2.1 Public Opinion and the EU Polity Formation

Since 1992, the politicization of European polity formation has brought public opinion into the picture of European politics (Hooghe and Marks Reference Hooghe and Marks2009). The 1992 referendum failure in Denmark, which rejected the Maastricht Treaty, and the reluctant ‘little yes’ uttered by French voters marked the end of the permissive consensus and elite-driven European polity formation. More than ever after a decade-plus of crises, the EU relies on different types of support from electorates. Polity building – as embodied by the many reforms and capacity building at the centre – needs to be fully supported by voters to be sustainable in the long run and not exploited by euro-skeptic party actors. We, thus, argue that mapping out potential conflicts – whether functional or territorial (Caramani Reference Caramani2015) – is vital for understanding where political frictions can appear and how they will influence the future of both European polity formation and the Russian invasion of Ukraine (since Ukraine relies on its European allies for crucial help).

While the argument of external threats inducing polity centralization that stems from the state-building literature (Riker Reference Riker1964; Hintze Reference Hintze and Gilbert1975; Tilly Reference Tilly1975; Kelemen and McNamara Reference Kelemen and Kathleen2021) has been chiefly focused on the supply side of politics (policymakers) and does not have much to say about public demand for polity centralization, we argue that the demand side is an important link in the chain going from the external threat to polity centralization (Truchlewski, Oana, and Moise Reference Zbigniew, Oana and Moise2023).

First, the Hintze–Riker–Tilly thesis was developed to explain state formation in the Middle Ages and the Renaissance when elites operated without much popular constrain and when military technology required economies of scale that needed to go beyond the feudal structure (Cederman et al. Reference Cederman, Galano Toro, Girardin and Schvitz2023). Modern democratic nation-states need to consider public opinion, as it may constrain or enable elite action. We know that a strong dissensus among the European public can constrain further political integration (Hooghe and Marks Reference Hooghe and Marks2009). Conversely, a strong consensus allows greater room of manoeuvre for politicians to steer the shape of the EU polity. At the Member State level, domestic policymakers are aware of the electoral consequences of their decisions and attempt to satisfy public opinion at home when making decisions on the EU stage. At the EU level, European policymakers have an interest in polity maintenance and, hence, avoiding backlashes from domestic audiences that could threaten the polity and bring about divisions that would undermine common decision-making (Ferrera, Miró, and Ronchi Reference Ferrera, Miró and Ronchi2021; Ferrera, Kriesi, and Schelkle Reference Ferrera, Kriesi and Schelkle2024).

Furthermore, we note that public opinion is more likely to exert pressure on politicians during times of high salience when voters follow what is happening and have more well-formed preferences. The present moment is, therefore, an opportunity to probe into demand-side dynamics at a time when the public is particularly attuned to the Russian invasion of Ukraine and the response of the EU. Footnote ² We started our panel survey at the onset of the invasion in 2022, a moment of very high salience for the Russian war in Ukraine and the economic and political response of the EU. Thus, while several of the policy domains associated with the invasion (foreign policy, energy, etc.) are usually considered too complex for individuals and of low salience, our timing allows us to examine them in a situation when the public is aware and engaged in discussions surrounding the implications of these policies. Indeed, several elections, such as those in HungaryFootnote ³ and Slovakia,Footnote ⁴ showed that policies concerning the war were crucial for electoral success. All in all, we argue that in case of an external threat, when highly salient policies take centre stage in the public sphere, consensus on the demand side becomes crucial for policymaking.

Second, and more generally, following the polity approach to the European Union (Bartolini Reference Bartolini2005; Ferrera, Kriesi, and Schelkle Reference Ferrera, Kriesi and Schelkle2024) that draws on the Hirschman-Rokkan model of state-building (Hirschman Reference Hirschman1970; Rokkan et al. Reference Rokkan, Flora, Kuhnle and Urwin1999), we look at bonding as one of the main elements characterizing a polity (alongside bounding–borders and binding–authority/capacity).Footnote ⁵ Bonding refers to the loyalty and solidarity that members of a polity have towards the polity itself and towards other members, a loyalty that ultimately constitutes the political community (Oana and Truchlewski Reference Oana and Truchlewski2024). Investigating how thin or volatile such bonding is in times of crises on the demand side is important as it speaks not only to the amount of resistance there is in the EU against pooling resources and decisions in particular policy domains but also to the potential for triggering more foundational conflicts over the raison d’être of the polity itself (Ferrera, Kriesi, and Schelkle Reference Ferrera, Kriesi and Schelkle2024).

Nevertheless, we do not yet have a good understanding of how the demand side of politics is affected by external threats and by the Russian invasion of Ukraine in particular. The evidence so far shows that similarly to the COVID pandemic (Altiparmakis et al. Reference Altiparmakis, Brouard, Foucault, Kriesi and Nadeau2021; Beetsma, Burgoon, and Nicoli Reference Beetsma, Burgoon and Nicoli2023; Bremer et al. Reference Bremer, Kuhn, Meijers and Nicoli2023), the Russian invasion of Ukraine triggered a rally-round-the-flag moment among European voters, who became more supportive of leaders and policies following the start of the war (Steiner et al. Reference Steiner, Berlinschi and Farvaque2023; Truchlewski, Oana, and Moise Reference Zbigniew, Oana and Moise2023; Nicoli et al. Reference Nicoli, van der Duin and Beetsma2024). Importantly, this implies that EU elites have greater room for manoeuvring in order to pursue certain forms of polity building. Conversely, if far-right parties are not able to capitalize on possible dissensions, they may prove unable to create political momentum against policies and polity formation, as postfunctionalism would predict (Hooghe and Marks Reference Hooghe and Marks2009). However, as we further argue in the next section, such a rally-round-the-flag is likely to vary across policy domains. Furthermore, support might be short-lived and give way to more dissensus as the crisis progresses. By focusing on the demand side of politics, our manuscript aims to shed light on these developments.

2.2 Beyond Integration: Conceptualizing Polity Support across Policy Domains

We argued that analyzing the demand side of EU politics in the aftermath of the Russian invasion of Ukraine is important given the war’s public salience and given public opinion’s crucial role in enabling or constraining policymaking. While this Element is definitely not the first to focus on public opinion in the EU in times of crises (De Vries Reference Vries, Catherine, Vries and Catherine2018; Kriesi et al. Reference Kriesi, Altiparmakis, Bojar and Oana2024; Truchlewski et al. Reference Zbigniew, Oana, Moise and Kriesi2025), it does start from two consequential theoretical assumptions inspired by the polity approach to the European Union (Caramani Reference Caramani2015; Ferrera, Kriesi, and Schelkle Reference Ferrera, Kriesi and Schelkle2024) that bring about unique analytical gains and novelty. First, rather than considering public support for the EU as uni-dimensional – more or less integration, we conceive of such support as playing out in two dimensions stemming from a distinction between ‘policy’ and ‘polity’ support. Second, we argue that crises are not monolithic threats but that they play out in different policy domains, and support for types of EU polity can vary across these domains. In this section, we elaborate on both of these arguments.

2.3 The Drivers of Support for EU Polity Types

The literature on public support for the European polity has already theorized its different potential drivers. Our aim in this manuscript is to bring them under the same comparative umbrella and analyse their relationship with our proposed measure of EU polity support across policy domains. In line with Caramani Reference Caramani2015 (but also more recent contributions to the literature on EU support or solidarity, Kriesi, Moise, and Oana Reference Hanspeter, Moise and Oana2024; Oana and Truchlewski Reference Oana and Truchlewski2024; Truchlewski, Oana, and Natili Reference Truchlewski, Oana and Natili2024) which points out to the importance of two types of cleavages when it comes to the EU: territorial cleavages between the Member States and functional cleavages which are transnational and cut across territorial lines, we also look at both types of factors throughout our analyses. With regards to territorial cleavages, we examine divisions in public support across Member States and how these vary across policy domains. In what regards functional divisions we examine the extent to which preexisting and deep-seated predispositions and attitudes such as ideology are related to EU polity support, but also look at the effect of crisis or policy-specific factors on such support. We focus on two such crisis or policy-specific factors which are rarely discussed under the same umbrella: output legitimacy (Scharpf Reference Scharpf1999; Jones Reference Jones2009; Schmidt Reference Schmidt2013), and security logics (Kelemen and McNamara Reference Kelemen and Kathleen2022; Natili and Visconti Reference Natili and Visconti2023; Ferrera and Schelkle Reference Ferrera and Schelkle2024) which have been put forward by various strands of literature on European polity formation. First, our framework integrating both territorial and functional divides speaks to the extent to which territoriality is dissipated within the EU across policy domain or replaced by cleavages cutting across these territories (Caramani Reference Caramani2015) as well as to the ways in which territorial and (post-)functional constraints could be re-enhanced, overcame, or bypassed altogether. Second, while the literature on European integration has classically been focused on internal drivers of polity formation, more recently scholars started paying attention to the mechanism of external threats influencing the EU (Kelemen and McNamara Reference Kelemen and Kathleen2021). Our framework aims to bring together and compare the impact of both internal and external drivers of demand for different polity types.

2.4 Design of the Study

2.4.1 Data

We further our empirical goals by using original public opinion data collected within the ERC Synergy project SOLID at three-time points (March, July, and December 2022) after the Russian invasion of Ukraine, forming an original three-wave panel in five countries: Germany, France, Italy, Hungary, and Poland. The second wave of our panel also included respondents from Finland and Portugal. The selection was these countries was guided by the idea of obtaining a wide amount of country heterogeneity in terms of reliance on Russian gas, political discourse related to sanctions, centrality in the EU, and geopolitical location that allows us to map preferences in a wide range of contexts. As further detailed in what follows, when comparing our data with other data sources with a wider geographical coverage on particular items of interest, the selected countries are fairly representative of wider European trends. Interviews were administered on national samples obtained using a quota design based on gender, age, macro-area of residence (NUTS-1), and education. Our total sample size is approximately 33,000 observations, while our panel respondents include 6,000 individuals surveyed over all three waves.

The first wave was carried out between 11 March and 5 April 2022, two weeks after the start of the war. This period captures attitudes at the very start of the war when the conflict dominated media channels across Europe, and thus the initial rally-around-the-flag effect (Truchlewski, Oana, and Moise Reference Zbigniew, Oana and Moise2023). EU member states came together in an unprecedented manner to form a common front against Russian aggression, applying sanctions, receiving refugees, and providing support to Ukraine. Attitudes from this period also tell us whether the EU may have missed a critical juncture for policy and polity change if such attitudes did not last. The second wave of our panel was conducted between 8 and 28 July 2022. Five months into the war saw a large decrease in salience, as other topics, including inflation and a looming energy crisis, took centre stage. At the same time, the conflict appeared in stale-mate after the spring, when the Ukrainians pushed Russian forces out from the capital, and the fighting concentrated on the South and East. The third wave was administered between 14 December 2022 and 4 January 2023. In between the second and third waves, the Ukrainian army made significant gains in the fall of 2022, showing the importance of Western-provided weapons and support. The timing of our third wave captured a period of calm in the conflict, while EU decision-making was focused primarily on energy policy and rising inflation.

While we focus on seven key member states over a roughly ten-month time-period following the start of the war, our findings speak to broader cross-country and over-time trends. Figure 1 shows how support for an integrated EU army fluctuates from 2018 to 2024 in fourteen EU countries.Footnote ⁶ What can be seen is that the war is a clear critical juncture, drastically shifting support in most countries starting in 2022, as already noted in the literature (Genschel Reference Genschel2022; Truchlewski, Oana, and Moise Reference Zbigniew, Oana and Moise2023). We see three types of countries. In France, Germany, Poland, and the Netherlands, we see stable support across time. In Denmark, Finland, Lithuania, Spain, Sweden, and the United Kingdom, we see strongly increased support in 2022, following the invasion, with some reversion in 2023–2024, but overall higher support. Lastly, in Italy, Hungary, and Romania, we see a decrease in support following the Russian invasion. Our sample of seven includes countries from each category, allowing us to study these dynamics in depth. Our analysis confirms the pattern of increasing initial support, followed by a slight reversion and then stability in most countries.Footnote ⁷ Importantly, our more detailed original panel survey allows us to show how these trends differ across policy domains. Secondly, our original panel data allows us to go more in depth in terms of the significance of these time trends. What does it mean that a higher or a lower share of respondents support an EU army? We combine this data with views on EU integration and unpack the main drivers behind these shifting dynamics.

Figure 1

Support for EU army across time and EU countries

Panel data has the unique advantage of tracking individuals over time in order to see how their attitudes shift in response to changing conditions and the continuation of the conflict. It also allows more in-depth causal exploration of the interplay between ideology, output legitimacy, and security conditions. In addition to our panel structure, our series of surveys also included several experiments, which aimed to shed further causal light on questions surrounding the support for policies and the European polity.

Preliminary analyses using part of the data have been published (Moise, Dennison, and Kriesi Reference Moise, Dennison and Kriesi2023; Truchlewski, Oana, and Moise Reference Zbigniew, Oana and Moise2023; Wang and Moise Reference Chendi and Moise2023) and inform our current theoretical framework and empirical goals. Our preliminary findings suggest neither a complete endorsement nor a complete rejection of the ‘bellicist’ logic. Instead of focusing on the question of ‘whether the war resulted in polity-building’ we focus on what types of polity-building are likely, given what the demand side of politics can support. Much remains to be explored, including the temporal dynamics across the three waves.

3.1 Descriptives: General Preferences for Polity Types

Figure 3 shows an overwhelming support for pooling and sharing resources in the refugee domain. The top plot shows that close to 40 per cent of respondents would prefer a centralized polity. The size of this group is slightly larger than in the energy and foreign policy domains, and considerably larger than that in the defence domain. What sets the refugee domain apart is that the second biggest group in our data is formed by those wanting a pooled polity who want to share the hosting of asylum-seekers in spite of not having a high loyalty to the polity (24.44 per cent). While in the other policy domains the two groups on the off-diagonal with consistent polity–policy attitudes (centralized and decentralized polity) were the larger categories, the high support for Ukrainian refugees in the current crisis makes preferences for pooling trump considerations of loyalty with most respondents located in the upper cells of Figure 3.

Figure 3

Polity versus policy attitudes on refugee policy

3.2 Territorial: Preferences for Polity Types by Country

When looking at preferences for polity types across the countries in our sample in Figure 4, we can see some notable divisions. To begin with, the group preferring a centralized polity is the largest in five out of the seven member states we consider: Germany, Finland, Italy, Portugal, and Poland. In most of these countries this group is followed by the pooled polity group, with the exception of Portugal in which the reinsurance group is the second biggest. France also closely follows the pattern of these five countries, with the centralized polity and the pooled polity groups being the largest two. Hungary stands out as the only member state among the ones analysed in which respondents preferring a decentralized polity constitutes the largest group. This group is followed by the group wanting a reinsurance polity and the group wanting a centralized polity. Strikingly, the group preferring a pooled polity obtains virtually no support in Hungary suggesting that among those respondents who are not loyal to the EU polity very few would prefer to pool resources and share the hosting of refugees.

Figure 4

Preferences for polity types by country

Thus, our expectations are only partly borne out. We indeed see high support for pooling in destination states: Poland, Germany, and Finland. However, we also note surprisingly high support in Portugal and Italy and remarkably low support in Hungary, despite it being a frontline state which saw a large influx of refugees in the beginning of the war. What might explain Hungarian preferences is the relatively low final number of refugees that stayed, and the anti-EU and anti-Ukraine stances of Hungary’s ruling party (Madlovics and Magyar Reference Madlovics, Magyar, Madlovics and Magyar2023).

3.3 Functional: The Effect of Individual Attitudes on Polity Types

Figure 5 shows the effects of satisfaction with refugee aid to Ukraine, support for such aid, and general immigration attitudes on support for polity types in the refugee domain. As expected, all these variables have a similar direction of effects. All three attitudes are strongly positively related to support for a centralized polity, with effect sizes varying between a 50 percentage points increase in the probability of support in the case of pro-immigration attitudes, to a 28 percentage points increase in the case of the refugee aid variable (when moving from end-to-end of the scale in each independent variable). They are also negatively related to support for a decentralized polity but with smaller effect sizes. The only difference is that while satisfaction with the EU’s performance is also negatively related to support for a pooled polity, the support for aid in general and the general pro-immigration attitudes have little to no effects on the probability to support this polity type. This is unsurprising given the large support for pooling observed in our sample.

Figure 5

Performance evaluations and support – refugees

Generally, we note that satisfaction with the way the national government handled refugee aid to Ukraine has quite different effects from these three variables. It is positively related to support for a pooled polity and negatively related to support for a centralized polity. This is important and in stark difference to our other three policy domains as we shall see in the following empirical sections. National satisfaction in the foreign policy, energy, and military domains was mostly negatively related to support for a centralized polity, an effect that we also observe here. However, in other policy domains national satisfaction increased preferences for a decentralized polity, while here it mainly increases the probability of supporting a pooled polity. In other words, in contrast to the other policy domains, satisfaction with national refugee policy does not imply opposition to organizing such a policy at the EU level.

In Figure 6, we move onto the effects of ideational and security factors (ideology, threat, trust, and economic vulnerability). In what regards ideology, the more to the right a respondent is the less likely they want a centralized polity and the more likely they want a decentralized one. Different to the other polity domains, we can see here that there is a small but significant effect on preferences for a pooled polity, preferences which nevertheless remain high along the whole ideological scale, and higher than preferences for a decentralized polity. In other words, even those on the very right would rather want a pooled polity, but followed closely by the decentralized group. Overall, when compared to other policy areas, we conclude that ideology has a weaker effect for refugee policy most likely due to the fact that attitudes on immigration (one of the other predictors in the model) are strongly correlated to ideological self-placement. Trust in Ukraine has the opposite effect to left-right ideology: it is positively related to preferences for a centralized polity and negatively related to preferences for a decentralized or a pooled one. Finally, threat and economic vulnerability have null to small effects net of the ideational and output legitimacy factors included in the model.Footnote ¹²

Figure 6

Ideology, threat, trust – refugees

3.4 Dynamics over Time

We now move on to the dynamic aspect of our panel data, namely observing how respondents change preferences between waves of our survey. Tables 2 and 3 present the raw data. The four columns in each table show the percentage of the members of each respective polity type that either maintain their preference (the diagonal, in bold) or change to other polity types. The final column shows the number of individuals initially in each group in wave 2. Several important patterns emerge. Between waves 1 and 2, similar to the other policy fields, the centralized and decentralized categories are the most stable, followed by the pooled category which is also much more stable in the refugee field compared to the other fields studied. Looking at changes between waves 2 and 3, the pooled category remains even more stable than the decentralized category. Between waves 1 and 2, 68 per cent of respondents stay in the centralized category, while between waves 2 and 3 this proportion increased to 77 per cent. By contrast, the share of respondents staying in the decentralized category decreases from 69 to 58 per cent, whereas in the pooled category it increases from 54 to 59 per cent. The reinsurance type of polity has 33–34 per cent of respondents staying between waves while the pooled polity has slightly below 60 per cent.

Next, the dynamics of these changes are important. We note that we see both horizontal and vertical changes in our original two-by-two. That is to say that we observe movement on both the polity and the policy dimensions of our typology. There are important flows going from the centralized polity type to and from both the pooled and reinsurance types, of roughly equal size. In other words, those who favour a centralized polity change both in their polity attitudes (switching in and out of pooled) and their policy attitudes (switching in and out of reinsurance). This is surprising, given how deeply entrenched we usually consider attitudes towards the EU polity to be. Crisis times put these attitudes in flux, allowing greater shifts in opinion, and it is the case here in the aftermath of the Russian invasion of Ukraine. We likewise note both types of shifts in and out of the decentralized group. Between the first two waves, the decentralized group gains in size from both the pooled category and the reinsurance category. By contrast, it loses support mainly to the pooled one. This implies that the dynamics of changing support differ for those who are loyal to the EU polity and those who are not. For the latter group, sharing resources may be a more fruitful first step of persuasion.

Finally, Figure 7 shows the dynamics of change in the form of a Sankey plot. Each column corresponds to each wave in our survey. The nodes represent the four groups, while the flows between them show where each group switches to and from. The figure suggests two additional patterns: a fading of the rally-round-the-flag effect and a bounce-back effect. On one hand, the literature would in general suggest that the initial measurement that we have in our panel is biased because it would already take into account the initial rally-round-the-flag: of course, respondents would want a more centralized polity in the aftermath of the Russian invasion of Ukraine. But what matters is whether this effect persists with other findings suggesting a fading over time (Truchlewski, Oana, and Moise Reference Zbigniew, Oana and Moise2023). The switches from waves 1 to 2 confirm such a fading effect. However, the pattern of change from waves 2 to 3 also suggests a bounce-back effect, with the polity categories gaining the most being the pooled and the centralized ones. There is thus evidence that over-time respondents consider that the European polity has to step in to solve a common problem, but this is not necessarily done at the cost of national sovereignty through centralization at the supranational level: the pooled polity works as a safety valve where anti-Europeans might take refuge and still preserve a modicum of sovereignty.

Beyond the descriptive analysis of change, we also ask what are the factors that determine such change with a particular focus on output legitimacy and security factors (ideational and other factors are controlled for and their effects can be explored in the Appendix, Section 2). For reasons of simplicity, we focus on the centralized group and ask which factors explain whether respondents remain in the centralized group or switch to other categories. While the full models, available in the Appendix, include both levels and changes in predictors, for ease of interpretation we focus here only on change. Thus, the predicted probabilities in Figure 8 can be interpreted as the probability to remain in the centralized group or switch to other groups, for varying levels of change in the predictors.

Figure 7

Refugee policy change and stability across waves^a

^a The number indicates the raw group size in that particular wave.

Figure 8

Analysis of change – refugees

While the security literature suggests that external threats and economic vulnerability should lead to increasing preferences for polity centralization, Figure 8 finds little evidence for such effects in the refugee policy field net of ideational and output legitimacy factors. What we find is evidence for endogenous polity formation: the more satisfied with the EU refugee policy respondents are, the more prone they are to stick to the centralized polity and the less likely they are to opt for other types of polities. The same applies to the national polity: if respondents are satisfied with their national governments’ refugee policy, they are still counter-intuitively prone to support a centralized European polity. This stands in stark contrast to the effects we observe for other policy fields where satisfaction with the national policy leads to a lower likelihood of staying in the centralized category. This difference in policy domains can be reflective of a common underlying positive attitude towards immigration (on top of the one captured by the variable on immigration attitudes).

3.5 Conclusion

Our evidence suggests that refugee policy has the highest share of individuals preferring a centralized polity, followed by a pooled polity (further integration in refugee policy without broader political integration). This stands in stark contrast with other policy fields which show a greater degree of dissensus. Also in contrast with other policy fields, we see fewer territorial divisions, as only Hungary does not have a plurality of respondents that are in favour of a centralized polity. Our analysis of the individual drivers of support shows that a higher satisfaction with the EU’s handling of refugees is strongly related to preferring a centralized polity, while satisfaction with national refugee policy increases the probability of being in the pooled category, as opposed to the decentralized one as observed in other policy fields. Overall, satisfaction and ideological variables have a strong impact in support for polity types, while security concerns, including threat and economic vulnerability, play a much more limited role.

4.1 Descriptives: General Preferences for Polity Types

Descriptively, in terms of the share of the four groups among our sample, we notice in Figure 9 that a plurality of the respondents prefers a centralized polity (39.5 per cent), and more so than in foreign and military policies, and only slightly less than in refugee policy. Likewise, 21.4 per cent of respondents opt for a pooled polity. In other words, almost 61 per cent of respondents voice a strong preference for one form or another of pooling and/or sharing of resources and decisions. These possible coalitions between respondents who prefer a centralized polity and pooled polity is the second biggest in the realm of energy policy (after refugee policy at 64.2 per cent, with foreign policy at 58 per cent and defence policy at 51.8 per cent following).

Figure 9

Polity versus policy attitudes on energy policy

By contrast, there are only 15.1 per cent of respondents who opt for the reinsurance polity in energy policy. Given that this type of respondent is loyal to the EU polity but wants resources and decisions to remain at the national level in this policy domain, they could be easily swayed to move into a pooled or a centralized polity, should the need arise (e.g., crises intensity increasing). If we factor in preferences for the reinsurance polity, the coalition that could support some form of further polity building in the realm of energy policy broadens to 76 per cent of respondents. Taking all the three polity preferences together, it seems that the postfunctional consensus for polity formation is pretty large and it mostly depends on how policymakers will balance the preferences of voters who prefer a pooled or a reinsurance polity.

Such a possible coalition between respondents of the centralized, pooled and reinsurance polities nonetheless faces a powerful minority expressing support for the decentralized polity: almost 24 per cent of respondents. This powerful minority is however not as consequential as in other policy domains: while it is bigger than in refugee policy domain (21.8 per cent), it is smaller than in the domain of foreign policy (26.4 per cent) and defence policy (27.6 per cent).

4.2 Territorial: Preferences for Polity Types by Country

The energy policy domain is marked by high heterogeneity between the countries in our sample given different levels of energy dependence on Russia. Figure 10 offers a visualization of this heterogeneity. The left-hand graph plots total gas imports from Russia (as share of all gas imports) against the burden of public debt (as a share of GDP). It shows how countries are dependent on Russia for their gas, and how fragile they are in terms of public capacity to fund alternatives and, perhaps most importantly, to cushion the blow of any sanctions via fiscal transfers for their electorates.

Figure 10

Public debt to GDP and Russian energy dependence across the EU

Data source: Data on energy are taken from Eurostat, which calculates the volume of energy imports compared to the volume of energy consumption. Note that some countries import more gas than they need because they re-export it. Data on public debt are taken from the World Economic Outlook of the International Monetary Fund (Version of October 2022). In the figures, ‘countries exposed to Russia’ means either bordering Russia or having in the recent past experienced conflicts with Russia (e.g., former communist countries).

In this two-dimensional space, we can group countries into four categories and we predict that these four categories should on average have different preferences for the EU polity in the realm of energy policy. The first group encompasses countries that are dependent on Russia for gas, but robust economically (bottom right corner: this group includes Germany, Poland, and FinlandFootnote ¹³ in our sample but also other countries such as Luxembourg, Netherlands, or Denmark). These countries should prefer, on average, a pooled polity that helps to withstand the cost of sanctions but does not centralize policymaking to such an extent that it curtails sovereignty and imposes costs on economically robust countries.

The second group comprises countries that are dependent on Russia for gas, and weak economically (top right corner: this group includes Hungary and Italy in our sample, but also others such as Greece, Austria, and Croatia). These countries should prefer a centralized polity that would help with solidaristic safety nets and socialize solidarity in order to spread the cost of sanctions.

The third group consists of countries that are not dependent on gas for Russia and economically vulnerable and thus could fall pray to a spillover of costs from energy markets to other markets or to inflation – and which could as a result back opposition to energy sanctions and/or more polity building in the field of energy (top left corner: Portugal and France, in our sample). These countries should, on average, display stronger profiles for the reinsurance polity, that is, a polity that intervenes in a decentralized manner only if the unforeseen need arises to deal with a vulnerability.

The fourth group includes countries that are neither dependent on Russia for gas nor vulnerable economically in terms of budgetary power (Malta, Ireland, and Romania). These countries should have a strong preference for a decentralized polity. However, none of these countries fall within our sample.

In terms of territorial divides, Figure 11 shows a fair amount of country heterogeneity and our hypotheses are only partially confirmed. Respondents in Italy, Portugal, and Poland all prefer a centralized polity to the other types, with the pooled polity category coming in as a far second. Hungary, Germany, and France stand out as the most divided countries. In these three countries preferences for a centralized polity and for a decentralized polity have similar values indicating a bigger societal divide in the kind of polity respondents want. Nevertheless, while in Hungary this societal divide really is just split between centralized-decentralized, in France and Germany, the pooled polity type also enters the mix indicating that in these countries, those that are not loyal to the polity are heavily split in their pooling preferences. Finally, in what regards energy sharing, Finland is the country that stands out as the only one in which respondents are more strongly in favour of a decentralized polity compared to all other polity types. Preferences for a decentralized polity are followed by preferences for a reinsurance one, indicating that respondents in Finland are heavily against the pooling of resources in the energy policy domain. This is quite surprising given the high reliance of Finland on Russian energy and the heavy consequences of sanctions on rethinking energy sources and mitigating price rises.

Figure 11

Preferences for polity types by country

4.3 Functional: The Effect of Individual Attitudes on Polity Types

We now move on from territorial divisions to functional divisions. We start with the output legitimacy. Figure 12 suggests higher satisfaction with EU sanctions increases preferences for a centralized polity at the expense of the decentralized polity (with an effect size of more than 20 percentage points if one were to move on the opposite sides of the satisfaction scale) and the less one wants a decentralized polity (with an effect size of almost 25 percentage points). Thus, EU measures have the potential to mitigate polity skepticism and consequently polity preferences are at least partially endogenous to European crisis responses. Satisfaction with the EU measures also slightly increases the probability of being in the reinsurance group and slightly decreases the probability of being in the pooled polity group, suggesting that satisfaction might be coupled more with loyalty towards the polity, than preferences for pooling means and decisions.

Figure 12

Performance evaluations and support – energy

When it comes to satisfaction with the actions of the Member States’ governments, we observe an inverse relationship between this and preferences for a centralized polity. Granted, if one is already satisfied with what the actions of the strong sub-units of the polity, the incentive to centralize is lower. In line with this, there is also a positive effect of national satisfaction on decentralizing (albeit smaller than the one on centralizing), while there are little to no effects on the mixed polity types – pooled and reinsurance. Finally, in terms of support for sanctioning Russia, irrespective of the level at which it is implemented, we observe results similar to those obtained for the satisfaction with the EU measures.

In Figure 13 we turn to ideational and security factors. The results show a strong effect of ideology and trust in Ukraine on polity preferences, while it suggests that security factors – economic vulnerability and threat perceptions stemming from the invasion – play a much smaller role even in the energy domain, a domain with strong asymmetric implications in what regards the economic outcomes of the policies taken. The top left plot shows the strong effect of ideology: the more a respondent self-places themselves to the right, the more they support a decentralized policy when it comes to sharing energy costs (25 percentage point increase along the 0–10 ideology scale) and the less they want a centralized policy (38 percentage point decrease along the 0–10 ideology scale). Along similar lines, the bottom right plot shows strong effects of trust in Ukraine: the more one trusts Ukraine, the more one prefers a centralized polity and the less one prefers a decentralized one. Furthermore, trust in Ukraine is also positively associated with preferences for a reinsurance polity and negatively associated with preferences for a pooled polity, suggesting a strong association with the polity loyalty of respondents in particular. By contrast, threat perceptions and economic vulnerability have no to small effects net of the other factors included in the model.

Figure 13

Ideology, threat, trust – energy

4.4 Dynamics over Time

We now move on to the analysis of change asking which polity category is most preferred and most stable over time, and if respondents change preferences, which polity category do they switch to? Our empirics suggest that the centralized and decentralized categories are the biggest and the most stable, while respondents tend to move in and out of the pooled and reinsurance polities to the other two categories. Most importantly, our empirics confirm to a certain extent the coalitional politics between groups preferring pooled, reinsurance, and centralized polities are very much possible, which indicates that policymakers have some leeway in shaping polity formation – provided they can convince their electorates.

These changes vary in our three waves. Let us have a look in turn, starting with switches from waves 1 to 2 (see Table 4). The most stable polity categories are centralized and decentralized: here, 67 and 63 per cent of respondents respectively stay in their initial categories. Most of the respondents who opted for centralized polity in wave 1 moved on to the pooled polity (15 per cent) and the reinsurance polity (12 per cent). Those who opted to move away from the decentralized polity rather moved towards the pooled polity (24 per cent) than the reinsurance polity. By contrast, very few people move between the centralized and the decentralized polity types (between 6 and 7 per cent of respondents). All in all, the change between waves 1 and 2 suggests that people either stick to their guns in terms of polity preferences or move towards more pooling in general. The change between waves 2 and 3 provides further evidence for the stability of the centralized and decentralized polity types and the flexibility offered to respondents by the pooled and reinsurance types (see Table 5). There is still very little movement between centralized and decentralized polity types, while the pooled and reinsurance types of polities function as a flexibility valve for respondents who change their minds on what polity type is best in their opinion in terms of energy policy.

Figure 14 shows that the largest change occurs between waves 1 and 2, as the centralized and reinsurance groups shrink and the pooled and decentralized grow. While there are also intense flows between waves 2 and 3, the relative sizes of groups do not change. What is visible, through the thickness of the flows, is the relative stability of the centralized and decentralized groups, compared to pooled and reinsurance. The majority of respondents in the former two groups retain their status, while for the latter a majority switch in and out. It is also noteworthy who switches in and out of the pooled and reinsurance groups. The centralized group contributes to both, but receives more support from the reinsurance group, suggesting that it is easier to persuade to pool resources and/or decisions those who already have loyalty towards the polity. Conversely, the decentralized group both loses and gains individuals from the pooled category, suggesting that policy attitudes are more malleable than polity ones.

Figure 14

Energy policy change and stability across waves

What determines then the change of polity preferences? In Figure 15, we focus on the effect of performance evaluations and security factors asking whether a change in polity preferences is endogenous to the evaluation of responses to a crisis (satisfaction) or whether it is driven by interactions with external factors (threat and economic vulnerability).Footnote ¹⁴ The results suggest that change between polity types is mostly driven by satisfaction, and to a much lower extent by the materiality of the crisis in energy policy itself. The higher the change in satisfaction with EU sanctions between waves 1 and 2 of our survey, the higher the probability of remaining centralized and slightly higher probability of switching to reinsurance, and almost zero probability of switching from centralized to decentralized. To a certain extent this pattern remains when we analyse the change from waves 2 to 3, with the only change being that more satisfied respondents now are less likely to switch to a reinsurance polity type. Likewise, satisfaction with national sanctions presents a mirror image: the more respondents become satisfied between waves 1 and 2 with national sanctions, to less likely they are to remain in the centralized polity and the more likely they are to switch to the pooled polity (although this pattern disappears in the change from waves 2 to 3).

Figure 15

Analysis of change – energy

The magnitude of the effects for threat perception and economic vulnerability are smaller, but their effects are still substantively worth noting. Increasing threat perceptions have a flat effect on remaining in the centralized polity but a positive impact on the probability of switching to a reinsurance polity. This makes sense, since greater threats increase the need for strengthening already existing national capacity, rather than transferring such capacity (and creating it ex nihilo) to the EU level. Interestingly, increased threat perception reduces the probability of switching from centralized to decentralized to almost zero. Respondents hence seem to prefer some form of engagement of the EU level under threat to having national-level policymaking only. Finally, while the effect of change in economic vulnerability is small, the pattern of remaining in the centralized polity and switching to other types of polities changes. Respondents who are more vulnerable economically are as likely to stay in the centralized polity but are also more likely to choose the pooled polity.

4.5 Conclusion

In the realm of energy policy, respondents clearly demand a centralized polity (with the pooled polity coming second in preferences), which is in line with other findings in the literature (Nicoli, Duin, and Burgoon Reference Nicoli, van der Duin and Burgoon2023) and signals that a European risk-sharing logic is desired by respondents. Similar to refugee policy domain, here as well preferences for polity types are strongly associated with performance evaluations in this domain, suggesting that the way the European polity reacts to a crisis deeply influences polity preferences. Likewise, over-time change in polity preferences is mostly influenced by changes in such policy satisfaction. Thus, the EU’s responses to crises can ease or harden the democratic postfunctional constrain on policymakers. This is surprising for energy policy, given the strong asymmetries in energy dependence and fiscal capacity to meet energy needs and compensate those affected. We note however, that there are important territorial divisions that might be hard to overcome, with respondents in Italy, Portugal and Poland being more in favour of a centralized polity, respondents in Finland being the least in favour, and Hungary, Germany, and France showing mixed preferences for polity types.

5.1 Descriptives: General Preferences for Polity Types

We begin by analyzing how respondents from our survey cluster in the four polity types that represent our dependent variable. Figure 16 presents the size of these four groups by operationalizing loyalty towards the polity with preferences for further European integration and policy attitudes with preferences for centralization in the foreign policy domain.

Figure 16

Polity versus policy attitudes on foreign policy

The graph indicates that when it comes to the EU’s foreign policy, the most preferred polity type when it comes to foreign policy is the centralized polity (policy centralization with high loyalty for the polity) with 39 per cent of respondents. The second most preferred category however is now the decentralized polity: these are respondents who neither support EU integration, nor would like foreign policy to be centralized. In the third category are those in favour of a pooled polity (centralization without strong loyalty) and the fourth and smallest category is now the reinsurance polity (coordination with strong loyalty). What is worth underlining from these descriptives is that a pro-centralization coalition between those respondents who are primarily in favour of a centralized polity and those who prefer a pooled polity represents a majority of respondents: 57.97 per cent. In Schattschneiderian terms (Hacker and Pierson Reference Hacker and Pierson2014; Busemeyer, Abrassart, and Nezi Reference Busemeyer, Abrassart and Nezi2021), it means that EU politicians could rest their policies on this coalition of respondents that all prefer pooling foreign policy. The calculus of the pooled polity group is more utilitarian than idealistic: therefore, as long as a centralized foreign policy delivers for national governments, these respondents are likely to remain supporters of a centralized foreign policy.

5.2 Territorial: Preferences for Polity Types by Country

It is difficult to predict the preferences of states when it comes to pooling their foreign policies. On the one hand, small countries should tend to bandwagon and align behind more powerful states as well as exploit the power gained by being able to co-decide the foreign policy of a bigger centralized polity. Giving up certain policy positions would then be compensated by greater power and coercive capacities at the European level. Big states, by contrast, should also embrace this policy because they can influence a common foreign policy much more. Thus both big and small states should be in favour of a centralized polity. On the other hand, however, small states may be reluctant to pool foreign policy especially if it would be too misaligned with their own preferences – for instance, the Baltics may fear that bigger countries would prefer to appease a revisionist power rather than confront it. Likewise, bigger countries would be wary of giving a seat at the table to small countries and allow them to ‘upload’ and ‘Europeanize’ their foreign policy goals. More importantly, big powers looking to coordinate their foreign policy should be wary of small states that can veto collective action.

We therefore have no strong theoretical expectations. Figure 17 suggests that the respondents most convinced about the centralized polity are to be found in Italy, Portugal, and Poland. The same applies to Hungary and Germany but here respondents are very much divided between the centralized and the decentralized polity, with the pooled polity coming a close third in Germany. Finland is even more divided as the preferences between the two extremes are almost equal and a sizeable share of the respondents (above 20 per cent) expresses preferences for the pooled polity. French respondents are the most idiosyncratic: their preferences for the centralized polity are the lowest (which goes against French policymakers’ aim of ‘strategic autonomy’) and their preferences for a decentralized polity are the highest (around 35 per cent of respondents). However, the French respondents wanting to build foreign capacity in the centre of the European polity are still a majority.

Figure 17

Preferences for polity types by country

5.3 Functional: The Effect of Individual Attitudes on Polity Types

The section on descriptives suggested that a majority of respondents is in favour of a centralized EU foreign policy, with their degree of loyalty for the EU polity varying. We now move on to a more inferential analysis, asking what predictors shape support for our four ideal polity types in foreign policy.

Figure 18 visualizes the effect of output legitimacy predictors and predictors related to general support for the policy (satisfaction with EU military aid to Ukraine, satisfaction with national military aid to Ukraine, support for military aid to Ukraine and support for increasing national military capacity). Satisfaction with EU military aid to Ukraine primarily impacts preferences for the centralized and the decentralized forms of polity: the more one is satisfied with the EU’s aid to Ukraine, the more one prefers the EU to take the shape of a centralized polity. Conversely, higher satisfaction decreases preferences for a decentralized polity (the same results apply to support for military aid to Ukraine). Satisfaction with national military aid has the opposite effect: the more national governments are seen as well-performing units, the more there is support for a decentralized but also for a pooled polity and the less support there is for a centralized polity. In other words, respondents do not necessarily think in binary terms about the EU: satisfaction does not lead to a trade-off  between a centralized and decentralized polity but leaves place for other types of polities to be supported (results are similar for support for increasing national military capacity).

Figure 18

Performance evaluations and support – foreign policy

Figure 19 vizualises the effect of ideational and security factors. Three variables go in a similar direction as far as preferences for a centralized polity in foreign policy go: the higher the perceived threat, the higher the economic vulnerability, and the higher the trust in Ukraine, the higher the preferences for a centralized polity in EU foreign policy and the lower the preferences for a decentralized polity. Threat and economic vulnerability show very small effect sizes, above and beyond trust in Ukraine and ideology. Concerning ideology, on the left respondents prefer most strongly a centralized polity and then a reinsurance polity, while right-wing respondents would rather have a decentralized polity first and a pooled polity second.

Figure 19

Ideology, threat, trust – foreign policy

5.4 Dynamics over Time

If in the previous section we looked at the static predictors of preferences for polity types, in this section we focus on explaining the extent to which our respondents change their polity preferences over time and the predictors associated with such change.

Table 6 and Figure 20 show the extent to which individuals in each polity type switched to a different polity type from waves 2 to 3 in our survey.Footnote ¹⁵ We first note the high stability of the centralized category, with 72 per cent of respondents maintaining their preference over time. Second, we also note substantive shares of individuals in the pooled and reinsurance groups who switch to the centralized category. These two categories are also the most dynamic ones in our data (losing and/or gaining more respondents over time).

Figure 20

Foreign policy change and stability across waves 2 and 3

Examining Figure 20 in more in-depth we also note that the centralized type gains equal and substantive shares of respondents from both the reinsurance category and the pooled category over time, whereas the decentralized type mostly gains respondents from the pooled category. This indicates that it is mostly those respondents without loyalty to the polity to begin with that can turn against pooling resources over time. Furthermore, while the gains and losses in the centralized category are rather symmetric, making the size of this group quite stable over time, the decentralized group ‘loses’ more respondents than it ‘gains’ (especially in what regards the pooled group) resulting in net decreases of this group in wave 3.

Figure 21 shows our multinomial analysis of change. Two findings from the figure are similar to the findings in other policy domains. First, the strongest predictor for remaining in the centralized category is satisfaction with EU military aid to Ukraine. Those respondents whose satisfaction with military aid increases are more than 75 per cent likely to remain in the centralized category. Furthermore, increases in satisfaction are coupled with decreases in switching to any other category. Second, similar to the other policy domains, we can see that changes in the threat stemming from the war and economic vulnerability are poorer predictors of change in polity types, at least when taking into account ideational and output legitimacy factors. This further underscores the idea that polity attitudes are heavily dependent on performance and output legitimacy, more so than on the threats stemming from the crisis. Finally, the impact of satisfaction with national policy is similar to that of its impact in waves 2 to 3 for military and energy policy, namely null. In the other policy fields, we saw that there was a negative impact of satisfaction with national policy on remaining in the centralized group, but that effect was mostly visible at the beginning of the crisis, from waves 1 to 2.

Figure 21

Analysis of change – foreign policy

5.5 Conclusion

Foreign policy falls in the middle of the policy domains we study when it comes to support for the centralized polity type, with slightly lower support than refugee and energy policy, and more support than military policy. Compared to refugee policy, foreign policy shows starker territorial divisions. The same three countries have the highest share of respondents in favour of a centralized polity: Italy, Portugal, and Poland, followed by Hungary and Germany who are also more heterogeneous. Finland and France show deeper divisions, with majorities in favour of a decentralized polity. In functional terms, we see that satisfaction with EU military policy plays a similar role to other policy fields, increasing the probability of favouring a centralized compared to a decentralized polity. National satisfaction shows the opposite, although less pronounced, trend. Interestingly, and differing from military policy, pacifism (proxied through lack of support for increasing national military capacity) does not play a role in shaping polity preferences when it comes to foreign policy. Ideology is an important driver of support with the left being more in favour of centralizing while the right is more in favour of decentralizing. Differently from energy and refugee policy, but similarly to military policy, threat and economic vulnerability have a significant, if small, effect. Higher threat and higher economic vulnerability are related to stronger support for a centralized polity. However, these results are not confirmed in our analysis of change, where changes in neither has an important impact in driving changes in support.

6.1 Descriptives: General Preferences for Polity Types

Figure 22 shows the size of the group supporting each polity type when it comes to the defence domain. Similar to the other policy domains, the respondents preferring either a centralized polity (33.18 per cent) or a pooled polity (18.64 per cent) constitute a majority of respondents. This indicates that respondents have a strong preference for an EU army, whether in a centralized or pooled form, suggesting that there is a relatively strong coalitional basis for such an institutional reform at the EU level.

Figure 22

Polity versus policy attitudes on military policy

Nevertheless, the results also show high levels of preference for the decentralized and the reinsurance groups: 27.65 per cent of respondents express preferences for the former and 20.53 per cent of them for a reinsurance polity. The reinsurance group, pro-Europeans but skeptical of pooling resources, is the largest when compared to the size of this group in the other policy domains and, together, with the decentralized group, while still a minority, amounts to the biggest opposition towards centralization.

6.2 Territorial: Preferences for Polity Types by Country

Theoretical guidance on country preferences on military spending highlights three structuring dimensions: the context and the presence of a threat, the geographical proximity of the threat and the trade-off between guns and butter – that is between social and defence spending (Bartels Reference Bartels1994; Williams Reference Williams2019; Barnum et al. Reference Barnum, Fariss, Markowitz and Morales2024). In general, military spending preferences can be seen as a function of perceived external threat, which varies in time and space. Defence spending during the Cold War was much higher and much more supported than during the golden age of globalization (1989–2014), during which the ‘peace dividend’ allowed countries to dramatically reduce defence spending (apart from the United States, engaged in two wars). However, the rise of China as a strategic challenger and the resurgence of a revisionist Russia with the invasion of Ukraine in 2014 have ended this ‘peace dividend’ and have even spurred some divergences within NATO. Countries which are geographically close to a potential threat, such as the Baltic states and Poland, should witness a higher support for military spending. Finally, the literature also points to the theoretical trade-off between guns and butter: public preferences and actual spending on defence should decrease when the welfare state has a prominent place in the political economy of a country, especially in times of austerity (Hübscher, Sattler, and Truchlewski Reference Hübscher, Sattler and Truchlewski2023). The evidence on this point is quite mixed however, as voters often internalize the Keynesian logic of defence spending (Williams Reference Williams2019): every dollar or euro spent on defence generates jobs and growth which benefit the broader political economy.

Figure 23 suggests that some of these patterns hold. First, as the Cold War started to vanish in the rear-view mirror of history, military spending did decrease significantly and converged among countries to lower levels of GDP, especially after 2000 (the grey area represents the mean and ± standard deviation of Western countries). Second, military spending decreased faster and to lower levels in some countries than in others: take for instance landlocked Hungary, whose defence spending dropped below 1 per cent of GDP in 2014, or Germany, whose defence spending never increased beyond 1.4 per cent of GDP since 1997 (even after Russia attacked Ukraine in 2014). Their military spending is even lower than that of Portugal and Italy, who should not be threatened by a resurgent Russia. By contrast, other countries, like Russia-weary Poland, tried to maintain defence spending at around 2 per cent of GDP despite the tribulations of the post-communist transition. France followed the same goal due to the necessity to maintain its nuclear deterrent capacities and its role in forging security in Africa and the Middle-East. Finally, Finland does not fit the geography argument: its military spending does not reflect the perception of a Russian threat, which should be exacerbated by Finland’s lack of NATO security until 2022. This is perhaps explained by the guns versus butter trade-off: countries with high social spending (Finland is in the top five, together with Italy and FranceFootnote ¹⁸) should tend to have lower defence spending.

Figure 23

Defence spending in Western countries, in % of GDP

Data source: SIPRI Military Expenditure Database. The shaded area represents the range of spending in Western countries (minimum and maximum) for context. Western countries include: Austria, Belgium, Canada, Denmark, France, Germany, Greece, Ireland, Italy, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States.

Figure 24 shows country-level preferences for polity types surrounding military policy. First, as suggested in the previous section, compared to other policy fields, respondents are much more lukewarm towards the centralized polity but still form the largest group. In terms of territorial divides within the EU, three countries have the largest group of respondents preferring a decentralized polity: Hungary, Germany, and Finland. In both Hungary and Finland the reinsurance type of polity comes second, while in Germany the centralized polity comes second, suggesting a polarized electorate. By contrast, respondents in a second group of countries – the same as for foreign policy, that is, Italy, Portugal, and Poland – express clear preferences for a centralized polity for defence. This is somewhat counter-intuitive: neither Portugal nor Italy are directly threatened by a resurgent Russia. Poland is indeed, but is always perceived as putting its bets on NATO and aligning with the only credible military power, the United States. France is once more the idiosyncratic case: this is the only country where the pooled polity comes first. Respondents are almost equally divided into three groups, which points to difficult coalition-building of voters on the national level and a harder postfunctionalist constrain (30 per cent of respondents prefer a decentralized polity). However, a push for a European level army could be feasible if policymakers cobble a coalition out of respondents preferring a pooled or a centralized polity (slightly above 60 per cent of respondents).

Figure 24

Preferences for polity types by country

All in all, countries where most respondents prefer a centralized or a pooled polity and form a majority that ease the postfunctional constraint are Italy (around 55 per cent), Portugal and France (around 60 per cent), and Poland (around 56 per cent). Countries where the coalition constrains on any form of pooling (decentralized and reinsurance preferences being highest) are Hungary and Finland (around 60 per cent).

6.3 Functional: The Effect of Individual Attitudes on Polity Types

We now move on to the multinomial results on the individual level. As for foreign policy, our previous results suggested that there is a significant amount of public support for an EU army which could be embedded into a centralized or a pooled polity. But which predictors on the individual-level shape support for each particular polity type?

We start with output legitimacy factors, displayed in Figure 25. Concerning satisfaction with military support for Ukraine, it appears that the more satisfied respondents are with aid, the more likely they are to prefer a centralized polity (when satisfaction increases from 0 to 10, support increases from a 20 per cent probability to a 50 per cent probability). In other words, support for a centralized polity increased 2.5 times. Conversely, support for a decentralized polity drops from 42.5 to 12.5 per cent, a decrease of 3.4 fold. Thus, satisfaction with EU aid to Ukraine has an asymmetric effects on polity preferences, the implication being that there is an important elasticity of satisfaction on polity type, as the potential support coalitions change dramatically in the two opposite ideal types (centralized vs. decentralized polity), but not in the pooled and reinsurance types of polity, which remain flat. We observe the same empirical patterns for general support for military aid to Ukraine, which reflects the results we obtained for the EU’s foreign policy. Naturally, almost the opposite is observed with satisfaction for national aid to Ukraine: the more respondents are satisfied with it, the more they support a decentralized polity at the cost of a centralized one. This suggests that the way a polity responds to a crisis influences the preferences for the polity itself.

Figure 25

Performance evaluations and support – EU army

Next, we continue with the ideational and security factors shown in Figure 26. First, examining ideology we see that respondents from the centre and the left opt for a centralized and a reinsurance polity, while respondents from the right prefer a decentralized or a pooled polity. Next, the greater the threat, the greater the demand for a centralized polity and the lower the demand for a decentralized polity (note that whatever the threat level, the decentralized polity is never preferred). Furthermore, vulnerability also has an impact on the type of polity preferred: the effect on the two ideal types of polities goes in the expected direction (more vulnerability leads to higher preferences for a centralized polity and lower preferences for a decentralized polity), while the effect on pooled and reinsurance polities is marginal. Trust in Ukraine has a similar impact (with the exception of the reinsurance polity, where the effect of trust in Ukraine is strong).

Figure 26

Ideology, threat, trust – EU army

6.4 Dynamics over Time

We now consider the dynamics of change in preference for polity types over the course of the crisis. Table 7 shows that from waves 1 to 2, 64 per cent remained in the centralized category, with 7 per cent switching to decentralized, 15 per cent to pooled and 15 per cent to reinsurance. In contrast to the other policy fields, when it comes to the EU army, the decentralized group is the most stable one. Thus, not only are respondents overall less likely to support a centralized polity for military policy, but they are also more likely to defect. However, similarly to other policy fields, we see a stabilization from waves 2 to 3 in Table 8, with the centralized group retaining its size and also gaining respondents from the pooled and reinsurance groups. These patterns can also be seen in Figure 27. Here we see the drastic decrease in centralized and increase in decentralized from waves 1 to 2, whereas waves 2 to 3 show a stabilization of centralized and a slight shrinking of decentralized. While in other policy fields reinsurance moved more towards centralized, in military policy in waves 1 to 2 more move from reinsurance to decentralized. This is reversed in waves 2 to 3, with both pooled and reinsurance switching more to centralized than decentralized.

Figure 27

EU army change and stability across waves

In terms of the predictors of change, Figure 28 shows the predicted probability plot for the output legitimacy and security relate variables: satisfaction with EU and national military aid to Ukraine (both from waves 1 to 2 and 2 to 3) and threat and economic vulnerability (only waves 2 to 3). We see similar effects to those already shown in the foreign policy and energy policy domains. Growing satisfaction with EU military aid to Ukraine is associated with a drastic increase in the probability to stay in the centralized group, and a shrinking probability to switch to the others. Satisfaction with national military aid has the opposite effect, lowering the probability of staying in centralized or switching to reinsurance, and increasing the probability of switching to pooled, and to a lesser extent, to decentralized. Increasing one’s satisfaction with national military aid to Ukraine therefore does not decrease support for the policy since respondents are more likely to prefer a pooled polity but less likely to prefer a reinsurance polity. Interestingly, and similarly to foreign policy and energy, satisfaction with national military aid is no longer predictive in waves 2 to 3. Unlike other policy fields, we note a substantive effect of change in economic hardship. Those who report being more greatly affected by rising energy prices in wave 3 compared to wave 2, are more likely to remain in the centralized category and less likely to switch to pooled.

Figure 28

Analysis of change – EU army

6.5 Conclusion

Military policy stands out as the domain with the largest dissensus among the ones we study. Both the centralized and the pooled groups are considerably smaller than in other policy areas, although together they still form a narrow majority. As such, our results differ from a more optimistic literature that is confined to Western Europe (Burgoon, Van Der Duin, and Nicoli Reference Burgoon, Van Der Duin and Nicoli2023). The small number of respondents favouring a pooled polity suggests that respondents cannot imagine pooling military competencies in a unified army without overall greater political integration. The fact that defence is a core state power may explain why it is much more polarizing for individuals. It may also explain why we see more territorial divisions. Our usual pro-centralized group of countries, Italy, Portugal, and Poland, contrast with Hungary, Germany, and Finland, which all have majorities for the decentralized type. France, on the other hand, is the most divided, with no clear majority. In terms of functional divides, we see similar patterns to other policy areas, as greater satisfaction with EU military aid increases the likelihood of supporting a centralized polity, while satisfaction with national aid decreases it. Differently from foreign policy, respondents who do not want to see any national military increase, that is, pacifists, are most in favour of a reinsurance polity, where they do not want to see military buildup at the EU level but support further political integration. Also differently from other policy fields, we see that both economic vulnerability and threat perception are positively related to supporting a centralized polity, albeit with smaller effects than other variables. When it comes to how opinions change over time, we note that military policy is also one of the most volatile. Only 64 per cent of those supporting a centralized polity in wave 1 remain supportive in wave 2. At the same time, a majority of those switching from supporting the pooled and reinsurance types go to the decentralized category. The result is a drastic shrinking of support for the centralized type and a growth of support for the decentralized one from waves 1 to 2. However, similarly to other policies, waves 2 to 3 see a stabilization and even a slight reversal as the centralized type gains in size. The analysis of change shows similar patterns to other policies, as satisfaction with the EU increases the probability of remaining supportive for the centralized type over time, while satisfaction with the national government decreases it. Differently from other policy areas, change in economic vulnerability has a positive effect on remaining in the centralized group.

7.1 Contributions to the Literature

In this Element, we set out to analyse the state of public opinion on the EU in the aftermath of the Russian invasion of Ukraine. We argued that focusing on public opinion is important given the politicization of the European polity and criticisms of its democratic deficit and weak voice channels. Conversely, a supportive public opinion can offer an enabling environment for policymaking at the EU level and could take the wind out of Euroskeptic parties’ sails.

While this Element is definitely not the first to focus on public opinion in the EU in times of crises, it aims to bring several important contributions to this literature. These contributions have been inspired by the polity approach to the European Union (Bartolini Reference Bartolini2005; Caramani Reference Caramani2015; Ferrera, Kriesi, and Schelkle Reference Ferrera, Kriesi and Schelkle2024). The key insight that we take from the polity approach is that the building of the EU polity does not necessarily need to imply a full transfer or new creation of ‘core’ institutions to the EU at the expense of the Member States. By contrast, the authoritative centre of the EU can be further ‘built’ by, for example, improving the coordination between member states and further developing the compound polity where member states play first fiddles (Ferrera, Kriesi, and Schelkle Reference Ferrera, Kriesi and Schelkle2024). Scholars have proposed concepts such as ‘extensive unification’ (Ferrera, Kyriazi, and Miró Reference Ferrera, Kyriazi and Miró2024; Truchlewski et al. Reference Zbigniew, Oana, Moise and Kriesi2025) or ‘coordinative Europeanization’ (Ladi and Wolff Reference Ladi and Wolff2021) to refer to forms of polity building that do not necessarily imply a full centralization of powers at the EU level. We leverage this insight related to the variety of pathways that EU polity formation can take to further conceptualize demand-side support for the EU in two ways.

First, starting from the idea of the diversity of pathways that the EU polity formation can take, we also went beyond conceiving public support for the EU as uni-dimensional – that is, more or less pro-integration. For this, we proposed a new typology of public support for the EU that would allow us to capture a more diverse array of attitudes while maintaining parsimony. This typology conceives of EU support as playing out in two dimensions stemming from a distinction between ‘policy’ (specific) and ‘polity’ (diffuse) support (Easton Reference Easton1975). By policy support, we refer to support for pooling resources and/or decisions at the EU level in specific policy domains. By polity support, we refer to a more general positive attitude toward the EU based on a deeper loyalty towards the polity. While the concepts of specific and diffuse EU support are not novel and have been related to one another in the literature, we argue that there is merit in studying their intersections rather than solely their (positive) correlation. Cross-tabulating these two dimensions opens a rich, four-fold analytical space in which public support for the EU can be categorized into four types: support for a centralized polity, a decentralized one, a pooled polity, or a reinsurance polity. While the existing literature assumes that the two attitudes go hand in hand, we have shown that what might be thought of as inconsistent views (wanting more EU integration but not wanting to centralize a specific policy or vice-versa) (Zaller Reference Zaller1992) are widespread. Indeed, we argue that these views are not inconsistent but point to alternative potential development pathways for the EU. A pooled polity would combine less direct political integration with more cooperation or coordination of specific policies. A reinsurance polity would combine more political integration with a decentralized approach to specific policies. We show that both groups are sizable and vary across policy fields. Our analysis, therefore, sheds light on two groups of citizens that are usually ignored in studies of EU support: those that support what we call a pooled polity and those that support a reinsurance type of polity.

The second theoretical assumption inspired by the polity approach that we started with and that guided our results was the fact that neither are crises monolithic threats nor do EU polity formation pathways need to be the same across policy domains. Our descriptive findings point to the utility of looking not only at attitudes towards the polity but also within each policy domain. Support for types of EU polity can vary across these domains as a function of the asymmetries that they exacerbate between countries and social groups, of the performance of European and Member state actors, or of previous attitudes. These factors drive out territorial divisions – between citizens in different Member States – and functional divisions – between groups of citizens across Member States.

The third contribution of the Element is to compare how respondents weight internal and external drivers of polity formation. The internal drivers refer to output legitimacy and ideology, while external drivers refer to security-related factors such as threats stemming from the invasion. As the classical field of European integration focused mainly on internal drivers, studying external drivers has gained prominence only more recently and particularly following the Russian invasion. As such, our contribution is a response to such recent debates about how wars shape polities and the European one in particular. Leveraging a classical argument in political science (Riker Reference Riker1964; Tilly Reference Tilly1975), Kelemen and McNamara (Reference Kelemen and Kathleen2022) have argued that European polity formation is incomplete and that external threats like war would spur further centralization and capacity building. By contrast, Genschel (Reference Genschel2022), Genschel, Leek, and Weyns (Reference Genschel, Leek and Weyns2023), Moise, Truchlewski, and Oana (Reference Moise, Truchlewski and Oana2024) argue that this is not necessarily the case due to the compound nature of the European polity (Ferrera, Kriesi, and Schelkle Reference Ferrera, Kriesi and Schelkle2024). Despite the deep intellectual aggiornamento that the Russian invasion in Ukraine has forced upon European elites (together with the American threat to withdraw from NATO), our results suggest that the external drivers of polity formation are still weaker than classical internal drivers. In what follows we further summarize these findings across policy domains.

7.2 Summary of Our Results

Our results underscore the complexity of the paths that lie ahead for the EU. When it comes to public opinion there are varying degrees of consensus for different types of polity formation. While across policy domains we see that the group preferring a centralized polity is the largest one, we show that respondents in each country are not necessarily split between preferring a centralized polity or a decentralized one, with a pooled polity or a reinsurance one being viable alternatives in all policy domains (see Table 11).

When comparing policy domains descriptively, we see that the refugee policy domain has the highest share of respondents supporting a centralized polity type (see Table 11). Importantly, for refugee policy, the pooled type is also the second-largest category. Taken together these groups imply that the refugee policy domain is the one where we can find the widest consensus in regards to pooling resources and/or decisions at the EU level. This domain is followed by energy policy, foreign policy, and, finally, military policy, which has the lowest share of respondents preferring a centralized polity and the highest share preferring a decentralized one relative to other policies (yet note that it still has centralized as the largest group). What makes for a more permissive consensus in refugee and energy policy? One aspect might be that they do not involve giving up core state powers, as would be the case in the foreign policy and military fields. In both arenas EU security is enhanced by the fact that most EU member states belong to NATO, which provides external security, thus lessening the functional pressure on the EU.

In terms of territorial divides, we note important country differences among our respondents across policy fields. Thus, Table 10 shows the country divisions in terms of preferences for refugee policy are relatively muted, with only Hungary having a majority of respondents in favour of the decentralized polity and France having a majority for pooled, while the rest of the countries all have a plurality of respondents in favour of a centralized polity. Across the other three policy fields, we see a consistent ‘pro-centralized polity coalition’ formed by Poland, Portugal, and Italy. In energy and foreign policy, they are joined by Hungary and Germany. Military policy remains the most polarizing across countries. Finnish respondents, perhaps surprisingly, have a plurality for the decentralized polity. This might be surprising given Finland’s geographical proximity to Russia and, hence, its higher threat level. However, as our individual results show, threat plays a minor role in driving support for polity types and its change over time.

When it comes to our static analyses, several patterns stand out (see Table 11). First, we note several similarities across policy fields. Support for polity types is driven not by external, security-related conditions, such as threat and economic vulnerability, but by internal, ideational and output legitimacy-related factors, including satisfaction with EU and national policy, ideology, and trust in Ukraine.Footnote ¹⁹ Importantly, quite stable across policy domains, satisfaction with the EU and national policy have strong effects on preferences for polity types and play opposite roles, the first being positively related to more polity centralization while the latter being negatively related. We note the same dynamic in our analysis of change between our three waves.Footnote ²⁰ Taken together, these results suggest that within countries, preferences for these four polity types are more strongly rooted in output legitimacy and deep attitudinal variables rather than in factors directly related to the security or economic threats raised by the war when taking all attitudes into account. One exception is military policy, where threat and economic vulnerability have significant effects even net of output legitimacy and ideational factors. In the military field, higher perceived threat and higher economic vulnerability increase the propensity towards the centralized polity type and decrease that towards the decentralized type.

Second, we also note important differences between policy domains. Most notably, refugee policy stands out as having the largest permissive consensus across ideological groups, as well as satisfaction with both the EU and national policy. Even at low levels of satisfaction with the EU and high levels of satisfaction with national policy, the centralized and pooled groups are the largest. At very high levels of satisfaction with the national government, the pooled group is similar in size to the centralized group. Military policy stands out as having the largest constraining dissensus and some of the largest heterogeneity by ideology and satisfaction, with right-wing respondents being more in favour of a decentralized polity and the left wing being split between centralized and reinsurance. Interestingly, pacifist respondents (who are completely against increasing national military capacity) are the only group that forms a majority for the reinsurance category. This is intuitive given that this group opposes not only national military increases but also increased military capacity and centralization at the EU level, while still supporting the EU polity in general. Since these individuals are more likely to be left-wing, this can explain the split we see in left-wing individuals in this policy field. The defence polity domain is, hence, the least likely case for centralization when looking at the demand side. Given that defence is a core state power and that (most) EU countries have external security guarantees through NATO, this is unsurprising.

Given the asymmetries in energy dependence and fiscal capacity, we note that energy policy also initially appeared to be a least likely case for high preferences for centralizing or pooling. However, as our analyses showed, security considerations played a minor role in attitude formation or change. This may explain why, despite the asymmetries, energy policy shows less conflict than what might be expected from an asymmetric policy, such as the case of the 2015 refugee crisis. The consensus on refugee sharing despite the asymmetric burden of Ukrainian refugees in 2022 shows that in certain instances, such asymmetries are not important when more powerful factors are at play such as output legitimacy, ideology, and views on the war.

Finally, when we look at our results over time we note that polity attitudes are not as stable as one might initially expect. Even in the short timeframe that we use here, from March to December 2022, we see intense volatility in polity preferences. The first months of the war saw many shifts in the political position of governments and the EU as policy adapted to the initial shock and to later developments in the war. Public opinion followed suit. Our results point to a decrease in support for a centralized polity from our waves 1 to 2, likely indicating an eroding rally-around-the-flag effect in line with previous results (Truchlewski, Oana, and Moise Reference Zbigniew, Oana and Moise2023). However, from waves 2 to 3, the size of support for each polity type begins to stabilize. One exception to this trend is again refugee policy, which saw a small reversal of the fading rally trend as the pro-centralized polity group grew larger in wave 3 while the pro-decentralized polity group shrank, ‘losing’ respondents to the pooled group.

Our temporal analysis of the centralized group confirmed our static findings from wave 2.Footnote ²¹ Growing satisfaction with the EU increased the probability of remaining in the centralized category, while growing satisfaction with national policy decreased it. In the case of refugee and energy policy, respondents who were very satisfied with the national government were more likely to switch from centralized to pooled, while for military policy this resulted in a gain for the decentralized group. Results for the change between waves 2 and 3 were broadly similar. For both foreign and military policy the second wave saw a fading of the effect of satisfaction with national military policy, with a small effect for energy policy. Similarly to our static analysis, the role of threat perceptions and economic vulnerability is limited. We note that growing economic vulnerability increases the probability of staying in the centralized group for military policy while increasing threat perception has a small positive effect on switching from centralized to reinsurance. Overall, however, it is again output legitimacy-related and ideational factors that are strongly related to changes in preference for polity types rather than external, security-related factors.

7.3 The Future of the EU after the Russian Invasion of Ukraine

What are the implications of our findings? First, our polity approach helps us to see that European politics are not frozen in a binary integration/anti-integration or pro/anti-EU cleavage, but rather that oftentimes respondents prefer different forms of polity formation that maximize the scale of the response to common problems without sacrificing sovereignty (e.g., pooled or reinsurance polity). A corollary of this proposition is that there is not necessarily a trade-off between European and national polity formation: strengthening one does not need to come at the expense of the other. Instead, we see a strong demand for the co-evolution of these two levels of the polity, where the European polity does not supersede the national one but enhances it.

Second and consequently, our results also underscore that there is a strong potential for coalitions among respondents. While the demand for a centralized polity is strongest, particularly, in those policy domains that do not involve core state powers (energy and refugee policy), in the foreign policy and defence domains Euroskeptical respondents do not necessarily all prefer the decentralized polity but demand, for instance, a pooled type of polity. This creates ample space for policymakers to ‘craft’ coalitions among their constituencies to gather support for policies that can shape the EU. For instance, policymakers and political elites seeing a need for further resource pooling in a policy domain can craft a coalition of respondents between the centralized and the pooled polity groups. Across all policy domains, these two groups together form a majority of respondents (as compared to the decentralized and reinsurance groups).

Third, such coalitions can lead to novel forms of centre building that do not necessarily entail a transfer of sovereignty but can rather involve novel forms of resource-pooling and risk-sharing. The key EU response to the COVID pandemic was the ‘Next Generation EU’ instrument which is a case in point of a compromise between those states that need solidarity and those that are reluctant to pay for it. The result is a temporary coordinated structure (for now) that pools common economic power to disburse fiscal stimulus to the national level (instead of centralizing it in Brussels). Likewise, military help to Ukraine is channelled through the European Peace Facility which reimburses Member States for the military hardware they transfer to Ukraine. Such flexibility in coalitions and forms of centre-building in the European polity enables a variety of polity pathways for the EU, pathways that can also vary across policy domains.

Fourth, the cost of this is that the EU is not necessarily bound to evolve as a coherent polity that can be designed ex-ante, but rather that the EU evolves piecemeal into a compound polity without a teleology or finalité along varying polity formation pathways by policy domains. This is the price to pay to assuage voters who may be skeptical of the EU and to avoid a strong backlash against integrationist leaps forward. The failed European Constitution of 2005 and the difficult ratification of the Lisbon Treaty have put such ‘big bang’ strategies to rest. However, the benefit of such a development is that the European Union can still form and reform as a polity even though crises may polarize its territorial and functional constituencies.

Finally, our results show that external threats do not automatically push public opinion to demand a more centralized polity in all policy fields. Despite the predictions of the ‘bellicist’ school of polity formation (Tilly Reference Tilly1975; Riker Reference Riker1964; Kelemen and McNamara Reference Kelemen and Kathleen2021), leveraging a classical argument in political science, we rather see that internal drivers of polity formation are equally contributing to centralized polity preferences in all policies (while threat for instance only matters for foreign policy and defence). The external threat of Russian imperialism and war-mongering thus mostly operates through other channels than the security logic, as we show in this Element.

The fact that public opinion favours different forms of polity formation that are not the centralized ideal type also has implications in terms of which public goods the EU produces. The pooled and reinsurance types of polities suggest that public goods like energy or military security would still be produced at the national level, not the European one which would rather act as either a backstop or a coordination device. In other words, member states are responsible for their own public goods. If we see international organizations as clubs creating public goods, which therefore raise the costs of staying out (Anghel and Jones Reference Anghel and Jones2024), the implication is that the impact of Russia’s war on Ukraine may perhaps undermine the procurement of centralized, truly common European goods.