Are cosmopolitan dispositions learned at home, at school, or through contact with others? Evidence from young people in Europe

ABSTRACT There has been a resurgence of interest in cosmopolitan theories of citizenship over the past two decades, as academics and policy-makers have sought to understand and conceptualise citizenship affiliations that transcend the nation-state. As part of these debates, a number of theories have been developed to try to explain the emergence of cosmopolitan dispositions, and in particular, why some citizens develop these dispositions and others do not. This article seeks to refine these theories by testing their assumptions on a youth sample, by drawing on the findings of the wider youth socialisation literature, and by comparing the relative merits of different ‘sources’ of cosmopolitanism (namely cognitive engagement; contact with cultural Others; learning a foreign language; and exposure to discussions about international issues in the public and the private sphere). To do so, the analysis draws on data from the 2009 International Civic and Citizenship Education Study (ICCS) and its survey of young people aged 14. Using these data, this article will show that each of these theories are positively associated with cosmopolitan sources, but that the strength of these relationships is variable and that not all sources of cosmopolitan dispositions are equal. Based on these findings, this article will argue that schools can play an important role in cultivating cosmopolitan dispositions at this particular life stage.


Introduction
Over the past 20 years, traditional conceptions of citizenship have been challenged as we have sought to come to terms with the theoretical and practical implications of globalisation, technological innovation, and increased mobility and migration. Historically, citizenship was seen as almost inextricably linked to the nation-state (Turner 2006), but this relationship has had to be re-considered as societies have become more diverse and individuals are more likely to come into contact with people from other cultural, ethnic, and religious backgrounds as a result of increased migration, travel, and/ or internationalised media. At the same time, the institutional and legal frameworks are also shifting, and individuals are increasingly governed not just by national laws and institutions, but also by regional and international governance frameworks. Awareness of these developments has triggered wide-ranging discussions about how we characterise citizenship and its sources (legal, affective, structural, and otherwise), and in the wake of these debates, it has become commonplace to conceptualise citizenship as a multiple, hybrid, nested, and/ or flexible construct, particularly when it comes to citizens' identities (Checkel and Katzenstein 2009). The extent to which these multiple citizenships exist among the general public continues to be debated (see Fligstein, Polyakova, and Sandholtz 2012), but it is nonetheless now accepted that citizens' affiliations can, at least in theory, transcend the nation-state and extend to other political or cultural communities.
The effort to understand and conceptualise citizenship affiliations that transcend the nation-state has, in turn, prompted a re-examination of cosmopolitan theories of citizenship and raised questions about how and why citizens develop cosmopolitan attitudes and dispositions. Much of this debate has been theoretical and normative (Calhoun 2008), but more and more empirical research is being developed to try to test the competing claims that have emerged. Despite this, much remains unknown about how or why these dispositions emerge. This article therefore seeks to contribute to these debates by testing four different theories of cosmopolitan disposition development among young people. This group is of particular interest, as young people are more likely to report cosmopolitan attitudes and identities (Woodward, Skrbis, and Bean 2008;Norris and Inglehart 2009), but there is a dearth of comparative research of cosmopolitan dispositions at this age or life stage (Hahn 2010, 17). Furthermore, the benefits of focusing the analysis on young people may not be limited to this age group; examining their formative experiences may help us to better understand the development (or lack thereof) of cosmopolitan dispositions in other age groups.
As we shall see in the next section, a number of theories have been developed to try to explain the emergence of cosmopolitan dispositions. This article seeks to refine these theories by testing their assumptions on a youth sample, by drawing on the findings of the wider youth socialisation literature, and by comparing the relative merits of different 'sources' of cosmopolitanism (namely cognitive engagement; contact with cultural Others; learning a foreign language; and exposure to discussions about international issues in the public and the private sphere). To do so, the analysis draws on data from the 2009 International Civic and Citizenship Education Study (ICCS) and its survey of young people aged 14. The analysis will focus in particular on European countries, and use European citizenship as a case study and exemplar of a cosmopolitan society. This focus has been selected for theoretical and practical reasons. First, European integration provides a clearly defined citizenship project that transcends the nation-state; as a result, it is an apposite site in which to explore empirical manifestations of cosmopolitan citizenship (Norris and Inglehart 2009;Pichler 2009, 707). Second, while European policies have often relied on a bounded version of cosmopolitanism, individuals do not necessarily adopt the same exclusionary approach; as we shall see, young people who display favourable dispositions towards European integration also tend to display interest and trust in international issues and institutions. Finally, the European Regional Module (ERM) of the ICCS data set includes unique variables that allow us to empirically test some of the key theories about the sources of cosmopolitan dispositions among young people.

Defining cosmopolitan dispositions
Theories of cosmopolitanism can be traced back to Ancient Greece (Heater 2004), but the concept remains fuzzy and contested, and there are a wide array of different ways of defining, theorising and researching this concept (see, e.g. Calhoun 2008;Woodward, Skrbis, and Bean 2008;Delanty 2012;Pichler 2012). Indeed, at times it appears that this term has been used to describe any type of citizenship that transcends the nation-state. However, a unifying theme is that this term refers to the emergence (real or ideal) of a world or universal community that shares values, political institutions, identities, and/or rights and responsibilities but also embraces cultural diversity (see Delanty 2009Delanty , 2012 for a more comprehensive discussion). This central tenet is reflected in the research on cosmopolitan dispositions at the individual level, in which the core characteristic is commonly conceptualised in terms of an openness to and appreciation of other cultures, values, and experiences. As evidence of this central characteristic, empirical research has tended to focus on examining whether citizens any of have the following dispositions: . a sense of belonging to/identification with an international community or community larger than the nation-state; . trust and tolerance of immigrants, ethnic minorities, and cultural Others; . support for equal rights for immigrants; . respect for human rights; . an interest in cross-cultural consumption of goods, culture, or media; and . support for international governance and/ or the globalisation of economies and societies (see, e.g. Woodward, Skrbis, and Bean 2008;Gustafson 2009, 30;Norris and Inglehart 2009;Pichler 2012). It is generally not assumed that individuals must exhibit all of these dispositions in order to be considered cosmopolitan citizens. Moreover, this list should not be seen as exhaustive, as the conceptual framework continues to be contested (see Delanty 2012). However, this list serves to show the ways in which scholars have tried to operationalise cosmopolitanism at the individual level, and that such measures reflect some of the different dimensions (political, cultural, ethical/ normative, and practical) of this complex construct.
Although the concept of cosmopolitanism is ancient, the emergence of these dispositions in the contemporary context is generally seen to be a recent phenomenon. As noted above, a nation-centric model of citizenship has dominated political communities, at least until very recently, and in many ways this governance model relied on cultivating dispositions among its citizens that represent the very antithesis of some of these cosmopolitan attributes (Keating 2014). The emergence of cosmopolitan citizenship models and dispositions is thus often characterised as a function of (or desired response to) the recent acceleration of global and regional integration. However, as the spread and depth of globalisation varies across countries and regions, cosmopolitan dispositions are far from universal, and one of the key questions that the latest empirical research has sought to address is which individuals or groups are more likely to exhibit these dispositions and why.
In the process, several common findings have emerged about the characteristics of citizens with cosmopolitan dispositions. First, it is commonly found that young people are more likely to exhibit cosmopolitan values, identities, and behaviours, particularly in their (positive) attitudes towards the impact of globalisation on personal consumption (Woodward, Skrbis, and Bean 2008, 221). This may reflect a life-cycle phase rather than a wholesale generational shift (Jung 2008), but it nonetheless suggests that age is an important variable. More broadly, individuals are also more likely to exhibit cosmopolitan dispositions if they are female, highly educated, high-income earning, working in a professional occupation and/ or living in an urban area (Gustafson 2009;Pichler 2012, 25). Socioeconomic status (SES) is a particularly prevalent predictor in these empirical studies, and there is a general consensus that the presence or absence of cosmopolitan dispositions is highly stratified. That said, cosmopolitanism is not just the preserve of the wealthy, the well educated, or the young. Religion, ethnic background, and political orientation have also been found to be statistically significant for some of these dispositions (Woodward, Skrbis, and Bean 2008;Pichler 2012, 25). Cosmopolitan citizens are also likely to travel abroad regularly for work or pleasure; to have frequent contact with friendship and family groups in other countries; to be proficient in the English language; and to regularly consume international news and cultural media (Weenink 2007, 495;Norris and Inglehart 2009).

How are cosmopolitan dispositions acquired?
Several theories have been developed to try to explain these patterns, but many of the most prominent and comprehensive theories suggest that the transmission of knowledge, understanding and values is central to understanding why some people have cosmopolitan dispositions while others do not. In brief, the transmission of knowledge ensures that individuals are better able to understand the issues at hand, are less likely to feel threatened by these developments, and better equipped to evaluate and navigate the new social, economic and political contours of this increasingly interconnected world (see, e. g. Inglehart 1970;Gabel 1998, 335). But while knowledge is seen as a necessary condition, it is not usually seen as a sufficient one. At the same time, values also have to be transmitted to, and internalised by, the individual (Inglehart 1970). The transmission of knowledge contributes to this process also, as the new information that is received challenges individuals' perceptions of their worlds and enables them to develop new meanings, understandings, attitudes, and identities (McIntosh, Hart, and Youniss 2007, 497).
These themes are common to several theories and what distinguishes them most is the varying emphasis placed on how knowledge, understanding and/or values are transmitted. In cognitive engagement theory (perhaps the most well known and well established of the theories that have emerged thus far) the requisite knowledge and skills are largely acquired as individuals progress through the formal education system. Empirical tests of this theory often focus on the educational level or qualifications of respondents and their general education (Pattie, Seyd, and Whiteley 2004), but recent research has started to try to identify which specific educational and/or life experiences foster the development of cosmopolitan dispositions (Bellucci, Sanders, and Serricchio 2012). For example, Verhaegen, Hooghe, and Meeusen (2013) suggest that schools provide cognitive resources not just through general education, but also by providing 'cognitive', classroombased opportunities for learning about other countries. This latter finding thus leads us to the first hypothesis, namely that young people are more likely to report cosmopolitan dispositions if they attend schools that provide opportunities to acquire information about social, political, and cultural issues in other countries (H1).
A second school of thought posits that knowledge and values are transmitted by contact with other cultures and communities. These 'contact' theories stem from the fact that individuals are more likely to report cosmopolitan dispositions if they travel abroad frequently and/or have lived abroad; or if they have a network of friends, relatives or work contacts that live in other countries and that they interact with frequently (Mau, Mewes, and Zimmermann 2008). These experiences, it is surmised, provide individuals with knowledge, experiences, and social contacts beyond their national or local communities, which in turn makes them more accepting of other cultural practices and more conscious of the potential benefits of global integration and a cosmopolitan community (Fligstein 2009;Fligstein, Polyakova, and Sandholtz 2012;Gustafson 2009;Ceobanu and Escandell 2010, 317). These theories thus echo Allport's (1954) theory of inter-ethnic contact (see also Pettigrew and Tropp 2006), which contends that inter-cultural interaction increases mutual understanding and reduces intergroup prejudice. 1 In light of this, our second hypothesis is that young people are more likely to report cosmopolitan dispositions if they have participated in school exchanges or other activities that have brought them into contact with people from other countries and/or cultures (H2).
The potential benefits of inter-cultural contact are likely to be linked to participants being able to communicate with one another in a common language. However, it has been posited that foreign language competence not only facilitates the exchange of information between individuals and cultures, but that the very act of language learning can also have 'spill-over' effects that change individual's social identities and encourage positive attitudes towards cultural Others (Byram 2008, 29, 127). Although there is some theoretical and empirical evidence to support these claims, the evidence base remains weak, in part because of the dearth of cross-national comparative data (Byram 2008, 140;Fligstein 2009, 147). The third proposition for this article, then, is that young people with high levels of foreign language competence are more likely to report cosmopolitan dispositions (H3).
The fourth and final proposition considered here is that young people are more likely to report cosmopolitan dispositions if they engage with media, parents' and/ or peers' discussions about international issues (H4). This hypothesis combines two theories that explore the relationship between civic attitudes and political discussion in the public and private spheres. We know that the media, for example, can shape public opinion by providing 'cues' about the appropriate attitudinal, affective, or electoral response to social, political, and civic issues (see, e.g. Hobolt 2009). Indeed, Norris and Inglehart (2009) suggest that simply being exposed to news can in and of itself encourage the development of cosmopolitan dispositions. Drawing on World Values Survey data, they argue that people who follow international and national news stories are more likely to report trust in and tolerance of outsiders, as regular exposure to news heightens awareness and understanding of other countries and cultures (at least in societies with media freedom) (Norris and Inglehart 2009, 12).
When considering attitude formation among young people, however, one must also consider the role of parents and peers. Parents, in particular, not only provide their children with information, but they also serve as role models of civic engagement and attitudes (Wilkenfeld 2009) and 'moral guides' that interpret events for their children (Flanagan 2013, 195). These civic attitudes and values are then internalised by young people (although one cannot, of course, assume that parental values will be automatically transferred to their children in a top-down fashion; McDevitt and Chaffee 2002). Political discussions with parents appear to play a particularly important role in this process; empirical studies have shown that youth discussions with parents are positively associated with political knowledge, news consumption, community service, and trust (McIntosh, Hart, and Youniss 2007;Flanagan 2013). Discussions with peers complement this process; they act as sounding boards and mediators of knowledge, and in this way may raise the salience of cosmopolitan issues and the acceptance of diversity (Flanagan 2013, 18).

Data and methods
To test these propositions, cross-national data were drawn from the 2009 ICCS and its supplementary ERM in which European youths were asked a series of questions about: their knowledge about the EU; their attitudes towards and affective ties to Europe; and their sources of information about the EU and Europe more generally. Twenty-four European countries participated in the ERM, but for the purposes of this article, analysis was limited to 20 countries: namely Austria, Bulgaria, Cyprus, Czech Republic, Denmark, England, Estonia, Finland, Greece, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Poland, Slovakia, Slovenia, Spain, and Sweden. 2 In total, 57,693 students from these countries participated in both the ERM and the ICCS surveys. In each of these countries, the participants were selected using a two-stage cluster sample design. First, schools were selected using PPS proceduresthat is, probability proportional to size as measured by the number of students enrolled in a school. Second, in each sampled and participating school, one intact class was randomly selected from the target grade. The target grade was usually grade 8, which was selected with a view to ensuring that the students participating in the study were approximately 14 years of age (see Schulz, Ainley, and Fraillon 2011, Chapter 6).
Data that is collected in this way are hierarchical, and this was born out by the intraclass correlations in the null models (discussed below). The data were thus analysed through a series of multi-level random intercept models where students (level 1) were nested within schools (level 2) and schools within countries (level 3). Multi-level linear models of this sort can be seen as an extension of regression, as it enables us to also take into account the fact that students are more likely to have similar traits and outcomes as their fellow classmates and, at the national level, as other students in their country (Goldstein 1999). The results of this analysis are discussed below, and in presenting these results, the discussion examines not only the statistical significance of the relationships, but also the comparative strength of the relationships (or what is sometimes known as the 'effect' size). Including these dimensions in the discussion of the results is particularly important because of the size of the ICCS sample; in large samples such as this, even small differences may show up as being statistically significant and thus the importance of these differences may be overstated (Strand 2004, 43).
Many of the variables used here were measured on different scales, making direct comparisons of their 'effects' difficult. Therefore to facilitate a comparison of the relative strength of these relationships, this article adopted Schagen and Schagen's (2005) post hoc method of normalising or standardising the coefficients, as this strategy allows us to review both the raw coefficient and the normalised coefficients. Following Schagen (2004), the coefficients were normalised using the formula n = 100 * b * s/S where β is the co-efficient, s is the standard deviation of X (i.e. the predictor of interest), and S is the standard deviation of Y (i.e. the outcome indicator of interest). In other words, a normalised coefficient (n) reflects the expected change in the standard deviation of the outcome (Y ) following one standard deviation change in the predictor (X ). The resultant 'normalised' coefficient is expressed as a percentage, and is not dissimilar from other methods of calculating effect sizes (Schagen and Schagen 2005, 315). For the purposes of this analysis, then, this approach was applied to all continuous variables in the models; for discrete variables (and again following Schagen (2004)), the formula was adjusted so that the normalised coefficient instead represents the expected change in standard deviation of the outcome (Y ) following a change in the predictor (X ) of one unit.

Variable selection
As noted above, cosmopolitan citizenship has been linked to a wide range of dispositions. In this article, however, we shall focus on just two of these dispositions, namely reporting a cosmopolitan identity and holding positive attitudes towards equal rights and opportunities for immigrants. Following Pichler (2009) and Norris and Inglehart (2009), this article is using the European context as a case study of cosmopolitanism, and the outcome variables thus reflect this. In particular, this article is taking students' sense of European identity as an exemplar of one type of cosmopolitan identity, and their attitudes towards equality of opportunities for European citizens as an indicator of their attitudes towards equal rights for all immigrants. It could be argued that European identity (and indeed, European citizenship more broadly) is too particularistic and regionally specific to be considered as an exemplar of a cosmopolitan identity. And indeed some European policies have sought to build exclusionary boundaries that are contrary to the core principles of cosmopolitanism (Benhabib 2002;Author 2014). However, data from European citizens suggest that cosmopolitanism and Europeanisation are mutually supportive rather than mutually exclusive, and Pichler (2008) found that Europeans who hold cosmopolitan attitudes are more likely to support European integration and to report nested identities that span the national to the global. Similar patterns are apparent in the ICCS data being used here. This data set does not include a measure of global or cosmopolitan identities, but preliminary analysis for this paper revealed that there is a positive correlation between student attitudes towards European and global institutions, student interest in European and international news, and student attitudes towards cultural Others regardless of whether they are from Europe or elsewhere. While European and cosmopolitan identities should not be conflated, this pattern supports the case for viewing European identity as an exemplar of one type of cosmopolitanism, and using this construct as a means of exploring how cosmopolitan dispositions more generally are formed.
In light of this, the analysis relies on two factors that emerged from principle component analysis (PCA). The first summarises student responses to six items about their attitudes towards European identity (Cronbach's α = 0.82), while the second summarises five items about student attitudes towards equal opportunities being extended to European citizens when they move to the student's home country (Cronbach's α = 0.844). A full list of the items that comprise these factors is provided in Appendix 1. The null models of these outcome indicators confirmed that the data are hierarchically structured and that multi-level modelling is appropriate. For our measure of European Identity, 6% of the variance was found at the country level and 5% at the school level. In the case of our Equal Opportunities measure, the corresponding figures were 2% and 4%. A variance of 2% is very low (Duncan and Raudenbusch 1999, 33) 3 but the three-level models were retained to facilitate comparability.
PCA was also instrumental in creating a summary measure for one of the four key predictors of interestnamely learning about international issues at school. In this case, the resultant factor includes six inter-related items that capture students' perceptions of the opportunities their school offers them to learn about European topics at school. The Cronbach's α for the resultant factor was 0.790, and the constituent items are listed in Appendix 1. The items in question focus on learning about European topics rather than international issues more broadly for both pragmatic and theoretical reasons. Not only is the European Union the case study for this analysis, but the main study of ICCS does not include any data on the type of international issues that are taught in schools, whereas the ERM asked students about the type of European topics that they learn about at school. This variable is still limited, as it relies on self-reported data from students, but at the present time there is no independent and comparable measure of the international or European topics that students learn about at school.
To operationalise the contact hypothesis and the discussion hypothesis, the analysis instead relied on two 'ready-made' summary variables that were part of the original ICCS data set. These two variables (namely EUPART and EUROCOM) were created using Item Response Theory (IRT) which, much like PCA, summarises student responses to multiple items that are conceptually related and provides a single overall measure for use in the analysis. In these variables, however, the scale is set so that each has a mean of 50 and a standard deviation of 10 (see Kerr et al. 2010;Schulz, Ainley, and Fraillon 2011). Again, a full list of the items that were included in each variable are listed in Appendix 1, but in short, EUPART summarises eight items that capture students' engagement in activities that involve contact with other European cultures or people from other European countries (e.g. through school trips, holidays, exchange programmes, and/ or art, music or sporting events at home or in other countries). EURCOM, for its part, composed of 10 items that reflect students' frequency of engagement with European news and in discussions with parents and peers about various European topics (social, cultural, and political). 4 In comparison, the foreign language competence hypothesis relied on a more simple measure. As part of the ERM, students were asked if they spoke a second European language, and if so, if they were able to communicate in this language well, very well, or not very well. Based on this, dummy variables were created to facilitate comparisons between individuals with high, medium, and low levels of language competence in other European languages.
Descriptive statistics for each of these key predictors are provided in Table 1, while a more detailed discussion of the descriptive statics for each of the individual components can be found in Author (2014) and Kerr et al. (2010). In short, these data indicate that, on average, young Europeans exhibit high levels of European identity and high levels of support for equal opportunities. European identity is the highest in Italy, Slovenia, and Spain, and the lowest in Latvia, England, and the Czech Republic. For the principle of equal opportunities for other Europeans, support is highest in Italy, Spain, and Greece, and lowest in Latvia, Austria, and the Czech Republic (Kerr et al. 2010, 132-138). Students in Bulgaria, Italy, and Malta are mostly likely to report learning about Europe at school, while students in Luxembourg reported the highest level of contact activities and students from Bulgaria reported the lowest (see Kerr et al. 2010, 133-134 and 119). In all of the ERM countries, participation in discussions about Europe (EURCOM) was relatively infrequent. Amongst the cases under review here, students in Bulgaria and Italy reported the highest frequencies, while students in England and Finland recorded the lowest (Kerr et al. 2010, 110). Finally, in terms of language competence, students in Luxemburg, Greece, and Denmark were most likely to say that they could communicate very well in another European language, while at the opposite end of the spectrum, the highest proportions of students reporting low proficiency were in Ireland, Lithuania, and England. In addition to these key predictors, a number of controls were also included in the analysis (see Model 1 in Tables A1 and A2). In particular, each model included key background variables such as gender, immigrant status, and SES, which, as noted above, previous studies have consistently found to be related to cosmopolitan dispositions. SES was measured using a summary variable (NISB), which combines the number of books in home with the highest occupational and educational status of parents (see Schulz, Ainley, and Fraillon 2011, 193). To account for the impact of more 'traditional' cognitive resources (see Pattie, Seyd, and Whiteley 2004), the models also included a measure of students' political knowledge and of their expectations of how long they would stay in education. The public opinion literature suggests that political pre-dispositions towards European integration may also shape attitudes towards the outcomes of interest, and hence the models also included controls for higher levels of interest in European political issues and of trust in the European Parliament (the latter being a proxy for EU support; McLaren 2006). In both cases, dummy variables were created to distinguish between higher and lower levels of interest and trust. Following Hooghe and Marks (2004, 415), the European identity model also took attitudes towards the nation-state into account (in the form of IRT-derived variable ATTCNT), whereas it was assumed that student attitudes towards equal citizenship opportunities were more likely to be influenced by their attitudes towards European identity (see Author 2014), and thus in this model, the European identity measure was included as a control rather than ATTCNT. Finally, the latter model also included a measure of students' general attitudes towards immigrants (namely IRT-derived variable IMMRGHT), to account for the possibility that student responses towards these items about European citizens were shaped by their attitudes towards immigrants in general, rather than their exposure to the predictors of interest.
At the school level, the models also included measures of school achievement and of classroom climate; the latter is particularly relevant as it has been found that students are more tolerant of others if they experience an open school climate in which they feel able to openly investigate issues and to discuss controversial issues with their teachers and peers (Flanagan et al. 2007). At the national level, the models included a measure of average public perceptions of the EU, 5 to control for the possibility that public attitudes towards the EU may frame the debate and trickle down to influence youth attitudes (Diez-Medrano 2003;Verhaegen, Hooghe, and Meeusen 2013). The fruits of this analysis are presented in the next section.

Results
Starting with the first hypothesis, this analysis suggests that there is a significant and positive relationship between learning about European issues at school and both of the outcomes of interest, namely students' sense of European identity and their attitudes towards equal opportunities for citizens from other European countries. What is more, in the case of European identity, the normalised coefficients suggest that the relationship is relatively strong; an increase in one standard deviation of learning about European issues at school is associated with a change in European identity equivalent to 15% of a standard deviation (see Table 2). In comparison, the change in student attitudes towards equal opportunities is weaker, but still notable. In this case, an increase in one standard deviation of learning about European issues is associated with a change in student attitudes equivalent to 9% of a standard deviation (see Table 3).
These patterns are echoed in the results from our second hypothesis, namely that contact with cultural Others increases cosmopolitan identities and attitudes. In line with expectations, there is a positive and significant relationship between student participation in cross-cultural European activities and their sense of European identity and their attitudes towards equal opportunities. However, a review of the normalised coefficients indicates that these relationships are relatively weak, particularly when compared with the effects associated with learning about Europe school. In this case, an increase in one standard deviation of contact activities is only associated with a change in European identity equivalent to 8% of a deviation (see Table 2) and with a change in equal opportunities attitudes equivalent to only 2% of a standard deviation (see Table 3). Our third hypothesis posited that language competence was related to cosmopolitan dispositions, and again, there is evidence to support this claim. In keeping with our assumptions, students who reported medium or high levels of competence in a second European language were more likely to report having a sense of European identity (see Table 2), and students with high levels of competence were also more likely to report positive attitudes towards equal opportunities (the relationship with medium language competence was not significant; see Table 3). As with our first hypothesis, the relationship between language competence and European identity is stronger than its relationship with student attitudes towards equal opportunities. Students with high levels of language competence are likely to have a European identity score that is 15% of a standard deviation higher than students with little or no competence in a second European language. When students have medium levels of competence, the equivalent change in European identity is 7% of a standard deviation. In the case of student attitudes towards equal opportunities, only higher levels of language competence are significantly associated with this outcome, and in this case, the effect size is around 9% of a standard deviation.
There is less evidence, however, to support our fourth and final hypothesis, which posited that there would be a positive relationship between cosmopolitan dispositions and engagement with media, parents', and peers' discussions about international issues. For one, these data suggest there is no relationship between discussions about European issues and student attitudes towards equal opportunities for other European citizens. These types of discussions are associated with European identity, but the normalised coefficients suggest that the relationship is comparatively weak; an increase in one standard deviation of discussion about European issues is only associated with an increase in European identity equivalent to 5% of a standard deviation. Of the four cosmopolitan sources considered here, then, discussion about European issues appears to have the weakest relationship with the cosmopolitan dispositions being considered here.
Finally, in addition to these key findings, the models also confirm that youth cosmopolitan dispositions are linked to background variables such as gender and immigrant status as well as to their affective ties, political interest, and political trust (see Tables 2 and 3). By contrast, the data indicate there is little or no independent relationship between our outcomes of interest and the classroom conditions that were included in the model (namely classroom climate and average civic knowledge in the classroom). More notably still, SES of a student is not related to either of the outcomes of interest. This contrasts with the findings of previous studies, which, as noted above, have tended to find that cosmopolitan dispositions are linked to income and occupation.
A review of the preliminary models provided further notable results about the relationship between SES and the outcomes of interest. As part of the preliminary analysis, separate multi-level models were created to examine the effects of the four key predictors independently of one another. This revealed that if contact activities are included in the models, the socio-economic status (NISB) of the student ceases to be significant (compare Models 1 and 3 in Tables A1 and A2 in Appendix 2). By contrast, NISB was statistically significant in the preliminary models that independently examined either learning about international issues at school; exposure to international discussions; or competence in foreign languages, at least when European identity was the outcome of interest. (When student attitudes towards equal opportunities was the outcome of interest, NISB was also not statistically significant when the effects of foreign language competence were modelled independently of the other key predictors examined here; see Model 4 of Table  A2 in Appendix 2). The implications of this are discussed in the next section.

Discussion and conclusions
This analysis sought to test and compare four possible experiences through which young people might develop cosmopolitan dispositions: learning about international issues at school; contact with cultural Others; competence in foreign languages; and exposure to international issues through public and private discussions (i.e. through the media, parents, and peers). Based on this analysis of the ICCS data, it would appear that there is very little support for the fourth hypothesis, namely that exposure to discussions about international issues in the public and private spheres will raise the salience of international issues for individuals and encourage them to develop cosmopolitan dispositions such as supranational identities and support for equal opportunities for citizens of other countries.
The evidence in support of the inter-cultural contact hypothesis is also mixed. While participation in European activities is linked to increased cosmopolitan dispositions, the relationships are weak, particularly when compared to some of the other predictors considered here. These findings are somewhat surprising, given that the positive effects of contact have been amply demonstrated in previous research on tolerance (Pettigrew and Tropp 2006). How, then, can we explain the weak relationships found here? One reason for this disparity may be that the type of contact activities that were considered here include school trips, exchange programmes, holidays, and attending sporting events or cultural festivals (in their local area or in other European countries). These activities are usually short-lived, and as a result, they may not involve the frequent interactions and equal conditions that underpin Allport's (1954) contact theory and that have led to the positive findings that have been reported elsewhere. A second possible reason, however, is that these counter-intuitive findings are not, in fact, surprising. Recent empirical studies in higher education have also found that cross-border mobility programmes (such as Erasmus) have a limited effect on student attitudes towards European identity or EU support (Sigalas 2010), at least among higher educated students (Kuhn 2012). Finally, as Delanty (2012, 336) points out, mobility is not an inherent element of cosmopolitanism: while travel is associated with increased cosmopolitan dispositions, one can still develop these dispositions without having travelled abroad.
Yet even if the effects of inter-cultural contact are relatively weak, these activities may have the potential to facilitate an equalising function that was not anticipated at the outset of this analysis. In short, these results indicate that activities that facilitate contact with cultural Others can obviate the link between SES and cosmopolitan dispositions, as students who have participated in contact activities are more likely to exhibit cosmopolitan dispositions, regardless of their SES. In previous studies, the relationship between SES and cosmopolitanism was partly linked to the fact that contact activities such as foreign travel are most likely to be undertaken by individuals who have the resources do sothat is, individuals with a high SES. However, these studies were largely undertaken among adults, and for a school-age cohort, there are perhaps more policy options to help overcome these obstacles. In particular, additional state resources could be allocated to allow schools to provide inter-cultural contact activities (such as school exchanges or attending European cultural events in the local area) for all students, irrespective of the income or cultural capital of the students' family. This option, while often expensive, could help to ensure that contact activities are not the preserve of the wealthy, and that all young people are provided with opportunities to develop cosmopolitan dispositions.
Of course, this is not the only (or indeed perhaps even the most important) way in which schools can contribute to the development of cosmopolitan dispositions. Schools also provide general education that, as this analysis confirmed once again, is positively associated with the cosmopolitan dispositions that were considered here. What this analysis has also been able to demonstrate, however, is that it is not just general education that is important, but also (and more specifically) students' exposure to education about international issues. As noted above, the analysis presented here shows that learning about international (or at least European) issues at school is positively associated with cosmopolitan dispositions, above and beyond the impact of general education. What is more, the strength of this relationship appears to be relatively strong and indeed in this analysis, stronger than the relationships our outcome indicators had with our measures of general education (namely civic knowledge or expected years of education; see Tables 2 and 3).
In addition, schools also have a potential role to play in supporting the results of the third proposition considered here. These results indicate that students with higher levels of competence in a second European language are more likely to report European identity and support for equal opportunities. The measure used here did not specify whether students had acquired their languages skills in school or elsewhere, but these findings do nonetheless suggest that governments and schools could play a positive and notable role in this process by increasing and improving foreign language teaching.
Channelling these activities through schools is not unproblematic. Schools have historically been bound up with state-and nation-building projects. In recent decades, the overtly nationalistic elements of school curricula have been replaced with a more postnationalist approach that incorporates cosmopolitan principles and values. Despite these reforms, the central logic of national education systems remains the same (to prepare young citizens for citizenship of the nation-state) and these new post-nationalist curricula have often brought with them new forms of exclusion (Author 2014). Citizenshipand state-building projects have long rested on creating boundaries and excluding peoples. Removing these barriers and creating a truly inclusive, tolerance and cosmopolitan view is extremely challenging.
Finally, while this article has been able to data on robust cross-national data, these results must still be treated with some caution. This analysis has only been able to show an association rather than causality, or even the direction of the relationship. As a result, we must continue to pursue empirical tests of cosmopolitan theories, ideally using mixed methods, experiments and programme evaluations of specific interventions at the school level (and not just at the university level). To this end, this article points to some of the experiences that may help young people to develop cosmopolitan dispositions, and highlights some of the key ways that schools may be able to contribute to this process in a positive way.

Disclosure statement
No potential conflict of interest was reported by the author.

Funding
This work was supported by the ESRC (grant ref: ES/K001620/2). Notes 1. Contact theory assumes that these interactions will result in a positive outcome, but there is an opposing view which suggests that increased contact can instead lead to negative outcomes and increased conflict; see Laurence (2011). 2. Switzerland, Liechtenstein, the Netherlands and Belgium-Flanders were excluded from the analysis due to limitations in the ERM dataset of ICCS. For example, insufficient data were collected in the Netherlands to facilitate school-level analyses, while in Switzerland and Lichtenstein, the ERM questionnaire did not include some questions about the EU. 3. Duncan and Raudenbusch (1999, 33) argue that variance of 4% at the school-or country-level can be considered a 'medium' difference, while variance at 8% is 'large' and over 14%, 'very large'. By contrast, if the level of variation is 2% or lower, one can conclude that 'system characteristics were unlikely to be relevant in accounting for the variation in our two outcomes of interest' (Janmaat and Mons 2011, 71). 4. Entering media engagement into the models as a separate variable from parental and peer discussions made no difference to the significance of the relationships, nor to the level of variance explained by the models. Furthermore, frequent use of the Internet to access news was not significant, once other contextual and experiential predictors were taken into account. 5. This variable was operationalised using data drawn from Eurobarometer 71 (2009), and in particular the country-level mean of responses to the question 'Taking everything into account, would you say on balance that [our country] has benefited or not from being a member of the EU?'. Variable name: EUPART (Student participation in activities or groups at the European level) Procedure: IRT WLE scores with mean of 50 and standard deviation of 10 for equally weighted countries. Stem: Have you ever participated in any of the following activities?