Prejudice and Cuts to Public Health and Education: A Migration Crisis or a Crisis of the European Welfare State and Its Socio-Political Values?

Abstract

The past few years have witnessed the development of prejudiced attitudes in some places in Europe. Biases alike are often considered a consequence of increased migratory movements to the continent and have also been connected to a more general crisis of the European Union political project. However, societies have diversely responded to migration even in countries presenting similar economic performances and immigrant inflows. Akin different reactions have raised some important questions: is prejudice connected to a broader European crisis and what does the latter consist of? This article responds to these research questions through a multilevel analysis of 24 European countries, and shows that the percentages of migrant population alone are not associated to anti-migrant sentiments. Such a situation has instead been the case only in those countries that have concurrently experienced cuts to the two key public sectors of education and health care, which constituted the pillars of the European Welfare State and one of the cores of the European Union’s political project.

1. Introduction

The year 2015 marked the starting point of what has been identified as a “Migration crisis” in Europe, and since then societies have witnessed the development of prejudiced attitudes in some places of the continent and segments of the population. One of the most widely used definitions considers that: “Prejudice is an antipathy accompanied by a faulty generalization” [1] (p. 821). So, “It concerns the sphere of those biased feelings and mental dispositions towards others that precede the real experience of facts” [2] (p. 90). Discriminatory postures alike are often considered a consequence of migratory movements to the region, which have grown in many places in Europe in past decades. However, societies have diversely responded to migration even in countries that presented similar economic performances and immigrant inflows. While some countries have experienced new xenophobic episodes and a revival of nationalist politics, others have more generally welcomed migrants. Conversely, a variety of countries have become quite polarized on the topic. As a result, harshly debated political campaigns have developed, often connected to the responsibility of the European Union in the supposed unease that migration has entailed for their societies, such as in the case of the Brexit campaign in the UK or the 2017 election campaign in France. These situations have illustrated that the post-national solidarity at the basis of both the European Union (EU) project, and the welcoming of persons of different origins, is seriously undermined in some places in Europe but prospers in others. Such mixed responses have raised some critical questions amongst scholars: why, in spite of a similarity of macro-level factors and mobility trends, have countries reacted differently towards migrants? Why have these reactions been connected with a larger crisis that involves the EU’s political project?

The literature in the past has already advanced that the crisis of the EU’s political project develops from the weakening of one of its fundamental pillars, the European Social Model [2]. Furthermore, García Faroldi showed that both the crisis of the EU’s inclusive post-national values and the lack of interculturalism in some segments of European societies go hand-in-hand with the presence of negative attitudes towards migration [3]. The cuts in public expenditures in some basic public services that are crucial for European citizens’ wellness and the lack of interculturalism together undermine the post-national solidarity in a way that could explain prejudiced attitudes towards immigrants. By considering the effects of long-term cuts to the public health and public educations sectors happened (2002–2012), this article proposes a multilevel analysis of 24 European countries in 2012 to consider if the percentages of migrant population have only affected the development of anti-migrant sentiments for those countries that have concurrently experienced cuts to the two key public sectors of education and health care. The choice of using the year 2012 for the multilevel analysis was made for both theoretical and methodological reasons. Theoretically, the analysis wishes to prove that such a situation preceded the 2015 migration crisis. Empirically, the 2012 European Social Survey round 6 includes more numerous and more relevant country data than the EES round 7, year 2014.

2. Multi-Level Analyses as Tools to Understand the Effect of a Political Crisis on the Development of Prejudice

For the past decades, many works in the social sciences have contributed to explaining prejudice as a consequence of both individual- and context-level variables. These studies have yielded todays’ greater understanding of the phenomenon. Since the mid-1990s, multilevel analyses have illustrated how individuals’ sentiments toward immigrants are not only dependent on individual characteristics, but also partly shaped by the context [4,5,6,7]. In one of the first analyses of context-level factors, Quillian proved that neither the economic context nor the size of immigrant population is relevant per se. It is only significant when the two elements are combined; so, when an economic recession coincides with an increase in the presence of migrants in a territory, the perception that migrants could represent additional threats to the labor market and for social services develops in the most vulnerable parts of a society [6]. In such a light, Quillian [6] demonstrated that individual attitudes towards migrants also depend on the interaction between economic variables and the size of the migrant population. Conversely, those who support the self-interest model consider immigrants to be actual and not perceived threats for less educated and skilled members of societies and, therefore, their tangible presence will entail the development of negative sentiments [7].

New multilevel studies have more recently illustrated the impact that new cultural and identity elements have on both individual and contextual responses to newcomers even in times of crisis [4,5,8]. Bello [4] suggests that when collective identities are constructed in inclusive ways, individuals are one-third less prejudiced towards migrants than in countries with exclusive identity constructions. Davidov and Meuleman [5] demonstrate that people who hold self-transcendent values show more positive attitudes towards migrants than conservative persons and that, in contrast with Quillian’s [6] findings, the economic performances of countries combined with the size of migrant population do not entail a prevalence of prejudice.

Recent studies akin to these have contributed to the opening up of new directions of investigation and have pushed towards the formulation of new explanatory theories [3,8], with a revival of interest in the cultural elements that affect prejudice. Namely, Bello [8] proves that, even in times of crisis, the development of intercultural values can moderate the negative impact of an economic recession by playing a major role in affecting the type of attitudes towards immigrants more than the one that economic variables imply. García-Faroldi shows that there is a relation between supporting the European Union and European identity and the development of positive attitudes towards migration [3]. In addition to this, by considering changes in public polices and consequences in the governance of migration in the last twenty-five years in Europe, it has been observed that prejudice has increased only in some segments of different societies because of unevenly distributed cuts to public sectors [2]. Therefore, prejudice in Europe seems to be connected to the crisis of the EU Social Model.

In such a light, it is possible to suggest that Quillian’s theory needs to be revisited to consider the ways through which societies are shaped, also including post-national elements such as in the case of the European Union. Recent research achievements urge the search for a comparative multilevel analysis that could prove the role played by these newly suggested context-level variables in finding what could explain these diverse country- and individual-level reactions to the presence of migrants. The current article responds to such a call by developing such an analysis that comprehends the contemporary migration crisis as a crisis of the EU’s political project and its inclusive post-national values, undermined by the cuts to the welfare state, which is identified with the two crucial public sectors of education and health.

Individual- and Context-Level Theories of Attitudes towards Immigrants

For more than half a century, sociological research has focused on the individual characteristics that affect people’s attitudes towards immigrants. Thanks to these studies, we know more thoroughly the individual-level variables influencing the development of prejudice.

Many studies have demonstrated that personal (age and gender), political (political orientation), socio-economic (status, income and education), and cultural attributes (religiosity and values), along with socio-psychological and social network factors (personal trauma, life satisfaction, alienation, social contacts and interactions), all explain attitudes towards immigrants [9].

Among the first theories of prejudice, Adorno et al.’s model of the “authoritarian personality” explains why certain personal features are linked to negative attitudes towards new persons perceived as “outsiders” or newcomers [10]. According to this theoretical framework, males, right-wing-oriented and older individuals—all qualities related to an authoritarian personality—are more likely to be prejudiced against immigrants and, therefore, are positively associated with prejudice [6,11].

Allport’s model [12], instead, focused on the irrational component of prejudice. Allport was the first to consider the effects that both ignorance about members of outgroups and faulty generalizations could have on attitudes towards immigrants. Building on this perspective, education and social interactions have always been correlated with more positive attitudes towards newcomers. Thus, Allport’s model1 has been inspiring for both the socio-psychological approach [11,13] and the social networks approach (i.e., [14]) in the study of sentiments towards those perceived as “non-citizens”. Allport’s model of prejudice has, therefore, connected personal trauma, negative affective behaviors, life satisfaction, and all manner of socially learned feelings of dislike, to aversion and negative attitudes towards immigrants in general.

Finally, the self-interest model has connected individuals’ socio-economic vulnerabilities and prejudice by outlining the role that economic interests play in determining hostility towards outgroups. In this theoretical framework, immigrants are considered competitors in the job market and the reason for the diminishing availability of social and welfare services. Therefore, low-skilled workers, those who have low incomes and a low-status occupation are more likely to show negative attitudes towards them [7]. In this case, the authors consider that the size of immigrant population will explain prejudice.

Bobo and Hutchings test these three different models by applying them in the framework of Blumer’s theory [15] of group position extended to a multiracial context. Their findings confirm that, depending on specific groups (Black, Latino or Asian people), different determinants included in these three models apply in the USA. For instance, while for Black people, social distance matters and income does not, the opposite is the case with respect to the other two groups [11]. These findings encourage the development of studies attempting to show that cultural and identity elements, along with context-level factors, affect attitudes towards outgroups.

Some scholars have more recently focused on the role that values play in the type of attitudes towards immigrants [5,16,17]. In particular, Ward and Masgoret [17] connect multicultural ideology and intercultural contacts as factors that both contribute to a decrease in the perception of threats and lead to positive attitudes towards immigrants and immigration policies. Similarly, Dandy and Pe-Pua [16], by investigating the role of multiculturalism and cultural diversity, identify ambivalence between these factors in dominant and non-dominant groups in three Australian states. Despite this ambivalence, their study demonstrates once more that the type of collective identity that dominant groups hold is linked to their attitudes towards immigrants and that, in general, intercultural contacts help the development of positive feelings towards minority communities. Furthermore, Davidov and colleagues [18], by focusing on the role of different values in European countries, differentiate between self-transcendent and conservative individuals. Their results show that values of self-transcendence—which include understanding, appreciation, tolerance and protection for the well-being of people and nature (also identified as universalism)—correlate with positive attitudes towards immigration policies [5,18].

Despite the great contribution of these studies on both individual- and context-level variables affecting attitudes towards immigrants, substantial questions, particularly across countries, remain unexplored [9]. This is particularly true due to the variety of studies that could not replicate Quillian’s findings [6] on the effects that a negative economic trend combined with the size of immigrant communities could imply for attitudes towards newcomers [5,7,19,20]. For this reason, in a review of the literature produced in the field, Ceobanu and Escandell [9] have highlighted the necessity to also focus on the institutional and socio-economic factors that could play a role in the existence of prejudice.

Among the newly suggested theories, Bello [4] has constructed an index that measures levels of inclusiveness in the process of a collective identity construction of countries. The index includes different macro-socioeconomic items that measure the performance of countries’ integration processes in the domain of citizenship, the labor market, and social activities and contacts, and has proved that these different levels of inclusiveness matter more than any other context-level variables when it comes to the composition of attitudes towards immigrants. The most inclusive country actually shows a general level of positive attitudes towards newcomers that is one-third higher than the most exclusive country [4]. In her study, at an individual level, the intercultural values of self-transcendence [18] remain the most powerful determinants of positive dispositions towards immigrants [4].

Studies of social exclusion have also highlighted the role of some of these macro-level factors in the creation of discriminatory attitudes and behaviors in European countries [21,22,23,24]. In particular, in a report on social exclusion in Ireland requested by the European Commission, Commins [22] explains that social exclusion is not only a result of a lack of citizenship rights but also a consequence of inadequate access to the societal institutions granting the concrete application of those rights. Namely, these societal institutions are the democratic and legal framework that promote civic integration, the labour market, the welfare system, and the family and community networks that provide individuals with interpersonal integration. As Robila [23] illustrates, social exclusion is the opposite of social integration. Social exclusion is actually a condition in which people suffer multiple disadvantages because they experience deprivation of basic needs, including education, health care, employment, housing and financial resources. This situation can indeed concern both long-term members of a society and newcomers and would eventually polarize the position of those who suffer, or feel threatened, by these conditions. Jordan [25] also notes that exclusion from health care, education and welfare reinforces the dynamics of polarization in a society. Atkinson [21] then clarifies that, as already claimed by previous studies [24,26], when persons and groups are poorly integrated in one or more of these societal institutions, this can lead to social isolation. Thanks to Allport’s work in the field of sociology, it is well known that social isolation is a strong component of prejudice [12].

Considerations akin to this have inspired new reflections on the role that governments could have in shaping the general perceptions of newcomers and minority groups more generally in a country. In a recent work that looked at migratory trends in the past twenty-five years and the development of prejudice in some specific places in Europe [2], the same connection between cuts in social welfare policies and the proliferation of different forms of exclusions could be noticed. In particular, it has been argued that, in some countries, the increasing presence of migrants has happened concurrently to cuts in public expenditures in two key public sectors—education and public health—which has, as a consequence, caused people started to blaming migrants for the worsening conditions of public services.

Developing these ideas further, the present article argues that the main context-level variable affecting prejudice in European countries today is not a result of a migration crisis per se, but is the combination of a migration crisis with a policy of cuts in public expenditure for the two basic services that are an important component of the European Welfare State: public education and public health care [27]. In such a light, it is possible to understand why prejudice is also connected to a larger crisis of the political project of the EU, which found one of its main components in the welfare state model that today is damaged by the attempt to cope with economic crisis by cutting the maintenance and wellbeing of these two basic services. Given this context, this article proposes an understanding of prejudiced attitudes towards migrants as a consequence of the crisis of the EU’s political project and its inclusive post-national values. The crisis of the EU’s political project developed from the weakening of the European Welfare State model and, in particular, cuts in public expenditures in these two public sectors. This situation entailed to the development of negative attitudes towards newcomers, because people associate the presence of migrants with the worsening conditions of these two public sectors. Therefore, the presence of a migrant population alone does not explain the prejudice faced by migrants. Only the interaction between the presence of migrants and a long-term policy of cuts in public expenditures in the two public services of education and health care cause negative attitudes to be developed towards migrants.

In the light of these theoretical arguments, the main hypothesis that this study proposes is as follows:

Hypothesis 1 (H1).

The presence of a migrant population alone will not be a significant context-level factor in explaining prejudice. Instead, its combination with a long-term policy of cuts to the two public sectors of education and health will be a significant context-level variable affecting attitudes towards immigrants.

3. Methodology

This work employs a multilevel analysis carried out on 24 European countries2. It aims to show that the percentages of immigrant inflows alone do not entail negative attitudes. Such a phenomenon is expected to take place only when, at context level, these inflows are concurrently combined with a long-term policy of cuts to the two public sectors of education and health care and, at individual level, with an absence of intercultural values (or values of self-transcendence in Davidov’s and Meuleman’s terminology). To test the hypothesis, this work employs context-level data for the period 2005–2012 to see the effects of long-term cuts to the public sectors of health and education. It then employs these data in a multi-level analysis for the year 2012 to observe how the phenomenon of prejudice emerged in Europe before the so-called “migration crisis” that Europe has witnessed from 2015 onward and that is still partly influencing societies’ and governments’ reactions to migration. The database Eurostat3 provides context-level data for the years 2005–2012. The database European Social Survey (ESS)4; in particular, round 6 for the year 2012 provides the individual-level data. The choice of round 6 for the individual dataset is explained by the fact that it is the ESS round that allows us to include more sample countries at the context level when combined with the Eurostat dataset used for the context-level variables, whose number is indispensable for employing a multilevel analysis. The 2012 European Social Survey round 6 (2012) includes more numerous and relevant country data than the EES round 7 (2014). As the intention of the present study is to prove that the prevalence of prejudice is associated to an interaction between the presence of migrants and concurrent cuts to the public expenditure in the two crucial sectors of health and education, and to contrast it with other models, the analysis must include at least 20 countries in the sample at context level. As suggested in the literature, there needs to be 5 context cases per variable at context level [28]. Therefore, the ESS round 6 (2012) allowed us to include at least 24 countries and considered the data available at context level from OECD and Eurostat. Instead, ESS round 7 (2014) allowed us to include only 19 countries cases for which data for the selected context-level variables that are available in the OECD and Eurostat databases.

The models have been estimated using the restricted maximum likelihood method (REML), as this method produces the least biased estimates5. Both individual- and context-level variables center around the grand mean6. This means that the estimates of each variable will be measured for an ideal-type individual with average characteristics considered in the entire sample composed of 24 countries. Each of these variables is also concurrently controlled by all the other individual- and country-level variables included in the model.

All variables show normal values and no problems of collinearity were found. Eventual problems of multi-collinearity in the model have been checked through variance inflation factors (VIF) in SPSS 15. All VIF values are extremely low, all below 2, so they are far from the threshold of 5 which is commonly used for large sample sizes such as the one used in this model.

To consider whether country level is a relevant context in the analysis of attitudes towards immigrants, the multilevel method employs an empty model, which has an intra-class correlation (ICC) showing the extent to which the context affects the dependent variable. The empty model is a model that includes a random intercept only, without explanatory variables. As the following formula illustrates, with [${\sigma }_{u0}^2$] being the context-level variance, and [${\sigma }_e^2$] being the individual-level variance, the ICC ($\rho$) shows the proportion of variance that the context level explains compared to the total variance:

$$\rho =\frac{{\sigma }_{u0}^2}{{\sigma }_{u0}^2+{\sigma }_e^2}$$

As Hox suggests, 10% is the minimum level of variance that the context level needs to explain in order to justify the employment of multilevel techniques of analysis [28]7. Therefore, $\rho$ = 0.10 is the rule of thumb.

3.1. Dependent Variable

Ajzen considers attitudes as “individuals’ dispositions to react with a certain degree of favourableness or unfavourableness to an object, behaviour, person, institution or event—or to any other discriminable aspect of the individual’s world” [29] (p. 41). It is difficult to define the term “immigrant”, as a country’s specific immigration policies define who is part of its immigrant population. The question is complicated further because researchers in the field have to deal with individuals’ disposition regarding people whom they “perceive” as immigrants [30,31]. Therefore, this work refers to attitudes towards immigrants as individuals’ dispositions regarding persons perceived as “immigrants” who come to live in the country, to work or to obtain better conditions of life.

This will be measured through the ESS battery items on migration normally included in migration studies: “Do immigrants make [country] a worst or a better country”, “Do immigrants enrich or undermine [country]’s cultural life”, “Is immigration bad or good for [country]’s economy”, measured with a ten-point scale, 0 (worst) to 10 (best)8. A principal component factors analysis proved the adequacy of the composite measure, with a KMO adequacy of 0.740 and a significant Bartlett’s test.

3.2. Individual-Level Variables Used in the Model

Age, gender, political orientation, authoritarian personality—According to Adorno and colleagues’ model [10], authoritarian, male and older persons will show negative attitudes towards immigrants. The two items, “it is important that government is strong and ensures safety” and “the importance to follow rules” (for both items the scale goes from 0 = “very much like me” to 6 = “not at all like me”), measure the least authoritarian individuals when it comes to its relation to the dependent variable9. Therefore, the least authoritarian individual will be, the more they will be positively associated with positive attitudes towards immigrants. Gender is also expected to be positively associated with positive attitudes towards immigrants. The left–right scale measures individuals’ political orientation from left to right, so it will be negatively associated with positive attitudes: the more politically oriented to the right, the more prejudiced towards migrants.

Sociability—According to Allport [12], the more sociable people are, the more they will be positively disposed towards immigrants. Sociability, measured with participation in social activities, is expected to be positively associated with positive attitudes towards immigrants.

Income—Personal economic status and qualified job, according to the self-interest theory, should be inversely related to prejudice: the higher their income and the skills, the less prejudiced people are against immigrants, because they do not see them as competitors. The ESS item used, instead of measuring actual income, measures people’s satisfaction with their life. People are less likely to lie about their personal situation when a question does not refer directly to the amount of money they earn [32]. This wording also allows us to capture those individuals who do not work, but whose perceptions are still influenced by their family status (e.g., young people or housekeepers).

Religiosity—According to Adorno and colleagues [10], the more authoritarian a person is, the more religious this individual will also be. Like in economic theories, a high degree of religiosity is, then, expected to be associated with negative sentiments towards migrants. However, O’Rourke and Sinnott [33] find that religiosity is correlated with the development of positive sentiments towards migrants. Indeed, some studies exploring the role of values in prejudice (i.e., [5]), expect the degree of religiosity to be correlated to tolerance and so to more positive attitudes towards migrants. This study also considers that religiosity will be associated with positive attitudes.

Self-transcendence and conservatism—Davidov and Meuleman [5,18] used some of the ESS items to draw up two indexes measuring the individual values of self-transcendence and conservatism and this analysis also includes an index for self-transcendence10. The authors demonstrate their validity using a structural equation modelling technique. This work has tested the index of self-transcendence with a principal component factor analysis (PCFA), which also proved an effective adequacy measure (KMO = 738, with a significant Bartlett’s test). According to Davidov and Meulemann, self-transcendence is associated with positive ATI, while conservatism is the reverse. For conservatism, this work does not include the index because both the importance attached to strong governments and the importance to follow rules were already employed to consider the role of the authoritarian personality. So, for the rule of parsimony, here only one item is included for conservatism: the importance attached to traditions.

3.3. Context-Level Variables Used in the Model

At the context level, the analysis has compared the hypothesis H1 with other two models: Quillian’s model [6] and the self-interest model [7].

To test H1, the model has introduced an interaction between the following three context-level variables: a variable that measures the mean variation in public expenditures for public education from 2002 to 2012 (OECD database), a mean variation in public expenditures for the public health care sector (OECD Database), and average variations of immigrant inflows (EUROSTAT). The decision to consider average variations for these two context-level variables has been taken in order to identify a long-term policy of cuts in public expenditures in the two public services of education and health care. For consistency, also for immigrant populations, the average immigrant inflows for the same time frame have been introduced in the model.

To test Quillian’s theory, the interaction between the variation in GDP growth per capita, at purchase power parity (EUROSTAT) for the years 2002–2012 and the migrant population size variations for the same years have been introduced in the model. To test the self-interest model, only the size of the immigrant population has been considered.

4. Findings

The empty model tested the ICC. The empty model tests the multilevel analysis on the dependent variable without including any independent variables. In such a fashion, it helps us to understand how the variance is distributed at both individual and at context levels, to evaluate whether an MLA is a valid technique to test the hypothesis. The minimum level of variance that needs to be present at context level to use an MLA is ρ = 0.10.

Table 1. Residuals’ inspection and variance explained by the different models at different levels of analysis (a).

Model	Residuals at Individual Level	Residuals at Context Level	Total Variance	ICC	Relative Variance Explained at Context Level	Relative Variance Explained at Individual Level
Empty	4.393	0.614	5.007	0.12
Model 1	3.723	0.447	4.170	0.107	0.272	0.152
Model 2 selftrasc.	3.662	0.418	4.081	0.102	0.318	0.166
Model 3a	3.505	0.337	3.84	0.09	0.451	0.202
Model 3 H1	3.505	0.278	3.78	0.073	0.550	0.202

a: The models were tested on a dependent variable composed of the ESS battery items on migration normally included in migration studies: “Do immigrants make [country] a worst or a better country”, “Do immigrants enrich or undermine [country]’s cultural life”, “Is immigration bad or good for [country]’s economy”, (0 = worst; 10 = best) ESS round 6: year 2012, with a multilevel analysis (REML applies) employing SPSS 15. Post-stratification weight applies. For descriptive statistics of all the variables included in the models, please see Table A1 in Appendix A.

shows the proportion of variance that the different models explain at each level of analysis and includes information on the intra-class correlation (ICC).

Table 2. Coefficients of fixed effects (a) per model tested in the multilevel analysis of attitudes towards immigrants.

Variables	Model 1	Model 2 Self-transcendence	Model 3a	Model 3b	Model 3c H1	Model 4 Quillian	Model 5 Self-Interest
Intercept	5.163 ***	5.129 ***	5.33 ***	3.44 (Non-sign.)	7.136 **	5.257 ***	5.201 ***
Individual-level variables
Gender (male; female)	0.025 (non-sign.)	0.070 **	0.093 ***	0.093 ***	0.094 **	0.093 *	0.093 ***
Age	−0.003 *	−0.003 *	−0.004 ***	−0.004 ***	−0.004 ***	−0.004 ***	−0.004 ***
Education	0.103 ***	0.095 ***	0.98 ***	0.98 ***	0.098 ***	0.098 ***	0.098 ***
Life satisfaction	0.135 ***	0.128 ***	0.125 ***	0.126 ***	0.126 ***	0.126 ***	0.126 ***
Sociability	0.123 ***	0.108 ***	0.114 ***	0.114 ***	0.114 ***	0.114 ***	0.114 ***
Political Orientation (Left–Right Scale)	−0.078 ***	−0.065 ***	−0.079 ***	−0.079 ***	−0.079 ***	−0.079 ***	−0.079 ***
Religiosity	0.043 ***	0.044 ***	0.051 ***	0.051 ***	0.051 ***	0.051 ***	0.051 ***
Authoritarian personality (rules), most to least authoritarian	0.057 *	0.081 **	0.074 **	0.074 **	0.074 **	0.075 **	0.074 **
Authoritarian personality (strong governments), most to least authoritarian	0.080 ***	0.158 ***	0.168 ***	0.169 ***	0.169 ***	0.169 ***	0.169 ***
Conservatism (most to least conservative)	0.099 ***	0.141 ***	0.130 ***	0.130 ***	0.130 ***	0.130 ***	0.130 ***
Self-transcendence		0.396 ***	0.414 ***	0.414 ***	0.414 ***	0.096414 ***	0.096414 ***
Context-level Variables
Percentages of Immigration Flows			0.224 (Non-sign.)	1.77 (Non-sign.)	0.69 (Non-sign.)	−0.335 (Non-sign.)	−0.169 (Non-sign.)
GDP Variation						−0.007 (Non-sign.)
Mean Imm Flows * GDP Variation						0.011 (Non-sign.)
Public education expenditure variation			−0.445 (Non-sign.)		−0.810 (Non-sign.)
Public health expenditure variation				−0.114 (Non-sign.)	−0.114 (Non-sign.)
Perc Imm Flow Mean * VariationPublEducation			−1.52 *		−29.86 *
Perc Imm Flow Mean * VariationPublHealth				0.124 (Non-sign.)	0.054 (Non-sign.)
H1: Perc Imm Flow Mean * VariationPublEducation * VariationPublHealth					−0.708 *
ImmFlow * GDPpercap (PPP)						0.000 (non sign.)
Imm Flow							0.000 (non sign.)

All variables are significant, except where differently stated. * sig. ≤ 0.05 ** sig. ≤ 0.01 *** sig. ≤ 0.001. The models were tested on the dependent variable composed of the ESS battery items on migration normally included in migration studies: “Do immigrants make [country] a worst or a better country”, “Do immigrants enrich or undermine [country]’s cultural life”, “Is immigration bad or good for [country]’s economy”, (0 = worst; 10 = best) ESS round 6: year 2012, with a multilevel analysis (REML applies) employing SPSS 15. Post-stratification weight applies. The OECD dataset has been used for public expenditure in public health and public education sectors for countries, and the EUROSTAT database has provided all other context-level variables.

illustrates the results of testing the different models.

As Table 1 elucidates, the empty model demonstrates that ICC is ρ = 0.12 and, consequently, it makes sense to employ multilevel techniques of analysis as 12% of the variance is at a context level. Therefore, a multilevel analysis is a proper statistical technique to test the hypothesis.

Model 1 introduces in the analysis the individual-level variables considered in the reviewed literature and listed in the methodological part. As the values in Table 1 indicates, this individual-level model presents a proportion of variance at a context level of roughly 45%. It is important to highlight here that the introduction of the individual-level variables already explains 27% of the initial context-level variance. This means that 27% of the variance at context level is explained by the different composition of countries at individual level. These individual-level variables explain 15% of the variance at individual level. They all show correlations in the expected directions, confirming those previous studies reviewed in this article. A very important variable is self-transcendence: this variable is significant, presents a coefficient of 0.39 and contributes to reducing almost 5% of the variance left at a context level after the introduction to the analysis of the initial individual-level variables and an additional 1% of the individual-level variance. This finding confirms the importance of intercultural values of self-transcendence to ensure that both societies and individuals are inclusive towards newcomers and all those perceived as “outsiders”. Intercultural values correlate with positive attitudes towards newcomers, strengthening those past research results, according to which, when societies and individuals frame their identity in inclusive ways, prejudice will not be affected despite the economic crisis [4,5,8].

The self-interest model and Quillian’s model both proved to be non-significant. Immigrants’ inflows do not affect attitudes towards migration, and nor does their interaction with the economic performance of the country. Their inclusion worsened the ICC and the fitness of the model, the reasons for which are not included in Table 1.

Model 3a,b include the single pair of interactions between immigrants percentages and cuts to the public education expenditure in one model, and cuts to the health expenditure in the other. Model 3a shows a significant coefficient. However, it reduces the unexplained variance at context level to 0.37.

Model 3c finally tests H1 by introducing the three-interaction model between immigrant percentages and cuts to public expenditures in the sector of both public education and health care11. The context-level variable is significant, and therefore there is support for the main hypothesis. As Table 1 illustrates, its introduction importantly reduces the unexplained variance both at context and individual levels. It explains 55% of all the unexplained variance at context level if compared to the empty model. The context-level variance remained unexplained after the introduction of all individual-level variables (including self-transcendence) decreased by 23% when the model testing the main hypothesis H1 was introduced in the analysis. The cuts to public expenditures in the sector of public education and health care did not significantly correlate with the development of prejudicial attitudes towards immigrants per se. However, when combined with the percentage of immigration inflows, their interaction was significant, and most crucially explained a very relevant part of the variance which was left unexplained by context-level variables.

The interaction between migrant inflows and cuts to public expenditures in education and health care sectors also accounts for roughly 5% of the variance left unexplained at individual level compared to the previous model, when only individual-level variables were included. This means that this interaction at context level also contributes to a better measurement of the impact of some individual-level variables previously introduced in the analysis, as it is possible to also consider this by inspecting the coefficients in Table 2. The context variable affects, albeit in a small way, the strength of the correlation of the individual-level variables, namely, the cuts operated in the public sectors of education and health prove to establish an environment in which the effect of individual predispositions are, in most cases, accentuated (see Table 2, model 3c). Such an effect is particularly important for the variables connected to the authoritarian personality, which pass from having a coefficient of 0.16 to 0.17, and self-transcendence, whose coefficient increases from 0.39 to 0.41.

The interaction between the context-level factors of the model presented here is associated with the prevalence of negative attitudes towards immigrants: as the coefficients of immigrant percentages are positive and those of the two variation expenditures are negative, their interaction entails that when there cuts to the two public sectors are higher, together with higher percentages of migration flows, they are associated with worse attitudes towards migrants.

Finally, the analysis tests Quillian’s hypothesis of perceived threats and the self-interest model of actual threats, but neither were significant, as Table 2 shows. This last finding finally confirms in full the theoretical framework presented in this study. The combination of economic performances and migrants’ inflows have not set a context for the development of prejudiced attitudes towards migrants before 2015. Instead, the undermining of the public sectors of health and education have definitely created a context that has affected individuals’ attitudes towards migrants.

5. Conclusions

The past few years have witnessed a development of prejudiced attitudes in some places in Europe. These exclusionary tendencies are often considered a consequence of migratory movements to the continent, which have increased almost everywhere in Europe in the last decade. However, societies have diversely responded to migration even among countries presenting similar economic performances and immigrant inflows.

For more than half a century now, many sociological works have contributed to explaining prejudice as a consequence of both individual- and context-level variables. These have yielded todays’ greater understanding of the phenomenon. However, substantial questions, particularly across countries, remained unexplored, such as why some countries, in spite of the similarity they have in economic and migratory inflow context dimensions, show very different reactions in terms of prejudice.

With a multi-level analysis of 24 European countries for the year 2012, this work has supported a hypothesis about what the background of the migration crisis consisted of and why it has been connected to a crisis of the EU’s political project in some places. First, this work has advanced and demonstrated through the data that the percentage of immigrant flows alone is not a significant context-level factor associated with prejudice. Instead, its interaction with a long-term policy of cuts to the two public sectors of education and health care that has been happening between 2002 and 2012 is a significant and important context-level variable that negatively affects attitudes towards immigrants. Therefore, there is support for the main hypothesis that this article has formulated a valid theory for the composition of prejudiced attitudes towards immigrants in Europe in recent years. Other models, and namely Quillian’s model of perceived group threats and the self-interest model of actual threats, were not corroborated by this multi-level analysis. The combination of economic performances and migrants’ inflows has not set a context for the development of prejudice before 2015. Instead, the current work shows that the undermining of the public sectors of health and education, combined with percentages of migration flows, definitely constitute a valid theory to explain individuals’ attitudes towards migrants. In particular, higher cuts to the public expenditure in the two crucial sectors of education and health, combined with higher migration flows percentages, are associated to the development of negative attitudes towards migration. Such a context-level interaction is able to explain a very relevant part of the variance at context level; namely, this model overall reduces 54% of the variance at context level and 20% at the individual level. Therefore, the model is a powerful analysis for accounting what affects prejudice. Further analyses should attempt to replicate the analysis for the following years to consolidate these findings.

Following the theoretical arguments hereof presented, this article affirms that the anti-migrant attitudes of some segments of European societies is not associated with a migration crisis per se and other previously identified variables. Prejudice is instead significantly correlated to the interaction of long-term policies of cuts to public expenditures to the education and health care sectors and percentage of immigrant inflows. The combination of these policies and the percentages of immigrant inflows has consequently crucially affected prejudice at both a context level, where the model alone accounts for 23% of the variance left unexplained once controlled for individual-level variables, and at individual-level, where it explains and additional 3% of the variance left, thus strengthening the explanatory power of individual-level variables, namely those connected to the theory of the authoritarian personality and self-transcendence. The latter is confirmed to be the most significant individual-level variable associated to positive attitudes towards immigrants, as past research results also showed [4,5,8]. These individual-level variables, together with gender, age, level of education, life satisfaction and the authoritarian personality, all contribute to explaining attitudes towards immigrants.

The two sectors of public education and health are considered the core elements of the EU’s political project and its welfare state. Their interaction with the percentages of migrants flows in countries is associated with the prevalence of prejudice. Such an analysis allows us to more clearly see the connection between attitudes towards migration and the wellbeing of the EU’s political project. When European citizens witness a degradation of public services of public education and public health, and concurrently growing migration inflows, such a situation becomes associated with negative attitudes towards migration. Without changing the presence of migration inflows, when instead the public expenditure in these sectors increases, the situation is associated with more positive attitudes towards migration. These important results should be taken into account in future studies of context-level factors that affect attitudes towards newcomers and all those perceived as “outsiders”. They contribute to understanding why prejudice towards migrants has been at the heart of political campaigns that mainly concerned the future of some European countries in the EU, such as has been the case of the Brexit campaigns and the 2017 election campaigns in France. These findings actually show that a longstanding policy of cuts to the two key public sectors of education and health care, which are important components of the European Social Model, that in some countries have been particularly harsh, have both undermined the EU post-national solidarity and led to the scapegoating of migrants for the worsening conditions of their countries.