The Role of Social Capital in Shaping Europeans’ Immigration Sentiments

Social capital as a theoretical concept is present within both sociological and economic traditions.

One of the first formal uses of social capital as an academic concept was laid out by the French Marxist sociologist Bourdieu (1986), which refers to economic resources that one gains from being part of a network of social relationships, including group membership.

The ideas of shared trust, norms, and values have also a long tradition in economics. Bruni and Sugden (2000) pointed out that in his Lectures on Jurisprudence, Adam Smith presents a theory of social capital that is quite similar to that of the modern economic sociological theories of Putnam (1993, 2000).

Smith argued that “... reputations for trustworthiness are transmitted through networks of trading relationships; the denser the network ... the greater is the value of reputation and so ... the greater is the degree of the trust.” (Bruni and Sugden, 2000, p. 33)

In similar vein, the scant literature on immigration sentiments and social capital focuses mainly on one aspect of the latter, that of trust. For example, the study of Herreros and Criado (2009) employed one dimension of social capital, that of “trust in people”, to study the role of social trust on attitudes toward immigration. The study argues that societies with high levels of social capital facilitate the integration of immigrants because those members with high levels of social trust will tend to have more positive attitudes toward immigration. This hypothesis is empirically tested in a sample of 16 European countries, drawing on the 2002–2003 wave of the ESS. Results show that, regardless of the impact of other factors such as levels of unemployment and percentage of foreign population, citizens with high social capital do exhibit more positive attitudes toward immigration than the rest of the population.

The role of social integration and its effects on anti-immigrant attitudes were also studied by Rustenbach (2010) among other explanations, namely, cultural marginality theory, human capital theory, political affiliation, neighborhood safety, contact theory, foreign investment, and economic competition. Results reveal that among all tested explanations, interpersonal trust was the strongest predictor of anti-immigrant attitudes. Individuals with high levels of interpersonal trust are more likely to overcome the uncertainty associated with the unknown. In addition, findings show that living in regions where interpersonal trust is high may create a positive context of perception toward immigrants and influence an individual toward a more positive view.

A prominent aspect of the social capital theory is the aggregate value of citizen participation in associations and organizations, social ties and networks, civic engagement, trust, and norms of reciprocity. After all, the whole point of social capital is to facilitate collective action. It is usually bandied as a cure for commons-type problems, such as littered parks and crime-infested streets, but there is no reason to assume that a part of society will deploy its social capital only for wise and benevolent ends. A neighborhood, for instance, that is adept at organizing litter patrols and crime watches will also likely be adept at organizing opposition to real estate developments that threaten to add economic or racial diversity. In this respect, the forces that bind communities together may also have a dark side, the so-called “blocking capital” (Portes and Landolt, 1996), i.e., the use of community cohesion to reinforce the exclusion of outsiders. In other words, social capital can act as a double-edged sword; social capital within a group is often built by excluding those outside that group.

The study of Côté and Erickson (2009) explored this aspect of social capital in studying the relationship between social networks, associations, and tolerance. By constructing an index of network and association diversity, the authors aim to unmask the potential negative aspects of social capital. Findings from the analysis of the 2000 Canadian federal election show that different forms of network diversity and voluntary associations have very different relationships with tolerance. Diversity of ties to middle-class people is found to be related to greater tolerance and is consistent with the higher education levels of middle-class people and their lower degree of economic competition with minorities, both of which contribute to higher tolerance. Diversity of ties to working-class people is associated with lower tolerance and is consistent with their lower education levels and greater economic competition with minorities. Voluntary associations exhibit more complex picture. Associations with well-educated members or a relatively high proportion of minority members positively relate to tolerance because they include kinds of people disposed to tolerance in contrast to associations with poorly educated memberships. However, often the associations as such have no effect at all.

Our study adds to the literature by considering more dimensions of social capital than the aforementioned studies. Along with other contextual factors, our paper explores the role of social capital in shaping immigration sentiments, bringing evidence for more than a decade and for a large panel of countries.

More particularly, we consider three wide dimensions of social capital, as proposed by Coleman (1988): (i) trust in people and institutions; (ii) interpersonal networks of informal type, such as meeting with friends, relatives, and colleagues; participating in associations, including churches; and voluntary work (so-called Putnam-type social activities) or formal type, such as work for political parties or professional organizations among others (so-called Olson-type social activities); and (iii) norms and effective sanctions. We include 13 aspects across the three dimensions of social capital to capture the multifacet nature of social capital. These aspects are trust in people, trust in institutions (trust legal system, state of health services, state of education system), networks (social interaction, church attendance, worked in political/action group, contacted politician/government, worn campaign badge/sticker, taken part in demonstrations, and member of trade union), and norms and effective sanctions (feel safe and victim of burglary/assault).

The ESS, a large-scale biennial survey of attitudes and values consisting of seven rounds/ waves that date back to 2002, provided information for all 13 aspects of social capital we investigated. The ESS data were gathered in face-to-face interviews conducted in the native language of the interviewees. The interviewees were asked questions about various dimensions of social capital aspects and had to provide answers that ranged on a scale from 0 to 10, depending on the specific question. Information on social capital is available at http://www.europeansocialsurvey.org and described in detail in the Appendix.

In this section, we argue that social capital, in terms of social and institutional trust, networks, and norms and effective sanctions, has on average a positive dual role on immigration sentiments – a direct one, as expressed by H₁ hypothesis and an indirect one, as expressed by H₂ hypothesis, both presented in the following.

A crucial dimension of social capital is trust. Social trust is defined as trust in strangers, trust in people with whom we are not previously acquainted. Institutional trust is placing confidence in the working and mechanisms of society’s institutions. Trust creates bonds between people, influences individual outcomes as well as social group interactions, facilitates cooperation and efficiency, and enhances the quality of the political, legal, and institutional environment (North, 1990), increasing thus society’s overall effectiveness (Paxton, 1999; Paldam, 2001). Nevertheless, a more pessimistic possibility is that social trust could be confined just to strangers close to the truster’s cultural background (Hero, 2003). We will empirically test this idea in our article, but we argue here, following theoretical justifications, that people can develop expectations of trust about other people from different cultural and ethnic backgrounds. Rather than behaving from an imperative moral rule that people are trustworthy, the social truster is willing to offer cooperation to a stranger and then evaluates his or her behavior depending on the result. This makes the social truster to be more exposed to risk than the distruster, but the truster can also achieve higher pay-offs if his or her cooperative moves are reciprocated by the stranger. As a result, trusters can obtain “social intelligence” as a consequence of their cooperative moves and learn to discern better which people are trustworthy. Consequently, they tend to avoid the use of cues and heuristics based on racist or cultural stereotypes, because his or her “social intelligence” teaches him or her that these cues and heuristics are systematically biased (Yamagishi, 2001).

Networks is another important aspect of social capital. Interpersonal social networks such as meeting colleagues, friends, or family lead to a greater social embeddedness of the individual and to a strong feeling of belonging in the society and thus enhance trust and cooperation. Arguably, societies with citizens who are more engaged in the community life are tolerant and respectful of others and their rights and show altruistic preferences and positive attitudes on behalf of others and of the wider community (Mansbridge, 1999).

A more pessimistic possibility, however, is that social trust is confined just to strangers close to the truster’s cultural background.

The negative flip side of the contact argument is the competition argument. Whereas cooperative contact fosters tolerance, competition and lack of personal contact foster intolerance, and this can trigger a darker side of social capital. When ethnic groups compete in certain settings, they perceive each other as threats.

For example, occupational groups, including labor unions and business and professional associations, bring together people experiencing similar job pressures and tensions.

Finally, another aspect of social capital is that of solid norms and transparent and effective sanctions, which reduce the incentives for criminal action; individuals who do not feel afraid but feel safe in the surroundings they live in develop stronger ties within their community, feel less threat, and are likely to overcome the uncertainty associated with the unknown.

Overall, rich social capital portrays societies that are more equalitarian, cooperative, and integrated (Putnam, 2000) with better governance (Bowles and Gintis, 2002) and higher economic prospects (Knack and Keefer, 1997). Consequently, immigration and ethnic diversity would be less of a challenge to community cohesion in societies where social capital is present. Therefore, the first hypothesis we aim to test is the following:

The literature on public opinion on immigration has analyzed various versions of two main hypotheses: the “perceived threat” hypothesis, which argues that racial prejudice toward immigrants is a response to a perceived threat to the economic and political privileges of the dominant ethnic group in a society (Blumer, 1958), and the “contact” hypothesis, which, in contrast, considers that the presence of large populations of out-groups provides members of the dominant group with first-hand experience of different people, which, under certain conditions, can diminish racial prejudices (Oliver and Wong, 2003).

Social capital may have an additional, indirect role on shaping immigration attitudes in a society; it can mitigate typical “perceived threat” conditions in a society such as country’s economic conditions (Oliver and Mendelberg, 2000) or personal economic circumstances (Olzak, 1992), and relative size of the subordinate groups (Blalock, 1967). Even in situations in which cultural or racial prejudices would be more salient, people would feel threat due to their own conditions, or in times of economic upheavals when usually people feel threat of losing privileges or being heavily taxed, societies with high levels of trust in people and institutions, effective social networks, and solid norms and transparent sanctions will be more immune to the saliency of such prejudices. Based on this claim, we can test the second hypothesis:

The task of the following sections is to develop the model and econometric techniques to test for these two hypotheses.

Suppose that stated immigration concerns depend on latent continuous concerns about immigration. If y* denotes latent concerns and y the stated concerns, then the following holds:

where i and t denote country and time, x is a vector of covariates, and β is the vector of regression coefficients to be estimated.

Furthermore, suppose that while we cannot observe y*, we can instead observe the categories of response:

To capture the complex and multifaceted nature of sentiments toward immigration, we consider the dependent variable (y*) to be a vector of interviewees’ responses in six different statements: three about the effects of immigration on economic, cultural, and welfare prospects of the host country and three about the ethnic/race background of the immigrant.

The vector x is a set of covariates that includes sociodemographic, social capital, macroeconomic and foreign participation variables. More analytically, the set sociodemographic contains characteristics of the individual, namely, gender, age, education, marital status, health status, net income, domicile, employment status, religiousness, political orientation, and the origin (family roots) of the interviewee. The literature in sociology has identified some of these characteristics as important ones in shaping attitudes (Semyonov et al., 2006; Hainmueller and Hiscox, 2007). Arguably, one would expect that older, low-income, low-educated, unemployed, and extreme right in political orientation individuals, for example, would express more xenophobic opinions about immigration. This set of variables has been commonly used in all micro-level analyses with evidence, on their statistical importance, to vary across studies.

The next set of variables is what the literature calls social capital and here we include three important dimensions: (i) trust in people and institutions, (ii) interpersonal networks, and (iii) norms and effective sanctions. As it has been argued in previous sections, one expects a positive association of all three dimensions of social capital and attitudes toward immigration. Empirical evidence corroborates to the important role of trust in shaping positive attitudes toward immigration (Herreros and Criado, 2009; Rustenbach, 2010).

The set macroeconomic includes variables that characterize the economic environment of a country that could also shape public assessments. Arguably, public attitudes toward immigrants become more positive in good economic times and more negative in economic downturns.

At economic downturns, the fiscal cost of immigration becomes more an issue of concern and raises worries among individuals, who are either at the bottom of the income distribution and thus are threatened of immigrant competition in the labour market and welfare cuts or further up of the income distribution and therefore worry about the potential tax implications of immigration-induced expansion of the welfare budget.

The set foreign participation contains a number of variables relevant to the multicultural and ethnic diversity of the host country. The presence of foreign stock in a country and country’s colonial tradition may influence immigration attitudes as well. According to a simplified version of contact theory, increased contact with immigrants should undermine xenophobic sentiment; people who live next to well-integrated neighbors tend to have positive immigration sentiments (Ha, 2010). In contrast, threat theory argues that intergroup contact intensifies conflict due to competition over scarce resources; from this perspective, in countries with larger immigrant populations, there is more of a perceived group threat, leading to more antipathy toward the out-group (Hjerm, 2009). Empirical evidence, however, is mixed as to the relationship between the presence of foreigners and individuals’ perceptions toward immigration.

For example, Schlueter and Wagner (2008) found that between regions, a larger size of the immigrant population increases negative reactions, but within regions, more immigrants increase intergroup contact and reduce immigrant derogation. In a similar vein, Markaki and Longhi (2013) found that regions with a higher percentage of immigrants born outside the EU and a higher unemployment rate among the immigrant population show higher probability for negative attitudes to immigration. In contrast, Rustenbach (2010) found no evidence.

Our dependent variable, attitude of the individuals toward immigration, is ordinal; that is, we can rank the values, but the real distance between categories is unknown. Respondents state that immigration is “bad for a country’s economy” if their latent concern exceeds a higher threshold c1, “good for a country’s economy” if their latent concern exceeds a much higher threshold c2, and “very good for a country’s economy” if their latent concern exceeds an even higher threshold c3.

Immigration is considered “very bad for a country’s economy” if individuals’ latent concern is below threshold c1.

A critical concern in any empirical analysis is endogeneity, which produces biased and inconsistent estimates. Omitted factors such as trade collaborations, networks (Eichengreen and Irwin, 1998), and historic linkages (Anjonami and Hariri, 1992) between countries of origin and destination not captured in our estimations may induce endogeneity issues. To alleviate such concerns, we use country and year fixed effects. Furthermore, aspects of social capital, especially trust in people can be endogenous and bias our estimates. To account for endogeneity issues, we apply a control function approach (Wooldridge, 2015) that has the advantage of primarily dealing with nonlinear models as it is our case. In doing so, we regress the instruments on our endogenous variable in the first stage, conditional upon the other exogenous variables of the original model, and recover the predicted residuals of this estimation to plug them, at the second stage, into our original model (without excluding our endogenous variables from the model). Our inference is based on bootstrapping over all two-step procedure and for 1,000 iterations. Statistically insignificant predicted residuals would be an indication that there is no statistically significant bias at the coefficients of social capital due to endogeneity.

We use the variable “feel fair treated” as an instrument for our endogenous variable “Trust People”, which is an important dimension of social capital. The choice of “feel fair treated” is justified based on theory and statistics. On the theory front, both fairness and trust are crucial to our willingness to cooperate with others. Smith and Lindsay (2014) argued that leaders aiming to build trust in relationships need to pay attention to the issue of fairness. The direct interplay between the two was studied by Müller et al. (2008) who explored the causal link between procedural fairness and trust. Their results demonstrated that changes in perceived fairness of a procedure lead to changes in the willingness to entrust. This finding illustrates that trust judgments and trustful behavior can be influenced by procedural fairness judgments. On statistical front, a good instrument needs to be strongly correlated with the endogenous variable under instrumentation (i.e., “Trust People” in our case) and little or no related at all with the error term and the depended variable (i.e., immigration attitude or immigration policy). In other words, “feel fair treated” is a reasonable choice for an instrument if it can only be correlated with the dependent variable, immigration attitudes, through its effect on the endogenous variable “Trust People”. Indeed, the correlation between the instrument (“feel fair treated”) and the endogenous variable “Trust People” is 0.57 and is statistically significant at 5%, while the correlation between the instrument and the estimated error is 0.03 and is statistically significant at 5%.

Table 2 shows the results. Columns (i) to (iii) report odd estimates of equation (1) on the interviewees’ attitudes toward “immigration is bad or good for country’s economy” (economy), “country’s cultural life undermined or enriched by immigrants” (culture), and “immigrants make country worse or better place to live” (place to live), respectively. Heteroscedasticity-adjusted standard errors are reported in parentheses.

The estimates above are odd ratios. One can read the odd ratios as follows: if the odd ratio, a, is bigger than one (a > 1), then the probability of an individual to express very positive sentiments toward immigration (y_it = 4, i.e., maximum level) increases by (a − 1)*100%, whereas the probability decreases by (1 − a)*100%, if the odd ratio is smaller than one (a < 1). The coefficient of the control term included tells us furthermore that the latent factor captured by the instruments is correlated with our dependent variable(s). Note, however, that the bias is small and the control term is not statistically significant, so endogeneity does not impose a serious concern.

In terms of individual characteristics, the educational level of the individual matters the most. Higher levels of formal education tend to be associated with positive attitude toward immigration’s effects on economy (70.3%), culture (73.5%), and making a country a better place to live (65.3%).

Higher educated individuals are less likely to express prejudice, negative stereotypes, and racism toward minorities (Herreros and Criado, 2009) for at least two reasons. First, according to the labour market competition theory, as immigrants mostly work in low-skilled manual jobs, they are more likely to be complement rather than substitute to highly educated natives (Bogard and Sherrod, 2008). Second, the link between education and attitudes is deeply rooted in the fact that educational systems tend to promote acceptance of different cultural values and beliefs (Hainmueller and Hiscox, 2007).

The next set of estimates analyze the role of social capital. Among its three dimensions, it is the trust in people and institutions (legal, health, and education) that has the largest weight in shaping positive sentiments toward immigrants. It has been argued that social trusters have altruistic preferences (Mansbridge, 1999). Social trust implies an expectation that strangers are trustworthy, and this expectation is extended to all people, including immigrants from different cultural backgrounds. The “mechanism” that links social trust and positive attitudes toward immigrants is that trusters exclude heuristics and cues based on racial or cultural stereotypes when forming their beliefs about other people’s trustworthiness. As our results show, the coefficient related to trust in people is significant and in the expected direction. Social trust is associated with more positive attitude toward immigrants and their impact on economy (179.2%), culture (230.2%), and quality of living (215.7%) than social distrust. This result is also confirmed by the study of Herreros and Criado (2009). Likewise, institutional trust, especially trust in the legal system, shapes positive attitudes to immigration. Trust in the legal system associates with 15.9% higher probability of positive attitude to immigrants’ effect on economy (14% and 6.6%, respectively, on culture and quality of living). Satisfaction from the health and education systems also associates with 31.4% and 4.4%, respectively, more positive sentiments to immigrants’ effect on economy. Institutional trust increases society’s overall effectiveness (Paxton, 1999; Paldam, 2001), makes people more confident for the society they live in and the way the society copes with stressful situations. If immigration is such, then trusters in country’s institutions, all other variables considered, tend to express more welcoming attitude toward immigrants compared to institutional distrusters.

Social networks of informal type (meeting with friends, relatives, and colleagues; participating in associations including churches; and voluntary work) or formal type (work for political parties or professional organizations among others) are also found to positively associate with immigration attitude index. Individuals who are socially and, especially, politically active tend to express more positive immigration sentiments compared to individuals who are not. Among these activities, being member or working for political parties/action groups, participating in campaigns and lawful public demonstrations, significantly increases the probability of having a more positive attitude toward immigrants. The only exception is the association with trade union. In times of widespread unemployment, competition between indigenous and immigrant workers might increase, making inclusive union policies difficult to maintain (Penninx and Roosblad, 2000). However, trade union estimates are statistically insignificant. Solid norms and transparent and effective sanctions reduce the incentives for criminal action; individuals who feel safe and not being victimized develop stronger ties within their community and feel less threat. Consequently, absence of criminality is highly associated with positive stance toward immigrants. This conjecture is confirmed by our estimates of feeling safe that carry the expected sign and are statistically significant.

Overall, we find a strong, positive direct association of almost all aspects of social capital on immigration sentiments of the European citizens. This finding validates our first hypothesis (H1) that social capital associates with positive attitudes toward immigration across European countries.

We continue with macroeconomic conditions. We find that countries with higher GDP per capita (D_HighGDPcap) are 35% more probable to associate with positive attitude toward immigrants’ effect on economy than countries with lower GDP per capita. The fiscal cost of immigration also appears to be an important aspect of attitude. Countries with high debt (D_HighDebt) are about 45% more prone to associate with anti-immigration sentiments compared to countries with low debt. For example, immigrants who do not find employment are a fiscal burden on the welfare state and thus for natives of all ages. In contrast to the literature that provides weak or even mixed evidence on the link of macroenvironment on public sentiments, we find strong and clear results.

Empirical evidence on the macroeconomic conditions, however, is mixed, mainly weak and sometimes with perverse results. For instance, Semyonov et al. (2006) found that higher GDP per capita implies significantly more negative attitudes, while Schneider (2008) and Hatton (2016) supported the opposite relation. Other studies (Rustenbach, 2010) found no impact at all.

The stock of foreigners contributes to the multicultural profile of a country. The latter may stir up both positive and negative reactions: it can encourage the decrease of prejudice or can increase the perception of threat to national identity. Our results show that citizens in countries with many foreigners (D_{HighForeignStock}) and refugees/asylum seekers (D_{HighAsylumSeekers}) are about 50.9% and 57.5%, respectively, more prone to show anti-immigrant feelings compared to citizens who live in countries with more ethnically homogenous population and less refugees (Markus, 2014). For some people, perhaps, there can be a threshold beyond which they do not think favorably toward new immigrants. This negative association between the size of immigrant (foreign) population and immigration feelings is also documented in the literature (Oliver and Mendelberg, 2000). The number of nationality acquisitions (D_{HighNationalityAcq}) and the colonial tradition (D_{ColonialTradition}) of the country are not statistically significant.

Nevertheless, the negative association between multiculturalism and immigration attitude may alter once we control for the level of economic development of the country. The coefficient of the dummy D _{HighGDPcap&HighForeignStock} captures exactly this. What we observe is that countries with a high percentage of foreign population and high GDP per capita sentiments toward immigrants are more positive, especially when it comes to the economic impact, than countries with a high percentage of foreign population and low GDP per capita; a finding consistent with Figure 2. Wealthier countries have better economic and institutional mechanisms to provide for and integrate different ethnicities/races. Independent of economic conditions, however, individuals’ prejudice may remain against foreigners, and this is indeed the case as the coefficient of D_{HighGDPcap&HighForeignStock} turns negative in the other two specifications, losing, however, its statistical significance.

Having analyzed all factors (and their interactions) that could possibly shape immigration sentiments, we can now embark on testing our second hypothesis (H2) that social capital smoothes out the negative effects of adverse social conditions on attitudes toward immigration. That is, irrespective of a country’s social conditions, when we consider a country’s social capital (here, level of social trust), the picture that emerges may contradict with the xenophobic stance documented above.

According to our findings, in countries that host big numbers of foreigners or asylum seekers and are rich in social trust, the sentiments toward immigrants tend to be more xenophile than in countries with low social trust. The coefficient of D _{HighTrust&HighForeignStock} captures this positive association – larger than 1 in all specifications and statistically significant in the last one. This is because countries rich in social capital are more cooperative and egalitarian (Putnam, 2000; Bowles and Gintis, 2002) and, therefore, able to cope effectively with accommodating new comers. Social trusters are also 15.9% (D_{HighTrust&HighAsylumSeekers}) more prone to have a positive stance on the impact of asylum seekers on the quality of living. Similar evidence emerges when we revisit the individual characteristics and particularly the ones that relate to the “threat effect”. This time, however, we consider how different is a social truster individual who is unemployed (D_{HighTrust&Unemployment}) or a social truster individual who has a very low income level (D_{HighTrust&LowIncome}) from equal peers who are social distrusters. The results alter dramatically. Unemployed trusters are not negative at all to immigrants compared to unemployed distrusters. In fact, they are 13.2% and 17.5% more probable to have positive attitude on the immigrants’ impact on culture and quality of living, respectively. Likewise, low-income social trusters are about 10.3% more positive to immigrants than low-income distrusters. Such findings corroborate with the claim of our second hypothesis on the positive indirect association between social capital and individuals’ attitudes toward immigration.

Overall, our empirical analysis documents the important role of almost all aspects of social capital on public sentiments about immigration; the evidence on social trust corroborates findings of related studies of the past (Herreros and Criado, 2009; Rustenbach, 2010). The two testable hypotheses find strong support. Social capital has a positive association with immigration attitudes via two ways, a direct way, social capital associates with positive attitude toward immigration, and an indirect way, social capital moderates the negative effects of “perceived threat” on people’s opinions about immigrants. Individuals with low income, unemployed people, or citizens in multicultural and ethnically diverse environments, where the risk of intergroup conflicts is high, will nonetheless exhibit generally positive attitudes toward immigration if they are social trusters.

A potential issue in our analysis is that the average positive correlation between associations and immigration sentiments could mask important variation across different types of associations. Erickson (2004) and Côté and Erickson (2009) argued that associations that are heterogeneous tend to relate to more social tolerance than homogenous ones. They proposed and constructed a measure of network and association diversity based on income class, occupation, and gender of their members. As we do not have information on the composition of associations and on particular types of associations (e.g., ethnic associations for minority ethnic groups) for the European citizens, we can only derive nuanced conclusions.

The literature argues that the specific consequences of social capital largely depend on political and social conditions, the existence of inequality and poverty, and liberal democracies, to name a few (Gargiulo and Benassi, 1999; van Deth and Zmerli, 2010; Villalonga-Olives and Kawachi, 2017). In our analysis, we included most of these factors to control the effect of social capital on immigration attitudes.

In this section, we turn our attention into examining whether sentiments toward immigrants differ depending on immigrant’s profile, that is, race/ethnicity and country of origin. Different attitude toward different types of immigrants reveals the concerns of the Europeans toward certain cultures, and it is an indication of individuals’ perception toward the immigration policy (stricter or looser toward certain immigrant profiles) their country should apply.

Table 3 shows the results. Columns (i) to (iii) report the odd ratios of equation (1) on the respondent’s attitudes toward “allow many/few immigrants of same race/ethnic group as majority” (same race), “allow many/ few immigrants of different race/ethnic group from majority” (different race), and “allow many/few immigrants from poorer non-European countries” (poorer/ non-EU), respectively. Heteroscedasticity-adjusted standard errors are reported in parentheses.

All sociodemographic variables that were important in the previous analysis pertain their statistical significance carrying the expected sign; however, in some cases, their coefficients change across different immigrant profiles as they become more negative to immigrants from different race/ethnicity and especially to those from poorer non-European countries; noteworthy exception is the highly educated and religious individuals who actually become more receptive to very different racial/ethnic background.

Furthermore, all three dimensions of social capital continue to greatly contribute to the xenophile attitude of the individuals, a finding that supports our first hypothesis. Our results show that when social capital is present, there is no differentiation at all across different origins of immigrants.

Macroeconomic conditions do not seem to shape Europeans’ sentiments favorably toward any group of immigrants, in particular as estimates tend to be statistically insignificant. Perhaps, macroeconomic conditions do not directly relate to feelings about the ethnic background of the immigrants but rather to their skills. Unfortunately, we have no information about the skills of the immigrants relative to those of natives. This information could shed some light as documented in the study of Mayda (2006). When it comes to the statistical significance of the interaction terms of social capital with the macroeconomic variables, the same picture emerges. Perhaps, this is because we have no deeper information about the social capital that would allow us to unfold different reactions based on the ethnic background of the immigrants. For instance, information on associations of ethnic minorities, their contact with the natives, and their embeddedness into the society would shed more light why native citizens are more (or less) friendly to different immigrant backgrounds. This requires detailed information that is not available at the moment. Future research could focus further on these issues, disentangling potential diversity in the effects of social capital.

When it comes to multiculturalism, it seems that it does not pose any threat when social trust is present as the coefficient of D_{HighTrust&HighForeignPop} indicates. In such societies, immigrants of all types are welcome. Furthermore, unemployed trusters do not actually differentiate among immigrants. Same holds for low-income social trusters. Both findings are in line with the claim of our second hypothesis and therefore validate it.

Finally, endogeneity is not a serious concern here as the control term is statistically insignificant.

In sum, a consistent finding that emerges is that public attitude toward immigrants appears to vary across different immigrant profiles. Europeans, on average, are slightly more xenophile compared to immigrants of the same race/ethnicity. Social capital tends to associate with pro-immigration attitudes and diminishes any differentiation across immigrant profiles.

For robustness purposes, we make an effort to instrument for the social capital as a whole and not just for the societal trust. In doing so, we construct, using factor analysis, an index of social capital based on all of its individual components. Then, we instrument for it. This is not an easy task, though. To find appropriate and exogenous instruments, we resort to the economic growth literature. Potential candidates would be historic data (the past affects the present, but not the other way around) about a country’s literacy rates and dominant religion in the ninetieth century (i.e., as early as we have data from the establishment of the modern nations). Historic educational endowments represent a good predictor of current institutional differences across countries of Europe. As Glaeser et al. (2004) had shown, historic levels of schooling are robustly correlated with changes in political institutions and economic development. Furthermore, Tabellini (2010) had demonstrated that literacy in 1880 represents an important source of variation for the cultural traits and level of social capital of European countries. Religion has also been a proxy for culture and widely used in the finance and economics literature (La Porta et al., 1999) as it is a key component of a system of beliefs. For instance, the literature argues that the way a religion is organized, i.e., hierarchical vs. equalitarian structures, reflects on how the society and government are structured (centralization vs. individualism; Stulz and Williamson, 2003).

We therefore adopt the same instrumental variables, i.e., countries’ literacy rates (percentage of population) and dominant religion (e.g., Catholic, Protestants, Greek Orthodox, Muslim) in the 1900s as an exogenous determinant of current social capital quality. Data are retrieved from the following sources: religion is obtained from the World Factbook (2019) and literacy rates from Roser and Ortiz-Ospina (2020). In addition, we also include, along with these two instruments, the “feel fair treated” variable to allow for individual (and not only country) variation in our instruments. We then apply the control function approach (Wooldridge, 2015). That is to say, we regress the instruments on our dependent variable in the first stage, conditional upon our instruments, and recover the predicted residuals of this estimation to plug them into our original model (without excluding the constructed index of social capital); inference is based on bootstrapping over all two-step procedure, 1,000 iterations. The results of this process are shown in Table A3 of the Appendix.

The coefficient of the control term tells us furthermore that the latent factor captured by the instruments is correlated with our dependent variable(s). Hence, endogeneity seems to cause a small bias in the social capital coefficients in our previous estimates. Note, however, that the bias is small and the control term is not significant, so the main conclusions of the analysis undertaken, so far, hold; that is, social capital positively relates to immigration sentiments and further mitigates the negative effects of perceived threat on people’s opinions about immigrants.

We also performed a battery of sensitivity analysis to sharpen the robustness of our results. We split our sample in two sub-periods, 2002–2008 and 2010–2014, to test whether financial crisis had an impact on public sentiments about immigration. The results did not alter significantly.

We also included various control variables, namely, the Gini coefficient for inequality and welfare benefits (% of GDP). The estimates are statistically insignificant, and the rest of the results are barely modified.

Overall, the results do not change in any significant way across different specifications, subsamples, and alternative definitions.