Caught between sovereignty and solidarity? A multidimensional revisitation of EU mass–elite congruence

Abstract This article undertakes a critical revisitation of mass–elite congruence on EU matters, taking stock of 30 years of research and addressing durable ambiguities flagged by recent scholarship. Its specific contribution leverages EUEngage elite and mass survey data gathered in 2016 in 10 European countries. Examining congruence at both the country and the party level, we carry out an uncommon multidimensional analysis that encompasses general European integration and certain key sub-dimensions. At both levels, we perform a distinctive systematization of multiple approaches to the assessment of EU issue congruence, probing the substantive consistency of ensuing results. The findings qualify and soften the conventional wisdom of a chasm between pro-European elites and lukewarm citizens. While most countries exhibit pro-EU elite bias in terms of averages and proportions alike, mass–elite alignment is the rule when the general dimension and its sub-dimensions are understood as binary. Party-level analyses display different outcomes, depending on whether party positions are derived from elites' self-placement or their voters' perceptions, yet discrepancies are generally lower than in past assessments. Altogether, ‘constraining dissensus’ chiefly emerges along sub-dimensions concerning decision-making authority, as opposed to sub-dimensions evoking solidarity and burden-sharing. The layered panorama of congruence and incongruence implies a dependence of mass–elite interplays on context and sub-dimensions, drawing attention to the mediating role of critical junctures and elite entrepreneurship.


Introduction
Recent accounts of opinion congruence have begun to take stock of the panoply of research produced over the last decade (Lefkofridi, 2020;Shim and Gherghina, 2020). This article, which specifically delves into mass-elite congruence on EU matters, starts by undertaking a critical review of the findings in related studies, revisiting 30 years of evidence in light of the latest transnational analyses. In so doing, we pinpoint a number of lingering gaps that prevent unequivocal substantive conclusions, and we tackle several of them by leveraging mass and elite survey data from 10 European countries. The added value of our work is that we carry out a comparatively rare multidimensional analysis of EU issue congruence, focusing on salient sub-dimensions of European integration alongside the general axis, while also performing a distinctive systematization of different methodological approaches usually employed in assessments of congruence on the European dimension. Our study yields a set of fine-grained outcomes that qualify and soften the conventional wisdom of a sweeping mass-elite chasm. While opinion gaps are conditional on context and measurement, 'constraining dissensus' chiefly emerges along sub-dimensions concerning decision-making authority, as opposed to sub-dimensions evoking solidarity and burden-sharing.
Traditionally, political scientists have held the basic normative assumption that mass-elite agreementat the country level, between elected legislators and their national populations, or at the party level, between political parties and their respective votersconstitutes a benchmark of democratic representation. Linked empirically to voter turnout and satisfaction with democracy (e.g. Stecker and Tausendpfund, 2016), congruence across dimensions and specific issues has also been argued to influence political parties' electoral fortunes (Van der Eijk and Franklin, 1991). In the words of a distinguished scholar, connecting country level and party level, 'low levels of policy congruence in a system with more and more issue voting citizens, not only implies dangers for the individual parties, it also implies dangers for the system as a whole' (Holmberg, 1989: 29).
Moreover, in the crisis-ridden European context of the 2010s, mass-elite congruence has ceased to appear as 'a rather abstract measurement that may have implications for the long-term development of political systems' (Karyotis et al., 2014: 436), taking up an existential character instead and prompting renewed inspection. In recent years, the field has burgeoned, developing innovations in methodology and measurement (Golder and Stramski, 2010;Lupu et al., 2017), covering democracies outside the traditional Western European setting (Rohrschneider and Whitefield, 2012;Bornschier, 2019), assessing congruence in issue salience rather than merely in issue positions (Giger and Lefkofridi, 2014;Reher, 2015).
Within this body of research, studies assessing 'EU mass-elite congruence' have gained special relevance. Not only have EU matters become a fulcrum of party competition and a source of intense discord among national populations: analyses of issue space configuration across European democracies have also confirmed that European integration constitutes an autonomous dimension of struggle (Bakker et al., 2012(Bakker et al., , 2018Costello et al., 2012). In fact, many accounts of representational strains in contemporary European political systems underline an opinion gulf on European integration (Mattila and Raunio, 2012). In accordance with the 'constraining dissensus' identified by postfunctionalist theory (Hooghe and Marks, 2009), more divided and tepid public opinions have restrained political elites in the troubled stage experienced by the process since the Maastricht watershed. In such circumstances, the preferences and the extent of agreement between masses and elites seem highly consequential (De Vries, 2018), insofar as further sovereignty transfers depend on intergovernmental, elite-level decisions to be subjected to national public scrutiny.
However, examinations of mass-elite discrepancies on EU matters have been besetas has congruence scholarship more generallyby ambiguities concerning the definition of 'masses' and 'elites', the identification of appropriate data, the selection of well-grounded metrics. In fact, research activity throughout the past decade has increased the diversity of approaches deployed (Schmitt, 2010;Mattila and Raunio, 2012;Müller et al., 2012;Dolný and Baboš, 2015), but attempts to bridge the resulting gaps have been in short supply (a partial exception being Real-Dato, 2017). Persisting conceptual and measurement pitfalls (Real-Dato, 2017;Shim and Gherghina, 2020) have engendered a 'peaceful coexistence of research results and conclusions', which 'could remain as long as different methods of comparing mass and elite attitudes were applied to different data sets' (Müller et al., 2012: 169-170). Such problems compound the fact that 'there is no natural metric with which to measure the gap and no certainty about what kind of divergence between elites and the masses will result in trouble' (Müller et al., 2012: 168). In brief, EU issue (in)congruence is as consequential as the ground on which its evaluation currently stands is unstable.
Aware of the challenge, our contribution addresses anew the question whether, in the words of Thomassen and Schmitt (1997: 181), 'voters and their […] representatives are living in different European worlds'. In so doing, our assessment is designed to address several gaps flagged by recent scholarship (Dolný and Baboš, 2015;Stecker and Tausendpfund, 2016;Real-Dato, 2017;Shim and Gherghina, 2020). First, we present an uncommon comparison between mass and elite preferences not just on general European integration, but also on sub-dimensions related to economic and migration governance. Furthermore, we pursue systematization of results at both the country and the party level, premised on the rationale for following each path of investigation. Assessing country-level congruence, we provide a first-ever concurrent overview of measures based on averages, majorities, distributions, and directions (Pareschi et al., 2019;Shim and Gherghina, 2020), expressly probing whether different aggregation methods lead to substantively similar conclusions. In turn, our assessment of party-level congruence is novel in that we compare mass-level attitudes both to 'direct' elite-level stances (based on politicians' self-placement) and to 'indirect' ones (based on party supporters' perceptions).
The findings are fine-grained and noteworthy. At the country level, 'pro-EU elite bias' does constitute the prevailing trend, yet its pervasiveness depends on sub-dimensions, national cases, and measurement approach. Relatedly, if general European integration and its subdimensions are understood as binary disputes, the ensuing directional judgements present mass-elite 'alignment' as the general rule. At the party level, 'direct' and 'indirect' analyses unveil distinct pictures. An overall pro-EU elite bias across Europe is detected along all scales by the former; the latter confirms such evidence for sub-dimensions evoking sovereignty, but in those summoning solidarity an 'anti-EU elite bias' is suggested instead. Moreover, in comparison with previous assessments, discrepancies appear reduced in size, while concerns about a spiralling 'under-responsiveness' of party elites are largely dispelled. We argue, on such bases, that the idée reçue of a broad-brush gulf between the represented and their representatives conceals a layered panorama of congruence and incongruence, whereby the impact of 'constraining dissensus' on European integration may be mediated by critical junctures and elite entrepreneurship.
The article presents four sections. First, we review the literature on EU issue congruence, focusing on its main results and weaknesses. Then, we illustrate the details of our dataset and data analysis strategy. The third section discusses our country-level and party-level empirical analyses. In the conclusions, we recapitulate the substance and implications of our findings and reflect on possible future pathways.
Thirty years of EU issue congruence Studies of issue congruence 1 have acquired a time-honoured position within political science. After the pioneering work by Miller and Stokes (1963) on congruence between US electoral districts and their representatives, attention to representational linkages soon spread to European politics, resulting in several single-country analyses (e.g. Irwin and Thomassen, 1975). In the European setting, the 'party dyad'the relationship between the positions of a party and its voterswas firmly established as the crucial unit of analysis, together with the country level itself (Dalton, 1985). Generally, correspondence and discrepancy 2 have been assessed along the overarching left-right continuum (Dalton, 1985(Dalton, , 2017, though increasingly coupled with other issues and dimensions, following its waning as a driver of electoral behaviour. It is unfeasible to recapitulate all key findings achieved on various policy issues, national contexts, timings, and measurement approaches. Broadly speaking, however, tight correspondence has been reported along the left-right axis (Rohrschneider and Whitefield, 2012), on issues closely related to it, highly politicized or widely emphasized by parties (Belchior, 2008;Andreadis and Stavrakakis, 2017;Costello et al., 2021), but not on issues without such traits.
Studies of EU issue congruence share a family resemblance with broader congruence scholarship. Their typical starting point is descriptive rather than explanatory. Acknowledging the multiple processes involved in issue representation from below and from above (Steenbergen et al., 2007), they often settle for agnosticism and evaluate mass-elite agreement 'regardless of its origins' 3 (Reher, 2015; see also Dalton, 1985;Van der Eijk and Franklin, 1991;Holmberg, 1997). Such works are mostly anchored in mass and elite opinion surveys. 'Opinions' ideally arise from questions worded identically at the two levels, the most common formulation being a general scale that measures whether 'European unification has already gone too far' (point 0 or 1) or 'should be pushed further' (point 10). Survey responses are aggregated for analysis by country, given the bearing of congruence on the process of European integration, or by party, as expectations of an impact of congruence on electoral success are renewed by recent interest in contemporary challenges to 'mainstream' parties; normative concerns about meaningful representation supply an additional rationale at both levels. However, elite surveys are not always available. Hence, party-level works may derive party positions from alternative data sources, such as expert surveys or citizens' perceptions from mass surveys (see Ray, 2007;Andeweg, 2011).
Since the early 1990s, several studies of EU issue congruence have used transnational surveys to cover numerous countries (or their parties) at once, thus providing a few virtually pan-European snapshots of mass-elite congruence on European integration. Yet, as we shall discuss, the mentioned family resemblance has not prevented the 'peaceful coexistence' highlighted by Müller et al. (2012) in methodsand thus in outcomes and conclusionsfor mass-elite comparison. As a result, 'there is still a lack of systematic understanding of the effects of different conceptualisation and measurement on the resulting congruence', which warrants the promotion of 'a deeper examination of the issues of conceptualisation and measurement, and a systematic comparison of their results, [as] another important area of future congruence research' (Dolný andBaboš, 2015: 1295). Thus, revisiting EU issue congruence crucially entails deliberate evaluation of the (in)consistency between distinct routes to measuring the mass-elite gap.
Insofar as some scholars (McEvoy, 2012;Real-Dato, 2017) have summarized the body of knowledge on EU mass-elite congruence despite these caveats, their overviews have read out as follows: substantive correspondence on general European integration has fallen from the 1990s to the 2000s, while the overall picture for the 2010s appears more blurred and inconsistent. And where multidimensional perspectives have underpinned assessments of congruence, a yet more nuanced picture has obtained. We situate our review of the literature at the minimum level of detail which, for the purposes of this article, is compatible with accounting for the ingrained heterogeneity of the body of research.
The trailblazing study by Van der Eijk and Franklin (1991) measuring citizens' stances through an index of EU-related items and comparing them with citizens' median perceptions of party positionsfound that very few parties across the EU-12 disagreed with their average voter in 1989. Referring to the same EU-12 in 1994, Schmitt and Thomassen (2000) also found party-level congruence on the general dimension, measured by correlating mass and elite average positions, to be as strong as congruence on the left-right axis. Yet, in specific domains such as a common European currency or the removal of national borders, political elites' mean positions proved dramatically more pro-European than their voters' positions, whether at party or country level Schmitt, 1997, 1999). Conversely, Hooghe (2003) thanks to a one-time survey of EU-15 elites interviewed in 1996juxtaposed aggregate preferences among elites and citizens regarding the 'Europeanization' of 13 policy areas, spotting 3 A cognate body of research has explored the (prevalent) direction of mass-elite 'linkages' across time. Scholars have provided empirical evidence of bottom-up dynamics (Schmitt and Thomassen, 2000;Carrubba, 2001), top-down dynamics (Hellström, 2008), and bidirectional dynamics (Sanders and Toka, 2013) also depending on contextual conditions (Steenbergen et al., 2007). Recent analyses of party responsiveness (Spoon and Williams, 2017; Conti et al., 2020) and the impact of party cues (Pannico, 2020) nuance our understanding, although the overarching question remains. different patterns instead of a generalized pro-EU elite bias. All in all, what evidence existed hinted at solid congruence on general European integration, remaining more elusive on sub-dimensions.
In the 2000s, in parallel with unexpected setbacks in the integration process, multiple partylevel studies dealing with general European integration corroborated Van der Eijk and Franklin's (2004) warning about a 'sleeping giant' of mass Euroscepticism. Such examinations consistently found an overall pro-EU elite bias, with congruence already lower than on the left-right dimension and progressively diminishing. These findings held across the EU, whether the parties' positions had been ascertained through party exponents' self-placements (Costello et al., 2012), party voters' perceptions Raunio, 2006, 2012), or Chapel Hill Expert Survey scores (McEvoy, 2012). Assessments of EU issue congruenceconsistently measured on the general dimension, at the party level and through comparison of averagesprompted grim evaluations.
Further investigations relied on the IntUne project surveys of citizens and politicians carried out in 2007 in 17 European democracies. Schmitt (2010) correlated average positions both by country (weighing national electorates by proportions of party support and MPs by their parties' seat shares) and by party, discovering close correspondence on general European integration. Müller et al. (2012) assessed country-level discrepancies on distinct policy sub-dimensions between the median voter, the median party in Parliament, and the government. All national electorates, Parliaments, and governments turned out to favour cohesion policy and a single foreign policy; on common tax and social security systems, overall sentiment leaned towards support, but certain countries displayed three-way alignment against further integration, or even some internal mass-elite misalignment, whose direction was not uniform. Comparing countrylevel mean positions for masses and elites on the European dimension, Sanders and Toka (2013) showed a generalized gap between supportive elites and lukewarm populations, but the discrepancies elicited by an index of the four mentioned policy areas proved small and country-specific instead. Hence, deepening methodological divergence stoked further debate on congruence in sub-dimensions of European integration.
Relying on data from the second IntUne wave of 2009, Dolný and Baboš (2015) brought even further variety to EU issue congruence studies, by applying a novel cumulative measure of 'many-to-many' correspondence between national electorates and their elected representatives (Golder and Stramski, 2010;Andeweg, 2011). On such bases, they found congruence on the EU unification scale to be even higheralmost in every member state consideredthan on the left-right axis. Later, a longitudinal analysis covering nine countries at three time-points (2007, 2009, and 2014) noticed a certain pro-EU elite bias, at both country and party level, on the basis of mass and elite averages and medians. Intriguingly, the work failed to uncover any pattern of systematic change in EU mass-elite congruence throughout the years of the European crises 4 (Real-Dato, 2017). Finally, a party-level study exploiting mass survey data and Chapel Hill party scores (from 2014) again diagnosed a sizable pro-EU elite bias on general European integration, fuelled by gulfs between mean elite and mass positions among all mainstream party families (Bakker et al., 2018). The reported degree of mass-elite (in)congruence on the general dimension clearly depended on the measurement and level of analysis chosen.
Ultimately, the sheer variety of approaches deployedcombined with erratic inspection of policy sub-dimensions, reliance on different elite-level data sources, and yet more fine-grained detailshas combined to hinder proper systematization of findings on EU issue (in)congruence and watertight assessments of diachronic variation. Leveraging a dataset whose unique 4 This result chimes with a recent analysis having similarly asked whether the unfolding of European integration has decreased congruence across Europe since the early 1980son various issues including European integration itselfwith unambiguously negative findings (Devine and Ibenskas, 2021). However, the authors relied on a different, bespoke measure of cumulative congruence within national party systems. characteristics offer the opportunity to pursue comprehensive research outcomes, this article is purposefully built to address the specific frailties we proceed to point out.
First, while research on elite and mass opinion towards the EU is moving past the hitherto dominant one-dimensional approach (Sanders and Toka, 2013; see Hobolt and De Vries, 2016), most studies of EU mass-elite congruence have not yet adopted a multidimensional perspective embracing both general European integration and its sub-dimensions. 5 This is unfortunate because, given the protean nature of the European dimension, '[w]hich specific issue items are more salient and relevant might depend on a member state's level of EU integration issues faced at the time of measurement' (Shim and Gherghina, 2020: 513; see Goldberg et al., 2020). In other words, multidimensional examinations could shed light on the idiosyncratic strains between masses and elites taking place in different member states at any given moment, paving the way for reflections on context-specific political dynamics.
Second, while what counts as 'elites' and 'masses' has rarely been problematized, even rarer are empirical attempts to probe the robustness of congruence assessments to distinct specifications thereof. Two points are relevant here. Regarding the masses, it is unclear in most works whether they are identified, at the country level, with eligible or actual voters, and, at the party level, with party supporters or party voters (Shim and Gherghina, 2020). As for elite positions in party-level studies, infrequent have been the contributions examining multiple ways of placing parties concurrently: specifically, no past transnational analysis has cross-validated 'direct' elite-level positions based on politicians' self-placement with 'indirect' ones based on citizens' perceptions 6 (Shim and Gherghina, 2020).
Third, while recent measurement innovations led some analyses of EU issue congruence to consider more than one indicator (Real-Dato, 2017), even such investigations have not simultaneously computed measures derived from the different conceivable aggregation methods in order to evaluate their substantive consistency. In this respect, we consequently argue that all-round assessments of opinion congruence on EU matters should precede endeavours to explain discrepancies on the basis of potentially causal factors. Perhaps not coincidentally, the few existing explanatory works on EU mass-elite congruencerelying on diverse conceptualizations and indicatorshave largely reported negative findings on the causal impact of country institutions or party system traits Raunio, 2006, 2012;Dolný and Baboš, 2015).
So as to shed more light on the variety of measurement approaches and to provide necessary contextualization for our endeavour, we conclude this section with a review, based on how such approaches extract information from comparable mass and elite opinion distributions: via measures of central tendency, majorities, directions, or the distributions themselves (Pareschi et al., 2019;Shim and Gherghina, 2020).
Central tendency measures have been the default for calculating mass-elite distances (Dalton, 1985;Mattila and Raunio, 2012) or condensing information into scatterplots and correlation coefficients (Thomassen and Schmitt, 1999). Most works, whether focusing on countries or parties, have settled for simple averages, which permit straightforward tests of statistical significance (Belchior, 2008;Karyotis et al., 2014) and variability checks through standard deviations (Irwin and Thomassen, 1975;Valen and Narud, 2007). Within countries and parties, majorities have also been used, by dichotomizing a scale and computing mass and elite proportions on each side. After all, there is no certainty that mass-elite differences in proportions will neatly mirror differences in average positions. While some bespoke indicators exist (e.g. Holmberg, 1997), the 5 Even recent 'multidimensional' approaches to party-voter incongruence and vote switching did not consider subdimensions alongside the general European dimension (Bakker et al., 2018(Bakker et al., , 2020. A further work on citizens' and parties' stances towards European integration did, yet it centred on their respective polarization, rather than congruence stricto sensu (Goldberg et al., 2020). 6 Outside the literature on EU mass-elite congruence, one exception is the study by Dahlberg and Holmberg (2014). Their statistical models included an individual's distance both from an 'objective' party position and from their perception of it.
plainest modus operandi compares mass and elite figures found on the same side 7 (Freire et al., 2014).
Comparisons of distributions have been infrequent until recently: Golder and Stramski's (2010) 'many-to-many' (MTM) measure, amounting to the overlap between cumulative distribution functions, sparked visible innovation. Several empirical studies have followed their methodology or proposed functional equivalents 8 : a measure based on non-cumulative distribution functions (Andeweg, 2011) or a measure drawn from computer science, the Earth Mover's Distance (Lupu et al., 2017). Conversely, assessments based on directions remain uncommon. 9 Analyses of EU issue congruence have diagnosed pro-EU elite bias based on the sign associated with differences in means and in proportions along a scale. 10 However, except for the work by Van der Eijk and Franklin (1991), neither means nor majorities have been employed to probe binary alignment, within countries or within parties. Yet, inasmuch as European integration and its sub-dimensions are understood as political antitheses between opposed camps, surely understanding whether citizens and elites substantively align on the same side matters in its own right.
While we cannot account here for rarer, more sophisticated indicators, we mention one final line of enquiry. 'Responsiveness', as adapted by Dalton (1985) to party-level analysis 11 following the original proposal by Achen (1978), consists of a simple regression of parties' average positions over their respective supporters' ones. Hence, a slope 'greater than 1.0 means that parties accentuate the opinion differences between voter groups' (Dalton, 2017: 612), whereas a slope lower than 1.0 carries the opposite implication. In this regard, past examinations of EU issue congruence have reported a progressively worsening trend towards elite 'under-responsiveness' to the diversity of mass opinions Raunio, 2006, 2012;Real-Dato, 2017).

Data and methods: the EUEngage dataset and its advantages
The dataset we employ presents certain advantagessuch as identical wording at mass and elite levels, a short time interval between the surveys, and comparable scaleswhile avoiding some of the aforementioned shortcomings with which past research was confronted. Assembled by Wave 1 of the cross-national project EUEngage, the dataset contains a set of questions posed to elected political elites and ordinary citizens across 10 member states diversified on multiple counts. Opinion surveys targeted samples of national populations (N = 21,820) and incumbent MPs and MEPs (N = 696). The fieldwork took place from June to July 2016 at the mass level, from April to October at the elite level.
Especially crafted to generate data for country-by-country analysis, the mass survey reached approximately 2,200 responding citizens per country. As samples assembled through online panels were nonprobabilistic, quota targets were imposed for each country on demographic traits 7 The scale's midpoint presents a complication: presumably equidistant opinions are conflated with undecided or indifferent ones. For comparison's sake, the midpoint proportion may be excluded (Holmberg, 1997) or split into equal halves (Irwin and Thomassen, 1975). 8 Notably, scholars have also begun to devise statistical solutions to test the discrepancy between distributions (e.g. Andreadis and Stavrakakis, 2017). 9 In fact, directional theories of issue voting forgo the normative desirability of tight mass-elite correspondence. Certain studies have sought to substantiate elite-driven models of representation by comparing elites' and citizens' averages or majorities, on multiple issues and/or at different time-points (Holmberg, 1997;Valen and Narud, 2007). 10 Several party-level works Raunio, 2006, 2012;McEvoy, 2012) employ two average-based measures labelled as 'bias', corresponding to a difference-in-means, and 'distance', corresponding to the former's absolute value. Conversely, this article refers to 'bias' in a general manner across aggregation methods, to denote any circumstance where a certain measure reflects a mass-elite discrepancy oriented in a certain direction.

11
The term 'responsiveness' may engender ambiguity. While we follow Dalton's (1985) use, taken up by numerous partylevel studies (see the review in Powell, 2004), in some contexts (policy) 'responsiveness' marks the relationship between policy preferences in public opinion and the policy outputs produced by the political system (Wratil, 2019; see Andeweg, 2011). such as age, gender, and region, together with multiple sets of weights (for detailed information see EUEngage, 2017a). The elite survey, despite the notoriously low response rates that characterize such enquiries (Müller et al., 2012), managed to gather a minimum of 59 responses for each country, except for the Netherlands (see EUEngage, 2017b).
Alongside the general EU unification item (ranging 0-10), the dataset includes six domainspecific scales morphologically identical to it, which similarly generate opinion distributions among masses and elites. These scales measure the preferred balance between the nation-State and the EU level on matters of economic governance (two items), migration governance (three items), and international security (one item). Their wording is neither so technical as to risk 'manufacturing' opinions among less sophisticated citizens, nor so simplistic as to be unpalatable to political elites. 12 Our empirical analyses refer to General European integration plus four sub-dimensional scales: the two economy-related items, asking which governance level should hold Authority over economic policy and whether national difficulties should be tackled through Pooling of economic resources, and two migration-related items, asking which governance level should have Authority over immigration policy and cover the Costs of providing asylum. Thus, we focus on issues of special salience since the Great Recession and the migration crisis (Börzel and Risse, 2018), which presumably led not just political elites but also many common citizens to develop meaningful, relatively stable preferences. Moreover, our choice of items is somewhat symmetrical: the first item of each domain concerns decision-making authority, the second item concerns solidarity and financial burden-sharing.
The wording of the five items appears in Table 1. For clarity, we recode all five scales to range from the most pro-integration position (point 0) to the most anti-integration position (point 10).
The dataset permits comparison both between national populations and their elected representatives (country level), and between the subsets formed by supporters and politicians of individual parties (party level), since each interviewed citizen was asked to identify the party they favoured. Additionally, on all domain-specific scales, although not on the general scale, each interviewee was also asked to locate their preferred party on the same continuum on which they placed themselves. This provides an unmatched opportunity 13 for our party-level analysis to compare mass positions both with the equivalent 'direct' elite positions, derived from party elites' self-placement (Real-Dato, 2017), and with 'indirect' elite-level positions, reflecting party supporters' perceptions of their parties' standpoints Raunio, 2006, 2012). Since the latter approach relies on the perceptions of party identifiers alone rather than the general population, it may be biased toward congruence: such respondents are less likely to hold perceptions independent of their own self-placement (Andeweg, 2011), as they may project their stances onto the party (Mattila and Raunio, 2012) or take their cue from it (Andreadis and Stavrakakis, 2017). A fortiori, cross-validation proves desirable.
Our data analysis strategy is predicated on the purpose to systematize methodological approaches traditionally employed to assess EU issue congruence, comparing and contrasting the respective outcomes they yield.
At the country level, the opinions of the whole pool of elected representatives from each member state are compared to those of a sample of the national adult population. In any case, the substantive findings of this article are unaffected if the voting population alone is considered. Among average-based indicators, we compute elite and mass averages and their differences. Among majority-based indicators, we compute elite and mass dichotomized proportions -12 Congenial wording allays but does not dispel 'differential item functioning', i.e. different respondentsespecially from different countriesexpressing the same underlying attitudes through different positions along the same scale and vice versa. Strictly speaking, for instance, comparing differences-in-means across countries only assumes the distances themselves to be comparable, not the substantive positions (Golder and Stramski, 2010). 13 A recent contribution explicitly regretted that 'no comparative survey exists that asks citizens about their own positions and their perceptions of parties and governments in a multidimensional policy space' (Stecker and Tausendpfund, 2016: 498). splitting the midpoint into equal halves (Irwin and Thomassen, 1975)and the ensuing differences. Among distribution-based indicators, we use Andeweg's (2011) widespread variant (ranging 0-100): a measure that adds up, for each point of the scale, the minimum between the elite-level and the mass-level proportion. Among direction-based indicators, we exploit the mean scores and the dichotomized proportions computed for each country to distil two parallel directional judgements. In each, four scenarios may arise: 'pro-EU alignment', 'anti-EU alignment', 'pro-EU elite bias', and 'anti-EU elite bias'.
At the party level, partisan sorting of MPs and MEPs is unproblematic. Two separate questions permit the sorting of citizens into 'party dyads', respectively rooted in declared party closeness (identifying party supporters) or prospective vote intention (identifying party voters). While we adopt the former option, the entire set of empirical assessments is robust to replication based on the latter. Our stated purpose is effectively served through measures of central tendency: mass and elite mean positions are employed to underpin both 'direct' and 'indirect' analyses in parallel. Furthermore, to probe whether parties are in effect 'drifting further apart' from their voters (Mattila and Raunio, 2012), we leverage such averagesboth direct and indirectto obtain analyses of 'responsiveness' along each scale, accompanied by the corresponding crossnational correlations and scatterplots.
As a preliminary move, we must clarify which partiesbelonging to the party systems of the 10 member statesshould be considered relevant and incorporated into the analysis. Sartori's (1976) criteria, 'coalition potential' and 'blackmail potential', appear ill-suited to ascertain party relevance at an exact time-point. Moreover, including all parties in Parliament, or all parties above a certain electoral threshold, is not viable either, inasmuch as some had no respondents in the EUEngage elite survey. Out of necessity, but in line with consensual practice in previous studies of issue congruence, we consider parties represented in our dataset by a minimum of X 1 elite interviewees and X 2 mass respondents, though well aware of the unavoidable 'tradeoff between the reliability of party estimates and the number of party dyads produced by thresholds' (Dalton, 2017: 620). With a threshold of 25 party supporters, our indirect analysisentirely based on citizens' self-placement and perceptionscomprises 77 parties. Settling for an additional threshold of two party elites, 14 we narrow our direct analysis to Each country should rely on its own resources to fix its economic problems (0) vs. The European Union's countries should pool resources to fix economic problems (10) Authority over immigration policy [COUNTRY] should decide for itself how many immigrants to accept each year (0) vs. The European Union should decide how many immigrants should be accepted by each Member State each year (10) Costs of providing asylum The costs of providing asylum should be shared among all the European Union's Member States (0) vs. Each country should bear the costs depending on how many asylum seekers it receives (10)

14
Since elite surveys often exhibit low response rates, as already mentioned, many party-level studies have relied on thresholds of even two or three elite respondents (e.g. Luna and Zechmeister, 2005;Belchior, 2008;Dalton, 2017). To maximize analysed party dyads, we also pool MPs and MEPs, although these legislators operate in different arenas and may show different socialization patterns. Ultimately justified by the common partisan 'glue', this move is also instrumental in including minor parties of interest (e.g. the British UKIP). 40 political forces. 15 The outcomes of the indirect analysis are robust to replication based on the smaller set of parties comprised in the direct analysis.

A comprehensive assessment of EU issue congruence
First of all, we tackle issue congruence on general European integration at the country level. Table 2 displays average-based, majority-based, direction-based, and distribution-based indicators for the 10 countries. The first two columns highlight average scores on the 0-10 continuum for masses and elites, respectively; their difference is displayed in the third column, while the fourth holds the attendant directional judgement. The next four columns follow the same logic with regards to majorities: after listing mass and elite proportions located on the pro-integration side, they present the ensuing difference and directional evaluation. The last column exhibits the MTM indicator (which is non-directional).
First, in average terms the elites turn out to favour integration more than their citizens in eight countries. In six cases the difference is statistically significant, while no difference-in-means in the opposite direction is. 16 Furthermore, this overall pro-EU elite bias is entirely mirrored by the evaluation based on differences in proportions. The precise figures for the MTM measure are less immediately interpreted, since the indicator bears no sign and cannot be judged against any obvious benchmark; still, consistency between measures appears throughout. Indeed, a matrix of correlations 17 (not displayed here) between the average-based, the majority-based, and the distribution-based measures as reported in Table 2 yields coefficients of at least 0.86. Indicators based on different aggregated points can occasion slightly different overall rankings and slightly different 'health checks' for an individual country, Italy being a case in point; yet, high-congruence countries, such as Portugal, the UK, and Poland, are clearly distinguished from low-congruence cases, like France, Germany, and Greece.
Directional judgementswhether anchored in averages or in majoritieslead to a radically divergent interpretation. Inasmuch as EU issue congruence is understood as binary, national alignment between the represented and their representatives becomes the near-ubiquitous rule (9 countries out of 10, with the sole exception of France). In particular, Germany and Greece show 'pro-European alignment' despite their rather tepid public opinions, as does Spain, notwithstanding the quantitatively sizable gap that separates its extremely pro-European elites from its still solidly pro-integration population. In turn, 'anti-EU alignment' characterizes Czechia, the UK, and the Netherlands. Ultimately, regarding the matter as inherently dichotomousrather than continuouscauses pro-EU elite bias to disappear almost entirely.
In a second step, we consider EU mass-elite congruence, again at the country level, along the four domain-specific scales. While in Appendix A we include a table akin to Table 2 for each sub-dimension, the following Table 3 condenses the main results of our examination of congruence in relation to Authority over economic policy, Pooling of economic resources, Authority over immigration policy, and Costs of providing asylum (with General European integration included for comparison). For each scale, the first three columns report the average difference-in-means across the 10 countries, the number of cases whose differences-in-means express quantitative pro-EU elite bias, and the number of statistically significant ones. 18 Three further columns describe our average-based directional judgement, specifying how many countries are 15 We exclude all Dutch parties from the direct analysis, since the elite threshold would rule out most of them. The Portuguese parties are excluded from all analyses due to a possible coding imperfection in our version of the dataset. 16 All calculations are performed with STATA/SE 13.0. As its ttest command prevents any usage of weights, we compute significance through the svyset routine and the commands mean and test.

17
Since differences in means and in proportions, unlike the MTM indicator, reflect directional discrepancy through their sign, their absolute values are employed when computing correlations. 18 The third figure is a subset of the second one, since no difference-in-means expressing anti-EU elite bias reaches statistical significance. Three stars (***), two stars (**) and a single star (*) respectively correspond to P < 0.001, P < 0.01, P < 0.05. characterized by 'pro-EU alignment', 'anti-EU alignment', or 'pro-EU elite bias' respectively. Then, two columns spell out the average difference in proportions and the number of cases displaying quantitative pro-EU elite bias on such bases. The next three columns display the outcomes of directional evaluations rooted in proportions. Finally, the last column contains the average value of the MTM indicator.
Regarding the strictly quantitative outcomes, pro-EU elite bias seems to be the rule along the four sub-dimensions, too. In either seven or eight countries, on each scale, political elites favour European integration on average more than their national populations do, and in four to six cases such differences are statistically significant (whereas no distance bearing the opposite sign is). Accordingly, the average figure across the 10 cases is also consistently positive. Judging on the basis of such evidence, the two migration-related scales appear to exhibit the widest mass-elite gapssimilarly to the general dimension of European integrationwhereas Pooling of economic resources has the lowest.
In terms of proportions, along every scale seven to nine national cases are characterized by quantitative pro-EU elite bias and the overall average score is noticeably negative (though less so in relation to Pooling of economic resources). Otherwise put, as with the general dimension, majorities largely mirror averages. Indeed, for each scale the mean-based and the majority-based differences correlate by at least 0.94, and across the four scales very few cases are marked by a difference in proportions whose sign is inconsistent with that of the corresponding difference in means. However, while the minimum correlation coefficient between the MTM indicator and the other two measures amounts to 0.85 along the two economy-related scales, it decreases to 0.60 for Authority over immigration policy and even to 0.43 for Costs of providing asylum. Thus, whether the MTM measure actually appraises the same observandum as its two counterparts is moot.
Moreover, the binary directional evaluations support narratives which do not coincide with those arising from simple differences in means and in proportions. According to both averagebased and majority-based binary assessments, on each scale mass-elite alignment prevails once again. At the same time, the directional criterion contributes to showcasing the greatest representational strains among the sub-dimensions. Despite the sizable quantitative discrepancies exhibited by several countries, Costs of providing asylumtogether with Pooling of economic resources and also General European integrationdisplays agreement in at least eight national cases, most of which have pro-EU alignment. Conversely, Authority over immigration policy reveals its contentiousness in full: dichotomous alignment characterizes six countries, but only the four where masses and elites converge on the anti-EU side are spared from wide quantitative gaps. Authority over economic policy may elicit comparatively limited mass-elite differences, but the clear divide is between four countries showing binary pro-EU elite bias and the remaining six marked by anti-EU alignment: the national elites embracing a pan-European perspective do so at their own risk.
We now turn to the third step of our assessment, addressing the whole array of party dyads. The first row in Table 4 displays Pearson's r, obtained by correlating the mass and elite average positions corresponding to the parties. The second row exhibits the slope b resulting from responsiveness analyses, which regress elite positions on mass positions. The third row comprises the average difference-in-means computed for each scale, with positive scores corresponding to quantitative pro-EU elite bias; the fourth row complements the previous one, accounting for the number of party dyads characterized by a positive score along the scale.
For illustrative purposes, we also display Figures 1-5. The first graph consists of the transnational party-level scatterplot for General European integration, with elite-level positions only defined in the direct way (since citizens' perceptions were not available for this scale). The other four graphs show the equivalent scatterplots, depending on the direct and indirect party positions, for the two migration-related items: Authority over immigration policy and Three stars (***), two stars (**) and a single star (*) respectively correspond to P < 0.001, P < 0.01, P < 0.05. Costs of providing asylum. For ease of interpretation, the same shape is used for the data markers representing all parties of each country. Along the general dimension, parties' standpoints correlate with their electoral constituencies' positions rather well (coefficient: 0.74). Nevertheless, visual inspection of Figure 1 suggests that the direct positions of parties span a wider range than their supporters' mean stances. Accordingly, responsiveness analysis yields a slope of 1.36, confirming that at party levelacross the European party systems considered as a wholethe elites are somewhat 'over-responsive' vis-à-vis the masses. The average difference-in-means assumes a positive value of 0.57; 29 party dyads out of 40 have positive differences between their elite and mass average scores. Such evidence warrants the presence of pro-EU elite bias in quantitative terms in the aggregate.
Again in direct terms, the findings concerning the four domain-specific scales are quite similar. The correlation coefficients are even higher (ranging between 0.76 and 0.88), while all four slopes resulting from regression analyses are decidedly higher than 1 (ranging between 1.44 and 1.87). Alongside the over-responsiveness characterizing the elites' position-taking, an overall pro-EU elite bias is also present, although its intensity varies slightly across scales. As a matter of fact, the average difference-in-means across the four sub-dimensions amounts to 0.46, 0.08, 0.58,  and 0.46 respectively, whereas the number of parties whose elites favour European integration more than their supporters consistently exceeds 50% of the cases (26, 23, 24, and 27 party dyads).
In indirect terms, the scenario is both strikingly different and internally heterogeneous. As expected, mass-elite discrepancies now appear heavily reduced: correlation coefficients approach perfection, attaining a minimum figure of 0.95, whereas all points symbolizing party dyads are much closer to the 45-degree line, as exemplified by Figures 3 and 5. Less plainly, the regression slopesnow amounting to 0.86, 0.86, 0.91, and 0.85are all lower than 1. Thus, party supporters throughout Europe actually attribute a certain under-responsiveness to their preferred political forces as a whole. Furthermore, in one decisive respect Authority over economic policy and Authority over immigration policy diverge from Pooling of economic resources and Costs of providing asylum. Along the former two scales, the average difference-in-means amounts to 0.17 and to 0.14, with as many as 58 and 53 cases out of 77 showing quantitative pro-EU elite bias. The reverse happens along the latter two, where the average difference is negative (−0.37 and −0.21) and the party dyads displaying positive scores are a minority (8 and 25). On matters entailing European solidarity, then, political parties are perceived to express a widespread, if quantitatively limited, 'anti-EU elite bias'.

Conclusions
Is there a veritable gulf between political elites and ordinary citizens on European integration? Scholarship on EU mass-elite congruence has surely attained noteworthy results over 30 years of research (among others, Van der Eijk and Franklin, 1991;Thomassen and Schmitt, 1999;Hooghe, 2003;Schmitt, 2010;Mattila and Raunio, 2012;McEvoy, 2012;Müller et al., 2012;Dolný and Baboš, 2015;Real-Dato, 2017;Bakker et al., 2018). At the same time, it suffers from conceptual and measurement shortcomings (Real-Dato, 2017; Shim and Gherghina, 2020) whose persistencefurther entrenched by the intense activities of the last decadehinders any search for unequivocal substantive conclusions (Müller et al., 2012;Dolný and Baboš, 2015). Against such an unsettled background, this article has undertaken an innovative revisitation of EU issue congruence, leveraging data gathered in 2016 in 10 European countries with a view to filling some gaps in the systematization of evaluations of mass-elite discrepancy.
We have addressed congruence both at the country level, between national populations and their elected representatives, and at the party level, between supporters and elected officials of political parties. The article has taken stock of the existing literature and it has undertaken a dialogue with its most recent findings, while contributing several novel aspects. First, we have performed an uncommon multidimensional evaluation, assessing mass-elite agreement not simply on general European integration, but also on sub-dimensions touching on matters of sovereignty and solidarity in relation to economic and migration governance. Second, at the country level we have conducted a first-ever concurrent assessment of measures based on four distinct aggregation methods averages, majorities, distributions, and directionspurposely to probe whether they produce consistent findings. Third, at the party level we have simultaneously compared party supporters' attitudes both to 'direct' elite-level positions, rooted in party politicians' self-placement, and to 'indirect' elite-level positions, rooted in citizens' perceptions. Fourth, comprehensive checks safeguard the robustness of our findings to different definitions of the mass level (the voting population rather than the whole adult population, party voters rather than party supporters).
At the country level, our focus on sub-dimensions and measurement yields outcomes that problematize the conventional wisdom of a broad-brush 'pro-EU elite bias'. Admittedly, along the general dimension as well as each sub-dimension, in strictly quantitative terms mass-elite differences in means and in proportions clearly identify pro-EU elite bias (often statistically significant) in most national cases. However, binary directional judgements, stating whether national populations and their representatives stand on the same side of the divide, dramatically downgrade mass-elite rifts, presenting national alignment as the rule along each scale. As a whole, our findings do not deviate from the country-level assessmentsalbeit more circumscribedperformed by Müller et al. (2012), Sanders and Toka (2013), and Dolný and Baboš (2015). While the overall mass-elite gap across Europe consists of a pro-EU elite bias, its size widely depends on measurement and is uneven across sub-dimensions, whereas some countries nearly eschew it altogether. On a final note, while average-based and majority-based measures display close consistency, a distribution-based indicator covaries loosely with them at least on some sub-dimensions, which calls into question the relation between these measures and a supposedly unique observandum.
At the party level, our cross-validation of direct and indirect elite positions produces remarkable results. As expected, congruence between parties and their supporters is higher in indirect terms, presumably because numerous citizens project their views onto their preferred party or vice versa. Other findings are surprising. First, our evidence contradicts the regression models presented by Raunio (2006, 2012) and by Real-Dato (2017), whereby the 'responsiveness' of parties to the spectrum of voters' positions on European integration is fragile and shrinking across time. Along each scale, in indirect terms, we only unearth a slight under-responsiveness of elite positions. Furthermore, directly computed average party positions throughout Europe are scattered across the 0-10 continuum more than their supporters' mean standpoints, so that European parties actually appear to be over-responsive. The divergence from preceding studies may depend on the subsequent emergence of parties steadfastly opposing Europe, but also on the radical pro-integration positions embraced by certain established political forces.
According to the direct analysis again rooted in average positions, along each scale political parties across European countries are marked on the whole by pro-EU elite bias vis-à-vis their supporters. The indirect assessment, conversely, separates the two sub-dimensions evoking matters of sovereignty, where an overall pro-EU elite bias is perceived by citizens, from those concerning matters of solidarity, where an aggregate anti-EU elite bias is perceived instead. Although our figures are not immediately comparable with previous analyses (McEvoy, 2012;Mattila and Raunio, 2012;Bakker et al., 2018), in substantive terms this article finds lower degrees of party-level incongruence, both direct and indirect. Recent studies, emphasizing an individuallevel mechanism whereby incongruence with one's preferred party on EU matters increases the probability of voting for a 'challenger' party, foresee further space for such political forces given the unresponsiveness of mainstream ones (Bakker et al., 2018(Bakker et al., , 2020. Although this article does not tackle the question directly, our data mitigates such a scenario, insofar as it scales down representational deficits across the pan-European party landscape. On the one hand, our results imply lessons referring to measurement. The degrees of inconsistency between different statistical approaches call for methodological reflexivity, demanding that evaluations of mass-elite discrepancy include multiple checks, whose import must be carefully explored. At the country level, where our evaluation has comprised several indicators of EU issue congruenceone for each aggregation methodbut by no means all existing ones, further work is warranted in this direction no less than in searching for explanatory factors. At the party level, too, the consistency between different congruence indicators is a significant question, also with a view to uncovering which ones may capture the ebbs and flows of electoral success. In this vein, we reprise the purpose of the special issue edited by Shim and Gherghina (2020: 511), namely to 'provide a springboard from which students of mass-elite congruence can reflect on and discuss the globally relevant methodological choices they make'. Still, it is also worth recalling that standard survey answers may not always 'display to the same extent the qualitiessincerity, consistency, stabilitythat justify their interpretation through various statistical operations of distribution, aggregation, comparison' (Aldrin, 2011: 19). Remedial strategies should be applied, but encouraging a relaunch of further research framing mass-elite comparisons in interpretative ways remains sensible nonetheless.
On the other hand, the findings of this study bring important substantive contributions. Most clearly, our multidimensional perspective on EU issue congruence reveals different dynamics that beyond the general European dimensionpit sovereignty against solidarity. Along the two subdimensions corresponding to decision-making authority, four countries show binary pro-EU elite bias, with at least four other national cases displaying alignment, but on the anti-EU side. Moreover, party-level analyses dealing with such sub-dimensions reveal an overall quantitative pro-EU elite bias, objective and perceived. On the contrary, when it comes to burden-sharing, member states marked by pro-EU alignment are in a solid majority, and citizens' perceptions demonstrate a conviction that the extent to which they favour pro-integration solutions is not reflected in full by political parties across Europe. All in all, the evidence we weigh spells trouble for the trajectory of European integration, and yet it fundamentally suggests a dependence of mass-elite interplays on context and sub-dimensions, spurring scholars to reflect on the relevance of critical junctures and elite agency in 'weaponizing' opinion discrepancies.
Funding. The research received no grants from public, commercial, or non-profit funding agency.
Data. The replication dataset is available at http://thedata.harvard.edu/dvn/dv/ipsr-risp. took part in the workshop 'Convergence versus Divergence of Mass-Elite Political Cleavages', in the framework of the ECPR Joint Sessions 2019. Finally, we thank the two anonymous referees for their valuable comments.