Perceived societal discrimination among immigrants across 17 European countries: extending the integration paradox? Open Access

This study contributes to the literature on immigrants’ perceived discrimination in two ways. First, while earlier work focused on personal or group-level discrimination, this paper develops and tests a theory of perceived societal discrimination, that is, immigrants’ perceptions of how prevalent different types of discrimination are in their society. Importantly, immigrants’ views of societal discrimination encompass ‘out-group empathy’, that is, beliefs about the frequency with which out-groups experience discrimination. Second, whereas existing literature relied on single-country designs, this study provides a cross-national examination, offering insights into the effects of country-level conditions on perceived discrimination. Based on survey data from 12,000 first- and second-generation immigrants in 17 European countries, results show differences between minority groups, with Muslim immigrants perceiving higher religious discrimination and black immigrants being more attuned to skin color discrimination in society. The study further reveals a positive association between immigrant integration and perceived societal discrimination. Evidence suggests that this ‘integration paradox’ holds for both Muslim and non-Muslim immigrants, black and non-black immigrants, and also for Muslim immigrants who wear traditional religious clothing and those who don't. Finally, the study supports the ‘Tocqueville paradox’, indicating that in countries with more extensive anti-discrimination laws, the relationship between integration and perceived societal discrimination is amplified.

As Western societies experience increasing ethnic diversity, understanding the dynamics of social cohesion and intergroup relations has become a prominent area of research (Drouhot and Nee, 2019). Studies consistently demonstrate that ethnic majority members often exhibit negative attitudes towards immigrant groups (Czaika and Di Lillo, 2018; Czymara, 2021), and discriminate against ethnic minorities across various domains such as the labor market (Zschirnt and Ruedin, 2016), housing market (Auspurg, Schneck, and Hinz, 2019), and daily interactions (Zhang, Gereke, and Baldassarri, 2022). Support for right-wing populism and anti-immigrant political parties is prevalent among substantial proportions of ethnic majority members (Berman, 2021).

However, a key dimension of interethnic cohesion remains unexplored in this perspective. It is important to account for the perceptions of exclusion and discrimination by ethnic minority groups (Lamont et al., 2018; Spörlein and Schlueter, 2020). Findings indicate that such perceptions have significant consequences. Immigrants who perceive higher levels of discrimination experience more mental health problems (Lewis, Cogburn, and Williams, 2015), identify less strongly with their host country (De Vroome, Verkuyten, and Martinovic, 2014; Esaiasson, Lajevardi, and Sohlberg, 2023), exhibit lower levels of trust in mainstream institutions (Esaiasson, Lajevardi, and Sohlberg, 2023; Maxwell, 2010; Röder and Mühlau, 2012), and are less likely to vote in local and general elections (Oskooii, 2020). Recognizing these perceptions is crucial for understanding the dynamics of intergroup relations and social cohesion in contemporary immigrant-receiving nations.

Research on the extent and determinants of perceived discrimination among immigrants is rapidly evolving in Europe (Diehl, Liebau, and Mühlau, 2021; Lindemann and Stolz, 2021; Lindemann, 2020; Schaeffer, 2019; Soehl, 2020; Tuppat and Gerhards, 2021) and the US (Dana et al., 2019; Gong, Xu, and Takeuchi, 2017; Marrow et al., 2022). Previous studies in this field have primarily focused on ‘personal-level’ or ‘group-level’ experiences with discrimination, investigating to what extent immigrants perceive that they themselves or other members of their in-group are discriminated against based on their origin. A key topic of discussion in this literature is the ‘integration paradox’, that is, the idea that immigrants who are more integrated into society perceive more personal and group-level discrimination (Lajevardi et al., 2020; Schaeffer and Kas, 2024; Steinmann, 2019; Verkuyten, 2016).

In this study, I elaborate on the emerging field on perceived discrimination among immigrant groups both theoretically and empirically. First, I study immigrants’ perceptions of societal discrimination, that is, people's perceptions of how prevalent different types of discrimination are in their society. People's views of societal discrimination also encompass what I term ‘out-group empathy’, that is, their beliefs about the frequency with which out-groups experience discrimination. Perceptions of societal discrimination are largely overlooked in the literature, which has focused on personal experiences of discrimination, as well as the perceived discrimination faced by their in-group.

In contemporary ethnically diverse societies, there can be different grounds of discrimination in society. In this paper, I examine three: ethnic origin, religion, and skin color. While the category of ‘ethnic origin’ applies to all migrant groups, discrimination based on ‘religion’ and ‘skin color’ apply to some minority groups, while not to others. By examining immigrants’ perceptions of religion and skin color discrimination in their society, I aim to steer the field in a new direction, as such perceptions about societal discrimination may relate to not only characteristics based on which immigrants themselves might be discriminated against (e.g., religion), but also types of discrimination that immigrant out-groups might face (e.g., skin color).

I develop a theory that explains immigrants’ beliefs about how prevalent various types of discrimination are in their society. Based on different mechanisms of belief formation, I argue that there should be a positive effect of integration on perceptions of societal discrimination. While extending the integration paradox in this way, I also argue that different minority groups hold different beliefs about which types of discrimination are prevalent in their society. I theorize about group differences in perceived societal discrimination, based on religion (Muslim vs. non-Muslim minorities) and race (black vs. non-black minorities), and about how these group characteristics moderate the integration paradox.

The empirical contribution of this paper is to provide a cross-national examination of immigrants’ perceived discrimination. A shortcoming of current work in this field is that most studies rely on single-country designs and/or focus on specific minority groups. Relatively much work on perceived discrimination among immigrants has been done on the Netherlands (Geurts et al., 2022; Van Doorn, Scheepers, and Dagevos, 2013; Van Maaren and Van de Rijt, 2020; Verkuyten, 2016) and Germany (Diehl, Liebau, and Mühlau, 2021; Lindemann, 2020; Steinmann, 2019; Tuppat and Gerhards, 2021), while less so on other European countries. Cross-national work on immigrants’ perceived discrimination is largely absent, or focuses on specific groups, such as Muslim minorities (Trittler, 2019; Yazdiha, 2019).

The lack of cross-national comparative work is an important shortcoming, because meta-analyses summarizing the state of the art in the literature reveal that the integration paradox is far from a universal phenomenon (Schaeffer and Kas, 2024). Specifically, there are some studies that find no support at all, while studies that do find support vary greatly in the strength of their relationships. The heterogeneity in findings across studies on the integration paradox could reflect differences in design, such as samples, measures, and models. But it could also be that the (strength of the) integration paradox systematically depends on country-level conditions. There is preliminary evidence supporting this idea (Schaeffer and Kas, 2024; Yazdiha, 2019), and in this paper I elaborate on these studies, offering new insights into potential scope conditions. I present evidence from the first cross-national study on the integration paradox, using survey data on 12,000 first- and second-generation immigrants in 17 European countries.

Central to the scientific literature on perceptions of discrimination among immigrants in Europe lies the following question: How does immigrants' integration into their host society influence their perceptions of personal and group-level discrimination? The concept of ‘integration’ in this context is often defined as indicating the degree of cultural similarity (‘cultural integration’), intergroup cohesion (‘social integration’), and similarity in realizing valued goals (‘structural integration’ or ‘economic integration’) between ethnic groups (Alba and Nee, 2003; Jonsson, Kalter, and Van Tubergen, 2018; Van Tubergen, 2020). Classical assimilation theory (Gordon, 1964) suggests that as immigrants integrate or ‘assimilate’ into the host society, their feelings of exclusion should diminish (Portes, Parker, and Cobas, 1980). The reasoning behind this hypothesis is that, over time, immigrants gradually become similar to the majority population of their host society. Gordon identified several dimensions of assimilation, emphasizing that these do not occur concurrently. He posited that the absence of discrimination, termed ‘behavior receptional assimilation’, is expected to ensue after immigrants have become more socially, culturally, and structurally integrated. Although Gordon's original model did not explicitly address ‘perceived discrimination’, it has been inferred that, aligned with the essence of this model, immigrants should report decreased feelings of exclusion as they integrate into the host society (Portes, Parker, and Cobas, 1980). Specifically, as immigrants become more integrated, they are likely to be treated more favorably by the ethnic majority. These positive encounters may then decrease immigrants’ perceptions of personal and group-level discrimination.

Conversely, scholars have argued that greater integration of immigrants might actually lead to heightened perceptions of personal and group-level discrimination (Diehl, Liebau, and Mühlau, 2021; Di Saint Pierre, Martinovic, and De Vroome, 2015; Steinmann, 2019). While different mechanisms underlying this hypothesis have been suggested in the literature, two are important in the context of the current paper: awareness and opportunities. First, integration makes immigrants more aware of being discriminated against, both personally and as a group. Second, integration increases immigrants' interaction with ethnic majority members, in settings such as workplaces or neighborhoods, potentially elevating the opportunity to be discriminated against. Empirically, the mechanisms often align. For example, key indicators of integration—such as having attained higher education or being fluent in the host-country language—capture heightened awareness of instances of discrimination, as well as more opportunities for intergroup encounters. Based on these mechanisms, the concept of the integration paradox has emerged, positing that immigrants who are more integrated in the host society tend to perceive greater discrimination against themselves as well as against the group to which they belong.

Empirical studies on the impact of immigrant integration on perceived discrimination have rapidly expanded. Studies have been done in the US (Lajevardi et al., 2020) and in several European countries, including Germany (Diehl, Liebau, and Mühlau, 2021; Lindemann, 2020; Steinmann, 2019; Tuppat and Gerhards, 2021), the Netherlands (De Vroome, Martinovic, and Verkuyten, 2014; Van Maaren and Van de Rijt, 2020), Belgium (Alanya et al., 2017), Switzerland (Lindemann and Stolz, 2021), France (Soehl, 2020), Spain (Flores, 2015), and Sweden (Esaiasson Lajevardi, and Sohlberg, 2023). A comprehensive meta-analysis by Schaeffer and Kas (2024) synthesizes the findings on the relationship between immigrant integration and perceived personal and group-level discrimination. Analyzing data from 42 studies, the meta-analysis finds no support for the classical assimilation theory. Instead, it establishes that higher educational attainment, a primary indicator of integration, is significantly and positively correlated with increased perceptions of discrimination (p < 0.001).

Despite the support for the integration paradox, the meta-analysis also uncovers evidence to the contrary, suggesting that the paradox is not a universally accepted phenomenon (Schaeffer and Kas, 2024). Firstly, the analysis underscores significant variability in study outcomes. While some studies refute the integration paradox, others that support it report varying strengths of the relationship between integration and perceived discrimination. Secondly, although the results for educational attainment align with the paradox, findings related to other indicators of integration, such as length of stay in the host country (p < 0.1), immigrant generational status (n.s.), and labor market success (n.s.), are less definitive. Lastly, the meta-analysis proposes that the integration paradox might predominantly pertain to immigrant groups with distinct, easily identifiable characteristics, such as those defined by race (skin color) or religion (distinct name, dress). In contrast, for immigrant groups that are less easily identifiable, the analysis found no substantial evidence supporting the integration paradox.

Against this background, I propose a theory of immigrants’ perceptions of societal discrimination. This theory incorporates the insights from the literature on the integration paradox, yet also adds new elements to this literature as the concept of study—societal discrimination—is different from personal and group-level discrimination. In the following, I outline the mechanisms and argue how perceptions of societal discrimination are affected by integration and differ between religious and racial minority groups. I then argue that the integration paradox is moderated by group membership and country-level conditions.

To begin, I argue that the integration paradox extends to perceptions of societal discrimination. A key micro-level assumption I make in this claim is the availability heuristic: when making judgments about how likely or widespread certain events or processes are, people tend to rely on information that is personally available to them (Schwarz et al., 1991). There are two types of information signals that immigrants receive: on the one hand about discrimination experienced personally or by other in-group members, and on the other hand about discrimination with which out-group members are confronted. I theorize that immigrants use both type of sources to make inferences about the prevalence of different types of discrimination in their society (e.g., religion, skin color, ethnic origin).

Based on the availability heuristic, I posit that as the integration process unfolds, immigrants update their beliefs about the frequency of the type of discrimination in society with which they are personally, or as a group, confronted. The literature on the integration paradox suggests that integration increases immigrants’ perceptions of personal and group-level discrimination. Thus, when immigrants are experiencing increased discrimination based on their religious identity, it should shape their perceptions of the prevalence of religious discrimination in their society. Hence, I assume that experiences of personal and group-level discrimination are thought to shape beliefs about societal discrimination related to the in-group.

Simultaneously, I argue that integration also increases out-group empathy, that is, immigrants’ awareness of discrimination faced by out-groups, who may be discriminated against on different grounds than their own group. This out-group empathy can develop through greater exposure to host-country news outlets and increasing interethnic friendships and contacts, such as in workplaces or neighborhoods. As integration unfolds, immigrants update their information of discrimination faced by out-groups. In turn, these personal sources of information are then used to make judgments about how widespread different types of discrimination are in their society.

In summary, I assume that integration leads to updating of information that is personally available on discrimination faced by in- and out-groups. This should result in increased perceptions of discrimination on different grounds—not only the grounds with which immigrants are themselves confronted. Therefore, the first hypothesis posits that the integration paradox should also be applicable to perceived societal discrimination:

While the integration process may heighten awareness of discrimination experienced by out-groups, I argue that immigrant groups continue to harbor distinct beliefs regarding the prevalence of various types of discrimination in society. I examine this notion by comparing different religious and racial groups. Large strands of the migration literature have identified religion, and particularly the boundary between Muslims and non-Muslim minorities, as a major social cleavage organizing a sense of ‘us’ versus ‘them’ in European countries of immigration (Adida, Laitin, and Valfort, 2010; Bansak, Hainmueller, and Hangartner, 2016; Drouhot and Nee, 2019; Foner and Alba, 2008). Muslim minority groups in Europe encounter more discrimination than non-Muslim groups (Blommaert, Coenders, and Van Tubergen, 2014; Di Stasio et al., 2021). Beyond religion, scholars have emphasized that social stratification patterns and subjective differentiations between ‘us’ and ‘them’ may exclude non-white minorities on racial grounds (Beaman, 2016; Flores, 2015; McAvay, 2018). Groups with distinct skin color, such as Sub-Saharan African immigrants, face pervasive discrimination in various aspects, including the labor market (Bartkoski et al., 2018; Thijssen et al., 2022). Importantly, some immigrants may belong to one minority category (e.g., race), while for others it may intersect. While I treat Muslim and black immigrants as distinct groups in my theoretical framework to highlight the unique sources of discrimination they face, my empirical analysis accounts for the overlap between these identities.

I hypothesize and test the idea that, despite the process of integration and updating beliefs about discrimination experienced by out-groups, Muslim and black immigrants have different beliefs about which types of discrimination are prevalent in their society. This difference, I propose, arises from asymmetries in information and sensitivity. Thus, Muslim immigrants are not only personally and collectively confronted with discrimination based on their religious distinctiveness but may also be more attuned to news and societal discussions about the unfair treatment of Muslim minority groups. Similarly, black immigrants, facing racial profiling and discrimination, might give more weight to news reports on skin color discrimination. Based on the availability heuristic, these sources of personal information—which differ between religious and racial minority groups—are then used to gauge the frequency of religious and racial discrimination in society. Hence, group membership may shape perceptions of the common types of discrimination in society, potentially leading to a certain lack of out-group empathy. Therefore, my second hypothesis is the following:

I also theorize that the positive effect of integration on perceived societal discrimination (i.e., the integration paradox) depends on group membership. I elaborate on Flores (2015), who argued that with increased integration, the emphasis shifts away from discrimination based on religion and ethnic origin, instead accentuating racial discrimination. This is because, according to Flores, two processes unfold at the same time: the amplification of discrimination awareness (i.e., the integration paradox) and the increasing cultural assimilation of immigrants with the majority group. As immigrants perceive themselves as culturally closer to the majority, other grounds of discrimination, particularly racial, gain prominence. Flores argued that this transition is especially pronounced among visible minority groups, such as those with darker skin, as race becomes the defining characteristic distinguishing them from the native population. Flores, studying perceived personal discrimination among immigrants in Spain, found support for this hypothesis. For example, with increasing length of residence in Spain, immigrants from Sub-Saharan Africa perceive less often personal discrimination due to religion or ethnic origin, while increasingly perceive racial discrimination.

Building on this idea I assume that, with increasing integration, immigrant groups develop stronger awareness of the grounds of exclusion that set them apart from other groups. While integration may lead to updated beliefs about discrimination experienced by out-groups, thereby increasing awareness of various societal discrimination types, I argue that integration particularly intensifies immigrants' awareness of their unique characteristics that constitute grounds for exclusion. Therefore, I anticipate that the impact of integration is especially significant for the characteristic that renders the group distinct. For black minority groups, this is obviously race, and for Muslim minorities, it is their religion. Muslim minority groups in Europe show little evidence of religious change (Drouhot, 2021; Soehl, 2017; Van De Pol and Van Tubergen, 2014), contrasting with a general trend of declining religiosity among non-Muslim immigrant groups (Kasselstrand and Mahmoudi, 2020; Molteni and Van Tubergen, 2022). Thus, even as Muslim migrants become more culturally similar (e.g., acquiring the language, values, norms), their religious identity continues to set them apart. Consequently, I do not foresee that Muslim minorities, as they integrate further, will diminish their emphasis on religious discrimination in favor of perceiving increased racial discrimination. Instead, I explore the notion that group distinctiveness, whether based on religion or skin color, becomes a more salient issue with increasing integration, leading to a divergence in societal discrimination perceptions among Muslim and black immigrant groups:

Finally, I theorize that the integration paradox depends on characteristics of the host country. One theoretical argument to expect a moderating effect of country conditions is based on the ‘Tocqueville paradox,’ which postulates that policies aimed at reducing unequal treatment may increase the salience of remaining inequalities (De Tocqueville, 1889). Countries that pay considerable attention to reduce discrimination, such as in the labor or housing market, may create a discourse in which ‘discrimination’ and the striving for equal opportunities are emphasized. Evidence suggests that individuals residing in countries with effective anti-discrimination legislation know more about their rights to equal treatment than those from countries with less effective policies (Ziller, 2014).

Consequently, anti-discrimination laws and policies may enhance cognitive awareness of persisting discriminatory practices, potentially reshaping immigrants' perceptions. I assume that this ‘discrimination discourse’—when prevalent in the host society—is picked up more quickly by the more-integrated immigrants, thereby amplifying the integration paradox. Preliminary evidence for this idea was found in two studies. Yazdiha (2019) examined Muslim minority perceptions of hostility towards Muslims in Britain, France, Germany, and Spain. The study reveals that in Britain and France—two countries with extensive anti-discrimination laws and policies—second-generation Muslim migrants perceive more hostility towards their in-group than first-generation Muslim migrants. By contrast, in Germany and Spain—countries with less-extensive anti-discrimination laws and policies than Britain and France—the study finds no difference between generations in their perceived hostility. Another study that finds support for the Tocqueville paradox is Schaeffer and Kas (2024), who report that, based on bivariate correlational meta-analysis, the integration paradox is more pronounced in countries with more extensive anti-discrimination laws and policies. While these studies focused on personal and group-level discrimination, theoretically I expect to see that it applies to perceived societal discrimination as well. Elaborating on these theoretical and empirical insights, I expect to see the following:

The data for this study are from the Second European Union Minorities and Discrimination Survey (EU-MIDIS II) and were collected during the period of 2015–16 (FRA, 2020). The survey encompassed 28 European Union countries and specifically targeted immigrants and descendants from Turkey, North Africa, Sub-Saharan Africa, and Asia, as well as recent immigrants from non-EU/EFTA countries. In addition, the survey included interviews with two non-migrant ethnic minority groups: Roma and individuals of Russian background. EU-MIDIS II comprised 41 surveys, each aimed at specific groups in individual countries, all employing the same questionnaire. The objective for each group was to gather nationally representative data. Interviews were conducted face to face by bilingual interviewers, and questionnaires were translated (FRA, 2017).¹

EU-MIDIS II reached a total net sample size of (N_i =) 25,515 interviews across the (N_j=) 28 EU member states. As the purpose of the current study is to study immigrants and their descendants, I excluded the two non-migrant ethnic minority groups (Roma and Russians) from the analyses (N_i = 9,366 and N_j = 9 omitted), as information such as immigrant generation is unavailable/irrelevant. I excluded Poland and Slovenia, as in these countries only ‘recent immigrants’ were surveyed (N_i = 833, N_j = 2) instead of the target groups. In some countries, one respondent per household participated in the survey, while in other countries one or two members were interviewed. To address these dissimilarities and avoid clustering of observations, I keep only one respondent per household (N_i = 2,657). I excluded respondents with missing information on the dependent variable (N_i = 371). After these selections, the sample for the analyses consists of 12,288 respondents in 17 countries: Austria, Belgium, Cyprus, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Malta, Netherlands, Portugal, Spain, Sweden, and the United Kingdom. A breakdown of the number of respondents per targeted group by country of residence is presented in the Supplemental Material (S1 Table).

The dependent variable in this study is perceived societal discrimination. Respondents were first introduced to the concept of ‘discrimination’ in the survey: ‘I would like to ask you a few questions about human rights. A basic right is to be treated equally. Still, some people might experience discrimination. By discrimination we mean when somebody is treated unfavourably compared with others because of their skin colour, age, sex, sexual orientation, disability, ethnic origin, religion or religious beliefs.’ Subsequently, respondents were asked the following question: ‘For each of the following types of discrimination, could you please tell me whether, in your opinion, it is very rare, fairly rare, fairly widespread, or very widespread in [COUNTRY]?’ Then, respondents were shown a card, at which they could indicate for discrimination on the basis of (a) skin color, (b) ethnic origin or immigrant background, and (c) religion or religious beliefs if it was in their opinion (1) ‘non-existent’, (2) ‘very rare’, (3) ‘fairly rare’, (4) ‘fairly widespread’, or (5) ‘very widespread’. The bivariate correlations between the three variables are the following: 0.61 (religion and skin color), 0.71 (religion and ethnic origin), and 0.82 (ethnic origin and skin color), and associations vary slightly across groups (Supplemental Material S2 Table). Table 1 presents the distributions of the dependent variables in the sample (unweighted).

I measure immigrant integration with three variables: (1) immigrant generation, (2) education level, and (3) proficiency in the destination language (‘L2 skills’). For various reasons, I focus on these measures. First, exactly these three measures are commonly employed in the literature on perceived personal and group-level discrimination—making comparisons of my findings to earlier work easier.² Second, in studying perceived discrimination with cross-sectional data, using more subjective measures of integration, such as subjective belonging, identification with the host country, intergroup attitudes, and norms/values, may be subject to endogeneity issues. While issues with causality also apply to the measures included in the present study, theoretically endogeneity issues seem less plausible. Third, the three measures that I use capture the broad spectrum of cultural, social, and structural integration. For example, higher educated migrants, as well as those who are second-generation migrants, have more similar values and norms to the mainstream culture, more interethnic ties, and a better position in the labor market. And proficiency in the second language is important to include for foreign-born immigrants, as it's critically associated with heterogeneity in cultural knowledge, intergroup ties, and labor market outcomes. All three measures, by affecting the broad spectrum of integration, capture the core mechanisms of awareness and opportunities.

The variable immigrant generation was derived from country of birth of the parents and the respondent. I treat those who were born in the host country but have at least one foreign-born parent as 2G immigrants. Within the foreign-born population, I differentiate between those who arrived before age 19 (‘1.5G’), and those who arrived after that age (‘1G’). L2 skills is based on self-assessment of the proficiency to speak in the language of the county. It is measured on a 6-point scale, ranging from (1) ‘no skills’ to (6) ‘mother tongue’. As there are multilingual countries in the dataset, respondents were asked to indicate their proficiency for each national language. I use the maximum proficiency level. Note that this self-assessment variable highly correlates with the interviewers’ assessment of respondents’ language proficiency (r = 0.75). Regarding education, I use the International Standard Classification of Education (ISCED) and distinguish between three categories: 0–2 (at most primary or lower secondary education), 3–5 (upper secondary, vocational, post-secondary, short cycle tertiary education), and 6–8 (tertiary education).

To identify Muslim immigrants, I rely on religious affiliation. Respondents were asked, ‘What is your religion?’, and I coded the answers into (1) Muslim and (0) other (Christian, other religion, no religion). To capture black immigrants, it would be ideal to have a self-assessed measure of skin color. As there are no such data provided in the survey, I rely on region of origin, similar to earlier work (Flores, 2015). Among the targeted origin groups in the survey, those from Sub-Saharan Africa are the most often dark-skinned, much more so than immigrants from other regions (Jablonski and Chaplin, 2000). Hence, I contrast this group with the other three groups that were targeted in the survey (i.e., those from Turkey, North Africa, and Asia). In the data, 61% of the black immigrants are non-Muslim, and 39% are Muslim. Among non-black immigrants, 13% are non-Muslim, and 87% are Muslim. Around 16% of the respondents are black Muslim minorities.

To measure anti-discrimination laws and policies in each country, I use data from the Migrant Integration Policy Index, MIPEX (Giacomo and Huddleston, 2020), which are commonly employed in migration research (Kanas and Steinmetz, 2021; Schaeffer and Kas, 2024). Specifically, I rely on country-specific ratings with respect to ‘definition and concepts’ (e.g., number of grounds covered by law), ‘fields of application’ (e.g., employment, housing), ‘enforcement’ (e.g., range of legal sanctions), and ‘equality policies’ (e.g., implementation of specialized agencies or positive action measures). The combined MIPEX of anti-discrimination policies ranges from 0 (no anti-discrimination laws and policies) to 100 (maximum anti-discrimination laws and policies). I use data from the year 2015, thus shortly before the survey was conducted.

I include control variables for age, gender, urbanization (three categories of population density), and country of residence (fixed effects). A variable capturing mixed marriage of the parents is included to better compare immigrant generations. This variable distinguishes between children of ethnically mixed parents (one native-born parent) and those with both parents born abroad.

The descriptive statistics of the (individual-level) independent and control variables are presented in the Supplemental Material (S3 Table).

I use ordinary least squares regression to model perceived societal discrimination. As there are few missing cases regarding the variables used (i.e., <0.32%), I deleted respondents with missing values in a listwise manner in the regression models.

I find that immigrant integration is positively and statistically significantly related to perceptions of societal discrimination (Table 2). Specifically, immigrants’ L2 skills are positively associated with increased ethnic (M1) and religious (M3) discrimination in society. With respect to immigrant generation, 2G report 0.4 higher perceived ethnic discrimination (M1), 0.4 higher skin color discrimination (M2), and 0.4 higher religious discrimination (M3) compared with 1G.³ I also find that higher levels of education are associated with increased perceptions of societal discrimination. Comparing individuals having ISCED 6–8 or higher compared with those with ISCED 0–2, the results reveal amplified perceptions of discrimination concerning ethnic origin (+0.2), skin color (+0.2), and religion (+0.3). Overall, I find that language skills, education level, and immigrant generation are positively related to perceived societal discrimination (see Supplemental Material S4 Table).

To gauge the practical significance of these effects, it is essential to consider the range from the perceived discrimination scale, which spans from 1 to 5. A typical individual characterized by higher educational attainment, second-generation immigrant status, and fluent L2 skills scores more than a full point higher on this scale compared with an immigrant who is lower educated, is foreign-born, and lacks proficiency in L2. These findings support the idea that the integration paradox can be extended to immigrants’ perceptions of societal discrimination (H1).

I argued that various mechanisms bring about this link between integration and perceived societal discrimination. While the data do not allow a close inspection of each of these pathways, I can explore one, namely that the effect of integration on perceived societal discrimination is partly driven by personal experiences with discrimination. In additional analyses, in which I include measures of perceived personal discrimination,⁴ I find that, in line with the theoretical model, there is a strong positive association between personal experiences and perceived societal discrimination when the perceived grounds are the same (Supplemental Material S11 Table).

For example, I find that when immigrants indicate that, in the past five years, they have been personally discriminated against based on their religion at work, they score significantly higher on their societal perceptions of religious discrimination (+0.455, p < 0.001), while not significantly changing their perceived levels of ethnic or skin color discrimination in society. Moreover, I find that the positive association between integration and perceived societal discrimination in society weakens when including personal experiences with discrimination. Overall, my additional analyses are in line with the idea that the integration paradox holds for personal discrimination (Supplemental Material S10 Table), and that, because of the availability heuristic, this partly drives the link between integration and perceived societal discrimination.

Regarding group differences, I hypothesized that Muslim and Sub-Saharan African migrants perceive heightened discrimination concerning the distinctive grounds that set them apart. In line with this, I find that Muslim immigrants perceive significantly higher levels of religious discrimination in their society than non-Muslim groups do (M3). When it comes to perceived ethnic and skin color discrimination in society, I do not find any difference between Muslim and non-Muslim minorities. Additionally, my analyses suggest that Sub-Saharan African immigrants do not exhibit significant differences in their perceptions of the prevalence of religious discrimination in their respective societies when compared with other immigrant groups. They are more likely than immigrants from Turkey to emphasize ethnic discrimination (M1a), but they do not differ from North African or Asian immigrants regarding that type of discrimination. They do, however, strongly differ in their perceptions of the prevalence of skin color discrimination in society: Sub-Saharan African immigrants score 0.41 higher than Turkish immigrants, 0.19 higher than North Africans, and 0.20 higher than Asians. These findings support H2.⁵

To check the sensitivity of my findings for H1 and H2 to modeling assumptions, I conducted several robustness analyses. First, as the dependent variable is a scale with ordered categories, I estimated ordered logit models. Second, the main analyses are weighted to account for survey design. Given the continuing discussion in statistics and econometrics about the pros and cons of using weights (Bollen et al., 2016), I probed the results by leaving weights out. Third, I included additional control variables (i.e., religiosity, unemployment, perceived economic hardship). Fourth, as an alternative to measuring education in ISCED categories, I use a measure of total years of schooling. I truncated this variable at 20 years (95% of the sample). Reassuringly, the conclusions for H1 and H2 remain the same (see Supplemental Material S5–S8 Tables).

I also postulated that group membership would moderate the integration paradox. Specifically, it was anticipated that the positive impact of integration on perceived religious discrimination in society would be more pronounced among Muslim immigrants compared with non-Muslim immigrants. Conversely, the positive effect of integration on perceived skin color discrimination in society was expected to be stronger among Sub-Saharan African immigrants than among other immigrant groups. To evaluate these predictions, I conducted analyses similar to those presented in Table 2, incorporating interaction effects between integration variables and the two distinct group variables.⁶

In a further exploration of the presumed impact of identifiability on the integration paradox, I investigated among Muslim minority groups whether the link between integration and perceived societal discrimination is influenced by wearing religious clothing (which makes religious minority status more easily identifiable). To gauge this, I relied on responses to the question, ‘Do you wear traditional or religious clothing when out in public that is different to the type of clothing typically worn in your country of residence?’ I created a binary variable, distinguishing Muslim respondents who answered ‘yes, always’ or ‘yes, sometimes’, (coded 1) from those who answered ‘no, never’ (0). To simplify the presentation of the results, I created a composite index of integration (based on immigrant generation, education, and language proficiency), which ranges from 0 (minimum) to 1 (maximum).⁷

It appears that, among Muslim immigrants, religious clothing does not moderate the relationship between integration and perceived societal discrimination (Table 3). Thus, my combined results suggest that the positive association between integration and perceived societal discrimination (integration paradox) is the same for Muslim and non-Muslim minorities, for black and non-black minority groups, and for Muslim migrants who wear religious clothing and those who don't. In other words, I find evidence that suggests that how easily minority groups are identifiable does not moderate the integration paradox. This result goes against H3.

To test whether the integration paradox systematically depends on country-level conditions, as predicted by H4, I apply meta-regression techniques. This method is recommended above multilevel models with random slopes and cross-level interactions when the number of level-2 cases (in this case countries) is small, that is, below 30 (Bryan and Jenkins, 2016; Liefbroer and Zoutewelle-Terovan, 2021). An additional advantage of this methodology is that it allows for a closer inspection and better visualization of model fit and country-level outliers.

In the first step, I estimated for each country the relationship between immigrant integration (using the same index as discussed above) and perceived societal discrimination. As the findings, discussed before, indicate strongly overlapping results for the relationship between immigrant integration and perceived ethnic, religious, and skin color discrimination, I combined these different grounds into an index of perceived societal discrimination. I estimated the coefficients for the link between the index of integration and the index of perceived societal discrimination for each country, while taking into account controls. In the second step, I then performed a meta-analysis with random effects using the metan and metareg commands in Stata 18 (Harbord and Higgins, 2008).

The analyses support the idea that countries with more comprehensive anti-discrimination policies (as indexed by MIPEX) indeed intensify the positive correlation between integration and perceived societal discrimination (b = 0.015, p = 0.017, one-tailed test). I arrive at the same conclusions when breaking down the data by the 27 groups that were sampled across the 17 countries using multilevel meta-analyses (Supplemental Material S4 and S5 Figures), and when I don't use survey weights (Supplemental Material S6 and S7 Figures).⁸ However, my analyses also highlight that this model does not uniformly apply across all countries. Notably, France, Italy, Ireland, and Portugal emerge as outliers in this context. Thus, while I find support for H4, my analyses also suggest that there are important deviations. This could imply that the model is misspecified and/or that these deviant cases are due to methodological issues (sampling, non-response, etc.).

Four major conclusions can be drawn from this study. First, the results support the idea that the integration paradox applies to perceived societal discrimination. I find a positive and statistically significant association between all three examined integration indicators (second-language skills, immigrant generation, and education) and heightened perceptions of societal discrimination. This pattern holds true across discrimination based on ethnic origin, religion, and skin color. Additional analyses reveal that this integration paradox is partly driven by the positive association between integration, personal experiences with discrimination, and perceptions of societal discrimination. Theoretically, this aligns with the notion that greater integration prompts immigrants to update their beliefs about societal discrimination through on the one hand personal and in-group experiences with discrimination and on the other hand intergroup contact and out-group empathy.

Second, my analysis shows that group membership significantly influences immigrants' perceptions of societal discrimination. Immigrants predominantly perceive societal discrimination grounded in the distinct characteristics that set their group apart. Muslim immigrants, for instance, perceive higher levels of religious discrimination in society, while black immigrants are more attuned to skin color discrimination. These findings emphasize the importance of acknowledging the unique experiences of different immigrant groups when studying perceived societal discrimination.

Third, I find evidence that the integration paradox, when studying societal discrimination, holds across religious and racial groups. Increasing integration is associated with a stronger emphasis of ethnic, religious, and skin color discrimination across groups in the same way. I find no evidence that the integration paradox is moderated by identifiability. Interestingly, this pattern deviates from earlier work on personal and in-group discrimination, which suggests that the integration paradox only holds among easily identifiable minority groups. In addition, my findings also do not suggest shifting boundaries, such that increased integration merely changes the grounds of discrimination people perceive rather than discrimination levels overall. Instead, my observations suggest that increasing integration heightens awareness of not only the discrimination directly affecting the immigrants' own group but also the discrimination faced by other out-groups. More generally, these findings suggest that perceived societal discrimination—and the dimension of out-group empathy it encompasses—is not merely driven by perceptions of personal and in-group discrimination, and therefore deserves to be studied more thoroughly.

Fourth, despite the overarching evidence for the integration paradox, I also find that, when analyzed per country, there appears to be substantial cross-national variation. My findings reveal that in countries with more comprehensive anti-discrimination laws and policies, the positive relationship between immigrant integration and perceived discrimination strengthens. This supports the ‘Tocqueville paradox’, that is, the notion that policies aimed at reducing unequal treatment may increase the salience of remaining inequalities. Anti-discrimination laws and policies may make cognitive awareness of the remaining discrimination easier and could thereby alter immigrants’ perceptions. Discrimination discourses, when more prevalent in the host society, may be picked up more quickly by more-integrated immigrants, thereby amplifying the link between integration and perceptions of discrimination in society. This finding affirms earlier tests of this paradox concerning personal and group-level discrimination (Schaeffer and Kas, 2024; Yazdiha, 2019) and provides valuable insight into its scope conditions.

However, these conclusions should be contextualized within the limitations of this study, which opens avenues for future research. First, the study relies on cross-sectional data, making it challenging to establish causality. I cannot conclude whether immigrant integration leads to increased perceptions of societal discrimination, and therefore longitudinal and experimental studies are encouraged. Second, while I focused on three measures of integration (second-language skills, immigrant generation, and education), I did not explore other aspects of integration partly because of potential endogeneity issues. Third, stronger tests of the hypothesis on identifiability are needed. In particular, the data did not include ‘non-identifiable’ European immigrant groups, which would be a better contrast to the more easily identifiable groups included in the survey. Moreover, I could not include an individual-level measure of skin color. Fourth, follow-up work is encouraged to test the mechanisms underlying the hypotheses more directly, as in this study I could examine only the presumed role of perceived personal discrimination.

Further comparative research is essential to understand the contextual variability in the relationship between integration and perceived discrimination. While my meta-analysis indicates that anti-discrimination policies moderate this relationship, country anomalies warrant a deeper understanding of potentially different underlying processes. On the one hand, follow-up research could take advantage of within-country changes in anti-discrimination policies in the last decade. As such policies have become more prevalent in Europe (Giacomo and Huddleston, 2020), one would expect to see that the integration paradox has become more widespread as well. On the other hand, expanding the comparative scope beyond Europe and the US (Lajevardi et al., 2020; Marrow et al., 2022) is encouraged to include more country variation, offering a broader perspective on the integration paradox.

I'd like to thank Lucas Drouhot as well as the editor and reviewers from European Societies for their feedback on an earlier version of this paper.

The author used ChatGPT solely for language editing. No content, data analysis, or scientific conclusions were generated by the AI. The author is fully responsible for the content of this manuscript.

The data underlying this article are available in GESIS at https://doi.org/10.4232/1.13514. The computer codes for data preparation and analyses are available at https://osf.io/c2p97/.

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

The study is approved by the Ethics Committee of the Faculty of Social and Behavioural Sciences of Utrecht University (number 23-0007).

Supplemental data for this article can be accessed online at https://doi.org/10.1162/euso_a_00004.