State of the art

In recent years, there has been an extensive scholarly debate within the social sciences on whether ethnic diversity reduces trust and solidarity within a welfare state. The empirical findings point in rather different directions and seem inconclusive at present (Putnam, Reference Putnam2000; Alesina and Glaeser, Reference Alesina and Glaeser2004; Torpe and Lolle, Reference Torpe and Lolle2011; Torpe, Reference Torpe2012; Uslaner, Reference Uslaner2012; van der Meer and Tolsma, Reference van der Meer and Tolsma2014).

However, the question of whether certain community values or shared identities impact trust and solidarity has received very limited attention. This is so in spite of the fact that, in particular, nationalist political theorists have argued that diversity is a threat to the shared (national) values underpinning social cohesion and redistributive justice. While this national identity argument has received a great deal of attention in academic debates, the impact of national identity on social cohesion has only been tested in a few studies (Citrin et al., Reference Citrin, Wong and Duff2001; Martinez-Herrera, Reference Martinez-Herrera2004, Reference Martinez-Herrera2010; Shayo, Reference Shayo2009; Theiss‐Morse, Reference Theiss-Morse2009; Johnston et al., Reference Johnston, Banting, Kymlicka and Soroka2010; Wright and Reeskens, Reference Wright and Reeskens2013). Whether a national identity can actually serve the assumed function is therefore very much an open question. Furthermore, as we pointed out in the introduction, several other community values are by various political theorists claimed to bind community members together in ways that are conducive to social cohesion. These other sets of values have more or less been neglected in empirical studies and insofar as they have been studied, they have primarily been included in studies measuring the impact of such values at the level of policy (Crepaz, Reference Crepaz2006; Torpe, Reference Torpe2012) and at the country level (Hjerm and Schnabel, Reference Hjerm and Schnabel2012).

The results from the handful of studies testing the national identity argument are rather inconclusive and contradictory (Miller and Ali, Reference Miller and Ali2014: 2). In one of the few cross-national studies testing this argument, Shayo (Reference Shayo2009) finds not only a negative correlation between national identification (measured as national pride) and support for redistribution at the individual level, but also a strong negative correlation between national identification and levels of actual redistribution on the country level. In a Canadian study in which national identity is measured as ‘closeness’ to the nation, Johnston et al. (Reference Johnston, Banting, Kymlicka and Soroka2010) find partial support for the argument put forward by liberal nationalists, as ‘national identity contributes to a sense of belonging and solidarity that transcends economic interest and cultural differences’, but only when it comes to certain sub-domains of the welfare state (Johnston et al., Reference Johnston, Banting, Kymlicka and Soroka2010: 350). A British study calls the liberal nationalist hypothesis into question as it finds no impact of national identification (measured as ‘support for the political community’) on support for the welfare state (Martinez-Herrera, Reference Martinez-Herrera2004). Furthermore, a couple of US studies find that national identity at the individual level has no significant impact on beliefs about social justice in terms of spending on health, education, and welfare (Citrin et al., Reference Citrin, Wong and Duff2001; Theiss-Morse, Reference Theiss-Morse2009).

Hjerm and Schnabel (Reference Hjerm and Schnabel2012) use xenophobia as a proxy for individual national identity and find that ethnically based national identities have a positive impact on support for the welfare state. And Wright and Reeskens (Reference Wright and Reeskens2013) distinguish between ethnic national identity, cultural national identity, and civic national identity, and find that (only) ethnic national identity is conducive to redistribution and that all three national conceptions induce welfare chauvinism against immigrants.

These studies illustrate some of the difficulties in testing the national identity argument. Thus, the contradictory nature of the results may to some extent reflect different analytical frameworks, statistical approaches, and measures applied. Furthermore, these studies rely on different notions of national identity. This raises the question of exactly what it means to have a shared national identity. As Miller and Ali (Reference Miller and Ali2014) argue, national identity can be measured in several ways, including as national attachment, national pride, and uncritical/critical patriotism, where results may very much depend on how it is measured.

A further complication is that these studies tend not to appropriately distinguish between different community values. This is due to the fact that national attachment, national pride, etc. do not tell us anything about what aspects of the nation it is that individuals feel attached to or proud of. So when, for example, Shayo (Reference Shayo2009) measures national identity in terms of national pride, it is not clear whether his results tell us anything interesting about the national identity argument as it is usually conceived by nationalists. To illustrate this point, suppose that French citizens generally identify with their nation in virtue of the French republican model. In that case, correlations between national identification and trust and solidarity may say more about the impact of republican community values than about the impact of what are strictly nationalist community values. The latter will, after all, emphasize cultural aspects of the nation.

Wright and Reeskens (Reference Wright and Reeskens2013) do go some way toward accommodating different notions of national identity. However, we would argue that their operationalization of ethnic, cultural, and civic national identity is fragile, not least because each national identity relies only on a single measure. For example, cultural national identity is measured as the importance respondents ascribed to ‘speaking the country’s national languages’. While they find this measure to be the closest available to capturing commitments of liberal nationalists (Wright and Reeskens, Reference Wright and Reeskens2013: 1450), it only rather weakly approximates liberal nationalist commitments.

The study of the relation between community values and social cohesion is in many ways still in its infancy and this study differs from those mentioned above in several ways. First, it explicitly differentiates between four different sets of community values, where each of them is based on a broad range of measures in order to strengthen reliability and approximation to the values claimed by political theorists to foster social cohesion. Second, we study not only correlations between specific values and trust and solidarity, but between sharing these values with others and these two aspects of social cohesion. Third, we use a wider range of measures of trust and solidarity than the studies discussed above. And finally, we study the relation between community values and social cohesion in a country where this has so far not been tested, namely Denmark.

Theory

Political theorists arguing for, for example, nationalism, liberalism, or multiculturalism will often have views not just about the justice of social arrangements, but also (empirical) views about the social conditions required for implementing justice. Indeed, they will often hold that the values they propose as the basis for just social arrangements will furthermore, insofar as they are shared among citizens, tend to promote social cohesion and in particular, trust and solidarity, and so make feasible the implementation of social justice. The sharing of values fosters identification with other members of one’s community, which again tends to facilitate positive attitudes toward them such as trust and solidarity (Miller, Reference Miller2004). Of course, the idea that shared values promote social cohesion is not unique to political theorists, as it is widely held in the social sciences (Hooghe, Reference Hooghe2007: 717), but political theorists tend to argue for more specific views about what these values consist of.

Let us briefly say a little about each of the specific sets of community values we examine in this study. Conservative nationalists hold that social cohesion is best promoted by sharing an entire, or at least large part of a, national culture (Burke, Reference Burke1790; Scruton, Reference Scruton1990). The idea is that stable, democratic political institutions require a thick, pre-political community, where community members identify with each other on the basis of a common history, common culture, and (more generally) common way of life, underpinning a sense of a shared fate. In particular, liberal democratic principles are thought too thin and abstract to foster the emotional attachments required for social cohesion.

Liberal nationalists share with conservative nationalists the commitment to a national culture, insisting that, for example, trust ‘is much more likely to exist among people who share a common national identity, speak a common language, and have overlapping values’ (Miller, Reference Miller1998: 48). However, liberal nationalists suggest that national identities are not static but transformed over time and need to be sufficiently open so that, for example, immigrants can realistically access them. Furthermore, they stress the importance of a shared commitment to liberal political structures as a basis for social cohesion (Miller, Reference Miller1995: Chs 4, 5; Kymlicka, Reference Kymlicka2001: 258). This means that there are aspects of national cultures that liberal and conservative nationalists may disagree about; for example, liberal nationalists may be wary of including a particular religion in the national identity (Miller, Reference Miller1995: 92), whereas conservative nationalists may be more inclined to do so (Scruton, Reference Scruton1990: 300, 315).

Liberal citizenship, as we construe this set of community values here, has both liberal and republican elements. The republican element pertains to ‘active citizenship’ and more specifically to democratic participation and being an active member of one’s community. Thus, according to republicans, solidarity is grounded in ‘a reflective acceptance of certain obligations and in practical engagement’ (Honohan, Reference Honohan2010: 94). Along similar lines, Putnam (Reference Putnam2000: 136, 137) suggests that civic engagement, trust, and solidarity may tend to mutually reinforce one another.

The liberal element in liberal citizenship involves a commitment to a set of basic (liberal) principles of justice. Such principles may include freedom of speech, freedom of religion, toleration, and equality of opportunity. For example, Rawls (Reference Rawls1971: 498, Sections 69, 76) argues that a society in which his liberal principles of justice are widely shared and (known to be) implemented in its basic structure will generate its own support, where a sense of justice and trust in one’s fellow citizens will tend to mutually support each other, and where trust in others to comply with the requirements of justice will tend to affirm one’s commitment to this value and the institutions in which it is realized. Furthermore, Uslaner (Reference Uslaner2002: 2, 3) argues that trust and solidarity are part of – and fostered by – a larger package of (roughly) liberal values, including equal standing, equality of opportunity, and opposition to hierarchy, and based on the belief that other people share these values.

Finally, according to multiculturalism, social cohesion is best (or adequately) promoted by sharing a commitment to the recognition of difference. In particular, such recognition may tend to promote the allegiance of minorities who would otherwise feel they receive insufficient opportunities to express their religious and cultural identities (Kymlicka, Reference Kymlicka1995: 184, 185), an allegiance that may again increase majority trust in, and solidarity with, minorities. Furthermore, multicultural commitments may raise majority sensitivity to minority concerns and in this way foster trust and solidarity across groups (Murphy, Reference Murphy2012: 117).

Here, we aim to investigate a particular mechanism through which community values may affect social cohesion, namely that commitments to specific values lead to trust and solidarity at the individual level. The arguments of conservative and liberal nationalists outlined above are clearly of this kind, since they rely on the idea that identifying with others on the basis of a shared nationality generates trust in – and solidarity with – them. It is less clear that, for example, the multicultural argument that recognition promotes the allegiance of minorities is of this kind, and so predicts an association of specific values and trust and solidarity at the individual level. Nevertheless, both the argument that multicultural commitments raises sensitivity to minority concerns and the argument that liberal values and trust and solidarity mutually reinforce each other at the individual level are of the relevant kind.

Note also a further respect in which explanations of the impact of community values on social cohesion may differ. Sharing a particular set of values may impact trust and solidarity simply because individuals who have these values are more likely to exhibit trust and solidarity, or because sharing these values makes an independent contribution to these aspects of social cohesion (Holtug, Reference HoltugForthcoming). In case of the former, we would expect effects to turn up in correlations between individual values and trust and solidarity. In case of the latter, we would expect effects (or at least full effects) to turn up in correlations between individual values in conjunction with the belief that these values are shared by other community members on the one hand, and trust and solidarity on the other. At least a standard explanation of why shared identities matter is that it is easier to extend sympathy and trust toward individuals with whom we identify, where identification is based on beliefs about common attributes (Miller, Reference Miller2014: 13, 14). Indeed, this particular explanation of the significance of shared identities finds support in social identity theory, according to which in-group favoritism and out-group prejudice can result from identifying with the former group (Voci, Reference Voci2006).

Therefore, in the following, we test correlations not only between commitments to the four sets of community values listed above and trust and solidarity, but also between commitments to each of these sets in conjunction with the belief that these values are shared with other community members and trust and solidarity.

Operationalization, data, and methods

The representative sample in our survey includes 1282 respondents above 18 years of age. Data for the survey were collected in October to December 2014 by Statistics Denmark (pilot study in September 2014). The survey sample was drawn from Danish registry data, covering all of the adult population in Denmark. Responses were collected both via internet and phone (log-in and password for the survey sent by letter, phone follow-up if respondents had not replied after 2 weeks). With a response rate of 57.0, the final survey sample is representative of the population in terms of gender, age, income, education, and socio-economic status (labor market position).Footnote ¹

To keep different reasons for identifying with a particular community apart, we have constructed value-indexes for each of the four sets of community values outlined in the last section in order to measure individual commitments to them. The main independent variables in our study are the four indexes expressing community values: conservative nationalism, liberal nationalism, multiculturalism, and liberal citizenship. The value-indexes are based on 25 different variables consisting in value-statements, which are listed in Appendix 1. All variables are Likert scale questions with five categories on which respondents have been asked to indicate their degree of agreement or disagreement (agree strongly, agree somewhat, neither agree nor disagree, disagree somewhat, disagree strongly). The resulting indexes are computed to range from 0 to 100, where a score of 0 means that the respondent disagrees completely with all variables included in the index, whereas a score of 100 means that the respondent completely agrees with all statements.

Briefly, the conservative nationalist index includes values such as believing that everyone in Denmark should follow Danish traditions and that Danish culture and history and Christianity should be fundamental in schools. The liberal nationalist index includes values such as holding that a commitment to Denmark and Danish culture is important for being a part of Danish society and that citizenship requires knowledge of Danish history, culture, and society. The liberal citizenship index indicates an emphasis on the importance of voting, equality of opportunity, freedom of religion and speech, and active participation in society. Finally, the multiculturalism index emphasizes that it is best if a country has a mix of religions and that different religions should be on an equal footing in school curricula and in terms of state support, that Muslim women should be able to wear a headscarf on the job and that mother tongue instruction should be available.

The three indexes of conservative nationalism, liberal nationalism, and multiculturalism are robust indexes with Chronbach’s α of 0.745, 0.626, and 0.740, respectively. This indicates that the variables making up each of these indexes correlate sufficiently with each other. The index of liberal citizenship is somewhat weaker with Chronbach’s α at 0.495. The median of the index of liberal nationalism is 75.0 (relatively high agreement with this index in the population in general), whereas the medians of the three other indexes are 50–57 (not shown). With a std. dev. of 36 (68% of the survey sample score up to 36 points above or below the index mean), the index of liberal citizenship has a relatively high degree of variance, whereas the three other indexes all have std. dev. between 20 and 23 (not shown).

The indexes are somewhat interrelated. Unsurprisingly, conservative nationalism and liberal nationalism are positively correlated, while both are negatively correlated with multiculturalism (not shown here). The index of liberal citizenship, however, is positively correlated with all the other three indexes, but most strongly with multiculturalism (Pearson’s correlation of 0.240). In other words, the index of liberal citizenship to some extent captures respondents with strong value orientations regardless of the content of these values (at least in relation to the three other indexes).

Presumably, the interrelation of the indexes reflects both that a commitment to more than one index is sometimes to be expected for theoretical reasons, but also that people’s commitments often do not track theoretical positions tightly. As regards the former, we would expect conservative nationalists to also endorse the attitudes of the liberal nationalism index and so, for example, hold that Danish culture is important for being a part of Danish society. Likewise, we would expect liberal nationalists to endorse liberal values.

We should emphasize that our selection of variables is based on the theoretically motivated hypotheses we want to test, and so that the empirical interrelations between the indexes does not threaten the theoretical coherence of the distinctions between them. Furthermore, as it turns out, the indexes produce interestingly different results suggesting that important differences between people are in fact captured by these indexes. Nevertheless, the indexes of conservative and liberal nationalism are highly correlated with a Pearson’s correlation of 0.635. Cross-validation with confirmatory factor analysis shows that it is acceptable to distinguish between the two nationalisms looking only at conventional standards for goodness-of-fit measures, but confirms that the two are highly correlated (not shown).Footnote ² This suggests that the scores on these two indexes are quite similar to each other for many respondents, raising the question of whether the two nationalisms are meaningfully different empirically.

The main worry is that many people with high scores on the liberal nationalism index also have high scores of conservative nationalism and so that correlations to trust and solidarity of the former index may be unduly influenced by conservative attitudes. On this basis, we also conduct a separate analysis of liberal nationalists with low scores of conservative nationalism.

A final concern is the issue of multicollinearity when both indexes are included in the same regression model. Multicollinearity means that highly correlated predictor variables (the two indexes) can produce unreliable coefficients and inflate variance (large standard errors). Therefore, all relationships between our four indexes have been checked for variance inflation (not shown here). The variance inflation factor is at no point above 2.0. Multicollinearity should therefore not be an issue.

Besides the four indexes of community values, the belief that one’s values are shared by other community members is also included as an important independent variable. The question put to respondents was this: ‘Generally – to what degree do you believe that your values are shared by other Danes?’. In the survey, the great majority of respondents perceive values to be shared to some degree. In total, 27% believe that values are shared to a ‘high degree’, whereas 62% reply ‘to some degree’. Only 10% reply ‘to a lesser degree’ and 1% say ‘not at all’.

As pointed out above, the dependent variables consist in different measures of trust and solidarity (three measures of each). What the best way is to measure generalized trust has been widely discussed (Delhey et al., Reference Delhey, Newton and Welzel2011, Reference Delhey, Newton and Welzel2014; Torpe and Lolle, Reference Torpe and Lolle2011). In particular, researchers have raised doubts about the classic, dichotomous measure of trust in ‘most people’. Based on these observations, Delhey et al. (Reference Delhey, Newton and Welzel2014) argue that, for example, trust in people you meet for the first time is a better proxy for generalized trust than the classic, dichotomous measure. In this study, we apply three different measures of trust, namely in-group trust (trust in ‘ethnic Danes’), generalized trust (trust in people you meet for the first time), and out-group trust (trust in immigrants). (For the exact wording of our three measures, see online appendix Table 2A.)

We operationalize solidarity as support for redistribution and employ three different measures thereof. The first concerns the extent to which one believes that, on the whole, the state should ensure that people have adequate living conditions. The second concerns the extent to which one believes that the state should increase its efforts for the worse off, even if it means higher taxes. And the third concerns the extent to which one believes that the state is responsible for securing adequate living conditions for immigrants and refugees (see online appendix Table A2). Here, the second measure responds to a concern, expressed by Miller (Reference Miller2006), that a strong sense of national identity is particularly important for sustaining support for welfare state policies targeted at the poor (see also Johnston et al., Reference Johnston, Banting, Kymlicka and Soroka2010: 352).

Linear regression is used to investigate the degree to which we can identify associations between the four indexes of community values on the one hand and trust and solidarity on the other. Standardized regression coefficients (ranging from −1 to 1) will be reported since they are better suited for comparing the correlations of different variables within the same model. These coefficients are standardized according to the different variances for each variable, that is, the regression coefficients should be read as the change, measured in standard deviations, on the dependent variable (trust and solidarity) given a change of 1 std. dev. in the independent variable. Besides the main dependent and independent variables explained above, we control for a range of variables typical to the literature on attitudes, including socio-economic background and ideological orientation (Svallfors, Reference Svallfors2010). These are gender, age, education, and ideological self-placement on a left-right scale. In addition, we include the aforementioned variable on the degree to which the respondent believes that his or her values are shared by other Danes.