1 Transnational solidarity in the European sovereign debt crisis. Combined evidence of the European Election Survey and laboratory experiments Theresa Kuhn, University of Amsterdam Hector Solaz, University of Birmingham Erika van Elsas, University of Amsterdam Abstract. The political turf wars surrounding the European sovereign debt crisis have underlined both the high political relevance and the fragile state of transnational solidarity in the European Union. This paper combines evidence from the cross-national European Election survey 2014 and from an original laboratory experiment conducted in the UK and Germany in 2013 to study transnational solidarity among European citizens. More precisely, we analyse the determinants of public support for institutional redistribution and of individuals’ actual willingness to share across borders. Our analyses provide strong support for the hypothesis that transnational solidarity in the EU is structured by cosmopolitan vs nationalist attitudes rather than by pure economic selfinterest or political ideology. Our paper has important implications for the political economy of the sovereign debt crisis and for the research on foreign aid support. Preliminary draft. Please don’t cite nor circulate. Corresponding author: Dr. Theresa Kuhn, Assistant Professor, Department of Political Science, University of Amsterdam, Nieuwe Achtergracht 166, 1018 WV Amsterdam, The Netherlands Theresa.Kuhn@uva.nl www.theresakuhn.eu 2 Introduction The European sovereign debt crisis and the political turf wars it entailed have underlined both the political relevance and the fragile state of transnational solidarity in the European Union. Policy makers are confronted with a dilemma. In view of increased international interdependence, the call for transnational redistribution has become more vocal. However, citizens do not necessarily adapt their allegiances to the transnationalization of their realm (Burgoon 2009). To the contrary, globalization also triggers counter-reactions such as ethnocentrism and parochialism (Margalit 2012; Roudometof 2005). Therefore, this paper seeks to understand transnational solidarity among European citizens. More precisely, it analyses the determinants of public support for transnational redistribution using crossnational data of the European Election Survey (EES) 2014, and it studies people’s actual redistributive behaviour in laboratory experiments conducted in the UK and Germany in 2013. Our results provide strong evidence for the hypothesis that transnational solidarity in the EU is a question of cosmopolitan vs nationalist attitudes rather than of utilitarian evaluations or political ideology. The main finding of the analyses of EES data is that while self-interest, captured by socio-economic status, has some effect on support for transnational redistribution, this relationship seems to be mainly mediated by cosmopolitan values and identity. Moreover, the laboratory experiments show that these values are not mere lip service, but that people subscribing to cosmopolitan values put their money where their mouth is when deciding whether to share money with other Europeans. They don’t discriminate between national and European recipients, whereas people who oppose immigration and European integration give significantly less to European recipients. The contribution of this paper is twofold. First, the project joins the debate (Bechtel, Hainmueller, and Margalit 2014; Beckert et al. 2004) on a crucial, but understudied question, namely to what extent Europeans’ solidarities extend beyond the nation-state by analysing people’s readiness to redistribute at the European level. While existing research has focused on single countries such as Germany (Bechtel, Hainmueller, and Margalit 2014), our study provides empirical evidence of the entire European Union. Second, research on redistributional preferences at the European (Burgoon 2009) and national level (Amat and Wibbels 2009; Fong 2001; Rehm, Hacker, and Schlesinger 2012) has mainly relied on survey data. This approach captures declared preferences, which do not necessarily translate into real behaviour. We solve this problem by combining observational and experimental data, and we show that the factors 3 related to support for institutional redistribution in the EU are also significantly related to people’s actual willingness to share with other Europeans This paper has important implications for current policy debates, as highlighted in the economic crisis and in the controversy around bailouts for member states such as Greece. Moreover, it places itself in the wider scholarly discussion on the tension between community and scale in multi-level governance (Hooghe and Marks 2009). Beyond the European realm, its findings are highly relevant for research on support for redistribution (Alesina and Glaeser 2004; Iversen and Soskice 2001; Kenworthy and McCall 2008) and support for foreign aid (Milner and Tingley 2013; Paxton and Knack 2012; Prather 2014). Finally, it provides empirical support for the expectation of Kriesi and colleagues (2008) that we are witnessing the emergence of a new political divide among an integration-demarcation cleavage that is related to cultural rather than economic attitudes. The paper is organized as follows. First, we summarize the state of the art on redistributional preferences and on transnational solidarity in the EU and then develop the hypothesis guiding our analysis. We then discuss the research design, methods and results of the cross-national survey analysis before turning to the set up and results of the laboratory experiments. Implications and avenues for further research are discussed. Transnational solidarity in the EU We borrow Stjerno’s definition of solidarity as ‘the preparedness to share resources with others by personal contribution to those in struggle or in need and through taxation and redistribution organised by the state’ (Stjerno 2012: 2). This definition underlines that solidarity goes beyond mere empathy and implies the readiness to give up part of one’s own resources for the sake of others. Therefore, two aspects seem to be of prime interest here. First, the question is to what extent people support institutionalized economic transfers from their own member state to other member states in dire economic times. Second, it is relevant to know to what extent Europeans are ready to personally share resources with other Europeans. In this study, we assess both aspects of solidarity. Transnational solidarity implies that people are willing to share their resources with members of other national communities. Consequently, the object under study has two relevant dimensions. The first one relates to people’s readiness to redistribute and their redistributional preferences. The second dimension relates to the territorial scope of solidarity. In the present paper, we analyse the European scope of solidarity. 4 With respect to the first dimension, ample research exists on people’s redistributional preferences (Alesina and Giuliano 2011; Amat and Wibbels 2009; Fong 2001; Rehm 2009; Rehm, Hacker, and Schlesinger 2012; Svallfors 1997). According to Paskov and Dewilde, solidarity can be motivated by calculating considerations: by helping others, people improve their own welfare (Paskov and Dewilde 2012: 417). Similarly, people’s position in society, and their likelihood to mainly benefit from, or contribute to, the welfare state seems to be a strong predictor of redistributional preferences (Iversen and Soskice 2001; Rehm 2009).On the other hand, Paskov and Dewilde speak of “affective solidarity”, which is based on feelings of sympathy and moral duty (Paskov and Dewilde 2012: 417). In fact, a couple of factors, such as religion (Stegmueller et al. 2012), perceptions of deservingness (Van Oorschot 2006) and reciprocity as well as beliefs about the causes of income inequality (Fong 2001) seem to impact people’s feeling of solidarity and support for redistribution. Equally, (racial) group loyalties influence redistributional preferences (Alesina and Glaeser 2004; Luttmer 2001). How these redistributional preferences translate into behaviour is an open question, however. Second, scholars have investigated to what extent solidarity exists beyond the borders of the nation state (Beckert et al. 2004; Ellison 2012; Fenger and Van Paridon 2012). As Beckert and colleagues (2004: 13) note, transnational solidarity is lagging behind and much more difficult to establish than transnational politics and economy. Citizens still see the national community as the predominant reference frame for social inequality (Whelan and Maître 2009). This is partly due to the fact that nation states have long been the main providers of social welfare and therefore could influence people’s understanding of who should be in or out. Moreover, solidarity is more easily achieved in culturally homogenous societies. In the context of large-scale immigration, research on welfare state chauvinism (Crepaz and Damron 2009; Van der Waal, De Koster, and Van Oorschot 2013) has shown that certain members of Western societies hold generally egalitarian values but nonetheless think that welfare state services should not be granted to immigrants. Such an opposition to welfare service provision to immigrants is often rooted in the feeling that migrants don’t deserve it, they have not (yet) contributed enough to the welfare state themselves, and it is hard to predict whether they will actually stay long enough to “pay it back”. In short, many people seem to be “parochially” altruistic (Bernhard, Fischbacher, and Fehr 2006) – they behave altruistically towards members of their own group but less so towards people outside their group. This raises the question of which factors help overcome the national boundaries of solidarity. Put differently, which individual characteristics make people more willing to redistribute transnationally in the European Union? We propose two sets of factors that are deemed to 5 influence people’s willingness to redistribute EU-wide: People’s socio-economic status, their attitudes towards inequality, their political ideologies and their support for European integration. First, support for national redistribution is a question of self-interest (Iversen and Soskice 2001; Rehm 2009). People who expect to profit from redistribution - generally those with lower socioeconomic positions - tend to be more supportive of national redistribution than others. By contrast, people with higher socio-economic status generally have to pay higher taxes and therefore have an incentive to oppose large social expenditures. In the context of our study, this relationship is more complex: Redistribution from one’s own country to another EU member state in economic difficulties results in fewer resources for one’s own welfare state. Moreover, existing research has shown that people with lower socio-economic status tend to be more eurosceptical (Gabel 1998; Hakhverdian et al. 2013; Kuhn 2012) and more critical towards and international cooperation (Hainmueller and Hiscox 2006). Therefore, people with lower socioeconomic position should be less supportive of transnational redistribution in the EU. H1: People with higher socio-economic status are more supportive of transnational redistribution in the EU, and more willing to share with people from other European member states. Next, there are good reasons to expect that people who have a more cosmopolitan mind-set are more supportive of transnational redistribution in the European Union. Rather than representing a conventional political issue, European redistribution is likely to be part and parcel of a greater public divide along an integration-demarcation divide that is structured by cultural rather than purely economic aspects (Kriesi et al. 2008). A considerable share of Europeans has developed a sense of European collective identity (Kuhn 2015; Risse 2010). In contrast to the “exclusive nationalists”, they do not only see themselves as part of their national community, but also identify as European. These people might see other Europeans as equally deserving of redistribution as citizens of their own country. If group identity is high, then collective and individual interest becomes interchangeable. It is to be expected that these people are more willing to redistribute across borders than people who hold a purely national identity. Existing research supports this claim, Kuhn and Stoeckel (2014) find that European identity is strongly related to people’s support for European economic governance in the sovereign debt crisis. In more general terms, a more cosmopolitan identity seems to be associated with preferences for transnational redistribution. In an experiment with nested public good provision games in several countries, Buchan and colleagues (2011) have shown that people with more cosmopolitan identities are more likely to cooperate on a global level and are more willing to contribute to global common goods. Bechtel and colleagues (2014) find in a survey experiment conducted in Germany that support for economic bailouts is structured by cosmopolitan vs nationalist 6 attitudes rather than by political ideology. In an online survey among Latino Americans, Prather (2014) finds that support for foreign trade is question of cosmopolitanism rather than collective interest. Moreover, people who are more supportive of European integration might see the European sovereign debt crisis as a negative side effect of an overall positive project that they endorse. Showing transnational solidarity might be perceived as the bitter pill that needs to be swallowed in order save the European integration process. In contrast, for people who have had reservations against European integration from the outset, the current crisis might confirm their critical position and they might be more reluctant to share resources across borders. We therefore propose the following hypothesis: H2: The more cosmopolitan an individual, the more supportive they are of transnational redistribution in the EU, and the more willing they are to share with people from other European member states. Research design To test our hypotheses empirically, we combine evidence from the EES 2014 and an original laboratory experiment conducted in the UK and Germany in 2013. This research strategy allows us to maximize the generalizability of our results and to assess actual behavior in addition to stated preferences. We start by reporting the cross-national survey analysis before turning to the experimental data. Cross-national survey analysis The EES 2014 was conducted in all 28 EU member states in the month following the European Parliament elections of 22-25 May 2014. It includes a question on financial aid to other EU countries in economic difficulties, as well as other attitudinal items required for testing the hypotheses. Variables The dependent variable, support for transnational redistribution in the EU, is measured by respondents’ agreement with the following statement: “In times of crisis, it is desirable for [our country] to give financial help to another EU Member State facing severe economic and financial difficulties.” A 4-point scale is used to distinguish between strong and moderate (dis)agreement. 7 Respondents were also offered a ‘don’t know’ option, which a total of 1.268 (or 4%) of the respondents opted for. These respondents are left out of the analysis. As indicators of socio-economic positions we include education and social class. Level of education is measured by the age at which a respondent finished full-time education, separated into three categories: 15 or younger, 16-19, and 20+. Respondents who reported that they were still in education were classified on the basis of their age: those of 20 years or older fall into the 20+ category, whereas those younger than 20 are excluded from the analysis (since we cannot determine the age at which they will eventually finish). To measure social class, we use a subjective measure, which asks respondents to locate themselves on an 11-point scale, where 1 corresponds to the lowest and 10 to the highest level in society. Three measures are used to empirically test the second hypothesis. First, attitudes towards immigration are measured by support for restrictive immigration policy. On an 11-point scale, answer categories range from “You are fully in favour of restrictive policy on immigration” to “You are fully opposed to a restrictive policy on immigration”. General EU-support is measured as evaluations of one’s country’s EU membership as good, bad, or neither good nor bad. This is a standard and widely used measure of EU support. European identity is measured by agreement with the statement that “You feel attached to Europe” (from “not at all” to “yes, definitely”, 1-4). Also this item has been widely used to operationalize European identity. It is worth noting that this item refers to Europe rather than the EU, and therefore has a primarily cultural rather than political connotation. Moreover, we include a number of control variables that are informed by existing research on support for redistribution and attitudes towards European integration. Support for national redistribution of wealth is measured on an 11-point scale ranging from fully opposed (0) to fully in favour (10). Support for transnational redistribution might also be structured by political ideology. We use the 11-point scale of left-right self-placement. Given that support for European integration is orthogonal to the left-right dimension in some member states, we also include its square. All models control for gender as women have been shown to be slightly more eurosceptical (Nelson and Guth 2000) and more protectionist (Burgoon and Hiscox 2004). At the country level, we include a control variable for GDP per capita in 2013, and a dummy variable for whether a country is Eurozone member. In additional models that are not shown here, we also included a dummy variable for the new member states that joined in 2004-2013 and for net contributor status. These variables did not change the effect of our independent variables. Due to high collinearity they are not included in the final models. Missing values were treated by list-wise deletion. 8 Method of estimation The ordinal nature of the dependent variable requires ordered logistic regression analysis. The data have a clustered structure with individuals nested in countries. To avoid type-I errors due to this clustering, the regression models are run with robust standard errors (clustered at the country level). To prevent spurious relationships due to composition effects at the country level, we include macro level control variables. Additionally, we check whether these macro controls sufficiently tease out composition effects by running a model including country dummies, which takes out all country level variance. This model leaves the individual level results unaffected (see table 1, model 7). Results Table 1 presents the multivariate models. Model 1 includes the two variables relating to socioeconomic status as well as macro-level control variables. Social class has a significant positive effect on support for transnational redistribution (b=.12). Moreover, people with higher levels of education are significantly more likely to support transnational redistribution. The effect of socioeconomic status remains robust also when including a set of political attitudes in model 2. So far, the analyses provide support for hypothesis 1 that people with higher socio-economic status are more supportive of transnational redistribution. [Table 1 about here] Models 3-5 include the three variables to test the hypothesis that more cosmopolitan individuals are more supportive of transnational redistribution. In model 3 we see that a more restrictive stance on immigration policy has a strong negative effect on support for transnational redistribution, providing strong evidence in favour of our hypothesis (b=-.38). Figure 1 displays this relationship for the four categories of the dependent variable (based on model 3). These graphs give an insight in the size of the effects. Going from least to most in favour of immigration restriction, the likelihood to fully agree with transnational redistribution to other EU member states increases with .14 (for a change of one standard deviation this is .05). A similar increase occurs for the likelihood of tending to agree. As we would expect, the inverse relationship is visible for the lower two categories of the dependent variable: as support for immigration decreases, people become more likely to oppose transnational redistribution. This is in line with the hypothesis that support for transnational redistribution is a question of cosmopolitan vs nationalist attitudes. [Figure 1 about here] 9 The analyses provide further support for hypothesis 2 by showing that attitudes towards European integration are strongly correlated with support for transnational redistribution. The effects of attachment to Europe (b=.53, model 4) and EU membership support (b=.59, model 5) are of a similarly large size, and both variables by themselves explain more variance (with a respective pseudo R2 of .05 and .06) than the set of political issue attitudes in model 2 (pseudo R2 of .03). The effects are robust to the inclusion of other attitudes. Their positive effect remains when including both simultaneously, indicating that they explain different parts of the variance. Figure 2 displays the predicted probabilities of support for transnational redistribution by EU attachment (based on model 4). Going from minimal to maximal EU attachment, the average change in the predicted probability is .20 (and .06 for one standard deviation change in EU attachment). [Figure 2 about here] With respect to the control variables, support for national redistribution of wealth does not increase the likelihood of supporting transnational redistribution. The coefficient is even negative, though insignificant, and remains insignificant throughout different model specifications. Additional analyses show that a significantly negative effect exists only when we do not control for both class and education (not shown here). Either of these socio-structural factors fully accounts for the effect of support for redistribution, indicating that support for national and transnational redistribution are related only spuriously. The effect of left-right placement is significant but very small. Turning to the country-level control variables, we find that in countries with a higher GDP per capita, support for transnational redistribution is higher. A possible interpretation is that citizens of less affluent EU member states feel that their country is less capable of aiding other countries. Eurozone member countries demonstrate clearly lower support for transnational redistribution. This effect appears only when we control for GDP, indicating that comparing two equally affluent EU member states, the one that is Eurozone member is less supportive of transnational redistribution. So far, the analyses provide support for both hypotheses, but which one is more convincing? Model 6 shows that the effect of education and class decreases once accounting for the cosmopolitan variables, while the effect of the latter remains robust. Moreover models 3-5 explain more variance than model 1, as shown by the pseudo-r2. Finally, it is instructive to consider the substantive effect sizes by inspecting the change in the predicted probability of support for transnational redistribution induced by the independent variables in table 2. An increase of 0,5 standard deviations of class increase the likelihood of support for transnational redistribution by 1 per cent, and people with high level of education have a 3 per cent higher 10 likelihood of supporting transnational redistribution. On the other hand, an increase of 0,5 standard deviations of immigration attitudes and of European attachment increase the likelihood by 4 per cent, and a 0.5 standard deviation increase of EU membership support does so by 5 per cent. All in all these findings suggest that cultural values beat self interest in the explanation of support for transnational redistribution. [Table 2 about here] Laboratory experiment So far, our analyses provide strong evidence in line with the hypothesis that support for transnational redistribution in the EU is mainly linked to people’s attitudes about immigration and European integration rather than motivated by self-interest or political left-right ideology. However, critics may argue that our dependent and independent variables simply measure the same latent trait, and that support for transnational redistribution might be mere lip service that does not translate into actual behavior. In other words, are these cosmopolitan, pro-European people simply responding in a socially desirable way or are they really willing to give up a piece of their cake? To answer this question empirically, we conducted laboratory experiments in four locations in Germany and the United Kingdom. Experiments are highly beneficial to examine the present question because they allow analysing actual rather than reported behaviour or opinions on redistribution, which might be heavily influenced by social desirability. Not only for this reason, they have become increasingly popular in political science research (Druckman et al. 2006). Due to the nature of the question, the laboratory experiments had to involve citizens of different countries and take place in different EU member states. Participants were linked to each other across locations. Only by doing so, we could analyse people’s redistributive behaviour across countries without deceiving experimental subjects. Fieldwork took place in four locations in April and May 2013: Oxford (n=63), Edinburgh (n=43), Berlin (n=68) and Munich (n=43). We opted for Germany and the United Kingdom because they differ with respect to public opinion towards European integration. We chose the respective cities because they represent the national core (Berlin, Oxford) and regions with strong subnational identities (Edinburgh for Scotland and Munich for Bavaria). The computer assisted experiments were conducted using the software ztree (Fischbacher 2007) and took place in experimental laboratories at academic institutions. The experiments took place in experimental economics laboratories at academic institutions. Experimental participants were recruited by the laboratories among university students of all 11 disciplines. Only German and UK citizens were allowed to participate in the German and UK locations, respectively. They received a show-up fee of 5€1 and could keep the pay-offs they had earned in the games. On average, participants earned 20€ in total. Due to the fact that participants were matched across locations and their earnings were hence dependent on the behaviour of people participating elsewhere, payoffs could only be determined and paid out after three weeks. Participants were informed about the payment procedures ahead of the experiments. The experiments are based on standard experimental procedures in experimental economics, but enriched with a multi-level design that reflects the multi-level politics in the European Union. Experiments consisted of four decision games, two of which form the empirical basis for the analysis here: Dictator Games as well as Public Goods Provision Games. They were followed by a 20minute questionnaire that tapped respondents’ socio-economic background, collective identities, political attitudes, and international experiences. Each experiment consisted of a number of decisions. In each decision, participants received an initial endowment, which would be paid out in cash at the end of the experiment, and had to choose whether to keep it or to allocate it to another anonymous and randomly chosen participant. Their payoffs depended on their own decisions and on other participants’ decisions2. Subjects were not informed about the decisions taken by their peers, nor did they know who they were matched with. It is important to reiterate that we did not deceive subjects at any point in the experiment. Throughout the experiments, two decision parameters were varied. First, participants received different information on where the recipient was from: Either from the same town, the same country or from another EU member state3. Second, the scenario was varied: In the Dictator Games, subjects had to decide whether to keep their endowment or to allocate part of it to another anonymous and randomly chosen participant who had not received any endowment. This game thus induced inequality between the donor and the recipient. In the Public Goods Provision Games, subjects had to choose whether to allocate their endowment to a common account that was shared with three other anonymous and randomly chosen participants or to keep it for themselves. If they kept it, they received the full amount at the end of the game in cash; if they allocated it, the total amount in the common account was doubled and equally distributed among the four participants and paid out in cash. This game highlighted mutual trust and cooperation. Variables 1 6 GBP in the United Kingdom. The exact wording of the instructions given to participants can be found in the appendix. 3 Note that no information about the exact member state was given. 2 12 The unit of analysis refers to decisions. Each participant took part in two games with three decisions each. Decisions referred to contributing the local, national, and European level, respectively, and the order in which they are presented to subjects was randomly assigned. We analyse each game separately and present pooled analyses of decisions nested in participants. We use total contributions per decision as a dependent variable, and the different information on whether the contributions go to a local, national, or European participant as independent variables. We interact these dummy variables with our independent variables. With respect to the independent and control variables relating to the respondents, we mirror the EES as closely as possible. Orientations towards European integration are captured by EU membership support (same wording as in EES) and by identification as European. Respondents were asked, “Do you see yourself as [country national] only/ [country national] and European/ European and [country national] / European only?”. Attitudes towards immigration are measured by agreement with the statement “Right now [country] is taking too many immigrants” on an 11-point scale. Socio-economic background usually refers to people’s class, education, and income. As the sample is drawn among university students, participants’ educational background cannot be varied. While we asked participants to indicate the monthly amount of money at their disposal, these data do not seem to be reliable. We therefore exclusively refer to participants’ self-reported class, ranging from working class to upper class. The following item refers to inequality aversion. “Please indicate to what degree you personally agree with the following statements: Right now, differences in incomes are too large in [country]”. Answer categories range from absolutely agree (1) to absolutely disagree (11). People’s political ideology is measured using the same 11-point left-right scale spectrum as in the EES. Additionally, the models control for age and gender. Table 2 shows the descriptive statistics of these variables. Results Table 3 shows the direct effect of a European cue versus a national recipient cue on contributions in the dictator game and in the public goods provision game. We find very little difference between contributions to national and European recipients. Thus, on average people don’t make a difference between giving to someone form their own country or from another European member state. However, this does not mean that the origin of the recipient does not play a role at all. In fact, some people might actually give more if they know that the money is going to someone in another member state, for example, because they might feel pro-European or they might think that people in their own country might need it less. 13 [Table 3 about here] We therefore introduce interaction terms in order to test our hypotheses on socio-economic status and cosmopolitanism. In each model in table 4, the dummy variable “European recipient” is interacted with the independent variable of interest. We can see that only the three variables referring to EU membership support, European identification and immigration support have a significant interaction effect with European recipient cue, whereas class membership and general inequality aversion don’t play a role. In model 2, when accounting for immigration attitudes, the European recipient cue has a strong and highly significant negative effect. People who most strongly oppose immigration give about 64 tokens less to European recipients than to recipients from their own country. This difference decreases as immigration support increases. [Table 4 about here] This relationship is further clarified in figure 3: People who oppose immigration to their country give significantly less to a European recipient than ta national recipient, while respondents in favour of immigration don’t seem to make a difference in their contributions. The relationship is less clear for the public goods provision game. We find a similar pattern when referring to European identification (model 4): In the dictator game, there is strong and significant negative direct effect of European origin cue on contribution, but it is almost entirely moderated by European identification. People with exclusive national identity give significantly less to recipients from another EU country, while participants who also identify as European don’t discriminate between recipients. The very few people who primarily or exclusively identify as European seem to give even more to European recipients, but the confidence intervals fan out, which is probably due to the low numbers of observations. While we find a similar relationship for the public goods provision game, this effect is not statistically different from 0. Finally, model 5 includes an interaction effect with EU membership support, People who think that EU membership is a bad thing contribute almost 100 tokens less in the dictator game, whereas participants who support EU membership don’t make a difference between EU and national recipients. This pattern also holds with respect to the public goods provision game. [Figure 3 about here] [Figure 4 about here] All in all, these results support the hypothesis that transnational solidarity in the EU is mainly a question of collective identity and attitudes towards internationalization. They also strongly suggest that reported European identification is more than lip service but translates into real 14 behaviour and renders people more willing to decrease their material welfare for the sake of other Europeans. Robustness checks We estimated a number of additional models so as to ascertain whether our findings are robust across model specification. Due to space limitations, the results are not reported here, but we summarize the main findings. First, we also included interactions of the origin cues with political ideology and inequality aversion. This did not yield any significant effects. Next, we used two alternative dependent variables: The first one, “total EU redistribution”, referred to the absolute amount of tokens transferred to participants in another EU member state in each game. To discount the possibility that people are contributing more at the EU level than others simply because they are generally more generous, the second operationalization, “net EU redistribution” measured people’s contribution at the European level minus their contribution at the national level. The results confirmed the findings reported in this paper. Moreover, we estimated additional models that included dummy variables for each experimental location (Edinburgh, Munich, Oxford, while Berlin served as the reference category) so as to capture potential contextual effects. While participants in Oxford and Munich contributed significantly less in some decisions, this effect was not robust and did not substantively change the individual effects. Finally, as each subject had to take three decisions per game (contribution to local, national or European subject, random order), they might understand the rationale of the experiment after the second decision. Hence, critics might argue that while decisions were incentivized, subjects might still have acted in a socially desirable manner and might have based their second and third decisions on their earlier contribution. To eliminate the possibility that our findings are driven by such an effect, we moved from a within-subject design to a purely between-subject design by analysing the first decision only. The results confirmed the findings reported here. Discussion and outlook The aim of this paper was to investigate who are the Europeans that display transnational solidarity in the European sovereign debt crisis. In line with Kriesi et al. (2008) and confirming existing single-country studies, we expected transnational solidarity to be structured by a cultural cosmopolitanism vs nationalism fault line rather than by political ideology or self-interest. To this aim, we assessed public support for transnational redistribution using the EES 2014 and we analyzed people’s readiness to redistribute to citizens from other EU member states in laboratory experiments in the United Kingdom and in Germany. Experimental participants were given an 15 endowment and were confronted with a series of decisions in which they could either keep their endowment or give (part of) it to participants from other EU member state. Both the cross-national survey analysis in the EU-28 and the laboratory experiment strongly suggest that people’s orientations towards European integration and immigration are the most powerful predictors, while self-interest played a less important role. EU membership support is significantly associated with higher levels of redistribution towards other European participants under all conditions analyzed, while the positive and significant effect of European identification was slightly less robust. Equally, people holding a more restrictive stance on immigration policy are less supportive of transnational redistribution, and they are also less generous towards other Europeans in incentivized laboratory experiments. In contrast, respondents’ political ideology and their attitudes towards inequality were significantly related to the dependent variables in very few models. This suggests that transnational redistribution in the European Union is not a “conventional” political issue that can be easily integrated in the domestic political spectrum. It also suggests that transnational solidarity in the EU is motivated by affective solidarity rather than by calculating considerations. There are some limitations to this study. The analyses documented in this paper refer to correlation rather than causation. It is impossible to assess causality in cross-sectional surveys such as the EES, and while the great strength of experimental research lies in random assignment to treatments, the hypotheses here related to subjects’ characteristics rather than to having been in a certain treatment group. This limits our ability to make causal claims, as we cannot say, for example that membership support causes redistribution. On the other hand, the fact that findings are very similar across data and methods employed increases our confidence in their validity. Moreover, it suggests that social scientists may be more confident in the external validity of laboratory experiments: while based on a small sample of university students, the relationships found in the lab are confirmed in a cross-national, representative survey in the EU-28. 19 Tables and figures Table 1: Ordered logit models (with clustered SE's) explaining support for financial help to other EU countries Socio-‐structural factors Age Male Class (subjective, 1-‐10) Low educated (ref: middle) High educated (ref: middle) Attitudes Support redistribution (z) Left-‐right (z) Left-‐right squared (z) Restrict immigration (z) Attachment to EU (z) EU membership support (z) Country level GDP per capita (2013) Eurozone member (0/1) 1 2 3 4 5 6 0.00 (.00)** 0.00 (.00)** 0.00 (.00)*** 0.00 (.00) 0.00 (.00) 0.00 (.00) 0.14 (.03)*** 0.14 (.03)*** 0.16 (.03)*** 0.13 (.04)*** 0.12 (.03)*** 0.14 (.04)*** 0.12 (.02)*** 0.12 (.02)*** 0.11 (.02)*** 0.08 (.02)*** 0.08 (.02)*** 0.07 (.02)*** -‐0.26 (.06)*** -‐0.26 (.06)*** -‐0.23 (.05)*** -‐0.18 (.06)*** -‐0.16 (.06)** -‐0.11 (.06)* 0.43 (.07)*** 0.42 (.07)*** 0.38 (.06)*** 0.33 (.07)*** 0.31 (.06)*** 0.24 (.06)*** -‐0.03 (.03) 0.01 (.03) -‐0.11 (.05)** -‐0.11 (.03)*** 0.01 (.02) -‐0.00 (.02) -‐0.38 (.05)*** -‐0.30 (.04)*** 0.53 (.05)*** 0.33 (.04)*** 0.59 (.04)*** 0.44 (.03)*** 0.01 (.00)*** 0.01 (.00)*** 0.01 (.00)*** 0.01 (.00)*** 0.01 (.00)*** 0.01 (.00)*** -‐0.36 (.17)** -‐0.37 (.17)** -‐0.32 (.15)** -‐0.31 (.17)* -‐0.45 (.16)*** -‐0.38 (.15)*** Constant cut1 0.03 (.30) 0.06 (.28) 0.10 (.29) -‐0.39 (.34) -‐0.39 (.31) -‐0.46 (.31) Constant cut2 1.37 (.31)*** 1.40 (.28)*** 1.47 (.30)*** 1.01 (.34)*** 1.05 (.31)*** 1.03 (.31)*** Constant cut3 3.60 (.33)*** 3.64 (.30)*** 3.76 (.32)*** 3.34 (.36)*** 3.40 (.33)*** 3.46 (.33)*** Pseudo R2 0,03 0,03 0,04 0,05 0,06 0,08 Observations 20,633 20,633 20,633 20,633 20,633 20,633 Source: EES 2014. Robust standard errors in parentheses. *** p<0.001, ** p<0.01, * p<0.05. Note: Model 7* includes country dummies 7* 0.00 (.00) 0.11 (.04)*** 0.05 (.01)*** -‐0.14 (.05)*** 0.23 (.05)*** 0.00 (.03) -‐0.10 (.03)*** 0.01 (.02) -‐0.30 (.04)*** 0.34 (.04)*** 0.44 (.03)*** -‐1.34 (.15)*** 0.18 (.14) 2.67 (.14)*** 0,10 20,633 20 Table 2 Substantive effect sizes: effect of change in X on change in predicted outcome Socio-‐structural factors Age Male Class (subjective, 1-‐10) Low educated (ref: middle) High educated (ref: middle) Attitudes Support redistribution (z) Left-‐right (z) Left-‐right squared (z) Support immigration (z) Attachment to EU (z) EU membership support (z) Change in X Model 1 Model 6 (+/-‐ .5 SD) (0 to 1) (+/-‐ .5 SD) (0 to 1) (0 to 1) 0,01 0,02 0,03 0,03 0,05 0,01 0,02 0,01 0,01 0,03 (+/-‐ .5 SD) (+/-‐ .5 SD) (+/-‐ .5 SD) (+/-‐ .5 SD) (+/-‐ .5 SD) (+/-‐ .5 SD) 0,05 0,03 0.00 0,01 0.00 0,04 0,04 0,05 Eurozone (0 to 1) 0,05 GDP (+/-‐ .5 SD) 0,03 Source: EES 2014. Note: Average predicted changes (across 4 categories of DV) obtained through stata's prchange command, based on models 1 and 6 in table 1. Table 3: Aggregate analysis of contributions towards local, national, and European recipients in lab experiments Local recipient European recipient Constant N (decisions) Dictator game 16.908 (1.92)* -‐4.742 (0.54) 265.940 (17.13)**** 651 Public goods provision game 3.940 (2.47)** -‐0.392 (0.25) 49.452 (19.01)**** 651 Source: own laboratory experiment. Panel data analysis. Individual fixed effects. Coefficients refer to contributions in each decision. Standard errors in parentheses. 2-‐tailed test,* p<0.1; ** p<0.05; *** p<0.01; **** p<0.001 21 Table 4: Treatment effects on contributions, interacted with participant characteristics Social class Participant characteristics Age Gender Class DG Support for immigration European identity PGG DG 2.68 (3.44) -‐0.35 (0.56) 3.17 (3.46) -‐0.20 (0.55) 14.02 (33.06) -‐6.94 (5.35) 15.52 (33.04) -‐37.53 (19.04)* -‐1.86 (3.18) -‐38.75 (18.40)* PGG DG EU membership support PGG DG PGG 0.88 (3.56) -‐0.68 (0.56) 2.70 (3.98) -‐0.01 (0.65) -‐6.48 (5.29) 25.33 (35.19) -‐1.72 (5.57) 23.42 (33.18) -‐5.82 (5.37) -‐2.81 (2.95) -‐46.12 (19.21)* -‐3.42 (3.04) -‐37.67 (18.52)* -‐2.11 (3.00) Ideology 3.16 (10.25) -‐0.25 (1.66) 8.76 (11.23) 1.49 (1.80) 7.01 (10.87) 0.15 (1.72) 6.60 (10.38) 0.47 (1.68) Inequality aversion -‐1.70 (6.60) 2.93 (1.07)** -‐2.14 (6.61) 2.79 (1.06)** -‐2.96 (7.07) 3.02 (1.12)** -‐3.63 (6.63) 2.63 (1.07)* Support for immigration European identity (0-‐3) EU membership support Recipient in game Local 21.73 (28.39) European -‐27.96 (28.39) Interactions with recipient 4.14 (6.27) 2.38 (1.03)* 35.67 (28.06) 9.86 (4.57)* 15.22 (27.88) 7.26 (4.66) -‐0.89 (6.20) 10.75 (18.82) 9.65 (4.16)* -‐1.97 (14.01) 4.81 (3.04) -‐57.71 (22.87)* 3.06 (5.10) 5.73 (6.20) -‐64.38 (18.82)*** -‐3.90 (4.16) -‐31.94 (14.01)* -‐1.94 (3.04) -‐96.25 (22.87)*** -‐11.22 (5.10)* Local*Class -‐1.90 (9.15) 1.71 (2.00) EU*Class 6.36 (9.15) -‐2.02 (2.00) Local*Inequality aversion EU*Inequality aversion Local* Immi. support 0.85 (2.70) -‐0.87 (0.60) EU* Immi. support 8.77 (2.70)** 0.58 (0.60) Local*European identity 20.80 (12.75) -‐0.94 (2.76) EU*European identity 25.81 (12.75)* 2.45 (2.76) Local*EU support 44.63 (13.04)*** 0.71 (2.91) EU*EU support Constant 302.42 (125.41)* N 552 52.87 (13.04)*** 6.736 (2.907)* 44.48 (20.44)* 246.9 (137.8) 22.00 (22.13) 326.0 (131.3)* 43.73 (20.81)* 274.1 (141.4) 23.93 (23.00) 552 552 552 519 519 543 543 Source: own laboratory experiment. Panel data analysis. Individual fixed effects. Coefficients refer to contributions in each decision. Standard errors in parentheses. 2-‐tailed test,* p<0.1; ** p<0.05; *** p<0.01; **** p<0.001 22 Figure 1: Predicted probability of support for transnational redistribution (1-‐4) by support for immigration Source: EES. Note: Average predicted change (from minimum to maximum) over the four categories of dependent variable is .14. Figure 2: Predicted probability of support for transnational redistribution (1-‐4) by EU attachment Source: EES. Note: Average predicted change (from minimum to maximum) across the four categories of dependent variable is .20 23 Figure 3: Effect of European recipient as immigration support increases in Dictator game (left) and Public Good Game (right) Figure 4: Effect of European recipient as European identity strengthens in Dictator game (left) and Public Good Game (right) 24 Appendix Appendix table 1: Descriptive statistics of experimental data Variable Age Gender Class Ideology EU membership support European identification Ntl. inequality is too high n 201 203 191 197 197 188 199 Mean 23.76 .43 2.96 4.34 2.64 1.88 7.93 Std. Dev. 4.62 .50 .93 1.76 .63 .67 2.96 Min 19 0 1 1 1 1 1 Max 52 1 5 10 3 4 11 25 Appendix 2 Wording of experimental instructions General instructions: Welcome to this experiment. This experiment is about how people make decisions. If you pay close attention to the instructions then you could make a significant amount of money. Feel free to ask the monitor questions as they arise. From now until the end of the session, unauthorized communication of any nature with other participants is prohibited. Please note that the consumption of food and beverages is not allowed during the experiment. This experiment consists of 4 modules and one questionnaire at the end. Instructions will be handed out at the beginning of each module. We ask that you plan on staying until the end of the session, which will last about 90 minutes. In this experiment you are going to be asked to take decisions that affect you and other people. Some will be in this city, but they may not be in this room now; some will be from other cities in the United Kingdom, and some will be from other member states of the European Union. At this point, some people may have already participated in this experiment, and other groups are participating in the same experiment these weeks. Your choices, and the choices by others, will be matched with the help of some colleagues at another university once the research is finished. You will be paid £6 in cash as a show up fee at the end of this session, and in three weeks time, at the end of this research, you will receive an email asking you to come to be paid in cash for the decisions that you and the people you have been matched with made. The same instructions are being given to other people in other countries. Everyone will get the same materials that you get, and is hearing the same thing you are, but in their own language. All of the decisions are similar, so please pay attention to these instructions. At the outset of each decision you will be given tokens. It will be important to keep in mind that 1000 tokens are worth £3.25 to you. We have taken care that the tokens are worth the same value in terms of what could be purchased with them in each participating country. Again, keep in mind that you are being matched with other people (some of whom are from around here, and others from other places in the UK or other European member states). Your decisions, and the decisions of the other participants, will affect how much you make. When the research is finished, our server will match the information about others’ choices in order to calculate each participant’s payments. This may take a few weeks, and that’s why you will receive your payoffs in three weeks time, but you will receive a £6 cash show up fee before you leave today. 26 Dictator game a) In this module you are going to make three independent decisions. b) Half of the participants will receive an Endowment of 1000 tokens (group A), and the other half will not (group B). c) Each participant who receives an Endowment (group A) will be randomly paired with another participant who has not (group B). You will not know the other person's identity, nor will they know yours. Nor will these identities be revealed after the session is completed. d) However, before the endowments are distributed and the pairing takes place, you may allocate the endowment between yourself and the other person as you wish if you were to receive this Endowment. e) Profits in this module will be calculated in the following way: i) Group A: Profits = Endowment – Amount Sent ii) Group B: Profits = Amount Received Decision 1 Remember that you don’t know yet whether you are in Group A or in Group B. How many of the 1000 tokens do you send to the other participant knowing that he/she is participating in another location in the UK? How many of the 1000 tokens do you keep for yourself (remember that the sum of both amounts have to be equal to 1000 tokens)? Decision 2 Remember that you don’t know yet whether you are in Group A or in Group B. How many of the 1000 tokens do you send to the other participant knowing that he/she might not be in this room, but is participating in this local area? How many of the 1000 tokens do you keep for yourself (remember that the sum of both amounts have to be equal to 1000 tokens)? Decision 3 Remember that you don’t know yet whether you are in Group A or in Group B. How many of the 1000 tokens do you send to the other participant knowing that he/she is participating in another member state in the European Union? How many of the 1000 tokens do you keep for yourself (remember that the sum of both amounts have to be equal to 1000 tokens)? [NB: all participants were asked to make all three decisions; decisions were in a random order] 27 Public goods provision game In this module you are going to make three independent decisions. You will be given 100 tokens in each one. Your task is to decide how to allocate your tokens between two different accounts. You can put your tokens into your “Personal” account or into, what we call now, the “Common” account. The number of tokens you put into any account is entirely up to you. Whatever you put into the “Personal” account is yours and will not be shared with anyone else. In other words, every token you put into this account is independent of the other participant’s decisions. Any token that you, and three other people you will be grouped with, put into the “Common” account will be doubled by me. That amount will be equally distributed among you and the three other participants. Before you make your decisions, I want to make sure that you understand how the amount you obtain is determined. Therefore I am going to give you three examples. Please pay close attention. Once I am finished with the examples we will test whether you have understood the instructions asking you a few simple questions. Your performance in the test will not affect your payoffs. For example #1, suppose that you put 100 tokens in your “Personal” account and the other three people put a total of 120 tokens in the “Common” account. In that case, the 120 tokens in the “Common” account will be doubled (to 240) and shared equally among you and the other three people (60 each). You would receive 100 from your “Personal” account that you kept, and 60 from your share of the “Common” account. You would end up with 160 tokens. You Personal Account 100 Common Account 0 Total 100 Others 120 Total 120 Increased Your share 100 100 240 60 160 To take another simple example (#2) suppose you put 80 of your tokens in the “Common” account and no one else put any tokens in the “Common” account. What would you receive? You would receive 20 tokens from your “Personal” account. Given that there were 80 tokens in the “Common” account that amount would be doubled to 160 and you would get an equal share, which is 40 tokens. The other people in your group also get 40 tokens. You would end up with 60 tokens. You Personal Account 20 Common Account 80 Total 100 Others 0 Total 80 Increased Your share 20 20 160 40 60 28 Finally, let me give one more example (#3). Suppose you put all 100 of your tokens in the “Common” account. Suppose that the other 3 people did the same thing. That means a total of 400 tokens in the “Common” account. How much would you receive? You Personal Account 0 Common Account 100 Total 100 Others 300 Total 400 Increased Your share 0 0 800 200 200 Now please answer the test [participants were asked to pass a computerized test that checked their comprehension of the module before making their decisions] Decision 1 As I mentioned you are going to take a decision that affects you and three other people who participate in your common account. These three other participants are randomly determined. All three participants live in this local community. This is why we call the common account now the local account. How many of your tokens do you give to the local account (Remember that the remaining amount goes into your private account)? Decision 2 As I mentioned you are going to take a decision that affects you and three other people who participate in your common account. These three other participants are randomly determined. All three participants live in the United Kingdom. This is why we call the common account now the UK account. How many of your tokens do you give to the UK account (Remember that the remaining amount goes into your private account)? Decision 3 As I mentioned you are going to take a decision that affects you and three other people who participate in your common account. These three other participants are randomly determined. All three participants live in the European Union. This is why we call the common account now the EU account. How many of your tokens do you give to the EU account (Remember that the remaining amount goes into your private account)?
© Copyright 2024