United against a common foe? The nature and origins of Euroscepticism among left-wing and right-wing citizens

Abstract In Western European democracies opposition to the European Union is commonly found at the ideological extremes. Yet, the Euroscepticism of radical left-wing and radical right-wing parties has been shown to have distinct roots and manifestations. The article investigates whether these differences are mirrored at the citizen level. Using data from the European Election Study (2009/2014) and the European Social Survey (2008/2012) in 15 West European countries, it is found that left-wing and right-wing citizens not only differ in the object of their Euroscepticism, but also in their motivations for being sceptical of the EU. Left-wing Eurosceptics are dissatisfied with the current functioning of the EU, but do not oppose further European integration per se, while right-wing Eurosceptics categorically reject European integration. Euroscepticism among left-wing citizens is motivated by economic and cultural concerns, whereas for right-wing citizens Euroscepticism is solely anchored in cultural attitudes. These results refine the common ‘horseshoe’ understanding of ideology and Euroscepticism.


2002; De
. The Euroscepticism of radical left-wing parties stems from their defence of welfare state arrangements and from their opposition to ongoing market liberalisation. Radical right-wing Eurosceptic parties on the other hand mainly take issue with the threat that European integration poses to national sovereignty and cultural homogeneity, and though some have become increasingly concerned with protecting the welfare state, these concerns are generally intertwined with exclusivist, anti-immigrant sentiments. 1 A second distinction, which is less often made, is that these ideological opposites direct their criticism towards different aspects of European integration. A recent study shows that 'moving from broad Eurosceptical stances to more specific preferences on the integration process, extreme left and extreme right [parties] express rather distinctive views' (Conti and Memoli 2012: 93). Most parties of the radical right categorically reject any type of integration that goes beyond the most basic economic cooperation. For most radical left-wing Eurosceptic parties, Euroscepticism constitutes a rejection of the EU's current composition and practice, rather than a principled rejection of European integration in and of itself (March and Rommerskirchen 2012).
While quite some research exists on how party positions on European integration are structured, surprisingly little is known about how these attitudes are structured in public opinion. There is evidence that citizens at the extremes are the most Eurosceptic (Lubbers and Scheepers 2010;Steenbergen et al. 2007;Van Elsas and Van der Brug 2014). What remains understudied, however, are similarities and differences in the nature and origins of Euroscepticism among citizens with a left-wing or a right-wing ideology. The main reason why we expect similar structures at the level of parties and citizens is that citizens and parties at similar positions on the left-right scale can be expected to share similar values and principles. To the extent that positions of citizens and parties on matters of European integration are driven by these values and principles, a similar structure would materialise. However, citizens' attitudes are expected to be less constrained by ideology than are party positions. This raises the question whether the differences between left-and right-wing Euroscepticism found at the party level are mirrored among citizens. 2 We focus on two possible forms of heterogeneity. First, public Euroscepticism is a multidimensional concept (Boomgaarden et al. 2011), and can be directed at different aspects of European integration. We distinguish between the ideal of European integration and specifically whether further integration is needed, and the real EU as it currently exists and operates (Kopecký and Mudde 2002). Citizens on the far left as well as those on the far right are expected to be more critical towards the EU than those at the political centre, yet we expect that this criticism takes different forms depending on their ideological background. Citizens on the far right are expected to be particularly opposed to further EU strengthening as they outright reject the ideal of an 'ever closer union' . Far left-wing citizens, on the other hand, should not so much oppose the ideal as such, but rather the current functioning of the EU. Second, we expect that different motivations anchor these two dimensions of Euroscepticism. Left-wing citizens, particularly those of the radical left, are expected to be most concerned with the economic consequences of the current setup of the EU, but these economic concerns do not necessarily imply opposition to the general ideal of European integration. Further strengthening of the EU -particularly in terms of social provisions -might actually be seen as a partial solution to overcome some of the EU's perceived socio-economic downsides. Right-wing citizens, and especially the radical right, are expected to put nationalist objections to the EU front and centre. Almost by definition, transnational solutions are unacceptable for citizens with mainly nationalist concerns. For citizens with monoculturalist and related attitudes further integration is seen in an even more negative light than the EU as it currently functions.
We present a twofold analysis of the structure of citizen attitudes on the basis of data from the 2009 and 2014 waves of the European Election Study (EES) and the 2008 and 2012 waves of the European Social Survey (ESS), focusing on 15 Western European countries. 3 First, we distinguish between two dimensions of Euroscepticism: dissatisfaction with the current EU (the 'real EU'), and opposition to further EU strengthening (the 'ideal of European integration'). Our findings show that this distinction is important, as these dimensions relate differently to left-right ideology. Left-wing citizens are more critical of the current state of the EU than right-wing citizens, whereas the right is more opposed to EU strengthening than the left. In the second part of the paper, we analyse whether the motivations undergirding citizensʹ Euroscepticism are conditional on their ideology. The results indicate that dissatisfaction with the current functioning of the EU is traceable to egalitarian attitudes, but only for left-wing citizens, while opposition to further European integration is related equally strongly to nationalistic attitudes for all citizens. Strong monoculturalism leads to an outright rejection of the European project per se among all citizens regardless of their ideology, while a strong preference for national redistribution is translated into dissatisfaction with the current EU, but only among left-wing citizens.
This study contributes to the ongoing debate on whether or not, and how, the issue of European integration continues to transform the existing political space. Do EU attitudes constitute a new dimension uniting Eurosceptics of different ideological backgrounds against a common foe? Or can we actually distinguish different types of Eurosceptics whose similarities at the surface mask fundamentally diverging views on the European project? The results of this paper support the latter conclusion. Left-wing and right-wing citizens not only differ in the issue base of their Euroscepticism, but also direct their objections towards different aspects of European integration.

Left-wing and right-wing Euroscepticism
Scholars are increasingly emphasising the multidimensional nature of Euroscepticism (e.g. Boomgaarden et al. 2011). Being Eurosceptic could entail that one distrusts the institutions that together form the European Union, that one opposes specific EU policies, or that one fundamentally rejects the very idea of regional integration. By no means do these different critiques of European integration need to go together (Kopecký and Mudde 2002;Taggart and Szczerbiak 2004). Yet some have pointed out that 'a systematic engagement with potential differences in the explanatory power of different antecedents for different dimensions of EU attitudes is still lacking' (Boomgaarden et al. 2011: 243). This is particularly relevant when studying the ideological embedding of Euroscepticism.
Most of the literature on Euroscepticism and its ideological underpinnings focuses on political parties. Studies relying on expert survey data (Hooghe et al. 2002) and manifesto data (Halikiopoulou et al. 2012) consistently show that parties located at the left and right extremes in Western Europe stand out as most Eurosceptic. The 'inverted U' (Hooghe et al. 2002) has become a forceful image to illustrate this relationship between Euroscepticism and leftright ideology, suggesting that parties at the left and right extremes are united in their Euroscepticism. However, this simple image obscures the fact that radical left and radical right parties diverge in the motivations as well as the objects of their Euroscepticism. Two dimensions of Euroscepticism -directed at different objects -should be distinguished when studying how ideology relates to Euroscepticism: (1) evaluations of the 'real' EU as it currently exists and functions, and (2) a principled attitude towards the 'ideal' of European integration, reflected in preferences regarding the further strengthening of the EU (Kopecký and Mudde 2002). This distinction is particularly important for parties of the radical left, which primarily oppose the EU for its expected negative consequences for national welfare states, and the subsequent social harm that might befall vulnerable groups in society. This does not necessarily imply a principled opposition to European integration. To the extent that socio-economic grievances can be remedied at the European level, otherwise Eurosceptic left-wing parties can even support further EU integration. In the programmes of radical left-wing parties across Western Europe, we find statements such as 'Change Europe' (French Communist Party, 1999), '100% Social: No against this EU' (Dutch Socialist Party, 2014), 4 and ' Another Europe is possible' (European party group GUE/NGL, 2014). The radical left thus tends to be sceptical of the EU in its current form, while not opposing the ideal of European integration as such (March and Rommerskirchen 2012). Exceptions are a few orthodox communist parties, most notably the Greek Kommounistiko Komma Elladas (KKE) and the Portuguese Partido Comunista Português (PCP), which more fundamentally reject European integration in both its principle and practice (Charalambous 2011).
Parties of the radical right oppose the European project primarily for socio-cultural and political reasons (Hooghe et al. 2002). They fear that European integration will have harmful consequences for national interests, traditions, and identities. The economic implications of the EU trigger diverse reactions among the radical right. Some parties on the far right recognise the merits of market integration and trade liberalisation (e.g. the Austrian FPÖ and Swiss SVP). At the same time, many radical right parties, particularly in the recent period, have put themselves forward as defenders of national welfare provisions, albeit on exclusionary grounds (Lefkofridi and Michel 2014). This economic protectionism is blended into their nationalist discourse centred on countering immigration and maintaining national sovereignty. By definition, this discourse of nationalism is antithetical to furthering integration, and thus primarily results in a principled rejection of the European project. However, since the focus of the EU in the early 1990s shifted away from market liberalisation to political unification, the current setup of the EU infringes on the national sovereignty of its member states. Many far right parties therefore not only condemn the furthering of European integration in itself, but also negatively evaluate the current functioning of the EU, where 'Brussels' often serves as an umbrella term for the all too meddlesome European institutions. Euroscepticism of (radical) right parties is thus in the first place a principled opposition directed against further integration towards (ideally) a federal union. Yet sovereignty-based objections also anchor dissatisfaction with the current EU.
We expect the fundamental ideological premises that structure parties' EU positions to be mirrored at the level of citizens for two reasons. First, left-right ideology is related to certain values and principles, which are shared by parties and citizens of similar ideological positions. To the extent that EU attitudes are also driven by these underlying values, we would expect parties and voters at similar left-right positions to partially share these attitudes as well. Second, various studies have shown that elites shape public opinion with regard to European integration (Gabel and Scheve 2007;Ray 2003;Steenbergen et al. 2007), while there is also evidence for bottom-up influences where citizens' preferences steer party positioning (Carrubba 2001). Even though patterns in public attitudes are less tightly structured than the positions of parties (Converse 1964), this mutual influence can be expected to reinforce the similarities between parties and citizens.
Only a few empirical studies examine whether Euroscepticism is a different attitude for left-wing as opposed to right-wing citizens. Lubbers and Scheepers (2010) distinguish three types of attitudes towards Europe. While their study does not theorise on the relationships between different kinds of Euroscepticism and left-right orientations, and while they employ a different typology than we do, their results suggest that distrust towards the current EU is stronger among left-wing citizens, whereas citizens on the right are more fundamentally opposed to European integration (Lubbers and Scheepers 2010). In a Dutch case study, Boomgaarden et al. (2011) distinguish as many as five dimensions of EU support. Of these five, only one appears to be related to left-right in a curvilinear way. All in all, the research on this topic remains scarce and the results are quite inconclusive.
Our general expectation concerning the relationship between citizen ideology and attitudes towards the EU is as follows: H1a: The relationship between left-right ideology and dissatisfaction with the current EU is curvilinear, with citizens at the far right and far left being least satisfied.
H1b: The relationship between left-right ideology and opposition to further EU strengthening is linear, with opposition being strongest among far right-wing citizens, and decreasing as citizens are positioned more to the left.
In addition to the object of citizens' opposition, there is also good reason to distinguish between the motivations of left-wing and right-wing citizens for being Eurosceptic. Van Elsas and Van der Brug (2014) show that left-wing and right-wing citizens in Western Europe are sensitive to different drawbacks of European integration, with the left being more fearful of losing social benefits than the right, and the right more afraid of eroding national identity than the left. Yet the survey items ask people directly how fearful they are that European integration will lead to a 'loss of social security' , a 'loss of national identity' , etc. Someone could state that they are not afraid that the EU will lead to a loss of national identity, because they think national identities will stay intact, or because they simply do not care about these identities. Moreover, the questions pertain to the EU in general, rather than to more specific forms of Euroscepticism. For a more definitive test of the origins of different EU attitudes among left-and right-leaning citizens, we should look at correlations between independent measures of Euroscepticism and economic and cultural attitudes respectively, and assess to what extent these correlations differ between left-wing and right-wing citizens. To our knowledge only one study directly examines this matter. In a French case study, Evans (2000) compares voters of the French Communist Party and the Front National in how their Euroscepticism relates to economic and cultural attitudes. He finds that for PCF voters Euroscepticism is more closely related to economic attitudes than for FN voters (vice versa for cultural attitudes). This indicates that Euroscepticism has different roots for radical left-and right-wing voters, but as of yet we do not know to what extent this finding can be generalised to other countries, to larger groups of citizens and to a more recent time period.
In general, we expect that for left-wing citizens economic concerns (i.e. regarding redistribution and the role of the government in the economy) are a relatively more important reason for being Eurosceptic than for right-wing citizens, whereas for right-wing citizens cultural concerns (i.e. regarding immigration and national identity) are a more important predictor of their views towards the EU. However, the attitudinal roots of Euroscepticism also depend on the specific dimension of Euroscepticism under consideration. Previous research on the relation between Euroscepticism and economic and cultural issue dimensions has shown that EU attitudes are consistently and strongly related to cultural positions (Kriesi et al. 2008; Van der Brug and Van Spanje 2009), and less so to economic attitudes (Costello et al. 2012;Garry and Tilley 2014). No study has however distinguished between different dimensions of EU attitudes. It is likely that their relationship to economic attitudes only exists for specific dimensions of Euroscepticism. As argued, economic objections to the EU are mainly directed at how the EU currently functions, but not at the principle of European integration in and of itself. On the other hand, cultural objections against the EU are directed at both the current EU and its possible future strengthening. Therefore, we formulate different sets of hypotheses for the two dimensions of Euroscepticism and how they are motivated by economic concerns (here conceptualised as attitudes regarding national income redistribution) and cultural concerns (conceptualised as anti-immigrant sentiments). Dissatisfaction with the current EU is expected to be related to both economic and cultural concerns, but to different degrees depending on ideological positions, as is formalised in the following hypotheses: H2a: Support for redistribution has a positive effect on dissatisfaction with the current EU, but this effect is strongest among far left-wing citizens, and decreases as citizens are positioned more to the right.
H2b: Anti-immigrant sentiments have a positive effect on dissatisfaction with the current EU, but this effect is strongest among far right-wing citizens, and decreases as citizens are positioned more to the left.
With regard to opposition towards further EU strengthening, we expect cultural concerns to be dominant for all citizens, for three reasons. First, the most obvious consequence of the deepening and particularly the widening of the EU is increasing diversity. The opening of borders facilitates labour migration within the EU, and the enlargement of the EU to member states in Central and Eastern Europe has facilitated labour migration to the old member states. In addition, in this context member states have less control on the influx of asylum seekers, which is of particular concern to citizens with strong monoculturalist attitudes. Second, opposition to further European integration is a more principled EU attitude, and is therefore likely to be related to one's value system, or worldview. Citizens with an internationalist, cosmopolitan worldview hold values of openness and cooperation beyond national borders. They view the national level as but one level to solve collective action problems, and are open to politics at a higher -European -level. On the other hand, citizens with a communitarian, nationalist worldview consider the nation to be the only political unit that protects their interests (Evans 2000;Teney et al. 2013). These opposed worldviews drive positions on the cultural dimension as well as support for the principle of further European integration, thus reinforcing the relationship between the two attitudes. Third, we expect opposition to EU strengthening not to be related to economic attitudes, precisely because with regard to redistributive concerns, strengthening can be seen as both a threat and a solution. As we have seen, many radical left parties aim at a different Europe rather than no Europe at all. Typical left-wing objections to the EU are thus not necessarily extended to the European project as such.
H3: Anti-immigrant sentiments have a positive effect on opposition to further EU strengthening, and this effect is equally strong for all citizens irrespective of their left-right positions.

Data
To test the hypotheses we draw on data from two waves of the European Election Study (2009 and 2014) and two waves of the European Social Survey (2008 and 2012). 5 We combine different data sources for two reasons. First, the EES and ESS have different assets in terms of the included items. The EES 2009 allows us to distinguish between two dimensions of Euroscepticism based on various survey items. These data are suitable to map the two EU dimensions and study their relationship to left-right self-placement (H1a and H1b). However, in these surveys the measures of attitudes towards immigration and redistribution have very skewed distributions. This makes them ill-suited for testing how their relationship with Euroscepticism differs between left-wing and right-wing citizens (H2a, H2b and H3). For this purpose we use the ESS 2008, which enables us to construct more refined and balanced scales to measure attitudes towards redistribution and immigration, and thus to assess how the relationship between these attitudes and Euroscepticism differs between leftand right-wing citizens. By necessity, in the ESS analyses we use single-item measures of the two dimensions of Euroscepticism identified in the EES 2009.
Second, recent years have been particularly turbulent for the EU and its member states. The euro crisis has left an imprint on a wide range of citizens' attitudes (e.g. Bermeo and Bartels 2013). Although the EES 2009 and ESS 2008 were conducted during the global financial crisis (from 2007 onwards), the subsequent debt crisis in the eurozone (from December 2009 onwards) has been found to have the greatest impact on EU attitudes (Braun and Tausendpfund 2014). Therefore, we present replications of all analyses on the basis of the EES 2014 and the ESS 2012 in order to assess to what extent the findings hold during the euro crisis. By necessity, these analyses are based on fewer items to measure core concepts. Finally, we assessed whether the results hold outside times of crisis by replicating the analyses for two pre-crisis waves (EES 2004 andESS 2004, again by necessity relying on fewer items). The results strongly resemble the main findings, particularly those of 2009. For reasons of conciseness, we refer the reader to Online Appendix C for these results.
The hypotheses are developed on the basis of the structure of Western European party systems, so the analyses are limited to 15 Western EU member states: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Italy, Ireland, Luxembourg, The Netherlands, Portugal, Spain, Sweden, and the United Kingdom. The ESS analyses are conducted on 12 countries, since Austria and Luxembourg were not included in either of the ESS waves, and Greece is included in 2008 but replaced by Italy in 2012.

Measuring Euroscepticism
Different operationalisations of Euroscepticism appear in the literature. In many instances, the choice of indicators depends to a large extent on the practical availability of survey items. Many studies have used the EU membership evaluation question, the principal advantage of which is its wide availability (Anderson 1998;Gabel 1998;Hakhverdian et al. 2013;Steenbergen et al. 2007). Other studies use support for further integration (Evans 1998;Kriesi et al. 2008) or the desired speed of integration (De Vries and Edwards 2009), while still others use some form of EU institutional trust (Armingeon and Ceka 2014). Given our theoretical argument, we expect the outcomes of these studies to be in part a function of the specific sub-dimension of Euroscepticism tapped by the item. Constructing an index of multiple items does not by definition solve this. If one combines items on confidence in the current functioning of the EU with items on further integration (see Garry and Tilley 2014 for a recent example), this is likely to introduce noise into the relationships with covariates such as ideology. Combining items that load on the same theoretical sub-dimension, however, will improve the reliability and validity of the findings, and this is what we aim for in our operationalisation.

EES 2009
On the basis of the EU-related questions included in the EES 2009, we discern two dimensions of Euroscepticism. First, dissatisfaction with the current functioning of the EU is measured by the following four items (Cronbach's α = 0.77): 'How satisfied are you with the way democracy works in the EU?' (1-4), ' Agree/ Disagree: You trust the institutions of the EU' (1-5), ' Agree/Disagree: The EU parliament considers the concerns of citizens' (1-5), and 'How much confidence do you have that the decisions made by the EU are in the interest of your country?' (1-4). Second, opposition to further European integration is measured by two items (α = 0.64): 'European unification has gone too far or should be pushed further' (0-10) and 'EU enlargement is good or bad' (1-3). All items were recoded in such a way that high scores reflect negative attitudes towards the EU. Subsequently, they were standardised and then added up to form scales. The resulting scales were again standardised, in order to obtain fully comparable scales representing the two EU dimensions.
Left-right ideology is measured by using the familiar left-right self-placement scale, which ranges from 0 to 10. We standardise the left-right scale and include it in the analysis together with its quadratic term.

EES 2014
Not all EES 2009 items reappear in the 2014 EES wave. In 2014 dissatisfaction with the current EU is measured by two items: 'You trust the institutions of the EU' (1-4), and 'The European parliament takes into consideration the concerns of European citizens' (1-4) (Cronbach's α = 0.77). These items have the same question wording, yet a slightly different response scale. 6 Opposition to EU strengthening is measured by a single item: 'European unification has gone too far or should be pushed further' (0-10). Left-right ideology is measured on the same 0 to 10 scale as in the EES 2009.

ESS 2008/2012
The ESS 2008 and 2012 both include two items that measure EU attitudes, which can be matched to the two dimensions distinguished in this study. As a proxy for dissatisfaction with the current EU, we use the item 'Distrust in the European Parliament' , measured on a scale from 0 (complete trust) to 10 (no trust at all). Conceptually, trust -like dissatisfaction -is based in large part on current evaluations (Hardin 1999). Though the European Parliament constitutes a pars pro toto, the item overlaps in content with the two EES items that are included in the 'dissatisfaction with the current EU' scale in 2009 and 2014 (on institutional (dis)trust and on the European Parliament). For opposition to EU strengthening, we use the item 'European integration has gone too far, or should go further' (0-10), which is very similar to one of the two items we used in the EES 2009.
Left-right ideology is again measured by the self-placement scale, running from 0 (left) to 10 (right). For measuring support for redistribution and anti-immigrant attitudes, the ESS 2008 enables us to use three-item scales. In 2012, the immigration scale consists of the same three items, but attitudes towards redistribution are measured by a single item (see Table 1). Items were standardised before combining them into a scale, which is constructed by adding up the items and dividing the resulting scale by three (the number of items). Higher scores on these scales represent supportive attitudes regarding redistribution and negative attitudes regarding immigration. Again, to enable the comparison of effect sizes of indicators within and across models, both the independent and dependent variables are standardised.

Control variables
Hypotheses H1a and H1b are concerned with the functional form of the relationship between left-right ideology and EU opposition, not with establishing any causal relationship between variables. We therefore do not include any control variables in the first part of the analyses based on the EES 2009/2014 (but see Online Appendix D for the same analyses with demographic controls, which show that the results are highly robust). Hypotheses H2a, H2b and H3, on the other hand, make predictions about the motivations leading to Eurosceptic attitudes. Here we do include a set of demographic controls, consisting of age, gender, level of education (measured in five categories) and social class (measured in nine categories). 7

Method
We study the relationship between Euroscepticism, ideology and issue attitudes by means of several linear regressions. Though H1a and H1b focus on relationships rather than causal effects, we model Euroscepticism as the dependent variable as this enables us to gauge the curvilinear nature of the horseshoe model, and ensures consistency with the analyses for H2a, H2b and H3. The data sets include respondents from 12 to 15 countries. We present pooled models including country fixed effects to control for the nested structure of the data, a strategy that fits our interest in relationships at the individual level. 8 Yet we are aware that the ideological underpinnings of Euroscepticism may differ between countries due to macro-economic factors (Garry and Tilley 2014), the presence of left-wing or right-wing Eurosceptic parties (De Vries and Edwards, 2009), or possibly also due to a general Eurosceptic 'mood' in a country. The present study

ESS 2008
ESS 2012 Support for redistribution Government should reduce differences in income levels (1-5) X X For a fair society, differences in standard of living should be small (1-5) X Differences in income are acceptable if to reward talent and effort (1-5) X α = 0.57 -Anti-immigrant sentiments immigrants make country a better or a worse place to live (0-10) X X country's cultural life is undermined or enriched by immigrants (0-10) X X immigration is good or bad for the economy (0-10) X X α = 0.86 α = 0.85 focuses on common patterns within countries rather than between countries, but we do present country-specific analyses to assess to what extent the pooled results hold for individual countries (see Online Appendix A). 9  less satisfied with the current EU than are right-wing citizens. The curvilinear term is significant and positive (b = 0.03), indicating that dissatisfaction is stronger among citizens at the ideological extremes. Figure 1a gives insight into the substantive strength of the relationships. The predicted value of a far left-wing citizen (0) on the EU dissatisfaction scale is 0.21, whereas this is -0.06 for a citizen of the far right (10). This gives partial support to H1a. Left-wing citizens are, as expected, more dissatisfied than the political centre. For citizens of the far right, however, there is no significant difference -but at least the far right is certainly not more satisfied than citizens in the centre. In 2014 there is stronger support for H1a. The linear relationship of dissatisfaction with the current EU on left-right is again negative, but no longer significant (b = -0.01), while the curvilinear term remains significantly positive (b = 0.02). As Figure 2a shows, this results in the familiar U-shaped relationship. Left-wing citizens are still the most dissatisfied, yet the differences between the far left and the far right have become less pronounced. Our analyses cannot tell whether this is because the left has become more satisfied or the centre and right have become less satisfied, as we cannot compare absolute levels of EU dissatisfaction between 2009 and 2014 due to measurement differences (see note 7). 10 However, recent studies have documented a strong rise in Euroscepticism during the eurozone crisis (Armingeon and Ceka 2014;Braun and Tausendpfund, 2014). This suggests that the shape of the relationship changed mainly because of rising dissatisfaction among the right and centre.

Left-right and the two dimensions of Euroscepticism
The results for opposition to EU strengthening show a very different pattern. coefficient of the curvilinear term is not significant. Thus, as Figure 1b shows, the relationship is linear, with the predicted values running from -0.13 for the far left to 0.11 for the far right. In 2014 the curvilinear term is significant and positive (b = 0.03). Yet, as Figure 2b shows, opposition to EU strengthening is still clearly stronger for the right. The predicted value of support for EU strengthening runs from -0.02 for the far left to 0.21 for the far right. We thus find strong support for H1b in both years: opposition to strengthening of the EU is predominantly a right-wing attitude. Yet it is important to note the tendency towards increased left-wing opposition in 2014. After the euro crisis, we see glimpses of a horseshoe pattern for both EU dimensions, while Euroscepticism among the left and right was clearly more distinct before the euro crisis broke out. We will return to this finding in the discussion. These pooled results mask some cross-national heterogeneity. Countryspecific graphs for 2009 (Online Appendix A1/A2) show that in the UK, and to some extent in Austria, the right is the most Eurosceptic on both dimensions. Both countries have witnessed a drastic shift of Euroscepticism from the left to the right in recent decades (Evans 1998;Pelinka 2004). In Portugal and Sweden, we find that the left is most negative on both dimensions. Nevertheless, the majority of countries conform to the patterns found in the pooled results. The country-specific results of 2014 (Online Appendix A3/A4) show less pronounced differences between the two EU dimensions. In Finland, Greece and again Portugal and Sweden, both dimensions spark mainly opposition from the left. In Italy we now see that, similar to the UK and Austria, the right is the most negative on both dimensions.

What motivates left-wing and right-wing Euroscepticism?
The second set of analyses assesses to what extent the motivations for Euroscepticism are different for left-and right-wing citizens. These analyses are conducted by means of pooled fixed effects regression models, employing the ESS 2008 and 2012, and include a set of demographic control variables. The crucial tests of our second set of hypotheses rely on the interaction effects between left-right ideology and two attitude scales on the two dimensions of Euroscepticism. These interactions are presented in Table 3.
We find support for H2a in both 2008 and 2012. Distrust in the European Parliament -as a proxy for current EU dissatisfaction -is positively affected by support for redistribution in both years (b = 0.04 and 0.03, effect for centrist citizens), and this effect is stronger for left-wing than for right-wing citizens, as is shown by the negative interaction term (b = -0.05 and -0.04). To facilitate interpretation, Figures 3a and 4a show these results graphically. In both waves we find that for left-wing citizens, support for redistribution has a positive effect on distrust in the EP. This effect decreases as citizens are more right-wing, supporting the expectation in H2a that economic concerns are more important predictors of current EU dissatisfaction among left-wing than among rightwing citizens. Interestingly, among far right-wing citizens we find an inverse (negative) effect. Support for redistribution makes them more satisfied with the current EU. This unexpected negative effect suggests opposition to the increased regulatory role of the EU that exists among right-wing citizens with strong neoliberal preferences. Additional inspection of the marginal effects by country (displayed in Online Appendix B) shows that this effect is particularly marked in the UK (a country where this critique is indeed salient), though in 2012 it is significant in none of the countries. The positive effect of support for redistribution on current EU dissatisfaction on the left is consistent across countries. In 2008 it is positive in all 12 countries, and significantly so in seven of them (Table B1a in Online Appendix B). In 2012 (Table B1b) the results   Table 3. interactions between left-right and support for redistribution and anti-immigrant sentiments.
note: Fixed effects model of 12 countries. left-right self-placement, support for redistribution and antiimmigrant sentiments are measured by standardised scales. standard errors in parentheses. one-tailed significance indicated by asterisks: *** p < 0.001; ** p < 0.01; * p < 0.05. source: ees (2009/2014). across countries are slightly less consistent, but we still find a positive and significant effect in six out of 12 countries. These findings thus generally support H2a, as they show that concerns about redistribution mainly play a role for left-wing (and to a lesser extent for centrist) citizens. Unexpectedly, for opposition to EU strengthening we find a similar -yet much less pronounced -pattern, indicating that for left-wing citizens this   dimension is affected by redistributive attitudes too, albeit to a much lesser degree. Figures 3b and 4b display the marginal effect of support for redistribution on opposition to EU strengthening alongside citizen ideology. The graphs look very similar between the two years. For left-wing citizens, there is a weak positive effect of support for redistribution on opposition to EU strengthening. Countryspecific results show that in 2008 this effect is mostly driven by Sweden, Finland, Denmark and France, where it exists among left-wing (and centrist) citizens (Online Appendix B, Table B2a). In 2012, the effect again appears among leftwing citizens in Sweden and Finland, and to a lesser extent in The Netherlands, Belgium and Ireland (Table B2b). Redistributive concerns are thus somewhat related to opposition to EU strengthening, but much less strongly and consistently than to dissatisfaction with the current EU. In most countries, there is no effect of support for redistribution on opposition to EU strengthening, for leftwing or for right-wing citizens. This lends credibility to the expectation under hypothesis H3 that opposition to EU strengthening as a more principled attitude is not so much related to economic concerns but rather to cultural attitudes (such as anti-immigrant sentiments) for all citizens regardless of their ideology. We find mixed support for H2b. Anti-immigrant sentiments have a strong and positive effect on distrust in the EP in both years (b = 0.28 and b = 0.30, effect for centrist citizens). However, the expectation in H2b that this relationship is stronger for right-wing citizens is only corroborated equally strong for left-wing and right-wing citizens. In 2012, however, the effect clearly becomes more strongly positive towards the right of the left-right scale (Figure 6a). Comparing the 2008 and 2012 graphs shows that this change is due to an increase in the effect on the right. 11 During the euro crisis, the EU opposition of right-wing citizens has become even more firmly anchored in cultural positions. Similar differences between 2008 and 2012 can be observed for opposition to EU strengthening and anti-immigrant sentiments (Figures 5b and 6b). The 2008 result is in line with H3 that anti-immigrant sentiments have an equally strong effect on EU strengthening attitudes for the left, centre and right, while in 2012 the effect is somewhat stronger on the right. Another important finding is that in both years the effect of immigration attitudes is much stronger with regard to opposition to EU strengthening than with regard to current EU dissatisfaction. This underlines the particular importance of cultural attitudes as a predictor for support or opposition to the ideal of European integration. Tables B3 and B4 in the Online Appendix show the effect of immigration attitudes on EU opposition for left-and right-wing citizens per country (in 2008 and 2012). These results demonstrate the remarkable consistency of the effects across the two EU dimensions, across left-right ideology, and across countries.

Conclusion
The issue of European integration is often regarded as a potential new line of conflict in Western European societies, cross-cutting the existing conflict dimensions and possibly forging new alliances between Eurosceptic actors. Indeed, when looking at very general measures of EU support and opposition, ideological extremes seem to agree in their Euroscepticism, as is commonly illustrated by the well-known 'horseshoe'-shaped relationship between leftright positions and Euroscepticism. However, party-level research has shown that the similarities between left-wing and right-wing Eurosceptic parties are rather superficial in nature. Though these parties can sometimes be united in their opposition to the EU, the nature of their Euroscepticism differs tremendously. This study has shown that the same applies to citizens. While at first glance citizens on the ideological extremes appear united in their opposition to the EU, substantial differences exist between left-wing and right-wing citizens in the objects of their Euroscepticism, as well as in the motivations that underpin their Euroscepticism. The first source of variation concerns the nature of Euroscepticism itself. Previous studies have shown that Euroscepticism is a multidimensional attitude; citizens can oppose the EU in some regards but not in others. We have shown that when we break down general Euroscepticism into two sub-dimensionsdissatisfaction with the current EU and opposition to EU strengthening -the horseshoe falls apart. Left-wing citizens are relatively more dissatisfied with the current EU than right-wing citizens, whereas right-wing citizens oppose future strengthening of the EU more than their left-wing counterparts.
Second, left-wing and right-wing citizens also differ in their motivations for being Eurosceptic. Again, the results depend on the specific EU dimension under study. For left-wing citizens, support for redistribution increases dissatisfaction with the current functioning of the EU, whereas these egalitarian attitudes have no effect on rejecting further EU integration per se. Still, the Euroscepticism of left-wing citizens is also to a large extent driven by cultural attitudes. Right-wing citizens are much less ambivalent in their Euroscepticism. Their opposition to the EU, be it in terms of its current functioning or possible future form, is anchored solely in cultural attitudes.
These findings have implications for our understanding of the origins and nature of Euroscepticism, and carry recommendations for future empirical research on the subject. To start with the latter, we have shown that when studying the correlates of EU attitudes, it is crucial to distinguish between different dimensions of Euroscepticism. Extant research relies mostly on available indicators of Euroscepticism, and to the extent that these indicators represent different dimensions of Euroscepticism, they can produce fundamentally different results. For instance, previous research has produced inconsistent findings on whether EU attitudes are best interpreted as part of a cultural dimension of conflict, or whether they relate to both the cultural and the economic dimensions of the political space. Though there is evidence that citizens' positions on a socio-economic dimension matter for their Euroscepticism (e.g. Costello et al. 2012;Garry and Tilley 2014), this relationship does not come out as strongly in other studies (Kriesi et al. 2008; Van der Brug and Van Spanje 2009). Our findings imply that studies using indicators reflecting more principled EU attitudes are likely to find weaker relationships to economic attitudes (and stronger relationships to cultural attitudes) than studies that operationalise Euroscepticism as dissatisfaction with the current EU.
Our results also give insight into how the EU issue produces ambivalence among the political left. Generally, the Western European left combines economically left-wing and culturally progressive positions. Our study shows that at the level of citizens, these attitudes are in conflict with regard to Europe: egalitarians are sceptical towards the current functioning of the EU, while their culturally more cosmopolitan and universalist attitudes lead to a positive evaluation of European integration as an ideal. Left-wing citizens might therefore reject the current EU, but can at the same time find themselves supporting a different and 'better' (i.e. more social) Europe. These findings at the citizen level have implications for parties as well. Radical left-wing Eurosceptic parties will have to pursue a rather complex and nuanced combination of critical positions on European issues, particularly when compared to the categorical opposition to Europe voiced by the radical right.
In addition, our findings tentatively indicate that the economic crisis might cause shifts in how Euroscepticism is ideologically embedded. The type of EU criticism held by left-wing and right-wing citizens appears to have become more similar in 2014 as compared to 2009 (and 2004, see Online Appendix C), as right-wing citizens are increasingly dissatisfied with the current EU, and leftwing citizens in some countries seem to have picked up some more principled EU opposition. At the same time, the motivations for their Euroscepticism remain distinct. Economic concerns remain relevant only to those on the left, whereas cultural concerns have become more important for the right than for the left. In the near future, it is certainly possible that shared foes provide a basis for temporary Eurosceptic coalitions. Yet, due to the differences in the underlying reasoning, the formation of a stable Eurosceptic front cross-cutting the left-right dimension seems unlikely.
Finally, the superficial nature of the horseshoe model of ideology and Euroscepticism speaks directly to current debates on rising levels of opposition to the EU across member states. Our analyses have shown clearly that there is not one uniform type of Eurosceptic voter. Rather, European citizens are in disagreement on what it is they are sceptical about, as well as their motivations for being Eurosceptic in the first place. Those on the radical right that categorically reject any form of regional cooperation beyond a bare minimum are culturally motivated in their actions and beliefs. In sharp contrast, citizens on the radical left can actually demand further integration in some areas to align European policy with their redistributive ideals. This means that no silver bullet exists for reducing the EU's democratic deficit and the lack of representation (e.g. Follesdal and Hix 2006). Rather, it calls for a more comprehensive form of policy contestation than the EU has hitherto seen. Notes 1. Evidently, the radical left and radical right are not polar opposites. Rightwing Eurosceptic parties are increasingly concerned with welfare issues, albeit through an exclusivist lens. Likewise, the economic protectionism of the left can also be seen as a specific form of nationalism (Halikiopoulou et al. 2012). Though economic and cultural concerns are thus sometimes blended, the relative emphasis on either fundamentally differs between Eurosceptic actors of left-wing and right-wing lineage. 2. The terms 'left' and 'right' have multiple meanings to citizens, as the left-right dimension functions as a 'super issue' that absorbs attitudes towards various issues. Certain attitudes are considered left-wing (e.g. support for redistribution, multiculturalism) whereas others are associated with the right (e.g. economic liberalism, monoculturalism). In case of conflicting issue positions in terms of left-right, left-right identification can be adapted based on the issues most salient to the voter (Weber and Saris 2015). Left-right is thus meaningful even if there is a pluralization of issues it is associated with. 3. Our research question does not apply to Eastern European countries, since in these countries Euroscepticism is not related to left-right ideology in a horseshoe pattern, but rather concentrated at the left end of the spectrum (Marks et al. 2006). 4. Translated from source language statements: 'Bouge l'Europe' (PCF (FR), EP elections manifesto 1999) and '100% Sociaal: Nee tegen deze EU' (SP (NL), Official publication of scientific bureau 4/2014). 5. The EES data of 2009 and 2014 can be accessed online via the EES homepage at http://eeshomepage.net/. The ESS data of 2008 and 2012 are accessible at http://www.europeansocialsurvey.org/. 6. The main difference is that the 2014 items have no neutral category. For our main purpose of analysing the relationship with left-right placement, this is not problematic. It does however make a comparison of the absolute levels more difficult. 7. Education is measured as the highest level completed (five levels based on the International Standard Classification of Education (ISCED), combining levels 5 and 6). Class is measured in nine categories based on the International Standard Classification of Occupations (ISCO08). 8. Hausman tests comparing fixed to random effects models are significant for five of the eight main models (based on both the EES 2009/2014 and ESS 2008/2012), indicating that fixed effects are in most cases preferable. To account for the nested structure, fixed effects are also preferable over clustered standard errors given that the latter method has been shown to produce over-conservative results (see Huang 2016 for a simulation study), particularly inflating the level-1 standard errors, and particularly when within-cluster sample sizes are large.
Online Appendix E shows that despite considerable inflating of standard errors, most findings are robust to the estimation of clustered standard errors (in fixed effects models). 9. Though all countries included are EU member states, they are not all eurozone members. Since particularly in times of crisis we might expect different opinion dynamics in eurozone and non-eurozone countries, we assessed whether the results hold equally in both categories of countries (see Online Appendix F for the results and a discussion). The results are highly robust for the eurozone countries, and moderately robust for the non-eurozone countries (though deviations can be explained by the particularity of these countries, which are Sweden, Denmark and the UK). 10. We ran additional models for 2009 where we used the same two-item scale ('Trust in EU institutions' and 'The European parliament considers the concerns of citizens') as in 2014. The results are very similar, yet still not comparable in absolute terms due to different response scales. 11. Strictly, we cannot draw this conclusion from these graphs as they are standardised within years. Additional analyses (not shown here) standardising the dependent variables across the two waves combined show a very similar pattern.

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