Introduction

Three features of a crisis are commonly identified: threat, urgency, and uncertainty (see Alexander-Shaw, Ganderson, and Kyriazi Reference Alexander-Shaw, Ganderson and Kyriazi2023; Boin et al. Reference Boin, t’Hart, Stern and Sundelius2005: 3–4; Ferrara and Kriesi Reference Ferrara and Kriesi2022; Lipscy Reference Lipscy2020). The crisis situation corresponds to an extraordinary moment of urgency and uncertainty that poses an immediate threat to the proper functioning of the policy domain concerned by the crisis and, if the threat is existential enough, to the population at large and the polity as a whole. Crisis situations are embedded in a context. They vary with the nature of the threat exerted by the crisis, the polity in which they occur, and the conflict structure that prevails in the polity. In the case of our two crises, the nature of the threat is given by the Syrian civil war and the large number of refugees it created and by the pandemic. A master tension between the supranational center and the member states characterizes the EU polity exposed to these threats (Ferrera, Kriesi, and Schelkle Reference Ferrera, Kriesi and Schelkle2024). This tension gives rise to center–periphery conflicts between the supranational center and the member states, territorial conflicts between member states (transnational conflicts), and functional conflicts between pro-integrationist (cosmopolitan) and anti-integrationist (nationalist) social groups within member states.

Crisis situations, in turn, shape the crisis-specific policymaking processes, the policy output, the policy outcome, and broader implications for the polity at large.Footnote ² Figure 1 presents the analytical building blocks of the argument. This study focuses on the link between crisis situation and crisis-specific policymaking and policy output. The essential claim of this study is that the crisis situation decisively shapes crisis-specific politics and policymaking, which then determines the policy output and effectiveness. The broader implications of crisis policymaking for the resilience of the polity will get little attention.

Figure 1 The analytical building blocks of the theoretical framework.

Source: Author’s elaboration.

The discussion of the theoretical framework is divided into two sections: The first section introduces the concepts for the characterization of the crisis situation and describes the two crises studied here in terms of the relevant characteristics; the second section formulates expectations about the impact of the crisis situation on crisis-specific policymaking processes and their output.

The Crisis Situations

Figure 2 presents the relationship between the three elements of the crisis situation as proposed by this framework. Problem pressure comes first. It derives directly from the critical context, particularly the threat exerted by the crisis. Political pressure on the policymakers is partly due to problem pressure. However, it can also be created by political actors intent on exploiting the crisis for their purposes, as argued by the emergency politics literature and others. Alternatively, political pressure can be alleviated by political actors coming to the rescue of those directly exposed to problem pressure. In any case, the effect of problem pressure on political pressure is partly mediated by the institutional context. In the specific context of the EU polity, the effect of problem pressure on political pressure at one level may be mediated by the institutional context at the other level – e.g., by EU authorities coming to the rescue of hard-hit member states and, thus, alleviating the political pressure exerted by the respective publics on their national governments. This means that problem pressure may be neither sufficient nor necessary for political pressure. In the final analysis, political pressure is a matter of perception and is not entirely structurally determined by problem pressure. Figure 2 presents the overall relationship between these basic characteristics of the crisis situation, which I discuss individually.

Figure 2 Relationship between the three essential characteristics of the crisis situation.

The Impact of the Crisis Situation on Policymaking and Policy Output

The dependent variables of this study are characteristics of policymaking and policy output that can be compared across crises. They include the salience of crisis-specific policymaking, the role (centrality) of the various actors in the policymaking process, policy support, and policy output. I discuss them one by one and formulate a series of hypotheses for each one of them. These hypotheses focus on the following aspects (between parentheses are mentioned the initials of the corresponding hypotheses):

• Crisis policymaking in general (H1-4)

• General crisis-specific effects (H1a-4a)

• Phase-specific effects (H1b-4b)

• Pressure-specific effects (H1c-4c)

• Policy-specific effects (H1d-4d)

• Actor-specific effects at EU (H1e-4e) and member state level (H1f-4f)

• Country-specific effects (H1g-4g).

General Methodological Approach

For the empirical analysis of the policymaking processes, I rely on policy process analysis (PPA) (Bojar et al. Reference Bojar, Kyriazi, Oana and Truchlewski2023), a comprehensive method for the data collection and analysis of policymaking debates. Policymaking has a core of puzzling and powering, which occurs in specific and not always transparent arenas. Yet, it also gives rise to many external signals reaching the public sphere, primarily through the mass media. Policy process analysis intends to capture this public face of policymaking. In its design, PPA draws upon protest event analysis (PEA) (Hutter Reference Hutter and Donatella Della Porta2014), contentious episode analysis (Bojar et al. Reference Bojar, Kyriazi, Oana and Truchlewski2023), and political claims analysis (Koopmans and Statham Reference Koopmans and Statham1999) but broadens its focus beyond contention to policymaking debates more generally. Similar to these other methods, PPA is an event-based methodology that focuses on identifying distinct policy actions in the media, undertaken by various actors, addressing particular issues, and how they unfold over time concerning the core of the policymaking process. Policy process analysis is also related to the Comparative Policy Agendas approach (Baumgartner, Green-Pedersen, and Jones Reference Baumgartner, Green-Pedersen and Jones2006). However, it goes beyond it because PPA is not limited to the agenda-setting phase but includes all significant steps in the policymaking process, from agenda-setting to implementation. The PPA data provides the essential characteristics of the policymaking process.

For the systematic analysis of the effects of the crisis situation on policymaking, the PPA data is aggregated and combined with data on the crisis situation and on policy output taken from various existing sources to create a dataset where the episode-month (refugee crisis) or the country-week (COVID-19 crisis) constitutes the units of analysis. I now detail the different steps in constructing the two basic types of datasets and the procedures to analyze this data.

Selection of the Period

Given the protracted period of the refugee crisis and the more intermittent policymaking process during this crisis, the SOLID project teamFootnote ⁴ decided to break it down into a set of crucial policymaking episodes, which salient policy proposals have triggered. Some of the policies the team has chosen are legislative acts, such as reforms to the countries’ asylum systems; others are administrative decisions and novel practices by state institutions, such as the reimposition of border controls in a heightened period of problem pressure. A policy episode comprises the entire policy debate surrounding these specific policy proposals that governments put forward, from when the proposal enters the public debate to when it is implemented and the discussion surrounding it is no longer salient. The episodes cover the period from the beginning of 2013 until the end of February 2020. There are forty-six policymaking episodes (six EU-level episodes and five episodes each for eight countries), which were spread over the pre-peak period (January 2013–July 2015), the peak period (August 2015–March 2016), which ended with the adoption of the EU–Turkey agreement, and the long-tailed post-peak period (April 2016–February 2020). For the present analysis, I focus on the peak period (first wave) and the post-peak period (second wave) and exclude the two pre-peak episodes (one from Italy and one from the UK) from the analysis.

For the COVID-19 crisis, the team adopted a different procedure because the policymaking processes were not coded retrospectively but in real time, which means it would have been hard to identify policymaking episodes in advance. The team coded all policy actions (irrespective of policy domain) identified in the national press and related to COVID-19 from March 2020 to December 2021. Based on the epidemiological development of the crisis, this period is split into three waves: first wave (March 2020–August 2020), second wave (September 2020–May 2021), and third wave (June 2021–December 2021). For specific analyses, the second and third waves are often combined.

Selection of the Countries

For the refugee crisis, the team selected eight countries: two transit states (Austria and Hungary), two open destination states (Germany and Sweden), two frontier states (Greece and Italy), and two closed destination states (France and the UK).Footnote ⁵ Figure 3 presents the development of problem pressure and political pressure for the selected countries. For the refugee crisis, problem pressure is measured by the monthly submissions of asylum requests as a percentage of the population; political pressure is measured as public salience by Google Trends data.Footnote ⁶ As can be seen from Figure 3a, the two transit states (Austria and Hungary) and the two open destination states (Germany and Sweden) took a hit immediately in the fall of 2015, while the two frontier states (Greece and Italy) and the two closed destination states (France and UK) were initially not confronted with a similar challenge. The Mediterranean frontier states were also directly concerned by the inflow of refugees because, under prevailing EU law and the Dublin regulation, they were responsible for registering and accommodating the incoming refugees. Initially, however, they were spared more severe consequences because of the secondary movements of refugees to other member states. Once these secondary movements were stopped by adopting the EU–Turkey agreement in March 2016, Greece, but not Italy, received more asylum requests, too.

As the figure shows for the refugee crisis, in the transit and open destination states, i.e., in the states most exposed to the secondary movements, public salience peaked with problem pressure early on in the crisis. After the initial peak, both problem pressure and public salience dropped in these countries. In the frontier states, by contrast, we observe multiple peaks of public salience. In Greece, it peaked twice, at the moment of the exogenous shock and then again at the end of the period covered by the crisis, due to internal conflicts on the islands and external pressure from Turkish President Erdogan (see Kriesi et al. Reference Kriesi, Bojar, Altiparmakis and Oana2024: 291–93).

Figure 3 Problem pressure and political pressure by crisis, country, and time.

Problem pressure: monthly submissions of asylum requests as a percentage of the population for the refugee crisis and new monthly deaths for the COVID-19 crisis; political pressure: public salience as indicated by Google Trends. For gathering the Google Trends data, we analyzed the salience in Google searches of the following topics and search terms: Immigration – Topic and Refugee – Search Term for the refugee crisis; coronavirus – Topic, Covid, COVID-19- Search Terms for the COVID-19 crisis. For four countries during the COVID-19 crisis, information is only available for the first period.

In Italy, public salience even peaked three times: pre-crisis, i.e., during the Mare Nostrum rescue operation, which was launched by Italy after an awful shipwreck in the fall of 2013, at the moment of shock in the fall of 2015, and then again in the summer and fall of 2018, when the Minister of the Interior of the new populist coalition government, Matteo Salvini, used the rescue operations of NGOs in the Mediterranean to mobilize public opinion against refugees. However, public opinion was flatlined in the two closed destination states, although a peak of public salience was also visible at the peak of the crisis in the UK.

For the COVID-19 crisis, we covered ten countries representative of the various transnational coalitions that shaped up during the first wave of this crisis: the German–French coalition, three members of the Frugal 4 (Austria, the Netherlands, and Sweden), the UK (still a member during the first two waves), two Southern European (Italy and Spain), and two Eastern European countries (Poland and Romania). For the later waves, we focused on the six largest: Germany, France, Spain, Italy, Poland, and the UK. Four countries have been included throughout both crises (Germany, France, Italy, and the UK) and can be relied upon for more focused comparisons.

For the COVID-19 crisis, I use the number of monthly new deaths to indicate problem pressure.Footnote ⁷ Google Trends data measure political pressure once again. This crisis developed in three waves, as can be seen (see Figure 3b) for the six countries (France, Germany, UK, Italy, Spain, and Poland) for which we have detailed information for all three waves. The timing of the crisis varied, however, by region. The crisis spared Eastern Europe (Poland) in the first wave but hit it all the harder in the third wave, while Southern (Italy and Spain) and Northwestern (France, Germany, UK) Europe were equally hit early on and in the second wave (in fall 2020). The four additional countries (Austria, the Netherlands, Sweden, and Romania), for which we have detailed data on the first wave only, confirm the timing effect: Western European countries all experienced the early shock, although Austria (and Germany) less so than others, while Romania, just as Poland, was spared by it. In economic terms, all countries were hit by the COVID-19 crisis, but Western Europe, especially Southern Europe, fared less well than Eastern Europe. In sum, if the problem pressure in the COVID-19 crisis was more symmetrical than in the refugee crisis, it still varied in timing and the extent of its incidence across countries.

Public salience was generally not closely aligned with problem pressure in the COVID-19 crisis. It peaked in all the countries right at the beginning of the crisis, when the exogenous shock hit. Still, after this initial peak, it oscillated heavily throughout the crisis, reaching similar peaks again during later waves in all countries except Germany, where it declined after the initial shock.

Policy Process Analysis

Having set the scope of our data collection, the first step in constructing the PPA dataset is defining and gathering the media corpus to be analyzed (Bojar et al. Reference Bojar, Kyriazi, Oana and Truchlewski2023). Depending on the level of policymaking, the SOLID team selected either international news sources (for the EU level) or, respectively, national news sources (for the level of the member states). The team used the news aggregator platform Factiva for document retrieval. This platform provided access to many media outlets, allowing for systematic multicountry comparison and transparent and replicable selection criteria on the source. Following good practice in working with media data from methods such as protest event analysis (Hutter Reference Hutter and Donatella Della Porta2014), the team also tried to engage with issues of selection bias (e.g., Earl et al. Reference Jennifer, Martin, McCarthy and Soule2004; Ortiz et al. Reference Ortiz, Myers, Walls and Diaz2006), that is, with the biases associated with news source selection and their coverage of debates, actions, or events. To mitigate such biases, the team adopted several strategies. First, it relied on various media sources, rather than a single source, to capture as many aspects of the policy debates as possible. Second, to mitigate biases related to newsworthiness and proximity, the team selected news sources that are proximate to the level of analysis: For EU debates, it focused on large news agencies and the international press (Agence France Presse, the Associated Press, Reuters, Financial Times, Euronews, ANSA, BBC, MTI), while for national debates it relied on national media. Third, to mitigate biases related to the political motives of the various sources and their potential impact on news coverage, the team selected news sources from different sides of the political spectrum. Consequently, for each country, the team aimed to choose one major newspaper left of center and one right of center regarding ideological leaning (with some minor exceptions related to data availability).

After selecting the news sources, the second decision related to corpus construction consisted in identifying the keywords used to retrieve articles. At this step, one of the primary considerations is achieving a balanced relevance ratio – the ratio between false positives (irrelevant articles that the keyword combination retrieved as positive hits) and false negatives (relevant articles that the keyword combination filtered out as negative hits). Since the data is manually coded, the team aimed for a relatively slim but robust corpus. That is, the corpus ought to be manageable in terms of the number of articles identified in order not to make the coding process too cumbersome and resource-intensive but should still allow capturing the full range of actions in a given country without missing relevant articles filtered out by a too restrictive keyword combination. In practical terms, the selection of keywords was performed by team members in close collaboration with a team of native-language-speaking coders (mostly comprised of political science Ph.D. students also knowledgeable about the subject at hand).Footnote ⁸ At this stage, the team took advantage of the capabilities of the news aggregator Factiva, which allowed us to construct complex search strings using Boolean algebra and its standard logical operators.

After constructing the corpus, the last step in the PPA coding process consists of action coding. As already mentioned, PPA is an event-based methodology; hence, the observation unit at the level at which the data is collected is an action, i.e., “an act or a claim by an actor with a prominent role in the political world that has a direct or indirect relevance for the policy debate” (Bojar et al. Reference Bojar, Kyriazi, Oana and Truchlewski2023). Therefore, actions can be steps in the policymaking process, verbal claims, or protest events within our framework. To measure the various features of actions, action coding is based on a common core of variables that are coded for each of the actions: the arena where the action takes place, its (procedural) form, its (substantive) type of engagement with the policy, its overall direction vis-a-vis the policy, its direction vis-a-vis target actors, the organizational characteristics of the actor undertaking the action and of the target actors, the issues it engages with, and the normative frames used by actors to present their positions to the public (Bojar et al. Reference Bojar, Kyriazi, Oana and Truchlewski2023). Note that while the lowest level of observation is an action, the unit of analysis can be pitched at any level of aggregation (actor types, issue categories, types of countries, etc.) as needed for the analysis.

Datasets

Overall, the corpus for the comparison includes 28,926 policy actions. Roughly, three-fourths relate to the COVID-19 crisis (21,245), and one-fourth (7,681) to the refugee crisis. Most of the actions (84 percent) in each crisis refer to the national level, roughly one-sixth of which are at the EU level. The distribution of the coded actions confirms the unique salience of the COVID-19 crisis, although it is also partly a result of the different procedures applied in the two crises. This corpus constitutes the basic “action file.”

Based on this file, I constructed two additional files for national-level analyses, one each for the two crises, including PPA data and context data for the two types of pressure and the policy output. I have already introduced the indicators used for the problem and political pressure in the previous section – monthly submissions of asylum requests as a percentage of the population for problem pressure in the refugee crisis, and new weekly deaths for problem pressure during the COVID-19 crisis, as well as monthly and weekly Google trends for political pressure in both crises. For policy output, systematic measures are only available for the COVID-19 crisis. There are three indicators: a general indicator for the government’s response, an indicator for the stringency of the government’s lockdown measures, and one for the economic response.Footnote ⁹ For the analysis, I recoded the corresponding indicators, like all the other independent variables, to the 0–1 range, which means that the results will show the maximum effects for each variable.

The unit of analysis in these additional files is the episode-month for the refugee crisis and the country-week for the COVID-19 crisis. I chose monthly units for the refugee crisis because, in this case, data on problem and political pressure are, at best, available every month. A more fine-grained analysis would also not have been possible because of the scarcity of the PPA data during the refugee crisis. We would have ended up with many units with no actions at all had we relied on weekly data. For the COVID-19 crisis, it is possible to use weekly data because all the context variables are available every week, and the PPA data are sufficiently rich for analyses by weeks. I use both a wide and a long form of this dataset. In the wide format, the unit of analysis is the country-week combination, and the PPA indicators are coded by policy domain. In the long version, there is a country-week combination for each policy domain, of which there are six in this crisis (see the next section). Overall, the episode-month file for the refugee crisis includes 551 units, with the number of units per month in the individual episodes varying from a minimum of 4 to a maximum of 34. The country-week file in the wide format for the COVID-19 crisis includes 96 weeks for the six countries we covered in all three waves, i.e., 574 country-weeks and 21 weeks for the four countries covered only in the first wave.

Operationalization of Key Dependent Variables

Previous analyses of policymaking based on PPA data (e.g., Bojar and Kriesi Reference Bojar and Kriesi2023) have mainly relied on indicators for politicization and its components (salience and polarization). I primarily rely on three indicators for the present analysis: salience, centrality, and support. The salience is a system-level property measured by the absolute number of policymaking actions in a given crisis, wave, or policy. Although it is hard to compare across crises because of the different approaches used to collect the data, salience still provides a rough idea of the order of magnitude of the extent to which a crisis dominated the policymaking process at a given time. For most analyses, I recoded the salience indicator per crisis to the 0–1 range. When focusing on the actors’ role in the policymaking process, it is more appropriate to measure the centrality of the actor, i.e., the relative salience of an actor in such a process, which corresponds to the share of all the policy actions in a given policymaking process for which the actor is responsible. I do not use the concept of polarization in the present analysis because it is primarily a system-level property. Instead, I replace it with the concept of support, which has the advantage that it is easily applied to both the system level and to individual actors. It corresponds to the average direction of the policy-specific positions taken by the actors in a crisis, a wave, or a policy domain at a given level, measured on a scale ranging from –1 (complete opposition to) to +1 (full support of) the policy proposal. For policy output, we only have the three indicators already introduced at the national level in the COVID-19 crisis. Otherwise, we have to rely on qualitative information.

I recoded the multiple issues into a limited set of crisis-specific policy domains, as shown in Table 2. For the COVID-19 crisis, I distinguish between six policy domains – economic, fiscal, lockdowns (restrictions and relaxations), public health, vaccination, and institutional questions (questions mainly concerning emergency powers, rules of civil liberties, suspension of institutional procedures such as parliamentary sessions, or postponement of elections). At the EU level, economic policies mainly refer to single market policies, while fiscal policies mainly concern the RRF. As shown in (a) in Table 2, in the COVID-19 crisis, economic and fiscal policies and vaccination policies predominate at the EU level, while financial, public health, and lockdown-border closure measures prevail at the national level. For the refugee crisis, we can distinguish between border closures and asylum-integration policies, which include the bulk of policymaking at the national level and the three most important EU-specific policy episodes (EU–Turkey agreement, relocation scheme, and Dublin reform).

Data Analysis

The data analysis mainly relies on straightforward descriptive tools, ranging from bar charts to OLS regressions. The independent variables include the type of crisis (COVID-19 vs refugee crisis), the polity level, the period (first to third waves), country, problem and political pressure, actor type, and policy domain. The basic actor typology distinguishes between eight types: EU-polity actors, governments, governing and opposition parties, local–regional authorities, experts, and civil society. EU-polity actors include the EU Commission, other EU actors (supranational actors), and the governments of (other) member states (transnational coalitions). When not mentioned otherwise, I use the action file. For analyses, including indicators for the problem and political pressure, I rely, depending on the crisis, on the episode-month and the country-week files. Given the prevailing short series in the episode-month file for the refugee crisis, the analyses report contemporaneous effects in a given month, i.e., I do not analyze longitudinal dynamics for this crisis. Given the extended series for the six countries covered throughout the three waves, relying on more complex dynamic panel analyses for the COVID-19 crisis is possible.

The results will be presented in three sections. The next section analyzes crisis policymaking from the perspective of salience and centrality, i.e., from the standpoint of the prominence of the overall process and the role played by specific actor types in the process. Section 5 analyzes policy support, i.e., the conflicts involved in the policymaking process and their consequences. Section 6 will focus on policy output.

Salience

To begin with, I want to document the great salience of the immediate response of policymakers at both levels in the two crises, especially at the EU level (H1), the overwhelming salience of the COVID-19 crisis (H1a), and the declining salience in the course of the two crises (H1b). Figure 4 presents the average monthly salience of policy actions during the three (COVID-19) and two (refugee crisis) waves of these crises. For the member states, I calculate the average across member states for which we have complete data for all waves. All expectations are confirmed. First, salience immediately surged early in the two crises (H1), especially at the EU level. EU policymaking during the first wave was much more salient than policymaking in the average member state during the same period. The results confirm that, in the EU polity, crises pose a particularly intense challenge for policymakers at the EU level: They have to come to the rescue of hard-hit member states (in asymmetric crises such as the refugee crisis) or of member states lacking the capacity to respond to the crisis (in symmetric crises such as the COVID-19 crisis) to maintain a level-playing field. They must coordinate the member states’ response and build additional supranational capacity to deal with the crisis.

Figure 4 Salience by level, wave, and crisis: number of actions.

Second, policymaking in the COVID-19 crisis was much more salient than in the refugee crisis (H1a). Third, in both crises, the salience in later waves declines as expected (H1b). The decline is particularly noteworthy at the EU level during the refugee crisis. After adopting the EU–Turkey agreement, which concluded the first phase of the crisis, EU policymaking became much less conspicuous. In the COVID-19 crisis, salience declines at the national level only in the third wave. But even if its salience declines, national policymaking remains much more prominent than ever during the refugee crisis. The second and third waves of the pandemic kept policymaking focused on the crisis. This impact documents the extent to which Europe was in the thrall of the virus for more than a year.

Figure 5 provides a more detailed time line for the salience of crisis-specific policymaking. It presents three-weekly (COVID-19) and three-monthly (refugee crisis) moving averages of the number of actions.Footnote ¹⁰ The vertical lines indicate the waves in the case of the COVID-19 crisis and the adoption of the EU–Turkey agreement in the case of the refugee crisis. We observe that, in the COVID-19 crisis, policymakers reacted immediately at both levels, even more so at the EU level, as the exogenous shock hit. Policymaking activity reaches its peak almost instantly and then declines throughout the first wave of the crisis, more rapidly at the member state level than at the EU level, where the negotiations about the RRF kept policymakers busy until the end of July 2020. Policymaking resumed again during the second wave, but now it is relatively more salient in the member states. While the EU provided short-term and even long-term economic support during the first wave in line with its essential competence in this domain – roughly two-thirds of its policymaking during this wave focused on economic policies (H1d), it turned to support in vaccination during later waves – almost half of EU policymaking in wave two and roughly a third in wave 3 concerned vaccination, a policy domain, where the EU had acquired new competences in the course of the crisis. During the third wave, policymaking at both levels drops off, but more so at the EU level, since the EU has now supported both policy domains of crucial concern during the COVID-19 crisis.

Figure 5 Time line of the salience of crisis-specific policymaking, by crisis and level: absolute weekly number of actions, smoothed.

In the refugee crisis, there was a first attempt at crisis policymaking at both levels in the spring of 2015. But this aborts because of political resistance. When the crisis hit in earnest at the end of the summer of 2015, policymakers responded immediately at both levels, especially at the EU level. At this level, salience is double-peaked during the first wave. The two peaks correspond to the solutions proposed by EU policymakers – the relocation scheme and the EU–Turkey agreement. First, the EU focused on the relocation scheme, which was responsible for almost half of the actions during the first peak. Only when this solution failed because of the resistance of the V4 coalition did the EU home in on its plan B, the EU–Turkey agreement, which is responsible for almost two-thirds of the action during the second peak. In the long second wave of the refugee crisis, we observe a minor third peak at both levels in the summer of 2018, which corresponds to the attempt of the German Chancellor Merkel to find a solution to her domestic intraparty conflicts at the EU level, an attempt that failed ignominiously (see Kriesi et al. Reference Kriesi, Bojar, Altiparmakis and Oana2024).

As we have seen, the member states experienced similar ups and downs in the levels of problem and political pressure during the COVID-19 crisis but faced different levels of such pressure in the refugee crisis (see Figure 3): Only the transit and open destination states experienced a cumulation of both types of pressure during the peak wave of the crisis. The frontline and closed destination states were less pressured. The question is whether this difference in pressure leads to differences in the salience of policymaking (H1g). Figure 6 provides a first answer to this question. It presents the time lines of the salience of crisis-specific policymaking by type of member state in the refugee crisis. As expected, policymaking picked up in response to the external shock in the transit and open destination states in the summer of 2015, and it decreased in these countries once the problem pressure was somewhat constrained after adopting the EU–Turkey agreement. However, during this peak period, policymaking also increased in the frontline and closed destination states, which have hardly been exposed to problem pressure (as measured by the number of asylum requests). In the frontline states, indirect problem pressure may have also existed during the peak period: In Greece, most notably, even if they did not stay on, hundreds of thousands of refugees traveled across the country toward northern European destinations. Other member states temporarily closed their borders so as not to be directly concerned, or they used the occasion to tighten their asylum policy. During the long second wave of the crisis, we find peaks of policymaking in both types of member states, too. They are even more pronounced in the frontline and closed destination states. This pattern of policymaking salience suggests that it is not exclusively responding to problem pressure but is also likely to be induced by political entrepreneurs (H1f).

Figure 6 Time line of the salience of crisis-specific policymaking in the refugee crisis by type of member state: absolute weekly number of actions, thrice-weekly averages.

We pursue this question more systematically by analyzing the impact of our indicators for the problem and political pressure on crisis-specific policymaking during the different phases of the two crises at the national level. For this analysis, I use the episode-month and the country-week files, which provide monthly/weekly data on the number of actions. I combine second and third waves and distinguish between border control and asylum-integration policies for the refugee crisis and between border control-lockdown measures and all other policies for the COVID-19 crisis. I interact the two types of pressure with wave and policy to identify wave- and policy-specific effects. In addition, I introduce wave- and country-fixed effects and country-wave interactions, which allow controlling for all wave and country characteristics that are not directly related to the two types of pressure as measured by our indicators. The effects indicate the pressure-related variation within waves and countries regarding the salience of crisis-specific policymaking. For this analysis, the salience measure has been rescaled to the 0–1 range, which allows us to compare the effects across crises. Given that the pressure indicators are also rescaled to the 0–1 range, the effect parameters show the maximum impact of the two types of pressure on the salience of policymaking.

Figure 7 presents the results of this analysis.Footnote ¹¹ Its first part refers to the variation of the effects across waves, while its second part presents the corresponding variation across policies. As is shown by the left-hand graph of Figure 7a, the direct impact of problem pressure on the salience of policymaking goes in the expected (positive) direction. Still, it is comparatively weak or nonsignificant in all waves during both crises. By contrast, the right-hand graph of this figure shows that the effect of political pressure (as measured by public salience) is substantial. This result confirms H1c. In the COVID-19 crisis, the impact of political pressure is most marked in the first wave and then decreases in the second and third waves. Although the effect dropped during the COVID-19 crisis, it was still quite sizeable in later waves: The renewed pressure during the later waves created new preoccupations in the public, which had a continued effect on the salience of policymaking. In the refugee crisis, the effect does not change much in the second wave, which is indicative of the impact of political entrepreneurs: In the absence of any effect of problem pressure, the continued impact of political pressure, as indicated by public salience, is very likely an endogenous effect of the politicization of the crisis in some countries – most notably in Hungary and Italy, but also in Germany.

Figure 7 Effect of problem pressure and political pressure (public salience) on the salience of policymaking: Average marginal effects.

Given that problem pressure and political pressure are linked, it is possible that problem pressure had an indirect impact on the salience of policymaking via political pressure (see Figure 2). An analysis of such a possible mediation effect (not shown) indicates that problem pressure does have a weak but significant indirect effect on the salience of policymaking during the peak phase of the refugee crisis and a highly significant impact during the second phase of this crisis, making for a still relatively weak, but significant total impact in the peak phase and a more substantial, but imprecise total effect in the second phase. In the case of the COVID-19 crisis, there are hardly any indirect effects at all, however. This means that even if we take indirect effects into account, the overall effect of problem pressure on the salience of policymaking is still comparatively weak, in any case, much weaker than the effect of political pressure. The weakness of the indirect effects of problem pressure contradicts our expectations and the structuralist reasoning.

Figure 7b shows that to the extent that problem pressure has any effect, it is limited to the salience of policymaking on lockdown measures in the COVID-19 crisis. It does not affect other policy measures in this or the refugee crisis. As for the impact of political pressure, it is equally important for both types of policies in the refugee crisis but makes a big difference in the COVID-19 crisis: It has a powerful effect on lockdown measures but a much weaker effect on all other policy measures. These policy-specific effects during the COVID-19 crisis are in line with H1d: The governments of every country immediately reacted to pressure from the public by taking border control and lockdown measures. In contrast, the remaining measures were less directly responsive.

Finally, I also find country-specific differences concerning the salience of policymaking (H1g), which are not captured by our other indicators. Thus, during the refugee crisis, policymaking was most intense in Greece and least intense in the UK and Sweden during the peak phase: Greece, indeed, faced an exceptional kind of problem pressure, which is not captured by our indicator since most of the refugee arrivals moved on and did not stay in Greece. As a nonmember of Schengen, the UK was least concerned, and Swedish policymakers reacted somewhat belatedly. We observe no country differences during the second wave of the refugee crisis. In the COVID-19 crisis, policymaking was most salient in Italy and Poland during the first phase. The pressure was extreme in Italy, where the pandemic hit first and particularly hard. As for Poland, eastern European countries generally reacted strongly in the first wave, i.e., were locking down early relative to the number of deaths caused by the virus. At the same time, they changed course later on, relaxing lockdowns and shifting to less restrictive measures with devastating effects on the mortality rates. Popic and Moise (Reference Tamara and Moise2022) explain this particular response pattern in Eastern Europe as a combination of weak healthcare systems (inducing the governments to choose early lockdowns) and vulnerable economies (causing them to relax lockdowns more later on). In the second and third waves, policymaking in Poland was, indeed, least salient. At the same time, it was most salient in Germany, where the government struggled with the member state governments to impose its policy (Alexander Reference Alexander2021).

The Role of the Different Actors

The second set of results refers to the role of different actors in crisis management, which can be studied at both levels of the EU polity. At the national level, we can only rely on the four countries for which we have complete data on both crises. The results confirm the vital role of executive decision-making in crisis policymaking. Executive actors account for almost half (48 percent) of all policymaking action, and government executives account for no less than one-fourth (25 percent). In the absence of a benchmark outside of the two crises with which to compare these values, a German episode from the refugee crisis may serve as a proxy: In 2016, German policymakers adopted a new integration law, which was an essential piece of legislation that introduced an innovative new law, but which was only tangentially related to the refugee crisis and which in many ways represented a typical German legislative process. This episode mainly led to partisan conflicts between the members of the grand coalition and between them and the opposition. Executive decision-making accounted for 15 percent of the overall action in the episode, and government executives alone were responsible for 12 percent of all the actions. Taking these values as benchmarks, it is possible to appreciate executive decision-making’s much more critical role in the two crises, confirming H2. The overall shares of this type of decision-making were very similar in both crises – 49 percent in the COVID-19 crisis and 45 percent in the refugee crisis. The corresponding shares for government executives were identical.

Figure 8 presents the percentage shares of executive actors – overall and governmental executives – in the two crises per wave and polity level. Comparing levels and crises jointly, we observe, as expected, that executive decision-making is more important at the EU level during the refugee crisis and more critical at the national level during the COVID-19 crisis. The differences are quite significant for the refugee crisis – 60 percent executive decision-making at the EU level compared to 42 percent at the national level – but more limited for the COVID-19 crisis – 43 percent at the EU and 50 percent at the national level. These results confirm that conflict management at the EU level primarily involves executive decision-making. In contrast, at the national level, parties play a more significant role in conflict management (H2a). Contrary to expectations (H2b), the share of executive actors generally remains quite stable across waves and crises, except for the EU level during the COVID-19 crisis. Even in later waves of a crisis, executive policymaking remains the prevailing decision-making mode.

Figure 8 Centrality of executive actors by crisis, wave, and level: percentage shares.

More generally, given the predominance of the intergovernmental coordination mode of decision-making, EU actors and national governments almost completely dominate crisis policymaking at the EU level. Their domination becomes apparent in Figure 9, which presents the centrality of actor types by crisis and level. These two actor types dominate in both crises. Still, their relative importance varies by crisis: EU actors were relatively more important during the COVID-19 crisis because this crisis concerned policy domains where the EU has greater competences (H2d). By contrast, governments were more significant in the refugee crisis, where they heavily clashed with each other, and transnational conflicts were more prominent (H2e). Among the remaining actors, only third countries and supranational actors, as well as civil society organizations, are of some significance at the EU level. Third countries and other supranational actors were of greater importance during the refugee crisis, which reflects the fact that the EU externalized problem solutions to third countries during this crisis, most notably to Turkey, but also to Libya, or Morocco, as well as the fact that the EU was heavily clashing with UNHCR. The European Parliament and the parties represented there played virtually no role in crisis management at the EU level.

Figure 9 Centrality of actor types by crisis and level: percentage share.

At the national level, governments are also the dominant actors, which align with the importance of executive decision-making. There are no differences in this respect between the two crises. National governments dominate crisis policymaking domestically, independently of the crisis. However, other actors are also important at the national level, albeit to different degrees, depending on the crisis. Thus, as expected by H2f, parties and EU-polity actors play a more significant role during the refugee crisis, reflecting the crisis’s more conflictual character. By contrast, in the COVID-19 crisis, regional and local actors, experts, and civil society organizations stand out.

It is possible to investigate the impact of the competence distribution on the role actors play at the EU level more closely by distinguishing between three aggregate actor types: EU authorities, member state governments, and all other actors. Figure 10 presents the centrality of these three types of actors for the most important policies at the EU level. The policies are ordered from left to right according to the policy-specific centrality of EU authorities. As it turns out, the centrality of the EU authorities does not quite align with the policy-specific competence distribution between the two levels, although the latter has an impact. Thus, EU authorities are most central in the single market domain, where they have most competence, while their role is most restricted in the case of border controls during the COVID-19 crisis, where the member states reigned supreme, benefiting from the exception rules in the Schengen scheme. The EU authorities also played a more limited role in the two most significant EU policies during the refugee crisis – the relocation scheme (plan A) and the EU–Turkey agreement (plan B), where member states enjoyed much greater autonomy. However, EU authorities have been central in public health, vaccination, and fiscal policy (RRF) – i.e., in domains with more limited competence. The reason for the differences between formal competence distributions and the observed centrality of EU authorities in the latter domains may be linked to “centralized coordination.” This mode of coordination, which enhances the role of supranational actors while preserving the monitoring capacity of the member states, played a vital role in the joint vaccine procurement (Becker and Gehring Reference Becker and Gehring2023), in capacity building in the public health domain (Ferrera, Kyriazi, and Miró Reference Ferrera, Kyriazi and Miró2022), and in the negotiations for the RRF (Schelkle Reference Waltraud.2021).

Figure 10 Policy-specific centrality of actors at the EU level: percentages.

The transnational coalitions are also expected to vary by crisis. Figure 11 presents the respective results. As we already know, EU actors dominate more in the COVID-19 than in the refugee crisis. The main reason for their lesser presence in the refugee crisis is the important role played by third countries and other supranational actors, who are hardly present in the COVID-19 crisis. However, the prominence of the member states hardly varies between the crises at first sight. The multiple coalitions are almost equally present in both crises. The main differences concern the German–French coalition and the V4 coalition. The difference concerning the former does not become visible in the graph since it consists of the changing relative importance of the two countries in the two crises: As expected, Germany and the V4 coalition played a more significant role during the refugee crisis, whereas France was significantly more important during the COVID-19 crisis. The critical role of France corresponds to the increasing importance of the German–French coalition during the COVID-19 crisis, and the critical role of Germany and the V4 coalition reflects the driving role of Germany in the search for a joint solution to the refugee crisis, which met with the staunch opposition from the V4 coalition. Next to the German–French coalition, the southern Europeans constituted the most crucial coalition in both crises. They consistently asked for more solidarity among the member states, first in the context of the reform of the Dublin regulation, and then in the context of the economic catastrophe created by the COVID-19 crisis, especially in southern Europe. Similarly, the Frugal 4 were also present in both crises but to a lesser degree than the southern European countries. Although the graph does not capture this, their role was undoubtedly more significant during the COVID-19 crisis. The UK consistently played a minor role, as did all the other member states.

Figure 11 Centrality of transnational coalitions at the EU level, by crisis: percentage shares.

Finally, we briefly look at the role played by the various actor types at the national level (H2f). Figure 12 shows the corresponding variations for the four countries for which we have detailed information on both crises. The government is the most critical actor in both crises in all countries, but its role is comparatively weak in Germany during the COVID-19 crisis. Its weakness is a result of German federalism, which was crucial in managing this crisis in Germany (see Alexander Reference Alexander2021; Truchlewski et al. Reference Zbigniew, Oana, Moise and Kriesi2023). Conversely, the government played a vital role in Italy during the refugee crisis. As already mentioned, this derives from its composition in the later phases of this crisis: With the accession to power of the populist government in the summer of 2018, the government or, more specifically, the Minister of the Interior, Matteo Salvini, from the Lega, took over the politicization of the refugee crisis in an attempt to exploit it for his political purposes.

Figure 12 Centrality of actor types at the national level during the two crises: Germany, France, Italy, and the UK compared – percentages.

Finally, party interventions were particularly numerous in Germany in both crises, but especially in the refugee crisis, where the governing parties were deeply divided over the policies to be followed. As a result, the German government was divided over its border control and asylum measures. At the same time, it supported the (external) border control and asylum measures at the EU level, which ought to allow it to overcome its internal divisions.

Conclusion

Regarding the salience of crisis policymaking and the centrality of specific actors in the policymaking process, there are many similarities between the two crises, even if policymaking during the COVID-19 crisis was much more salient than during the refugee crisis. In both crises and at both levels of the EU polity, the policymakers immediately responded to the exogenous shock. In both crises, they continued to respond to political pressure (i.e., public salience) throughout the crises, while the effect of problem pressure on policymaking was much more limited. Executive decision-making has characterized both crises, too. Due to the prevailing decision mode, EU authorities and member state governments dominated policymaking at the EU level. At the national level, it was the government that dominated. Finally, in both crises, EU-level policymaking was much more salient than policymaking in the average member state at the outset of the crises.

Differences between the two crises may appear secondary compared to these similarities. In later phases of the crises, policymaking tended to decline – more so during the refugee crisis than in the COVID-19 crisis, where later pandemic waves kept the policymakers busy. At the EU level, the relative importance of EU authorities and governments varied by crisis and policy domain, depending on the policy-specific competence distribution – a greater role for EU actors in economic policymaking during the COVID-19 crisis, a greater role of the governments (especially the German and V4 governments) and third countries during the refugee crisis. At the national level, the types of actors who played a central role in addition to the governments varied considerably: EU-polity actors and parties in the refugee crisis versus local–regional governments, experts, and civil society actors in the COVID-19 crisis. Moreover, political entrepreneurs played a significant role in some countries during the later phases of the refugee crisis, while their role was more limited in the COVID-19 crisis. As it turns out, these differences were highly consequential for policy support, which we will study in the next section.

The Role of Executive Actors and Exogenous Pressure

Following Schattschneider (Reference Schattschneider1975), I expect policy support to increase as a function of the extent to which executive actors control the policymaking process (H3). This expectation is based on two claims. First, it assumes that policy support by executive actors is decisive for overall support. In line with this assumption, the correlation between general and executive actors’ support is very close. At the EU level, it varies between .84 for the six episodes of the refugee crisis and .98 for the six-issue domains in the COVID-19 crisis. The correlations for the episodes at the member state level are .89 and .92, respectively, for the two crises. Second, it assumes that executive actors generally lend more support to the policies than other types of actors. The first row of graphs in Figure 14 shows that this assumption holds for the national level in both crises but not at the EU level. At the EU level, there seems to be little difference between the executive and other actors in this respect in both crises. As the two graphs in the second row show, however, the support by executive actors varies by policy at the EU level, depending on the contentiousness of the policy. Thus, in the case of the two most contentious episodes during the refugee crisis, the reform of the Dublin regulation and, even more extreme, the relocation scheme, executive actors were more opposed to the proposal than the other actors. This results from conflicts at the EU level, mainly between governments of member states, represented by the top executives of transnational coalitions in joint decision-making. By contrast, executive actors favored the consensual EU–Turkey agreement more than other actors. In the consensual COVID crisis, support by executive actors was higher than support by different actors in two domains – vaccination and border control; in the other policy domains, there was hardly any difference in the support between the two.

Figure 14 Support of crisis-specific policies by crisis, level, and executive actors: average level of support.

At the national level, it is possible to systematically analyze the effect of problem pressure, political pressure, salience, and the role of actors on policy support using the episode-month and the long country-week files. For the COVID-19 crisis, but not for the refugee crisis, we expect problem and political pressure to enhance policy support by policymakers throughout the crisis (H3c). In contrast, more significant pressure in the refugee crisis will likely lead to more intense politicization and enhance the underlying conflict. The systematic analysis bears out these expectations. However, as shown in Figure 15Footnote ¹² for the COVID-19 crisis, greater support only occurs when problem pressure and public salience combine to exert a powerful thrust on policymakers. In the absence of public salience, problem pressure hurts policy support. However, if both types of pressure are intense, policymakers rally strongly around government proposals in the COVID-19 crisis. As expected, for the refugee crisis, we do not find any effects of pressure on support, whether positive or negative.

Figure 15 Effects of problem pressure, public salience, and crisis characteristics on the support of policymaking, the refugee crisis, and COVID crisis, predicted contemporary (refugee crisis) and long-run (COVID-19 crisis) support.

In addition, this analysis confirms that executive actors enhance the overall policy support in both crises (H3). I did not formulate any hypotheses regarding the effect of salience of policymaking, but the results indicate that it is not associated with policy support. The two vary independently: Intensely debated policy proposals may be equally supported/resisted as proposals that pass virtually unnoticed. Finally, these results also show that in the refugee crisis support is more significant in later waves, contrary to the rallying effect, but in line with the notion that this crisis hit a policy domain that was already highly politicized initially. Moreover, asylum-integration policy was less supported than border control measures during this crisis.

Policy Support at the EU Level

I have argued that member states are the critical actors at the EU level and that their resistance is policy-specific and organized into transnational coalitions (H3e). We can approach this issue in two steps: First, I show the policy-specific support by three types of actors – EU authorities, member state governments, and other actors, the role of which I already analyzed in Figure 10. Then, I document, in more detail, the support of the different policies by the transnational coalitions. Figure 16 presents the average policy-specific support by the three actor types and clarifies who was the most critical for EU proposals. The policies are the same as in Figure 10, but they are now ordered from left (low support) to right (high support) according to the support they received from member state governments. Unsurprisingly, EU authorities supported all the proposals except border control during the COVID-19 crisis, where they did not have much to say. In this exceptional case, they could only warn the member states that border closures would run the risk of fragmentation of the single market and undermine free movement and supply chains.

Figure 16 Support of EU policies by three types of actors: average support.

Support by member states, however, varies substantially from the relocation scheme and the Dublin reform, which were opposed by a majority of the member states or received not much support from them, to the border control policies in the COVID-19 crisis, and the EU–Turkey agreement, which they widely supported. The contrast between the relocation scheme (plan A) and the EU–Turkey agreement (plan B) during the refugee crisis is striking. Having rejected the EU’s plan A, the member states came around in support of the EU’s plan B, but they stuck to their opposition concerning the relocation scheme and the long-term Dublin reform. They supported a stop-gap solution (the EU–Turkey agreement, as well as the ECBG and the hotspot approach (not shown in Figure 16)), which alleviated the problem pressure but reduced the incentives for long-term reform. In contrast, they endorsed proposals in all policy domains in the COVID crisis, contributing to successful problem-solving.

Figure 17 presents more details regarding the positions of transnational coalitions in selected policy domains. I have chosen the six most important policy domains: single market policies, the RRF, vaccination policy for the COVID-19 crisis, the relocation scheme, the EU–Turkey agreement, and Dublin reform for the refugee crisis. The figure presents the average policy-specific support by the various transnational coalitions, provided they undertook more than three actions in a given wave and policy at the EU level. For the Dublin reform and vaccination policy, I do not distinguish by waves because most of the action in these domains centered on the second/third waves. I keep Germany and France apart because they do not always share the same positions.

In the generally consensual COVID-19 crisis, the single market policies were most consensual across transnational coalitions, except for Germany, which appears least supportive of the economic measures at the EU level. The German exception is primarily due to the German Constitutional Court ruling, which squeezed the ECB because of its bond-buying programs.Footnote ¹³ In the case of the RRF, the situation is more mixed (see Table 3). There was, after all, opposition to adopting the RRF, which rendered the negotiations at the July 2020 summit most difficult (Schelkle Reference Waltraud.2021; Truchlewski et al. Reference Zbigniew, Oana, Moise and Kriesi2023: Chapter 8). The Frugal 4 opposed the proposal in the first wave, and during the ratification phase (second wave), the V4 coalition opposed it even more vigorously. However, the RRF received strong support from the EU authorities, the German–French leadership, and the southern European member states. In the negotiation process, the Frugal 4 coalition gave up its opposition to the grant component of the fund and no longer objected to the proposal in the ratification phase of the RRF, as shown in Figure 17.

Figure 17 Policy-specific support by actors and wave, at the EU level: average support.

Hungary was the main protagonist in the conflict around the ratification of the RRF, which was dominated by the rule-of-law debate (Bohle, Greskovits, and Naczyk Reference Bohle, Greskovits and Naczyk2023). Poland was its closest ally, and the rest of the member states opposed both countries. Together, the two countries objected to linking the attribution of grants and loans in the framework of the RRF to the rule-of-law mechanism. This link was proposed by the Commission and supported by most member states and the European Parliament (EP). Resistance also came from other actors, particularly opposition parties in the first wave. But this opposition was negligible. Third, the vaccination graph shows broad support for the joint procurement policy. The UK government was the only actor opposed to this policy, but its opposition did not matter since it was already on the way out of the Union.

The relocation scheme starkly contrasts with these three policy domains from the COVID-19 crisis. As the respective graphs show, it was widely supported by the member states (in addition to help from the EU authorities), at least during the first wave of the refugee crisis, i.e., the peak phase. However, it met with extreme and very vocal opposition from the V4 countries, which boycotted any implementation of the joint scheme. Once the scheme had failed, support for it deteriorated during the later phases of the crisis among all member states, even in Germany. Contrary to the relocation scheme, the EU–Turkey agreement did not meet with strong resistance. Germany was the driving force behind it, and the southern frontline states supported it. Many member states did not take strong positions on this issue: France, for example, is not represented in the graph because it hardly at all participated in the debate on the deal with Turkey. The V4 coalition was also somewhat opposed to this agreement, but its opposition was not as vocal as in the relocation scheme. Many critiques came from the Frugal 4 (essentially from Austria), but this critique came late and was not decisive. Austria warned against dependency on Turkey, criticized human rights violations in Turkey, and urged an end to accession talks with Turkey, calling them a “diplomatic fiction.” Finally, the reform of the Dublin regulation met with broad resistance from the southern European frontline states, northern destination states, and the V4. This cumulated resistance gave any reform proposal very little chance.

The refugee crisis, in particular, demonstrates how domestic and EU policymaking may be closely linked in a crisis, with reciprocal spillover effects (H3e). Figure 18 illustrates this point. It presents the estimated policy support of various actor types for each crisis at the two levels.Footnote ¹⁴ First, the figure documents the German predicament during the refugee crisis. Because of its internal divisions, the German government (“Germany”) was not supportive of its national policy proposals but highly supportive of the EU-level proposals. It strongly promoted the EU proposals as a substitute for its contentious national policies. As the German government was internally divided and struggled to deal with the crisis domestically, it pressured the other member states to get the relocation quota passed in the Council of the EU to alleviate its burden. However, the adoption of the relocation scheme, pushed through by the Germans in September 2015, turned into a fiasco, from which European asylum policy never recovered (van Middelaar Reference van Middelaar2017: 110). From this point on, in Eastern Europe, accepting refugees was viewed not as a humanitarian act but as a submission to Berlin. It was only once Germany had failed to obtain a working relocation scheme from its fellow member states that it turned to an agreement with Turkey as the second-best solution. In the summer of 2018, when the border control issue returned to German politics, German policymaking once again spilled over to the EU level (see Figure 5).

Figure 18 Policy support by actors for national and EU policies during the COVID-19 and the refugee crisis: predicted support.

Figure 18 shows the opposite combination for the V4 countries, where the governments strongly endorsed their domestic policies during the refugee crisis but forcefully opposed the corresponding policies at the EU level. While the German predicament undermined the government’s leadership role at the EU level, the solid domestic stance of the V4 governments strengthened their critical hand at the EU level. The figure also shows that the Frugal 4 (i.e., primarily Austria) were generally critical of the EU’s policies during the refugee crisis. At the same time, their governments enjoyed strong support at the national level. In addition, the figure indicates comparatively low support of national policies by EU actors and the opposition of transnational actors to national policies during the refugee crisis. Both types of actors from the EU polity had to deal with the negative externalities of unilateral policies, as already pointed out before. Compared to the refugee crisis, such spillover effects were much more contained during the COVID-19 crisis because of the reciprocal character of the externalities of member state actions. In the COVID-19 crisis, policymaking at the two levels of the EU polity was segmented, which is an additional reason for the absence of conflict among policymakers in this crisis. The only two exceptions concern the spillover of the rule-of-law crisis into the ratification of the RRF (which could be contained) and the unilateral British vaccination policy (which had little effect since the UK was already on its way out).

Policy Support at the Member State Level

Moving to the domestic level, I analyze the support by the various actor types in the two crises based on the four countries, for which we have complete data on both crises. The overall results, controlling for country, wave, and policy domain, are presented in Figure 19 for six actor types: EU-polity actors, governments, regional and local authorities, governing and opposition parties, and civil society actors.Footnote ¹⁵ For governments and ruling parties, I additionally distinguish between Germany and the three other countries because of the specific intragovernmental conflicts in Germany during the refugee crisis. I do not show results for the COVID-19 crisis for EU-polity actors because there are too few cases – a result of the just mentioned segmentation of the two levels in this crisis. The results essentially, although not wholly, confirm H3f, which expects greater support among all nonexecutive actors during the COVID-19 crisis: regional and local actors, civil society actors in general, and the German government/German governing parties in particular are, indeed, more supportive of the government in the COVID-19 than in the refugee crisis. For governments and governing parties other than the German ones and opposition parties, there are only minor differences between the two crises: Governments generally support their proposals independently of the crisis in question, and opposition parties oppose them, undaunted by whether the crisis poses an existential threat to the country or not. However, in line with expectations, opposition parties are less vocal during the less conflictual COVID-19 crisis – they account for only 7.3 percent of all actions, compared to 12.4 percent during the refugee crisis. As we have already seen, EU-polity actors are somewhat critical of national policies in the case of the refugee crisis.

Figure 19 Policy support by actor type and crisis: predicted positions.

Finally, I turn to country differences (H3g). As shown in Figure 20, the government’s policy proposals found more support in the COVID-19 crisis than in the refugee crisis in three out of the four countries. The differences are particularly stark in Germany, where the government proposals found broad resistance in the refugee crisis but (weak) support during the COVID-19 crisis, and in Italy, where the support of the government’s proposals was also fragile during the refugee crisis. The exception is France, where the government’s proposals were equally supported in both crises. In the two crises, the French government received roughly average support (indicated by the dashed line). This is not surprising as far as the COVID-19 crisis is concerned. For the refugee crisis, the relatively strong support of the French government is likely to be explained by the fact that France was a closed destination state, which was not directly concerned by the inflow of refugees in 2015–16. The same applies to the UK, which was also a closed destination state and which, as a nonmember of the Schengen area, was even less concerned by the refugee crisis than France. In the case of the UK, what is somewhat surprising is the government’s strong support during the COVID-19 crisis. This strong support is surprising because the UK initially faced a worse pandemic than other European countries, partly because of the government’s fault. The British government did change course after initially accepting the virus’s spread, which would build herd immunity. We can only speculate that the high level of support for the British government’s policies was linked to the Brexit negotiation process, which overlapped with the COVID-19 crisis: The UK officially left the EU on January 31, 2020, which opened a transition period that was going to last until the end of 2020, i.e., deep into the second COVID-19 wave, during which the two parties – the EU and the UK – negotiated the Trade and Cooperation Agreement that was to regulate their future relationship (Laffan and Telle Reference Laffan and Telle2023: Chapter 8). It is likely that, in the UK, this ongoing negotiation process reinforced the rallying effect of the COVID-19 crisis beyond the first wave.

Figure 20 Policy support by crisis and country, mean support.

Figure 21 presents the support of the government’s proposals per wave in the two crises. We observe a rallying effect early on during the COVID-19 crisis in all four countries. For reasons I have just speculated about, this rallying effect was kept up in the UK during the later waves of the crisis. The continued rallying effect in the UK may have also resulted from the fact that the UK was initially more effective in vaccine procurement than the EU (Becker and Gehring Reference Becker and Gehring2023). By contrast, there was no rallying effect during the refugee crisis in any country. The seeming rallying effect in Italy is primarily a consequence of a change in governments during the crisis and of the fact that it was most concerned with reforming the Dublin regulation, which only became an important issue during the second part of the crisis.

Figure 21 Policy support by crisis, country, and wave: mean support.

Conclusion

Regarding policy support, the differences between the two crises prevail. The existential threat in general, the relative symmetry of the crisis incidence transnationally, and the comparative absence of partisan politicization domestically made the COVID-19 crisis much more consensual than the refugee crisis. The consensus manifests itself at both levels of EU policymaking, especially early on when there is a rallying effect in the COVID-19 crisis, but not in the refugee crisis. To be sure, we also observe a rallying effect at the EU level during the refugee crisis, but it is deceptive because it extends only to the second-best solutions, whereas the preferred solution met with powerful resistance from some transnational coalitions. At the national level, combined problem pressure and public salience enhanced policy support in the COVID-19 crisis but not in the refugee crisis. Moreover, all nonexecutive actors supported the government’s proposals more during the COVID-19 pandemic than during the refugee crisis, except for the partisan opposition, which was, however, less vocal during the pandemic than during the refugee crisis. Policy support at the two levels is closely linked by spillover effects in the refugee crisis but not in the COVID-19 crisis, where policymaking at the two levels was broadly segmented. The spillover effects during the refugee crisis reflect the conflictive character of policymaking in this crisis. In contrast, the segmentation of policymaking contributes to the consensual character of policymaking in the COVID-19 crisis.

There is one similarity between the two crises: Policy support increases as the executive actors control the policymaking process – at the national level. The situation is more complicated at the EU level since the national governments call the shots at the EU level, and their top executives represent them. Executive support at the EU level depends on the contentiousness of the policy proposal. Thus, in the highly conflictual policy proposals at the EU level – the relocation scheme and the reform of the Dublin regulation – executive control is associated with a lack of support.

Policy Output at the EU Level

At the EU level, the favorable policy output in the COVID-19 crisis, which contrasts with the minimal policy output in the refugee crisis, confirms the role of support in facilitating joint solutions (H4, H4a). Given the more favorable conditions of the crisis situation, EU authorities’ support was more significant in the COVID-19 crisis. Joint action became possible in the public health and the economic policy domain (see Jones, Daniel Kelemen, and Meunier Reference Jones, Daniel Kelemen and Meunier2021; Quaglia and Verdun Reference Quaglia and Verdun2023; Rhodes Reference Rhodes2021; Schelkle Reference Waltraud.2021; Truchlewski et al. Reference Zbigniew, Oana, Moise and Kriesi2023). As we have seen, transnational coalitions opposed the RRF’s proposals, but the German–French leadership overcame this resistance (H4e). The most significant advances were made in the economic policy domain, where the EU already had the most critical competencies – the integrated market is one of its core competencies, and monetary policy is in the hands of the ECB. The response came early in the economic policy domain and diminished in later waves, as H4b expected. In this domain, the crisis induced an accommodating monetary policy and regulatory measures, such as suspending SGP and competition rules, to give the member states free reign in their fiscal, economic, and social policy responses to the crisis. In addition, it also led to centralized capacity building in the form of SURE, an instrument for temporary support to mitigate unemployment risks in an emergency, and above all, in the form of the RRF, an investment fund above all designed to improve the economic capacity of the weaker member states. The SURE instrument helped more fragile member states finance short-time work schemes to deal with the immediate fallout of the economic shock of the pandemic. The fiscal permissiveness of the EU during the COVID-19 crisis indirectly also increased the capacity of the weaker member states to come to terms with the economic shock. The RRF, an investment fund of unique proportions, is expected to create a level playing field between the member states in the long run. Even if it is too early to tell whether this example of centralized capacity building is of a permanent or only of a temporary nature (Fabbrini Reference Fabbrini2022), and even if its implementation may turn out to have been less effective than initially intended (Boeri and Perrotti 2023), the adoption of the RRF in July 2020 has been a watershed moment in EU fiscal governance.

In the public health domain, the timing of the policy response, which was most important in later waves, contradicts H4b to some extent. The minimal competencies of the EU may explain this unexpected result in this policy domain, which limited its policy responses early on. The continued pressure exerted by the crisis helped to extend its competencies in later waves (H4c). In the first COVID-19 wave, the EU had a limited impact on national decisions to close borders and control entry. The lack of EU coordination encouraged member states to go it alone or copy their peers under the pressure of the crisis. As the pressure was kept up in the later waves, the EU engaged in comprehensive coordination with new instruments (e.g., the Green Pass and traffic light systems for traveling). Initially, the EU also failed to prevent unilateral measures by member states in the public health domain (e.g., unilateral export bans for medical). Still, it eventually made up for its initial failures with its joint vaccine procurement program. Although lagging behind the United States and the UK in its common vaccine procurement strategy, the EU quickly caught up in the spring and summer of 2021. EU member states managed to coordinate in pooling funds for access to vaccines (apart from Hungary and Slovakia, which also unilaterally relied on Chinese and Russian vaccines). In addition, in the realm of health policy, the EU also proceeded to build institutional capacities by creating components of a “European Health Union” – HERA (European Health Emergency Response Authority) and an EHU (European Health Union), while also expanding the mandates of the ECDC (European Center for Disease Control) and the EMA (European Medicines Agency) (Ferrera, Kyriazi, and Miró Reference Ferrera, Kyriazi and Miró2022).

COVID-19 also contributed to EU capacity building, i.e., polity maintenance, beyond the policy domains directly concerned by the crisis. Thus, it affected the “rule-of-law” crisis, which originated from two member states – Hungary and Poland – that introduced illiberal institutional reforms in violation of central European values. The EU proceeded to sanction these two deviant member states by not allowing them to benefit from the RRF funds attributed to them as long as they did not reform their legal institutions. Moreover, in procedural terms, as already mentioned, Ladi and Wolff (Reference Ladi and Wolff2021: 36) argue that a policymaking mode emerged from the EU’s institutional response to the COVID-19 crisis, best described as “coordinative Europeanization.” This coordination mode is neither purely intergovernmental nor exclusively supranational since it involves direct consultations between member states and the Commission to elaborate policies that would work for everyone. Interestingly, similar coordination mechanisms have been applied in the joint vaccine procurement during the COVID-19 crisis (Becker and Gehring Reference Becker and Gehring2023), in the capacity building in the public health domain (Ferrera, Kyriazi, and Miró Reference Ferrera, Kyriazi and Miró2022), and in the Brexit negotiations (Laffan and Telle Reference Laffan and Telle2023). Ferrera, Kyriazi, and Miró (Reference Ferrera, Kyriazi and Miró2022) use “expansive unification” to characterize these new coordination mechanisms.

The policy output at the EU level during the refugee crisis starkly contrasts with the production during the COVID-19 crisis. The combination of asymmetrical incidence and joint competence between EU and member states with partisan politicization at the domestic level proved to be particularly critical for common solutions. As I have shown, such a setting renders joint policymaking initiatives and collective action difficult and, instead, leads to unilateral reactions on the part of member states, the spillover effects of which unleash and exacerbate transnational conflicts and give rise to a complex web of cross-level and transnational interactions to come to terms with these conflicts. As a result of these difficulties, the EU failed to adopt the relocation scheme, which was blocked by the V4 coalition, and, as was already pointed out in the previous section, it had to rely on the stop-gap solution of the EU–Turkey agreement, which was made possible by the joint efforts of the German Chancellor and her allies among the EU authorities. Moreover, despite the pressure exerted by the crisis, the EU proved unable to reform its defective asylum policy for many years after the peak of the crisis. Instead, it limited access to the Common European Asylum System (CEAS) (Lavenex 2018) by reinforcing its external borders and by externalizing the problem solution to third countries. The EU has been buying time by relying on “defensive integration” (Schimmelfennig Reference Schimmelfennig2021). Still, the incapacity to reform the standard asylum policy led to reactivating the conflict potentials linked to asylum and migration policy at subsequent times. In this policy domain, the EU seems stuck in a “sub-optimal equilibrium” (Hix and Hoyland Reference Hix and Hoyland2022: 363).

Alexander Shaw, Kovarek, and Schelkle (Reference Shaw, Kate and Schelkle2024) asked a series of experts on European integration – academics and policymakers – about the consequences of four crises – the COVID-19 crisis, the refugee crisis, the Euro Area crisis, and the Brexit crisis – for European integration. Their results align with the summary of the two crises I have compared here. They show that the refugee crisis is overwhelmingly perceived as having had negative consequences for integration (by 85 percent of respondents). In comparison, the COVID-19 pandemic is overwhelmingly seen as having a positive effect on integration (by 70 percent of the respondents). The other two crises are situated in between.

Policy Output at the Member State Level: Refugee Crisis

At the national level, too, continuity prevailed in the refugee crisis. The measures introduced during this crisis were consistent with an approach that can be traced back over two decades (Geddes, Hadj Abdou, and Brumat Reference Geddes, Hadj Abdou and Brumat2020). The intense partisan politicization of the policies involved limited policy support and prevented any significant steps forward (H4, H4a). At the national level, the crisis pressure during the refugee crisis primarily led to the reintroduction of domestic border controls and further retrenching of asylum policy across member states. The national policy response varied, above all, according to the member state type (H4e). Thus, frontline states turning a blind eye to secondary movements of refugees created border conflicts with neighboring states (e.g., Italy with Austria and France or France with the UK). Unilateral border closures created conflicts with humanitarian organizations and other member states, such as in the case of the Italian port closures for NGO ships saving refugees or in the cases of Hungarian and Austrian fence building at their southern borders. Border issues poisoned domestic politics in Germany, where they openly divided the government throughout the crisis. As for retrenching asylum policy, it generally attempted to make the member states less attractive for asylum seekers in different ways. Thus, Sweden, the paradigmatic open destination state, introduced temporal limits (three years) to residence permits for asylum seekers, aligning itself with other member states that already had such limits. Austria introduced an “upper limit” for the number of asylum seekers it would accept annually. Germany adopted two asylum packages to make life more difficult for asylum seekers. France and the UK, closed destination states, generally aimed at creating a “hostile environment” for asylum seekers (see Kriesi et al. Reference Kriesi, Bojar, Altiparmakis and Oana2024).

The impact of the national policy output on the intended outcome is difficult to gauge, i.e., the impact of these measures on their intended goal to stop the inflow of asylum seekers on the national territory. It is the EU–Turkey agreement that essentially stopped the inflow in the 2015–16 crisis: The member states were rescued by the EU in the short run, in an illustration of Alan Milward’s general thesis (Milward Reference Milward2000). The national measures were not particularly relevant to stopping the inflow but remained essentially symbolic politics. Thus, to mention but this example, the preset Austrian upper limit – a European precedent, the provision for a quantitative limit to grant a human right (Gruber Reference Gruber, Barlai, Fähnrich, Griessler and Rhomberg2017: 51) – has never been reached, meaning the corresponding decree was never applied.

Importantly, however, the refugee crisis had a more general political impact (see Kriesi et al. Reference Kriesi, Bojar, Altiparmakis and Oana2024). The political parties on the right have been reinforced by a general drift toward the right resulting from this crisis, even if the crisis did not give rise to a wholesale transformation of party systems in any country, as was the case in the Euro Area crisis. In contrast to the Euro Area crisis, which caught several actors by surprise or forced them to adopt untenable and unpopular positions, the refugee crisis, with its cumulative and expected nature, allowed much more room for strategic choices by parties that were able to anticipate the potential political impact of the crisis and react strategically to the country-specific crisis situation. The most salient transformation pattern is a drift toward the right, as more parties on this side of the spectrum rushed to capitalize on the issue and prioritize it in their campaign discourse. Right-wing actors who persisted in their anti-immigration message enjoyed electoral gains at the expense of their proximate competitors and of the left. In some countries, such as Hungary, Austria, the UK, and Greece, nationalist conservative parties displaced the radical right, while in others, such as Italy, Germany, France, and Sweden, the radical right increased its vote share at the expense of the mainstream right. The drift to the nationalist–conservative right, the exacerbation of the conflict between nationalists and cosmopolitans at the domestic level, and the transnational conflicts between a sovereignty coalition and a core coalition bent on further integration suggest that the refugee crisis has generally undermined the solidarity between member states in the EU.

Policy Output at the Member State Level: COVID-19 Crisis

At the beginning of the COVID-19 crisis, the member states reigned supreme. The fear and panic generated by the images from Italy sent shockwaves throughout the Union. Border control was in their competence, as was the public health policy. Given the great uncertainty at the outset of the pandemic, governments thought they could buy time by closing borders to slow the spread of the virus in the first wave. Except for a few controversial outliers, notably Sweden (and the Netherlands at the beginning of the pandemic), member states adopted widespread lockdowns through regulatory convergence. They required various forms of social distancing, independently of how hard the virus hit them. They quickly adopted high-cost policies such as work and school closures. In contrast, practices later deemed relatively low-cost and highly effective at reducing the spread of the virus, such as screening and contact tracing, were comparatively slower to implement (Mistur, Givens, and Matisoff 2023). This regulatory convergence occurred spontaneously rather than as a consequence of the EU and its coordination mechanisms (Truchlewski et al. Reference Zbigniew, Oana, Moise and Kriesi2023). Lockdowns shut down economic and public life at the national level. Again, the member states were the first to react to the economic catastrophe these measures implied. If solidarity was the word of the hour, it was solidarity within the nation-states. The nation-state made an economic comeback. Across Europe, member state governments opened the fiscal spigots: fiscal activism replaced austerity. The measures they adopted were temporary emergency measures, ranging from liquidity support for enterprises to short-term wage schemes and schemes for the income of the self-employed. When the emergency subsided, they were phased out, only to be renewed, extended, and adjusted as the pandemic struck once again in subsequent waves. At the end of the first wave, short-term recovery came sooner and has been far faster than expected initially. As lockdowns ended and economies recovered, it became essential to shift policies toward promoting recovery and to prevent the mistake of the 2008 financial crisis of switching too soon from support toward fiscal consolidation and monetary tightening. In the end, one needed to move from recovery to long-term expansion.

For the analysis of national policy output during the COVID-19 crisis, I can rely on three well-defined indicators. As mentioned in the design section, they include a general indicator for the government’s response, an indicator for the stringency of the government’s lockdown measures, and an indicator for the economic response, each ranging from 0 to a maximum of 100. These indicators allow us to systematically analyze the link between the crisis situation, policymaking, and policy output. You will find the time line of the responses for the three indicators in Figure 22, by country and as a country mean. As seen from these time lines, the response pattern did not vary much from one member state to the other. They immediately responded in public health and economic terms when the crisis hit. At the end of the first wave, they all relaxed their public health and stringency response while keeping up the economic support. After an intermediate period, the public health and stringency response was stepped up again in the second wave and only relaxed a second time in the second half of 2021. The continuing high level of policy response contradicts H4b, which expected declining policy output throughout the crisis. It is the continuing pressure that keeps up the pace of policymaking: Problem pressure increases again massively at the beginning of the second wave, and public salience remains high throughout the three waves (see Figure 3b).

Figure 22 Time line of policy response of member states’ policy response, by country and overall mean. The indicators have been smoothed by running averages over three weeks.

Poland is the country that deviates most in the less restrictive direction. It stands here as an example for Eastern European countries. As already pointed out in Section 4, this particular response pattern in Eastern Europe can be explained by the vulnerable economies of these countries inducing them to relax lockdowns more in later phases of the pandemic (see Popic and Moise Reference Tamara and Moise2022). On the other side, Italy has the most robust overall response. It was hit early on, reacted faster and more strongly than the other member states, and kept up its general reaction until the end of 2021.

The similarity of the policy response patterns across member states during the COVID-19 crisis already points to the significant impact of the problem and political pressure on the polity output: The crisis situation did not leave the policymakers much room for choice of their reactions, it seems. This hunch can be tested with a more systematic analysis of the determinants of the policy output during this crisis. To predict the overall government response, I use the pressure indicators and the general characteristics of policymaking – its salience, executive dominance, and policy support as predictors. For the stringency of the lockdown and the economic policy response, I rely on the corresponding attributes of the lockdowns (restrictions and relaxations)/public health domains, as well as the economic policy domain. Figure 23 presents the effects of the pressure indicators and the characteristics of policymaking for the three types of policy response.Footnote ¹⁶ The effects shown are maximum since the predictors have been rescaled to the 0–1 range.

Figure 23 Determinants of policy output during the COVID-19 crisis: long-run effects.

As can be seen, both pressure and policymaking strongly affect the policy output at the national level. Regarding pressure, public salience has a powerful effect on the two more specific indicators – stringency and economic response: In line with H4c, political pressure greatly facilitates policy output. In the COVID-19 crisis, public attention is focused on the existential threat of the pandemic. The public expects the government to do something about this threat, and the government’s response is immediate (with one week’s lag). The effect on the overall response indicator is less pronounced, which may be attributed to this indicator’s lower precision and the broad characterization of policymaking (across all policy domains). Controlling for public salience, problem pressure as measured by new weekly deaths also has a considerable impact on stringency measures but a lower impact on the overall response and no impact on the economic response. Its effect is, to a large extent, mediated by public salience.

Controlling for external pressure, the policymaking characteristic with the most significant impact on the policy output for all three indicators is the salience of policymaking: It dramatically increases the policy response. Executive dominance weakly contributes to lockdown measures and slightly to the overall response but not to the economic response. However, policy support does not affect the three output indicators, contradicting the general hypothesis H4. This is to say that national policy output during the COVID-19 crisis directly resulted from the intensity with which policymakers dealt with the crisis and, to a limited extent, of prevailing executive actors. However, increasing levels of support among policymakers have not facilitated it. At first sight, this result may surprise. But it seems more plausible on reflection, as already discussed in the theory chapter. First of all, as we have seen throughout this study, policymaking during the COVID-19 crisis was rather consensual, i.e., under the tremendous exogenous pressure, policymakers rallied around the governments. The level of consensus was comparatively high throughout the crisis, and the partisan politicization of policymaking remained relatively limited. The limited variation in support of the government’s proposals throughout the crisis is illustrated by Figure 24, which presents the time lines for policy support. As this figure shows, on average, policy support is always positive. The average varies rather little across time, even if, as we have already seen (Figure 20), support varies across countries and sometimes even turns negative in countries such as Poland or Germany. As we have seen in Section 4 (Figure 5), the salience of policymaking is much more variable, on average, and its variation is more closely reflected in the policy output indicators. Moreover, as argued previously, exogenous pressure may force executive policymakers to come up with a response, even if broad sections of the political elite do not support them. To the extent that there was resistance against the government’s policies in the COVID-19 crisis, resistance mainly targeted the lockdown measures (Kriesi and Oana Reference Kriesi and Oana2023), it did not affect policy output. In the face of the enormous problem pressure and the equally vast pressure exerted by public opinion, the governments had little maneuvering room for taking the resistance exerted by specific groups into account.

Figure 24 Time line of policy support in member states, by country and overall mean. The indicators have been smoothed by running averages over three weeks.

Analyzing the determinants of policy output by wave (combining the second and third waves) for the stringency indicator, we find that problem pressure and public salience constitute the strongest determinants of lockdown measures in both waves (see Figure 25.)Footnote ¹⁷ This finding confirms the notion that the continued pressure kept up policy output during the COVID-19 crisis. Moreover, during the first wave, policy output was only a function of the effect of pressure, which is to say that early on in the COVID-19 crisis, policymakers at the national level just reacted to the pressure exerted by the crisis situation. In later waves, the salience of policymaking and the dominance of executive actors also played a role, but policy support remained insignificant.

Figure 25 Determinants of policy output during the COVID-19 crisis for stringency indicator, by wave: long-run effects.

Conclusion

The two crises differed markedly regarding policy output: At both levels of the EU polity, it was much more positive and consequential in the COVID-19 crisis than in the refugee crisis. This difference is primarily a result of the different crisis situations. In the COVID-19 crisis, the combination of overwhelming problem pressure and pressure from public opinion made for immediate and robust responses on the part of the leading actors. The other policymakers broadly supported this response, and even if they did not support it, the resistance was not very vocal, and the top policymakers could overcome it (at the EU level) or ignore it (at the national level). In contrast, in the refugee crisis, the strong opposition at the EU level led to the failure of the preferred solution in the short run and the inability to reform the asylum policy in the long run. At the national level, the highly conflictual situation did not allow significant reforms either, leading mainly to symbolic politics. But, more fundamentally, the crisis exacerbated the conflict between nationalists and cosmopolitans, leading to a rightward shift in the party system with lasting consequences for national and EU politics.