Eurovisions: An Exploration and Explanation of Public Preferences for Future EU Scenarios

Abstract

Public opinion on the EU has received growing attention in the last decades, with an ever‐increasing number of studies examining various aspects of it. Surprisingly, most studies focus on attitudes towards the past and present of the EU, yet we know very little about public attitudes towards the future of the EU. This study helps to fill this research gap by examining attitudes towards the EU's long‐term future using a novel approach. We developed eight concrete future EU scenarios based on an inductive analysis of qualitative survey data. Subsequently, respondents (in an independent survey) ranked their top three scenarios according to individual preferences. Using multidimensional unfolding, we show that these preferences form three clusters ordered along a more versus less EU dimension. In a second step, we used multinomial logistic regression to examine not only who supports which scenario (socio‐demographics) but also which EU attitudes lead to which future preferences. The analyses identify distinct characteristics and attitudes that drive people's preference for a given scenario. Overall, we find that factors such as occupational levels or left–right attitudes are strong determinants of preferences for the future of the EU, and that specific EU support (performance and utilitarian evaluations) is more important than diffuse EU support (identity and affect).

Introduction

The British electorate's vote of 51.9 per cent in support of a British EU exit shocked many politicians and pundits alike. Even if the prevalence of euroscepticism among the British population has long been acknowledged, many did not foresee the translation of this eurosceptic sentiment into the concrete, real‐world consequence of an EU exit. But what kind of scenario for the future UK–EU relation did voters actually have in mind when casting their vote? If anything became apparent in the aftermath of the referendum, it is that there are many different possible scenarios. Among the options, a ‘hard Brexit’ means cutting the ties with the EU altogether, including access to the internal market and the customs union, while a ‘soft Brexit’ refers to any scenario where Britain keeps some of its ties with the EU. It is also possible for some to have voted ‘Leave’ out of discontent with the current EU and a wish for reforms, but not to actually see Britain leave the EU altogether. The binary nature of the Brexit vote, however, gives us no insight into preferences for future scenarios of the UK inside or outside the EU.

This is not a peculiarity of the British case, but a general issue in EU public opinion research. We know very little about what EU citizens want for the future of the EU. Knowledge of such preferences and even more so, the underlying reasons for them, though, could be of utmost importance to understand and anticipate events such as Brexit in other EU countries.

Aiming to understand such future EU preferences, we developed a set of eight concrete and possible scenarios for the future of the EU (partly based on a content analysis of original qualitative survey data). Subsequently we asked citizens to rank these future scenarios of their country inside or outside the EU, and assess how such preferences are structured. This scenario‐ranking approach addresses criticism that there is a mismatch between the task of policymakers and available public opinion data. Van Knippenberg and Daamen (1996, p. 70) argue that ‘whereas policy makers are usually faced with the choice between several alternative courses of action, opinion surveys often assess opinions about each separate alternative without actually requesting respondents to make a choice between alternatives’. Our newly designed survey question enables us to assess priorities among alternative scenarios. In a second step, we analyse which citizens opt for which scenarios, and why. This latter question is addressed by analysing how citizens' evaluations of the current EU explain their preferences for future EU scenarios.

Our study contributes to the literature on EU public opinion in at least three ways. Firstly, we use a novel scenario‐based approach to examine public preferences toward the EU, thus bridging the gap between public opinion and the choices faced by policymakers. Secondly, we explicitly address future preferences, while most public opinion studies have focused on evaluations of the past and present EU. Thirdly, we link these future EU preferences to commonly used EU attitude dimensions (referring to the present or past), thereby adding to our knowledge of the possible political consequences of current evaluations of the EU.

We examine the case of the Netherlands using original survey data (N = 2,648). While the Netherlands, as one of the founding members of the EU, has traditionally supported the EU, euroscepticism has come to the mainstream in recent decades (Lubbers and Scheepers, 2010), and this has had real consequences for voting behaviour. The Dutch voted down both the EU Constitution in 2005 and the EU–Ukraine Association Treaty in 2016. Even before the Brexit vote, there were calls for a Nexit referendum in the Netherlands, particularly from right‐wing populist actors – while the main left‐wing populist party called for a referendum on major EU reforms. ¹ Despite this, we still know little about which (future) scenario Dutch citizens have in mind when they are more supportive of the EU or less so, and why. Our study takes the first step into examining the variety in such preferences and explaining what motivates them.

1. Theory

1.1. Examining Future Visions of the EU

Despite the richness of literature about European integration and public opinion towards the EU, ‘there is surprisingly little research that actually deals with the EU's long‐term future’ (Meyer, 2018, p. 29). In recent times, events such as Brexit have laid bare the fact that such future plans or visions for the EU are necessary to keep the whole European project going. Although for a long time there seemed to be a simple binary choice between having more or less European integration, a recent survey‐based report by Raines et al. (2017) shows that both the elite and the public have very broad and diverse opinions towards the EU, which go beyond the discussion between ‘more’ versus ‘less’ Europe. The authors claim that as a result ‘a richer, broader and perhaps even a more conflictual debate’ is needed for a political renewal in Europe, and one that does justice to the variety of public preferences (Raines et al., p. 40).

Another issue that Brexit has revealed is the missing link between opinions or attitudes towards the current EU and preferences for any future changes in the EU that follow from it. For instance, a vote in favour of an EU exit does not come with a clear prescription of what alternatives should be negotiated. Even if citizens agree that the EU is not functioning well, they may have very different preferences for what should be altered. These diverse preferences are not represented in the simple yes/no format of a referendum. Similarly, the countless studies examining EU public opinion in general (Gabel, 1998; Hooghe and Marks, 2005; McLaren, 2006) and euroscepticism in particular (De Vries and Edwards, 2009; Lubbers and Scheepers, 2010) are helpful in evaluating the present level of support or criticism of the EU and the underlying reasons for it. However, most existing measures entail uncertainties: for example, someone who wants her country to stay in the EU but supports reforms in the future may respond either positively or negatively on a general scale of EU support depending on the precise considerations she has in mind when responding. Furthermore, existing measures do not ask citizens what their ideal future EU should look like. In the aftermath of Brexit and the wave of rising euroscepticism across the continent, especially scenarios of ‘less’ Europe may have developed in the minds of (more sceptical) citizens. In line with these considerations we aimed to examine especially scenarios implying less Europe.

Providing citizens with concrete future scenarios has been done in the past using methods such as deliberative polling or information‐and‐choice questionnaires (Luskin et al., 2002; Neijens and De Vreese, 2009; Neijens et al., 1992; Van Knippenberg and Daamen, 1996). The basic idea of these alternatives to mass opinion surveys is that if citizens receive information or enter in a process of deliberation where they encounter different arguments they are able to develop an informed opinion before eventually choosing one of the proposed scenarios. Despite these being good methods for representing public preferences after informing the public, the resulting preferences of both methods do not represent public preferences without such additional information (Luskin et al., 2002; Van Knippenberg and Daamen, 1996). We borrow from this research tradition the use of concrete scenarios that function as alternatives, but we refrain from providing detailed information about the consequences of each of them. This is because we aim to capture the true, albeit potentially uninformed, preferences of the public for future EU scenarios. ² It has to be noted, though, that in the aftermath of Brexit the Dutch media have covered the ensuing discussion on the future of Europe and that we can thus assume that information about possible future trajectories were available to most citizens. ³

Asking survey respondents for their preferred future scenarios goes beyond common past‐oriented or present‐oriented survey measures that ask citizens if membership of the EU is a good or a bad thing, if their country has benefited from the EU, or if they tend to trust the EU. Furthermore, existing opinion measures mostly remain on an abstract level without addressing the real‐world consequences of the attitudes measured. Although they are thus helpful for capturing the existing general mood of the public toward the EU, policymakers gain little information about what they could and should do to accommodate existing attitudes among citizens. Our measurement approach, which asks respondents to rank future scenarios according to their preferences, also goes beyond existing (policy) studies that rely on surveys with more detailed, though separate evaluations on possible (policy) options without forcing people to choose one or some of the alternatives (see Leiserowitz, 2006, in the context of climate change policies).

An important ongoing advance in the EU literature is what one could call the multidimensional turn in EU attitudes. While earlier work on euroscepticism among parties and in public opinion had focused mainly on explaining pro versus anti‐EU positions, scholarly attention has shifted to developing a more fine‐grained conceptualization of EU attitudes (Boomgaarden et al., 2011; Hobolt and Brouard, 2011). Among the first attempts to do so was the distinction between ‘hard’ and ‘soft’ euroscepticism introduced by Taggart and Szczerbiak (2004). Hard euroscepticism ‘implies outright rejection of the entire project of European political and economic integration and opposition to [one's] country joining or remaining members of the EU’, whereas soft euroscepticism involves ‘contingent or qualified opposition to European integration’ (2004, pp. 3–4). This definition implies there is a link between eurosceptical attitudes and their real‐world consequences, yet this link has so far not been made explicit. Some studies explicitly examine public support for specific aspects of the Union's future, for instance for future strengthening of the EU (Boomgaarden et al., 2011) or enlargement of the EU (De Vreese and Boomgaarden, 2006). Others have used indicators such as the desired speed of European integration as a general measure of EU support (Hooghe and Marks, 2005), without addressing its distinctive future‐oriented outlook. A study that directly compares support for different future EU scenarios is – to our best knowledge – thus missing from the EU literature.

One first major goal of our study is to develop concrete scenarios for the future of the EU on the basis of a qualitative approach. We developed a total of eight scenarios covering the whole spectrum of future EU scenarios. Still, our focus was on more sceptical scenarios that could be ordered along a hard‐intermediate‐soft logic. The scenarios developed ranged from cutting all ties to the EU – or even ending the EU altogether – to staying in the EU under the condition of some reform, with several intermediate alternatives. ⁴ The latter include scenarios such as changing the EU's composition (a smaller union) or reducing EU's competencies by focusing on economic aspects only. Some of those intermediate scenarios lean more to the soft and others more to the hard side (borrowing Taggart and Szczerbiak's [2004] terminology) in terms of the degree to which they propose a break with the EU. Yet, whether the proposed scenarios can be categorized into hard, intermediate and soft scenarios and how exactly they are ordered along this dimension (or, possibly, along multiple dimensions) is an empirical question that is analysed hereafter.

1.2. What Are the Explanations for Future EU Preferences?

Our second goal is to examine the underlying reasons for preferring a certain scenario. The scholarly literature generally conceptualizes EU public opinion as symmetrical (with attitudes ranging from complete opposition to the EU to complete support for it) that can be explained as linear effects (that is, higher values on X lead to more opposition or support for it). Though intuitively sound, this conceptualization forces explanations to matter symmetrically. Extrapolating findings of studies that take different dimensions of EU support as their dependent variable (Boomgaarden et al., 2011; Van Elsas et al., 2016), some people may prefer an intermediate future scenario, where the EU is reformed but not completely dissolved, over more extreme hard or soft scenarios – and different variables will explain these preferences. For instance, a citizen who is dissatisfied with the current EU's performance may support moderate reform (a soft scenario) but at the same time oppose a full break with the EU (a hard scenario); lower performance evaluations may thus induce support for some change, yet not for drastic change. Most existing studies, however, have treated EU support as a linear variable and therefore do not permit us to observe such nuances. Such an analysis is only possible when treating the different categories of the dependent variable as nominal, thus allowing explanations to affect only particular types of scenarios (for example, some explanations may matter only for intermediate scenarios).

In a first step, we examine which citizens support a hard, soft or intermediate EU scenario by studying a set of standard socio‐demographic explanations. Following earlier research looking at economic positions (Gabel, 1998; Gabel and Palmer, 1995), a recurring and overarching finding is that so‐called winners of globalization are more positive and losers more negative towards European integration (Hobolt, 2016). Winners are characterized as younger, well educated and working in 'professional' occupations. In contrast, losers are characterized as older, less educated and belonging to the working class or being unemployed (see also Teney et al., 2014). Studies on the Brexit referendum have noted that the dichotomy between winners and losers of the globalization divide was a key driver of the vote (Hobolt, 2016), with EU exit support markedly stronger among globalization losers. We do not know, however, how milder eurosceptic preferences are driven by the winner/loser divide. Winners might just as well prefer reform of the current EU, but given their fundamentally more pro‐EU position we would expect them to support softer reforms, if at all. We thus expect losers to favour hard EU scenarios, and winners to favour soft scenarios.

Another important variable is the political ideology of a person in terms of being more left or right. The well‐known horseshoe model at the party level argues that both parties on the (extreme) left and right rally against Europe, just for very different reasons (Hooghe et al., 2002). Whereas ideologically right‐wing citizens tend to dislike the EU because of cultural opposition, as they fear a loss of national sovereignty and community, left‐wing citizens dislike the EU more in economic terms, as they dislike the neoliberal character of the European project (De Vries and Edwards, 2009). Furthermore, citizens on the left mainly criticize the current functioning of the EU, rather than fundamentally challenging the project of European integration. The latter is more common among radical right‐wing voters (Van Elsas et al., 2016). Based on this, we expect radical right‐wing citizens to prefer hard EU scenarios, and left‐wing citizens to prefer soft scenarios. Adherence to a left or right ideology may also distinguish between intermediate and soft scenarios. As left‐wing citizens dislike economic integration, intermediate scenarios, including a focus on economic links between EU countries, should be preferred less by left‐wing citizens than by right‐wing citizens.

In a second step, we analyse which EU attitudes explain preferences for a hard, soft or intermediate EU scenario. Recent work on the multidimensionality of EU attitudes has shown that, rather than simply liking or disliking the EU as a whole, citizens evaluate it on different aspects – among which its performance, the benefits they derive from it, but also their emotional attachment or identification with the Union (Boomgaarden et al., 2011). All such attitudes and evaluations refer to the EU at present, and are likely to be important grounds for determining one's preference for the future of the EU. Relying on the framework developed by Easton (1975) for (general) political support as well as on the conceptualization of Boomgaarden et al. (2011), we distinguished between more diffuse attitudes (identity and affect) towards the community and the principles of the regime, and more specific attitudes towards the current regime and authorities (performance and utilitarian evaluations). The four EU dimensions we identified have been validated over time as well as cross‐nationally (De Vreese et al., 2019). As a fifth dimension, the authors distinguished support for EU strengthening. We left out this dimension as it has a future outlook and thus overlaps with the scenarios that form our dependent variable.

Diffuse political support is a principled form of support, which in its strongest form refers to a basic attachment to the political community (Norris, 1999). Attitudes such as identification with the EU and negative emotions towards the EU are highly diffuse, and are most likely extended to the EU as a whole, not to specific aspects of it. For instance, when a person has negative emotions towards the EU, it is likely that this is not because of some minor aspect of the EU which this person dislikes, but rather relates to a wholehearted rejection of the European project. The same holds for identity considerations, as citizens with a low attachment to Europe tend to oppose the European integration process (Hooghe and Marks, 2005; McLaren, 2006). We thus expect diffuse EU opposition to lead to preferences for hard EU scenarios.

Finally, the specific attitudes towards utilitarianism and performance of the EU evaluate the current or past functioning or set‐up of the EU without necessarily putting the European project as such into question. Citizens may, for instance, criticize the EU's policy output, or its (lack of) democratic accountability, or they might evaluate the EU's economic performance. In all cases, such criticism does not automatically translate into a fundamental rejection of the EU. Hence, such specific critiques may be related to a wish for minor or moderate reform in certain domains, such as the composition of the EU, its competencies or democratic set‐up, rather than ending EU membership of one's own country'. Consequently, we expect specific EU opposition to lead to preferences for intermediate or soft EU scenarios. Table 1 summarizes our expectations.

Table 1.

Summary of expected preference explanations

Explanatory variable	Characteristic	Scenario preference
Globalization winner/loser	Loser	Hard
	Winner	Soft
Political ideology	Right‐wing	Hard/intermediate
	Left‐wing	Soft
EU attitudes	Diffuse opposition	Hard
	Specific opposition	Intermediate/soft

2. Data and Method

2.1. Data

The analysis is based on variables from two waves from a panel study in the Netherlands (Goldberg et al., 2019). The first wave was collected in September 2017 and the second in January 2018. The original sample was drawn from the TNS NIPO Netherlands database. This actively managed database consists of 124,000 respondents (from 65,000 households) that were recruited through multiple recruitment strategies, including telephone and face‐to‐face recruitment. It matches the population profile well on the dimensions of gender, age, education, region and urbanization. Quotas (on age, gender, region and education) were enforced in sampling from the database. The subsequent survey was conducted using computer assisted web interviewing. Of the original 3,026 respondents who participated in wave one (RR = 71.8 per cent), 2,648 respondents also participated in the second wave (RR = 88.6 per cent). Whereas the main variables of interest stem from the second wave of the panel survey, some of the independent (and control) variables stem from the first wave, which we explain in more detail below. Generally, we did not capitalize on the panel structure of our survey data, that is, we did not model change, but relied on two survey waves because of data availability.

In addition, in September 2016, following the EU Bratislava summit discussing the consequences of the then recent Brexit vote and the future of the EU, we administered a survey as part of a (different) ongoing panel (N = 1,722), probing respondents with an open question to elaborate on what future they would prefer for their country in (or outside) the EU. We used these qualitative data for developing the scenarios in our survey question, as we explain below.

2.2. Operationalization

Our dependent variable measures eight future scenarios of (the Netherlands in or out of) the EU. These scenarios were inspired not only by the public debate after the Brexit vote, but also by the aforementioned open responses about the preferred future of Europe. A manual content analysis ⁵ of these data revealed that a plurality of 20 per cent of the answers favoured increasing national sovereignty at the expense of EU integration, while eight per cent of the answers supported further unification of – and transfer of authority to – the EU. Aside from these general stances, we distinguished preferences for more specific scenarios. We found that eight per cent of respondents wanted a different composition of the Union (statements such as ‘no eastern or Muslim countries’, or ‘two separate unions for northern and southern Europe’). Another eight per cent wanted to restrict the EU to a trade union (‘like the EEC’), and nine per cent felt the EU should be fairer in terms of contributions, advantages and disadvantages (‘our country pays too much’). Around seven per cent preferred a Nexit, or an end of the EU altogether. ⁶

We developed the subsequent eight scenarios to reflect this variety of preferences. The survey question asked respondents to rank these scenarios according to their personal preferences. Specifically, we asked ‘Recently there has been discussion about the future of the Netherlands in or outside the European Union. Considering the various possibilities, which would have your preference? Please make a top three out of the following options’. The eight (randomized) options were:

1. The EU should dissolve into completely independent countries (EUdissolve).

2. The Netherlands should leave the EU and become completely independent (NLexit).

3. The EU should dissolve and the Netherlands should aim for a smaller union with select countries (smallunion).

4. The Netherlands should leave the EU but keep economic ties (that is, a status comparable to Norway and Switzerland) (economy).

5. The Netherlands should use a potential Nexit to enforce special advantages for staying (Nexitthreat).

6. The Netherlands should stay in the EU, but actively try to reform it (reform).

7. Everything should stay as it is (statusquo).

8. The EU should become one country (onecountry).

We designed scenarios that were plausible (regardless of how likely they were); differentiated, that is, covering the whole spectrum of possible futures and as mutually exclusive as possible (cf. Meyer, 2018). As mentioned previously, we expect to find a greater variety of preferences leaning towards less Europe, which are, therefore overrepresented among the scenarios. Scenario 1 is the hardest conceivable scenario: that of ending the EU altogether. Scenario 2, that of cutting all ties to the EU, is slightly less drastic, yet close to the original potential outcome of hard euroscepticism. From there, alternative scenarios inspired by the qualitative data address the EU's composition (3), propose returning to an economic union (4), or to enforce a better deal for one's country (5) following the UK rebate example. Scenario 6 on active EU reform represents the softest form of change, and may also be interpreted in a more positive sense, that is, reform could also mean that there should be more rather than less integration. Finally, we have included two scenarios that cover a preference for the status quo (7) and a more idealist scenario in favour of complete unification (8).

As independent variables we use four socio‐demographic variables and four EU attitudes. As socio‐demographics we included age (simple and squared term), education (recode of International Standard Classification of Education (ISCED11) into three levels) occupation (recode of ISCO=International Standard Classification of Occupations (ISCO) major groups and current employment status into 11 categories) and self‐positioning on the left–right scale (simple and squared term). As EU attitudes, we included four of the five multidimensional EU attitudes by Boomgaarden et al., (2011), that is, negative affect and European identity as diffuse indicators, plus utilitarianism and performance of the EU as specific indicators. All these attitude dimensions consist of three items that are merged and resulted in seven‐point scales each.

As controls for common explanatory models of EU support we include three more variables. For utilitarian considerations (Gabel, 1998; Gabel and Palmer, 1995) we include a variable measuring the economic perceptions of the Netherlands in a retrospective and prospective way, namely, how respondents evaluated the general economic situation (on a combined seven‐point scale). As identity‐based variable we include an anti‐immigration measure comprising five items (on a combined seven‐point scale) (McLaren, 2006; de Vreese and Boomgaarden, 2006). To control for the use of national heuristics we include satisfaction with the government measured with three questions (on a combined seven‐point scale) (Van der Eijk and Franklin, 1996). We further include a control for gender (female dummy). For an overview of the exact operationalizations, including the wordings of questions and items and answer categories for all variables, see Table S1 in the Appendix.

2.3. Multidimensional Unfolding

In a first step, we examine whether our future EU scenarios order along a hard‐intermediate‐soft logic, as theoretically discussed or any other logic. For this analytical step we use multidimensional unfolding (MDU). Unlike Mokken or Rasch scales, which assume a hierarchy (dominance) among items and subjects, unfolding analysis focuses on proximity relations. Multidimensional unfolding analysis creates (graphic) configurations between subjects, that is, respondents, and items, that is, future EU scenarios, relying on pairwise preferences between them (our description of the analysis strongly relies on Busing, 2010). The respondent's point in this configuration represents the so‐called ideal point. Moving away from this point decreases their preference for an item, meaning that the larger the distance between a respondent and an item, the lower the respondent's preference for this item. As a result, a (graphic) solution of unfolding analysis aims to represent the points of respondents and items in the respective configuration, with the smallest distances possible between respondents and their preferred items. In statistical terms, multidimensional unfolding minimizes the sum of the squares of the residuals between the (reported) item preferences and the distances in the unfolding solution.

As the name implies, multidimensional unfolding enables the detection of single dimensions, for example, representing different degrees of change along a hard‐intermediate‐soft scenario logic, but also multiple dimensions. To examine this possibility, we ran one‐dimensional and two‐dimensional solutions to our data (our N of eight scenarios would be too demanding for even more dimensions). In practical terms, we used the procedure of preference scaling (‘prefscal’ in SPSS) with the respondents' first three picks having a value of 1, 2 and 3 and all unselected scenarios receiving a value of 8. As a method we used an ordinal untied transformation to release all unselected scenarios, which ensures that non‐chosen scenarios that all receive the same value of 8 are not considered to belong together. The latter is important so that for a person choosing intermediate scenarios, the non‐chosen hard and soft EU scenarios are not seen as theoretically similar options.

2.4. Regression Models

For our second research goal, the examination of underlying reasons for future EU preferences, we rely on multinomial logistic regression models. ⁷ For these models we first reduce and categorize the future scenarios into groups following the hard‐intermediate‐soft logic, which serve as the dependent variable. This happens on the basis of the findings from the multidimensional unfolding analysis, which allows us to detect not only the order but also the clustering of the categories. The first scenario preference of respondents then determines their respective classification; that is, respondents are classified into the hard, intermediate or soft category based on respondents' first choice scenario. We subsequently run two different model specifications, the first with socio‐demographics only and a second model that adds the four EU attitudes plus the attitudinal controls. To ease interpretation of the multinomial logistic models, we present average marginal effect plots for each of the independent variables of interest.

3. Results

3.1. Citizens' Preferences for Future EU Scenarios

In a first step, we look at the preference distributions across the proposed eight future scenarios in a descriptive way. Figure 1a displays the first choice of respondents (out of three). When looking at all respondents, the huge majority prefer staying in the EU and either keeping it as it is (24 per cent) or reforming it slightly (39 per cent). Another significant proportion of respondents would like to leave the EU, but keep economic ties with it (16 per cent). Less than 10 per cent of respondents each chose one of the remaining scenarios. The more extreme options of leaving the EU (three per cent) and dissolving it altogether (four per cent) on the sceptical side and preferring the EU to be one big country (three per cent) on the positive side are very seldom the primary preference.

Figure 1.

Scenario choice among all respondents a: First Scenario Choice b: Three First Scenario Choices Combined

Looking at the preference distribution of all three chosen scenarios at the same time – independent of the order – relaxes the clear pattern somewhat (Figure 1b). When aggregating the first, second and third preference, the reform (70 per cent) and status quo (59 per cent) options are still the most popular ones and are included in the preferred scenario set for more than half of all respondents. However, as the length of the bars nicely display (though be aware of the different scale of the y‐axis), the relative preference differences to the other scenarios are now much smaller. Still, the clearly least occurring preferred scenario is the Netherlands leaving the EU and becoming independent (13 per cent).

In order to test for an ordering and classification of the eight scenarios, we now turn to the results of our multidimensional unfolding analysis. Figure 2 displays the corresponding configuration of a two‐dimensional solution, which provides a very good model fit (a more detailed explanation is provided in the Appendix). Notwithstanding the importance of two (instead of one) dimensions, the configuration strongly resembles a so‐called 'horseshoe solution', which is commonly found in cases of one‐dimensional solutions with two extremes (see dotted line that we added). One of the extremes is the positive scenario with the EU becoming one country in the top left corner, which is detached from any other scenario. Given that few respondents chose this scenario, we dropped the respective respondents from the following multinomial model. Notwithstanding the empirical proximity of the status quo and reform scenarios, for theoretical reasons we keep both scenarios separate in further analyses. While the status quo may or may not imply some sort of criticism of the EU, the reform scenario clearly indicates the wish for some soft change due to a critique of at least some aspects of the EU. The remaining scenarios on the right form two clusters, so that we combined the intermediate scenarios of an economic union, a smaller union and threatening the EU with the Nexit as well as the two hard scenarios of dissolving the EU and a complete Nexit. For the following explanatory analysis, we continue with the reduced four clusters, as summarized in Table 2. ⁸

Figure 2.

Column Objects of Two‐Dimensional Unfolding Solution

[Correction added on 29 May 2020, after first online publication: Figures 1a and 1b have been updated in this version of the article.]

Table 2.

Clusters of future EU preferences

Cluster	Scenarios
Hard	1. The EU should dissolve into completely independent countries.
	2. The Netherlands should leave the EU and become completely independent.
Medium	3. The EU should dissolve and the Netherlands should aim for a smaller union with select countries.
	4. The Netherlands should leave the EU but keep economic ties.
	5. The Netherlands should use a potential Nexit to enforce special advantages for staying.
Soft	6. The Netherlands should stay in the EU, but actively try to reform it.
Status quo	7. Everything should stay as it is.

3.2. What Explains Differences in Future EU Preferences?

In a second analytical step, we examine the individual determinants of future EU preferences. Following the established categorization in Table 2, our dependent variable in the multinomial logistic regression models comprises four categories: status quo (7), soft (6), medium (3, 4 and 5) and hard EU scenarios (1 and 2). Instead of presenting and discussing all regression coefficients in detail (see all regression results in Tables S2 and S3 in the Appendix), we use marginal effect plots for illustration purposes.

We begin with the effects of being a globalization winner/loser, displayed in Figure 3. Starting with education (a), in line with our expectations we see that citizens with less education prefer hard scenarios and oppose soft ones – which is exactly the opposite for highly educated citizens (both compared with the group with medium levels of education). Medium scenarios are particularly less preferred by citizens with higher education. Interestingly, for maintaining the status quo, we see no educational differences. The plotted age effects (b) are not in line with the winner/loser expectations. Young citizens (that is, winners) indeed seem to be more positive towards the EU, as is shown by their status quo preference. Importantly, however, they are also more likely to vote for hard scenarios than older people, although their overall likelihood of doing so is rather low. Older respondents, unexpectedly, show a clear preference for soft scenarios, compared with the other three options and compared with younger people. Here, we expected them to have much stronger preferences for the hard, or at least intermediate, scenarios.

Figure 3.

Average Marginal Effects and Adjusted Predictions For Socio‐demographics (based on Table S2) (a) Education (ref. category middle).(b) Age; (c) Occupation (ref. category professionals). 95% confidence intervals

The third winner/loser variable – occupation – mostly confirms our expectations. Figure 3c shows that, compared with the reference category (professionals), there are strong preferences for hard EU scenarios among manual and unskilled workers, as well as among the unemployed. Preferences for hard and medium scenarios are, however, hardly distinguishable for these groups. Yet they clearly dislike the soft scenario, which in turn is preferred more by those with higher occupations such as professionals. It is important to note the significant preference for the hard scenarios also among technicians, service workers, freelancers and pensioners. It is thus mainly the professionals, legislators, and clerical workers who stand out by not preferring hard scenarios.

The effects of political ideology, shown in Figure 4, partly match our expectations. We find a subtle U‐curve for hard scenarios, indicating that related preferences are slightly stronger at the extremes, which disconfirms the expectation that mainly right‐wing citizens support hard scenarios. We do, however, find a marked distinction between left and right on the soft and medium scenarios. Left‐wing citizens have a higher likelihood of supporting soft scenarios, while right‐wing citizens wish for the more drastic changes expressed in the medium scenarios. This shows that even the most right‐wing citizen thus prefers keeping some form of tie to the EU over its complete abolition.

Figure 4.

Adjusted Predictions for Left–right Ideology (based on Table S2), 95% confidence intervals.

Figure 5 displays the effects of the EU attitudes based on the full model (see Table S3). Similar to the findings so far, the different dimensions of EU attitudes affect the scenarios in different ways and to varying degrees. To start with the two diffuse dimensions – EU identity and negative affect – we expected that these would primarily lead to preferences for hard EU scenarios. But this is not what we find, as both negative affect and low EU identity mainly lead to support for medium scenarios, while citizens with more positive affect or a stronger sense of EU identity particularly prefer soft scenarios. Yet, these effects are substantively rather small.

Figure 5.

Average Marginal Effects for EU Attitudes (based on Table S3). 95% confidence interval.

In comparison with diffuse attitudes, utilitarianism and performance of the EU display much stronger overall effects. These are partly in line with our expectation that specific EU opposition leads to support for soft or medium scenarios. More negative evaluations of performance indeed lead to soft and (to a lesser extent) medium scenarios, while positive performance evaluations lead to supporting the status quo. Yet for utilitarianism we find that negative evaluations lead to supporting medium or hard scenarios; citizens who do not see benefits in EU membership are more likely to want drastic reforms or reject the EU altogether. Positive utilitarian evaluations are actually more likely to be found among citizens aiming at soft reforms. ⁹ Importantly, although we modelled EU attitudes alongside socio‐demographic factors and interpreted the effects independently, both factors may be more closely linked in the sense that EU attitudes mediate effects of socio‐demographic factors.

We ran two robustness checks that strengthen our findings. In a first robustness check, we used all scenarios separately as dependent variables. The separate effects (Table S4) show highly similar values as in our main models using combined scenarios. This further shows the usefulness of the multidimensional unfolding approach to reduce complexity. In a second robustness check, we additionally included political interest and knowledge as significant predictors of scenario preferences (Figure S1). While highly interested and knowledgeable respondents are especially in favour of soft scenarios (and opposed to the status quo), it is crucial that the effects of our main variables of interest remain unchanged in the respective models (full regression results available upon request).

4. Conclusion

By investigating public preferences for the future of the EU in a novel and encompassing way, our study has yielded several insights. Firstly, the descriptive analysis of respondents' preferences for the eight concrete future EU scenarios demonstrate that such preferences are rather diverse, and that public opinion is by no means drawn towards the more extreme options of full integration or full disintegration of the EU. A full‐blown Dutch EU exit is actually the least preferred scenario. Secondly, the multidimensional unfolding analysis of the preferences for the eight scenarios demonstrates there is a meaningful structure behind the responses, as scenarios are empirically ordered in a soft‐intermediate‐hard logic. This indicates that citizens hold fairly grounded opinions about the future of the EU, rather than being indifferent or holding non‐attitudes. Thirdly, the explanatory models of these different types of scenarios have shown the merit of a discrete distinction between scenario types, as most explanatory factors stand out in distinguishing one or two scenarios from the others.

The findings show that preferences for the future of the EU are driven by common factors found in the EU literature, but in some cases differently than expected. We broadly find support for the expectation that globalization losers in particular support hard future scenarios (that is, a break with or dissolution of the EU), which is in line with the findings of Hobolt (2016) on the Brexit vote. Citizens in lower occupations such as manual and unskilled workers and the unemployed have the strongest preferences for hard scenarios, and the same holds for citizens with little education, although the effect is relatively small. However, several of these groups also strongly prefer intermediate scenarios, which imply large reforms but not a rejection of the EU. As winners of globalization, highly educated citizens as well as those in professional occupations are the typical proponents of soft scenarios (that is, staying in the EU but pursuing reforms). As for age, the results are less straightforward and are not in line with the winner/loser argument. Whereas younger citizens have the strongest preferences for maintaining the status quo, older citizens stand out by strongly preferring a soft scenario. There is little age difference in preferences for intermediate and hard scenarios. Finally, we expected right‐wing citizens to prefer harder scenarios than left‐wing citizens. While this is true for soft and intermediate scenarios, with left‐wing citizens having a strong preference for only minor changes and (far) right‐wing citizens opting for medium scenarios with more drastic changes, we found them to have an equally strong preference for hard scenarios, although this was low in absolute terms.

In a second step, we explained future EU preferences by looking at evaluations of the current EU, by distinguishing common EU attitude dimensions. The main finding here is that specific evaluations of the EU's utility and performance have particularly strong effects on what kind of EU future citizens prefer, yet in different ways. Negative performance evaluations trigger a desire for reforms, but do not lead to a wish to leave or dissolve the EU, exactly as theorized. Negative utilitarian evaluations, in contrast, lead to supporting intermediate as well as hard scenarios – if the EU is not perceived as beneficial, citizens might thus give up on it altogether. Different kinds of specific EU evaluations can thus have different consequences for the future outlook on the EU. A second important finding is that diffuse attitudes of identity and negative affect have much smaller effects on future EU preferences than specific attitudes. And, unexpectedly, these diffuse attitudes do not explain support for hard EU scenarios, but mostly distinguish between the soft and intermediate scenarios. Thus, such fundamental attitudes may have less extreme implications for the EU's future trajectory than one might expect, as they do not necessarily lead to an outright rejection of the EU.

Our study contributes to the literature on EU public opinion in various ways. Firstly, its approach is novel, compared with existing studies, as it casts a future‐looking perspective on EU public opinion, which so far has been mostly implicitly present in the literature (for example, in the work by Taggart and Szczerbiak (2004) in the context of euroscepticism). To date, most studies on EU public opinion have relied on present‐oriented or past‐oriented measures. Secondly, the results of our study show that future EU preferences may be less extreme than what is suggested by binary or simplified questions on EU membership. While confirming several common findings, such as that there are generally more negative EU preferences among globalization losers (Gabel, 1998; Gabel and Palmer, 1995; Hobolt, 2016) or right‐wing voters (De Vries and Edwards, 2009; Van Elsas et al., 2016), we show that such negative preferences are not necessarily at the extreme of the spectrum and often result in preferences for profound reform rather than a full‐blown rejection of the EU. The same holds for the influence of multidimensional EU attitudes (Boomgaarden et al., 2011; Hobolt and Brouard, 2011), as our study shows that negative evaluations of the EU can lead to support for soft or intermediate scenarios rather than hard scenarios.

A third contribution of our study is that it sheds light on a question that has become all the more urgent given the Brexit referendum: what do European citizens actually want for the future of Europe? Abstract survey measures (such as the EU membership question) or binary referendum questions leave politicians and policymakers in the dark as to what future European citizens envision for the EU. Our study gives a highly concrete image of preferences among the Dutch public, in the hopes of informing the future debate on which road the EU should take to cater to public preferences. If anything, the results show that these preferences are highly varied and well‐structured, lending confidence to the possibility of a public debate on Europe's future.

This study should be considered as a first attempt to investigate public preferences toward the EU's future. As a first attempt, it has its limitations. Firstly, notwithstanding the new insights gained by our scenario‐ranking approach, this method has the potential to delve even further into public EU opinions. For instance, one could develop different and more scenarios than the ones tested here, although to keep it manageable for survey respondents, a restricted number of scenario options is advised. Secondly, following the discussions around Brexit, our current analysis has mainly focused on negative scenarios with regard to EU integration, but similar concrete scenarios could also be developed for more positive views of the integration process, distinguishing different trajectories of further integration. Finally, as usual with single‐country studies, the results are limited regarding their generalizability. However, this limitation mainly holds for potential differences in the absolute levels and answer patterns of citizens' future EU preferences, and less so for the examined underlying explanations, which are rather universal in nature and do not depend on any Dutch‐specific conditions. Still, similar analyses should be conducted in a large set of member states, which would not only help to reduce the information gap about citizens' future EU preferences, but also to confirm the explanations from the Dutch context in a more general setting.

Supporting information

Appendix S1. Online appendix

Table S1: Overview of variables and their operationalizations

Table S2: Multinomial logistic regression results (socio‐demographics model)

Table S3: Multinomial logistic regression results (EU attitude model)

Table S4: Multinomial logistic regression results (all scenarios separately)

Figure S1: Average marginal effects for political interest and political knowledge

Goldberg, A. C. , van Elsas, E. J. , and De Vreese, C. H. (2021) Eurovisions: An Exploration and Explanation of Public Preferences for Future EU Scenarios. JCMS: Journal of Common Market Studies, 59: 222–241. 10.1111/jcms.13057.