Introduction

Recent research shows that better educated and structurally integrated immigrants do not articulate more positive attitudes toward the native-born and ethnic majority than immigrants who have lower levels of educational attainment (de Vroome et al., 2014; Teije et al., 2013; Tolsma et al., 2012; Verkuyten, 2016). These seemingly counterintuitive findings have been described as evidence of an “integration paradox” (Schaeffer & Kas, 2023) and are attributed to higher levels of perceived discrimination among better-educated immigrants (van Doorn et al., 2013), which negatively affects how they feel about the members of the host society, as well as their sense of belonging (Geurts et al., 2020).

While these empirical findings have important implications for theories of immigrant integration, they stand in stark contrast with theories of intergroup relations. For instance, they contradict intergroup contact theory (Allport, 1954); insofar as better integrated, highly educated immigrants have greater contact with the native/ethnic majority (Martinovic et al., 2009), they should hold more positive attitudes toward them. Importantly, these findings also challenge the strong theoretical expectation that higher levels of education generate more positive intergroup attitudes (Hainmueller & Hiscox, 2007; Hello et al., 2006; Hjerm, 2001; Hooghe et al., 2013). Indeed, the inverse relationship between education and prejudice is arguably the most consistent empirical finding in studies of inter-ethnic relations (Vogt, 1997) and anti-immigrant sentiment (Ceobanu & Escandell, 2010). This strong correlation is most noticeable when discerning between individuals with a tertiary degree and those without one (Hainmueller & Hiscox, 2007).

How can we understand these divergent empirical results? Previous research on prejudice has primarily focused on the relationship between levels of educational attainment among native-born, ethnic, and/or racial majorities and their attitudes toward ethnic minorities and immigrants. Although studies on the integration paradox examine the association between educational attainment of ethnic minorities and their attitudes toward the native-born and ethnic majority, this literature has neglected the relationship between immigrants’ level of education and their attitudes toward other ethnic minority groups, and how they develop over time. Thus, it remains unclear the extent to which this educational effect is “universal” i.e., whether it is also present among immigrants and ethnic minorities.

The presence of an integration paradox for highly educated immigrants is worthy of attention, since they would be the most likely to pass exams measuring language skills and knowledge of the host society. This paradox has not been investigated systematically in other countries (Verkuyten, 2016), thus, the Netherlands provides an opportunity to further investigate this by building on previous literature, and contributing by analyzing to what extent immigrants’ level of education from their country of origin shapes their attitudes toward the native Dutch and other ethnic minority groups in a longitudinal perspective.

To do this, I rely on data from four waves of the ‘New Immigrants Survey Netherlands’ (NIS2NL), collected in the years 2013–2018 for recent immigrants to the Netherlands from Spain, Poland, Bulgaria, and Turkey. The Netherlands, like most Western democracies, has seen a considerable change in its ethnic composition in past decades, with immigrants making up a considerable share of EU residents (Van Der Zwan et al., 2017). As of 2019, immigrants in the Netherlands, including second generation immigrants with at least one parent born abroad, made up 23.6% of the population: 12.5% first-generation and 11.1% second-generation (Statistics Netherlands, 2019). This makes inter-minority attitudes something important to study, which is often neglected in migration and integration research.

In this study, I investigate the role of education from the country of origin on recent immigrants’ attitudes toward both the ethnic majority and other ethnic minorities in the Netherlands. First, I investigate whether the relationship between education and outgroup attitudes differs depending on whether the outgroup is the native majority or other ethnic minorities. Second, I examine how attitudes toward outgroups change over time depending on different levels of educational attainment. By doing this, I can speak to what extent this relationship holds among immigrants, and if it varies depending on the ethnicity of the outgroup. By examining if education matters for immigrants’ attitudes toward different ethnic outgroups in the host society, I provide greater insight into the universality of the educational effect.

Theory and Hypotheses

Integration Paradox

Research on the integration paradox postulates that there is a negative relationship between structural integration and level of education, and positive attitudes toward natives and the host society. One possible explanation for the integration paradox is the “theory of exposure,” (van Doorn et al., 2013) which suggests that highly educated immigrants experience more discrimination since they consume more media and have more contact with natives in the labor market and associations, therefore being more exposed to discrimination. A second argument is that highly educated immigrants hold more sophisticated cognitive abilities (Bobo & Licari, 1989; Wodtke, 2012) and are therefore more aware of discrimination in the destination country and their lower status as immigrants. Moreover, highly educated immigrants develop higher expectations which are not always met (Verkuyten, 2016; see also Thijssen et al., 2021), thus leading to a feeling of relative deprivation when comparing themselves to the group that they most often have contact with, i.e. natives with similar levels of education.

Studies in the Netherlands have shown that although highly educated minorities have more opportunities for contact with natives, they do not necessarily have more positive contact and meet less discrimination (Tolsma et al., 2012). In other words, minorities with higher levels of education are more likely to experience interpersonal contact with members of the ethnic majority, but these interactions lead minorities to view the majority less positively. Moreover, the relationship between minorities’ perceived discrimination/less societal acceptance and their attitudes toward the native population and the host society is strongest among those with higher levels of education (de Vroome et al., 2014; Teije et al., 2013). This also seems to be the case in Germany (Steinmann, 2019; see also Tuppat & Gerhards, 2021), implying that this phenomenon is not endemic to the Netherlands. In addition, research has shown that a longer duration in the destination country is associated with more discrimination (McGinnity & Gijsberts, 2018), which suggests that the attitudes of immigrants with the most education should also worsen over time.

The integration paradox has also been found to exist among smaller immigrant groups in the Netherlands (Geurts et al., 2020; van Doorn et al., 2013), suggesting similar integration processes (or lack thereof) among various immigrant groups. It is therefore not specific to the four largest immigrant groups, i.e., Turks, Moroccans, Surinamese, and Antilleans. Considering this, higher levels of education among immigrants should not result in more positive attitudes toward the ethnic majority, although this may differ toward other ethnic minorities. This finding, however, has failed to be replicated among adolescents in the Netherlands, indicating that the integration paradox is limited to migrants who do not grow up in the host society, or that young immigrants’ attitudes toward the ethnic majority drastically change after high school (van Maaren & van de Rijt 2020).

The absence of the relationship consistently found in studies of natives’ attitudes toward immigrants and ethnic/racial minorities challenges the notion that more education always means more positive attitudes toward outgroups. This is key since more education is often regarded as a way to integrate immigrants and improve their commitment and identification with the host society. And yet, there is evidence that higher levels of education among the immigrant population do not necessarily translate into more positive attitudes toward the native ethnic majority, mainly due to perceived discrimination, which likely increases over time (McGinnity & Gijsberts, 2018). Therefore, I hypothesize that due to the integration paradox:

  • H1a: Higher levels of educational attainment are not associated with more positive attitudes toward the ethnic majority in a destination country.

  • H1b: Attitudes toward the ethnic majority become less positive over time.

Education and outgroup attitudes

Research that has focused on the effect of education and prejudice (e.g., Creighton et al., 2023; Hello et al., 2002, 2006; Meeusen et al., 2013; Mulders & van Tubergen, 2022; Velásquez & Eger, 2022) has largely looked at the inverse relationship between education and prejudice among ethnic and racial majorities and native-born. Meanwhile, empirical studies on the effect of education on immigrants’ attitudes toward natives and toward other immigrants and minorities is scarce (Roth & Kim, 2013), leaving the question open as to whether this robust statistical relationship also exists among immigrants.

Yet, previous studies that have investigated inter-minority attitudes have not focused on education. In the United States, Philip et al. (2010) looked at the differences in attitudes toward African-Americans among Indian immigrants and Indian-Americans, and showed that Indian immigrants hold more negative attitudes toward African-Americans, due to lack of contact, as well as beliefs about social hierarchies, inter alia. Despite this, ethnic minority groups at times may share their minority position in society and the feeling of being “in the same boat,” which could trigger a common disadvantaged minority ingroup identity that results in more positive attitudes toward other minorities (Craig & Richeson, 2012).

In a study of 23 European countries, Van Der Zwan et al. (2017) found that perceived migrant threat among immigrants was stronger among those who had lived longer in the destination country, and the more education a migrant had, the weaker their threat perceptions were toward other migrants. In Belgium, Meeusen et al. (2019) showed that Belgians of Turkish and Moroccan descent were not exempt from having negative attitudes toward immigrants in general, and these were more negative when it specifically concerned Eastern European immigrants, with a slight majority believing Eastern Europeans take their jobs and ruin the reputation of other immigrants. In a study among secondary school students in the Netherlands, Verkuyten et al. (1996) found that students with Turkish, Moroccan and Surinamese background considered the Dutch the most preferred outgroup, and that Turks and Moroccans had more favorable attitudes toward each other than to the Surinamese. Similarly, Hindriks et al. (2014) found that more social distance was reported by Turks and Moroccans toward Surinamese/Antilleans than toward each other.

These studies suggest two important things: First, similar groups are evaluated more positively (McPherson et al., 2001); Second, the ethnic majority population (at least in the Netherlands) is the most positively viewed outgroup by ethnic minorities, after their own group. Yet, these studies have not particularly focused on education and its effects on inter-minority attitudes, and on attitudes toward the ethnic majority. Research has also not investigated whether these attitudes change over time, despite existing evidence that the effects of education may vary across groups. For instance, Hindriks et al., (2014) found that higher levels of education are associated with less social distance toward other minority groups but not toward the native Dutch. In fact, this association is reversed toward the native Dutch, albeit the effect size is rather small and weak. In contrast, Meeusen et al. (2019) found that Belgian Turks and Moroccans show no significant differences in their anti-immigrant attitudes by level of education. The authors speculate that this is likely due to the majority of respondents in the sample being in low socio-economic positions.

To summarize, the few studies on inter-minority attitudes have not focused on the effect of education on these attitudes, while evidence suggests it may operate differently toward other minority groups than toward the ethnic majority. Still, education is consistently shown to be a strong predictor of pro-immigrant attitudes (Ceobanu & Escandell, 2010; Hainmueller & Hiscox, 2007; Hello et al., 2002). Therefore, due to strong theoretical expectations on the educational effect on attitudes toward immigrants, I hypothesize:

  • H2a: Higher levels of educational attainment are associated with more positive attitudes toward other ethnic minority groups in a destination country.

In addition, since there are no strong theoretical expectations for inter-minority attitudes to change over time by level of education, as there is toward the ethnic majority, I hypothesize:

  • H2b: Attitudes toward other ethnic minority groups are stable over time.

Data

Data come from the ‘New Immigrants Survey Netherlands’ (NIS2NL) for the years 2013–2018. This is a survey designed to analyze early integration processes of recently arrived immigrants to the Netherlands from Spain, Poland, Bulgaria, and Turkey (Lubbers et al., 2018). Information for the first wave was collected in November 2013 and March 2014 for migrants who had registered in a Dutch municipality less than eighteen months before September 2013. The mean response rate was 32.3% leading to a total of 4,808 respondents in the first wave. In the second, third and fourth waves, respondents were approached in March and May 2015; September 2016; and January 2018, respectively.

In the original dataset, 4,808 individuals were interviewed. However due to missing values mainly on the dependent variables, the sample includes a total 4,323 individuals: 1,254 (29%) Spaniards; 1,622 (37.5%) Poles; 687 (15.9%) Bulgarians, and 760 (17.6%) Turks. The number of respondents is 4,323 in wave 1; 2,069 in wave 2; 1,230 in wave 3, and 901 in the last wave. In order to test hypotheses H1a and H2a, I only use the first wave since the highest level of education is from the immigrants’ country of origin and therefore time-invariant in the analyses. The reason for not using highest level of education attained in the Netherlands is due to data limitations that do not capture adequate educational attainment in the destination country as well as a high attrition rate. The dropout between waves can be selective to a certain degree, due to the nature of immigrants’ mobility and wishes to return to their country of origin or migrate to other countries. While testing hypotheses H2a and H2b, I complement the analyses with a balanced sample (Table A7 in Appendix) with respondents who participated in all waves, which can better represent the migrant population that eventually stays in the country.

Dependent Variables

The dependent variables are based on “feeling thermometer” questions asking respondents how they feel toward different ethnic groups: native Dutch, Spaniards, Poles, Bulgarians, and Turks. Feeling thermometer questions are a reliable way to measure attitudes (Alwin, 1997) and are commonly used in ethnic studies, including research in the Netherlands (Coenders et al., 2008; de Vroome et al., 2014; Geurts et al., 2020; Velthuis et al., 2021; Weber, 2022). Respondents are asked to imagine a thermometer and can specify a temperature between 0 to 100 with 10-degree intervals representing how they feel about a given group. In the original dataset, the responses are coded on a 0–10 scale with 0 degrees representing very negative attitudes, and 100 degrees (coded as 10) meaning they feel very positively toward the group. As robustness checks, I also create two alternative dependent variables which are presented in Table A6 in the Appendix. For the first alternative dependent variable, I subtract ingroup attitudes from attitudes toward the ethnic majority. Similarly for the second one, I subtract ingroup attitudes from the average attitudes toward other ethnic minorities. By doing so, I am able to “isolate changes in the relative views of the targeted group and shifts in a willingness to apply stereotypes generally are thus deliberately removed” (Hopkins & Washington, 2020: 8).

Main Independent Variable

The main independent variable is highest level of educational attainment in the country of origin. Many of the respondents in the sample have not pursued higher education in the Netherlands and graduated with a degree (around 92%), and a vast majority of those who did pursue higher education already had a university degree. To test my hypotheses, I use a three-category version of education: 0 “Less than Secondary” 1 “Secondary” 2 “Tertiary.” This three-category variable was constructed from a 9-category variable ranging from “less than primary” to “Doctoral” education. This was done due to very low numbers in some categories and even zero for others (e.g., no Bulgarians with a master’s degree, or Turks with a doctoral degree).

Controls

In order to isolate education’s effect from other SES indicators, I use immigrants’ employment status (main activity) in the Netherlands: Unemployed; Employed; Non-employed; In Education. The first category captures individuals who are currently unemployed. The second category consists of individuals who are currently working, including those who are temporarily on parental leave. The third category ‘non-employed’ consists of individuals who are long term sick or disabled, retired, or looking after the home/children as their main activity and are not searching for a job. The fourth category refers to individuals who are currently enrolled in any type of education. Due to a substantial number of missing values for income (N≈700), this control is not added in the main analyses, but in additional analyses in the Appendix (see models in Tables A1-A7).

As basic demographic controls, I include the continuous measure of age, and a binary measure of sex. Other relevant controls included are perceived group discrimination and contact with native Dutch. Perceived group discrimination is captured by answers to the question: “Some say that people from [country of origin] are being discriminated against in the Netherlands. How often do you think [country of origin] people are discriminated against in the Netherlands?” I then constructed a 4-category variable: Never/Almost Never; Sometimes; Often/Very Often; Do not know. Contact with native DutchFootnote 1 consists of 6-categories: Never; Less often; Several times a year; A few times a month; Several times a week; Every day, answering to the question “How often do you spend time with Dutch people in your free time?”.

Estimation

I restrict the sample to exclude ingroup attitudes in the models. In this manner, the number of respondents becomes 3,069 for attitudes toward Spaniards; Poles 2,071; Bulgarians 3,636; and Turks 3,563. In additional analyses using alternative dependent variables (Table A6 in the Appendix), the number of respondents is 4,323 for both attitudes toward the ethnic majority, and toward ethnic minorities.

Methods

To test the hypotheses, I analyze data with ordinary least squares (OLS) regressions for hypotheses H1a and H2a due to the time-invariant nature of the key independent variable. In order to test H1b and H2b I use mixed, multilevel repeated measurement models. These are hierarchical models, with observations (time) nested in individuals. This approach estimates fixed-effects and two random-effects, corresponding to the between-individual (intercepts) and within-individual (residuals) variances. The generic model using a single equation looks like this:

$${y}_{ij}={\beta }_{0}+{\beta }_{1}{x}_{1}+\dots {\beta }_{n}{x}_{n}+{u}_{j}+{e}_{ij}$$

Where \({y}_{ij}\) is the dependent variable measured for i th level-1 unit (observations) nested within the j th level-2 unit (individuals). The fixed-part of the model consists of the intercept \({\beta }_{0}\) and covariates \({\beta }_{1}{x}_{1}\)+… \({\beta }_{n}{x}_{n}\). The random-part consists of \({u}_{j}\), the residual variance of the level-2 units (individuals) and \({e}_{ij}\), the residual variance of the level-1 units (observations).

This can also be written as a set of hierarchical equations:

$$\boldsymbol L\boldsymbol e\boldsymbol v\boldsymbol e\boldsymbol l\boldsymbol\;\mathbf1\boldsymbol\;\boldsymbol e\boldsymbol x\boldsymbol p\boldsymbol r\boldsymbol e\boldsymbol s\boldsymbol s\boldsymbol i\boldsymbol o\boldsymbol n:y_{ij}=\beta_0+\beta_1x_1+e_{ij}$$
$$\begin{array}{cc}\boldsymbol L\boldsymbol e\boldsymbol v\boldsymbol e\boldsymbol l\boldsymbol\;\mathbf2\boldsymbol\;\boldsymbol e\boldsymbol x\boldsymbol p\boldsymbol r\boldsymbol e\boldsymbol s\boldsymbol s\boldsymbol i\boldsymbol o\boldsymbol n\boldsymbol s:&\beta_{0j}=\gamma_{00}+\gamma_{01}w_j+u_{0j}\\&\beta_{1j}=\gamma_{10}+\gamma_{11}w_j+u_{1j}\end{array}$$

With \({\gamma }_{00}\) referring to the overall intercept, \({\gamma }_{01}\) the overall regression coefficient, \({w}_{j}\) a level-2 predictor (individual), \({u}_{0j}\) the random error component for the deviation of individuals’ intercepts from the overall intercept, \({\gamma }_{10}\) the overall regression coefficient between the dependent variable and level-1 predictor, \({\gamma }_{11}\) the regression coefficient for covariate \({w}_{j}\) and \({u}_{1j}\) the error component for the slope. A random slope for time is also included in the analyses for purposes of cross-level interactions (Heisig & Schaeffer, 2019).

Results

Descriptive statistics

The descriptive statistics for this study are presented in Table 1. The average response to the first dependent variable capturing attitudes toward the ethnic majority is 7.24 over the four waves. The average attitudes toward ethnic minorities (excluding ingroup attitudes) are Spaniards (5.44), Poles (5.29), Bulgarians (4.58), and Turks (4.46). About 54% of the sample has tertiary education, 30% secondary education, and 16% less than secondary education. Figure 1 below shows a fine-grained version of educational attainment in the sample by immigrant group. The majority (62%) are employed, 15% are unemployed, 8% are “non-employed,” and 15% are enrolled in some type of education. There are more women (55%) than men (45%), and the average age of respondents is 32 years old. Most respondents in the sample come from Poland (37%), followed by Spaniards (29%), Turks (18%), and Bulgarians (16%).

Table 1 Descriptive statistics
Full size table
Fig. 1
figure 1

Educational attainment (%) in the sample within each immigrant group at wave 1

Full size image

Figure 1 shows the distribution of educational levels by immigrant group. The vast majority of Spaniards in the sample have tertiary education (81.2%). 45.8% of Poles have tertiary education; 33.8% of Bulgarians; and 47% Turks. Respondents with tertiary education form the largest educational category for each immigrant group, except for Bulgarians who mostly have secondary education (58.2%) as their highest level of completed education in their country of origin.

Figure 2 illustrates mean attitudes by country of origin, including attitudes toward their own ethnic group, for comparison purposes. It is noteworthy that all immigrant groups regard the native Dutch more positively than other ethnic groups, even their own, except for respondents from Spain, who regard their own group the most favorably. The next most favored group is the ingroup, followed by Spaniards, Poles, Bulgarians, and Turks.

Fig. 2
figure 2

Attitudes toward ethnic outgroups and ingroup by country of origin

Full size image

Empirical Analyses

In Table 2, I report results from the test of hypotheses H1a in Models 1–2 and H2a in Models 3–10. I begin with model 1, including only the trichotomous education variable. Results show support for H1a in that having higher levels of education from the country of origin are not associated with more positive attitudes toward the ethnic majority. These analyses are also corroborated in Table A1 in the Appendix, where I analyze separately each immigrant group’s attitudes toward the ethnic majority. In model 2, controls are added showing that being employed, non-employed or in education does not seem to have an effect on attitudes toward the ethnic majority when compared to those who are unemployed. In addition, respondents who perceive more group discrimination hold more negative attitudes toward the ethnic majority. Respondents who have more contact with the native Dutch, however, hold more positive attitudes toward them.

Table 2 Highest level of education from country of origin and attitudes toward the ethnic majority and ethnic minorities (OLS regressions)
Full size table

In the following models (3–10), attitudes toward the different ethnic minorities are shown excluding ingroup attitudes. These models show a consistent relationship between higher levels of education and more positive attitudes toward other ethnic minorities, thus supporting hypothesis H2a. In addition, separate analyses were conducted for each immigrant group in Tables A1-A5 in the Appendix. These analyses show a consistent relationship between higher levels of education and more positive attitudes toward other ethnic minorities among all immigrant groups, except for Poles’ and Bulgarians’ attitudes toward Turks, where no effect of education is found (Tables A3-A4).

Models 3–10 in Table 2 also show respondents who are currently in some type of education to hold more positive attitudes toward the other immigrant groups, except for Turks, when compared to those who are unemployed. Similarly, as in previous models, those who perceive more discrimination toward their own ethnic group hold less positive attitudes toward other ethnic minorities, while those who have more contact with the native Dutch, hold, on average, more positive attitudes.

In Table 3, results for hypotheses H1b and H2b are presented. In model 11, time as a continuous variable is added. This shows a negative linear trend over time for attitudes toward the ethnic majority, providing support for H1b. In model 12, I interact level of education with time and this result is illustrated in Fig. 3. Additional analyses using an alternative dependent variable, and a balanced panel (Tables A6-A7 in the Appendix) confirm these results.

Table 3 Multilevel repeated measurement models of attitudes toward the ethnic majority and ethnic minorities
Full size table
Fig. 3
figure 3

Attitudes toward the ethnic majority over time, by level of education

Full size image

In models 13–20, similar analyses were conducted showing attitudes toward other ethnic minorities are, for the most part, stable over time. Attitudes toward Bulgarians seem to become slightly more positive over time, whereas attitudes toward Turks become slightly more less positive over time. These time trends, however, are not as strong and pronounced as those captured toward the ethnic majority. Additional analyses were conducted using an alternative dependent variable and in a balanced sample in Tables A6-A7 in the Appendix, showing stability in attitudes toward other ethnic minorities, thus showing support for H2b.

Conclusion

Previous research has consistently shown a relationship between higher levels of education and lower levels of prejudice. However, studies have mostly relied on analyzing this relationship among ethnic and racial majorities and native-born, leaving the question open as to whether the educational effect is indeed present among immigrants toward the ethnic majority. In addition, research that has focused on inter-minority attitudes has also failed to investigate whether the educational effect is present among immigrants and ethnic minorities. In this study, I have sought to contribute to this gap in the literature by analyzing attitudes toward different outgroups among immigrants in the Netherlands, consisting of attitudes toward the ethnic majority (native Dutch), and toward other ethnic minorities in a longitudinal perspective.

Using four waves from the ‘New Immigrants Survey Netherlands’ (NIS2NL), my main objectives were: (1) to investigate to what extent educational attainment was associated with more positive inter-ethnic attitudes among four recent immigrant ingroups in the Netherlands, as it has repeatedly been shown to be the case among the native-born and ethnic majority (e.g., Hainmueller & Hiscox, 2007); and (2) to examine if these attitudes had different trajectories over time. Hypotheses H1a and H1b tested the effect of education from the country of origin on attitudes toward the ethnic majority. Consistent with the integration paradox, I found support for both hypotheses, inasmuch as higher levels of educational attainment are not associated with more positive attitudes toward the ethnic majority, and that these attitudes become less positive over time. Hypotheses H2a and H2b, were derived from strong theoretical expectations of the universality of the educational effect on outgroup attitudes. I hypothesized that higher levels of educational attainment were associated with more positive attitudes toward other ethnic minorities, and that these attitudes were stable over time. The results also provided support for these hypotheses.

I acknowledgements several limitations of the present study. First, this research cannot rule out potential self-selection mechanisms into higher education (e.g. Lancee & Sarrasin, 2015), and the effect of field of study on prejudice (Feldman & Newcomb, 1969; Guimond et al., 2003; Guimond & Palmer, 1996). Due to data limitations, I am not able to address these issues, which become more complex when taking into account the different educational systems from the immigrants’ country of origin. Moreover, it is also fair to assume there are selection processes that drive migration, something that is endemic to this type of study. For example, it could be the case that migrants who already were more positively disposed toward the native Dutch, moved to the Netherlands in greater numbers. Particularly, this applies to individuals from countries that were outside the European Union at the time of migration, due to being required to pass an exam measuring their knowledge of the Dutch language and society (Joppke, 2007).

However, even if this was the case, selection effects are not able to explain the worsening of attitudes toward the native Dutch (regardless of level of education) on the one hand, and on the other hand, the stable educational effect found toward other ethnic minorities over time. Indeed, the whole premise of the integration paradox is that better structurally- and socially-integrated migrants do not exhibit more positive attitudes toward the ethnic majority, as one would expect. Perhaps, this is due to a “honeymoon” effect, where migrants quickly become disillusioned in the host society, after a period of high spirits, due to, among other things, their low sense of belonging, their (unmet) cultural, economic and political expectations of the host society, experiencing exclusion in the host society, and feelings of attachment to their country of origin (Geurts et al., 2021).

In conclusion, these analyses reveal that higher levels of education among immigrants only have a so-called “liberalizing” effect toward subordinate groups in a host society, but not toward the ethnic majority. This sheds new light on the universality of the educational effect by delineating scope conditions to this specific phenomenon, highlighting the importance of ethnic group positions in the social hierarchy. For example, while Hagendoorn and Hraba (1987) assessed an ethnic hierarchy in the Netherlands for the first time among the native Dutch; their results show a striking resemblance with how immigrants view other groups in this study (Fig. 2), with respondents expressing the most social distance toward Turks, and the least toward Spaniards, among others. Subsequent studies have also shown that the ethnic hierarchy of the Dutch toward other ethnic groups has not changed much since the 1980’s (Hagendoorn, 1995; Hagendoorn & Pepels, 2003) with Turks and Moroccans evaluated the most negatively and Northern Europeans the most positively (Coenders et al., 2008; Verkuyten et al., 1996). This ethnic hierarchy is likely derived from the racialization of migrants from certain countries (Garner, 2007). This process of racialization of migrants is not just limited to physical features but may deem certain migrants as unassimilable due to notions of cultural incompatibility related to the idea of who constitutes a nation and national identity (Balibar, 1991; Schnapper, 1994, 1995), and who is included as a full member of society (Ivarsflaten & Sniderman, 2022; Velásquez et al., 2023). This may also explain, to some extent, why higher levels of education among Poles and Bulgarians were not associated with more positive attitudes toward Turks.

With these persistent ethnic group status hierarchies in mind, the results of the current study imply that higher levels of education has the potential to improve attitudes toward ethnic outgroups that are deemed lower in status by members of a given ethnic group. When individuals of a certain ethnic group evaluate another ethnic group similarly to theirs, or higher in status, then an educational effect is absent. Future research would benefit by investigating the possible mechanisms underpinning the integration paradox (e.g., Geurts et al., 2021, 2022), including immigrants’ negative and positive experiences (Geurts & Lubbers, 2022), as well as exploring whether these attitudes differ by educational attainment between immigrants in a receiving country and citizens in their countries of origin. This would not only shed light in the relationship between education and prejudice, but also on whether immigrants as a group are more likely to hold a particular set of attitudes as compared to those who do not migrate.