Introduction

Large-scale immigration has diversified electorates in representative democracies around the world. One in five Dutch eligible voters currently have roots abroad; nearly 2.5 million of them were born outside the Netherlands or have at least one parent who was (Lubbers and Spierings 2021). Such changing demographics seldom translate directly or unambiguously into political systems (Lee 2008; Hochschild and Mollenkopf 2009). This hastens the need to study political participation divides between citizens with and without an immigrant backgroundFootnote 1 and to identify the main factors that cause such gaps. What is more, studies have shown that across political contexts, racialized minorities are less likely to vote and less likely to be or feel equally represented than the majority (Leighley and Vedlitz 1999; Jacobs and Skocpol 2005; Ramakrishnan 2022; Griffin and Newman 2007; De Rooij 2012; Bird et al. 2011; Fraga 2018), and it is these groups we refer to when talking about immigrant background citizens.

Studies suggest that more traditionally examined factors, such as socioeconomic status, education, and age (Verba and Nie 1993), help explain lower levels of political participation among citizens with an immigrant background (Leighley and Vedlitz 1999; De Vroome et al. 2013; Spies et al. 2020). However, to better understand political divides, we also need to take into account how this group experiences and relates to the political system in contexts in which they experience exclusion and in which immigration and diversity are politicized. For instance, Fraga (2018) shows that individuals are likelier to vote when they expect to be able to influence the political process (cf. Jeroense & Spierings 2023). Accordingly, if citizens with an immigrant background feel they do not belong to or have a stake in society—that they have “nothing to fight for”—they will refrain from participating in that system whatsoever (De Rooij 2012; Kranendonk et al. 2018). So in an immigration context, political participation, political representation, and external efficacy can rarely be fully understood without considering the role of religious–cultural identification and the (negative) reception, if not rejection, of minoritized groups in such processes.

Drawing from a setting characterized by such divides—Western Europe—this article delves into the indirect effects of inclusion and exclusion processes on political engagement (Pérez 2015). We use the concepts of inclusion and exclusion to synthesize research on the “migration gap” in political engagement (De Vroome et al. 2013; Heath et al. 2013; Spies et al. 2020) and research that addresses racialized immigrant communities’ diverse positions and experiences (e.g., Geurts 2022; Kranendonk and Vermeulen 2019; Pilati & Morales 2016; Spierings 2016; Tillie 2004). Doing so, we seek to understand the complex relationship between factors related to cultural processes of inclusion and exclusion and political engagement. Analyzing two pairs of linked concepts, we set out to (1) provide new theoretical insights through integrating different literatures and (2) contribute an understanding of these complex processes by analyzing them simultaneously.

The first conceptual pair comprises subconcepts of political engagement: citizens’ evaluation of politics and their electoral participation. While levels of political evaluation higher than or equal to that of citizens without an immigrant background are often considered signs of political integration, the general literature also shows that negative evaluations can spur participation (e.g., Jeroense and Spierings 2023). Studying these outcomes simultaneously, we can better understand the processes impacting political engagement.

Our second pair concerns perceived cultural and religious inclusion and exclusion. De Vroome et al. (2013) encourage investigation of cultural explanations, notably religiosity, and perceived group discrimination to explain political differences between voters with and without an immigrant background. It is therefore important to examine whether membership and embeddedness in religious organizations correlate not only to political participation, as has been studied extensively (Fennema and Tillie 1999; Sobolewska et al. 2015), but also to the evaluation of politics, so far hardly studied (De Vroome et al. 2013). Uniting these different cultural factors and dimensions of political engagement, we arrive at so far largely untested explanations. Testing them provides a look at how cultural and migration factors contribute to or counter gaps in political engagement (as has been urged by, e.g., Bird et al. 2011; De Vroome et al. 2013).

To test our expectations, we drew data from the 2021 Dutch Parliamentary Election Study (DPES) (Jacobs et al. 2021) in combination with the 2021 Dutch Ethnic Minority Election Study (DEMES) (Lubbers et al. 2021). The nature of the data enabled establishing the political engagement gaps among various groups of voters with an immigrant background and to see whether explanations hold across groups from different parts of the world, including the understudied ones (see Sanders et al. 2014; Pilati & Morales 2016). The Dutch case entails a salient polarized context with three credible ethnic-minority interest parties credibly competing in the national elections (see Lubbers et al. 2022) and debates on racism, the blackface character Zwarte Piet, migration, and Islam. If a participation gap is found in such a context, we believe it is likely to be found in other Western democratic countries as well. The same may not hold true for political trust, however, because strong signals of not belonging fuel this very issue’s debate. Notably, in 2021, the Dutch government collapsed over a social benefits scandal involving statistical discrimination of citizens with an immigrant background.

Theoretical background

Gaps between citizens with and without an immigrant background in terms of both political participation and evaluation are attributable to differences in socioeconomics and demographics, but gaps remain even after accounting for such differences for some contexts and groups (De Vroome et al. 2013; Fraga 2018). Since traditional explanations do not always account for the engagement gap, focus has shifted to migration-related explanatory factors (Bird et al. 2011; Bloemraad and Vermeulen 2014). We have heeded suggestions for investigating cultural reception-related variables (De Vroome et al. 2013) to explain the engagement gaps. Congruently, we draw theoretical connections between literature on religious organization and ethno-religious discrimination and literature on the political engagement gap.

Before moving onto those explanations, we offer two hypotheses, which capture the explorative nature of our study and acknowledge that gaps might not exist for all groups.

H1a

Citizens with an immigrant background, or from a particular national background, have lower electoral participation likelihood than citizens without one.

H1b

Citizens with an immigrant background, or from a particular national background, evaluate politics more negatively than citizens without one.

The mobilizing capacity of religious networks

Embeddedness in social and organizational networks enhances political engagement among immigrants and non-immigrants (La Due Lake and Huckfeldt 1998; Fennema and Tillie 1999; De Rooij 2012; Spies et al. 2020), and this impacts the relevance of religious–cultural integration. Being embedded enables individuals to acquire abilities (civic skills, cooperation), political information, and the mentality (political interest, trust) that spur political engagement and participation (Van der Meer and Van Ingen, 2009). Formalized religious networks affect political participation by providing strong recruitment networks, resources, and incentives for members’ involvement in civic and political matters (Campbell 2013; Hirschman 2004; Kranendonk and Vermeulen 2019). Cesari (2014), Kranendonk and Vermeulen (2019), and Just et al. (2014) found that Muslim immigrants who are exposed to religious organizations participate significantly more in politics than those who are not. We also expected formalized religious networks to increase feelings of political trust and efficacy. Because in Northern and Western Europe, on average, religious embeddedness is higher among immigrant than among non-immigrant groups (Diehl et al. 2009), we expect that embeddedness in religious communities (i.e., being active in a religious intermediary groups or social network, for instance signified by attending mosque or church) suppresses or obscures a larger engagement gap.

H2a

Negative electoral participation gaps between citizens with an immigrant background and without one are suppressed by immigrants’ stronger embeddedness in religious communities.

H3a

Negative evaluation gaps between citizens with an immigrant background and without one are suppressed by immigrants’ stronger embeddedness in religious communities.

However, we also consider that involvement in an immigrant-origin or religious organization might not affect political participation positively. Such organizations primarily connect or bond like people, generating in-group rather than generalized trust. Consequently, political engagement is not enhanced beyond the group.

Bonding might also lead to political disengagement from society at large, likely as a result of retention processes in the context of perceived group discrimination (cf. Brown 2000; Röder and Spierings 2022). A group who experiences rejection through discrimination and hostility may, as a countering mechanism, try to generate a positive self-image by emphasizing a group’s distinguishing characteristics more positively. Religious embeddedness facilitates in-group networks that can strengthen feelings of political apathy and inefficacy. Accordingly, Kranendonk and Vermeulen (2019) found that membership in an immigrant or religious organization negatively correlates with voting of Turkish and Moroccan immigrants in the Netherlands. Again, while studies have focused on participation, part of this mechanism seems rooted in how groups evaluate politics in the first place.

H2b

Negative electoral participation gaps between citizens with an immigrant background and without one are partly explained by immigrants’ stronger embeddedness in religious communities.

H3b

Negative evaluation gaps between citizens with an immigrant background and without one are partly explained by immigrants’ stronger embeddedness in religious communities.

Effects of perceptions and experiences of rejection

As already touched upon above, discrimination can increase or decrease the levels of political participation among voters with an immigrant background. Political discrimination encompasses laws, policies, and political campaigns and discourse that deprive some groups of collective resources and rights. Social discrimination entails negative actions in the form of verbal or non-verbal albeit physical intimidation against minorities (Oskooii 2020). Distinguishing these interrelated dimensions helps us conceptualize how the experience of politics, next to social experiences, itself matters, something seldom specifically studied. It allows us to introduce and theorize the role of feeling represented as part of an ethnic minority under the broader umbrella of discrimination.

Perceived group discrimination is generally expected to have a positive reactive effect and enhance political participation (Schildkraut 2005; Kranendonk et al. 2018). It is argued that immigrants experiencing group discrimination may seek to improve group status by engaging in collective action that affirms their group’s worth, thereby increasing the levels of political participation (Sanchez 2006; Spies et al. 2020). Exclusionary practices by political elites can strengthen in-group identities and consciousness among minorities and the perception of shared grievances intensifies group-based emotions, such as anger, which can move individuals to collective political mobilization, particularly when experiencing, attacks on sacred values and antireligious practices (Pérez 2015; Wald et al. 2005). Simultaneously, experiencing racial and ethnic discrimination leads to lower levels of trust and confidence in political institutions (Brehm and Rahn 1997).

However, it has also been theorized that perceptions of discrimination lead to lower levels of political engagement. Discrimination can activate or aggravate feelings of inferiority, insecurity, powerlessness, and depression (Schildkraut 2005; Spies et al. 2020). Depleted psychological resources correlate with lower turnout rates (Spies et al. 2020). Yet, this reinforces the idea that discrimination partly explains the evaluation gap, implying that discrimination feeds into a negative emotional relationship with the collective. By this logic, feelings of less belonging arise and thus there is less likelihood of a favorable evaluation of the political process (Leighley and Vedlitz 1999) and/or perceptions of lower levels of political representation (Junn and Masuoka 2008). For instance, Schildkraut (2005) found that in the US perceptions of discrimination are particularly damaging because they promote both behavioral and attitudinal alienation (non-voting, lack of political trust or representation) (see also Bird et al. 2011; Modood and Thompson 2022).

We thus expect perceived discrimination to impact the evaluation gap in one direction and the participation gap to impact it in the opposite.Footnote 2

H4

The negative evaluation gap is stronger when comparing citizens without an immigrant background to citizens with one who perceive strong group discrimination (4.1) or clear underrepresentation (4.2).

H5a

The negative electoral participation gap is stronger comparing citizens without an background to citizens with one who perceive strong group discrimination (5.1a) or clear underrepresentation (5.2a).

H5b

The negative electoral participation gap is stronger comparing citizens without an immigrant background to citizens with one who do not perceive strong group discrimination (5.1b) or clear underrepresentation (5.2b).

Based on the literature, it is not far-fetched to expect differences between groups of citizens with an immigrant background when it comes to political participation and evaluation, perceived discrimination, and representation. The same would go for explanatory-related factors such as networks and organizational structure; ethnic and religious identification; and sense of belonging (Portes and Zhou 1993; Lee 2008; Heath et al. 2013; Tillie 2004; De Rooij 2012; Fraga 2018). Groups with extended social networks or well-developed organizational structures—such as the Turkish community in the Netherlands and, for that matter, across Northern and Western Europe—are likely to be exposed to political conversations and recruitment and mobilization networks as they become embedded in religious–cultural communities. Similarly, organizational infrastructures, recruitment and party mobilization mechanisms among immigrants are stronger in neighborhoods with more inhabitants of an overlapping or culturally proximate immigrant background, but only for groups with high levels of groups-based resources (Vermeulen et al. 2020). These examples align with our expectation of variation across groups (Hypotheses 1), although theorizing per group goes beyond this article’s scope. Still, the data let us look at various communities, including thus far the understudied ones in the Netherlands. The discussion section reflects on our outcomes, generating theoretical insights on which shared and distinctive group categories might explain patterns across groups.

Data and methods

DPES and DEMES

The 2021 DPES (Dutch Parliamentary Election Study; Jacobs et al. 2021; Sipma et al. 2021a) and 2021 DEMES (Dutch Ethnic Minority Election Study; Lubbers et al. 2021; Sipma et al. 2021b) offer a unique set of data. These studies combine an oversampling of citizens with African, Asian, and Latin American backgrounds with data on various political attitudes and behaviors, including novel measurements of citizens’ feelings of descriptive and substantive representation. Moreover, DEMES includes items specifically on integration and inclusion.

The studies were undertaken around the 2021 Dutch parliamentary election. DPES data were collected via a random sample via the LISS infrastructure and via an original sample supplied by Statistics Netherlands (CBS) based on the Dutch Personal Records Database (BPR). DEMES was collected via an original sample among first- and second-generation “non-Western” immigrants who hold Dutch citizenship—and thus the right to vote in national elections—drawn from the BPR by CBS. “Non-Western” was the term used by the Dutch government to classify people who have immigrated from Africa, Asia (except Indonesia and Japan), and Latin America or whose parents have. Using this demarcation, we could include a wider range of backgrounds than most Dutch immigrant-based samples, which often focus on Dutch Caribbean, Moroccan, and Turkish groups, and contrast them to Dutch citizens without a migration background.Footnote 3

The Dutch case

The Netherlands presents a common political and electoral system in Western Europe. It is a parliamentary proportional representation system, albeit one characterized by a high degree of proportionality and many smaller parties (Spierings and Jacobs 2019), leading to a relatively high number of Muslim MPs (Aktürk and Katliarou 2021) as well as multiple ethnic-minority interest parties (Lubbers et al. 2022). Around one million Dutch citizens with an immigrant background are estimated to be Muslim (Huijnk et al. 2022) and a nearly equal number are estimated to be Christian (De Hart and Van Houwelingen 2018).

Immigration, discrimination, and the integration of Islam in society are salient issues in Dutch politics, similar to elsewhere in Western Europe, so it is unsurprising that voters with an immigrant background feel political exclusion (Vermeulen 2018) despite relatively strong numerical representation at the national level. Again, these experiences can differ across groups or communities, though self-reports of perceived discrimination and feeling it is hard to be accepted as “really Dutch” are considerable across groups. Among Dutch citizens with an immigrant background, approximately 4 in 10 indicate that they are discriminated against as a group and over half say it feels hard to be accepted as “really Dutch.” The latter is highest among the groups problematized most in the public discourse—Moroccan-Dutch and Turkish-Dutch (approximately 7 in 10)—and lowest among Caribbean-Dutch and Latin American-Dutch (approximately 3 and 4 in 10, respectively), with other African-Dutch and Asian-Dutch groups ranking in between (Geurts et al. 2022). Interestingly, Dutch citizens with Moroccan and Turkish backgrounds are more often numerically overrepresented in Dutch national parliament, which is not the case for groups classified as “other non-Western” (Van der Zwan et al. 2019); this could be partly ascribed to higher turnout and in-group voting among Dutch-Turkish communities who have the mobilizing capacity of a closer-knit network of Turkish organizations (Fennema and Tillie 1999; Vermeulen and Berger, 2008).

Data selection

The non-immigrant sample in DPES includes 3,331 respondents.Footnote 4 DEMES includes 735 respondents with an identifiable “non-Western” immigrant background. These numbers refer to the pre-election wave of the survey. Actual electoral participation and several core explanatory items were, however, included in the post-elections surveys. In them, respectively, 2847 and 394 of the initial respondents participated, giving most of our models a base sample of 3241 (80%) respondents. To account for potential selection bias, we controlled all models for confounding demographic factors.Footnote 5

Immigrant background

DEMES includes information on respondents’ birth country and that of their parents.Footnote 6 For origin country, respondents born abroad were classified according to their birth country; for those born in the Netherlands, birth country of their parents was used.Footnote 7 Next, we used these origin countries to distinguish as many groups as possible while still assuring reasonable statistical power. Categorizing and grouping countries to represented preferably at least 50 respondents in the pre-election survey sample this yielded eight meaningful categories: former Dutch Antilles (55), Morocco (105), Suriname (171), Turkey (128), other Middle East and North Africa (MENA) (96),Footnote 8 sub-Saharan Africa (62), other Asia (75), and other Latin America (43).

Dependent variables: electoral participation and evaluation of politics

Electoral participation was measured using the post-election recall variable, with the answering options “yes” or “no.” The participation rate on the survey item was 94% and of that 94%, 94% answered the question positively.Footnote 9

The evaluation of politics variable is based on eight items from the post-election survey tapping the subconcepts of political trust, external political efficacy, and satisfaction with the function of democracy. Item wordings, means, and standard deviations are given in Appendix C (Electronic supplementary material); mean scores and standard deviations are highly similar within each subconcept. Factor analyses show that each subset of items represents a singular underlying factor.Footnote 10 To optimize the number of cases, we calculated the means per subconcept if at least one indicator for it was present. An exploratory factor analyses including the three indices returned one factor (all loadings > 0.6) and Cronbach’s alpha was 0.670.Footnote 11 We therefore combined the three indices, scaled to run from 0 to 3, into one final index, based on the available means. Our dependent variable on the evaluation of politics covers 3,154 respondents (97%).

Core independent variables

Respondents’ embeddedness in religious communities was measured by two separate indicators: attendance of religious services (pre-election survey) and being a member of a religious organization (church, mosque, synagogue, temple, or other religious organization) (post-election survey). The first is included as an interval-like variable with five categories: 0 “(almost) never”; 1 “several times a year”; 2 “once a month”; 3 “2 or 3 times a month”; and 4 “once a week or more.” Membership is a dichotomous variable, with 1 indicating that a person is a member of such an organization.Footnote 12

Feelings of rejection by Dutch society were operationalized in terms of perceived social discrimination and perceived political underrepresentation of citizens with an immigrant background. Both were only included in DEMES. To assess impact on our outcome variables, we divided the immigrant background groups according to those who more frequently perceive exclusions and those who do so less. Including these groups separately and comparing them to respondents without an immigrant background let us assess whether the engagement gaps were stronger or mainly found among those not experiencing discrimination. Accordingly, we created indices for both subconcepts and subsequently dichotomized them by applying substantive and statistical arguments.

First, we dichotomized whether a respondent perceives discrimination of people with an immigrant background by Dutch citizens without an immigrant background. We did this by splitting the variable, placing “never,” “almost never,” and “now and then” on one side and “often” and “very often” on the other. This produced the most equal halves and the most meaningful distinction. Of the sample, we classified 43% as strongly perceiving discrimination.

Second, we used two novel questions on experienced representation, drawn from a small-scale pilot (see Peelen et al. 2022). Building on the seminal conceptualization spelled out by Pitkin (1967), we focused on the two dimensions most central to current debates. Experiencing descriptive representation was tapped by asking: “Do you think that there are too many or too few MPs with an immigration background currently? [far too few; too few; exactly the right number; too many; far too many].” and substantive representation by: “Do you think that the interests of people with an immigration background are sufficiently represented in the Dutch parliament? [far too little; too little; exactly as should be; too much; far too much].” For both items, “don’t know” was a preset answer option, which we considered to indicate at least that respondents did not feel that migrant-background respondent were underrepresented; this option was selected by, respectively, 32% and 23% of respondents, yielding a total of 380 respondents for inclusion.Footnote 13 Objectively, these two dimensions are interlinked though cannot be equated (Wängnerud 2009). However, in subjective terms—how people perceive the situation—they are strongly connected and part of the same discourse. Our coding concluded that as many as 85% of respondents fell in the same category on both dimensions.

For parsimony’s sake, the combined index served as the mainstay in our analyses. To create this index, the equivalent of feeling less than “neither too little, nor too much” represented on both items was coded 1 (46%) and the rest were coded 0 (54%). In addition, we reran the models with both dimensions separately to test results robustness and explore how elements of subjective representation might have different implications [see Appendix D (Electronic supplementary material) and results section].

Control variables

As control variables, we included characteristics that might influence political participation or evaluation and that might simultaneously interconnect with migration background (see Bird et al. 2011; De Vroome et al. 2013; Spies et al. 2020). Age was included as self-reported age, with 18 being set to equal 0. Education was measured by self-reported highest obtained degree, included as an interval-like variable running from 0 to 7. Sex was included based on self-reported sex, where 1 denotes “male.”Footnote 14 The surveys also include self-image of social class, broken down into four categories (working, middle, upper-middle, upper).Footnote 15 This was included as an interval-like variable, running from 0 to 3, along with its quadratic term, based on additional assessment of the linearity of the relationships.

Modeling

The evaluation of politics is a linear and normally distributed variable (skewness -0.281; kurtosis -1.030). We used standard OLS models to assess the explanatory power of potential evaluation gaps by including mediating variables (embeddedness in religious communities) or splitting the immigrant groups (discrimination).

Electoral participation is a dichotomous variable, thus suggesting that logistic regression would be most suitable. However, as convincingly argued by Mood (2010), simple model comparisons—for instance, when assessing mediating effects—can be problematic using logistic models, particularly, if there is skewed distributed. Therefore, we reverted to linear probability models (LPM) below. To ensure our main effect was correctly estimated in terms of significance and relative effect size, we also estimated logistic variants of the main models [Appendix B (Electronic supplementary material)].

Results

Varying engagement gaps

Table 1 shows results for assessing the engagement gap in the evaluation of politics and electoral participation, both in a general collapsed form and split by origin country. Models 1 and 3 [as well as all robustness tests in the appendices (Electronic supplementary material)] show the gaps expected by Hypotheses 1a and b: Citizens with an immigrant background are significantly less likely to vote in parliamentary elections and evaluate politics significantly more negatively than citizens without an immigrant background. While these overall gaps are very robust, they vary in degree and mere existence across groups; notably, the two gaps seem to be driven by different groups. More negative evaluation of politics is only consistently present for citizens of Moroccan and Turkish backgrounds. Given subsample sizes, it might be argued that we should also consider the size of the statistically insignificant coefficients, which shows negative but not significant gaps between citizens without an immigrant background and citizens with Surinamese, other Latin American and sub-Saharan African backgrounds. However, besides not being significant, these gaps are considerably smaller than for citizens with Moroccan and Turkish backgrounds. No substantive negative gap was found for groups with Antillean, other MENA and other Asian backgrounds.

Table 1 OLS regression assessing the ‘engagement gap’ between ethnic-majority and ethnic-minority citizens
Full size table

Thus, we can conclude that an evaluation gap exists, but it is foremost driven by a more negative evaluation among citizens of Turkish and Moroccan backgrounds. That said, we note that public and political debates give attention mostly to these two groups, who are not underrepresented, but rather overrepresented, in parliament. Our interpretation is that the negative evaluation might be fostered in part by the groups being politically visible and organized as their dissatisfaction gets channeled into politics.

For electoral participation—even when considering the variation in the robustness tests [Appendices A and B (Electronic supplementary material)]—we can conclude that the gap is rather consistently negative and significant for citizens with MENA (excluding Morocco), Latin American (excluding the Antilles and Suriname), and Asian (excluding Turkey) backgrounds. In other words, the electoral participation gap is most strongly present for groups with weaker ties to the Netherlands. These are also groups who are less well-organized and less visible in political and public debates. That we find particularly strong effects for these subsamples is noteworthy, these groups are overwhelmingly the smallest and most diverse in terms of origin country, which generally decreases the likelihood of finding statistically significant differences.

Altogether, we can draw four preliminary conclusions on the inclusion of a more diverse set of citizens with an immigrant background. First, using a general collapsed migrant/non-migrant approach is likely to obscure important differences between immigrant groups. Disambiguating them is necessary for better understanding how engagement gaps manifest themselves. Second, differences between groups are not the same across subconcepts of political engagement, indicating that studying them simultaneously, albeit separately, sheds more light on what is going on. Third, the differences in gaps do not seem random: those who more negatively evaluate politics are all larger immigrant groups who receive more attention in Dutch politics, media, and academia, while those showing lower voting likelihood mainly comprise smaller immigrant groups who are less salient in Dutch societal debates. Finally, our results thus suggest that focusing on the main, most salient or most commonly analyzed historical ethnic-minority groups in a country likely leads to an overestimation of the overall evaluation gap and an underestimation of the overall participation gap.

Participation in religious communities

Table 2’s results help answer to what extent embeddedness in religious communities explains political engagement gaps. First, we re-estimated the subgroup models from Table 1 concerning the cases for which religious attendance and organization membership were present; then we assessed whether including the religious embeddedness variables decreases or aggravates the gaps found.

Table 2 OLS regression assessing the role of integration in religious subcommunities to explain ‘the engagement gaps’
Full size table

By and large, Table 2 helps us conclude that embeddedness in a religious subcommunity fosters political engagement, as expected by Hypothesis 2a, which is a prerequisite for any factor mediating (or suppressing) the found gaps. While two religious variables are not multicollinear,Footnote 16 but dropping the membership variable from the electoral participation model does lead to a significant positive effect for attendance.

Crucially, comparing Model 2 to Model 1 and Model 4 to Model 3 hardly shows any change in the coefficients for an immigrant background. This suggests that embeddedness in religious communities is neither responsible for nor suppresses the electoral participation gap or the evaluation gap.

We are cautious, however, not to reject our expectations too quickly because religious communities serve different roles across ethnicities. The Dutch context is one of an ethnic majority living in a paradigm of secularization, similar to many European countries, but that does not hold equally for groups with an immigrant background whose religious infrastructure is often conflated with their more general ethnic network (Huijnk et al. 2022). Applying this to our modeling, the impact of religious community integration might differ between ethnic-minority and ethnic-majority members, which we assessed by interacting religious embeddedness with migration variables. On the evaluation of politics we did not find a differential effect, leading us to conclude that differences in religious embeddedness neither explains nor obscures a gap for this subconcept of political engagement.

For voting, the result is strikingly different as well as remarkably significant for the small-n subgroups. We visualized the results of these additional analyses in Fig. 1. With all other factors included being equal, the bars show to what extent voting likelihood differs between ethnic-minority and ethnic-majority voters and separately for religious organization members and non-members. Solid bars indicate that the difference is statistically significant. The light gray bars (mainly on the right side) clearly show that among religious organization members, there is no voting gap between ethnic-minority and ethnic-majority citizens.

Fig. 1
figure 1

Voting gap by membership of religious organization, per ethnic-minority group (full bars P < 0.05)

Full size image

However, for five of the eight groups, we find a significantly and substantially lower voting likelihood than among ethnic-majority members, represented by the black solid bars on the left. This leads to two more general observations. First, electoral participation gaps are only found among those who are not members of a religious organization, suggesting that such a membership helps close the gap among citizens with an immigrant background. Since we found no similar effect for evaluating politics and we observed that the impact of religious organization membership varies according to background, we surmise that citizens with an immigrant background internalize information and the norm of voting as a civic duty via these organizations. In other words, religious organizations serve to promulgate democratic values within immigrant communities in the Netherlands.

Second, despite a clear overall pattern, we found no voting gap for all immigrant groups. However, the split between the groups for which we did and did not find such a gap does not fall along lines that collectivize groups with backgrounds from Muslim-majority societies and communities. If we found this effect only or more strongly for Christian-based or Confucius-based groups, one might interpret support for arguments that Islam is incompatible with democracy; however, with strong effects for people with Moroccan and MENA backgrounds, this logic can be unequivocally refuted. These results underline the importance of not generalizing findings from one particular immigrant group (e.g., the largest) to immigrant populations in general. Differences in migration history and societal position seem to shapes the formation and perpetuation of political engagement gaps.

The role of exclusion

As a last step, we split the immigrant groups according to those who perceive significant discrimination and those who do not as well as between those who feel clearly underrepresented and those who do not.

For evaluation of politics, Models 1 and 2 (Table 3) show that the more negative evaluation among ethnic-minority citizens is only found for certain subsets: those with Moroccan, Surinamese, and Turkish backgrounds who perceive more discrimination against their group and feel strongly underrepresented; the same groups found in Table 1. For underrepresentation, the pattern seems less strong, but for groups with Moroccan and Surinamese backgrounds, we found relatively large coefficients with relatively low p-values (p = 0.102/p = 0.104); this is in line with the pattern for our discrimination variable. Moreover, by including the dimensions of representation separately, we found similar results for descriptive representation and for subjective substantive representation the effects were more pronounced and even more in line with those for discrimination [Appendix D (Electronic supplementary material)]. Signals of not being accepted or not belonging to society seem to push segments of communities with immigrant backgrounds affectively away from the political system and its actors, confirming Hypothesis 4 (both 4.1 and 4.2) for these groups. Notably, we mainly find this effect for the groups central to societal debates on migration and integration.

Table 3 OLS regression assessing the ‘engagement gap’ by discrimination and representation
Full size table

Models 3 and 4 show similar analyses for electoral participation, suggesting rather different processes. All of 10 statistically significant (p < 0.05) coefficients indicate lower electoral participation among groups with an immigrant background; however, nine of 10 are found among groups who do not perceive strong discrimination and particularly those who do not experience underrepresentation. So while those who experience discrimination and underrepresentation affectively distance themselves from politics in evaluative terms, they are not the ones who participate significantly less; it is those who experience fewer problems or less political exclusion that distance themselves behaviorally. In other words, perceived discrimination and underrepresentation mobilize voters with an immigrant background to vote and close the participation gap, as expected by Hypotheses 5.1b and 5.2b.

Conclusion

This article examined the gap in political engagement between Dutch citizens with and without an immigrant background. Most studies look at either political participation or political trust for all citizens with an immigrant background together or some specific groups, focusing on traditional demographics and socioeconomic explanations (De Vroom et al. 2013; Leighley and Vedlitz 1999; Ramakrishnan 2022; De Rooij 2012; Bird et al. 2011; Fraga 2018; Spies et al. 2020). First, we argued that considering political evaluation and electoral participation together and distinguishing different groups sheds light on the dynamics of political engagement gaps. Second, following some calls in the literature (Bird et al. 2011; De Vroome et al. 2013), we argued that to understand participation and evaluation gaps between citizens with and without an immigrant background, we need to theoretically and empirically apply inclusion and exclusion as an analytical lens (see Heath et al. 2013; Spies et al. 2020). We argue that to study Western Europe, we need to focus specifically on cultural inclusion and exclusion, which manifest as religious embeddedness and experienced exclusion or discrimination on the basis of immigrant or religious background (De Vroome et al. 2013). The 2021 Dutch election studies (DPES, DEMES) allowed us to assess variation in engagement gaps across multiple of Dutch immigrant communities, including smaller, less typical ones regularly overlooked in existing studies.

Our results illustrate that a dichotomous immigrant/non-immigrant approach is likely to obscure important differences, which are not the same across aspects of political engagement. And, to better understand engagement gaps, it is indeed necessary to focus on specific immigrant origins. Moreover, the differences in the political engagement gap do not seem to be random, but show a rather clear pattern: evaluation gaps are found for groups who get more attention, often negative, in Dutch society. Groups displaying negative participation gaps, however, represent smaller, often less discussed Dutch migrant groups. Thus, focusing on the in the literature most common ethnic-minority groups in a country likely leads to an overestimation of the evaluation gap and an underestimation of the participation gap.

Culture-related explanations seem important. Embeddedness in religious communities, both Christian and Muslim, turns out to be especially influential on the electoral participation of Dutch citizens with an immigrant background. Gaps are only found among those who are not members of religious organizations, suggesting that such membership helps close the participation gap. Since we found no similar effect for political evaluation and that the impact of religious organization membership also varies by background, the most likely mechanism at work is the transmission of information and the norm of voting as a civic duty.

Regarding impact of perceived group discrimination (see Schildkraut 2005; Spies et al. 2020), our study suggests that discrimination affectively pushes away some voters with an immigrant background vis-à-vis their positive evaluation of Dutch politics while simultaneously mobilizing them. This holds for perceived group discrimination as well as the novel concept of perceived political underrepresentation in descriptive and substantive terms. Particularly, experienced substantive representation seems to matter, even though the formulation is in general “immigrant background” terms rather than referring to specific origin countries or ethnic groups whom respondents identified with. These results suggest that a further unfolding of subjective representation in survey research might be fruitful for understanding the migration–political engagement link. In the Dutch relatively open political system it seems that groups that evaluate politics most negatively but who do participate, are also the groups who are mobilized by, and attracted to, new minority political parties (see Lubbers et al. 2022). For these citizens political participation is a promising strategy as these new parties openly represent their negative political evaluation in the Dutch parliament.

Our results show that members of highly visible and politicized groups, such as Muslims and people of color, evaluate politics more negatively. This notably holds for those who feel discriminated and underrepresented, and their frustration seems channeled through higher levels of voting turnout, as seen among those who experience group discrimination. Less visible minority groups display a different pattern: they might experience social discrimination and exclusion on an individual basis, though neither appears to affect their evaluation of Dutch politics nor mobilizes them. Future work on these groups specifically can shed more light on the dynamics at play.

It could be argued that the presence of engagement gaps for some groups, but not for others, is attributable to a relatively small sample sizes in DEMES, potentially leading to false negatives. While this possibility holds true in general, we found no gaps for all larger groups of citizens with an immigrant background and substantive effects were absent for several groups (discussed in the results section). More research with a larger subgroup sample would provide ample new opportunities for study, a springboard for which we hope this article provided. Future research with richer data would do well to include and separately analyze more groups, such as citizens of Chinese-Dutch and Indonesian-Dutch or Moluccan-Dutch backgrounds. Partly due to sampling definitions, these groups are largely overlooked even though their post-colonial relations are distinct and significant, recalling that our results suggest historical connections and processes of minoritization matter.

Finally, more in-depth analyses of the impact of discrimination are needed. This would entail delving into the more methodological question of how to include similar questions on discrimination and representation for the ethnic-majority population. Can we establish a measurement equivalence in that respect? This study explored whether it was possible to assess how discrimination is linked to the engagement gap. Our strategy facilitated new insights showing how perceived discrimination mobilizes but also politically detaches citizens, notably those from ethnic minorities. We conclude therefore that perceived discrimination thus both explains and obscures engagement gaps.