual hostility is also likely to be greater on individuals in early adolescence, which is an important transition period for civic development (Metz and Youniss 2005; Eckstein, Noack, and Gniewosz 2012). Moreover, it is a time period where diffuse forms of support such as political community support typically arises (Easton 1975). We also know, in line with the impressionable years hypothesis, that experiences in young adolescence can shape adult political beliefs (Beck and Jennings 1982; Krosnick and Alwin 1989; Pacheco 2008). The influence of contextual hostility is, in other words, not only likely to be stronger among the younger generation, it also risks sticking into adulthood. Filling the gaps in previous studies on contextual hostility, political community support and immigrant youths, I therefore to test the expectation that:

H1: Contextual hostility decreases political community support among immigrant youths.

2.2. The assimilation theory and paradox of social integration

Turning to the group exposed to contextual hostility, immigrants with non-Western born background are more likely to be targeted than those with Western born background.¹ The non-Western born immigrant group is still highly diverse, theoretically consisting of immigrants as well as those who are native born with immigrant parents, the so-called first and second generation of immigrants. While the second generation have not migrated themselves, anti-immigrant attitudes generally tend to disregard such distinctions, targeting individuals with both direct and indirect migrant background.

The consequences of contextual hostility may however differ between the first and second generation of immigrants, and there are contrasting theoretical expectations regarding how the groups ‘should’ be influenced. Thinking of political integration in more general terms, the assimilation theory expects the second generation of immigrants to be closer to their native peers, converging in identity and socio-political integration (Alba and Nee 1997). This concerns both attitudes and engagement, where the gap between natives and immigrants in national attachment and political participation, for instance, is expected to decrease as time in the host country increases. From the perspective of the assimilation theory, the integration trajectory is thus linear to some extent with increased conversion between natives and immigrants with each new generation. Applied to the framework of this paper, the assimilation theory would expect the political community support of the second generation to resemble that of natives and the exposure to contextual hostility to have a weaker impact, since the second generation are assumed to identify less with their immigrant background. In line with the assimilation theory, hypothesis 2a thereby proposes that:

H2a: Contextual hostility has a weaker (or non-existing) impact on political community support among the second generation of immigrants in comparison to the first generation.

The assimilation theory however acknowledges that there can be differences in assimilation depending on contextual factors (Alba and Nee 1997; Platt 2014), which may explain empirical findings indicating less political integration then expected among the second generation of immigrants. A potential reason for the lack of convergence is provided by the paradox of social integration (Heath 2014), which suggests the second generation of immigrants to be more aware of discrimination and prejudice than their parents. The second generation of immigrants are, from this perspective, more likely to compare themselves with their native peers and therefore react stronger when treated in an unjust way than first generation immigrants, who are expected to have a different reference frame, comparing their life circumstances instead with the country of origin² (Rumbaut 2008; Heath 2014; Platt 2014). The social integration paradox closely resembles the more well-known integration paradox (de Vroome, Martinovic, and Verkuyten 2014), which focuses instead on education as the driving factor. According to the integration paradox, highly educated immigrants perceive discrimination to a greater extent than others and therefore turn away from the host society rather than integrate in to it. Similarly, the social integration paradox would posit the second generation to react stronger to contextual hostility than the first generation, decreasing political community support further. Testing for the paradox of social integration, hypothesis 2b in contrast to H2a posits that:

H2b: Contextual hostility has a stronger impact on political community support among the second generation of immigrants in comparison to the first generation.

3. Data and methods

The hypotheses are tested using within-country comparison in Sweden. I use data from the Swedish part of the 2009³ International Civic and Citizen Education Study (ICCS), a cross-national study aiming to collect information regarding the civic development of young people by testing the knowledge, values and engagement of youths in 8th and 9th grades. The study is conducted by the International Association for the Evaluation of Educational Achievement (IEA), an international cooperative consisting of national research institutions, government research agencies, scholars and analysts.⁴ The school classes are selected using a two-stage sampling procedure, where a random sample of schools are selected, followed by the selection of one or two intact target grade classes (Brese et al. 2011).

The ICCS data are advantageous since it enables testing the fine-grained context of school classes, where I can tap into contextual hostility in the immediate surrounding. Another benefit with the survey is that it reaches a large group of youths, considering secondary school attendance is mandatory in Sweden. This a point worth highlighting, since it means selection bias is limited in comparison to surveys distributed to high school students, where attendance is voluntary. The sample should therefore be representative of the Swedish youth population.

The Swedish sample in total includes 6979 students in 169 schools, with a response rate above 90 per cent for the selected students (Skolverket 2010). Out of those, 517 are first generation and 817 are second-generation immigrants (born in Sweden with two foreign-born parents), resulting in more than 1300 respondents in 255 school classes available for the analyses. While not able to distinguish further on origin, immigrants from non-Western countries make up a significant part of the immigrant reception in Sweden (Statistics Sweden 2019). The immigrant measurement is thus likely to capture youth from the most targeted group.

Following previous studies on political community support, I operationalize the concept using questions related to national pride (Dalton 1999; Klingemann 1999; Norris 1999, see also discussion about dimensions of national pride in De Figueiredo and Elkins 2003). The ICCS poses two statements about national pride: ‘In Sweden, we should be proud of what we have achieved’, and ‘I am proud to live in Sweden’. The original response scales range from 1 = strongly agree to 4 = strongly disagree. These are reversed and combined into a national pride index, with values ranging from 1–7 so that 1 = low pride and 7 = high pride (Cronbach’s α: 0.76).

The frequency distribution of national pride is presented in Table 1. As shown, the pride levels are somewhat higher among natives than the first and second generation of immigrants. The pride levels are however generally high in all groups, with most students ranking on the top levels of the pride index.⁵ This distribution of national pride goes in line with the cross-national findings of Klingemann (1999), where 85 per cent of Swedish citizens reported high political community support in terms of national pride and willingness to fight for the country. The low variation in pride should set for conservative tests of the hypotheses when attempting to trace the contextual impact.

Table 1. Frequency distribution – National pride index.

	1st generation		2nd generation		Natives
National pride	Frequency	Per cent	Frequency	Per cent	Frequency	Per cent
1 (Low pride)	24	5	21	3	62	1
2	8	2	16	2	40	1
3	32	7	44	6	211	5
4	74	15	90	11	426	9
5	179	37	323	41	1715	37
6	75	15	113	14	839	18
7 (High pride)	98	20	179	23	1281	28
Total	490	100	786	100	4574	100

Note: Samples include first generation immigrants, second-generation immigrants, or natives with two native born parents.

Turning to the measurement of contextual hostility, I use the students’ responses to statements about immigrants’ rights and opportunities, taking into consideration that the context is created by the attitudes of all students, not only those who are natives. The responses are combined to an index and aggregated to the class level, which is a form of compositional contextual variable (Brooks and Prysby 1991) indicating the mean level of anti-immigrant attitudes per class. The contextual measurement includes statements related to immigrants’ rights, for instance the right to vote in elections, continue their own customs and speak their own language (Cronbach’s α: 0.88).⁶ The response scale ranges from 1 = strongly agree to 4 = strongly disagree, so that a higher number equals a more restricted position towards immigrants.⁷ The values of the additive index theoretically range from 1 to 16, but the empirical range in the sample roughly goes from 2 to 9 with a fairly normal distribution (see Figure 1 in Appendix for a histogram portraying the distribution).

In order to clarify that I am capturing the impact of aggregated hostility rather than a correlation between the student’s own perception of immigration and national pride, I include the attitudes towards immigrants on the individual level as a control. Considering the individual and aggregated measurement of anti-immigrant attitudes correlate, I follow the approach by Campbell (2008) and Persson (2015) and ‘purge’ the measurements from any correlation. The contextual measurement of hostility is cleaned from individual influence by regressing the average attitudes on the individual level attitudes. I then save the residuals from this model and use them as the contextual measurement of hostility. This means each individual receives their own value on the contextual measurement depending on their attitudes towards immigration. Likewise, I regress the individual level attitude on the average attitude and use the residuals from this model as a measurement of the individual’s attitudes towards immigration.

As with all cross-sectional studies, the risk of endogeneity should be taken into consideration. A potential objection to my theoretical argument is that low levels of pride among immigrant students could increase anti-immigrant attitudes, rather than the other way around. An important contextual control to include here is the share of immigrants within each class, which is likely to influence the level of anti-immigrant attitudes as well as the political community support. Including the share of immigrants should also address the potential risk of reversed causality, considering the unlikeliness that low levels of pride among immigrants would increase anti-immigrant attitudes if the presence of immigrants does not, since the presence of immigrants in itself should be more visible than the levels of national pride among the present immigrants. The share of immigrants is calculated using the number of foreign-born students divided by the total number of students per class.⁸ As shown in the correlation matrices (see Appendix, Tables 8 and 9), there is a negative and significant correlation between the share of immigrants and anti-immigrant attitudes⁹, indicating that the presence of immigrants decreases contextual hostility. The risk of reversed causality should thus be limited.

Additional variables are included on the contextual level in order to control for potential confounding factors. Capturing the socioeconomic status of the school class, I aggregate students’ approximation of number of books in their home. This measurement is frequently used as a proxy for the socioeconomic status of the family (Wolbrecht and Campbell 2007; Campbell 2009; Lopes, Benton, and Cleaver 2009; Persson 2015). The responses are ordered in categories from 1 to 6, where 1 ≤ 10 books at home and 6 ≥ 500 books at home. The variable is purged from any correlation with the individual response to books using the same procedure as for the anti-immigrant attitudes. Finally, I use survey responses from the principals to measure the community size and type of school (public/private) for each school class.

At the individual level, I include gender and the individual response to number of books in their home (again, purged from correlation with the contextual measurement of books). I also include self-reported experiences of impartial treatment. The purpose here is to control for a subjective perception of discrimination, in order to capture a more unique effect of contextual hostility unrelated to the subjective perception. In addition, experiences of impartiality have previously been argued to have a positive impact on social trust (Kumlin and Rothstein 2010), which also relates to support for the political system. Experiences of impartiality are measured with the student’s perception of fairness in teacher’s treatment, where a higher number equals perception of unfair treatment. Finally, tapping into integration specific factors, I use a dummy variable indicating the language that is mostly spoken at home (Swedish or other), taking into account that native speaking can correlate with national pride. Moreover, the level of integration within each class may impact the attitudes towards immigration. Descriptive statistics of the variables included in the analyses are presented in Table 2. A description of all questions and answer alternatives and correlation matrices are shown in the Appendix, Tables 8–10.¹⁰

Table 2. Descriptive statistics.

Statistic	N	Mean	St. Dev.	Min	Max
Individual level
National pride	1301	5.141	1.452	1	7
Origin (1 = Foreign-born)	1334	0.388	0.487	0	1
Anti-immigrant attitudes	1310	2.928	2.514	1	16
Teacher treatment (4 = Unfair)	1346	1.923	0.813	1	4
Gender (1 = Girl)	1348	0.487	0.500	0	1
Language (1 = Swedish)	1293	0.300	0.458	0	1
Books (6 = >500)	1347	3.072	1.345	1	6
Class level
Mean anti-immigrant attitudes	1359	3.924	1.313	1.440	9.400
Share of immigrants	1359	16.728	13.025	0	62.500
Mean books	1359	3.430	0.619	2.180	5.110
School level
School type (1 = Private)	1256	0.173	0.378	0	1
Community size (5 = Large city)	1256	3.552	1.028	1	5

Note: Sample including first and second generation immigrants.

3.1. Model specification

I test the hypotheses using multilevel linear analyses. The model includes three levels, with students (i) embedded in classes (j) and schools (k). The full model is specified below, where $Ant{i}_{jk}$ is the main level two variable, the random intercept is specified as $U_{0jk}$, and matrix of control variables as ${\mathit{X}}_{ijk}$ (individual level), ${\mathit{X}}_{jk}$ (class level) and ${\mathit{X}}_k$ (school level).$\begin{array}{rl}Prid{e}_{ijk}& =\alpha +\gamma 10\left(Ant{i}_{ijk}\right)+\gamma 01\left({\overline{Anti}}_{jk}\right)+\gamma 20\left({\mathit{X}}_{ijk}\right)+\gamma 02\left({\mathit{X}}_{jk}\right)+\gamma 001\left({\mathit{X}}_k\right)\\ & +U_{0jk}+R_{ijk}\end{array}$For H2a and H2b, I also compare the correlations with political community support between first and second-generation immigrants by including a cross-level interaction between origin and the aggregated anti-immigrant attitudes, written as $\left(Origi{n}_{ijk}\right)\left({\overline{Anti}}_{jk}\right)$. The slope of origin is allowed to vary between classes, specified as $U_{1jk}\left(Origin\right)$ rendering the following model.$\begin{array}{rl}Prid{e}_{ijk}& =\alpha +\gamma 10\left(Ant{i}_{ijk}\right)+\gamma 20\left(Origi{n}_{ijk}\right)+\gamma 01\left({\overline{Anti}}_{jk}\right)\\ & +\gamma 2001\left(Origi{n}_{ijk}\right)\left({\overline{Anti}}_{jk}\right)+\gamma 30\left({\mathit{X}}_{ijk}\right)+\gamma 02\left({\mathit{X}}_{jk}\right)+\gamma 001\left({\mathit{X}}_k\right)+U_{0jk}\\ & +U_{1jk}\left(Origin\right)+R_{ijk}\end{array}$

4. Empirical findings

Results from the analyses are presented in Table 3. Models 1 and 2 show the variation in national pride on the full sample of students, including natives, first generation of immigrants, and second generation of immigrants. The purpose here is to test if there is, on the outset, any variation in national pride between natives and immigrant youths without the inclusion of contextual hostility. As shown in Model 1, the coefficient for the binary variable capturing first generation of immigrants is negative but falls below the standard threshold of statistical significance (p = 0.08). This indicates that the first generation of immigrant youths are less proud than their native and second-generation peers, but the finding is uncertain considering the low level of statistical significance. Model 2 presents the results when replacing the measurement of origin, now indicating if the student is second-generation immigrant. A significant result would here render initial support for the social integration paradox (H2b), signalling lower political community support within the second generation. The coefficient however remains statistically insignificant, indicating that the political community support among second-generation immigrants on average resembles that of natives, which is more in line with the assimilation theory (H2a).

Table 3. Effects of contextual hostility on national pride.

	1	2	3	4	5
	Full sample	Full sample	1st gen	2nd gen	1st and 2nd gen
Individual level
AI attitudes	0.003	0.001	−0.068*	−0.047	−0.060**
	(0.007)	(0.007)	(0.028)	(0.024)	(0.018)
Teacher treatment (unfair)	−0.301***	−0.298***	−0.556***	−0.336***	−0.418***
	(0.024)	(0.024)	(0.085)	(0.064)	(0.051)
Gender (girl)	−0.044	−0.038	0.310*	0.082	0.170*
	(0.035)	(0.035)	(0.133)	(0.103)	(0.082)
Language (Swedish)	0.387***	0.428***	0.146	0.345**	0.278**
	(0.063)	(0.066)	(0.164)	(0.114)	(0.093)
Books	−0.075***	−0.076***	−0.179**	−0.085	−0.130***
	(0.017)	(0.017)	(0.068)	(0.047)	(0.038)
Origin (1st generation)	−0.133
	(0.075)
Origin (2nd generation)		−0.016			0.018
		(0.067)			(0.090)
Class level
Mean AI attitudes	−0.032	−0.036	−0.216**	−0.102	−0.154**
	(0.022)	(0.022)	(0.069)	(0.057)	(0.059)
Mean books	−0.217***	−0.226***	−0.574**	−0.296*	−0.407***
	(0.058)	(0.059)	(0.188)	(0.124)	(0.103)
Share of immigrants	0.003	−0.006	−0.184	−0.053	−0.090
	(0.028)	(0.028)	(0.096)	(0.066)	(0.056)
Mean AI attitudes x Origin					0.031
					(0.067)
School level
School type (private)	−0.121	−0.117	−0.313	−0.084	−0.128
	(0.069)	(0.069)	(0.210)	(0.153)	(0.123)
Community size	0.013	0.013	−0.039	0.073	0.021
	(0.023)	(0.023)	(0.076)	(0.061)	(0.048)
Constant	5.630***	5.576***	6.155***	5.501***	5.774***
	(0.103)	(0.103)	(0.318)	(0.270)	(0.205)
Random intercept	Yes	Yes	Yes	Yes	Yes
Random slope	No	No	No	No	Yes
Classes	309	309	178	179	234
Schools	152	152	116	112	137
Observations	5771	5708	422	658	1080
Log Likelihood	−9,648.011	−9,541.909	−713.714	−1,106.665	−1,827.741
Akaike Inf. Crit.	19,326.020	19,113.820	1,455.427	2,241.329	3,687.483
Bayesian Inf. Crit.	19,425.930	19,213.560	1,512.057	2,304.178	3,767.238

Note: *p < 0.05; **p < 0.01; ***p < 0.001. The samples are limited as follows: models 1–2 includes full sample of students, model 3 includes first generation immigrants, model 4 includes second generation immigrants, model 5 includes first and second-generation immigrants.

Moving on to the main test of H1, I limit the sample including only the first generation of immigrants and introduce the variable capturing contextual hostility to test whether the national pride changes in contexts with higher levels of anti-immigrant attitudes, where I expect increased degrees of contextual hostility to decrease immigrant youths’ political community support. Model 3 presents the result, where the main finding is the negative and significant coefficient for mean anti-immigrant attitudes at the class level, indicating a contextual effect of hostility on political community support under control for the individual level attitude. An immigrant student is thereby less supportive of the political community if attending a school class where the average level of anti-immigrant attitudes is higher, taking the additional individual and contextual variables into consideration. The results hold also when testing the contextual correlation with the two original separate outcomes, proud of living in Sweden and proud of achievements in Sweden (see Table 12 in the Appendix), showing negative correlations with contextual hostility and each respective proud measurement.

The empirical finding can be further portrayed with an illustrative example. An immigrant student attending a school class where contextual hostility is at the observed lowest value (1.4) is predicted to score a 6 on the pride index. The level of pride is expected to decrease by 2 steps if the hostility within the same class would increase to the empirical maximum (9.4), a change equivalent to more than one standard deviation in pride levels. The level of political community support among immigrant students in friendly school contexts thus mirror the support of natives, whereas it decreases in schools where immigrants’ rights are questioned, rendering strong support for H1.¹¹

Turning to a test of the assimilation theory (H2a) and the paradox of social integration (H2b), I run the same analysis as above but on a new sample including only the second generation of immigrants, testing the potential correlation with contextual hostility and national pride within this subgroup. The expectation according to the assimilation theory is that contextual hostility has a weaker or non-existing influence among the second generation in comparison to the first generation, whereas the paradox of social integration instead suggests a stronger effect of hostility on the second generation. The results are presented in Table 3, Model 4. As shown, the association between average anti-immigrant attitudes and national pride is negative but weaker than in the previous analysis on the first generation of immigrants and falls just below statistical significance (p = 0.07). The correlation is thus less statistically robust than among the first generation of immigrants, but the result to some extent suggests that contextual hostility decreases the political community support also among the second generation of immigrants.

Testing the comparison between the first and second generation further, I widen the sample and include both first and second generation of immigrants, introducing an interaction with origin and contextual hostility. Origin is coded as 1 if the respondent is second-generation immigrant and 0 if first generation. Results are presented in Table 3, Model 5, showing a positive but insignificant coefficient for the interaction term, indicating there is no significant difference in contextual effects between the first and second-generation immigrants. This result supports the findings from previous analyses, rendering no support for H2b, where the second generation was expected to react stronger to contextual hostility than the first generation. While not fully supporting H2a either, the result is more in line with the assimilation theory. The impact of contextual hostility is somewhat present also within the second generation, but it is weaker and less robust than within the first generation.

4.1. Robustness tests

Additional analyses were conducted in order to test the robustness of the results. First, I re-ran the analysis of H1 using a sub-sample including only students who are native with two native born parents. While political community support among natives also could be influenced by the surrounding attitudes towards immigration, for instance as a sign of solidarity towards their immigrant peers, the initial findings would be strengthened if the impact is non-existing in this group, considering a lack of impact among natives indicates a more unique effect of contextual hostility on the targeted group. As shown in Table 4, Model 1, the coefficients for the individual as well as the mean level of anti-immigrant attitudes are insignificant. The political community support among native youths is thereby not influenced by the surrounding attitudes towards immigrants.

Table 4. Effects of contextual hostility on national pride, robustness tests.

	1	2	3	4	5	6	7
	Natives	1st gen	2nd gen	1st and 2nd gen	1st gen	2nd gen	1st and 2nd gen
Origin (2nd gen)				0.045			0.051
				(0.086)			(0.095)
AI attitudes	0.011	−0.061*	−0.055*	−0.061***	−0.078**	−0.042	−0.065***
	(0.008)	(0.027)	(0.024)	(0.018)	(0.027)	(0.024)	(0.018)
Mean AI attitudes	0.008	−0.198**	−0.103*	−0.177***
	(0.025)	(0.063)	(0.049)	(0.053)
Mean AI attitudes (natives)					−0.151**	−0.055	−0.104*
					(0.048)	(0.039)	(0.046)
Mean AI attitudes x Origin				0.076
				(0.064)
Mean AI attitudes (natives) x Origin							0.026
							(0.058)
Constant	5.997***	5.957***	5.680***	5.764***	6.250***	5.420***	5.774***
	(0.091)	(0.192)	(0.144)	(0.124)	(0.307)	(0.279)	(0.203)
Random intercept	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Random slope	No	No	Yes	No	No	Yes	No
Classes	298	194	198	255	167	168	223
Schools	148				111	108	113
Observations	4079	458	713	1171	370	540	910
Log Likelihood	−6,732.744	−781.599	−1,203.278	−1,991.892	−612.629	−903.910	−1,525.299
Akaike Inf. Crit.	13,491.490	1,585.198	2,428.555	4,013.785	1,253.258	1,835.820	3,090.598
Bayesian Inf. Crit.	13,573.560	1,630.594	2,478.820	4,089.769	1,308.047	1,895.902	3,186.867

Note: *p < 0.05; **p < 0.01; ***p < 0.001. Individual and contextual controls included. Model 1 includes a native sample. Models 2–4 excludes school level variables. Models 5–7 includes alternative measurement of the anti-immigrant attitude index, including only the attitudes of natives. The samples are limited as follows: model 1 includes native students, models 2 and 5 includes first generation immigrants, models 3 and 6 includes second generation immigrants, models 4 and 7 includes first and second-generation immigrants.

As a second step, I ran the analyses excluding the school level variables. The participation rate for principals is 92 per cent (Skolverket 2010), which means a loss of observations for those classes lacking principal participation. Community size and type of school were therefore excluded in order to test whether the results hold on the larger sample of students and classes. Moreover, I tested the robustness of the anti-immigrant attitude index by aggregating the attitudes of students that are native with two native born parents only instead of using the attitudes of all students. The results are shown in Table 4, Models 2–7, closely resembling those of previous tests when it comes to the impact of contextual hostility among the first generation of immigrants (H1).¹² There are however slight differences in the outcomes from the analyses run on the second generation of immigrants when testing H2a and H2b. Model 3 presents the results when the school variables are excluded. As shown, the coefficient for mean anti-immigrant attitudes is now statistically significant. When changing the measurement of anti-immigrant attitudes to that only including natives, the coefficient again drops below statistical significance (Model 6). The statistical significance thus varies depending on model specification, but leans toward support for H2a considering the association is consistently either non-existing or weaker among the second generation compared to the first generation of immigrants.¹³ Overall, the results from the robustness tests mirror those of the original analyses.

5. Conclusion

Low political community support among immigrant youths entails a legitimacy problem for the democratic system, and a potential obstacle for long-term political integration. This study shows that the hostility facing immigrants in their surrounding environment can play an important role, influencing the levels of political community support within the targeted group. Setting out to fill the gaps related to contextual hostility, immigrant youths and political community support, my study finds that anti-immigrant attitudes in school classes decreases immigrant youths support for the political community, captured with a sense of national pride. Immigrant youths are, in other words, less supportive of the political community if exposed to a more hostile school context. This finding is especially relevant considering the analyses show immigrant youths to have similar levels of political community support as natives in general. The legitimacy problem with low political community support thus only exists in hostile contexts.

The association between contextual hostility and political community support is to some extent present also among the second generation of immigrants, but it is either weaker than among first generation immigrants or non-existing depending on model specification. The empirical findings are thereby mainly in line with the assimilation theory, where the second generation of immigrants are expected to converge with the majority in terms of identity and socio-political attitudes and thus become less influenced by hostile expressions. The social integration paradox which suggests a stronger impact of hostility on political community support among the second generation of immigrants in comparison to the first generation is consequently not supported, indicating a less far-reaching influence of hostility than expected.

Moving forward, more attention should be paid to the complexity of the targeted immigrant group, clarifying the potential importance of intersectionality. Hostility towards immigrants does not exist in isolation, and the consequences of exposure to contextual hostility likely depend on intersections with aspects such as gender and class. While I control for these in the empirical analyses, future studies should take the next step and theorize and analyse further on how and why intersectionality matters.

It is important to note that my study captures something different than perceptions of discrimination, frequently studied in previous work (Waisman 2008; Heath and Demireva 2014; Pilati 2018). Exposure to contextual hostility is another form of personal experience, including discrimination but also covering more indirect expressions of hostility. It is also worth reminding that the study is conducted in the context of school classes, where values such as equal treatment and solidarity are highlighted and hostile attitudes are expected to be countered. The exposure to contextual hostility may therefore be mitigated in these settings, indicating a stronger effect in other environments.

The empirical findings of my study underline the notion that context matters, and that the surrounding environment is important to consider when explaining individual political beliefs. More specifically, my study contributes to the literature on hostile expressions and political beliefs (Jones-Correa 2001; Heath and Demireva 2014; Just 2017; Pérez 2015), highlighting the negative impact of contextual hostility on a previously unexplored group and political outcome, namely immigrant youths and political community support. The impact of hostility on immigrant youths is particularly problematic considering they are part of the next generation of political citizens, and the development of political community support in these ages shapes political beliefs in adulthood. Exposure to contextual hostility in early adolescence thus risks eroding future levels of support for the political system.

The findings of contextual hostility and low political community support however indirectly also provide a way to increase societal cohesion, indicating higher political community support for those who are part of contexts where immigrants’ rights are promoted. An alternative takeaway is thereby that political integration prospers in more inclusive contexts.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability

The data that support the findings of this study are openly available in figshare at https://figshare.com/articles/_/6974114

Notes

1 See for instance Rydgren’s study (2006) on ethnic discrimination in the Swedish labour market.

2 Second generation immigrants with non-Western born background in Sweden are, for instance, more likely to be unemployed and have lower wages than natives, potentially due to discriminating structures (Rooth and Ekberg, 2003; Rydgren 2006)

3 The ICCS has been conducted in 1971, 1999, 2009 and 2016.

4 www.iea.nl

5 The frequency distribution of each original variable included in the national pride index resembles that of the index, tables presented in Tables 5 and 6 in the Appendix.

6 All statements are listed in the Appendix, Table 7. All items load above 0.4 after varimax rotation when running a polychoric factor analysis. See Table 7 in the appendix for factor loadings. I tested alternative indices, where I included the items loading either below or above 0.7 into two separate measurements, potentially indicating different dimensions of anti-immigrant attitudes. The inclusion of these indices yielded similar results in the analyses as those of the full index. An additional statement about immigration was left out of the study due to low correlation with the other questions about immigrants, also shown in Table 7.

7 The additive index includes five questions, the scale is recoded from 5–20 to 1–16 for a more intuitive interpretation.

8 I tested another version of the share of immigrants measurement in the analyses, where I calculated the share of first and second-generation immigrants per class. The main results remained when including this measurement.

9 The correlation between the share of immigrants and anti-immigrant attitudes was also tested on a limited sample including only the native population, with the same finding as in the full sample.

10 Descriptive statistics from the full sample of students is shown in the Appendix, Table 11.

11 Analyses illustrating the association between contextual hostility and national pride with controls added stepwise are shown in the Appendix, Table 13.

12 The hypotheses were also tested using multilevel ordered probit analyses, with results mirroring those of the linear models.

13 I tested an alternative measurement of the second-generation immigrants, including those who are born in Sweden with one foreign-born parent. The interaction term remained statistically insignificant, but the coefficient for anti-immigrant attitudes in the limited sample including only second-generation immigrants was negative and statistically significant.