Voter Preferences for EU Asylum Policies: The Role of Government Cues

Abstract

We study whether and how governments influence public opinion about immigration policies in Europe. At the European level, conflicts about policy are generally territorial in nature – that is, they involve conflicts between member states, which are represented by their governments. Distinguishing between four types of situations, depending on whether the national governments support or oppose EU policy proposals, we formulate and test hypotheses concerning the positions of incumbent and opposition voters/non-voters on four different asylum-policy proposals in 16 European countries. We test both direct effects of incumbent cues on voters' preferences, and moderating effects, where the cueing affects the way in which individual attitudes to immigration and European integration translate into specific preferences for EU asylum and immigration policies. Our results suggest that voters, indeed, follow the cues provided by their governments when forming their preferences on EU policies.

The relationship between public opinion and government policy is a two-way street. On the one hand, public opinion imposes constraints on governments. The dynamic model of representation (e.g. Stimson 2004; Stimson et al. 1995), for instance, posits that government policy typically responds to public opinion. On the other hand, public opinion also responds to government policy, as is argued, for example, by the thermostatic model of representation (e.g. Soroka and Wlezien 2010; Wlezien and Soroka 2012). Marco Steenbergen et al. (2007), who have studied this relationship in the context of European integration, find evidence for such reciprocal causation whereby party elites both respond to and shape the views of their supporters.

In this article, we take a more one-sided view and examine how governments influence public opinion about public policies. Specifically, we analyse the impact of government policy positions on the public opinion about asylum policy in the EU. Drawing on rich original survey data collected in the summer of 2021, we focus on four key EU policies in the asylum domain that have received scant attention so far in the well-developed literature on voters' preferences for immigration and asylum: (1) the relocation of asylum seekers in the EU; (2) the Dublin Regulation, which assigns the asylum seekers to the country of first entry; (3) the reinforcement of the EU's external borders through the creation of a strong border and coastguard; and (4) the externalization of the accommodation of asylum seekers to third countries such as Turkey and Libya.

These policies have all been of crucial importance in the EU's attempt to stem the tide of asylum seekers flowing into the EU during the refugee crisis of 2015–2016. In this crisis, it was up to the EU to come to the rescue of the hardest-hit member states. It did so with very unequal success: its relocation policy failed dismally, and it has not been possible to revise the dysfunctional Dublin Regulation right up to the present time. On the other hand, external border control was successfully reinforced and the EU–Turkey agreement, the main example of externalization, proved to be highly successful in stemming the inflow of new refugees (Kriesi et al. 2024). In spite of the importance of voters' views, the existing research provides limited insights into views about these concrete policies and their determinants.

We address this question in two ways. First, we provide novel and detailed data on public opinion in a large number of European countries, going beyond existing research that focuses either on general immigration or asylum-policy preferences (e.g. Hangartner et al. 2019; Nordø and Ivarsflaten 2022) or single specific policies (e.g. Bansak et al. 2017; Heizmann and Ziller 2020 on relocation). Second, we examine how voters' preferences are shaped (directly and indirectly, as we detail below) by the national political elites.

Our key assumption, on which we elaborate below, is that there is typically high complexity and low information available on EU issues and policies. Consequently, policymaking and policy outcomes represent crucial informational cues for citizens' opinion formation. National governments have thus considerable leeway in shaping domestic debates and opinions. This is not to say that the national governments do not generally respond to public opinion. Instead, we argue that in the particular policy areas we examine in this article, there is considerable room for national elites to frame conflict, define the contours of national interest and shape voters' opinions.

In support of this assumption, we turn to two strands of research: (1) the general research on the nature and origins of public opinion; and (2) research on EU support. First, long-standing scholarship on public opinion suggests that information from political elites influences voters' opinions and beliefs on a variety of issues (e.g. Zaller 1992), including European integration (e.g. Gabel and Scheve 2007; Steenbergen et al. 2007) or immigration (e.g. Harteveld et al. 2017; Hellwig and Kweon 2016). Particularly on EU issues and policies, elite cues serve as a heuristic to compensate for typically high complexity and low information availability (Pannico 2020). Elite cues therefore matter for citizens' preferences regarding EU immigration and asylum policies, and the extent to which national elites adopt consensual or polarized policy positions should affect the degree to which voter constituencies converge or diverge in their policy preferences (Zaller 1992).

However, while cue-taking is generally theorized to occur along partisan lines (Bisgaard and Slothuus 2018; Brader and Tucker 2012; Druckman et al. 2013; Zaller 1992; see also Harteveld et al. 2017; Hellwig and Kweon 2016), political conflict on European integration and EU issues occurs along territorial lines, too (Marks 2004). As Gary Marks (2004: 246) pointed out, it does so because, among other things, European integration ‘engages national governments in a process of ongoing bargaining over a range of issues that formerly were determined within, rather than among, national states’. National governments play a key role in managing the linkages between the EU and its member states, and their role is likely to become particularly visible and salient in moments of crisis, like the eurozone crisis or the 2015–2016 refugee crisis, paving the way for an increased capacity to shape public opinion (Ares et al. 2017). As national governments ‘compete by representing distinct territorial communities’ (Marks 2004: 251), they are in a privileged position to shape the public discourse around the contours of national interest and the distributive consequences of different EU policy options.

Therefore, we argue that cues from national governments are likely to impact voters' preferences for EU asylum and immigration policies. In particular, we expect the cues from national governments to affect the policy preferences of the incumbent voters directly and more strongly, in line with research revealing a perceptual bias underpinning incumbent evaluations and responsibility attributions (e.g. Hobolt et al. 2013). As a consequence, we should see a strong alignment between the national governments' policy positions and the incumbent voters' policy preferences. Voters of opposition parties should be less affected by cues from the national government, particularly when the opposition and the incumbent are deeply divided. Importantly, in addition to a direct effect, we also expect a moderating effect of elite cues. More specifically, the capacity of cues from national governments to affect individual policy preferences is likely to depend also on pre-existing individual attitudes on immigration and European integration. Cues consistent with pre-existing attitudes are not only more effective in shifting policy preferences in the direction of the cues,¹ but cues generally can also serve two additional purposes. First, they can provide new information for citizens to better connect their pre-existing beliefs with specific policy options. Experimental evidence suggests that voters are responsive to information they receive from their national governments, as opposed, for instance, to EU-sourced information (Hobolt et al. 2013). Second, they can legitimize prior beliefs, which is particularly relevant when these are socially sanctioned, like radical-right views (e.g. Bischof and Wagner 2019; Jost 2019). As a result, in addition to direct effects, cues should also affect the way in which attitudes to issues translate into specific policy preferences.

Our article is structured in the following way. The theory section develops our theoretical argument and details our specific hypotheses. The next section introduces the configuration of national political elites in EU asylum policy. The third section is devoted to a brief description of the design of the study. Then we turn to the presentation of the results and conclude with a short summary and discussion of our findings.

Theory: the role of elites in shaping public opinion on public policies

In general terms, as John Zaller (1992) observed some time ago, the issue-specific configurations of the political elites are highly consequential for the opinion formation of the general public: if elites are in agreement, the most sophisticated voters who receive the most political information tend to accept this information and follow the consensual elites. If, however, political elites are polarized, as in the case of some of the key policies pursued at the EU level during the refugee crisis, then public opinion polarizes as well, with the most sophisticated citizens polarizing the most. James Druckman et al. (2013) and Martin Bisgaard and Rune Slothuus (2018) have confirmed that when elites polarize, citizens follow and polarize as well, both in the US and in Europe. More specifically, studies of the effects of party cues on support for the EU have demonstrated that information received from parties strongly influences citizens' opinions about the EU (Gabel and Scheve 2007; Maier et al. 2012; Steenbergen et al. 2007; Stoeckel and Kuhn 2018). Given the low availability of information about the EU and the high complexity of the debated issues at the EU level, parties are largely able to shape citizens' attitudes about the EU (Pannico 2020). Party cues also increase the public's competence on the EU. These effects are stronger than those of cues on domestic politics (Torcal et al. 2018).

In these studies, cue-taking occurs along partisan lines. However, at the EU level, polarization also occurs along territorial lines (Marks 2004). This is to suggest that we need to focus on cues not only from parties, but in particular from national government representatives who link national politics to EU policymaking. Indeed, there are a number of studies which show that support of the EU is conditioned by support for national institutions – that is, there is a spillover effect from the domestic to the EU level with respect to institutional support (Armingeon and Ceka 2014; Harteveld et al. 2013; Hobolt 2012). One study in particular has shown that these spillover effects rely not only on partisan cues but also on cues from national governments (Ares et al. 2017). The latter are especially relevant in critical moments, such as the eurozone crisis, when the role of the national government in the integration process becomes exceptionally salient. As a consequence of their cue-taking from national governments, we expect citizens to polarize along territorial lines.

There is a debate in the literature as to the actual extent of cue-taking on migration-related issues among citizens. On the one hand, a number of studies relying on panel data from several European countries find that individual attitudes to immigration tend to be rather stable (Kustov et al. 2021; Lancaster 2022; see also Dennison and Geddes 2019 for a similar argument). This should make them resilient to external stimuli (Druckman et al. 2012). Thus, Mathias Mader and Harald Schoen (2019), studying the particularly interesting case of Germany during the refugee crisis 2015–2016, found no evidence of cue-taking among voters.² Based on panel survey data, the authors show that the refugee crisis did indeed trigger substantial intra-individual change in partisan support, but, instead of following party cues, voters responded to the flow of events. Increasing attention on immigration triggered voters to change party support in line with their attitudes on immigration – that is, previous Christian Democratic Union (CDU) supporters abandoned the party in favour of the immigration-critical Alternative for Germany (AfD).

At the same time, there is scholarly research suggesting that voters do take cues from the parties they support. For instance, Timothy Hellwig and Yesola Kweon (2016) studied the effect of elite cues on voters' attitudes about migration, based on cross-sectional European Social Survey (ESS) data and panel data from Denmark. They point out that immigration is fundamentally a multidimensional issue, which means that public opinion about immigration ought to be susceptible to issue framing by trusted political elites. More specifically, they consider that the more educated should be especially influenced by such cues. They find that, indeed, individuals are more likely to oppose (support) immigration when their preferred party adopts restrictive (accommodating) positions on immigration and that the corresponding effects are stronger among the highly educated (see also Vrânceanu and Lachat 2021). While (repeated) cross-sectional studies may have a hard time arguing that they tap into a cueing process (see e.g. Hellwig and Kweon 2016: 721), the fact that panel data have corroborated cueing effects is reassuring. Along this line, Eelco Harteveld et al.'s (2017) analyses of panel data from the Netherlands and Sweden confirm voters' tendency to take cues from the political parties they support (particularly those parties adopting radical policy positions). Similarly, the effect of the 2015–2016 refugee crisis on the Norwegian public's opinion about immigration and refugees might also be attributed to cueing by political elites precisely because of its limited duration (Nordø and Ivarsflaten 2022). Moreover, the party cueing effect among partisans has also been confirmed in experimental settings in the US (Druckman et al. 2013) as well as Europe (Brader and Tucker 2012; Brader et al. 2013).

We do not aim to settle this debate here. We argue instead that, even if parties may have a hard time influencing immigration attitudes, policy-specific opinions are more likely to be malleable and subject to framing effects, especially if the policies in question are policies decided at the EU level. In fact, Mader and Schoen's study did not involve policymaking, but national electoral politics. Note also that Hellwig and Kweon's argument about the immigration issues' multidimensionality bears notable implications for policy preferences. Thus, Sabina Avdagic and Lee Savage (2021) showed, based on survey experiments in Germany, Sweden and the UK, that negative framing of immigration has a strong and pervasive effect on welfare support (although the effect of positive framing is shown to be considerably weaker).

With regard to a given EU policy in the context of territorial conflicts at the EU level, we can distinguish between four types of situations, depending on: (a) whether the government of the member state supports the policy or not, and (b) whether national elites are divided or not with regard to immigration policy in general. Assuming that voters take cues from their preferred party, incumbent voters will take different cues from opposition voters, if the national elites are divided. By contrast, both incumbent and opposition voters will take the same cues – either both positive or both negative – if the national elites are consensual (see Table 1). In the consensual situation, where both government and opposition support the policy, government voters may, however, still be more supportive of the EU policy than opposition voters, because the government represents the country's position in the EU. This implies that the government's position is likely to be more visible for the public. These considerations can be summarized by three hypotheses (H1a–H1c).

Hypothesis 1a: If elites are divided and the government strongly opposes EU policy, incumbent voters are more opposed to EU policy than opposition voters.

Table 1. Four Theoretical Elite Configurations at National Level

Hypothesis 1b: If elites are divided and the government strongly supports EU policy, incumbent voters are more supportive of EU policy than opposition voters.

Hypothesis 1c: If elites are consensual and both government and opposition support EU policy, the gap between the positions of incumbent and opposition voters should be smaller, but incumbent voters may still be more supportive of EU policy than opposition voters, because it is the government which officially represents the position of the member state in the EU.

Note that we formulate the hypotheses in relative terms, because elite cues are not the only factors determining public opinion with regard to the policy in question. Even if the national elites are divided on the policy, for other reasons the overall preferences of the voters with respect to the specific EU policy may be skewed either in the government's or the opposition's direction. There are other factors at both the national and the individual level which may influence individuals' attitudes about EU policies.

At the individual level, Boris Heizmann and Conrad Ziller (2020), relying on Eurobarometer data collected in September 2015,³ find that anti-immigrant attitudes are negatively related to support for better redistribution of asylum seekers in all countries. We would argue that this finding applies to any policy which facilitates the accommodation and integration of asylum seekers. In fact, anti-immigration attitudes can be expected to be positively associated with policies which render access to asylum more difficult. Interestingly, the effect of anti-immigration sentiments is smaller in top asylum-seeker receiving countries. Heizmann and Ziller (2020) suggest that humanitarian concerns might be responsible for this result. Instead, we propose that the government's position may affect this tendency – that is, we expect government cues also to moderate the effect of immigration attitudes.

The mechanism behind this expected effect is linked to source credibility and system justification. On the one hand, voters tend to take their cues from credible sources (Druckman 2001), and their governments constitute a particularly credible source, as is shown by experimental evidence which suggests that voters are responsive to information they receive from their national governments, as opposed for instance to EU-sourced information (Hobolt et al. 2013). This is likely to apply even more so for voters for the incumbent party. On the other hand, cues can legitimize prior beliefs, which is particularly relevant when these beliefs are socially sanctioned, such as the beliefs of the radical right (e.g. Bischof and Wagner 2019). In our context, this means that cues in favour of restrictive policies from the national government should embolden citizens with anti-immigration views to express support for this type of policy explicitly, whereas cues in the opposite direction would cast doubt on the social acceptability of anti-immigration views.

System justification theory also suggests that governmental cues can legitimize prior beliefs. It argues that people are motivated (to varying degrees, depending on dispositional and situational factors) to defend and justify the status quo of existing social, economic and political systems, sometimes even at the expense of individual and collective self-interest (Jost 2019). Since the acting government stands for the status quo, we would argue that the motivation to justify the status quo also implies the readiness to take cues from the acting government. More specifically, this motivation ought to be particularly pronounced among conservative or right-wing voters, as studies all over the world reveal that system justification is almost always associated with the endorsement of conservative or right-wing ideologies (Jost 2019: 286). Accordingly, in member states where the government opposes the redistribution of asylum seekers, or for that matter any kind of EU asylum policy, the opponents of immigration are expected to feel justified to oppose the measure without restraint, and vice versa if the government supports an EU policy. The effect should, however, be more important for opponents of immigration (who are typically also endorsing conservative or right-wing ideologies), given the asymmetric tendency to resort to system justification.

Heizmann and Ziller (2020) also show that Euroscepticism is negatively associated with favourability towards redistribution of asylum seekers among all EU member states, though some nuances seem to be at play. We also expect Eurosceptic voters to be less favourable towards EU asylum policies, regardless of the specific content of the policy. More importantly, however, we also expect government cues to moderate the effect of Euroscepticism on policy preferences; that is, we expect this effect to be more substantial if the government opposes the policy in question. If the national elites are consensual, the individuals' attitude towards EU integration is likely to make little difference. If the national government opposes the policy, however, Eurosceptics have an additional reason to oppose EU policy. In summary, we expect two additional moderation effects:

Hypothesis 2a: The effect of anti-immigration attitudes on policy preferences is stronger when the government opposes a policy than when it supports an EU policy.

Hypothesis 2b: The effect of Eurosceptic attitudes on policy preferences is stronger when the government opposes a policy than when it supports an EU policy.

The baseline, however, is that Europeans do not want to accommodate additional asylum seekers or that they want to do so only under limiting conditions. Thus, Kirk Bansak et al. (2017) study the acceptance of additional asylum seekers across 15 European countries in February–March 2016, at the height of the refugee crisis. They find that not one single country has a majority of the population willing to accept more asylum seekers. Once this reluctance to accept more asylum seekers is established, they test with conjoint survey experiments the conditions under which acceptance might be improved: they find that a large majority of citizens prefer an allocation mechanism that foresees a redistribution proportional to the country's capacity (defined in terms of population, GDP, past applications and unemployment), as opposed to the current Dublin Regulation, which places emphasis on the country of first entry. This relative support is maintained even when respondents are made aware that a proportional allocation mechanism would imply an increase in the share of asylum seekers assigned to their country (except in three countries: the Czech Republic, Poland and the United Kingdom, which shift to being more supportive of the status quo in relative terms). In other words, the current first-entry rule is highly unpopular across Europe.

Anne-Marie Jeannet et al. (2021), who also rely on conjoint experiments to examine European citizens' preferences for refugee and asylum policies in eight European countries, suggest that citizens impose limits and conditions on their support of protective measures for asylum seekers (e.g. by setting annual limits to asylum applications; providing at most limited/low resettlement; or allowing family reunification only if the refugee can pay for their cost of living). Importantly, citizens (except in Italy) are more supportive of national decision-making on asylum applications. Finally, Alina Vrânceanu et al. (2023) rely on conjoint experiments to understand what preferences citizens in two European countries considered in their study (namely Germany and Greece) and in Turkey have for cross-country cooperation between European countries and Turkey on migration management and refugee protection, building on the 2016 EU–Turkey agreement – the key policy to address the refugee crisis of 2015–2016. They find that public preferences are driven by a mix of humanitarian and instrumental concerns, as well as reciprocity and burden-sharing considerations. Interestingly, citizens in the two European countries appear to be supportive of several features of the status quo (i.e. the existing 2016 EU–Turkey deal).

The configuration of the national political elites in EU asylum policy

Among the four policies we study in this article, two are highly contested among the member states, while two are more consensual. The two highly contested policies refer to the redistribution of asylum seekers among the member states, and the assignment of the responsibility for asylum seekers to the member state of first entry. Compared to these two policies, the enhancement of external border controls and externalization have been much more consensual among the policymakers. For the analysis of the individual attitudes towards these four policies, we propose to classify the member states into five types.⁴ This classification is based on the two criteria which we introduced for the theoretical configurations (see Table 1): the polarization between incumbent and opposition parties (polarized vs undivided) with regard to immigration issues in general, and the policy positions of the respective governments with regard to the two contested policies – the relocation of asylum seekers in the EU and the first-entry rule – in the context of the 2015–2016 refugee crisis and its aftermath.

Table 2 presents an overview of the classifications – one each for relocation (a) and first entry (b). Relocation policy is supported by the Mediterranean frontier states, by the destination states and by the UK, which is treated as a separate category, since it has not been part of Schengen or of the Common European Asylum System (CEAS) and participated in relocation on a purely voluntary basis.⁵ This policy is opposed by the Visegrád Group (V4) coalition (represented by Hungary and Poland), which aggressively attacks EU asylum policy and constitutes the category of ‘adversaries’, as well as by Austria and Latvia, which operate in a less openly hostile manner. The first-entry principle is also supported by the UK and the destination states, but the frontline states (Italy, Greece and Spain) are their direct opponents on this policy, while the V4 countries are rather indifferent in this respect. The frontline states push for EU policy reform, especially of the principle that the country of first entry should be responsible for the asylum seekers, which they want to see abolished. In their push for reform, they are partly supported by the destination states, which agree with their demand for burden sharing, but insist that the frontline states should continue to be responsible for managing the asylum requests. In addition to the four theoretical types, empirically we also need to introduce a residual category for states which do not take an explicit position on the policy in question and which we call bystander states.

Table 2. Classification of the Member States

Notes: The numbers in brackets indicate the mean positions for the respective government's actions in the policymaking process on relocation at the EU level. +1 means full support, –1 means full opposition. Countries with less than three actions are coded as ‘no position’. For more details, see Kriesi et al. (2024).

The polarization between incumbent and opposition parties on immigration issues, our indicator for divided elites, is based on 2017–2019 data from the Observatory for Political Conflict and Democracy (PolDem) electoral campaigns data,⁶ 2019 data from the Chapel Hill Expert Survey and the most recent data from the Comparative Manifesto Project. These sources do not always agree with each other (see Table A1 in the Supplementary Material). We classify a country as polarized if at least two of the three indicators show relatively high polarization between incumbent and opposition parties with respect to immigration issues. Interestingly, the party systems of the frontline states and the V4 states are all polarized, as is the party system of the UK, which was still an EU member state at the time of the crisis.

The policy positions of the member-state governments on relocation are based on the Political Process (PPA) dataset on the refugee crisis (see Bojar et al. 2023; Kriesi et al. 2024). This dataset allows us to construct quantitative indicators for the governments' positions in the relocation policy debates, to the extent that they took explicit positions which were reported in the international press. The corresponding positions on first entry (the Dublin Regulation reform) are based on qualitative information about the actions taken by the respective governments in these debates. Policy positions may be subject to change if a new government replaces an old one. In the asylum-policy domain, however, we observe a surprising continuity with regard to the fundamentals across governments since the refugee crisis of 2015–2016. In five countries, the government composition has not changed between 2015 and summer 2021 (Hungary, Poland, Germany, Sweden and the UK). In another five countries, it has changed partially – that is, some parties were always part of the government coalitions (Austria, Latvia, the Netherlands, Ireland and Italy). In the remaining six countries, the government changed between centre left and centre right (France, Finland, Romania, Portugal, Greece and Spain). In four of these countries, the elites are undivided on immigration policy, while in two of them – Greece and Spain – the elites are, indeed, polarized in this regard. However, with respect to the most contested policies at the EU level, the centre left and the centre right in both of these countries share a common national interest, too.

With respect to the relocation scheme, Hungary and Poland, as members of the V4 coalition, which also included the Czech Republic and Slovakia, aggressively opposed any such scheme.⁷ Hungary and Slovakia went as far as to appeal to the European Court of Justice (ECJ) against the September 2015 decision of the Council of Ministers to introduce relocation quotas, an appeal that ended up being rejected by the ECJ on 3 October 2016. In addition, the Hungarian government organized a referendum over the relocation issue, a referendum which eventually failed to reach the quorum, not least because the national opposition had called for a boycott of the referendum. In other words, the national elite in Hungary was deeply divided on the issue. Poland, too, refused any kind of mandatory quota early on in spring 2015. In September 2015, the EU twisted the arms of its outgoing Civic Platform (PO) government, which ended up accepting the relocation scheme, but the incoming Law and Justice (PiS) government strongly opposed any such scheme and refused to accept any refugees. Subsequently, the PO, now in opposition, was less certain about how to respond to the issue and took ambivalent positions (Szczerbiak 2017). Still, it opposed the government's outright anti-European stance.

In addition to Hungary and Poland, among our countries, Austria and Latvia also rejected relocation schemes. In March 2018, the new Austrian prime minister, Sebastian Kurz, hid behind the eastern European countries by claiming that there was no point in discussing the relocation quotas, since the eastern countries would not accept them anyway. He pleaded for an alternative system with harsh border controls and returns to countries of origin. As for Latvia, although it rejected the relocation schemes, it was not part of the militant V4 coalition and hardly intervened in the EU-level policymaking process at all.

On the other side of the relocation issue, there are the frontline states – Italy, Greece and Spain – which pleaded for such a scheme in order to allow them to share the burden of the inflow of asylum seekers over the Mediterranean. These frontline states openly criticized the V4 countries for their resistance to the scheme. Italian Prime Minister Matteo Renzi went as far as to suggest that eastern European countries which refused to help frontline states in Europe's refugee crisis should have their EU funding cut. The destination states – Germany, Sweden, the Netherlands, France and the UK – supported such schemes in principle. Thus, Germany had already thrown its weight behind a relocation system in April 2015, before the crisis started in earnest, and so did Sweden. Later on, in December 2018, Germany dropped its demands that all EU countries should accept refugees during a crisis in a bid to break the deadlock over efforts to overhaul the CEAS. It put its hopes on what came to be called ‘flexible solidarity’: the possibility for a country to make financial contributions instead of taking in refugees. France, in turn, had always taken a more ambivalent position with regard to relocation. In May 2015, Prime Minister Manuel Valls (under President François Hollande) rejected quotas. President Hollande prevaricated by rejecting quotas but agreeing with the need to distribute asylum seekers across Europe. In September 2015, France would support a push for mandatory quotas⁸ but remained somewhat ambivalent subsequently.

The line-up of the member states is different with respect to the reform of the existing CEAS, which is based on the Dublin Regulation that stipulates the first-entry principle. As we have already mentioned, the cornerstone of the latter is that the state where the asylum seekers first entered the EU ought to be responsible for accommodating them. The V4 coalition also opposed reforming the system, in particular it refused any reforms including relocation schemes. With regard to first entry, however, it did not take an explicit position. Austria, however, defended the first-entry rule. During the refugee crisis, Austria had strongly insisted on this rule and had provoked several incidents with its neighbours over this rule, such as the Brenner incident with Italy.⁹ By contrast, the frontline states continued to demand a reform of the system which would no longer make them responsible for incoming asylum seekers. This is a point which has been made by a series of Italian governments since 2015. But on this point, destination states oppose the frontline states: while they support the need for reform and are in favour of relocation mechanisms, they prefer to stick with the responsibility of frontline states for the incoming asylum seekers. Thus, at the EU summit on 28–29 June 2018, the new Italian prime minister, Giuseppe Conte, clashed with German Chancellor Angela Merkel on this issue. He was ready to understand the asylum issue as one concerning the whole of Europe, but he refused to accept that the obligation to rescue people at sea implied the obligation to treat their asylum requests in the name of all of Europe. At this summit, Merkel could not get what she wanted to defend her national asylum policy against domestic challengers. France remained ambivalent on this reform as well, whereas the other countries remained in the background in this respect.

In terms of additional country-level factors, we would expect that the support for EU policies depends on whether the country benefits from such policies, or whether it suffers from additional costs imposed by them. In line with this expectation, Heizmann and Ziller's (2020) results suggest that, on average, people who live in a country with high pressure from the problem – that is, with a higher share of asylum seekers – support a better EU-wide redistribution of asylum seekers, whereas those who live in countries with low immigrant stocks (as measured in 2010) and less generous integration policies are less likely to support such a mechanism. Even if we do not make problem pressure a criterion for our typology, the resulting categorization in Table 2 closely mirrors categories previously proposed in the literature which reflect, among others, problem pressure – categories such as frontline, destination and bystander states (Kriesi et al. 2021). This is not surprising since problem pressure correlates with the positions of governments on the contested policies which we focus on.

Design of the study

For the individual-level data we rely on an original online survey put into the field in 16 EU member states (see Table 2) in June 2021. For each country, quota samples of at least 2,000 respondents have been interviewed. This survey included the following question about asylum policy:

In recent years, many refugees came to the EU. By refugees, we mean people who come to the EU on the grounds that they fear persecution in their own country. To what extent do you agree or disagree with the following statements about refugee movements? Please, select a value from 0 to 10, where 0 means ‘completely disagree’ and 10 means ‘completely agree’.

This introductory question was followed up by six statements about EU asylum policy, four of which we use for this study:

• • Each EU country should be required to accommodate a share of refugees (relocation)

• • Refugees should be accommodated by the country through which they first entered Europe and in which they were first registered (first entry)

• • The EU should invest in reinforcing external borders by building a strong border and coastguard (external borders)

• • The EU should collectively pursue deals with third countries (such as Turkey and Libya) via financial and other incentives to crack down on trafficking networks and reduce the flow of asylum seekers across the Mediterranean (externalization).

To explain public support for these four policy statements, we use our typology of member states (see Table 2) and three key indicators at the individual level. The first individual-level indicator distinguishes between incumbent and opposition voters based on the respondents' vote choice in the last national elections (previous to June 2021). Non-voters are included among the opposition voters. The second indicator refers to immigration attitudes. For this indicator, we use eight statements about immigrants, which we have taken from the ESS. The first four statements deal with the consequences of immigration for the economy (good or bad), for cultural life (undermined or enriched), for problems of crime (made worse or made better) and for the country at large (make it a worse or a better place to live). The other four statements refer to different groups of people who should be allowed or not to come and live in one's country: people of the same/different race or ethnic groups, and from poorer countries inside/outside Europe. The eight statements form a strong factor (EV = 4.3). For the analysis, we have dichotomized this factor at the mean, distinguishing simply between pro- and anti-immigration positions. The third indicator refers to trust in the EU, measured on an 11-point scale. We also dichotomized this indicator at the mean for the analysis, distinguishing again simply between high and low trust positions. All three indicators are interacted with the country typology to measure the moderation effects of elite cues.

In addition, we control for problem pressure (low, medium, high), self-placement on the left–right dimension (recoded into three categories – left, right, neither), as well as for the party family which one voted for (recoded into eight categories – radical left, greens, social democrats, liberals, conservatives/Christian democrats, radical right, others and non-voters). These controls allow us to obtain the impact of the government's position independently of the government's partisan composition – that is, we control for coalition cabinets. We also control for political interest (ranging from 1 (not at all interested) to 4 (very interested)), level of education (low, medium, high), age and gender. Table A2 in the Supplementary Material presents the summary descriptives of all the variables included in the analysis.

To test our hypotheses, we estimated ordinary least squares (OLS) regressions with standard errors clustered by country, separately for each policy. Specifically, to test H1a–H1c, we include interaction terms between the dummy indicator measuring support for incumbency and the indicator variables for the EU member-state type (frontline, destination states and so on), in addition to the constitutive terms. By evaluating the policy preferences of incumbent and opposition supporters, we are able to see whether these map onto the configurations of elites' positions in the corresponding EU member-state types. To test H2a–H2b, we include in our models the variable measuring respondents' immigration attitudes (trust in EU) and its interaction with the indicator variables for the EU member-state type.

Results

Figure 1 presents the shares of voters who support the four policies, by type of member state. The classification of member states for relocation is slightly different from that for first entry (see Table 2). For the consensual policies we apply the classification for first entry. For this presentation, we have recoded the 11-point scales to 0 (0–5: opposition) and 1 (6–10: support). As the figure shows, the consensual policies are equally supported by roughly two-thirds of the voters in each type of member state. By contrast, the support of the contested policies varies considerably from one type of member state to the other. Relocation is supported by majorities in countries where incumbents support the policy (destination states, frontline states and the UK) or do not take a clear position (Portugal, Ireland and Finland), but it is clearly rejected by the voters in countries where incumbents oppose the policy (Hungary and Poland, Austria and Latvia). The first-entry principle fares even worse among the voters of all types of countries except those of the UK, who are not concerned by it. This principle is rejected by two-thirds of the voters in countries where incumbents oppose the policy (frontline states). It is also heavily rejected in Hungary and Poland (where the government is not explicitly opposed to it) and still rejected, even if to a less pronounced extent, in the destination and undivided bystander states. These support patterns correspond more or less to what we would have expected, given the corresponding patterns among the political elites, except for the fact that the first-entry principle finds relatively little support even among types of states which would benefit from it.

Figure 1. Public Support for EU Asylum Policies, Separately by Policy and Member-State Type

We now turn to the question of whether voters take cues from their respective elites; that is, whether incumbent voters differ from opposition voters and non-voters with regard to their support of the various policies. Figure 2 provides the answers. This figure and the following figures are based on the regressions presented in Table A3 in the Supplementary Material. Part (a) of the figure shows the predicted probabilities of support – that is, the average preference for the four policies by incumbent voters and opposition/non-voters in the various member-state types. To facilitate the comparison, part (b) presents the difference in the predictive margins between incumbent and opposition voters together with the confidence intervals. The vertical line in part (a) indicates the sample mean. In line with our hypothesis H1a, for relocation we find the strongest effect for the two adversarial member states, Hungary and Poland. In these two countries, where the governments strongly oppose relocation policy and the government and opposition are divided, both incumbent and opposition/non-voters are opposed to relocation schemes. However, the opposition of the incumbent voters is considerably more pronounced (with a difference of −1.8 on the 0–10 scale). This occurs in spite of the fact that we control for party choice, left–right self-placement, immigration attitudes and trust in the EU. This effect supports H1a. Most likely, it is due to the cues the respective governments have provided to their followers in a very aggressive way over a period of several consecutive years since the 2015–2016 refugee crisis (there has been no change in the government since then). For voters in Austria and Latvia, where the government also opposes relocation, but the elites are not divided, we do not find a similar effect. Here, we see that voters tend to support relocation schemes less than the average voter in Europe, but we do not see significant differences between incumbent and opposition/non-voters. In Austria, grand coalition governments in the past are likely to have contributed to the blurring of the relevant cues. In the bystander countries, where the average position corresponds to the overall average in the pooled sample, incumbent and opposition/non-voters do not differ from each other either, as a result of the reduced salience of the issue and a more consensual stance among their elites.

Figure 2. Incumbent and Opposition/Non-Voters' Policy Support, by Policy and Member-State Type: Top (a) Predicted Probabilities of Support. Bottom (b) Differences in the Predictive Margins of Incumbent (inc) and Opposition (opp) Voters

Notes: The labels in this graph refer to the classification of member states in Table 2. For relocation: against–divided = adversarial states (HUN, POL); against–undivided = Austria, Latvia; bystander = POR, IRE, FIN; pro-undivided = destination states (GER, FRA, NEL, SWE, ROM); pro-divided = frontline states (GRE, ITA, ESP). For first entry: against–divided = frontline states; bystander–divided = HUN, POL; bystander–undivided = POR, IRE, LAT, FIN, ROM, FRA; pro-undivided = destination states (NEL, SWE, GER, AUT).

On the other hand, in line with H1b/H1c, with the exception of the UK, in the countries where the governments have taken a supportive stance with respect to relocation schemes, voters are generally likely to support them and incumbent voters are more likely to support them than opposition/non-voters. The differences are not that large (0.4 and 0.6 respectively in the destination and frontline states), but they are highly significant. Even if the elites have been generally supportive of relocation schemes in these countries, the government's position has been more salient than the opposition's position. We find only very small and statistically not significant differences in the UK.

Concerning first entry and the other two policies, we find the most pronounced differences between incumbent and opposition/non-voters in the UK. In line with H1b, British incumbent voters support these policies more strongly than opposition voters. There are no corresponding effects in the frontline states, where the government opposes the first-entry rule. Opposition voters in these countries were just as opposed to this policy. Partially in line with H1c, however, in (divided and undivided) bystander states, incumbent voters were slightly, but significantly, more supportive of first-entry policies than opposition voters. The lack of any effect among frontline and destination states in the case of first entry may seem surprising. It is likely a result of the fact that, in these states, there is a consensus among all the parties that the first-entry rule is either unacceptable (in frontline states), or that one should stick to it (in destination states). For the consensual policies, the differences between incumbent and opposition voters are, with the exception of the UK, generally very small, even if some of them are significant. As Table A3 in the Supplementary Material indicates, our model is best able to explain support for relocation, the most contested policy (R ² = 0.28), while it performs rather poorly for first entry and externalization (R ² = 0.08 and 0.06 respectively), with intermediate results for external border control (R ² = 0.18). The poor performance for first entry and externalization may be due to the low visibility of the policies in some member states.

Next, we turn to the moderation effects of the government's position on the association between immigration attitudes and support for EU policies. We only show these results for the contested policies, for which cue-taking is most likely (see Figure A2 for the remaining policies). Figure 3 presents the results in two ways. First, using the same presentation mode as in the previous figure, it shows the policy support of citizens with pro- and anti-immigration positions. To test H2a, we are mainly interested in seeing whether anti-immigration voters are in fact more likely to oppose a given policy when the government opposes that same policy. It again adds a second subgraph that shows the difference in the predictive margins between pro- and anti-immigration positions. The purpose of this subgraph is to facilitate the assessment of the magnitude and the statistical significance of these differences. As these graphs show, the differences between the pro- and anti-immigration constituencies are especially stark with respect to relocation policies in all countries. But the effects vary between country types. In line with expectations for both the relocation and first-entry policy, anti-immigration voters are more opposed to the policy when their government has spoken out against such schemes or does not explicitly support them (bystander states) than when their government supports these policies. This supports H2a. There is only one exception to this generally supportive pattern, for relocation: in countries where government and opposition jointly support the policy, the moderating effect of immigration attitudes is as strong as it is in countries with governments opposing the policy or taking no positions. Figure A1 in the Supplementary Material shows that there are no such differences in the case of the two consensual policies.

Figure 3. Effect of Immigration Attitude on Policy Support, by Policy and Member-State Type

Instead of comparing the effects of anti- and pro-immigration positions within countries, we can also compare the effects of anti-immigration attitudes across countries. If we do so (see results in Figure A2 in the Supplementary Material), we find that, indeed, anti-immigration attitudes have stronger effects if the government opposes the policy than when it supports the policy. The effects are also typically stronger in divided countries.

We have also tested the moderation effects for trust in the EU. Figure 4 presents the results in a similar way to Figure 3. The pattern of effects for the relocation policy is very similar to that for immigration attitudes. In other words, as expected, the effect of trust in the EU on the policy assessment is stronger for member states opposed to EU policy, which confirms hypothesis H2b. This also applies to the first-entry policy: the effect of trust in the EU is strongest for frontline states, which are strongly opposed to this rule. However, with respect to first entry, the effect of EU trust in undivided bystander states is not significantly different from the effect in frontline states. Comparing across countries leads to the same results as we already presented in Figure A2 in the Supplementary Material, since these are the effects for voters with anti-immigration positions who also do not trust the EU.

Figure 4. Effect of Trust in the EU on Policy Support, by Policy and Member-State Type

Conclusion

Our results provide support for our hypotheses in the case of the relocation policy and first-entry policy, the two contested EU asylum policies. They are weaker and less systematic for the consensual policies, even if they are overall in line with expectations for these policies, too. We found that, when their governments support EU policy, voters are generally likely to support them and, if the elites are divided, incumbent voters are especially likely to follow their governments to a greater extent than opposition voters or non-voters. By contrast, when elites are divided and governments of member states oppose EU policy, incumbent voters oppose the corresponding policy to a greater extent than opposition and non-voters. Moreover, the effect of two key attitudes related to EU asylum policy, attitudes towards immigration and towards the EU, is reinforced among incumbent voters compared to opposition and non-voters, when their government opposes an EU policy.

Given that our analyses control for the political orientations and the immigration attitudes of the public, these results provide evidence for both cue-taking and moderation effects of government cues in the two-level EU polity. In their preferences for EU policies, especially in the case of highly contested policies such as the relocation schemes, voters, indeed, seem to take their cues from their governments. When their governments oppose the EU policy and elites are divided, as was the case in Hungary and Poland with regard to the relocation schemes, not only do incumbent voters adopt their government's position, but opponents of immigration and Eurosceptics more generally feel justified to oppose EU policy. Given the importance of the territorial channel of representation in the EU polity (Bickerton et al. 2015; Fabbrini 2017; Puetter 2014), this mechanism is likely to be of crucial importance for the success of EU policies. When acting at EU level, member-state governments not only respond to their public opinion (Hagemann et al. 2017), but they are able to shape it to a considerable extent. The impact of their cues is likely to be of particular importance in EU policymaking, where the remoteness of this process from the experience of ordinary voters opens a large manoeuvring space for ‘interpretative’ effects of government policy positions (Pierson 1993: 619–624).

Our study is not without limitations, an important one being the difficulty of disentangling the causal direction (bottom-up vs top-down) of the influences. To enhance the credibility of our claim that top-down elite cues are effectively at play, we leverage the temporal ordering, as elite positioning and polarization are measured earlier than voters' preferences (fieldwork in 2021). Yet, we acknowledge that this strategy is imperfect and encourage future research that relies on panel data or experimental designs, to further probe the extent to which elites can contribute to shaping citizens' specific EU policy preferences. Our study is also limited in its ability to specify to what extent the national media play a role in this process, either by shaping elites or voters, or by mirroring elite discourses and debates on the specific policies we focus on here (Marquart et al. 2018; Vliegenthart and Mena Montes 2014). This is an important avenue for future research.