Zooming in on the ‘Europeanisation’ of national politics: A comparative analysis of seven EU countries

This article empirically revisits and tests the effect of individual distance from parties on the EU integration dimension and on the left–right dimension for vote choice in both national and European elections. This analysis is based on the unique European Election Study (EES) 2014 survey panel data from seven EU countries. Our findings show that in most countries the effect of individual distance on the EU integration dimension is positive and significant for both European and national elections. Yet the effect of this dimension is not uniform across all seven countries, revealing two scenarios: one in which it is only relevant for Eurosceptic voters and the other in which it is significant for voters of most parties in the system. The first is mainly related to the presence of a ‘hard’ Eurosceptic party in the party supply, but the second, which indicates a more advanced level of Europeanisation of party systems, is not explained by most current theoretical and empirical contributions. We conclude by proposing two additional explanations for this latter scenario in which the EU integration dimension is present for most voters in both type of elections, including those voting for the main parties. Our findings and further discussion have implications for the understanding of the Europeanisation of national politics and its relationship with vote choice.


INTRODUCTION
What is the effect of individual positioning on the EU integration dimension of vote choice? Does this dimension shape vote choice both in the EU and national elections, or are its effects only present for the supranational elections? Scholars have examined whether the EU is a salient dimension in individual vote choice ever since the first elections for the European Parliament (EP). Yet, and despite a significant number of empirical contributions, the question remains not fully and satisfactory answered.
The lack of a definitive answer is partly explained by the changing nature of the European project, which forces scholars to continue revisiting their theoretical and empirical expectations. Thus, the study of the effect of the EU issues on vote choice has been frequently framed using the second-order elections (SOE) model. This framing posits that European elections have been less relevant to the electorate because, together with other factors, the issue at stake, Europe, does not matter to voters (Schmitt and Toygür 2016). Yet this understanding has traditionally coexisted with several studies showing that, under some circumstances, European issues matter (Ferrara and Weishaupt 2004;Reif 1984;van der Eijk 1996). However, no overall conclusion was reached and many scholars still concluded that the EU dimension did not matter or was largely irrelevant (Hix and Marsh 2007).
Over the last few years, this debate has once again gained momentum by the contributions of an important and growing literature on the politicisation of Europe. More concretely these contributions have started to show that European issues are increasingly present in national public opinion, gaining space in people's discussions and interests, and structuring national political competition (Spanje and Vreese 2011;Wilde and Lord, 2016;Ares et al., 2016;Hooghe and Marks 2018). There are contributions pointing at that direction such the one by Hobolt et al. (2008) showing that voters defect from governing parties because the government is generally far more pro-European than they are. Similarly, Hobolt and Wittrock (2011) concluded that while voters base their EP vote choices primarily on domestic preferences, those having additional information about the European integration dimension are also more likely to vote on this basis. As Hernández and Kriesi (2015) more recently pointed out the so-called 'Europeanisation of National Politics' is gaining traction among specialists.
However, despite all these significant contributions, the empirical evidence is still inconclusive, especially when it comes to comparing the micro-level explanations of party support behind this process.There have certainly been several empirical attempts to show how 'Europe matters' in national elections (Gabel 2002;de Vries 2007), but the literature has yet to fully explore the effect of individual positioning on the EU dimension visà-vis that of the traditional left-right dimension which traditionally has been driving the competition at the national arena. In other words, there is still missing a conclusive and comprehensive cross-national study comparing the effects at the individual level of both dimensions of party competition for both types of elections (for an exception see van der Eijk and Franklin 2004).
This article aims at extending previous literature by revisiting the (relative) effect of individual positioning on the EU integration dimension of vote choice across different European countries using a unique panel dataset, the European Election Study (EES) 2014 survey panel that includes two waves in seven then-EU member states 1 : Germany, Italy, Spain, Poland, Sweden, the United Kingdom and Greece. One of the waves took place as part of the 2014 European post-electoral survey and the other as part of a national post-electoral survey of the same seven countries. This dataset allows, for these two types of elections, an assessment of the significance and magnitude of the EU integration dimension vis-à-vis the left-right dimension, the other dimension against which the EU dimension is usually compared for both elections and for the same set of respondents. Crucially, we also extend this analysis to examining whether the EU dimension equally matters for the electoral support of all parties in the party system.
Our empirical analysis proceeds in two steps. In the first part, we analyse whether the European dimension drives people's vote in all seven EU countries. If the process of Europeanisation referred to by the literature is taking place, we should observe that individual positioning on the European dimension shapes people's vote choice, with its effects being similar to that of the leftright dimension. Our results show that individual positioning on the dimension has an effect on vote choice in all seven countries under analysis and it shapes people's political behaviour not only in European elections but also in national elections to a quite similar degree. Yet the effect of the European dimension is still much smaller than that of the left-right dimension.
In the second part, we look more closely at the variation in the effect of individual positioning on the EU dimension, both across countries and between parties. On the one hand, we suggest that the EU integration dimension will likely be more salient to voters in party systems that have 'hard Eurosceptic parties' Marsh 2007, 2011;Hobolt and de Vries 2016). Therefore, the EU integration dimension becomes salient, but only to voters who support these political formations. For the remaining voters, the left-right dimension, strongly related to national issues, is still the only one that matters. However, in other countries, the EU integration dimension seems to be relevant also for the individual 1 For the first time, the EES 2014 includes an online panel component which consists of a number of online panel surveys that are administered in eight EU member countries for national and European elections. This dataset does not allow us to extend our approach to all EU countries. Yet the seven countries selected for the analysis (Austria is excluded due to the lack of some relevant variables for the analysis) vary both in the presence of Eurosceptic parties and the type of party system. For more information about the role of the EES 2014 project, go to: http://europeanelectionstudies.net/european-election-studies/ees- 2014-study/panel-study-2014. vote choice of most parties and for both types of elections, suggesting a more advanced stage in the Europeanisation of national party systems. We suggest here that current explanations about the party supply are not able to explain that more advanced stage we observe in countries such as Germany or Greece, proposing to focus future research on explaining the contextual factors that might contribute to the creation of this more advanced scenario of Europeanisation of national party systems. More concretely, we suggest two alternative explanations to be considered in further research on this topic. These two contextual explanations are related to potential perceptions that the national vote in both elections might have an impact on the country's fade in the European Union.

THE EUROPEANISATION OF NATIONAL POLITICS
Recent studies have shown an increase in the politicisation of EU governance in national arenas (Brouard et al. 2012;Ares et al. 2016;Hooghe and Marks 2018). As has recently been claimed, this change is empirically observed in three factors: the growing salience of European governance, opinion polarisation on EU issues and an increase in the number of actors and audiences engaged in monitoring EU affairs (van der Eijk and Franklin 2004). It has been suggested that this process may ultimately be changing the role of EU issues in structuring public opinion on EU affairs and voting mechanisms in EU elections (Spanje and Vreese 2011; Hobolt and Wittrock 2011;Ares et al. 2016;Hobolt and de Vries 2016).
Because of the increasing preponderance of EU issues in the political debate, the 'Europe Matters' school is once again trying to understand its implications for vote choice or political attitudes towards the EU (van der Eijk et al. 2006;Koepke and Ringe 2006;de Vries and Tillman 2011). Since this process might still be in its inception, recent research suggests that national issues still have more weight in citizen voting processes and the European elections remain of second-order significance, a characteristic that was already apparent in 2014 (Schmitt and Toygür 2016).
Despite the prolific number of recent works on the topic, this literature does not fully examine the effect of the individual positioning on the EU integration dimension, vis-à-vis that of the left-right dimension, of vote choice for the most recent wave of European elections and the most remarkable attempt to do so has involved two cross-sectional studies (de Vries 2007;van der Eijk and Franklin 2004). This empirical gap is even more remarkable if we consider that since the 2008 financial crisis and the refugee crisis that started in early 2014, EU institutions have increasingly assumed, or been requested in other cases to play, a prominent role in political decisions, triggering an intense debate about the extent and limits of EU integration. Indeed, several recent studies have shown that the 2008 economic crisis had important effects on people's vote choice (Hernández and Kriesi 2015). In recent years, thus, the politicisation of the EU might have reached remarkable levels, potentially leading to an increase in the effect of EU issues on vote choice across the board.
Nonetheless, this process might depend on the differential degree of EU politicisation across elections. The current system of EP elections inevitably links national issues, political parties and EU issues (Clark and Rohrschneider 2009, 660). For instance, De Vries (2007) shows, after comparing the UK, Denmark, the Netherlands and Germany that EU issue voting is more likely to occur in elections in which both the extent of partisan conflict over European integration and the degree of EU issue salience among voters are high.
This Europeanisation of vote choice is not confined to European elections but is also affecting national elections (de Vries 2007), with the European debate increasingly present in national election campaigns (Kriesi et al. 2006, Grande andHutter 2016). Several studies have shown an increase in national parliamentary questions about the EU (Senninger 2016).
Following the previous theoretical discussion, our first expectation is the following: H1: Following the 'Europe matters school', if the European integration dimension matters for voting choice in European elections, it should matter equally in national elections.

THE HETEROGENEOUS IMPACT OF THE EUROPEANISATION OF NATIONAL POLITICS
Following the previous discussion, it is consequentially convenient to examine whether individual positioning on the EU integration dimension shapes people's vote choice, to compare the magnitude of the effect with that of the left-right dimension and examine whether the EU integration dimension matters for both EU and national elections. After this starting point, however, we also delve into the differential effect of the EU integration dimension within and across countries. As we review below, there are theoretical reasons to expect the effect of this dimension to matter when voting for some parties and not others and to differ across countries.
More concretely and following previous studies, we test the main hypotheses that might account for this heterogeneity: whether the differential impact of the EU integration dimension of vote choice in a party system is related to the presence of a significant 'hard Eurosceptic party'. The factor of the party supply explanation might increase the saliency of this issue among voters, media and public opinion in general.
As shown by previous studies, the consensus on the EU integration dimension at the party level is essentially broken due to the emergence of small or ideologically extreme left-wing and right-wing Eurosceptic parties, which offer very differentiated policy choices regarding the country's permanence in the EU (Anderson 1998;Vasilopoulou 2011;Hobolt and de Vries 2016) and which position themselves far from the consensual status quo (remaining in the EU). These are the so-called 'hard Eurosceptic' parties (Taggart and Szczerbiak 2004). Following the logic of the spatial model literature, the party differential should increase the effect of this dimension of vote choice. The empirical implication is that, once these hard Eurosceptic parties are present, the EU integration dimension becomes relevant and salient. This effect might likely be larger for voters of Eurosceptic parties than for the rest, 2 but its salience might also affect the other parties in the party system, given the priming effect it might have in the media and general public opinion. Conversely, we should observe that, in contexts in which these parties are not present, the effect of the EU integration dimension should be negligible for all parties in the system.
To sum up, we should observe the following: H2: If a 'hard Eurosceptic party' is in the party supply, the EU integration dimension matters for both national and European elections for all voters.

DATA AND METHODS
Our empirical analysis is based on the individuallevel panel survey data that formed part of the 2014 EES (see footnote 1). These panels included two post-electoral waves, covering one national election and the 2014 European election. These data provide a perfect tool for analysing the effects of the EU integration dimension on both types of elections in a more reliable and valid way. In addition, the order of elections across the seven countries varies: while in some, the observed European election takes place after the national one (Italy, Sweden and Germany), the opposite happens in the others. This allows us to rule out potential patterns of spill-over effects, especially to check whether the EU integration dimension matters more when the EU elections are close to the national ones.
The outcome of interest is vote choice for the 2014 European post-electoral wave and vote choice for the national post-electoral wave. We exclude regional parties and parties with a very low percentage of votes at the national level.
The main explanatory variables are individuals' ideological distances based on respondents' self-reported positions and reported party positions on the ideological scale (0, 'extreme left-wing', to 10, 'extreme rightwing') and the EU integration dimension (0, 'Unification should go further', to 10, 'Unification has already gone too far') 3 (for the distribution of these variables among the seven countries, see Figures A1-A6 in the online Appendix).
We follow the literature and conceptualise a voter's utility as the distance between the party's policy position and the respondent's self-placement on the same scale. This means that voters derive a larger utility as they get closer to a party's policy position (Downs 1957).
Since neither the ideological distance nor the distance along the EU integration dimension is the same for each party alternative, we employ a conditional logit model. This is considered the correct procedure for estimating discrete vote choice in multi-party systems (van der Eijk et al. 2006). We estimate the effects of alternative-specific variables (i.e. distances between a voter and each candidate) separately, which in the next section will allow us to compare the effect of each alternativespecific coefficient on vote choice. In statistical terms, this is important as voters are likely to offer different (perceived) ideological positions for different alternatives (parties in our case).
Our statistical specifications also take other confounders into account. First, we control for individual-specifi c characteristics, such as gender and age. Second, we control for party identifi cation as the most relevant non-spatial factor (Th urner and Eymann 2000). Table A1 in the online Appendix provides a summary of the main explanatory variables and the other controls included in the models for each of the seven countries.
Finally, in the second part of the study, we are also interested in calculating the eff ect of individual distance on each of the two dimensions (the left -right and the EU) of vote choice across all parties in each of the seven countries. To estimate these effects, we follow Greene (2012) and Mauerer et al. (2015) by splitting the coeffi cients into as many alternative-specifi c coeffi cients as there are parties in the political system. Th is model allows us to test whether voters' distance from each party on both dimensions varies across parties in a conditional model framework. In other words, this model allows to capture whether the eff ect of perceived distance towards one party may be diff erent from another party for each respondent.

DOES EUROPE MATTER FOR VOTE CHOICE?
Following our initial hypothesis, we test whether the EU integration dimension drives an individual's vote choice and, if this is the case, we examine whether its eff ects are as present in EU elections as in national elections. In terms of the magnitude, recall that the baseline expectation, as posited by the SOE model, is that the eff ect of the EU integration dimension should be lower than that of the left -right dimension for vote choice, which remains the most relevant dimension. To test this, we ran conditional logits and plotted the (standardised) eff ect of each dimension for each country. Th e right panel on Figure 1 shows the average eff ect of respondents' distance from each party on the EU dimension and the left -right dimension of vote choice in the context of the 2014 European election. Th e panel on the left shows the eff ects for the same variables in each country's national election. Th is fi gure also indicates those countries where an EP election took place aft er the country's national election (Germany, Sweden and Italy) or otherwise (the rest). Th is is important as one might argue that the eff ect of the EU integration dimension on vote choice is only felt in the national arena when the EU election takes place before the national one. If the order is the opposite, the national debates might cloud the European election even more.
As expected, Figure 1 shows that the distance based on the left-right scale has a significant and positive eff ect on vote choice, both in the EU and for each coun-try's national election. Across all contexts and regardless of whether the EP election took place before or aft er the national one, its eff ect is larger than that of the EU integration dimension. Notwithstanding this pattern, the eff ect of this dimension is not negligible. As Figure  1 shows, the eff ect of the EU integration dimension on vote choice in EU elections is statistically signifi cant in all countries except Spain (and marginally in Poland). Most crucially, this dimension not only shapes vote choice in the 2014 EP elections but is also an almost equally important factor in national elections. In other words, regardless of whether the national elections took place before or aft er the EP elections, the EU dimension was similarly integrated into voters' decision-making logic when casting a ballot both in the European and in national elections. Th ese results mostly confi rm H1, although we need to look at these results for individual parties in more detail.
To strengthen our fi ndings, we take advantage of the panel structure of the data and replicate the analysis by using respondents' self-positions on the EU and the left -right scale at t-1. Th is design ameliorates concerns about the endogenous relationship between people's and parties' policy position on a given dimension and allows us to further corroborate that the EU integration dimension matters, even when we isolate context-specific debates that might increase the salience of EU issues temporarily. Th e results show the exact same pattern: Th us, even when we isolate context-specifi c debates by measuring our key explanatory factors at t-1, the eff ect of the EU integration dimension on vote choice remains. Most importantly, it remains positive and significant regardless of whether the EU election took place before or after the national election (see results in Figure A7 in the Appendix).

THE EFFECT OF EUROSCEPTIC PARTIES
The previous section shows the EU integration dimension matters for vote choice in the majority of the countries included in the analysis which allows accepting Hypothesis 1. In other words, the EU integration dimension matters for vote choice in the majority of the countries included in the analysis. The effect size is smaller than the left-right dimension, but it turns out to be important in both national and European elections.
Next, we unpack the effect of the EU dimension by analysing the heterogeneous impact it might have for the different parties in the party system and across the seven countries for which we have data. As we discuss in the theoretical part, the expectation is that the EU integration dimension will be relevant in contexts in which the party supply incorporates a 'hard Eurosceptic party'.
Before we enter into the next set of results, Table 1 shows the descriptive statistics of the EU party position of the different national parties included in our analysis. As can be seen, both the presence and strength of Eurosceptic parties varies within each of the seven EU countries showing the presence of soft and hard Eurosceptics in each of the seven countries under study.
For all seven countries, we plotted the coefficients of the ideological and EU dimensions on vote choice for each of the political formations for both the EP and national elections (for the coefficients and intervals represented in all the figures, see Tables A2-A7 in the online Appendix; the complete models with all the vari- A positive coeffi cient indicates that the spatial proximity to a particular party along the ideological dimension signifi cantly increases the probability of voting for that particular party.

Th e eff ect of the absence of a signifi cant 'hard' Eurosceptic party
In Spain, we have focused on the fi ve parties that competed in the 2014 EU elections: the national incumbent, the Popular Party (PP); the main opposition party, the Socialists (PSOE); the traditional left -wing political formation, United Left (IU); and two emerging platforms, Ciudadanos (Citizens-Cs) and Podemos (Yes We Can). Th is party system does not include any 'hard' Eurosceptic political formation. Podemos was clearly, at least at that time, a 'soft ' Eurosceptic party with a score of 4.4 (see Table 1). Figure 2 plots the coeffi cients extracted from the conditional logit model for this country. Th e results show that none of the party-varying coeffi cients for the EU integration dimension are signifi cant for either of the two elections, while the ideological distance coeffi cients are signifi cant for all parties and for both elections.
We do not have data on more countries in our sample that are similarly characterised by the total absence of hard Eurosceptic parties, but according to Freire and Santana-Pereira (2015), this also seems to be the case for Portugal, where the EU integration dimension also seems to have a weak impact on vote choice.

Th e heterogeneous eff ects of the presence of a signifi cant 'hard' Eurosceptic party
In this sub-section, we will discuss and present the results of the eff ects of the presence of some signifi cant 'hard' Eurosceptic parties in the party supply. As we will discuss, the eff ect of the EU integration dimension on voters' choice of parties is not homogenous for all party systems (countries). We will distinguish between two types of scenarios: a) the EU integration dimension is only relevant for the support given to hard Eurosceptic parties; b) the EU integration dimension is relevant also 4 Additionally, Figure A8 in the online Appendix displays the results of the same analysis but using the Chapel Hill expert survey to establish party location to compute the proximity scales on both the left -right and the EU scale. Th is measure is less likely to be aff ected by endogeneity. As we can observe, results are exactly the same.
for most political parties in the system including mainstream parties. a) Th e EU integration dimension is only relevant for the support given to hard Eurosceptic parties Figure 3 contain the results of the model for two cases that represent this scenario. Th e Italian case is the fi rst one. We replicate the same analysis for Italy for both types of election. In this case, we take four parties into account: the Democratic Party (PD), Forza Italia (FI), Movimento 5 Stelle (M5S) and the Lega Nord (LN), although FI ran as part of a coalition for the national elections. 5 Th e party system, in this case, includes two clear 'hard' Eurosceptic political formations: the LN and, most notably, the M5S with scores of 1.1 and 1.4, respectively (see Table 1). Th e confi guration of the party supply is therefore likely to increase the salience of the EU integration dimension. As shown in Figure 2, neither of the coeffi cients for PD shows statistically signifi cant eff ects for either election. Only FI seems to be signifi cant for the 2013 national elections, but with a much reduced coeffi cient (0.019) and a p value of 0.037, which seems to be remarkable given that the campaign in this particular election took place in the middle of the debate on a potential fi nancial rescue of Italy.
In contrast, the distance between the self-reported position and the party policy position along the EU dimension is signifi cant for the two 'hard' Euroscep-5 Forza Italia ran with another platform, Il Popolo della Libertà (PdL) -People of Freedom, in the preceding national elections. tic parties, LN and M5S, for both types of election. Th e coeffi cients for the eff ect of the ideological distance are positive and signifi cant for the four parties and tend to be larger for voters opting for FI and PD. All in all, the analysis for Italy confi rms that the ideological dimension still has more weight, although Europe also matters-but only for those voters opting for 'hard' Eurosceptic parties (Giannetti et al. 2017).
Poland in 2014 is characterised by the presence of a signifi cant 'hard' Eurosceptic party. Results from estimating the same model are similar to those of Italy, as Figure 3 also illustrates. Th e eff ect of the individual distance to parties in the EU integration dimension is signifi cant for voters for the 'hard' Eurosceptic Congress of the New Right (score 1.1 in Table 1), but it is not signifi cant for the remaining parties including 'soft ' Eurosceptics (at least at that time) Law and Justice (PiS) and United Poland. Th e only exception is the then incumbent Civic Platform and only for EU elections (not for national ones). Th is can probably be explained by the intense confrontation that the issue of Poland-EU relations triggered between the incumbent party, a party that has a pro-European stance and other signifi cant parties adopting more critical views against the government during the run-up to the 2014 European election. b) EU integration dimension is relevant for the support given to most of the main parties We replicate the same analysis for another four countries: Germany, Greece, Sweden and the UK. All of them are countries with important 'hard' Eurosceptic parties, but unlike what we observe in the preceding cases, the EU integration dimension is signifi cant when it comes to explaining the support of most main parties in both types of elections.
Germany has one significant 'hard' Eurosceptic party, AfD (Alternative für Deutschland) with a score of 1.6 in Table 1. However, as we see for Germany in Figure 4, the coeffi cients for the CDU/CSU (Christian Democratic Union and Christian Social Union), SPD (Social Democratic Party), Die Linke (The Left), Die Grünen (the Greens) and, fi nally, AfD (Alternative für Deutschland), are signifi cant for both the ideological and the EU dimensions. Th e coeffi cients for both dimensions for nearly all the parties are signifi cant for both types of election. Th e exception is Die Linke, which displays a non-signifi cant coeffi cient for the EU integration dimension. In all the other more relevant parties in the German party system-CDU/CSU, SPD, Die Grünen and AfD-the EU dimension shaped vote choice for both types of election. Th e EU distance eff ect is therefore relevant, although once again the magnitude of the coeffi cients is smaller than the ideological distances of the coeffi cients.
Another quite similar case was that of the UK in 2014. Eurosceptic views were relevant in UK politics even before the 2008 economic crisis with at least a signifi cant 'hard Eurosceptic party' (UKIP score in Table  1 is 1.1). Results displayed in Figure 4 also confi rm that the eff ect of the EU integration dimension is positive and signifi cant for both the EU and national elections for all parties and not only for the UKIP. Th e eff ect is weaker than that of the traditional left -right dimension. Th e positive eff ect of the individual proximity scale on the EU integration dimension persists in the national election. Interestingly, however, only voters opting for regional parties give less salience to the EU integration dimension in the national election, probably due to the greater relative salience of other issues (Hobolt 2016).
In Figure 5, we display the results of the same models for Greece and Sweden. The Greek party system transitioned from the previously very stable two-party system dominated by the Socialist PASOK and the conservative New Democracy (New Democratia) to a more fragmented system with bipolar competition between the latter and the more left ist party SYRIZA. Th is party is a 'soft ' Eurosceptic party (3.4 on the CHSE EU scale) (Hobolt and de Vries 2016, 511) although its score is far from that of the pro-EU parties (with an average on the same scale of 2.9 -see Table 1). Th e Greek party system also witnessed the electoral consolidation of two signifi cant right-wing 'hard' Eurosceptic parties: Golden Dawn and ANEL (both with a score of 1.1 in Table 1). As we can observe again in Figure 5, and similar to the two preceding cases, the EU dimension became relevant across all the political formations in Greece (except for Olive Tree voters). Interestingly, even in the 2015 national election, when the debate over the EU was less salient and Greek politics was returning to a normal setting, the EU dimension remained signifi cant for most of the voters of the main parties.  In Sweden, also in Figure 5, the far-right Sweden Democrats constitute a clear 'hard' Eurosceptic party (1.3 score in Table 1), although the party has linked its attacks on the EU to the refugee crisis. However, as in Germany and the UK, results show that the EU dimension constitutes a significant factor not only for vote choice among far-right voters but also for vote choice among other political formations, both conservative (i.e. Moderate Coalition Party) and left-wing (Left Party), with significance in both EU and national elections. The most significant exception is the Social Democratic Labour Party and only for national elections. All in all, Sweden displays a scenario in which the anti-EU discourse of a Eurosceptic party spills over to voters of the other parties in the political system, making the EU dimension a significant factor for the overall vote choice in both types of elections.

THE UNEXPLAINED HETEROGENEOUS EFFECTS OF THE PARTY SUPPLY
The preceding section has shown how the presence of a 'hard' Eurosceptic party might produce two different scenarios in 2014. In the first scenario, the EU integration dimension only seems to influence the party preference for this type of party for both types of elections (Italy and Poland). The second scenario seems to display more of a spread effect of the EU integration scale for voters of most of the relevant parties in the party systems (Germany, Greece, Sweden and the UK). This second scenario seems to reflect the fact that most parties' voters are influenced by the EU integration dimension when choosing how to vote.
It is obvious that the presence of 'hard' Euroscepticism in the party supply and/or their consequential presence of Eurosceptic content during the campaign and in the media are not sufficiently adequate to explain the scenarios detected in countries those last four countries. This is why we suggest two additional and complementary explanations for the presence of this more comprehensive stage in the Europeanisation of national party systems 6 : the pivotal role of the national authorities in the decision-making of the EU and the perception that EU decisions are heavily affecting important issues with strong national salience. In both cases, voters need to feel that their voting choices at the national level have clear consequences for both levels of governing. 6 These two hypotheses are not to be understood as opposite expectations. It might be that a country has both a high pivotal role and the presence of a hard Eurosceptic party (or the contrary). Yet this empirical challenge is partially circumvented due to our case selection.
First, then, we suggest that the EU integration dimension matters because of the country's role within the EU. As Clark and Rohrschneider (2009, 660) suggest, national issues may dominate EU issues in EP elections due to the perception that EP elections are disingenuous instruments of accountability. Following this argument, this logic might be the opposite, for instance, in countries in which national governments play a prevalent pivotal role in the EU (see Antonakakis et al. 2014), which might result in more EU citizens being aware of the role of national governments in EU decisions. A highly pivotal role might increase citizens' perception that their votes, both at the national and at the EU level, might alter European policies. This ultimately enhances the importance of the national arena as a mechanism for EU accountability. This argument fits with the literature that suggests that when voters believe governing parties are better positioned to influence EU issues, they might also be more inclined to look to EU issues in deciding to punish or reward these parties (Clark and Rohrschneider 2009).
Consequently, the main observational consequence of a country's 'pivotality' is that citizens integrate the structure of power present at the European level into their voting decision processes. In some EU countries, citizens might think that voting for certain parties in the system may have consequences in terms of the political, economic and social model that is implemented at the European level. Citizens thus might perceive that they have greater leverage to directly or indirectly alter European policies with their votes. If this is the case, the EU integration dimension will also be significant in shaping their vote choice. Finally, it is important to stress again that, if this logic is correct, the EU dimension could be relevant for both EU and national elections. 7 This could clearly be said to be the case for Germany and the UK in 2014. As we can see in Table 2, which shows the country distribution in 2014 of the Bargain Power Index 8 in EU decision-making, a proxy measuring the pivotal role of each member state, Germany holds a prevalent pivotal role in EU institutional arrangements 7 It is important to highlight here that, in the immediate aftermath of the 2008 economic crisis, national governments reverted to conventional intergovernmental diplomatic relations to resolve the euro and fiscal crisis, side-lining the EU institutions (Jones et al. 2016) and, as a result, highlighting in the eyes of their national citizens the importance of the relative power of the member states in EU decision-making processes. 8 The Shapely-Shubik power index measures the powers of players in a voting game. It is based on the ratio between the number of times each country plays a pivotal role compared to the total number of times all players (together) play pivotal roles. Sources of bargaining power include different indicators of state, institutional and individual capacity. This index has been applied to explain the distribution of power in several EU institutions, such as the Council and the European Parliament. and plays a key role in EU politics, which might explain the significance of EU individual distance in shaping vote choices for all parties. In other words, German citizens are aware that their vote can have an impact on EU policies because of their pivotal position in the EU decision-making process. Thus, the effect of the European dimension is significant for voters opting for the rightwing populist and 'hard' Eurosceptic political party Alternative für Deutschland (AfD) and for those choosing other parties, including mainstream parties.
The effect of the EU integration dimension in the UK is similar to its effect in Germany. At the time of our study, the UK happened to hold with Germany an important institutional role in the decision-making of the EU (third in the ranking with only 3.7 points difference with Germany but far distant from the rest; see Table 2). These two countries have a substantive and prevalent difference in the pivotal role of their national authorities and representatives in the decision-making process in the EU Commission, Council and even the EP (Milushev 2019).
The importance of the EU integration dimension in the vote choice for most parties in the Greek and Swedish cases might be situated in a different contextual factor: as a consequence of an EU intervention in a relevant national issue. Greek national politics suffered a significant earthquake due to the 2008 economic and financial crisis. The Eurozone authorities forced Greek governments to implement harsh austerity policies (Teperoglou and Tsatsanis 2014), making this intervention the main disputed element in national politics. This exceptional context, together with other internal factors, resulted, as discussed previously, in an important party system change. Thus, after the 2012 Greek national elections, the country entered 'a phase of triangular polarisation marked by centrist pro-European forces (represented by the old major parties), anti-austerity forces on the left and xenophobic anti-bailout forces on the right' (Teperoglou and Tsatsanis 2014, 238). The political debate revolved around polarising questions, including EU integration, the austerity measures and the bailout negotiations with the 'Troika' (ECB, FMI and EU Commission). As a consequence, the EU dimension might have subsequently played a role in levels of support for all the parties in this new 'trivotal' party system, in which all the parties from SYRIZA to Golden Dawn brought EU policies to the fore.
Something similar can be observed for Sweden. In this country, EU issues have almost always been in the political debate but especially so because of the immigration issue. This centrality has been exacerbated by discussion on how the EU has been handling immigration, which has become prevalent since the refugee crisis, making this EU issue often come to the fore in Swedish politics (Odmalm 2011), as in the UK (Hobolt 2016)

CONCLUSIONS AND DISCUSSION
In light of the findings, can we claim that individual positioning on the EU integration dimension was relevant for vote choice in the 2014 European elections? The answer is positive but with important nuances. First, the effects of the EU dimension are substantially lower than those of the left-right dimension. Second, we observe substantial heterogeneity across countries even in those countries where we observe this Europeanisation of national politics.
These findings are relevant for two reasons. First, we have shown that 'Europe matters' in explaining vote choice. Its relevance is still secondary to the traditional left-right conflicts (as predicted by the second-order model), but its effects are equally present in national and EP elections. This suggests that, to study the effect of the EU dimension, it is essential to note that the Europeanisation of national politics is not about the type of elec-  tion but the nature of the issue itself, as a growing body of literature has suggested (Kriesi et al. 2006;Hooghe and Marks 2018). Future research should continue with the focus on how the EU issue creates challenges and opportunities for an established party competition (De Sio et al. 2016), leaving aside the argument about the type of elections. Our results also offer a tentative description of the effect of the presence of Eurosceptic parties in the party supply and why, once they emerge in EU elections, they tend to remain competitive in national elections.
Crucially, we also noticed that there are still crossnational differences in the process of Europeanisation of national politics. In one scenario, the EU integration dimension is relevant only for the support given to Eurosceptic parties which can be explain by the presence of strong 'hard' Eurosceptic parties in the party supply. However, we have also detected some other countries where the EU integration dimension is also relevant for the rest of the main parties.
We appeal to and suggest two additional factors to explain this last scenario. The first one is the pivotal role of the national authorities in the decision-making of the EU. Europeanisation of national politics on voting preferences and competition might depend on a more systemic institutional configuration (Clark and Rohrschneider 2009), such as member states' bargaining power in the EU decision-making process (Antonakakis et al. 2014). This contextual factor might affect the entire party system and the average voter. The cases of Germany and the UK suggest that the EU integration dimension can become a relevant factor shaping the whole party system when voters perceive that their voting decisions have the power to alter the EU's decision-making process.
The second contextual factors might be resulting from the heavy intervention of the EU in an issue that have salient presence in national politics, generating the perception that EU decisions heavily affect it. We think that this contextual factor might also facilitate this second step of the Europeanisation of national competition making 'Europe' a central issue of the national debate, as it the case of Greece (national economy) or Sweden (immigration policies).
We are, however, aware of the limitations of our research. First, our conclusions are based on a limited number of cases. The data on which this research is based are unfortunately not present for the remaining EU countries. Second, we have proposed two contextual factors to explain cross-country variations on the presence of the EU integration dimension in voters' choices, but they both present clear limited evidence. To start with, the indicator used to measure the pivotal role of the member states is an institutional and objective one, while our argument implies the importance of subjective citizens' perceptions about such a role. About the second one, we do not have any convincing individual level data to prove not only the importance of such issue but also if citizens attribute the responsibility of handling it to the EU authorities.
Unfortunately, no survey measure does a satisfactory job capturing them. Thus, future research needs to delve into this process even further, by expanding this study to other outcomes or other countries. Finally, future studies will need to further investigate the dynamics of these multi-dimensionality conflicts in voters' preferences by paying attention to whether parties converging on one scale trigger heterogeneous effects on other scales (see van der Eijk and Franklin 2004  Note: Due to variables not being present, the same controls are not consistently included in all the models. Spain Ideological self-placement EU self-placement 0 "Extreme left" -10 "Extreme right" Nº of observations: 3,916 0 "Unification should go further" -10 "Unification has already gone too far" Nº of observations: 3,285          Standard errors in parentheses + p<0.1, * p<0.05, ** p<0.01, *** p<0.001 Zooming in on the 'Europeanisation' of national politics: A comparative analysis of seven EU countries Standard errors in parentheses + p<0.1, * p<0.05, ** p<0.01, *** p<0.001   Standard errors in parentheses. + p<0.1, * p<0.05, ** p<0.01, *** p<0.001. Figure A9. Th e eff ects of ideological distance and EU distance in the 2014 EP election and in each country's national election (Chapel Hill Survey).