The reshaping of political conflict over Europe: from pre-Maastricht to post-‘Euro crisis’

Abstract Party competition over European integration is structured by two main dimensions of political conflict: a socio-economic dimension (market liberalisation vs. a more regulated economy) and a socio-cultural dimension (libertarian, cosmopolitan values vs. authoritarian, nationalist values). This article investigates the relationship between these conflict dimensions and parties’ positions towards EU issues across time and space, in particular focussing on two ‘critical junctures’ in the European unification process. For this purpose, analysis is made of the election manifestos of parties competing in European Parliament elections (Euromanifestos) from 1979 to 2014. First, it is found that the key moment of the Maastricht treaty significantly reshaped party competition over Europe. After Maastricht, positions towards European integration have become less connected to the economic dimension and much more related to the cultural dimension in Western Europe. Second, it is contended that the Euro crisis has not dramatically restructured political conflict over European integration.

traditional, authoritarian, nationalist values) by Hooghe et al. (2002), whereas others labelled it the 'integration-demarcation' (Kriesi et al. 2006(Kriesi et al. , 2012 or 'cosmopolitan-communitarian' conflict (Teney et al. 2014). Second, the European integration issue has become an increasingly relevant object of public controversy and party competition (Marks 2004). Despite a more and more Eurosceptic public opinion, mainstream parties downplayed the EU integration issue for too long, turning it into a 'sleeping giant' (van der Eijk and Franklin 2004). As a consequence, political entrepreneurs and challenger parties from the ideological extremes exploited this situation and mobilised electoral support (de Vries and Edwards 2009), especially during the European economic and sovereign debt crisis (Hobolt and Tilley 2016). 1 In this article, we address these different streams of literature and describe the evolving nature of the relationship between underlying political conflicts and party competition over the issue of European integration. Building on the findings of Prosser (2016), who shows that party positions on EU issues can be largely explained by a 'two-dimensional approach' and that 'the relationship between ideology and European integration has changed substantially over the history of European integration' (Prosser 2016: 731), we continue to investigate this subject across time and space. We do so by analysing the impact of specific 'critical junctures' 2 in the European unification process, which led to major 'readjustment[s] of issue associations' (Rovny and Whitefield 2019: 9), especially between political conflict dimensions and EU integration positions.
More specifically, we argue that the exact association between party positions towards European integration and underlying political conflicts has readjusted twice during the integration process. A first readjustment should have occurred in the direct aftermath of the Maastricht treaty, when the nature of the unification process changed from an economic cooperation to a genuine political project. Party positions on European integration in Western Europe, which were originally strongly influenced by socio-economic issues, should have consequently become much more related to the non-economic, socio-cultural conflict dimension. A second readjustment should have occurred during the European economic and sovereign debt crisis, when the pathways of party competition diverged strongly between EU member states. In the most hard-hit countries in Southern Europe, economic determinants of EU issue positions should have experienced an unexpected comeback as a consequence of austerity measures and embodied by radical left parties (Hutter et al. 2018;Otjes and Katsanidou 2017). In Western Europe, in contrast, the crisis should not have reinforced the old economic conflict, but rather accelerated a general 'cultural backlash' (Norris and Inglehart 2019) that further strengthened the explanatory power of the cultural conflict dimension.
We test these hypotheses by analysing a unique long-term, crossnational dataset comprising the election manifestos of parties competing in European Parliament elections (Euromanifestos) from 1979 to 2014 . For that purpose, we apply a novel empirical approach developed by Prosser (2014) to extract ideological dimensions from party manifestos. In summary, our comprehensive analysis finds that the watershed moment of the Maastricht treaty has dramatically altered party competition over European integration, but we contend that this has been much less the case for the events surrounding the Euro crisis.

The origins of party positions towards European integration
Scholars who investigate the origins of party positions towards European integration sometimes disagree whether these positions are a function of 'strategy or ideology' (Kopeck y and Mudde 2002: 319). Ideological approaches, on the one hand, view positions on specific policy issues as being derived from a party's ideological profile and, hence, from existing social cleavages and political conflicts (Lipset and Rokkan 1967;Marks and Wilson 2000). Contemporary cleavage theory identifies two major conflict dimensions that structure political competition in (Western) Europe: first, a socio-economic dimension stemming from the old class conflict between capital owners and workers, and second, a more recent value-based, socio-cultural dimension 3 with different scholarly denotations, as mentioned above (see e.g. Kriesi 2010). 4 From cleavage theory 5 perspective, party positions are historically rooted and relatively stable. Changes in party competition happen rather slow-paced and 'it usually takes epochal change to redraw the battle-lines.' (Hooghe 2007: 6).
Initially, European integration scholars focused on the relationship between the general left-right axis and voters' or parties' preferences towards European integration. This led to the famous 'inverted-U curve' (Aspinwall 2002;Hix and Lord 1997;Hooghe et al. 2002) that describes 'support for European integration among centrist parties, and opposition among parties toward the extremes of both left and right' (Marks 2004: 239). Prosser (2016), however, demonstrates that distinguishing the economic left-right dimension from the cultural liberal-conservative dimension offers greater explanatory power than a one-dimensional approach when relating political conflicts to EU issue positions. Moreover, the author stresses the need to conceive of European integration as a dynamic process when studying the structuring power of ideological conflicts on parties' policy positions.
Strategic approaches, on the other hand, understand party positions as more flexible and affected by short-time considerations. From this perspective, parties choose and change their positions strategically to mobilise supporters and maximise their vote shares (Downs 1957). Strategic accounts explain how and why individual parties deal with policy issues in electoral competition, for example, why they decide whether to ignore or downplay an issue (Green-Pedersen 2012), why they act as 'issue entrepreneurs' (de Vries and Hobolt 2012), why their government participation matters (Sitter 2001) or why they are responsive to public opinion and preferences among their voters (Ezrow et al. 2011). Numerous empirical studies show that political parties change their positions towards European integration on the basis of strategic considerations. Mainstream parties, for example, react to the behaviour of Eurosceptic competitors (Williams and Ishiyama 2018) and become more negative towards Europe when being confronted with public support for anti-EU challenger parties (Meijers 2017;Spoon and Williams 2017). And even far-right parties change their positions towards Europe as a response to the changing public mood in times of the EU's multiple crises (Braun et al. 2019).
We believe that both approaches offer relevant insights into the origins and determinants of party positions towards the issue of European integration. EU issue positions reflect long-term ideological conflicts that structure party competition (Arnold et al. 2012;Marks et al. 2002;Marks and Wilson 2000), but they can also vary in the short term based on strategic considerations. This is particularly the case since EU issues have become electorally relevant. This finding is supported by empirical research on elite-mass linkages in the context of European integration, which demonstrates that the relationship between political parties and voters is of reciprocal nature. While parties respond to the changing preferences of their electorates, they also shape the view of their supporters (Carrubba 2001;Schmitt and Thomassen 2000;Steenbergen et al. 2007).
In line with these findings, we combine ideological and strategic explanations of party positions towards European integration when laying out our theoretical expectations in the following section. By doing so, our paper offers a comprehensive approach to understand the reshaping of political conflict over Europe from the pre-Maastricht era to the post-'Euro crisis' period and across different EU member states. Importantly, however, we do not aim to explain positional changes of individual parties, but to provide a more general and descriptive picture of the relationship between underlying political conflicts and issue positions in European party systems.

The changing relationship between political conflict dimensions and party positions towards European integration
Evidently, the nature of European integration has changed considerably over the course of the unification process, which is why the EU is sometimes described as a 'moving target' (Hooghe and Marks 2008: 123). In this chapter, we argue that major institutional changes and external shocks have altered party competition over European integration. These two forms of events have, first, affected the way in which European integration can be evaluated from the perspective of the parties' core ideologies and, second, provided opportunity structures for political entrepreneurs to mobilise supporters.
During its first decades, the European project was mainly an economic project of market liberalisation, harmonisation and deregulation. Accordingly, party positions on the European unification issue essentially reflected parties' stances on the economic conflict dimension between proponents of free market capitalism and supporters of workers' rights. Economic integration during the creation of the Single European Market enhanced the movement of capital between European countries and put pressure on labour costs. Therefore, political parties located towards the right of the socio-economic dimensionsuch as liberal, conservative, and Christian democratic partiesmainly favoured European integration. In contrast, parties on the economic leftsuch as socialist and socialdemocratic partiesoriginally opposed or only reluctantly supported the integration process, as it threatened social achievements on the national level by intensifying international economic competition (Featherstone 1988;Marks 2004;Marks and Wilson 2000;Ray 1999). 6 Over time, the character of European integration changed from a mainly economic cooperation to a genuine political project. Both the widening of the union and its constant deepening produced a unique multilevel system of governance with significant legislative powers in many policy domains, which makes the political influence of the EU much more tangible than that of the former European Communities (EC). These institutional changes had important implications for party competition over European integration in Western Europe, especially due to the simultaneous rise of the value-based cultural conflict dimension, i.e. the societal division between those who share progressive, libertarian, cosmopolitan values and those who uphold traditional, authoritarian, communitarian values. While the former group sees globalisation and Europeanisation in a more positive light, the latter group is much more averse to these internationalisation processes and, thus, reacted with resistance to the changing nature of European integration. Since many traditional mainstream parties were unable to adapt rapidly enough to the structural changes and the shift in their electorates' preferences, the number of political parties increased significantly (Hooghe and Marks 2018). Many newly emerging parties located themselves at the 'TAN' end of the cultural conflict and opposed the European integration project as core part of their ideological portfolio. As a result, these new Eurosceptic parties completed the 'inverted-U curve' in EU party systems that reflects the two main political conflict dimensions described above. While Euroscepticism of left-wing parties still originates in their opposition to ongoing market liberalisation and, thus, in their position on the economic conflict dimension, right-wing parties take issue with the threat that European integration poses to national sovereignty and, thus, derive their Euroscepticism from their stance on the cultural conflict dimension (see also Lachat 2018;van Elsas et al. 2016).
Taken together, we infer that party positions on the two basic political conflict dimensions are essential for understanding party competition over Europe. Over the course of the European unification process, however, the two main political conflicts in Western Europe have not always been related to party positions on European integration in the same way. We argue in the following sections that specific external events have altered these relationships at particular moments of the unification process.

A post-Maastricht readjustment hypothesis
Our first hypothesis rests on the assumption that party stances towards Europe became aligned to the new cultural conflict dimension in Western Europe towards the end of the 20 th century, although they used to be more strongly influenced by the economic conflict dimension at the beginning of the European integration process (Hooghe et al. 2002;Kriesi et al. 2006). As described above, the nature of political conflict over Europe changed, when the European integration process evolved from a mainly economic cooperation into a genuinely political unification project. Increasingly, the issues connected to European integration became more political than economic. Even more importantly, the deeper political integration promised some sort of regulated capitalism on the European level and thus a solution to the problems created by the liberated market forces, which made many centre-left parties readjust their stance on European integration (Marks andWilson 2000: 442-8, Marks 2004: 240-1). This readjustment should have occurred most visibly during the 1990s after the ratification of the Maastricht treaty, as it reformed the structure of the EC through the establishment of a political union with its famous three-pillar structure that represented the most obvious epitomisation of the EU's institutional transformation (van Elsas and van der Brug 2015). As Marks (2004) phrased it, 'Maastricht' took political integration a great step further, shifting attention 'from creating a market to regulating it' (Marks 2004: 258). 7 In response to the institutional changes and the increasing influence of EU policy-making, many citizens became more sceptical towards the integration process. This phenomenon has become known as the 'post-Maastricht blues' (Eichenberg and Dalton 2007), which implies that the former 'permissive consensus' of the citizenry with the European project (Lindberg and Scheingold 1970) became replaced by a 'constraining dissensus' (Hooghe and Marks 2009). Additionally, the relationship between ideological orientations of EU citizens and their attitudes towards European integration changed significantly. 'Maastricht made citizens aware of the implications of the EU for national interests, sovereignty, and identity, thereby giving right-wing citizens a reason to become Eurosceptical' (van Elsas and van der Brug 2015: 202). In this regard, not the least important feature of the treaty was the introduction of EU citizenship. It cemented the idea of a European identity and thus offered a catalyst for the difference between those who felt part of a European community and those who saw European and national identities as incompatible (Carey 2002). Moreover, the debates around the common currency and the creation of the European Central Bank represented further challenges to national sovereignty and eroded European identification among EU citizens (Eichenberg and Dalton 2007;Popa and Dumitrescu 2017;Scheuer and Schmitt 2009). Overall, the Maastricht treaty led to increasing worries that the newly consolidated EU constituted a threat to national sovereignty, thus bringing the cultural dimension for structuring position-taking around European integration to the forefront.
Until the mid-1990s, public opinion researchers mainly identified economic considerations as main determinants of citizens' EU preferences (Anderson and Kaltenthaler 1996;Eichenberg and Dalton 1993). But soon after, the focus in this research area switched and scholars discovered new central explanatory factors, such as national identity (Carey 2002), perceived cultural threats (McLaren 2002), and fear of immigration (de Vreese and Boomgaarden 2005). Evidently, this new set of determinants is rooted more strongly in the cultural dimension of political conflict than in the economic dimension. For existing as well as newly emerging political parties, this development in the demand side of political competition meant that they faced electoral incentives to strategically adjust their own positions to the preferences of potential voters.
Considering both the ideological roots and the strategic incentives of political parties, we hypothesise that party positions towards European integration in Western European party systems should have been associated more with cultural than with economic issues in the aftermath of the Maastricht treaty. Hence, our post-Maastricht readjustment hypothesis postulates a dual development that can be expressed as two separate hypotheses: H1a: In the aftermath of the Maastricht treaty, the relative impact of the economic conflict dimension on party positions towards European integration has decreased.
H1b: In the aftermath of the Maastricht treaty, the relative impact of the cultural dimension on party positions towards European integration has increased.

A post-'Euro crisis' readjustment hypothesis
Arguably, a second watershed period of the European integration process was the European economic and sovereign debt crisis between 2009 and 2014: a 'transformational moment in the history of the EU' (Otjes and van der Veer 2016: 243) with unprecedented consequences for the politicisation of the European integration issue. 'The Euro Crisis has produced new conflicts between member states, in particular between so-called creditor and debtor countries; and it has intensified distributional conflicts within the member states, most certainly in the southern European debtor countries that have been hit by the crisis most severely' (Grande and Hutter 2016: 16). Due to the austerity measures imposed on the countries in crisis and the financial bailouts by the creditor countries, it soon evolved into not only an economic but also a legitimacy crisis of the EU political system. The associated negative effects of EU policy-making on the welfare of its citizens led to a stark decline in public support for the EU, especially in the most severely affected member states (Braun and Tausendpfund 2014;Sch€ afer and Gross 2020).
In the face of such severe socio-economic and political circumstances, most national governments in the Eurozone were soon voted out of office (Bosco and Verney 2012). In addition to rising government instability, Eurosceptic and populist challenger parties on both sides of the ideological spectrum experienced great electoral gains all over Western Europe (Hern andez and Kriesi 2016;Hobolt and Tilley 2016). Several studies (Hobolt 2015;Hobolt and de Vries 2016;Nicoli 2017) show that the crisis significantly affected citizens' decision to vote for Eurosceptic parties, even in countries that were hit less by the crisis. However, country differences need to be considered when making inferences about political conflict and electoral competition over Europe (Kriesi and Pappas 2015;Magalhães 2015;Sch€ afer and Debus 2018).
In the South European debtor countries, particularly in Greece, the European integration issue was closely connected to economic considerations, especially owing to the strict austerity measures imposed by the socalled 'troika' (Hutter et al. 2018;Otjes and Katsanidou 2017). In these contexts, the political conflict over European integration became highly politicized (Charalambous et al. 2018). Many parties responded with growing Euroscepticism to the external economic interventions (Turnbull-Dugarte 2019) and radical left parties, which campaigned strongly against the austerity regime, experienced great electoral gains. Hobolt (2015: 7) underlines that the support for left-wing Eurosceptic parties 'was not driven by a rejection of the European project, but by discontent with austerity policies and a desire for more European solidarity.' Beaudonnet and Gomez (2017) support this conclusion by demonstrating that economic evaluations played a more important role than before for voters of radical left parties after the outbreak of the crisis. Such findings indicate that the Euro crisis has reinforced the explanatory power of the economic conflict dimension for party positions towards European integration in the South European countries.
However, the Euro crisis also made European integration a more salient issue in the North and West European states that experienced no economic intervention (Turnbull-Dugarte 2019). But in contrast to Southern Europe, the political conflicts in the creditor countries were less about economic issues. Here, the transfer of funds to other EU members triggered questions over national sovereignty and responsibility, resulting in great electoral success of radical right Eurosceptic parties (Hobolt 2015). 'In less hard-hit countries, populist radical right parties, as the ultimate defenders of national interest, were inclined to express opposition to the bailing out of financially troubled EU members and the transfer of sovereignty to the European level in response to the crisis' (Pirro and van Kessel 2017: 407). This is supported by the study of Bechtel et al. (2014), who find very limited effects of economic considerations on people's support for international redistribution. 'Instead, the bailout debate is better understood as a foreign policy issue that pits economic nationalist sentiments versus greater cosmopolitan affinity and other-regarding concerns' (Bechtel et al. 2014: 835). More generally, we see the response to the crisis in the creditor countries as part of a general 'cultural backlash' in Western societies (de Vries 2018; Norris and Inglehart 2019), which has been aggravated by the Euro crisis (Hooghe and Marks 2018). Therefore, we argue that the cultural dimension has gained importance in the West European creditor countries.
Taken together, our post-'Euro crisis' readjustment hypothesis is a combination of two separate hypotheses that account for the country differences during the Euro crisis.
H2a: After the outbreak of the Euro crisis, the relative impact of the economic conflict dimension on party positions towards European integration has increased in the Southern European countries most affected by the crisis.
H2b: After the outbreak of the Euro crisis, the relative impact of the cultural conflict dimension on party positions towards European integration has increased in the North-Western European countries less affected by the crisis.

Empirical analysis
We empirically test our theoretical expectations by conducting a multivariate multilevel regression analysis based on a content analysis of all election manifestos (Euromanifestos) published by political parties that competed in the eight European Parliament elections between 1979 and 2014 . In the following, we describe the unique value of election programmes for our purpose, the operationalisation of our main variables, and the methodological approach, before presenting the empirical results of our analysis and their interpretation.

Data
As the Euromanifestos dataset combines the advantages of longitudinal and cross-sectional data, it is ideally suited for analysing the relationship between party preferences across time and space. 8 More generally speaking, election programmes are a valuable source for studying parties' ideological profiles and their policy orientations, because they constitute central programmatic statements covering a wide range of issues. In most cases, manifestos are debated in depth and subsequently ratified at a party convention. Thus, they are most likely to represent the collective internal expression of policy preferences of a party. Although using manifesto data to measure policy positions is not without controversy (Dinas and Gemenis 2010), they are widely used in research on party politics (see, e.g., Ezrow et al. 2011), also when it comes to EP elections (see, e.g., Spoon and Williams 2017). We can therefore assume that the manifestos reflect the official positions a party takes on specific issues and that they are used by party elites as well as activists during the election campaign. 9 Additionally, their obvious advantages over other measurements include their objectivity and the possibility of clearly separating a party's political preferences from their actual behaviour. Put differently, manifesto-derived measures of party positions are intentional, while expert survey-derived positions are perceived as actual positions based on past behaviour. 10 The full dataset contains 942 election manifestos from all countries that were EC/EU members at the time of the respective EP election. Included in this dataset are parties that were represented in the European Parliament at least once and continued to be represented for two or more consecutive legislative periods. We restrict our empirical evidence to 638 election manifestos from the 12 countries that had already been a member of the EC before the Maastricht treaty. We do so, because, first, our interest lies in the long-term relationships between policy positions and their ideological determinants, and second, we aim to exclude the influence of new member state entries on the identified effects. To analyse the content of the manifesto, the election programmes are broken down into 'quasi-sentences', which are defined as political arguments and represent the coding units. Every single unit is assigned a content category (issue or policy area) and an evaluation of its connotation (positive or negative).

Operationalisation of variables
Regarding our dependent variable, we create an indicator measuring a party's position towards European integration ranging from 'totally in favour of' to 'totally against' European integration. In order to deductively construct this scale, we use the items of the pro-anti-EU scale provided by the Euromanifestos project . 11 Hence, we include all arguments in favour of and against the EC/EU in general (per_108), the competences of its institutionsthe European Parliament (per_306), the European Commission (per_308), the Council (per_310), European Court of Justice (per_312), the European Central Bank (per_3141), and others institutions (per_314) -, and the transfer of powers to the European level (per_3011). Moreover, we also include manifesto categories regarding central European policies, such as the single market (per_4084), structural cohesion funds (per_4011), further EU enlargement (per_316), an EU constitution (per_203), and the common currency (per_4086). Lastly, we also account for negative statements regarding the complexity of the EC/EU's political system (per_318a), national contributions to the financing of the EC/EU or its policies (per_1081b), and EU integration in general (per_6021b). 12 To construct the scale, we use Lowe et al.'s (2011) logit scaling technique, which 'combines the advantages of both additive and ratio scaling methods for manifesto data, whilst avoiding the problem of polarisation found in ratio scales, with the additional benefit of a diminishing impact of repeated emphasis, mirroring natural language' (Prosser 2016: 739). On the final scale, positive values represent pro-integration positions, and negative values represent anti-integration positions. 13 More specifically, the scale is calculated with the following formula, where Pro and Anti represent the total (raw) number of quasi-sentences from the manifesto categories listed above: 14 In order to construct our two main independent variables, we employ a strategy developed by Prosser (2014), which is laid out in the following. For selecting the indicators that best represent the two dimensions of political conflict, we first create a logarithmic salience measure for every manifesto category (h I ). The reason to transform the original (raw) measure into this logarithmic structure is 'the diminishing importance of subsequent mentions of a policy' (Prosser 2014: 95). Lowe et al. (2011) propose the following formula, where S is the number of quasi-sentences for each relevant category in a manifesto and N is the total number of quasi-sentences in a manifesto: In a second step, we correlate these logarithmic salience measures with exogenously provided starting scales for both dimensions, which is done to achieve a high degree of content validity. 15 More specifically, we use the non-zero cases 16 of all potentially relevant manifesto categories, i.e. all categories available for the time span 1979-2014 that relate to the ideological dimensions in question. 17 Third, when graphically inspecting the (absolute) correlations between our salience measures and the exogeneous scales, we can identify breaking points in the distributions (see Figure A-2 in the online appendix), which are then used as thresholds to construct new (exogeneous) scales. 18 This procedure is repeated 'until the scale reaches a stable equilibrium point where all components with an exogenous correlation above the threshold are included in the scale and all components with an exogenous correlation below the threshold are excluded' (Prosser 2014: 96). 19 The final manifesto categories that are included in our indicators by this procedure are displayed in Table 1.
Lastly, the identified manifesto categories are used to construct the two final ideological scales by applying Lowe et al.'s (2011) logit scaling technique, as we already did for the dependent variable. Here, higher values represent more economically right positions (h Economic Dimension ) and culturally authoritarian positions (h Cultural Dimension ). Hence, in the following formulas, Right/Authoritarian and Left/Libertarian represent the total (raw) number of quasi-sentences in the manifesto components belonging to the 'right'/'authoritarian' and 'left'/'libertarian' sides of the respectiveeconomic or culturalscale 20 : Apart from integrating our two main independent variables into a regression analysis, we can be parsimonious with other independent party-level variables. This is because our aim is not to disentangle the effects of different ideological and strategic factors for party behaviour, but to provide a general description of the relationship between basic political conflicts and positions towards the issue of European integration over time and across countries. By doing so, we build on previous research with the same objective (Hooghe et al. 2002;Marks et al. 2006;Prosser 2016). Still, we control for two basic party features of political parties that seem to fundamentally change the way that parties position themselves towards European integration (Sitter 2001;Taggart and Szczerbiak 2013). This is, first, the size of a partymeasured by its vote share in the last national parliamentary electionand, second, a party's government participation, which we operationalise by including a dummy that is 1 when the party is member of the national government in the year of the election, and 0 otherwise.  (per_7053a) Note: If the connotation is not clearly stated, the component represents 'positive' mentions; the code in brackets refers to the variable in the dataset; 'UMG' is an abbreviation for 'underprivileged minority groups'.

Results
In order to test our hypotheses, we estimate two linear multilevel regression models, in which observations are nested in country-years (i.e. specific elections in each country). We include dummy variables for the two 'critical junctures' under investigation by constructing a factor variable with three categories: pre-Maastricht (1979( -89), post-Maastricht (1994) and post-'Euro crisis ' (2014). We further differentiate between the effects in the 'EU-9' (i.e. the nine oldest EU member states in Western Europe) and in 'PGS' (i.e. the three Southern European member states Portugal, Greece, and Spain) by including dummy variables for each country and estimating three-way interaction effects for one conflict dimension per model. This serves to identify whether effects found for the different time periods and country groups reach standard levels of statistical significance. 21 The results of the two models are displayed in Table 2.
As interaction effects are often not easily interpretable, 22 we plot the marginal effects of the two dimensions across the three periods for both Note: Displayed are REML estimates of multilevel linear regression models with standard errors in parenthesis; significance levels: Ã p < 0.1, ÃÃ p < 0.05, ÃÃÃ p < 0.01. The observations are parties nested in country-years, i.e. specific EP elections in the different EU member states; the coefficient labelled 'dimension" represents the economic dimension (model 1) and the cultural dimension (model 2) in the respective models; both conflict dimension variables are group mean-centred within country-years (i.e. election within countries); the baseline categories for the factor variables are 'post-Maastricht (1994-2009)" and 'EU-9". country groups -EU-9 and PGSin Figures 1 and 2. We note that economically right-wing preferences (i.e. pro-market liberalism) were related to pro-integration party positions in the EU-9 during the pre-Maastricht period. However, this positive and statically significant effect fades away afterwards, as it is close to zero after the 1994 EP elections. Importantly, the drop in the effect of the economic dimension between the pre-and the post-Maastricht period is statistically significant (see both Table 2 and Figure 1, right-hand side). 23 This result suggests that the role of the economic dimension in structuring positions on European integration has substantially decreased once the Maastricht treaty cemented the future direction of the EU by 'solving' the initial conflicts around the economic goals of the Union. These findings support our first hypothesis about the declining role of economic issues in parties' positioning on European issues (H1a). At the same time, conflicts around the cultural dimension played a more important role in structuring party position on European integration once the EU transformed from an economic community to a political union. This is shown in the left-hand panel of Figure 2, in which the cultural scale exerts a more substantial impact during the post-Maastricht years, especially for the older EU member states. While the marginal effect of the cultural dimension is barely different from zero before the ratification of the Maastricht treaty, it is strongly negative and highly significant in the two later periods for the EU-9 countries. After Maastricht, preferences related to traditional, authoritarian and nationalist values are connected to anti-integration positions of political parties. However, since the confidence intervals overlap between the pre-and post-Maastricht period, our second hypothesis (H1b) is only partially confirmed. Although the effect size of the cultural dimension also increased after the Euro crisis, this is not statistically different from the pre-crisis (post-Maastricht) period (1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009). Hence, we do not find conclusive empirical evidence for an increased importance of the cultural dimension after the Euro crisis in the Western European creditor countries and, therefore, reject our third hypothesis (H2a).
The marginal effects presented on the right-hand side of Figure 1 show that the economic dimension has always played a substantial role in party competition over Europe in the three countries that joined the EC in the 1980s. In contrast, this is much less the case for the cultural conflict dimension, as can be observed on the right-hand side of Figure 2. However, we do not find any significant differences in the effects of the economic dimension between the pre-crisis and post-crisis periods. In the face of this lacking empirical evidence for any readjustment effects after the Euro crisis in Southern Europe, we also reject our fourth and final hypothesis (H2b). 24 We further note that both Figures 1 and 2 reveal substantial differences between the Western and Southern European member states. While the position-taking of parties towards European integration was exclusively shaped by cultural issues after Maastricht in the EU-9 countries, the opposite pattern can be observed in Southern Europe. Here, it is the economic dimension that shaped parties' positions towards the issue of European integration, while the cultural dimension never played a substantial and statistically significant role.

Conclusion
We analysed the relationship between party positions towards European integration and the two basic conflict dimensions that ultimately define the political space in Europe's modern history: the socio-economic leftright dimension and the socio-cultural libertarian-authoritarian dimension (Kriesi et al. 2012). In a first systematic and comprehensive empirical analysis drawing on the integrated Euromanifestos database , we investigated whether the associations between these main conflict dimensions and parties' EU positions have readjusted during two 'critical junctures' of the integration processthe Maastricht treaty and the Euro crisis. Moreover, we asked whether these readjustments followed different pathways between the nine oldest Western European member states (EU-9) and the three Southern European countries that joined the Union in the 1980s. Our results for the EU-9 imply that the two dimensions were associated to European issues in a similar way prior to the Maastricht treaty, but this changed after the mid-1990s. While the cultural dimension has become the main driver of parties' EU positions ever since, the economic dimension barely plays any role for political conflict over Europe anymore. In stark contrast, economic concerns always dominated cultural issues in the three Southern European countries and even became increasingly relevant over time. Much to our surprise, the Euro crisis that hit the EU member states after 2009 has not exerted similar readjustment effectsneither in the North-Western creditor countries nor in the Southern countries most affected by the crisis.
Against this background, the findings of our study have several important implications for scholars of European integration. First, the two most relevant political conflicts of modern societies shape party competition over European integration in highly context-specific ways. Across time and space, the structuring power of the economic and the cultural conflict dimensions have varied significantly and continue to do so. Second, the watershed moment of the Maastricht treaty truly represents a 'critical juncture' for the relationship between political conflict dimensions and party positions towards European integration in Western Europe. Since Maastricht, the cultural dimension has replaced the economic dimension in structuring party competition over Europe in the older EU member states. Third, although the European economic and sovereign debt crisis is considered a 'transformational moment in the history of the EU' (Otjes and van der Veer 2016: 243) by some, party competition over European integration has not been dramatically restructured when it comes to underlying political conflicts. We thus conclude that the 'external shock' of the Euro crisis has not fundamentally changed the nature of the integration process in the same way as the Maastricht treaty has. It therefore cannot be regarded as a 'critical juncture' in the relationship between underlying political conflicts and party competition over Europe.
Last but not least, we would like to reiterate that the goal of our analysis was to provide a first systematic and comprehensive description of the basic political conflict dimension and positions towards the issue of European integration across different electoral contexts. We neither aimed to disentangle ideological from strategic factors of party behaviour nor did we investigate positional changes of individual parties, as those shifts are just one potential source of a changing relationship between dimensions of conflict and issue positions (others would be, for example, the emergence of new parties and the disappearance of others). However, explaining the changes we observed in our analysis seems like a promising future route for scholars of party politics and European integration.

Notes
1. Yet, the nature of both phenomenathe changing structure of political competition and the politicisation of European integrationvaries across time and space (Enyedi and Deegan-Krause 2010;Hutter et al. 2016). 2. We understand a critical juncture as a 'major event or confluence of factors [which disrupts] the existing balance of political and economic power' (Acemoglu and Robinson 2012: 106). Although we consider those watershed moments as 'critical' for the relationship between conflict dimensions and party competition, we leave a more in-depth discussion of the 'critical junctures' concept to the historical institutionalism literature (see, e.g., Capoccia 2016). 3. It should be noted, however, that cultural cleavages did exist before, such as the religious cleavage (Lipset and Rokkan 1967 Bartolini (2005). To avoid misunderstandings, we therefore stick to the term 'dimensions of political conflict'. 6. The German Social Democrats, for example, originally opposed the Schumann Plan, denouncing it as 'a conservative-catholic-capitalist project' (Egle 2011). The French Socialists remained critical towards the integration process until the presidency of Francois Mitterrand. Even more critical, the British Labour Party opposed the EEC membership in the 1960s and, after the UK's accession, wanted to withdraw from it until the mid-1980s (Almeida 2012). However, all three parties revisited their positions towards 'Europe' later on, partly because of their own general programmatic changes and partly due to the changing nature of the European integration process. 7. This paradigm shift is closely related to the shift from 'negative integration' to 'positive integration', in which the European integration process is pursued (Scharpf 1999). The former form refers to the elimination of custom and trade barriers, whereas the latter is concerned with the exercise of economic policy and regulatory powers at the EU level. While negative European integration has been pursued mainly by neo(liberal) forces, positive integration is more favourably viewed by actors who aim to intervene in the capitalist economy (see also Scharpf 2008). 8. Although Braun and Schmitt (2018) show that EU party positions do not substantially vary between elections at different governmental levels, we believe that they are most clearly expressed in manifestos relating to elections on the EU level. 9. It is certainly the case that few voters actually read manifestos. However, their content is widely spread via the campaign communication of elites and activists of a party as well as by mass media. 10. Even though parties' own programmatic positions are not always correctly perceived by experts or voters, our manifesto measures (which are derived in the following) correlate significantly with positional measures that are derived from surveys. For example, the correlation between our dependent variable and the perceived party position towards EU integration taken from the European Election Studies (EES) 2014 voter survey (Schmitt et al. 2016) is 0.62 (statistically significant at p < 0.001) for the parties of our analysis. 11. However, we exclude the manifesto category 'National way of life' from the indicator, as it is included in one of the independent variables. 12. In order to empirically confirm our deductive reasoning, we performed a confirmatory factor analysis (see Table A-1), which shows that most categories (11/16) load on a common first factor. An index that exclusively relies on these categories correlates highly with our own indicator (r ¼ 0.96). However, since we agree with certain reservations against using factor analyses for manifesto data (Prosser 2014) and since positions towards European integration are multidimensional (Boomgaarden et al. 2011), we rely on our (theoretically deduced) index. To our benefit, the results of the analysis are robust to other operationalisations of the dependent variable, such as an index constructed on the confirmatory factor analysis (see Table  A-12, Figure A-17 and Figure A-18) or a measure only using the most general category per_108 (see Table A-11, Figure A-15 and Figure A-16). 13. This EU integration scale approximates a normal distribution and is shown in Figure A-1 (see online appendix). 14. For the period between 2004 and 2014, the correlation between this scale and the EU integration scale computed by the Chapel Hill Expert Survey (CHES) is 0.61 (significant at p < 0.01). 15. We tested three different exogeneous starting scales, namely the ones by Prosser (2014), Benoit and Laver (2007) and Bakker and Hobolt (2013).
Eventually, all three scales resulted in the same final indicators for both variables. See the online appendix (Table A-3) for a detailed summary of the exogeneous starting scale provided by Prosser (2014). 16. Using only the non-zero cases of the manifesto categories avoids the problem that many categories of the manifesto coding scheme are not used at all. This 'structural zero problem' is an important reason why the manifesto components do not co-vary in a meaningful way, which also bears important implications for the usefulness of factor analyses when trying to extract ideological scales from manifestos (Prosser 2014: 94-96). 17. For an overview of all manifesto categories tested for the scale construction see online appendix (Table A-4). 18. This means that we only use the manifesto components that correlate most strongly with the exogenous scale. 19. This step could be repeated until the two thresholds do not have to be adjusted anymore, but for our scales, this is already the case after the second iteration ( Figures A-3 and A-4). 20. For the period 2004-14, the correlations between our scales and the ones computed by the CHES are 0.68 (significant at p < 0.01) for the economic dimension and 0.57 (significant at p < 0.01) for the cultural dimension. 21. Estimating a linear regression on the dataset (pooled across countries and elections), in which we interact the year of each election with our two dimensions (see Table A-7, Figures A-7 and A-8), reveals the same pattern of results. 22. However, it shows that both control variables are statistically significant, which implies that larger parties and governing parties have, on average, been more positive towards European integration across all electoral contexts. 23. Although the pre-and post-Maastricht confidence intervals slightly overlap (for the EU-9) in Figure 1, we performed a t-test, which shows that this difference is statically significant (t ¼ 2.456, p < 0.05; pre-Maastricht effect size ¼ 0.285, se ¼ 0.089; post-Maastricht effect size ¼ 0.034, se ¼ 0.050). 24. The lack of change in the post-crisis period compared to the pre-crisis period in both the EU-9 and PGS is confirmed by replicating the analysis using the 1999-2014 CHES trend file (Table A-8 and Figures A-9 and A-10).

Funding
This research was supported by the EUENGAGE project funded by the Horizon 2020 Research and Innovation Programme (grant agreement number 649281-EUENGAGE-H2020-EURO-2014-2015/H2020-EURO-SOCIETY-2014). Moreover, collection and coding of Euromanifestos was supported over the years by additional research funding institutions, among them the German Research Foundation, the Thyssen Foundation, the European Union's 7th Framework Programme, and the Volkswagen Foundation.

Notes on contributors
Constantin Sch€ afer is a postdoctoral researcher in the H2020 project 'RECONNECT Europe' and research associate at the University of M€ unster (WWU