Thirty years of EU issue congruence

Studies of issue congruenceFootnote ¹ have acquired a time-honoured position within political science. After the pioneering work by Miller and Stokes (Reference Miller and Stokes1963) 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, Reference Irwin and Thomassen1975). In the European setting, the ‘party dyad’ – the relationship between the positions of a party and its voters – was firmly established as the crucial unit of analysis, together with the country level itself (Dalton, Reference Dalton1985). Generally, correspondence and discrepancyFootnote ² have been assessed along the overarching left-right continuum (Dalton, Reference Dalton1985, Reference Dalton2017), 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, Reference Rohrschneider and Whitefield2012), on issues closely related to it, highly politicized or widely emphasized by parties (Belchior, Reference Belchior2008; Andreadis and Stavrakakis, Reference Andreadis and Stavrakakis2017; Costello et al., Reference Costello, Toshkov, Bos and Krouwel2021), 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., Reference Steenbergen, Edwards and De Vries2007), they often settle for agnosticism and evaluate mass–elite agreement ‘regardless of its origins’Footnote ³ (Reher, Reference Reher2015; see also Dalton, Reference Dalton1985; Van der Eijk and Franklin, Reference Van der Eijk and Franklin1991; Holmberg, Reference Holmberg1997). 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, Reference Ray2007; Andeweg, Reference Andeweg, Rosema, Denters and Aarts2011).

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. (Reference Müller, Jenny, Ecker, Best, Lengyel and Verzichelli2012) in methods – and thus in outcomes and conclusions – for 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 and Baboš, Reference Dolný and Baboš2015: 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, Reference McEvoy2012; Real-Dato, Reference Real-Dato2017) 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 (Reference Van der Eijk and Franklin1991) – measuring citizens' stances through an index of EU-related items and comparing them with citizens' median perceptions of party positions – found 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 (Reference Schmitt and Thomassen2000) 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 (Thomassen and Schmitt, Reference Thomassen and Schmitt1997, Reference Thomassen, Schmitt, Schmitt and Thomassen1999). Conversely, Hooghe (Reference Hooghe2003) – thanks to a one-time survey of EU-15 elites interviewed in 1996 – juxtaposed aggregate preferences among elites and citizens regarding the ‘Europeanization’ of 13 policy areas, spotting 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 party-level studies dealing with general European integration corroborated Van der Eijk and Franklin's (Reference Van der Eijk, Franklin, Marks and Steenbergen2004) 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., Reference Costello, Thomassen and Rosema2012), party voters' perceptions (Mattila and Raunio, Reference Mattila and Raunio2006, Reference Mattila and Raunio2012), or Chapel Hill Expert Survey scores (McEvoy, Reference McEvoy2012). Assessments of EU issue congruence – consistently measured on the general dimension, at the party level and through comparison of averages – prompted grim evaluations.

Further investigations relied on the IntUne project surveys of citizens and politicians carried out in 2007 in 17 European democracies. Schmitt (Reference Schmitt2010) 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. (Reference Müller, Jenny, Ecker, Best, Lengyel and Verzichelli2012) 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 country-level mean positions for masses and elites on the European dimension, Sanders and Toka (Reference Sanders and Toka2013) 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 (Reference Dolný and Baboš2015) 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, Reference Golder and Stramski2010; Andeweg, Reference Andeweg, Rosema, Denters and Aarts2011). On such bases, they found congruence on the EU unification scale to be even higher – almost in every member state considered – than 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 crisesFootnote ⁴ (Real-Dato, Reference Real-Dato2017). 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., Reference Bakker, Jolly and Polk2018). 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 deployed – combined with erratic inspection of policy sub-dimensions, reliance on different elite-level data sources, and yet more fine-grained details – has combined to hinder proper systematization of findings on EU issue (in)congruence and watertight assessments of diachronic variation. Leveraging a dataset whose unique 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, Reference Sanders and Toka2013; see Hobolt and De Vries, Reference Hobolt and De Vries2016), most studies of EU mass–elite congruence have not yet adopted a multidimensional perspective embracing both general European integration and its sub-dimensions.Footnote 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, Reference Shim and Gherghina2020: 513; see Goldberg et al., Reference Goldberg, van Elsas and de Vreese2020). 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, Reference Shim and Gherghina2020). 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' perceptionsFootnote ⁶ (Shim and Gherghina, Reference Shim and Gherghina2020).

Third, while recent measurement innovations led some analyses of EU issue congruence to consider more than one indicator (Real-Dato, Reference Real-Dato2017), 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 congruence – relying on diverse conceptualizations and indicators – have largely reported negative findings on the causal impact of country institutions or party system traits (Mattila and Raunio, Reference Mattila and Raunio2006, Reference Mattila and Raunio2012; Dolný and Baboš, Reference Dolný and Baboš2015).

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., Reference Pareschi, Baldini and Giglioli2019; Shim and Gherghina, Reference Shim and Gherghina2020).

Central tendency measures have been the default for calculating mass–elite distances (Dalton, Reference Dalton1985; Mattila and Raunio, Reference Mattila and Raunio2012) or condensing information into scatterplots and correlation coefficients (Thomassen and Schmitt, Reference Thomassen, Schmitt, Schmitt and Thomassen1999). Most works, whether focusing on countries or parties, have settled for simple averages, which permit straightforward tests of statistical significance (Belchior, Reference Belchior2008; Karyotis et al., Reference Karyotis, Rüdig and Judge2014) and variability checks through standard deviations (Irwin and Thomassen, Reference Irwin and Thomassen1975; Valen and Narud, Reference Valen and Narud2007). 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, Reference Holmberg1997), the plainest modus operandi compares mass and elite figures found on the same sideFootnote ⁷ (Freire et al., Reference Freire, Teperoglou and Moury2014).

Comparisons of distributions have been infrequent until recently: Golder and Stramski's (Reference Golder and Stramski2010) ‘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 equivalentsFootnote ⁸: a measure based on non-cumulative distribution functions (Andeweg, Reference Andeweg, Rosema, Denters and Aarts2011) or a measure drawn from computer science, the Earth Mover's Distance (Lupu et al., Reference Lupu, Selios and Warner2017). Conversely, assessments based on directions remain uncommon.Footnote 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.Footnote ¹⁰ However, except for the work by Van der Eijk and Franklin (Reference Van der Eijk and Franklin1991), 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 (Reference Dalton1985) to party-level analysisFootnote ¹¹ following the original proposal by Achen (Reference Achen1978), 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, Reference Dalton2017: 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 (Mattila and Raunio, Reference Mattila and Raunio2006, Reference Mattila and Raunio2012; Real-Dato, Reference Real-Dato2017).

Data and methods: the EUEngage dataset and its advantages

The dataset we employ presents certain advantages – such as identical wording at mass and elite levels, a short time interval between the surveys, and comparable scales – while 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 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., Reference Müller, Jenny, Ecker, Best, Lengyel and Verzichelli2012), 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 domain-specific 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.Footnote 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, Reference Börzel and Risse2018), 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).

Table 1. EUEngage survey questions – preferences on European integration

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 opportunityFootnote 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, Reference Real-Dato2017), and with ‘indirect’ elite-level positions, reflecting party supporters' perceptions of their parties' standpoints (Mattila and Raunio, Reference Mattila and Raunio2006, Reference Mattila and Raunio2012). 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, Reference Andeweg, Rosema, Denters and Aarts2011), as they may project their stances onto the party (Mattila and Raunio, Reference Mattila and Raunio2012) or take their cue from it (Andreadis and Stavrakakis, Reference Andreadis and Stavrakakis2017). 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 – splitting the midpoint into equal halves (Irwin and Thomassen, Reference Irwin and Thomassen1975) – and the ensuing differences. Among distribution-based indicators, we use Andeweg's (Reference Andeweg, Rosema, Denters and Aarts2011) 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, Reference Mattila and Raunio2012), we leverage such averages – both direct and indirect – to obtain analyses of ‘responsiveness’ along each scale, accompanied by the corresponding cross-national correlations and scatterplots.

As a preliminary move, we must clarify which parties – belonging to the party systems of the 10 member states – should be considered relevant and incorporated into the analysis. Sartori's (Reference Sartori1976) 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 X₁ elite interviewees and X₂ 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, Reference Dalton2017: 620). With a threshold of 25 party supporters, our indirect analysis – entirely based on citizens' self-placement and perceptions – comprises 77 parties. Settling for an additional threshold of two party elites,Footnote ¹⁴ we narrow our direct analysis to 40 political forces.Footnote 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).

Table 2. Country-level measures of mass–elite congruence – General European integration

Three stars (***), two stars (**) and a single star (*) respectively correspond to P < 0.001, P < 0.01, P < 0.05.

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.Footnote 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 correlationsFootnote 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 judgements – whether anchored in averages or in majorities – lead 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 dichotomous – rather than continuous – causes 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.Footnote ¹⁸ Three further columns describe our average-based directional judgement, specifying how many countries are 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.

Table 3. Synthetic country-level evidence on mass–elite congruence along the general dimension and the four sub-dimensions

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 gaps – similarly to the general dimension of European integration – whereas 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 average-based 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 asylum – together with Pooling of economic resources and also General European integration – displays 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.

Table 4. Synthetic party-level evidence on mass–elite congruence along the general dimension and the four sub-dimensions, in direct and in indirect terms

Three stars (***), two stars (**) and a single star (*) respectively correspond to P < 0.001, P < 0.01, P < 0.05.

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 Costs of providing asylum. For ease of interpretation, the same shape is used for the data markers representing all parties of each country.

Figure 1. Party-level scatterplot representing pairs of elite and mass average positions, in direct terms (General European integration).

Figure 2. Party-level scatterplot representing pairs of elite and mass average positions, in direct terms (Authority over immigration policy).

Figure 3. Party-level scatterplot representing pairs of elite and mass average positions, in indirect terms (Authority over immigration policy).

Figure 4. Party-level scatterplot representing pairs of elite and mass average positions, in direct terms (Costs of providing asylum).

Figure 5. Party-level scatterplot representing pairs of elite and mass average positions, in indirect terms (Costs of providing asylum).

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 level – across the European party systems considered as a whole – the 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 slopes – now amounting to 0.86, 0.86, 0.91, and 0.85 – are 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’.