Between Fears, Contacts and Family Dynamics: the Anti-Immigrant Attitudes in Italy

Abstract

Empirical research on sources of prejudice and anti-immigrant hostility has grown considerably over the years, and the literature on the topic has become substantial firstly in classical receiving countries and in recent years also in Europe. This work analyses the anti-immigrant attitudes in Italy. The main results of this study are three. The first is the importance of individual characteristics in explaining anti-immigrant hostility: much of the variability in anti-immigrant hostility is due to individual characteristics, and in particular to the educational level achieved. Second, contacts have opposite effect on hostility: real contacts (i.e., real interpersonal contacts) decrease anti-immigrant hostility, while abstract contacts (i.e., contacts mediated by mass media) increase hostility towards migrants. Third, anti-immigrant hostility is lower among issues concerning family immigration dynamics.

Introduction

According to the well-known aphorism of the Swiss writer Max Frisch, the controversial relationship of receiving societies with immigrants can be expressed as a contrast between labor demand and resistance against the social consequences of immigration: “We asked for workers. We got people instead.” Students of ethnic inequality and race relations have long been interested in understanding the social mechanisms underlying prejudice and discrimination against immigrants. Empirical research on sources of prejudice, discriminatory attitudes, xenophobia, and hostility towards ethnic minorities has grown considerably over the years, and the literature on the topic has become substantial.

This work focuses on anti-immigrant attitudes in Italy, providing three contributions to the existing literature. First, this article focuses on social relations between immigrants and natives (see Pettigrew 2015) distinguishing two types of contacts: real and abstract contacts: Real contacts means that natives interact with real people, with names, faces and needs, while abstract are those contacts mediated by the mass media, which frequently describe immigrants as threatening ghosts and create and support anti-immigrant populism, thus fuelling prejudice and hostility. Second, this work studies different dimensions of hostility, analyzing not only on those issues traditionally related to anti-migrants attitudes (economic and cultural/nationalistic issues), but also on some aspects of family immigration (family reunions, mixed couples, second generation), which have been less studied in the literature (Tolsma et al. 2008). Finally, literature on anti-migrants attitudes has greatly developed firstly in traditional receiving countries and in recent years also in the old European receiving countries (see, for instance, Fetzer 2000), but there is not much about the newer receiving ones (Campani 1993). Although Italy has often been covered by most comparative studies (e.g., Scheepers et al. 2002), to the best of our knowledge, no one has yet systematically examined anti-immigrant attitudes in this country, despite the dramatic growth of international migration in recent decades (Ambrosini 2013a). This growth leads to the another reason that makes analysis of the Italian case interesting. Differently from data collecting in other older receiving countries, the surveys conducted in Italy in recent years have measured anti-migrant attitudes when immigration reached its maximum growth, and then when it has been at the center of the public debate: consider, for example, the political discussion that followed the recent landings of migrants on the Southern Italian coasts.

The structure of the paper is as follows: after this introduction, the section “Previous Literature: Group Competition, Social Identity, and Vulnerability” describes the main empirical findings on the topic, and the section “The Interethnic Contacts and Family Dynamics” focuses on the effect of contacts and family dynamics on anti-immigrant attitudes. The section “Expected Evidences” presents a set of research hypotheses. The “Data, Variables, and Methods” section describes data and modeling techniques, the “Empirical Evidence” section the empirical findings, and the “Conclusions and Discussion” section concludes.

Previous Literature: Group Competition, Social Identity and Vulnerability

The literature on anti-immigration attitudes is very large, and the topic has been extensively studied in various disciplines, such as sociology, economics, and social psychology. The main literature divides factors affecting hostility towards immigrants into two strands, which refer to two different levels of analysis. The first includes studies adopting a microlevel perspective focusing on individual characteristics such as education, gender, economic condition, or political affiliation (O’rourke and Sinnott 2006). Within this research stream, individual aversion towards immigrants can be analyzed with two influential and complementary arguments: the ethnic competition theory and the social identity theory (Scheepers et al. 2002). The former explains anti-immigrant attitudes through the group conflict theory (Sherif 1967), which claims that attitudes and behaviors among groups reflect the fact that social groups have different interests and objectives. In a context of limited resources, the incompatibility between the objectives of different groups creates social conflict: if aims and interests are in conflict, social conflict arises, which leads to openly hostile and discriminatory attitudes and behaviors.

For this reason, members of certain social categories are more prone to perceive ethnic threats and develop more exclusionist attitudes: essentially because they feel themselves to be more in competition with immigrants. Since this competition is stronger in the lower strata of the social hierarchy, individuals who are socially and economically vulnerable—people from the working class, from older cohorts, and those with a lower educational levels—are more likely to express discriminatory and exclusionary attitudes towards members of outgroup populations (Semyonov et al. 2004; Carvacho et al. 2013). This dynamic is similar to the “working-class authoritarianism” described by the classic studies on social stratification and mobility: working-class individuals are exposed to negative experiences—such as competition for limited resources—that often produce deep-rooted hostilities expressed through ethnic prejudice and political authoritarianism (Lipset 1959).¹

Conflict among groups is not only the result of an economic competition, but it can also involve other non-monetary resources such as power or social status. Hostility against immigrants is often based on an endeavor to defend the values and the cultural and ethnic homogeneity of the society. In this sense, the anti-immigrant attitude includes “more general and sociotropic evaluations of interest resulting from a broader sense of community or national good” (Markaki and Longhi 2013: 315–316). Finally, besides these factors, a growing bulk of literature stresses that cultural orientation is the single most important individual factor predicting anti-immigrant attitudes. For instance, Lancee and Sarrasin (2015) claimed that differences between groups in anti-migrant attitudes are mostly due to selection effects and not to the alleged liberalizing effect of education.

The second broad strand in the literature uses a macrolevel approach—or combines micro and macro levels—to analyze the effect of country and/or regional characteristics on anti-immigrant hostility. The rationale is that structural environment around the individual creates a sort of filter which conditions the perceptions of immigrants: the type of neighborhood, the city, and the county where an individual lives determine both the size and the characteristics of the immigrant groups that he or she meets every day (Stein et al. 2000). Hence, the degree of ethnic competition has been conceptualized not only at an individual level but also at a macrolevel focusing on countries, regions, municipalities, or communities.

At macrolevel, prejudice and discriminatory attitudes are mainly studied with two indicators (Scheepers et al. 2002). The first is the size of the outgroup population, which has been considered to be the main indicator of threat and thus as a crucial determinant of prejudice and discrimination. The rationale of this indicator is straightforward: the increase in the size of immigrant population means greater competition for limited resources, which creates a “greater challenge to the actual interests and prerogatives of the dominant population” (Semyonov et al. 2006: 428). Hence, a dominant group (e.g., whites) becomes more racially hostile as the size of a proximate subordinate group increases, which putatively threatens the former’s economic and social privilege (Blalock 1967). The second macropredictor is the economic condition of the region of residence, because a worse economic condition creates restrictions in the labor market and thus greater competition among social groups.

Most of the studies focusing on the macrolevel have used multilevel modeling techniques, controlling the heterogeneity in individual attitudes across countries (or regions) using fixed or random effects (Markaki and Longhi 2013). Also, cross-national comparisons are widely used, mainly because of the increased availability of comparative dataset containing information on anti-immigrant attitudes (i.e., European Social Survey (ESS)). Despite the agreement on the statistical tools to be used, however, the empirical results are somewhat contradictory. The effect of migrants’ group size has been found in several studies on the relationship between whites and blacks in the USA, but the size argument has been judged inconclusive and not obvious outside the US context (Hjerm 2009). Also, the effect of economic conditions is unclear: while the theory predicts that limited resources increase hostility, the empirical evidence suggests that regional unemployment rates decrease anti-immigrant attitudes (Rustenbach 2010; Markaki and Longhi 2013). Another shortcoming of macroresearch is the broad and vague distinction drawn between conflicts on material resources, on the one hand, and values and identity on the other, so that it is difficult to disentangle economic and cultural interpretations (Schneider 2009: 54).

Moreover, cross-country comparative researchers often do not take account of the fact that the self-selection of immigrants differs among countries. This selection may have a crucial effect on anti-immigrant attitudes, because it determines the “type” of immigrant that natives meet in their everyday lives. Finally, another weakness concerns the fact that the size of immigrant group is a weak indicator of competition among groups, since it also measures the possibility of contacts between the native and immigrant population, which may have opposite effects on anti-immigrant attitudes. For this reason, the next section considers the role of contacts between migrants and natives separately, from both a theoretical and an empirical point of view.

The Interethnic Contacts and Family Dynamics

Beside micro (age, education, occupation, etc.) and macro (unemployment rate, immigrant group size, etc.) factors, other two issues might have an independent effect on anti-immigrant attitudes: the contact between immigrants and natives and immigrant family dynamics. While there has been a rich tradition of research on intergroup contacts in social psychology (see Pettigrew 2015) and in sociology (see, for instance, Dixon 2006; Savelkoul et al. 2011; Laurence 2014), the analysis of immigrant family dynamics has been less studied by previous literature (Tolsma et al. 2008). Nevertheless, the role of both factors—which can be defined as two mesolevel factors, because they are in-between the individual and the structural characteristics typically analyzed by other studies—is not still clear in the Italian setting.

Regarding the former, the contact theory claims that as the outgroup grows, so the opportunities for majority group members to have positive interactions with immigrants increase, reducing prejudice and stereotypes (Wagner et al. 2006). This hypothesis is based on the pioneering social-psychology studies, which claimed that intergroup contacts alleviate tensions between groups only under optimal conditions—such as common aims, intergroup cooperation, equal status, and authority support (Allport 1954), but in the recent years, scholars commonly agree that, even in the absence of these supportive conditions, intergroup contact typically improves intergroup attitudes (Pettigrew and Tropp 2008). Hence, face-to-face interaction, and the formation of social ties between different ethnic groups, leads to the formation of positive intergroup attitudes (Hewstone 2009), helping to produce lower levels of threat perception in contexts of high immigration (McLaren 2003).

However, the lack of data on intergroup contacts prevents in-depth study of the role of interpersonal relations between migrants and natives. For instance, to date, only a handful of studies have set out to synthesize and test group threat theory and intergroup contact theory simultaneously, and the actual evidence that outgroup size is a contextual characteristic which affects (positively or negatively) anti-outgroup attitudes is still inconclusive (Scheepers et al. 2002). Moreover, contacts are often analyzed in a static way, although interpersonal relations are likely to change over time. One of the few studies based on longitudinal data on contacts shows that more educated immigrants tend to develop more contacts with natives over time because they attribute less importance to their ethnic network (Martinovic et al. 2009).

Previous literature considered social contact as real contacts, meaning that that natives interact with real people, with names, faces, and needs. Social contacts, however, can also be abstract, in the sense that the natives can establish a relationship with migrants without a real interpersonal relation. Abstract contacts are often managed by the mass media, which frequently describe immigrants as threatening ghosts and create and support anti-immigrant populism, thus fuelling prejudice and hostility (Boomgaarden and Vliegenthart 2007).

The concept of abstract contacts is a novelty in the literature. In psychology is widespread the concept of imagined contact (Crisp and Turner 2009), but it has differently understood from the concept of abstract contacts. In the case of imagined contact, in fact, participants of experiments are asked to imagine contact defined as a positive interaction with a member of an outgroup. Abstract contacts, on the contrary, means that individuals create a certain imagination of immigrants on the basis of the information managed by mass media. Actually, abstract contacts do not refer to any form of contact which would involve some kind of interaction with another person, for example talking with each other or being involved in the same activity or even sharing the same space. This kind of contacts indicated an exposure to information and, possibly, selective memorization of this information. However, it could be helpful to consider this type of exposition at mass media news as a different type of contacts between natives and immigrants, highlighting the difference between a real interpersonal contacts and abstract contacts with phantasmagoric entities.

The effect of mass media on anti-migrant attitudes has been deeply analyzed in the literature, where a long-standing proposition advanced in the group threat literature is that perceived threats and anti-minority sentiments are centrally shaped by negative mass media portrayals of minority groups (Allport 1954). Also, in the recent years, researchers widely agree that the mass media depict immigrants in a predominantly negative tone and as a threat to the host society (e.g., Brader et al. 2008). This has an effect on anti-migrant attitudes, because the number of negative immigration-related news reports increases perceived group threat over and above the influence of immigrant group size (Schlueter and Davidov 2013), increasing the spread of the populism and populist parties which sustain strong anti-immigrant policies (Mazzoleni 2003). Moreover, it has been shown that the effect from negative immigration-related news reports on perceived group threat increases with a smaller immigrant group size, hence when people have less direct contact (i.e., real) with the outgroup members (Newton 2006).

To our knowledge, no studies have investigated the association between exposure to television messages and the formation of anti-immigrant attitudes in Italy, even if also the Italian mass media often give more emphasis to crimes committed by immigrants than by natives and tend to reproduce a stereotyped and distorted image of immigration (Osservatorio Europeo sulla Sicurezza 2013). Also, current affairs programs play an important role in this regard, for instance by emphasizing the Islamic terrorism threat or choosing guests in an unbalanced way: such programs often have guests from parties with clear anti-immigrant positions but small political constituencies compared with other political forces (Cere 2010). This crucial issue, therefore, deserves to be analyzed in more detail.

Beside real and abstract contacts, this study focuses on another mesolevel factor: immigrant family dynamics, namely the fact that the migration decision and the socioeconomic consequences of migration are embedded into the family (Ballarino and Panichella 2017). As opposed to the anomic and fragile condition of immigrants living alone, family reunions and the spread of immigrant families are often viewed as integration factors implying the assumption of rights and duties on the part of immigrant families. Immigrant families also become potential actors of experiences of participation and citizenship from below, and in everyday life, they come into contact in many ways, creating reciprocal practices of peaceful coexistence (Baumann 1999). Nevertheless, in recent years, more critical positions have been taken on immigrant families, which are seen as social entities which defend and perpetuate the cultural difference with respect to natives, and in which patriarchal oppression, gender inequalities, forced marriages, and female genital mutilation are reproduced (van Walsum 2011; Grillo 2011). These criticisms often derive from “feminist” arguments and have then been used by various political parties. The criticism of multiculturalism, defined as a political orientation that “is bad for women” (Okin 1999), is a point of encounter between the feminist movements and parties against immigrants and cultural diversity. For instance, prohibitions on veiling have been justified, in France and elsewhere, with the aim of protecting young women against the impositions of traditional families, and they have gained wide political consensus.

Also, political institutions have managed family reunion and other issues related to family immigration in contrasting ways. On the one hand, an emblematic example is the case of the Netherlands, where the government has favored family reunification with the explicit aim of increasing the positive effects of living in a family environment on the social integration of immigrants (Bonjour 2011). On the other hand, there are several examples of opposition to family reunions. In European post-war migration, for instance, reunions were not officially admitted because immigration was considered to be a temporary episode (e.g., guest-worker programs). Migrants were recruited to meet certain requirements of the host economic systems but not allowed to settle permanently in the receiving countries. The closure to entry by immigrants’ family members represented a guarantee of persisting social roots of immigrants in their places of origin, while reassuring natives about the provisional nature of migration and avoiding social costs arising from the acceptance of persons not economically active but receiving welfare services.

Another aspect of family migration dynamics that may affect attitudes to migrants is the spread of mixed couples, although in this case, there is general consensus that they diminish hostility. Marriage between individuals from different social groups is a classic proxy for the openness of a given social structure. Since people prefer to have a partner with similar values and worldview, the spread of mixed marriages has been considered as an indicator that cultural boundaries and the hostility between immigrants and natives have become weaker (Gordon 1964).² Moreover, the progeny of mixed marriages often take on multiple ethnic identities, which, in successive generations, further erode ethnic distinctiveness while potentially hastening the process of acculturation (Qian and Lichter 2001).

However, recent empirical studies have redefined these ideas. In Italy, for instance, mixed unions may play a pivotal role as drivers of socioeconomic integration, rather than being effects of it (Guetto and Azzolini 2015). In fact, immigrants, in most cases, the women, seem to “trade” some of their valuable traits—e.g., young age and high education—when marrying “lower rank” members of the native population—e.g., older and less educated individuals, in most cases men (Guetto and Azzolini 2015).

Expected Evidences

Starting from microlevel, the literature discussed at the outset of this discussion suggests that negative attitudes towards outgroup populations are likely be more pronounced among socioeconomically vulnerable populations. Thus, your hypothesis is, hostility against immigrants is higher among less educated (hypothesis 1). The effect of education on anti-immigrant attitudes can be direct, in the sense that highly educated individuals are more open-minded, are less influenced by stereotypes, etc., or indirect, because they are more likely to be employed in the lower strata of the occupational structure (Ballarino and Panichella 2015, 2017), where competition with immigrants is greater. In the first case, we expect that the effect of education on hostility against immigrants remains substantial even when occupational condition is controlled for (hypothesis 1a).³ On the contrary, if the effects of education are mediated by the position on the individuals in the occupational structure, we expect that differences among individuals with different educational titles substantially diminish once occupational condition has been controlled (hypothesis 1b).

Regarding contacts between immigrants and natives, the literature suggests that real and abstract contacts may have opposite effects on hostility. Thus, we expect that people who have real contacts with immigrants have less anti-immigrant attitudes, while those with only abstract contacts show higher levels of hostility (hypothesis 2). Moreover, the effect of both types of contacts may change according to the educational level of natives. On the one hand, highly educated individuals are generally more open-minded towards persons of different origins; therefore, they are open-minded and they are also more likely to redefine negative stereotypes. On the other hand, they are more likely not to be influenced by negative stereotypes fueled by the media. In technical terms, there may be a significant interaction effect between education and contacts, where the positive effect of real contacts is stronger among more educated individuals, while the negative effect of abstract contacts is stronger among lower educated ones (hypothesis 2a). We also expect the effect of abstract contacts to be stronger on those issues more emphasized by the mass media, such as criminality, safety, and so on (hypothesis 2b). The last expected finding regards family dynamics. Since security, defense of cultural identity, and preservation of scarce public welfare resources are the main issues which justify the closure of natives towards new residents (Ambrosini 2013b), while families are important for easing integration in the host country, we expect that natives have lower hostility on issues regarding immigrant families (family reunion, mixed couples, second generation at school, etc.) than on other issues (hypothesis 3).

Data, Variables and Methods

Data and Variables

Empirical analyses are based on data collected by CISF (Centro Internazionale Studi Famiglia) in 2013, which includes detailed and reliable information on feelings of hostility towards immigrants among a representative sample of the Italian population. After excluding cases with missing information and the first and second generation of migrants, the analytical sample size included 3330 cases (80.7% of the total sample).

The dependent variable was an additive index of anti-immigrant attitudes. This total index of hostility was the mean of a set of 12 0–4 indicators (0 = strongly disagree; 4 = strongly agree) of opposition against migrants. We then analyzed three other dependent variables, which were constructed in the same way but focused on three specific dimensions of total anti-immigrant attitudes (Table 1).⁴ The first was general hostility, which captured a wide-ranging negative view of immigration through five indicators (see Table 1). Basically, this dimension includes the issues most emphasized in the mass media debate, especially on television. Hence, following hypothesis 2b, we expected it to be strongly affected by abstract contacts. The second variable measured a cultural/nationalistic dimension of hostility emphasizing claims that immigrants must be culturally similar to the standards of the host society. As well as by the media, these issues are often stressed by right-wing political parties, which often address issues related to immigration using nationalistic arguments. Finally, the third dimension controlled hypothesis 3 and regarded the concern about some aspects of family immigration, with particular reference to family reunions and mixed couples.⁵

Table 1 Indicators of anti-immigrant attitudes (% of value 4)

Since our dependent variables were constructed in a slightly different way with respect to other studies (e.g., Schneider 2009; Rustenbach 2010), we replicated the analysis using a different operationalizations. First, the sum of the indicators instead of their mean was analyzed. Second, we also performed parallel analysis using an ordinal logit and logit model on an ordinal and dichotomous (0 = above the mean; 1 = over the mean) and ordinal dependent variable, respectively. Results of these additional analyses, which are available on request, confirm the results presented in the next section.

The main independent variables concern contacts between immigrants and natives. The CISF questionnaire includes two questions measuring the contacts with immigrants. The first is, “how often do you have direct contacts with immigrants?.” This question measures the real contacts with immigrants, and the related variable was operationalized with a 0–4 scale (0 = never; 1 = very often), measuring how often subjects had contacts in their daily lives with immigrants (mean 1.97; SD 0.98). Similarly, the questionnaire contains information for abstract contacts (“how often you have heard of immigrants on television, radio, or read about them in newspapers?”), which were measured with another 0–4 scale (0 = never; 4 = very often) variable, but measuring how often subjects had contacts with immigrants through television (mean 2.53; SD 0.86). The correlation between these two variables is 0.23.

Individual’s education is the first control variable, which includes three categories: (a) lower secondary or less, (b) upper secondary, and (c) tertiary or higher. Models also control for occupational status, distinguishing inactive, unemployed, and employed individuals. The employed were further divided according to their social class using the EGP (Erikson–Goldthorpe–Portocarero) class scheme as presented in Breen (2004). We ended up with a variable with six categories: (a) inactive, (b) unemployed, (c) employed in the service class (EGP I-II), (d) employed in the white collar class (EGP IIIab), (e) self-employed with fewer than five employees (EGP IVabc), and (f) employed in the working class (EGP V-VI-VIIab). Other control variables were gender, age class (15–35, 36–55, > 55), and size of municipality (fewer than 10,000 residents; 10,000–250,000 more than 200,000). We also included 17 regional fixed effects in the model in order to control for the structural characteristics of the region of residence (unemployment rate, size of immigrant group, and so on.).

Methods

The research hypotheses were tested by means of OLS regression, where the dependent variables (Y _i ) were the total index of anti-immigrant attitudes and its three dimensions.

Five different models were estimated:

$$ {Y}_i=\alpha +{\beta}_i{Cr}_i+{\beta}_i{Ca}_i+{u}_i $$

(1)

$$ {Y}_i=\alpha +{\beta}_i{Cr}_i+{\beta}_i{Ca}_i+{\beta}_i{EDU}_i+{u}_i $$

(2)

$$ {Y}_i=\alpha +{\beta}_i{Cr}_i+{\beta}_i{Ca}_i+{\beta}_i{EDU}_i+\beta {OCC}_i+{u}_i $$

(3)

$$ {Y}_i=\alpha +{\beta}_i{Cr}_i+{\beta}_i{Ca}_i+{\beta}_i{EDU}_i+\beta {OCC}_i+{\beta}_i{Z}_i+{u}_i $$

(4)

$$ {Y}_i=\alpha +{\beta}_i{Cr}_i+{\beta}_i{Ca}_i+{\beta}_i{EDU}_i+\beta {OCC}_i+{\beta}_i{Z}_i+{\beta}_i{REGIO}_i+{u}_i $$

(5)

The first model estimates the gross effect of abstract and real contacts on anti-immigrant attitudes. The second controls for own education (EDU_i), while the third also controlled for occupational condition (OCC). The aim of models 2 and 3 was to control hypotheses 1a and 2b, determining whether the expected higher negative attitude of less-educated individuals remained significant even when their employment status was controlled for. Model 4 added a vector of individual sociodemographic characteristics (Z _i ) including gender, size of municipality of residence, and age groups. Finally, model 5 included a set of regional fixed effects (REGIO _i ). We used the likelihood ratio (lr) test to evaluate the goodness of fit among these nested models.

In order to control whether the effects of the contacts change among individuals with different educational levels, it has been estimated and additional model (model 6) including an interaction between real and abstract contacts and educational titles. These interactions are estimated simultaneously, but results did not change if they are estimated separately, namely in two different models.

$$ {Y}_i=\alpha +{\beta}_i\left({Cr}_i\times {EDU}_i\right)+{\beta}_i\left({Ca}_i\times {EDU}_i\right)+\beta {OCC}_i+{\beta}_i{Z}_i+{\beta}_i{REGIO}_i+{u}_i $$

(6)

Empirical Evidence

Immigrants and Their Families in the Eyes of Natives

Empirical analysis starts by exploring the individual characteristics associated with anti-immigrant hostility and its dimensions. Table 2 reports a set of linear regression models on anti-immigrant attitudes measured with our first dependent variable: the total additive index.

Table 2 Determinants of total anti-immigrant attitudes. Linear regression models: beta, significance level and confidence intervals at 95% in brackets, log-likelihood ratio test

Results confirm that different types of contacts have opposite associations on anti-immigrant attitudes. Real interpersonal contacts significantly decrease hostility towards immigrants, while abstract contacts mediated by the mass media are strongly and positively associated with anti-immigrant attitudes. Results also show that education is a key variable in explaining the differing hostility towards immigrants among the population, and its association remains statistically significant even when occupation (model 3) and other regressors are included in the model (models 4 and 5). Thus, it is difficult to attribute the negative association of education on hostility to the fact that education provides skills by which individuals do not have to compete with immigrants in the lowest levels of the occupational structure; it is more likely to be the consequence of other factors, such as the greater tolerance and the broader international outlook of highly educated individuals. This result is consistent with other empirical studies, which showed that the link between education is not affected by labor market competition, but it is mainly driven by difference among individuals in cultural values and beliefs, since more educated place greater value on cultural diversity than do their counterparts (Hainmueller and Hiscox 2007).

The third important result is that the regional fixed effects do not change the associations between individual factors and anti-immigrant attitudes.⁶ When individual characteristics are controlled for, the variability of anti-immigrant attitudes among different Italian regions is very limited. Also the likelihood-ratio test confirms the limited weight of macro regional factors, since it was not significant between model 4 and model 5, when regional fixed effects were included in the equation (LR chi² = 28.8, p = 0.12). This is an interesting result because it shows that, when individual factors are kept under control, the variability between the different regional contexts is very limited and non-significant, even if Italian regions are very different from each other as regards both the presence of immigrant populations and economic conditions.

If we estimate model 5 considering different dimensions of total anti-immigrant attitudes, the results are substantially similar (Table 3). An interesting result concerns the abstract contacts, which are associated with higher level of hostility concerning the general and the nationalistic dimensions, as expected under hypothesis 3, but they are not associated with hostility concerning family immigration issues. This result is the consequence of the type of information given by the mass media, which fuel a general sense of hostility towards immigrants, for example by giving greater prominence to crimes committed by immigrants, but it does not address issues on immigrant families, like the spread of mixed couples, family reunions, and the rise of the second generation of immigrants.

Table 3 Determinants of general, nationalistic and familial dimensions of anti-immigrant attitudes Linear regression models: beta, significant level and confidence intervals at 95% in brackets

It is also interestingly to note that, although the models used the same specification for the three dimensions, the constant of the model of family immigration is lower. This means that hostility towards immigrant family issues is lower compared with economic and cultural issues. This evidence emerges more clearly in Fig. 1, which reports the predicted values (and the relative 95% confidence intervals) of different dimensions of hostility by educational level, estimated with the same specification as the models reported in Table 3.

Fig. 1

Predicted values of hostility among different dimensions by education

Although on all three dimensions more educated individuals meet a lower level of hostility than less educated ones, the results clearly show that family issues arouse less apprehension. To sum up, if we consider the relative value of hostility among people with different educational levels, individuals with tertiary educations have a lower level of hostility on all three dimensions; however, if absolute values are taken into account, the results clearly corroborate hypothesis 4, confirming that hostility towards family issues is lower than towards other issues.

The Contacts Between Immigrants and Natives

This section focuses more closely on the two factors with the highest association on hostility towards immigrants: education level and contacts with the immigrant population. Figure 2 shows the predicted values of total anti-immigrant attitudes estimated with an additional model, which interacted the frequency of real and abstract contacts with education.⁷ The aim was to control hypothesis 2 by studying whether the association of both types of contacts change according to education.

Fig. 2

Effect of contacts (real and abstract) on total index of hostility, by education. Linear regression model: predicted value

If we focus on the total anti-immigrant index, the results further confirm the crucial importance of contacts: independently from own education, real contacts decrease while abstract contacts increase the total anti-immigrant attitudes. However, whilst the signs of the effects of contacts do not change among people with different educations, their magnitudes are different. The predicted total anti-immigrant attitudes among those with tertiary educations and 0 point (never) in real contacts is 1.65, while the corresponding value for those with 4 points (very often) is 1.04, with a significant decrease of 0.61 points. On the other hand, for those with compulsory educations, the reduction is smaller (0.23), although it is significant. These results confirm that real contacts decrease hostility, but this decrease is more pronounced when more educated people interact with immigrants. Hence, even when highly educated people have a negative attitude to immigrants, when they come into contact with them, they are abler to question their own prejudices and stereotypes. This result might be also the consequence of the fact that types of contacts might change shifting from low- to high-educated individuals: neighborhood contacts might prevail among the low-educated, while friendships, possibly related to the school and university environments, might be more frequent among the high-educated. Unfortunately, available data do not allow to analyze these distinctions.

The association between abstract contacts and anti-immigrant attitudes by education is opposite. It is significant among persons with compulsory schooling: when contact is equal to 0, the predicted value of total hostility is 1.60 and then increases to 1.97 when the contact is equal to 4. By contrast, abstract contacts have a limited effect among tertiary educated individuals because the relative difference (− 0.09) in the predicted total anti-immigrant attitudes between 0 and 4 point is smaller and not significant. Thus, the less educated not only have more difficulties in questioning their negative attitudes through real contacts, but they are also more susceptible to messages of alarm and fear conveyed by the mass media. More educated individuals, by contrast, have better means to face and destructure the way in which immigration is portrayed by the mass media.

Figure 3 shows the effect of the same interactions on different dimensions of hostility (confidence intervals are not shown in the figure but are available on request). While for the general/economic and cultural/nationalistic dimensions the results are similar to those of the total index, two differences can be noted for family dynamics issues.

Fig. 3

Effect of contacts (real and abstract) on different dimensions of hostility, by education. Linear regression model: predicted value

First, although real contacts reduce hostility also on this dimension, the magnitude of the reduction is in general lower than that observed for the other two dimensions. This lower reduction is due to a sort of “floor” effect, because the average of this kind of hostility is smaller (see above). Second, it is confirmed that abstract contacts do not affect hostility towards immigrant families, independently from the educational levels achieved by individuals. For instance, the difference in the predicted hostility among tertiary educated individuals with 0 and 4 points in abstract contacts is very small and not significant (− 0.04). The same applies to individuals with upper secondary education (− 0.04), while the increase among those with compulsory educations is limited and barely significant at the 10% level (0.13). In conclusion, if it is true that abstract contacts mediated by the media have a strong and important impact on attitudes of hostility towards immigrants, it is also true that this result is confirmed only when economic and nationalistic dimensions are considered and their effect is greater among people with lower levels of education. By contrast, abstract contacts do not have any effect on hostility when issues related to family migration dynamics are considered, independently from the educational level of respondents.

Conclusions and Discussion

This study has examined factors affecting hostility to immigrants in Italy by considering two factors that needed to be investigated further: contacts between immigrants and natives and the role of immigrant family dynamics. The first result was the importance of individual characteristics in explaining aversion to immigrants. At least in Italy, much of the variability in anti-immigrant hostility is due to individual characteristics, and in particular to the educational level. Another interesting result regards the role of contacts. As expected, real contacts—which include real interpersonal contacts with immigrants—decrease anti-immigrant attitudes, especially among more educated individuals. On the other hand, when contacts are not real because they are mediated by the mass media, and are therefore abstract, they tend to increase anti-immigrant hostility. It is also interesting that exposure to the mass media has no effect among highly educated individuals but has a strong effect on hostility among those with less education.

The opposite effects of real and abstract contacts confirm that the views of respondents, and especially their critical assessments of immigration, are distinct from their direct and personal relations. Daily interpersonal interactions, in fact, have the effect of “normalizing” the settlement of immigrants in the host society, creating spaces and opportunities for reciprocal help between immigrant and natives (Ambrosini and Bonizzoni 2012). Thanks to interpersonal relations, immigrants begin to become neighbors, parents of classmates, customers of the same shops, users of the same urban spaces, and so on. All these interactions reduce stereotypes, and immigrants lose their menace, taking on less disturbing and more familiar traits: individuals move from stereotypical views (“immigrants,” or at most “Peruvians” or “Pakistanis”) to a more composite one. We can conclude thus, the less you know them, the more you avoid them and are afraid; if you begin to know them, maybe fear diminishes and you begin a relationship.

The third result regards the fact that Italians have more hostile positions towards immigration on those issues which imply competition with foreigners, such as priority in the labor market, access to public housing, and the right to vote, or when immigrants are perceived as threatening the dominant culture. In regard to family issues, by contrast, positions become milder, and hostility towards immigration decreases for all social groups. Indeed, immigrant families tend to acquire residential stability and greater identification with the place where they settle; they need economic stability to meet family needs. They send their children to school, encouraging participation in extracurricular educational activities. Families significantly reduce the risk of involvement in deviant activities (Portes and Rumbaut 1996). It is also interesting to note that hostility towards family issues is not affected by abstract contacts, because the mass media tend to emphasize other issues related to migration (i.e., crimes committed by immigrants) and more rarely focus on family reunion, the rise of the second generation, and the spread of mixed couples.

There are still some points that deserve further investigation. First, further studies might consider the issue of the self-selection, namely the fact that people most hostile towards immigrants tend to avoid contacts with them, while those who are more benevolent tend to have more contacts (Pettigrew 1998; Schlueter and Scheepers 2010; Thomsen and Birkmose 2015). The same argument also applies to abstract contacts: people select the type of television shows that they watch, the newspapers that they read, and so on. Hence, since no credible causal estimate can be retrieved with the using data, other studies are clearly needed. Second, our results show that macroregional factors have a limited effect on anti-immigrant attitudes in Italy. Most of the variability of the anti-immigrant attitudes, as well as its specific dimensions, are largely affected by the individual characteristics and the types of contacts that natives establish with immigrants. Also, this conclusion should be taken with caution, because CISF data do not allow robust estimation at macrolevel. To have a complete answer on how, ceteris paribus, the hostility towards immigrants changes among Italian regions, other analyses based on surveys with larger sample sizes are needed.

Finally, the Italian case is interesting because of the strong socioeconomic cleavage between northern and southern regions (Ballarino et al. 2014). Indeed, between the 1950s and the first half of the 1970s, about four million Italians moved from the rural areas of the South to the more industrialized northern regions (Panichella 2014). It is possible that this migration flow has in many ways come into contact with the more recent immigration from abroad, for example in the working-class districts of the northern cities (Turin, Milan) or in the unskilled and/or irregular labor market, creating a peculiar situation of downward competition and mutual hostility between old internal immigrants and foreign newcomers. Despite its importance, few studies have analyzed this issue. Unfortunately, CISF data do not contains information on the region of origin, hence it was not possible to study whether the anti-migrants attitudes is higher among Italian internal immigrants.

Appendix

Fig. 4

Distribution of total hostility index and its dimensions. Red line = mean of total index