Policymaking in the EU under crisis conditions: Covid and refugee crises compared

We study how crises situations shape the political decision-making structure of the EU and the responses adopted by European policy makers by comparing EU decision-making in the first wave of the COVID-19 crisis (March 2020-July 2020) and in the refugee crisis (2015 to 2019), based on a new data-set on policymaking. Similarities between the two crises include: comparable polarization and conflict intensity, executive dominance, a greater role of EU institutions in policy domains where the EU has higher competence, more conflict and greater resistance by coalitions of member states in domains where it has lower competence, and minority coalitions of critical member states have been crucial for possible solutions in both crises. The key difference between the two crises lies in the fact that, in the refugee crisis, the opposing coalition was able to prevent any kind of reform, while in the first wave of the Covid crisis, the opposition was more amenable to a joint solution in the crucial fiscal policy domain, where the conflict became most intense. We suggest that this key difference is ultimately rooted in the character of the original problem pressure and the different distribution of spatial incidence in the two crises.


Introduction
We compare the decision-making structures in the EU during two recent crises-the first wave of the Covid-19 crisis and the refugee crisis. In the EU, the shared multilevel exercise of authority has introduced a novel type of political co-production, in which national executives fully participate in central policymaking along with relatively weak supranational institutions. In this multi-level polity with a relatively weak center, the grand theories of European integration locate the power alternatively in the supranational agencies-the Commission (neofunctionalism) or the European Council (new intergovern mentalism), or in the member states (liberal intergovern mentalism, post-functionalism). Hooghe and Marks (2019), who compare the contributions of each theory to the analysis of the recent crises, come to the conclusion that they all discipline our thinking about the behavior of key actors and advance our understanding of these crises. We intend to specify their ecumenical conclusion based on a novel dataset to study the decision-making structures in the EU under crisis conditions to shed new light on the role played by supranational actors, governments of member states and other types of actors. In accounting for decision-making under crisis conditions, we follow the lead of new intergovernmentalist scholars who provide evidence that national preference formation in the EU has become an inherently transnational process which involves governments of member states (e.g., Kassim et al. 2020;Fabbrini 2022). In crisis situations where uncertainty and urgency prevail, national preference formation and European level bargaining tend to become simultaneous processes with policymakers being involved and negotiating at the national and the EU level at one and the same time and transnational coalitions shaping policymaking processes in decisive ways at the EU level (Fabbrini 2022).
In focusing on how the crises shaped the political decision-making structure of the EU and the responses adopted by the policy makers, we contend, following Ferrara and Kriesi (2021), that the specific combinations of crisis pressures at the origin of the crisis create different decision-making scenarios. Ferrara and Kriesi (2021) characterize the crisis pressures by the combination of the policy heritage and the supranational capacity, on the one hand, and the spatial distribution of the pressure and the politicization of national identities, on the other hand. Combining the two dimensions, they arrive at four scenarios, two of which concern the two crises we are focusing on here: the refugee crisis and the Covid-19 crisis. According to their schematic classification, both crises hit policy domains of low EU competence, but the situations of the two crises differed importantly with regard to the spatial distribution of the pressure and the politicization of national identity. In the case of the refugee crisis, the pressure was asymmetrical and national identities loomed large as debates related to immigration, especially from outside the EU, touch upon the very core of national self-conceptions of "us" versus the politically constructed "other". By contrast, in the case of the Covid-19 crisis, the pressure was more symmetrical and the joint exposure coupled with universal concerns related to healthcare made the activation of national identities less conspicuous. However, as we shall see, crises are multifaceted phenomena and there is within-crisis variation by policy domain with respect to these basic characteristics, with important consequences for the policymaking process in the crisis situation.
We shall show how these distinct crisis conditions shaped the policymaking and the corresponding conflict structures during the crisis at the level of EU decisionmaking. With respect to the Covid-19 crisis, we focus on the first wave up to July 2020, while we cover the refugee crisis over a more extended period. For the empirical analysis, we rely on a new method-Political Process Analysis-which allows us to chart the public face of the decision-making 1 processes as they were reported in the news. We proceed by first presenting our concepts and theoretical expectations. Next, we briefly introduce our method and the operationalization of the theoretical concepts. Third, we present our results in three steps-proceeding from the overall politicization of the two crises to the predominant type of decision-making and the emerging conflict structures. Finally, we summarize our argument and conclude.

Concepts and theoretical expectations
To characterize the decision-making scenarios, we rely on a number of key concepts-politicization, conflict intensity, pace, executive decision-making and coalition formation between member states. Our broadest concept is the concept of politicization, which builds on Schattschneider's (1975) notion of the 'expansion of the scope of conflict within a political system' (see Hutter andGrande 2014, 1003). We distinguish between two conceptual dimensions which jointly operationalize the concept of politicization: salience (visibility) and actor polarization (conflict, direction). 2 The concept of conflict intensity is closely related to actor polarization, for which it provides a more detailed account. We borrow the concept of pace (or tempo) from Grzymala-Busse (2011) and Bojar et al. (2023), who both recognize that the speed at which the policy debate moves forward is an essential feature of such a debate. Among other things, it constrains the room for deliberation and consultation between the participant actors (Grzymala-Busse 2011, p. 1269). The concept of executive decision-making is introduced to characterize the key actors in the policymaking process. It refers to policymaking that is not confined to policyspecific subsystems, but becomes the object of 'macro-politics', i.e., is taken over by the political leaders who focus on the issue in question. In the terminology of the punctuated equilibrium model of policymaking, executive bargaining is the result of 'serial shifts' (Jones and Baumgartner 2005). Finally, we follow Fabbrini (2022) in underlining the importance of coalitions between member states.
Our key argument is that the characteristics of the crisis situation-the distribution of crisis pressures and of competences in the multi-level EU polity-condition the decision-making scenarios. To account for the crisis situation, we rely, as already pointed out, on Ferrara and Kriesi's (2021) characterization of crisis pressures, which combines the competence distribution across the levels of the EU polity with the distribution of the pressure across the member states to come up with four types of crisis conditions. To the extent that the crisis situations of the two crises are similar or not, we expect to find similarities or differences in the decision-making scenarios of the two crises.
Let us first address the differences. Crises are highly salient public events. But even in this exceptional category of news events, the Covid-19 crisis is in a class of its own, given that it was one of the biggest news stories ever (The Economist, 19/12/2020). Only the world wars rival Covid-19's share of news coverage. It is worthwhile to state this at the outset, since the exceptional prominence of the Covid-19 crisis is likely to influence any results about decision-making processes that are based on the analysis of news reports. The extraordinary character of the Covid-19 crisis thus serves as a useful benchmark against which to compare other crises, such as the refugee crisis.
In addition, the Covid-19 crisis hit the EU member states as an unexpected exogenous shock to which they were more or less jointly exposed. Even if the depth of the public health crisis varied from one country to the other, they were all hit by the virus, as the pandemic proved to be a global phenomenon. As suggested by Ferrara and Kriesi (2021), such a decision-making scenario is characterized by an "enabling consensus". The refugee crisis, by contrast, arguably had a strongly asymmetrical incidence: frontline states and transit states were exposed to a very different crisis than destination states, and some member states of the EU were not really concerned by the crisis at all. As pointed out by Ferrara and Kriesi (2021), such a decisionmaking scenario is consistent with the post-functionalist notion of "constraining dissensus".
These different crisis situations lead us to expect different policymaking dynamics in the two crisis. First, the policymaking dynamics are expected to differ as a result of the different perceived urgency of the problem pressure. While we acknowledge that the perceived urgency of the problem pressure may be endogenous to the policymaking dynamics as political actors attempt to influence the salience of the issue at hand, we maintain that exogenous shocks like the crises we study here have an independent effect on issue salience, which determines the agenda-setting by political actors, at least in the short-run (Dennison and Geddes 2019). Though the peak of the refugee inflow in fall 2015 conveyed a sense of emergency to the wider European public, the refugee flows (see Fig. A-1) had been accumulating for some time and policymakers had anticipated them as evidenced by the EU's Agenda on Migration launched in May 2015 and early re-bordering efforts in the summer of 2015 in Hungary. By contrast, the spread of Covid-19 from Asia to Europe in a matter of weeks came as a sudden shock to the European public in the beginning of 2020. At first, European policy makers equivocated, exemplified by reports that British premier Boris Johnson wanted to "ignore" the virus. It was not until mid-March and the first full-fledged nationwide lockdown in Italy that policy makers started to confront the inevitable. Based on this difference in urgency and preparedness between the two crises, we expect a more rapid pace of policymaking in the Covid-19 crisis compared to the refugee crisis.
By contrast, we expect a greater conflict intensity and greater polarization in the case of the refugee crisis. In this case, opposition by member states to joint solutions was likely to be reinforced by the fact that they have been asymmetrically affected by the crisis. In such a situation, opposition to collective EU action is likely to be raised by relatively unaffected member states which are not ready to share the burden (Bauböck 2018). In addition, joint action was equally constrained by the politicization of national identities produced by the uneven distribution of crisis pressure.
In the case of the Covid-19 crisis, we also expect that the symmetrical exposure to the crisis made consensus among member states and joint solutions more likely, especially in the economic domain. The common threat induced public opinion and elites to become more supportive of a "federal bargain" to preserve the integrity of the EU polity. As argued by Genschel and Jachtenfuchs (2021), despite the initial divisions among EU member states, the common sense of crisis generated an extension of in-group feelings inside the EU. This reciprocal empathy provided a window of opportunity for an "enabling consensus" pushing EU policymaking in an unexpected direction of collective action in the fiscal domain. In line with this scenario, we expect more leeway for supranational actors in the Covid-19 crisis than in the refugee crisis, better coordination among them, and more success for their proposals for a "federal bargain".
Turning to the expected similarities, we first argue that, under crisis conditions of high urgency and uncertainty, executive decision-making is generally likely to become the key policymaking mode. In the multi-level polity of the EU, executive decision-making primarily involves the EU Commission, the European Council, and the governments of the member states. Under pressure of urgency and uncertainty, the Commission is expected to focus its agenda-setting power on the policy domains concerned by the crisis, while the governments are expected to scramble to come to terms with the crisis individually at the national level, and jointly in EU summit meetings. In EU decision-making, member states are represented by their top executives who play a pivotal role in the multi-level EU decision-making even under routine circumstances. In the two-level game of EU politics they constitute the pivot linking domestic politics to EU decision-making. Accordingly, we expect the governments of the member states and their key executives to generally play a crucial role policymaking at the EU level. In a crisis situation, their importance is likely to be even further enhanced, because member states still control some core state powers (fiscal and coercive resources), which the EU needs for the management of the crisis. In line with the enhanced role of member state governments, the European Council is likely to become central in crisis situations.
Second, we expect the pre-existing EU competences in the policy domain concerned to generally determine the extent to which the EU will intervene and joint solutions will be possible. In line with intergovernmentalist predictions, supranational entrepreneurship is generally constrained in policy domains where the EU has little competence (Moravcsik 2005: pp. 362-363). Under such conditions, the policy proposals by supranational actors to come to terms with the crisis run into resistance by some member states which use their autonomy to resort to unilateral action. In this respect, it is significant that the two crises hit policy domains where the EU has overall low competence, which serves to enhance the role of member state governments and their executives in both crisis. The asymmetrical incidence of the refugee crisis is expected to have enhanced their role even more.
However, we should point out that both crises are multifaceted phenomena with parallel debates unfolding on multiple issues and sub-issues involving various policy domains. 3 This is important, because EU competence and the politicization opportunities for member states vary by policy domain. In the case of the Covid-19 crisis, two issues dominated the policy agenda at the EU level: public health and the economic fallout from nationwide lockdowns measures. The public health crisis involved, among other things, issues of border control, procurement and export bans on medical equipment, research, development, procurement and distribution of vaccines, contact tracing techniques and related issues of privacy rights. The economic crisis involved issues such as exemptions from single-market rules, especially with regard to state aid to ailing sectors, monetary policy and fiscal policy to support the banking sector and the member states, and steps toward longer term fiscal burden sharing culminating in the July 2020 agreement on the Recovery and Resilience Facility (NGEU Fund). While internal market rules, agriculture and monetary policy are core competences of the EU, public health and fiscal policy mostly fall under the competences of the member states. Accordingly, we expect a greater role of EU agencies in the former policies than in the domains of public health and fiscal policy.
The refugee crisis above all concerned asylum policy, but it still had multiple facets: issues of external and internal border control, as well as asylum rules about reception, distribution, deportation, and integration. External border control (Frontex, negotiation of agreement with third countries, support of frontline states) required joint action, while the granting of asylum rights remained in the exclusive competence and responsibility of the member states. It was the latter which was unequally distributed, given that the Dublin regulation delegates responsibility for examining an asylum application to the frontline states, where asylum seekers first entered Europe. Accordingly, we expect a greater role for EU agencies in the domain of external border control than in the domain of asylum rules.
The opportunity for coalitions of opposing member states to politicize policymaking also varies by policy domain. The coalitions among member states are crucial for any joint solution to emerge. Even if a majority in the EU supports intervention on behalf of strongly affected member states, the opposition from a coalition of little affected member states may be strong enough to prevent such intervention, exploiting the super-majority requirement to change the status quo. Moreover, even if a majority of member states share a common problem and support EU intervention, it may not be forthcoming if the EU lacks the corresponding capacity to intervene. Such a capacity may have to be created by joint decision-making in the first place. To overcome the resistance of coalitions of member states against joint solutions, the European Council has become the key actor. Given the lack of a genuine and free-standing polity-holding center, the EU can be kept together only by extraordinary political investments of national leaders. In both crises, coalitions of member states opposing joint solutions are known to have played a key role, but we expect their position to have been weaker in the case of the Covid-19 crisis given the symmetrical distribution of the public health crisis, and we expect stronger conflicts between the coalitions confronting each other in the case of the refugee crisis. With respect to the specific policy domains, we expect opposing coalitions to be strongest and conflicts to be most intense in the decision-making concerning asylum rights, where the asymmetrical incidence is combined with low EU competence, but more moderate in external border control, where asymmetrical incidence combines with medium EU competence. Similarly, in the Covid-19 crisis, we expect opposing coalitions to be stronger and conflict more intense in the decision-making concerning public health and fiscal policy, characterized by symmetry and low/medium EU competence, than in monetary policy and single market-related policies, where symmetry and high EU competence prevail. Table 1 summarizes our expectations with regard to the impact of crisis conditions on the decision-making structures.

Design of the study
To assess the hypotheses we laid out in the previous section, we build on a novel data collection and data analysis technique: Political Process Analysis. A detailed description of the method is presented in Bojar et al. (2023). We only summarize the main features here which are directly relevant for the present analysis. PPA is a middle ground approach to study policy debates on various levels of policymaking, taking place in multiple arenas, following in the footsteps of similar middle ground approaches in the study of contentious politics (Kriesi et al. 2019;Bojar et al. 2023). The idea of the middle ground refers to the study of policy debates via a systematic collection and coding of media data that simultaneously enables the reconstruction of narratives of the policy debate (the "process-tracing approach") and the construction of substantively meaningful quantitative indicators to characterize policy processes and subject them to statistical analysis [in a fashion similar to protest event analysis (Hutter 2014)].
For our purposes, the relevant level of analysis is the EU. The text corpus we used for the coding was constructed based on a combination of crisis-specific keywords applied in the news aggregator platform Factiva, relying on a list of international media outlets. For the refugee crisis, we constructed this corpus around six policy episodes that constituted the core of the EU's response to the refugee crisis: the EU-Turkey agreement, the Resettlement Scheme, the establishment of Hotspots in frontline states, the reinforcement and transformation of Frontex into a permanent European Border and Coast Guard, the EU-Libya deal and the reform of the Dublin regulation. Our coding for this crisis covers the period from April 2015 to February 2020. During the Covid crisis, policymaking proceeded in a parallel fashion in various domains, making it more problematic to split up the coding process into distinct episodes. Therefore, we coded the entire timeline of the policymaking response using a more general set of Covid-related keywords, starting from the first local lockdowns in Italy in March 2020, until the late-July 2020 summit of the European Union where a broad agreement on the NGEU Fund was reached. We consider this summit as the end of the first wave, both because of the substantive importance of the agreement and because it coincided with the bottom of the infection-and the mortality curve between the first and the second waves. In the coding of both crises, we restricted the codes to those actions that had an international component, either via the actors or the issues involved.
We divided up both crises into distinct policy domains. For the refugee crises, we distinguish between border control, redistribution of asylum seekers, receptionrelated issues, reform of the EU's asylum regime, and a smaller category of institutional issues, such as rule of law-related issues in Turkey. In the Covid crisis, we distinguish between border control measures, public health measures (including social distancing rules, stay-at-home requirements, closures and openings of institutions, and other related measures), economic measures in the regulatory and monetary sphere (including issues related to state aid), fiscal policy measures with the NGEU Fund debate in the center, institutional measures (such as emergency powers) and other Covid-related policy domains. We use the following classification for EU competences in the respective policy domains: high EU competence-single market-and monetary policy in the Covid crisis; medium level EU competence-external border control in the refugee crisis, and fiscal policy-related matters in the Covid crisis; low EU competence-all the other policy domains, where member states rule supreme.
Our unit of analysis is the action undertaken by a participant in the policymaking process. Overall, the present analysis is based on 3013 actions, 1254 (= 41.6%) for the refugee crisis and 1759 (= 58.4%) for the first wave of the COVID crisis. These actions include both policy actions (proposals, negotiations, decisions, implementations) and policy claims (verbal statements) about the policy process. We coded the following aspects of the actions and claims in the policy debate based on a detailed codebook with specific operational rules and instructions to a team of coders: the arena and form of the action, the substantive type of engagement with policies, the actors undertaking the action, the target actors of the actions in the subset of cases where such target actors could be identified, the policy domain (issue), the actors' policy positions (issue direction), the actors' positions toward target actors (actor direction), and the narrative frames used by the actors to justify their action. Importantly, for each action, we also coded a descriptive string which provides the gist of the action in a short summary and constitutes the building block for narrative details. We refer the reader to the Online Appendix A2 for a concise version of the codebook.
The next step in our empirical approach consisted of building empirical indicators for our key concepts. First, we conceptualize the idea of politicization as the product of issue salience and actor polarization around the particular issue (Hutter et al. 2016). We use a simple monthly count of actions as a proxy for the first component, salience, because the frequency of action reported by the international media sends a strong signal on the importance of a given policy domain to the general public. We rescaled the salience scores to the 0-1 range. For the second component, polarization, we multiplied the monthly share of actions in support of the underlying policy proposals with the corresponding share of actions opposing the proposal, and rescaled the resulting product to the 0-1 range. We then calculated the product of the salience and polarization scores and also standardized it to the 0-1 range across the two crises, policy domains and months, in order to make the level of politicization comparable both across crises but also across policy domains.
While politicization describes the overall importance of a given crisis and issue in the public mind, it says little about how heated the policy debates gets between the actors involved. With a second indicator, we aim to capture the intensity of the conflict based on a combination of actor direction and policy action codes, which jointly determine the level of conflict between any given pair of actors. In operational terms, we assigned to each action a conflict intensity score ranging between 0 and 1 on an ordinal scale, with actions opposing target actors and/or involving conflictual policy claims (such as claims opposing the proposal in substantive terms or claims criticizing/denigrating opponents) scoring closer to 1, and actions supporting target actors and/or involving conciliatory policy claims (such as claims in full support of policy proposals, or steps toward implementation) scoring toward 0. The conflict intensity score for a given unit of analysis-which in our framework, is typically the crisis-issue-month-is then the average of the individual conflict intensity values across all observations within the same unit.
Two further indicators are required to test our hypotheses relating to pace and executive-dominance in the policy-debate. We adopt a simple operationalization for pace: the average number of days lapsed between two issue-specific actions. However, we distinguish between the pace of the overall debate and the pace of the policymaking process itself: while the former measures the average days lapsed between any two actions, the latter restricts the analysis to what we call "policymaking steps" (excluding verbal policy claims) and measures the average number of days lapsed between two such consecutive actions. Finally, for executive dominance, we rely on a combination of the institutional background of the actors and their position within their institutions. In particular, at the EU level we consider the European Commission, the European Council, and the European Central Bank as executive institutions, while at the member states' level, it is the national governments which we regard as institutions of this type. We provide two indicators of executive dominance: we code actions as executive (a) only when they are undertaken by actors in top executive positions within these institutions (heads of governments, presidents of the Commission, the European Council and the ECB) and (b) when they are undertaken by all actors in executive positions in these institutions (including all ministers and Commissioners). If, however, the action is undertaken by an actor in a non-executive position within any institution we do not regard it as an executive action. Once these executive dummies are assigned, we then average them over crises, issues and months to get a comparable measure of executive dominance between policy domains and over time.
In the last step of our empirical exercise, we derive the actor configurations in the respective crises and the underlying issue-specific debates. We do so by constructing a distance matrix based on the issue-and actor-position codes. For each actor-issue pair, we calculated the average distance between of the actor's positions on the issue in question. We then weighted these distances with the salience of the corresponding pairs. For visualization, we analyzed the resulting matrix of dissimilarities with a multi-dimensional scaling procedure (mds). Such a procedure allows to represent the overall actor configuration in a low-dimensional space, in our case in a twodimensional space. Actors who take similar positions in the crisis policymaking are placed close to each other in the resulting space, while actors who opposed each other in substantive terms are located at some distance from each other.

Overall pace and politicization
We start with the timing of decision-making. The difference between the two crises in terms of timing is most clearly reflected in the different pace of policymaking, as shown in Fig. 1. Broken down by policy domain, the differences are glaring. Even in the most rapid policy domain in the refugee crisis, border control measures, the average pace of the debate was considerably slower than in the Covid crisis. The policy debate in the first wave of Covid-19 was extremely rapid: in all policy domains, the average number of days lapsed between two consecutive actions was less than one, underscoring the exceptionally packed nature of the policymaking timetable. In fact, when we restrict the analysis to policymaking steps only, the comparative picture is not much different. In the refugee crisis, the average number of days lapsed between two policymaking steps amounts to almost a month (26.5 days), while in the Covid crisis it is a mere 1.5 days. Essentially, policymakers were involved in some sort of official step in the policymaking process almost every day in most Covid-related policy domains. Second, we present the overall politicization and conflict intensity of the two crises. As Fig. 2a shows, there was a vast difference in the average monthly level of politicization of the two crises. As expected, the Covid-19 crisis is in a category of its own as far as the monthly salience of policymaking dealing with the crisis is concerned. By contrast and contrary to expectations, with respect to the second component of politicization, polarization, the two crises are comparable. The same applies to the average monthly conflict intensity. Although the problem pressure in  the Covid-19 crisis was more symmetrical than in the refugee crisis, the overall conflict intensity of crisis management during the two crises was comparable as was the overall polarization between the actors involved. This means that the greater symmetry of the incidence of the crisis does not necessarily prevent conflicts to arise between member states. Nor does it necessarily lead to greater empathy between member states. It does, however, weaken the opposition to specific joint solutions in these conflicts, as we shall argue below. Compared to the first wave of the Covid crisis, the refugee crisis was much more drawn out. If we compare only the initial peak of the crisis with the first wave of the Covid crisis, and we do so for the two policy domains which are comparableborder controls and institutional reform-then the differences between the levels of politicization are no longer as stark as they appeared to be so far (Fig. 2b). In the first, peak phase of the refugee crisis, which lasted until the conclusion of the EU-Turkey agreement in March 2016, politicization levels in both policy domains reached roughly half of the monthly levels in the Covid crisis. This makes the Covid crisis still exceptional, especially if we keep in mind that the two comparable policy domains were the least politicized during the Covid crisis, but it indicates that the comparatively low level of politicization of the refugee crisis is to some extent due to its much longer duration.

Type of decision-making
We now turn to assess our hypothesis on executive dominance, which anticipates that in both crises, executive actors prevail. We first present the overall executive  Fig. 3. Although we lack an appropriate benchmark, the role of executives appears to be truly impressive. We present two indicators for executive dominance: one including all executives of the EU and member state governments and one including only their top executives. The average executive dominance scores including all executives for the refugee and Covid crises are .50 and .47, respectively, suggesting that roughly half of all the actions undertaken by the EU or member state governments are attributable to executive actors. The top executives, in turn, are responsible for roughly a third of EU and governmental actions in the refugee crisis (.30) and a fifth (.20) in the Covid-19 crisis.
The right-hand graph of Fig. 3 allows us to assess the relative importance of executive actors from EU agencies and from member state governments. In both crises, executive actors played a much greater role for the member state governments than for the EU. This is to confirm that, at the EU level, member state governments are essentially represented by executive actors. If the role of executives from member states is key in both crises (with shares of .76 and .71 of all actions of member state governments, respectively, in the refugee and Covid-19 crises), the role of top executives is more pronounced in the refugee than in the Covid-19 crisis: as expected, the comparative lack of EU competence and the asymmetry of the shock enhanced the role of top executives of member states beyond what we observe in the Covid-19 crisis. Thus, in the refugee crisis, top executives were responsible for almost half (.47) of the actions of member state governments, compared to one-third (.30) of the corresponding actions in the Covid-19 crisis.
Zooming in on the actors driving executive policymaking, we assess the relative weight of EU actors and member state governments in the different policy domains, relying on our broad issue categorization in terms of the level of EU competence and the distribution of the underlying shocks across member states. As we have argued, we expect the potential for joint EU-level action to be higher in domains where the EU has relatively high policymaking competences vis-à-vis the member states and/ or the shock is felt symmetrically across member states. Table 2 presents the results. The patterns of the relative shares of EU and member state actions largely conform to our expectations. On the one hand, the role of EU actors in policymaking has been generally more important in the symmetric Covid crisis than in the asymmetric refugee crisis. On the other hand, with one exception, we find that the lower the level of EU competence, the larger is the role of the member states in the policymaking process and vice-versa. The only exception to this pattern concerns public health and some miscellaneous other policy domains in the Covid crisis where the EU has low competence. While the member states prevailed as expected in the domain of border control-a domain where the EU had little competence-in the Covid crisis, 4 in terms of public health and miscellaneous other issues, EU actors turn out to have been more prevailing than expected. Thus, in spite of the fact that the EU had only limited competences in the public health domain, EU actors played a rather important role level in public health: it was above all the Commission which fought against national export bans on medical equipment, organized the procurement of medical equipment (masks and ventilators) and took steps to support vaccine research and improve diagnosis and treatment options.
We also expected opposing coalitions to be most polarizing and conflict to be most intense in policy domains with low EU competence, but more moderate in policy domains with medium/high EU competence, irrespective of the symmetry of the crisis incidence. As shown in Fig. 4, there is some evidence for such an effect of the competence distribution. While the differences are rather small in the case of the refugee crisis, where polarization and conflict intensity were generally rather high, Fig. 4 Polarization and conflict intensity in the refugee and Covid crises, by level of competence irrespective of the competence distribution, in the COVID-19 crisis, both polarization and conflict intensity were lower in the policy domains where the EU had medium (fiscal policy) or high (monetary and economic policy) competence.

Emerging coalitions
To characterize the conflict structure in each crisis, we first need to identify the adversaries in the various policy dimensions of the crises. As Fabbrini (2022) has pointed out, coalitions between member states have been shaping up in the COVID-19 crisis. In the highly contested fiscal policy during the first phase of the Covid crisis, what Fabbrini (2022) calls the "Solidarity coalition" of nine member states led by France which asked for joint debt instrument opposed the "Frugal coalition" of the "Frugal Four", joined by the Finnish government. By contrast, in the public health domain, the "Frugal Four" took a neutral position, while opposition mainly originated from the UK, which, although still a member state at the time, opted out of the joint procurement schemes for urgently needed medical equipment and for vaccines. The Frugal Four rejected the development of the EU into a "transfer union", originally in the form of Coronabonds and, at a later stage, in the form of grants. Unable to prevent the option of grants, they attempted to limit the funds attributed to grants, and to keep the overall fund smaller than proposed by the Commission. They also insisted on stricter conditionality in terms of rule of law and economic reforms. Finally, they softened their stance and accepted the compromise hammered out at the European Council meeting at the end of July, mainly as a result of the combined pressure from France and Germany (which had somewhat unexpectedly joined the Solidarity coalition (Krotz and Schramm 2022)).
The Visegrad four (V4) countries (Czech Republic, Slovakia, Hungary and Poland), Fabbrini's "Sovereignty coalition", did welcome the Commission's proposal, but had a series of objections which made them the least supportive group of member states apart from the "Frugal Four". 5 The V4 were also generally opposed to the "Frugal Four", whom they considered, in the words of Poland's PM Mateusz Morawiecki, "a group of stingy, egoistic states". Opposition to fiscal policy initiatives was not only voiced by coalitions of member states, but also by civil society and opposition parties in the member states. The latter attempted to politicize national identities in the case of the Covid crisis, too, even if it did not lend itself at first sight to such politicization. Thus, at the time of the letter of the nine member states calling for a common debt instrument (March 25), the media in Southern Europe were slamming Europe's foot-dragging and the leader of the main Italian opposition party, the Lega's Matteo Salvini, declared that any deal using ESM funding would be "an attack on our country". Once the German government had changed sides and joined France to make a common proposal for EU debt (instead of joint member states' debt), it was the turn of the German and French radical right opposition to criticize the proposals.
In the refugee crisis, opposition was mobilized from two sides. On the one hand, in the name of national sovereignty and security, the Visegrad four (V4) adamantly opposed any burden sharing in the form of the resettlement scheme proposed by the European Commission. Hungary and Slovakia appealed to the ECJ against the resettlement decision of the Council, and Hungary later on organized a referendum over the relocation quota. Eventually, the ECJ upheld the decision in September 2017, and the Hungarian referendum held on October 3, 2016, failed to reach the quorum due to opposition boycott. However, the implementation of the resettlement scheme proved to be elusive. On the other hand, civil society, supranational organizations (mainly UNHCR) and the opposition in the member states criticized the closing down of the national borders, the inhuman reception conditions in the refugee camps on the Greek islands and the conditions for refugees in Turkey, which became the key partner of the EU in the EU-Turkey agreement. Overall, we expected the opposition to be weaker in the Covid crisis than in the refugee crisis. Assessed by the average conflict intensity, this expectation does not hold, as we have seen: the average intensities are almost identical in the two crises. However, the conflict intensity of key opponents has been higher in the decisive policy domains of the two crises. Thus, the opposition by the V4 coalition was more intense in the resettlement conflict in the refugee crisis (with a score of .72) than the corresponding opposition by the Frugal Four coalition in the fiscal policy domain in the Covid crisis (a score of .53). The expectation also holds once we consider the outcomes. While the EU-Turkey agreement provided the EU with a provisional solution for stopping the flow of refugees, the resistance by the V4 coalition prevented any joint solution to the internal burden sharing of receiving and integrating asylum seekers. By contrast, in the fiscal policy domain of the Covid crisis, the NGEU Fund provided for a joint solution to the economic fallout of the Covid crisis.
In Fig. 5, we present the overall conflict configurations in the two crises. For this analysis, we have excluded some actor types (e. g. the national opposition in the Covid crisis) that do not have at least five mentions as actors and as target actors. In the refugee crisis, the resulting solution clearly distinguishes between the core decision makers and the two types of adversaries. The core includes both the European Commission and the European Council, Germany (the key member state in the decision-making process at the EU level), and the member states most concerned by the crisis (the southern European frontline states and France). Turkey, the EU's key external partner, and the other EU member states with the exception of the V4 are close to the core. The two adversaries are the "humanitarian" coalition of national opposition parties, civil society actors and supranational actors (above all UNHCR) and the Sovereignty coalition of four member states (V4). All policy domains are closely situated to the core, but note that redistribution (resettlement quota) and systemic reform (of the Dublin regulation) are most removed from the V4, who would not want anything to do with burden sharing between member states.
In the Covid crisis, the core coalition is made up of all the EU agencies, France and business. In this crisis, all EU institutions joined forces in a rather unique way. Although Germany, together with France, had paved the way to a joint solution in the fiscal policy domain with its common proposal on May 18, it is not located in the core. Instead, together with Italy and the other signatories of the letter of the nine, it is situated between the core and the two opposing coalitions on the fiscal policy-the Frugality and the Sovereignty coalitions. The UK, which took an opposing position on public health policy, appears to be furthest removed from the core, which is closely associated both with the economic policy at the EU level and with public health.

Conclusion
Our comparison of the EU policymaking during the two crises has shown that the Covid crisis during the first wave was, indeed, in a class of its own-in terms of the salience of EU decision-making, and in terms of the pace of the policymaking. Although the refugee crisis initially (until the adoption of the EU-Turkey agreement in spring 2016) also gave rise to highly intense policymaking, it was much more drawn out and even in its initial phase EU decision-making did not reach the intensity of the first wave of the Covid crisis.
However, decision-making during the two crises was comparable in a number of respects. First, and unexpectedly, in terms of polarization and conflict intensity, the Covid crisis was not exceptional. As a matter of fact, during the first wave the greater symmetry of the incidence of the Covid-19 crisis did not prevent the development of polarization and conflict intensity to levels comparable to the refugee crisis. Instead, the politicization of the policymaking process proved to be heavily dependent on the policy domain in the first wave of the Covid crisis. In domains where the EU has greater competence, not only did we observe a greater role of EU institutions, but also less conflict and less resistance by coalitions of member states. Thus, across crises but especially in the Covid crisis, policy domains where the EU's competences are restricted are more conflict prone among member states.
Second, we observed an executive dominance in both crises, and above all among member state governments. Remarkably, executive actors representing member state governments were more dominant than executives representing EU institutions, especially in the refugee crisis-a result which confirms the pivotal role of key leaders from member state governments in EU decision-making, especially in policy domains where the EU has low competence.
Third, coalitions of critical member states have been crucial for possible solutions. Strikingly, in both crises, large majorities of actors involved in the domainspecific policymaking supported the EU proposals on the table. In both crises, however, minority coalitions of member states became decisive for the adoption of joint solutions in the policy domains, where the EU has comparatively weak competences and/or where such coalitions benefit from the asymmetric incidence of the crisis. Given the super-majority requirements of EU decision-making in the Council and unanimity requirements in the European Council, these coalitions got decisive leverage and exerted outsize influence. Depending on the policy domain and the type of crisis, opposition came from different coalitions of member states, but, remarkably, the V4 Sovereignty coalition constituted a critical opposition in both crises.
Finally, there is a key difference between the crises: in the refugee crisis, the opposing coalition was able to prevent any kind of reform to the joint asylum policy and allowed only for a temporary solution provided by the EU-Turkey agreement. In the first wave of the Covid crisis, the opposition was more amenable to a joint solution in the crucial fiscal policy domain, where the conflict became most intense. We suggest that this key difference is ultimately rooted in the character of the original problem pressure, and the different distribution of spatial incidence in the two crises. The higher urgency of the problem pressure, the more limited politicization of national identities (although there were attempts by national oppositions to politicize them in the Covid crisis as well), and the more symmetric distribution of the crisis incidence (above all in terms of public health) constrained the role of top executives of member state governments compared to the refugee crisis, which facilitated the task of the core coalition to impose a joint solution against the minority opposition, even if the conflict intensity and polarization of positions was also very high in the fiscal policy domain.