Recent research on the consequences of ethnic diversity for social cohesion indicates that the effects of diversity are not necessarily universal. In this article we hypothesise that the rhetoric of political parties conditions whether diversity negatively affects generalised trust. Political campaigns might highlight the salience of cultural diversity issues in their discourse or, moreover, use a divisive rhetoric of ‘nationalistic’ positions. Thus political mobilisation might heighten the perceived conflict between those who are native born and immigrants, especially in diverse societies. In order to test this argument, we investigate the influence of political rhetoric framed on cultural diversity issues, that is, nationalism and multiculturalism – obtained from the Comparative Manifestos Project– on generalised trust in 21 European democracies. We find that the negative impact of ethnic diversity on trust is particularly strong when these issues are mobilised by political parties. It does not, however, matter whether these issues are presented in a positive or negative light.
In recent years, concerns about immigration, increasing numbers of refugees and asylum seekers in Europe, and the rising visibility of ethnic and racial minorities, have triggered an expanding debate about the consequences of diversity for community and social cohesion in Western democracies. Several empirical studies have indeed shown that racial and ethnic diversity and generalised trust do not go well together (Putnam, 2007). In the American context, generalised trust, which has been considered an important measure of social cohesion, is not just lower among minority groups themselves, but is also suppressed among majority populations when they face diverse surroundings (Alesina and La Ferrara, 2002; Soroka et al., 2006). These insights from the American local context have been applied to cross-national scrutiny, where the relationship between ethnic diversity and various indicators of social cohesion, such as trust, welfare state solidarity and civic engagement, have been found to be less stable. Whereas some studies confirm the negative relationship in cross-national settings (Delhey and Newton, 2005; Soroka et al., 2007), others indicate that it is either weak or conditioned by economic, institutional or other variables (Hooghe et al., 2009; Kesler and Bloemraad, 2010; Sturgis et al., 2011).
Thus, before concluding that racial and ethnic diversity affect the social fabric of society, we need to extend current research. We hypothesise that the rhetoric of political parties conditions whether diversity negatively affects social cohesion. Political campaigns might use a divisive rhetoric of ‘nationalistic’ positions with the potential of heightening the perceived conflict between, for example, those who are native born and immigrants. In other words, the way the reputation of immigrants is shaped in political debates has an impact on the diversity–trust nexus as it increases the perceived risks that one takes in dealing with others. If cultural diversity is presented as a problem and the advantages of cultural homogeneity are stressed, high rates of immigration decrease the trust we have in others. More particularly, on the one hand, we look at the effect of the increasing salience of cultural diversity debates and, on the other hand, at the extent to which debates with dominant anti-diversity arguments (compared to pro-diversity arguments) negatively shape the diversity–trust nexus.
The most obvious political actors mobilising cultural diversity issues are populist right-wing parties (Howard, 2009; Messina, 2007; Mudde, 2007, pp. 278–82; Skenderovic, 2007; Skenderovic and D'Amato, 2008). In many West European countries these parties warn of the dangers of cultural diversity and thus create an atmosphere of concern about immigration. Their rhetoric on how diversity threatens the functioning of liberal democracies could lead to declining social cohesion and trust, particularly in a context of existing high levels of immigration and diversity. Geert Wilders from the Freedom Party in the Netherlands or Marine Le Pen from the National Front in France, for example, regularly warn about the dangers that especially Muslim immigration causes for the functioning of liberal nation states. A case in point is the initiative against the construction of minarets in Switzerland which was accepted by the Swiss population in November 2009. The representatives of this campaign argued among others that minarets are a symbol of power and thus threaten the religious peace in Switzerland (Fetzer and Soper, 2012; Neue Zürcher Zeitung, 2007).
Similarly, the current political debates on how to define national identity in France generally relate cultural diversity to the functioning of a liberal democracy and social cohesion. This link is made explicitly by some politicians, for example representatives of the government party Union pour un Mouvement Populaire (UMP), who are in favour of this initiative and emphasise that appeals to nationalism and patriotism help integrate all social classes and thus strengthen national cohesion. They fear that otherwise such questions are left to fringe actors such as the radical right or the intellectuals (Le Monde, 2009). On the other side, many left-wing politicians and intellectuals oppose such nationalistic debates and consider them a danger to the functioning of a liberal democracy (Janoski, 2010). They fear that such a debate stigmatises immigrants and thus opens and strengthens social conflicts (Le Monde, 2009; Libération, 2009).
No matter how we look at it, in many Western democracies the public discourse on issues of cultural diversity has increased over recent decades. Many of these debates take a nationalistic and anti-diversity character and the question is whether they, particularly in diverse settings, affect generalised trust. To answer this question we use data on immigration rates gathered by the United Nations and data on generalised trust, measured in the first wave of the European Social Survey (2002–3). To measure political rhetoric we obtained data from the Comparative Manifestos Project, which collected information on party manifestos published in the context of national election campaigns for the period 1945–2003 for mainly Western countries and Eastern European countries since 1990 (Budge et al., 2001; Klingemann et al., 2006). In the following section, we will briefly review the recent developments in the literature on diversity and social cohesion and explain why we can expect political mobilisation to have an impact on this relation. In the third section, the theoretical arguments on the effect of political mobilisation are presented. The data and methodology are introduced in the fourth section. In the fifth section, the results are presented.
Diversity and Social Cohesion
The current debate about rising cultural diversity in Western democracies and its consequences for social cohesion has been reinvigorated in an article by political scientist Robert Putnam (2007), in which he claims that ethnic and racial diversity dampen various aspects of social capital. Based on the results of the Social Capital Community Benchmark Survey, Putnam argues that at least in the short run ethnic diversity is likely to reduce social solidarity, social capital and particularly various forms of social trust between citizens at the local level, including trust in one's own ethnic or racial group. The causal mechanisms for this effect are still debated, but if one takes the view that trust enables individuals to take risks in dealing with others (Levi, 1998, p. 78), one option is to focus on the importance of previous experiences with or judgements about unknown others in establishing trust (Hardin, 1993). In other words, to assess the odds of betrayal in a society, the reputations of the ‘unknown other’ are taken into account. It can thus be expected that trust decreases in a culturally heterogeneous environment if the discourse around ethnic groups is salient, or if they have been ascribed negative characteristics in the public sphere.
It is no coincidence that these mechanisms have been studied at the local level first. It is there where experiences of diversity between those who are native born and immigrants occur. Thus the effects of cultural diversity have also been studied in the public housing (Fong and Shibuya, 2005; Musterd, 2008) and inter-ethnic contact literature (Schaeffer, 2013). Eric Uslaner (2012, pp. 219–22) asks to what extent public housing policies allow the state to reduce the negative effects of cultural segregation on social cohesion, but comes to the conclusion that most findings in the field are inconclusive so far.
Follow-up studies have been concerned with some extensions of this work. The first extension tests the negative effect hypothesis at the local level in other countries and locations outside the US. Usually these are limited to single-country studies which reveal different results, depending on the context. For example, in Australian neighbourhoods, ethnic fractionalisation is not related to generalised trust, while linguistic fractionalisation shows only a modest negative effect (Leigh, 2006). Another null finding of local diversity on trust was reported in the context of Flanders (Reeskens and Hooghe, 2009). In Britain and the Netherlands, analyses have demonstrated that socio-economic deprivation and not cultural diversity are most harmful for trust (Sturgis et al., 2011; Tolsma et al., 2009). Natalia Letki (2008) found for British neighbourhoods that diversity is not related to social cohesion when the socio-economic status of the neighbourhood is taken into account.
However, other studies contradict these trends. In Canada the percentage of visible minorities in a neighbourhood has a negative effect on generalised trust among the majority and to a lesser extent the minority populations (Soroka et al., 2007; Stolle et al., 2008), indicating that diversity affects various groups of the population differently. Edward Fieldhouse and David Cutts (2010) provide one of the few country comparisons of the effects of localised diversity on neighbourhood norms and participation in the UK and US and also find a negative effect of diversity in both cases.
Another group of studies tests the diversity thesis in the cross-national context and highlights the effects of national forms of ethnic diversity. National levels of diversity might serve as a general proxy for some aspects of the overall climate in which the diversity discourse is placed. Across Europe, rising immigration levels have been met with increased prejudice, outspoken political opposition, racial discrimination and more restrictive views of citizenship requirements (Pettigrew, 1998; Wright, 2011). Marc Hooghe et al. (2009) show that 26 different indicators of diversity are mostly negatively (though not always significantly) related to trust across about twenty European countries. Moreover, in a study across a larger group of developed countries ethnic fractionalisation was related to lower levels of generalised trust (Anderson and Paskeviciute, 2006).
These somewhat mixed and partially negative findings have led scholars to question whether the effects of diversity on social cohesion might be conditional in part. The point is that threats of diverse others might be strengthened, or overcome, under certain conditions. Whereas conflict group theory predicts that ethnic in- and out-groups do stand in constant struggle with each other, the main assumption underlying this article is that there are political circumstances that help to mobilise negative out-group feelings and distrust towards other people, while others might help to moderate them.
One such moderating variable is certainly related to the character of integration policies. The idea is that potentially inclusive and egalitarian integration policies might be more apt in fostering social cohesion than divisive and targeted policies, especially in diverse societies. First evidence shows that this assumption might be right. Christel Kesler and Irene Bloemraad (2010) looked at the moderating effects of economic insecurity and inequality, the presence of multicultural policies and the way grievances are channelled through the political system. They show that in economically more equal societies, in more corporatist societies and in societies with more elaborate multicultural policies, that is, where cultural group rights are more easily recognised, the negative effects of cultural diversity on trust and political engagement are mitigated or even reversed. They conclude that there is no general link between diversity and collective-mindedness. Rather, the direction and strength of the relationship depend on institutional contexts (see also Hooghe et al., 2007).
Besides these institutional aspects we think that potentially short-term political effects also play an important role. Indeed, the role of political mobilisation through political parties has been completely left out of this discussion so far. We believe that the rhetoric of political parties influences how diversity is perceived and whether diversity negatively affects social cohesion. Remember that the risks taken in dealing with others depend on the reputation these others have. It thus can be expected that the reputation of immigrants is at least in part shaped in political debates, which in turn influence the diversity–trust nexus. Political campaigns might heighten the salience of discussion around immigration, and they might even use a divisive rhetoric that is nationalistic and anti-immigration in character. Both processes might potentially sharpen the perceived conflict between those who are native born and immigrants, emphasise threat and thus induce distrust in the overall population, particularly in diverse societies.
Political Mobilisation and Trust
Various studies have shown that people are most often unaware of their cultural surroundings for various reasons. There is first the innumeracy problem and the difficulty for ordinary citizens to know how their cultural surroundings look. Various studies have shown that correlations between perceived and actual cultural diversity in one's neighbourhood and country are rather low (Alba et al., 2005; Chiricos et al., 1997; Citrin and Sides, 2008; Nadeau et al., 1993; Sigelman and Niemi, 2001; Wong, 2007). Moreover, it is not always clear whether natives perceive ‘the other’ as such. James Habyarimana et al. (2010) have observed that members of an ethnic group often do not know that close friends and colleagues are from another ethnic group. More generally, this touches on the question of how to define and imagine a nation (Anderson, 1983). Cultural homogeneity or diversity are not given facts but negotiated and defined in political debates. As other studies have already shown, individual perceptions are heavily influenced by the arguments of political actors or, more generally, public debates (Boomgaarden and Vliegenthart, 2007; Hjerm and Schnabel, 2010; Hopkins, 2010; McLaren, 2010). While some studies do find that ethnic diversity might dampen social cohesion, the intention of this article is to qualify this insight and add that people react to diversity mostly when they receive salient arguments which tell them what the problem is and how to think about it.
While integration policies and other aspects of the welfare state might potentially be successful vehicles for channelling perceptions and images of diverse groups in society (Nannestad, 2008; Rothstein and Stolle, 2008), we would like to add that these views might also be mediated or exacerbated through direct political mobilisation. Every social issue is politicised, problematised and framed in a specific way by a large variety of political actors in the course of public debates. Political parties actively seek to convince ordinary people of the advantages or disadvantages of cultural heterogeneity. By doing so, they create and enforce cultural distinctions and categories, give them a certain meaning and may even provide arguments of how cultural diversity affects other social aspects such as social cohesion.
The role of political parties in politicising immigration issues has heavily increased over the last two decades. Hanspeter Kriesi et al. (2008; 2008,2012) have shown that immigration became the most polarising issue in the electoral arena during the 1990s as well as the most salient one in protest politics, and has remained in this position ever since. Immigration has become the favourite issue among populist right-wing parties (Betz, 2004; Martin, 2000, pp. 256–65; Skenderovic, 2007, pp. 157–60). This stands in contrast to the time up to the mid-1980s, when immigration was one of the least politicised issues on the political agenda of European countries (Hammar, 1985; Layton-Henry, 1992; Saggar, 1992; Schmitter Heisler, 1988) and immigration policy was decided behind closed doors and without public debate (Guiraudon, 2000; Lahav, 2004).
Various interests are involved in immigration politics and opposing positions are defended. It is often argued that a strong mobilisation from the far right leads to more restrictive policies in the arena of immigration (Howard, 2006; 2006,2009; Messina, 2007; Mudde, 2007, pp. 278–82; Skenderovic, 2007; Skenderovic and D'Amato, 2008). On the other hand, there are political actors who are expected to push towards more liberal policies, such as left-wing or post-materialist parties, for example (Janoski, 2010).
The impact of the mobilisation of political parties and especially populist right-wing parties on generalised attitudes towards other people has not been much researched (Givens, 2005; Skenderovic and D'Amato, 2008, pp. 221–4), and the role of political party discourse for generalised trust has so far not been addressed. Some researchers doubt the impact of party mobilisation, as a large part of the population has already rather negative attitudes towards immigrants (Fetzer, 2000) and related distrust. Others emphasise that political parties heavily shape public debates in the course of which they ‘create’ social problems in order to present themselves as those who are able to solve them (Skenderovic, 2007).
So far, it is mainly the impact of the politicisation of immigration issues on attitudes towards immigrants or immigration issues that has been investigated (Boomgaarden and Vliegenthart, 2007; 2009; Hopkins, 2010; 2010,2011; Walgrave and de Swert, 2004). Moreover, these studies merely looked at the salience of immigration issues and not at how they were framed. Thus, their hypotheses concerned the impact of increasing debates about immigration irrespective of whether they are in favour of or against immigration. Some argue that such differentiation is not necessary as most often immigration issues are debated or reported in terms of criminalisation and problematisation (Fitzgerald et al., 2012; Lubbers et al., 1998). More generally, it could be argued that immigration issues are owned by populist right-wing parties. In other words, immigration issues are most often directly related to those political parties that are known for taking restrictive and discriminatory positions (Mudde, 2007, pp. 63–78).
According to Daniel Hopkins (2011), the politicisation of immigration issues and the impact on individual attitudes can work through the following mechanism: salient political rhetoric serves a priming role; among the endless pieces of information to which people are exposed every day, those that are emphasised more intensively by political actors are considered more important, regardless of the actual interpretation of the issue. Salient political rhetoric promotes common knowledge and provides people with a ranking of what is important. This salience argument is an important starting point for our own analyses. Accordingly, in our first hypothesis we test whether issue salience of debates related to cultural diversity strengthens the impact of diversity on attitudes of generalised trust, which are widely recognised as the main indicator of social cohesion and social solidarity (Uslaner, 2002). In short, we hypothesise that political mobilisation leads to interaction effects with ethnic diversity, strengthening the effect of diversity on generalised trust.
In a second step we extend the salience argument as we think that it is useful to differentiate between the salience of issues and the positions political actors take, especially in light of the notion that ideas of multiculturalism and the celebration of diversity, for example, can work as counterweights and are thus not owned by the populist right-wing parties (see Alonso and da Fonseca, 2012). It thus might be that we observe different effects in countries where left-wing parties take positions in favour of multiculturalism compared to countries where the immigration field is primarily occupied by populist right-wing parties mobilising against cultural diversity.
By doing so, we are in a position to juxtapose the effect of increasing salience across any kind of diversity issues with the relative weight of positively and negatively framed diversity issues on trust. Thus, according to our second hypothesis we assume that political discourse that emphasises exclusive and nationalistic claims and the concept of cultural diversity as a threat will also negatively affect the diversity–trust nexus. In other words, we argue that not only the emphasis on cultural diversity issues in party-initiated debates but also the articulation and mobilisation of anti-diversity arguments lead to a strengthening of the potential negative effect of diversity on generalised trust. On the other hand, in a political environment that focuses on the idea of a multicultural society and diversity as a benefit, for example, there is no reason to believe that high levels of diversity lead to less trust.
By looking at the effects of issue salience of political parties’ positions this article improves on earlier work on the relationship between cultural diversity and measures of social cohesion in the national context: the consequences of cultural diversity are seen as embedded in patterns of political mobilisation by political parties. In this sense, we bring political parties back into the discussion about when, why and how diversity might have negative effects on social cohesion.
Data and Methodology
Dependent Variable: Generalised Trust
To assess the impact of political mobilisation on the relation between diversity and generalised trust we will make use of various data sources. First, for the dependent variable we will rely on the indicators for generalised trust as they have been measured in the 2002 wave of the European Social Survey (ESS).1 The ESS is a biennial cross-national survey programme carried out in more than twenty countries and has been praised for its high standards in comparative survey design and fieldwork.2 In this survey programme, generalised trust has been assessed by the classic three Rosenberg (1956) items that measure the extent to which respondents think that people are in general trusting, helpful and fair.3 While there is an ongoing interest in the validity of these survey items (Delhey et al., 2011; Glaeser et al., 2000; Sturgis and Smith, 2010), an analysis of the cross-national equivalence has demonstrated that the three-items scale is nevertheless sufficiently valid for comparative research efforts (Reeskens and Hooghe, 2008; Sønderskov, 2011). As Table A1 in the Appendix demonstrates, there is considerable geographical interdependence in the distribution of generalised trust across Europe (Delhey and Newton, 2005): while the Nordic countries top the list of most trusting societies, Eastern and Southern European countries have rather distrusting populations.
Main Independent Variables: Political Mobilisation and Cultural Diversity
With regard to the main independent variable of interest, that is, political mobilisation, data from the Comparative Manifestos Project (CMP) (Budge et al., 2001; Klingemann et al., 2006) have been used as it is the only data set available for a large range of countries that measures the overall positions of political parties towards issues concerning cultural heterogeneity.4 This data set has already been used by a range of important studies that were interested in political parties’ positions towards cultural diversity issues (Arzheimer, 2009; Arzheimer and Carter, 2006; Alonso and da Fonseca, 2012; Green-Pedersen and Krogstrup, 2008; Hjerm and Schnabel, 2010).5 In each of these studies the manifesto data turned out to be very useful. Moreover, a series of validity tests have shown that the manifesto data lead to similar results as data from expert or population surveys (Helbling and Tresch, 2011; Marks et al., 2007; Netjes and Binnema, 2007; Ray, 2007).6
The CMP data set includes information that has been retrieved from party manifestos and published in the context of national election campaigns. Information on 56 issues has been collected for the period 1945–2003 mainly for Western countries and for Eastern European countries since 1990. In this analysis of the impact of political mobilisation on the diversity–trust nexus, we will collapse the information provided by two issues that are covered in the CMP data set, namely ‘multiculturalism’ and ‘national way of life’. For each of these two categories there is a positive and a negative formulation: ‘multiculturalism: negative’ (MultiNeg) is defined as ‘enforcement or encouragement of cultural integration’ (Volkens, 2002, p. 35). On the other hand, ‘multiculturalism: positive’ (MultiPos) is defined as ‘favorable mentions of cultural diversity, communalism, cultural plurality and pillarization; preservation of autonomy of religious, linguistic heritages within the country including special educational provisions’. ‘National way of life: positive’ (NatPos) is defined as ‘appeals to patriotism and/or nationalism; suspension of some freedoms in order to protect the state against subversion; support for established national ideas’. Finally, ‘National way of life: negative’ (NatNeg) means opposition to patriotism and/or nationalism or to the existing national state. To create an indicator that measures the salience of cultural diversity issues we simply added the four issues:
This indicator ranges from 0 (no mention of such issues) to 100 per cent (no other issues are mentioned in the manifestos). Given that the manifesto approach is based on saliency theory, all data entries are percentages (standardised by the total number of quasi-sentences in a given manifesto). In theory, this variable ranges from 0 to 100 per cent. However, it is almost impossible to find a party manifesto that is dedicated exclusively to one issue or set of issues in particular. Remember that there are 56 categories and thus the percentages of individual categories are relatively low. Sonia Alonso and Sara da Fonseca (2012) have shown for their study that the mean salience score of the 56 categories is 1.62 with a standard deviation of 1.47.
To analyse the overall position of a party or the political climate in a country on the two respective issues, we need to create a position indicator. There are two different ways of doing this. The first one, namely the additive indicator, follows Michael Laver and Ian Budge's (1992) measurement of left–right positions and subtracts the percentage of positive sentences from the percentage of negative ones (or vice versa) (see also Carruba, 2001). This variable theoretically ranges from −100 per cent for a manifesto exclusively dedicated to pro-diversity categories to 100 per cent for a manifesto totally devoted to a negative view of cultural diversity.
The problem with this version is that the indicator does not simply reflect the relative weight of negative issue categories with respect to positive ones but is also influenced by the total content of the manifesto. In other words, two manifestos may share the same number of pro-diversity versus anti-diversity sentences and still obtain different position scores if one of the manifestos is much longer in its total number of sentences than the other. For this reason and following other studies we use the second position indicator, namely the ratio indicator, which divides the net measure (of the additive indicator) by the sum of all negative and positive sentences (Ray, 2007, p. 16; see also Marks et al., 2007, p. 24):7
The ratio measure ranges from a value of 1 for anti-diversity manifestos to a value of −1 for pro-diversity manifestos. To test our hypotheses, for each country in our analysis we created salience and position indicators by averaging the values of national parties in the election campaign that was closest to the year 2002 when the first ESS survey was conducted.8 We were also considering accounting for a longer period of mobilisation. However, mobilisation varies quite a lot from one election to another (Alonso and Da Fonseca, 2012). Moreover, various studies have shown that mobilisation or framing effects are rather short-lived (De Vreese, 2004; Druckman and Nelson, 2003; Tewksbury et al., 2000). For this reason we decided only to account for a rather short period of mobilisation.
Since we are interested in the effect of political mobilisation in diverse societies, it is necessary to add a measure for cultural diversity at the national level to the model. In order to retain as many countries as possible from the ESS, data regarding the share of immigrants for the year 2000 have been obtained from the United Nations Global Migration database (see Table A1).9
To assess the effect of political mobilisation around the issues of nationalism and multiculturalism, it is necessary to control for factors that might also explain differentials in generalised trust. Research thus far has already described that this type of trust is fostered by both individual- (Brehm and Rahn, 1997; Putnam, 2000) and country-level determinants (Bjørnskov, 2007; Delhey and Newton, 2005; Hooghe et al., 2009). At the individual level, an extensive set of covariates is included in the model, namely age, gender, being of foreign origin, years of education, employment status, frequency of informal social meetings with friends, relatives and colleagues and religious attendance. Because we are interested in the effect of national-level political mobilisation on individual-level generalised trust, we also included individual political controls, including political interest, newspaper reading and television watching (see Equation 1). At the country level, the decision was made to control for national wealth, measured by GDP per capita (in US$1,000 PPP) for reference year 2000 (UNDP, 2002), which is the same year as the UN data regarding the share of foreigners (see Equation 2). The low group sample size, that is, a cross-national investigation on around twenty countries, simply does not allow for an extensive set of country-level controls (Kreft, 1996; Meuleman and Billiet, 2009) in addition to the main effects of ethnic-cultural diversity and political mobilisation and the interaction between the two.
The analysis technique used in this article is multi-level multivariate regression (Gelman and Hill, 2006; Hox, 2002). We run random intercept models. Concerning the advantages of the technique, the multi-level regression model first of all allows controlling for the nested data structure of the ESS, namely the fact that respondents are sampled within countries and therefore share a considerable level of variance that might be related to the outcome variable of interest, generalised trust. Second, multi-level models enable us to assess the impact of country-level variables on individual-level outcomes under control of covariates at the lower and higher levels. Since authors have warned about problems of statistical power making use of only around twenty cases (Maas and Hox, 2005), we will be parsimoniously dealing with the country-level controls.
We present our findings in Tables 1 and 2, which display the effects of the salience and position indicators, respectively. Having only 21 cases at hand for the upper level we have to be careful with outliers that might heavily distort our results. Our more detailed analyses have shown that with regard to immigration rates, Switzerland and Luxembourg constitute extreme cases with rates of 21.8 per cent and 32.2 per cent, respectively. In other countries they vary between 2 per cent and 12 per cent (see Table A1). With regard to the mobilisation of cultural diversity issues, Denmark constitutes an extreme case as mobilisation in this period is at least two to three times higher than in other European countries. For these reasons Tables 1 and 2 offer different models excluding these outliers and thereby investigating to what extent they drive or distort our results.
Table 1. Effect of Salience Indicator on Generalised Trust (Standard Errors in Parentheses)
without CH, LU
without CH, DK
Notes: Reference categories are ‘employed’ for employment status and 'married' for marital status. Levels of significance: ***p < 0.001; **p < 0.01; *p < 0.05.
With regard to the individual-level determinants, all our models in both Tables 1 and 2 show that the effects are in line with relations discovered in earlier work (Brehm and Rahn, 1997; Hooghe et al., 2009; Putnam, 2000; 2000,2007). Moreover, it becomes clear that these results are robust as the effects are the same in all models – irrespective of the countries excluded and the indicator in question (salience versus position indicators). For example, on average, the elderly are more trusting than younger respondents (the effects are, however, extremely small). As has been widely discussed, individual socio-economic resources, measured in years of education and employment status, are essential for generalised trust. Finally, the frequency of attending religious activities matters for generalised trust.10 Turning to the trust determinants at the country level, national wealth measured by GDP per capita is relevant: countries with higher levels of GDP also accommodate more trusting people.
Let us now turn to the impact of ethnic-cultural diversity and the mobilisation thereof on generalised trust (shaded areas in Tables 1 and 2). In Table 1 we observe that the negative but not significant parameter of the percentage of immigrants (in societies with low mobilisation) is in line with recent cross-national research outcomes (Hooghe et al., 2009; Tolsma et al., 2009) that have demonstrated that trust is only weakly (if at all) related to ethnic diversity.
However, and largely in line with our argument, the slightly negative effect of cultural diversity on generalised trust becomes stronger when combined with the political mobilisation by political parties (interaction effect). At first glance, mobilisation thus tends to strengthen the impact of diversity on the social fabric of European societies. However, given the small-N nature of the sample an outlier analysis needs to be performed in order to understand the robustness of this result. While in both tables model 1 includes all cases, models 2 and 3 exclude the countries with unusually high immigration rates and model 4 excludes an outlier on the political mobilisation measure (Denmark).
As becomes immediately clear, Switzerland is the only case that somewhat distorts the main result. When Switzerland is excluded the interaction effect is stable and significant in all models; additionally excluding Luxembourg – another country with a high proportion of immigrants on its territory – does not change this result. In a separate model not shown here we excluded only Luxembourg but not Switzerland, and this leads to the same results as in model 1.1. Thus the inclusion of Switzerland dampens the interaction effect, which however remains negative and only about one-third smaller than in models 1.2 to 1.4; the distortion effect of Switzerland is therefore rather small.
Yet it is reasonable to consider Switzerland as an outlier regarding immigration rates. Remember that in Switzerland far more people live with immigrant status than in other European countries (see Table A1). Due to its very restrictive naturalisation regime (Helbling, 2008; Koopmans et al., 2005) the immigration rate in Switzerland is to a large extent composed of people who were born in Switzerland. The immigration rate thus does not reflect cultural diversity to the same extent that it does in other countries where many former immigrants have been naturalised.11
If Denmark is excluded (model 1.4) – a country characterised by massive political mobilisation on cultural issues – the relatively high mobilisation of cultural diversity issues does not change the principal finding. We can thus conclude that the interaction effect identified here is relatively robust with the exception of the Swiss case and that mobilisation in conjunction with high levels of diversity plays an important role. This finding is in line with the results of other studies that have investigated the impact of politicisation of cultural diversity issues on attitudes towards immigrants or immigration issues in individual countries (Boomgaarden and Vliegenthart, 2007; 2009; Hopkins, 2010; 2010,2011; Walgrave and de Swert, 2004).
To understand better the interaction effects in models 1.1 to 1.4 we reproduce them in Figure 1, where the marginal effects of cultural diversity on trust for different levels of mobilisation are displayed. All four graphs show that the more salient the mobilisation of diversity issues becomes, the stronger the negative effect of diversity on trust. Overall, the effects become significant for values above the mean of political mobilisation (indicated by the vertical line), except in the first graph, which includes Switzerland.
While the hypothesis pertaining to the salience of the mobilisation of diversity issues is confirmed, we find no such effects for the position indicator as shown in Table 2. None of the interaction terms is significant, irrespective of the cases we include or exclude in models 2.1 to 2.4. It seems that Marcel Lubbers et al. (1998) were right when they argued that immigration issues are most often debated or reported in terms of criminalisation and problematisation. Or as Jennifer Fitzgerald et al. (2012, p. 481) put it: ‘Politicians in many countries have publicly linked immigration and crime’. This might lead to a situation where these issues are perceived in a negative light irrespective of how the net tone is presented by political parties. And it might also indeed be that immigration issues are most often directly related to political parties that are known for taking restrictive and discriminatory positions (Mudde, 2007, pp. 63–78).
Discussion and Conclusion
The aim of this analysis was to investigate whether the effect of national-level ethnic diversity on generalised trust is moderated by political mobilisation framed on cultural diversity issues. The literature on generalised trust has thus far demonstrated that the national context is important for the generation of trust and cohesion, and consequently it was expected that mobilisation on cultural issues might be a conditional factor here which erodes generalised trust particularly in diverse societies.
We assumed that a higher share of immigrants would have inhibiting effects on trust, mainly in societies with either a salient share of cultural discourse or a discourse that has a particularly negative tone. The first hypothesis was fully confirmed by the analysis. Political mobilisation, which includes the share of discussion of cultural issues, does not bode well for trust in diverse societies. The combination of societies with such a discourse and also a high number of immigrants leads to a backlash in generalised attitudes. Our results seem to underline recent findings in the literature as well. Whereas Kesler and Bloemraad (2010) have already emphasised that the context of multicultural policies has more positive effects on various aspects of social capital particularly in diverse societies, in this article we show that the intensity of the discourse of political parties on cultural diversity issues subtracts from generalised values in diverse societies.
Having shown that mobilisation has an impact on individual attitudes, some might argue that the causal direction goes the other way round. In our case, however, and on the basis of previous research we are quite confident that an inverse causality is hardly possible. It might very well be that political parties mobilise diversity issues in countries where there is a potential for xenophobia. However, we know of no theory that stipulates or any empirical study that has shown that low trust leads to party mobilisation against cultural diversity. Moreover, we should not forget that we do not postulate a direct effect, but an interaction effect of mobilisation with immigration rates on individual attitudes. An inverse causality would therefore mean that individual attitudes have an effect not only on party positions but also on immigration rates. Not only is this hardly conceivable, but if there was indeed such an effect it would lead to lower rather than higher immigration rates.
Finally, it seems so far to be consensus in the literature that mobilisation has an effect on attitudes and not the other way round, even if the empirical proofs are not always very strong. In their study on nationalist sentiments, Mikael Hjerm and Annette Schnabel (2010, p. 530) argue by referring to some classics of the nationalism literature (Breuilly, 1982; Smith, 1998) that nationalism is not always a top-down phenomenon but that ‘elites play a very important role in supporting or constraining the level of publicly pronounced nationalist sentiments’. Especially with regard to party manifestos it has been shown that voters seem to take them into consideration for their voting choices (Hjerm and Schnabel, 2010, p. 530).
In his overview of the impact of populist radical right parties (the main mobilisers of immigration issues) Cas Mudde (2007, p. 291) comes to the conclusion that ‘with a few notable exceptions’ (particularly Schain et al., 2002), ‘studies of populist radical right parties often claim significant impact upon policies (immigration) and society (violence), but provide very little empirical evidence for those claims’. More recent quantitative studies have, however, shown that there is indeed a strong top-down effect from political parties, media salience and policies on individual attitudes (Boomgaarden and Vliegenthart, 2007; Hopkins, 2010; McLaren, 2010). Hajo Boomgaarden and Rens Vliegenthart (2007), as well as Hopkins (2010), show especially with advanced time-series models that salient rhetoric and media coverage of immigration issues have an impact on xenophobic attitudes and vote intentions for anti-immigrant parties. Unfortunately, the manifesto and survey data we use overlap for only one period in time. It is therefore not possible to run time-series models. The evidence put forward by other studies, however, seems to confirm our insights about the importance of top-down effects that result from political mobilisation in diverse societies.
While we observed that political context indeed plays an important role and that the effects vary across countries, one could even go a step further and ask whether mobilisation affects all people equally. It might be argued that some people are more exposed to political mobilisation as they follow the news more closely or are more interested in politics. On the other hand, it might also be, for example, that people with inter-ethnic relationships have been able to form an overarching identity and trust that are potentially immune to the effects of political rhetoric in diverse settings.
Therefore we tested whether the interaction effects between cultural diversity and mobilisation vary depending on political interest, media consumption and number of immigrant friends the respondent has. Overall, our analyses show that people with many immigrant friends and people who are less exposed to political rhetoric are less affected by political mobilisation (results available from the authors). The three-way interactions, however, were not stable and therefore it would be too early to publish them. These first results foreshadow the future direction of these research questions on diversity, political mobilisation and social cohesion in a larger cross-national setting.
Table A1. Univariate Distribution of Variables of Interest across Europe
Generalised trust (0/10 scale)
Percentage of immigrants
Salience indicator (in per cent)
Position indicator (−1/+1 scale)
GDP per capita (US$)
Note: Individual-level data are obtained from the European Social Survey. The GDP is for reference year 2000 and obtained from the Human Development Report. The percentage of immigrants is for reference year 2000 and obtained from the United Nations Global Migration Database.
We only included the first 2002 ESS wave in our study, as this is the only wave that overlaps with the manifesto data (CMP) that have been collected until 2003 (see below).
The participating countries in the 2002 wave of the European Social Survey are Austria (AT), Belgium (BE), Switzerland (CH), Czech Republic (CZ), Germany (DE), Denmark (DK), Spain (ES), Finland (FI), France (FR), United Kingdom (GB), Greece (GR), Hungary (HU), Ireland (IE), Israel (IL), Italy (IT), Luxembourg (LU), Netherlands (NL), Norway (NO), Poland (PL), Portugal (PT), Sweden (SE) and Slovenia (SI).
We used the three following items: ‘Most people can be trusted or you can't be too careful’, ‘Most people try to take advantage of you, or try to be fair’, ‘Most of the time people are careful or are mostly looking out for themselves’.
The unit of analysis is a quasi-sentence that can be defined as an argument, that is, the verbal expression of one political idea or issue in a party manifesto. In its simplest form, a grammatical sentence is the basic unit of meaning. In many cases, however, arguments are combined and related in one sentence. Take the following sentence: ‘Party X is in favour of immigration because its country needs cheap labour forces but is against the construction of mosques as it fears an Islamisation of its country’. Two arguments are made in this grammatical sentence, one in favour of immigration and one against the construction of mosques.
The validity of the CMP data is sometimes criticised. Kriesi et al. (2008, p. 67), for example, have argued that voters do not read party manifestos and can therefore not be influenced by them. Even if this might be true, manifestos nonetheless provide the basis for statements given by politicians during political campaigns (see Robertson, 1976, p. 72). There is no reason to assume that parties change their positions between writing their manifestos and defending their positions in public. Helbling and Tresch (2011) have shown that manifesto data measure very similar dimensions to media data regarding what political actors say in public debates. There is no category for immigration issues in the narrow sense. This points to one of the main weaknesses of the CMP data. As the 56 categories have been fixed for the entire period since 1945, new issue categories are not covered, such as immigration, which has become an important political issue on the agendas of Western states over the last two decades (Kriesi et al., 2008, p. 66). This is not such a big problem for this article as we are more generally interested in the question of cultural heterogeneity which is covered by the two issues ‘multiculturalism’ and ‘national way of life’.
Some might argue that the CMP data measure party rhetoric and thus the articulation of a position rather than mobilisation. We like to point out that the aim of this articulation is to mobilise people in favour of or against cultural diversity. In liberal democracies people are first and foremost mobilised by means of discourses. Moreover, we use data from election periods and thus from a period where parties are particularly concerned to mobilise people.
To illustrate the difference between the two indicators, take the following example of the two manifestos A and B. In both manifestos there are ten sentences in favour of cultural heterogeneity (positive) and five against cultural heterogeneity (negative). Overall, there are 100 sentences in manifesto A and 200 in manifesto B. If you subtract the percentage of ten positive sentences (= 10 per cent) from the five negative sentences (= 5 per cent) in manifesto A you get the value −5. In manifesto B you however get the value −2.5 as the negative and positive sentences make up 2.5 per cent and 5 per cent of all sentences, respectively. If you divide the respective values by the total percentage of positive and negative sentences you get the value −0.33 for both manifestos [A: (5–10)/15, B: (2.5-5)/7.5].
The data have been weighted by the electoral strength of the individual parties.
Ideally, we would like to have the data for 2002. Yet the UN statistics on migration are only available every five years. A test of association between the UN data and the OECD share of foreigners data, which has been used in other analyses (e.g. Hooghe et al., 2009), showed that there is a correlation of .92 between the two indicators, while the correlation between the UN estimation and the share of immigrants reaches .98. This comparison excludes Slovenia, because it does not exist in the OECD statistics. The UN data source reflects the Kesler and Bloemraad (2010) design that analysed the mediating roles of income inequality, corporatism and multicultural policies with cultural diversity on social capital. Given the outlier position of Israel (37 per cent) and the fact that its high share of immigrants is mainly due to religious settlement we excluded this case from our analyses.
More detailed analyses have shown that there is a significant effect for Protestants but not for Catholics. We also included interaction effects and observed that the interaction effect for religious Catholics becomes significant. It thus appears that religiosity as such plays an important role. The inclusion of these additional indicators does not affect our main results.
In Switzerland in 1998 around 60 per cent of all persons with an immigrant status were either born in Switzerland or had stayed in the country for more than twelve years, the minimum requirement to apply for Swiss citizenship (Office fédérale de la statistique, 1999, p. 19).
Marc Helbling is Head of the Emmy-Noether research group ‘Immigration Policies in Comparison’ (IMPIC) at the WZB Berlin Social Science Center and faculty member of the Berlin Graduate School of Social Sciences (BGSS) at Humboldt University. His research fields include immigration and citizenship politics, nationalism and xenophobia/Islamophobia. He has recently edited the volume Islamophobia in the West (Routledge, 2012) and co-authored Political Conflict in Western Europe (Cambridge University Press, 2012). Marc Helbling, WZB Berlin Social Science Center, Reichpietschufer 50, 10785 Berlin, Germany; email: email@example.com
Tim Reeskens is a Research Fellow at the Department of Political Science at the University of Amsterdam, and a Postdoctoral Research Fellow of the Research Foundation – Flanders (FWO) at the Centre for Sociological Research at KU Leuven. His research interests concern the cross-national study of social capital and generalised trust, national identity and support for welfare redistribution. His research has appeared in a number of peer-reviewed journals, including Comparative Political Studies, Journal of European Public Policy and International Journal of Comparative Sociology. Tim Reeskens, University of Amsterdam, OZ Achterburgwal 237, 1012 DL Amsterdam, the Netherlands; email: firstname.lastname@example.org
Dietlind Stolle is Associate Professor in Political Science at McGill University and Director of the Inter-University Centre for the Study of Democratic Citizenship. She conducts research and has published on voluntary associations, trust, institutional foundations of social capital, political mobilisation and new forms of political participation. Her newest book (with Michele Micheletti) is Political Consumerism: Global Responsibility in Action (Cambridge University Press, 2013). Dietlind Stolle, Department of Political Science, McGill University, 855 rue Sherbrooke Ouest, Montréal, Québec H3A 2T7, Canada; email: email@example.com