Differentiated integration in the EU – What do the member states ‘think’ about it?

A growing literature is studying the phenomenon of differentiated integration (DI) in the European Union (EU). Empirical studies have focused on charting the degree of existing differentiation, often understood as exemptions from common EU rules. The present report seeks to complement this literature in two ways: first, the report develops a holistic conceptualization of DI by distinguishing between polity and policy differentiation. Similarly, we distinguish two mechanisms through which differentiation can be realized, one satisfying demand for the status quo (‘opt-outs’) and the other satisfying demand for more integration (‘enhanced cooperation’). Second, the existing literature has paid relatively little attention to what the EU member states actually ‘think’ about DI. We contend that member states’ preferences about DI cannot be read off of differentiated EU rules, given that institutional rules and international bargaining processes intervene. Therefore, the report poses two empirical questions: What positions do EU member states take on DI? And why? To answer these questions, the report uses new data from 27 member states on expressed governmental preferences about differentiated integration between 2008-2020. With regard to the first question (WHAT), our findings are threefold: First, polity differentiation is seen slightly negatively on average by the member states. Second, there is significant cross-country variation in member state preferences toward polity differentiation. Third, the mechanism of ‘enhanced cooperation’ is viewed much more favourably than the ‘opt-out’ mechanism. With regard to the second question (WHY), our findings are twofold: First, member states’ positions are not only or predominantly determined by their structural characteristics (e.g. size, wealth, identity), but rather by the expected impact of DI on polity legitimacy and effectiveness. The second key finding is rooted in the inductive identification of factors that have not yet been prominently discussed in the literature. We find that ‘integration experience’, different ‘visions of the EU’, and the ‘shadow of polity DI’ play a role in how EU member states assess differentiated integration. Overall, these factors point to the conclusion that the formation of member state preferences does not happen in a vacuum but is strongly influenced by existing or expected European interdependences.

Integrating Diversity in the European Union (InDivEU) is a Horizon 2020 funded research project aimed at contributing concretely to the current debate on the 'Future of Europe' by assessing, developing and testing a range of models and scenarios for different levels of integration among EU member states. InDivEU begins from the assumption that managing heterogeneity and deep diversity is a continuous and growing challenge in the evolution of the EU and the dynamic of European integration.
The objective of InDivEU is to maximize the knowledge of Differentiated Integration (DI) on the basis of a theoretically robust conceptual foundations accompanied by an innovative and integrated analytical framework, and to provide Europe's policy makers with a knowledge hub on DI. InDivEU combines rigorous academic research with the capacity to translate research findings into policy design and advice.
InDivEU comprises a consortium of 14 partner institutions coordinated by the Robert Schuman Centre at the European University Institute, where the project is hosted by the European Governance and Politics Programme (EGPP). The scientific coordinators of InDivEU are Brigid Laffan (Robert Schuman Centre) and Frank Schimmelfennig (ETH Zürich).

Introduction *
The origins of European integration can be traced to the collective desire to promote peace and prosperity among European nation-states after the devastation of the Second World War. This was certainly no small order even among the six original member states of the European Economic Community. However, over the last seven decades, the emerging European community has both widened its membership and deepened its ambitions. In squaring the circle of growing mutual dependence and increasing diversity, the integration process has become less uniform over time. The literature refers to this phenomenon as differentiated integration (i.e. Dyson and Sepos 2010;Fabbrini and Schmidt 2019;Kölliker 2001;Leuffen, Rittberger, and Schimmelfennig 2013;Schimmelfennig and Winzen 2020a and b;Stubb 1996).
Differentiated integration (DI) is commonly understood to mean that the rules of the European Union (EU) do not apply equally to all member states. For example, the rules which govern the eurozone apply mainly to member states which have adopted the common currency. As such, DI enables integration to proceed where we would otherwise see deadlock due to the divergent interests and capabilities of the member states. EU member states which are not ready to commit to more integration can negotiate exemptions for treaty provisions. These opt-outs help to accommodate cultural, political, and economic differences between the member states. They can contribute to the perception of self-determination in the respective state and lead to greater acceptance of the EU (Schraff and Schimmelfennig 2020b). The case of Brexit, however, shows that this process may also lead to differentiated disintegration (Leruth, Gänzle, and Trondal 2019;Schimmelfennig 2018). In addition to differentiation among EU member states, several non-EU states have decided to selectively follow EU rules and regulations. For example, Norway and Switzerland apply the rules of the Schengen Area of free movement. Such external DI can help to bind non-member states closer to the Union and realize mutual benefits even in the absence of full membership. Arguably, for non-member states, this entails the risk of 'domination' as they had no voice in shaping the EU rules which apply to them (Fossum 2015). In short, DI is a policy tool with advantages and disadvantages, so it is worth exploring how member states perceive this tool and position themselves towards it.
Research has shown that differentiation that the differentiation of the EU's primary and secondary law has increased over time (Leuffen, Rittberger, and Schimmelfennig 2013;Schimmelfennig and Winzen 2020b). Moreover, increasing differentiation is often seen as an important driver of the integration process. Without differentiation, the argument goes, agreement on the European Monetary Union or the Schengen Area of free movement would have come later, if at all. But does the increasing differentiation of the EU suggest that member states view differentiated integration positively? We contend that it is not straightforward to read integration preferences off of policy outcomes, as the intervening process of intergovernmental bargaining is likely to necessitate compromises. Given the advantages and pitfalls of DI, which we described above, it is surprising that little research has inquired into what EU member states actually think about differentiated integration (but see Leruth 2015). This paper seeks to address this gap in the existing literature by using new empirical data on member states' expressed preferences about differentiated integration.
The paper asks two questions: First, what positions do member states have about DI? Second, which factors might explain their positions? To answer these questions, the paper goes beyond the extant literature's focus on exceptions from EU rules. On the one hand, we distinguish DI rooted in demand for the status quo (or less integration) from DI rooted in demand for more integration. 'Opt-outs' and Robert Schuman Centre for Advanced Studies Working Papers 'Enhanced Cooperation' are two mechanisms through which these different demands can lead to DI 1 . On the other hand, we distinguish the diverse integration preferences of member states with regard to specific policies (policy level) from what member states think about the impact of DI on the EU as a whole (polity level).
The paper is structured as follows: Section two reviews the conceptual and explanatory literature on DI and formulates a series of expectations. The third section presents our data and methodology. In the fourth section, we show the results of our analysis. Section five summarizes and discusses our findings. The last section draws some conclusions from our results and points to future research avenues.

Theory -Member states' positions on differentiated integration
This paper seeks to map the positions EU member states take on DI. We understand this focus as complementary to research centred on differentiated outcomes in EU rules (Schimmelfennig & Winzen 2020a and b). The gap between what member states want and what they get in terms of DI may be indicative of the balance of bargaining power among the member states. Therefore, in this section we first outline our conceptualization of DI and the mechanisms through which DI can be realized. We then review the literature for factors that may determine member states' positions on DI.

What is differentiated integration and how is it realized?
The existing literature has discussed differentiated integration, both as an outcome of the diverse integration preferences of the member states and in terms of its impact on the EU as a whole. We capture these different angles by introducing a conceptual distinction between policy differentiation and polity differentiation.
Policy differentiation is understood as the differentiated participation of member states in EU policies. As will be discussed in more detail in the next section, policy differentiation is seen as originating in the diversity of member state preferences concerning policy integration (Schimmelfennig & Winzen 2020b). Consequentially, policy differentiation is most likely to occur in policy domains where the centripetal forces of functional integration contrast with the centrifugal forces of sovereignty. These domains are the so-called 'core state' powers (Genschel and Jachtenfuchs 2016), such as security (i.e. police, defence, rule of law) and resource extraction (fiscal and redistributive policies). Therefore, our analysis will focus on EU policies affecting core state powers.
Polity differentiation refers to the impact of differentiation on the nature and the functioning of the European Union as a whole. This perspective builds on Alexander Stubb's (1996) influential categorization of differentiated integration along the dimensions of time, space, and matter. Differentiation by time means that not all EU member states need to pursue deeper integration at the same time. While member states can follow different pathways to integration in this perspective, the expectation is that an 'ever closer Union' remains a central aspiration of the member states. Ultimately, all member states will arrive at a state of deeper and uniform integration. This type of polity differentiation can be summarized as 'multi-speed Europe'. By contrast, differentiation by space and matter suggests the possibility of a permanent difference in the level of integration between the member states. Permanent differentiation can take two forms. On the one hand, it can refer to the difference between a deeply integrated 'core Europe', surrounded by 'concentric circles' of peripheral member states. On the other hand, it can refer to the co-existence of multiple overlapping regimes of integration in which member states can pick-and-choose policies. This type of polity differentiation can be 1 While the two mechanisms are analytically similar in that they accommodate diverse integration preferences, we submit that they are likely to be perceived as serving different purposes by the member states. More on this distinction below.

Differentiated integration in the EU -What do the member states 'think' about it?
European University Institute 3 summarized as 'multi-end Europe'. Hence, the two models (i.e. multi-speed and multi-end Europe) point to a fundamental difference in the impact of DI on the nature and functioning of the EU.
Importantly, the positions member states hold about polity DI are not necessarily congruent with their positions on policy DI. There are several reasons for this: First, policy and polity objectives may not overlap. While policy differentiation can achieve immediate policy objectives, the accumulated effect of a large number of exceptions may undermine the effectiveness and legitimacy of the European Union as a whole. Second, due to domestic electoral pressures, member states may pursue exceptions from EU rules even if they view the long-term effects critically. For example, while Ireland has sought opt-outs from various EU policies to ensure alignment with the UK, Irish governments are nevertheless highly critical of a 'two-speed Europe' as they worry about losing influence in what they call the 'slow lane'. Third, member states may seek to protect the effectiveness and legitimacy of the EU polity by demanding the exclusion of new member states from common policies in cases where the latter are deemed incapable of adequately implementing the common policy. In short, due to conflicting objectives, time inconsistency, and capacity concerns, member states are likely to engage in more policy differentiation than their polity preferences would suggest.
Next, we turn to the question of how differentiation is realized. There are two different perspectives on this issue. On the one hand, de facto differentiation may exist despite uniform rules due to differences in implementation (Sabel and Zeitlin 2010) and compliance (Falkner et al. 2005;Tallberg 2002). On the other hand, research has focused on de iure differentiation of EU rules. In this paper, we consider two legal mechanisms through which de iure differentiation can be realized: 'opt-outs' and 'enhanced cooperation'.
Frank Schimmelfennig andThomas Winzen (Schimmelfennig andWinzen 2014, 2020b) have investigated to what extent member states have obtained exceptions from EU rules. They distinguish between voluntary and discriminatory opt-outs from community rules. Voluntary opt-outs are rooted in member state preferences and usually 2 occur in the context of EU treaty change (deepening integration). For this reason, they are conceptualized as 'constitutional differentiation'. Discriminatory opt-outs are rooted in concerns about the capacity of a member state to implement a common policy. They usually occur in the context of EU enlargements (widening integration) and are, thus, conceptualized as 'instrumental differentiation' (Schimmelfennig and Winzen 2014).
While such exceptions are an important aspect of differentiated integration, this perspective neglects differentiation resulting from demand for more integration. Bruno de Witte (2018, 2019) identified various legal mechanisms that can accommodate differentiated demand for more integration. 'Enhanced Cooperation' is the most relevant of these mechanisms 3 . It allows member states to pursue deeper integration in a small avant-garde group, usually requiring the consent of the non-participating states. Analytically, 'opt-outs' and 'enhanced cooperation' are similar because in both cases there are some member states demanding more integration and others demanding less integration 4 . However, we argue that they differ with regard to their (perceived) purpose and nature: 'opt-outs' have usually been used to achieve unanimity on treaty change. As such, 'opt-outs' are understood as exceptions from a general agreement on more integration. By contrast, 'enhanced cooperation' is a regular treaty instrument allowing some member states to do more than others. The purpose is not to overcome resistance to treaty change but to enable deeper cooperation within the current treaty framework. Hence, we suggest that 'opt-outs' are likely to be perceived as an exception to the general desire for more integration. By 2 The Cameron re-negotiation (2015-2016) of the UK's position in the EU may be seen as an example of a MS seeking exemptions from common rules without treaty reform. 3 They include so-called inter-se agreements between EU member states outside the framework of the EU, such as the Fiscal Compact. Moreover, the EU can conclude external agreements with non-EU states which involve the export of EU rulesexamples are the Trade and Cooperation Agreement with the United Kingdom and the Eastern Partnership. While being beyond the scope of this report, future research should cover 'inter se agreements' and 'external agreements'. contrast, 'enhanced cooperation' is expected to be perceived as an exception to the status quo of integration in favour of deeper integration. In this perspective, 'opt-outs' are associated with resistance to more integration, while 'enhanced cooperation' is associated with the realization of more integration. In sum, while the two instruments are analytically similar, we except that the member states perceive them as serving different purposes and, therefore, assess them differently.

Determinants of member states' positions on differentiated integration
Arguably, one of the most fruitful research agendas about DI has been pursued by Frank Schimmelfennig and his various co-authors (Holzinger and Schimmelfennig 2012;Leuffen, Rittberger, and Schimmelfennig 2013;Schimmelfennig and Winzen 2020b;Schraff and Schimmelfennig 2020). Their perspective distinguishes between demand-side and supply-side factors in explaining the emergence of differentiation.
Demand for differentiation is rooted in the heterogeneity of member states' preferences, capacities, and dependence. Concerning preferences, some member states may not wish to integrate more and can opt out from further integration in specific policy areas. An example is the Danish opt-out from the Eurozone. Concerning capacity, certain member states may wish to integrate more but are perceived as incapable of implementing common policies properly. An example is the continued exclusion of Romania and Bulgaria from the Schengen Area. Concerning dependence, some member states may need common solutions less than others. Such member states may try to 'go it alone' and refuse to cooperate. Arguably, the domestic perception of low dependence on European solutions has played a role in the UK's bid for differentiated disintegration (Schimmelfennig 2018).
In their latest book, Schimmelfennig and Winzen (2020b) build on the three grand theories of regional integration to locate the sources of relevant heterogeneity among the EU member states ( Table  2).
The first is realist intergovernmentalism. This theory holds that states are most interested in their own autonomy and security. The relevant source of heterogeneity in this regard is the size of the member state. For Europe, the authors assume that the United States provides military security. Hence, in the EU, the relevant indicator of size is not purely military capability, but rather the overall size of the economy. The expectation is that bigger statesbut not smaller statescan credibly threaten to reject integration if their terms are not met.
The second is liberal intergovernmentalism. This theory proposes that states care mainly about the economy and the welfare of their citizens. Accordingly, the relevant source of heterogeneity is the wealth of the member states. The expectation is that wealthier statesbut not poorer statesprefer high labour and environmental standards and can afford a robust welfare state.
The third is post-functionalism. This last theory puts the focus on democratic self-determination. In democracies, voters have a voice in determining the boundaries of their political community and deciding on the policies enacted within these boundaries. Hence, the relevant source of heterogeneity is Differentiated integration in the EU -What do the member states 'think' about it?
European University Institute 5 the political identity of the voter. The expectation is that voters who identify as more national (as opposed to more European) will prefer to protect national sovereignty over further European integration.
In sum, states care about strategic autonomy, the economy and citizens' welfare, and selfdetermination. The relevant sources of heterogeneity are size, wealth, and identity. This perspective implies that we should expect variation in the demand for integration between smaller and bigger MS, more and less wealthy MS, and member states with more or less Europeanized identities. On the supply-side, Frank Schimmelfennig and Thomas Winzen identify three factors that impact the realization of differentiated integration. First, differentiation requires that a critical mass of member states pursue integration to sustain the necessary institutions. Second, differentiation is more likely to occur if the consequences for non-participating member states are small. Highly negative externalities provide a strong incentive for initially reluctant states to join the avant-garde group. Highly positive externalities provide a strong disincentive for the avant-garde group to pursue deeper integration. The third supply-side factor is the institutional context. This covers decision-making rules (unanimity enhances bargaining power of status quo states), the role of supranational actors (which prefer uniform integration over differentiation), and path-dependence (prior DI creates incentives for more DI). These supply-side factors are the main reason why we cannot read member states' preferences about differentiated integration off of the reality of differentiated EU rules.
Another strand of research investigates the conditions under which differentiation can become an acceptable mode of cooperation among democratic states (Bellamy and Kröger 2019;Fossum 2015). In this view, the acceptability of DI depends on two main factors: On the one hand, differentiated integration is more acceptable if it increases the effectiveness of the EU to respond to common challenges. When policy challenges have an asymmetric effect on the EU, differentiated integration can help achieve an effective response by the more affected member states. However, to the extent that differentiation leads to the creation of new institutions or differences in the rights and duties of member states (and citizens), DI can increase transaction costs and negatively impact on the EU's / MS's capacity to respond effectively.
On the other hand, DI is more acceptable if it does not negatively impact the exercise of democratic self-determination. This issue is most acute when differentiation involves high costs for nonparticipating states (Kölliker 2001). John Erik Fossum (2015) suggests that if states have no say in designing policies that impose negative costs on them, differentiation can lead to 'domination'. For example, because Norway is not an EU member state, it does not sit at the table when EU rules are determined. However, Norway is subject to EU rules as a member of the European Free Trade Association participating in the European Economic Area. In short, differentiation should be acceptable if it does not negatively impact the effectiveness and legitimacy of the Union and its constituent members. Robert Schuman Centre for Advanced Studies Working Papers This perspective gives less weight to structural factors, such as size, wealth, or identity. Instead, assessments about the expected impact of DI on polity efficiency and legitimacy are key in determining member states' DI positions. A variety of factors may influence these assessments. On the one hand, it might be the case that general assessments of DI improve over time, as it becomes a 'normal feature' of the EU. On the other hand, party ideology may play a crucial role in how member states assess the impact of differentiation.

Expectations
This section integrates our conceptual framework with the literature on explanatory factors and formulates a series of expectations regarding MS positions on DI.

Member states' positions on polity differentiation
At the polity level, 'multi-speed Europe' and 'multi-end Europe' constitute two divergent models for how differentiation may affect the nature and functioning of the EU. First, we look at the expected effect of DI on polity effectiveness and legitimacy. The 'multi-speed Europe' model assumes a continued commitment to the notion of an 'ever closer Union', while the 'multi-end Europe' model stresses the accommodation of domestic preferences for non-integration in particular policy fields. As such, a 'multiend Europe' would permanently increase transaction costs between the various regimes of integration and involve a greater risk of disintegration as functional interdependencies weaken. In the view of realist and liberal intergovernmentalism, member states are keen to preserve the political (i.e. peace) and economic (i.e. prosperity) gains of integration. Hence, member states should value a 'multi-speed Europe' more than a 'multi-end Europe'. From the perspective of postfunctionalism, however, domestic sovereignty concerns would tilt member states favouring a 'multi-end Europe'. We formulate two competing expectations: H1: Member states assess the 'multi-speed Europe' model more positively than the 'multi-end Europe' model because the latter constitutes a permanent challenge for the effectiveness and legitimacy of the EU.
H2: Member states favour the 'multi-end Europe' model because it accommodates national preferences for non-integration.
Next, we look at the expected impact of size, wealth, and identity on member states' positions about polity differentiation. First, we look at the 'multi-end Europe' model. We expect that member states with a greater share of citizens identifying as 'exclusively national' (henceforth 'more national MS') are more favourable of the multi-end EU model because it is perceived as preserving national sovereignty. In the realist intergovernmentalism perspective, member states with a larger national economy (henceforth 'larger MS') are generally less dependent on integration than small member states. Thus, the 'multi-end Europe' model allows larger MS to selectively reject integration on issues where it is not perceived to be in the national interest. By contrast, small member states are often more dependent on integration. For them, the multi-end model involves the risk of a permanent loss of international political leverage in the EU institutions. Lastly, we expect that member states with a higher GDP/capita (henceforth 'wealthier MS') view 'multi-end Europe' more positively than less wealthy MS, as it allows them to move ahead with integration in more demanding policy fields. Similarly, we expect poorer member states to be critical of a multi-end EU due to fears of a persistent 'second-class membership' originating in lacking policy-implementation capacity: H3: Member states with a higher national identity, or greater size, or greater wealth view multi-end DI more positively than member states with a higher European identity, or smaller size, or lower wealth.
Next, we look at the 'multi-speed Europe' model. Member states with a more European identity are expected to be more favourable towards the multi-speed model than MS with a more national identity, as it is perceived to move integration forward. As above, smaller and poorer MS should view the 'multispeed Europe' model more negatively due to concerns about declining international political leverage and 'second class' membership status: H4: Member states with a lower national identity, or greater size, or greater wealth view multispeed DI more positively than member states with a higher national identity, or smaller size, or lower wealth.

Member states' positions on policy integration
At the policy level, we use assumptions from the three grand integration theories outlined above to formulate expectations about how size, wealth, and identity matter for the integration preferences of the member states. In contrast to our expectations about polity differentiation, the focus here is not what member states think about policy differentiation but what they think about policy integration. In doing so, we build on the research by Schimmelfennig and Winzen (2020b) and formulate hypotheses about how size, wealth, and identity influence demand for policy integration in three different core state power' policy fields: defence policy, fiscal policy, and the rule of law. Moreover, the focus is on policies where differentiation has actually materialized. While differentiation can and did occur in other policy domains, this focus has the advantage of consistently integrating 'core state power' theory with the assumptions of three grand integration theories and applying it to actual cases of policy differentiation.
Realism holds that size matters for external security. Being small is dangerous in the international arena. Hence, we would expect that smaller member states are more positively inclined towards the EU's Permanent Structured Cooperation (PESCO). Liberalism holds that wealthy member states seek higher regulatory standards. Hence, we expect that wealthier member states are more positively inclined toward the Fiscal Compact, which sets stricter standards for eurozone governance. Finally, postfunctionalism holds that identity matters for issues of self-determination. Hence, we expect that member states with a higher share of citizens identifying as 'exclusively national' are critical of the European Public Prosecutors Office (EPPO), which can be seen as an external constraint on national selfdetermination 5 .
H5: Smaller member states view cooperation in defence policy as a way to benefit from pooled resources and view PESCO more positively.
H6: Wealthier member states seek stricter standards for the eurozone governance and view the Fiscal Compact more positively.
H7: Member states with a more national identity seek to reduce external constraints on domestic political processes and view EPPO more negatively.

Member states' positions on DI mechanisms
Next, we look at the mechanisms through which DI can be realized. We start with expectations related to the effects of differentiation on polity effectiveness and legitimacy. Comparing the two DI mechanisms, we expect that member states are more positively inclined towards 'enhanced cooperation' than 'opt-outs'. This expectation is rooted in the assumption that 'opt-outs' are seen as unity-threatening preferential treatments for reluctant member states. In contrast, enhanced cooperation is seen as a way to move integration forward, realizing mutual gains.
H8: Overall, member states view 'enhanced cooperation' more positively than 'opt-outs'. 5 Other relevant initiatives would be Rome III or matrimonial property regimes. Unfortunately, our data does not allow us to include them in the analysis. Robert Schuman Centre for Advanced Studies Working Papers Moreover, the literature on the acceptability of differentiation suggests that positive assessments of the impact of DI on the effectiveness and legitimacy of the EU should result in more positive assessments of differentiation.
H9: Member states which expect 'enhanced cooperation ' and 'opt-outs' to positively impact EU effectiveness and legitimacy view the respective DI mechanisms more positively than those who assess the impact negatively.
Relatedly, Schimmelfennig and Winzen (2020b) find that prior opt-outs are likely to lead to more optouts, supporting Jensen and Slapin's (2012) notion of 'cascading' differentiation. However, the distinction between voluntary and discriminatory opt-outs is crucial. We expect that member states which are subject to discriminatory opt-outs view opt-outs more negatively 6 .
H10: Member states with discriminatory opt-outs view opt-outs generally negatively. Schimmelfennig and Winzen (2020b: 42 and 46) expect that states with "a stronger national identity", "a larger size", and "higher affluence" are more likely to demand and achieve voluntary treaty opt-outs. Moreover, the authors expect that discriminatory opt-outs will mainly "apply to poorer new member states". Hence, the former group is expected to have a more positive perspective on 'opt-outs'.
H11: Member states with higher national identity, or greater size, or greater wealth view 'opt-outs' more positively than member states with a lower national identity, or smaller size, or lower wealth.
We assume that wealth, size, and identity play out differently concerning states' positions on 'enhanced cooperation'. Realism suggests that smaller (and poorer) member states are more dependent on European integration as they would benefit disproportionately from pooled resources. In addition, regional integration is a way for small and poor states to increase their international political leverage. Finally, member states with a lower share of the population identifying as 'exclusively national' are expected to be more favourable towards enhanced cooperation, because they see it as a way to realize mutual gains and move integration forward.
H12: Member states with lower national identity, or smaller size, or smaller wealth view 'enhanced cooperation' more positively than member states with a higher national identity, or greater size, or greater wealth.

General expectations
As differentiated integration is becoming a more 'normal' feature of the EU, we expect that member states' assessments improve over time. Crucially, this is likely to apply to all member states independent of their time of accession. The reason is that after the eastern enlargements of the 2000s, diversity in the EU has reached unprecedented levels. The subsequent succession of major European crises (Zeitlin, Nicoli, and Laffan 2019) has demonstrated the paralyzing implications of 'unity in diversity' to the member states. In this context, all member states are expected to become more open to DI as a tool for overcoming deadlock in the EU. This is likely to apply at the more abstract level of DI models and DI mechanisms, but not to DI instances.

H13: Assessments of DI models and DI mechanisms improve over time.
Several considerations speak against a central role of party ideology in determining member states' DI positions. First, European main-stream parties are generally Europhile, independent of their ideological orientation. Second, DI is not a highly salient political issue on which parties would take clear positions. For these reasons, it appears more plausible that member state perceptions of DI are based on functional cost-benefit analyses rather than on party ideology.
H14: Party ideology is not central to member states' assessments of DI. 6 A third category of 'informal voluntary opt-outs' may be introduced for those four member states (Czechia, Hungary, Poland, Sweden) which have a treaty obligation to join the eurozone but are reluctant to do so.
As stated before, the ambition of the report is not to provide conclusive evidence for each individual hypothesis. Rather, its purpose is to formulate a wide range of theoretical expectations and to let our available empirical data speak to them. We hope the report opens new avenues of research that can further refine both the theory and the empirics.

Data and Methods
This paper explores what EU member states think about differentiated integration. It does not look at the existing degree of legal differentiation as a measure for differentiated demand for integration. Instead, it looks at 'expressed preferences' about DI, i.e. what member states say they want. In other words, the paper turns the focus back in the political cycle, focusing on outputs instead of outcomes. The study covers all 27 EU member states. Temporally, it is limited to the time period between 2008 and 2020.

Data
This paper is based on the results of 27 country reports produced in the framework of the InDivEU project 7 . Here we describe how the data for these reports were collected and prepared for this comparative study.
There are different ways in which data on member state preferences about DI can be obtained. First, we need to clarify what we mean by member state preferences. Member states are not unitary actors. Even in centralist states, a plethora of economic and societal actors pursue a multitude of agendas. Nevertheless, for the purpose of this report, we suggest that the role of the government is paramount. Crucially, it is the governments of the member states which ultimately decide on and shape European affairs. Moreover, governments serve as important conduits for sectoral interests towards the EU. Therefore, our analysis focuses on the expressed preferences of member state governments. We use a broad definition of government, understood as a set of institutional actors. This includes the executive (i.e. Prime Ministers and ministers) as well as the legislative (i.e. Parliament). However, we do not distinguish between consecutive governments. For this reason, our unit of analysis is the member state.
One obvious way is to solicit the information directly from public officials via expert interviews or surveys. This approach proved unfeasible for several reasons. First, securing access to high-level public servants in all 27 EU member states was deemed unlikely, especially because differentiated integration is not limited to any one policy area but can potentially occur in all policy areas 8 . Hence, it would have been necessary to identify multiple experts for each of the twenty-seven member states. In addition, it was assumed that differentiated integration is not a 'bread and butter' issue for most governments. Hence, government representatives may not provide clear positions about the various aspects of DI this paper is investigating.
Due to these constraints, we opted for document analysis as our method of choice. The method has several advantages. First, it allowed us to retrieve the official governmental position (to the degree that a position existed) as expressed in various key documents, such as government programs or speeches and statements by Prime Ministers. Second, by expanding the scope of documents to parliamentary minutes, document analysis also allowed us to obtain 'unfiltered' positions as they emerged in domestic 7 The reports can be found here in the EUI repository (https://cadmus.eui.eu/handle/1814/812) and on the InDivEU project website (http://indiveu.eui.eu/integrated-database/). Unfortunately, the United Kingdom is not part of this study. 8 Qualification: Differentiation is expected to pertain especially to so-called "core state powers" (Genschel and Jachtenfuchs 2016). However, core state powers comprise a wide range of policy areas, such as defence policy, fiscal and budgetary policy, and police and security policy, or the rule of law. Robert Schuman Centre for Advanced Studies Working Papers political debates between governments and the opposition 9 . Finally, all these types of documents are relatively easy to access in all EU member states 10 .
The choice for document analysis allowed us to implement the study in all 27 member states within the given time and resource constraints. Each country analysis was conducted by one or two country experts, which were recruited based on language skills and familiarity with EU integration theory. They were provided with detailed guidelines for data collection and analysis. Additionally, an online portal was created to facilitate continuous communication among the contributors and the project organizers. To ensure comparability of results across countries, the guidelines indicated specific keywords to be searched for within the selected documents 11 . All keywords were translated by the country experts into the relevant local languages 12 .
This approach produced rich qualitative data on all 27 member states. As such, the 27 member state reports constitute the main source of data for this paper. However, the 'richness' of our qualitative data makes it difficult to systematically compare member states. Therefore, we opted to condense the qualitative accounts into quantitative measures for the key parameters of interest. To this end, we designed a survey questionnaire 13 which operationalized DI positions, using a 5-point ordinal scale as proposed by Benjamin Leruth (2015) for this purpose. Under this scale, the stances of countries towards a specific dimension of DI are coded as (1) very negative, (2) negative, (3) neutral, (4) positive, or (5) very positive. The questionnaire enabled us to quantify the qualitative information contained in the 27 reports. We asked the authors of all 27 reports to complete the survey based on the data they have gathered for their country report. To increase reliability, the survey was repeated by an internal coder based on the information contained in the member state reports. The authors of the country reports were then contacted to resolve the remaining conflicts.
To answer how a member state's size, wealth, and identity impact its positions on DI, we followed Schimmelfennig and Winzen's (2020b) indication. For size, we used the size of the economy as measured in overall GDP (logged) 14 . For wealth, we used GDP per capita 15 . For identity, we used Eurobarometer data 16 on the share of the population identifying as "[Nationality] only". In each case, we used the average value for the time period 2010-2019. In sum, our analysis is based on three types of data: qualitative member state reports, survey results, and measures for member state size, wealth, and identity.

Limitations
One important disadvantage of our approach pertains to language and vocabulary. First, not all English keywords have closely related terms in all local languages. In such cases, the country experts did their best to find the most similar local expression. Second, the keyword approach risks that references to DI 9 See Appendix 1 for an overview of the document categories. 10 This applies particularly to government programs and key speeches by Prime Ministers. With regard to parliamentary minutes, differences in the quality and accessibility of repositories were expected. The implementation of the study showed that these problems were manageable in most cases. 11 See Appendix 2 for an overview of the keywords. 12 The use of keywords has an additional advantage. By counting the frequency of keywords over time, we obtain a measure of the salience of DI in governmental debates. The assumption being that that the more a concept / keyword is being used, the more importance is given to the policy to which it refers. The results can be found in the individual member state reports. 13  are missed because other terms are used in the debate. Therefore, the country experts were asked to engage in a close reading of selected key documents to identify local keywords as well as indirect references to DI, which do not use specific DI-concepts 17 .
Due to resource constraints, we further limited the analysis to the most common expression in the most likely moments of the debate. To begin with, within the time interval 2008-2020, the analysis of government position was limited to critical years 18 . They were selected according to two criteria. First, they had to cover the same time period in all member states. Second, they correspond to years in which we observe alleviated debates on DI in many member states. Additionally, only the two most prevalent keywords for each DI model were included in the position analysis. The selected references to DI models and mechanisms were coded manually as negative, neutral, or positive. In the country reports, these coded references are supplemented with a narrative section that illustrates how member states' positions have evolved over time.
Finally, we asked the contributors to provide one overall score for the period since the Lisbon Treaty (2008 -2020). This means that the survey data we have obtained does not adequately take the temporal dimension into account. Between 2008 and 2020, however, the EU held three elections to the European Parliament and saw three Commissions. In addition, severe policy challengessuch as the euro crisis, the refugee crisis, an increasingly assertive Russia, Brexit, and Covid-19not only transformed the EU but also impacted member states differently. We sought to address this limitation by asking two additional questions in our survey. First, whether the government position has changed over time. Second, whether party ideology played a role in the government's assessment of DI. Lastly, we use our qualitative data to trace change over time.

Methods of analysis
We use the survey data to provide an overview of the distribution of member states' positions on various aspects of differentiated integration. We also use the survey data to show how DI positions relate to the wealth, size, and identity of the member states. We use simple scatter plots to provide an overview of where individual countries are positioned in this regard. The scatterplots serve as a starting point for testing our hypotheses. However, due to the low number of data points, we do not investigate statistical relationships between the variables. Instead, we triangulate the survey data with qualitative information, drawing on the descriptive analyses in the 27 member state reports. In other words, we use ordinal differences in our variables of interest to pre-structure a narrative country comparison. According to James Mahoney (1999), the combination of ordinal and narrative methods of causal appraisal allows us to take the interactions between multiple contextual factors into account. The advantage is a less deterministic view of causal relations. The disadvantage is a loss of parsimony in causal explanations. As the purpose of this paper lies in the explorative charting of new territory for future researchrather than focussing on an in-depth investigation of specific causal relationsthe combination of ordinal and narrative techniques of comparison was chosen as most appropriate.
We employ two types of narrative country-comparisons: First, we compare countries that seem to fit our expectations with countries that appear to contradict them, based on their positioning on the scatterplot. These comparisons help us to tease out how structural factors (wealth, size, identity) interact with ideational factors (i.e. visions of the nature of the EU) and cost-benefit analyses by governmental actors (i.e. expected effect of DI on legitimacy and effectiveness). 17 These measures yielded a limited number of 'inductive' keywords. Most of them relate to a "core-periphery" distinction.
As such, the term "core Europe" was substituted by "directorate" in Portugal, by "founding states" in Italy, and by "BeNeLux" in Slovenia. In Hungary, "second class membership" was used as one corresponding term of "two-tier Europe". Second, we contrast most-likely and least-likely countries with each other. For example, Hypothesis 3 suggests that large, wealthy, and identity-national member states are expected to be more positive about the 'multi-end Europe' model than small, less wealthy, and identity-European member states. Hence, we identify countries combining all of these attributes and see whether their position on 'multiend Europe' is congruent with our expectations. For example, we would contrast France (wealthy, large, national) with Slovakia (poor, small, European).

Analysis and results
This section presents the results of our analysis. First, it looks at polity differentiation, exploring member states' positions on the two DI models -'multi-speed Europe' and 'multi-end Europe'. Second, the section centres on three differentiated policy areasdefence policy, fiscal governance, and the rule of law. Third, member states' positions on two DI mechanisms -'opt-outs' and 'enhanced cooperation' are analysed. Finally, the analysis turns to the influence of party ideology and change over time. In each sub-section, we first present the positions (the 'What' question) of the member states and then explore possible explanations (the 'Why' question). For each of our hypotheses/expectations, we first provide an overview of the distribution of member states' positions using our survey data. We then use our qualitative data, drawing on 27 member-state reports, to shine a light on cases that do (not) conform with our expectations.

Polity differentiation
Polity differentiation is related to different visions of the nature and functioning of the European Union. The 'multi-end Europe' model suggests that DI is permanent as individual member states can selectively pick and choose policies in which they want to participate. In the 'multi-speed Europe' model, DI is temporary as all member states ultimately strive to arrive at deeper uniform integration. While the former is rooted in divergent (and relatively stable) preferences, the latter is rooted in presumably temporary differences in policy implementation capacity.
We begin the analysis by contrasting two competing hypotheses (H1 and H2). If member states view DI as a tool to overcome temporary capacity concerns, they should have a more positive view of the 'multi-speed Europe' model (H1). If member states view DI as a tool to permanently accommodate divergent national preferences, they should consider the 'multi-end Europe' model more positively (H2).
Panel A in Figure 1 shows the positions of 24 member states 19 about the idea of a multi-end Europe. While six countries view the multi-end model positively, twelve countries view it negatively or very negatively. Panel B in Figure 1 shows the positions of 26 member states 20 on the 'multi-speed Europe' model. Only three member states have a favourable view of this model of DI, while twelve countries view it negatively or very negatively.

Panel A Panel B
On average, the multi-speed model is viewed slightly more positively (2.61) than the multi-end model (2.54). Both models enjoy moderately negative views on average. However, there are significant differences between countries, ranging from very negative to positive assessments for both models. This leaves us with mixed results concerning our first two hypotheses.
The slightly more positive assessment of the 'multi-speed Europe model indicates that member states may view DI primarily as a tool to further integration rather than as a tool for accommodating special wishes. However, the difference in the assessments of the two models is small, while differences between countries on individual models are substantial. More important than the difference between the two models is the fact that both models were assessed in a slightly negative fashion on average. At the same time, only a small group of countries have positive views about either model. This observation supports the view that the member states view DI in general only as a 'second-best' option to uniform integrationindependently of the respective model of DI.
Next, we investigate whether differences in country size, wealth, or identity correlate with assessments of the two DI models. To this end, we look at two complementary hypotheses. Member states with a higher national identity (greater size, greater wealth) view multi-end DI more positively (H3). By contrast, member states with a higher European identity (greater size, higher wealth) view multi-speed DI more positively (H4). Figure 2 shows the distribution of member states' positions for the 'multi-end Europe' model. suggests that economic size is not strongly associated with member states' positions on 'multiend Europe'. However, Germany is something of an outlier here: it is the largest EU country and one of only three member states that view the DI model very negatively. Hence, excluding Germany shifts the distribution in the expected direction, especially since France and Italy take a favourable position on 'multi-end Europe'. Similarly, Slovakia does not conform to our expectations. Hence, we zoom in on these two countries by comparing them with structurally similar countries with different positions on the 'multi-end Europe' model.
The distance between Germany (very negative) and France (positive) is noteworthy. Both countries are large and wealthy member states, with Germans being somewhat more European in their identity. Does the small identity-related difference between French and German citizens explain the difference? A closer view reveals that the critical difference lies in divergent conceptions about the future of the European Union. Recent French governments adhere to an 'exclusivist' vision of the EU, comprising a highly integrated 'core' and a less integrated periphery. By contrast, German governments have a more inclusive vision of the EU, particularly with regard to Central and Eastern Europe. Both countries seek more integration. But France is focused on deepening integration while Germany is focused on keeping everyone on board (Michel 2020;Nagel 2020).
Similarly, Slovakia and Romania are structurally similar, but differ in their assessment of 'multi-end Europe'. Both countries are smaller and less wealthy member states. Slovak citizens feel somewhat more European than Romanian citizens. The former is more sympathetic towards 'multi-end Europe', while the latter tends to hold more negative views. The explanation for their divergent positions appears to be rooted in their accession experience rather than in identity differences. Slovakia has adopted the common currency and is in the Schengen Area, while Romania has neither of these. In Slovakia, support for 'multi-end Europe' is rooted in its experience and aspiration of joining the European 'core' (Janková 2021). By contrast, in Romania "any sort of European integration involving different speeds or different shapes are perceived as a sign of discrimination that would leave Romania in Europe's periphery" (Badulescu 2021: summary).
Concerning wealth, there are some indications that richer countries view the DI model more positivelyprovided we exclude Luxembourg, Ireland, and Germany from the picture. Luxembourg has a very negative view on 'multi-end Europe' because of its strongly European outlook and the associated ambition to "act as an example of European Integration and incentivise other Member States to follow its path" (Michel 2021: summary). Besides Luxembourg, the position of Ireland also does not correspond to our expectations. The negative Irish perspective on 'multi-end Europe' is rooted in concerns that the country may lose its influence in the European arena due to differentiation. On the one hand, Irish governments have worried that the Irish opt-outs (which were intended to ensure close functional alignment with the UK) may relegate Ireland into the 'slow lane of integration'. On the other hand, Ireland views the mechanism of 'enhanced cooperation' critically, as it would undermine Irish veto powers at the EU level. Besides, the failed Irish Lisbon Treaty referendum created concerns that 'enhanced cooperation' may be used by other member states to complete the treaty ratification without Ireland (Telle 2020). In Germany, 'multi-end Europe' is viewed negatively due to its 'inclusive vision' of the Union (Nagel 2020).
Moving on to the effect of identity, we find little evidence that a greater share of citizens identifying as 'exclusively national' is positively correlated with greater support for the 'multi-end Europe' model. Indeed, if we exclude Luxembourg from the analysis, member states with a higher share of more nationally minded citizens seem to take more negative views on the multi-end DI modelagainst our expectation. There may be two explanations for this. First, the postfunctionalist mechanism (that more national identity leads to more demand for national sovereignty which, in turn, leads to more support for DI) requires the politicization of European integration in the mass political arena in the member states. Research has shown, however, that the politicization of European integration remains relatively low in most member states with episodes of higher politicization (i.e. Hutter, Grande, and Kriesi 2016). Hence, we may not see a positive tendency because the postfunctionalist identity mechanism is not active due to low levels of politicization of the EU. But why do we see the opposite tendency? One explanation might be that countries with a large share of citizens with exclusively national identities also have other characteristics which draw them in the opposite direction with regard to differentiated integration. In fact, the majority 21 of those countries are also new member states, relatively poor and small. Romania and Bulgaria fit this description well (high national identity, relatively small, less wealthy, new MS) and also show the most negative assessment of 'multi-end Europe'.
Finally, we contrast most-likely and least-likely countries concerning their positive position on the 'multi-end Europe' model. The most-likely cases combine a big size with high wealth and an aboveaverage share of citizens identifying as 'exclusively national'. France (positive) and Italy (positive) fit the description best. The least-likely cases are small, less wealthy, and less exclusively national. Slovakia (positive) and Hungary (negative) fit this description best. Overall, this comparison provides some support for our hypothesis. Only Slovakia's positive position contradicts our expectations. As shown above, this deviation might be explicable by Slovakia's positive integration experiencewhich is perceived as a transition from the periphery into the core of the EU. Figure 3 shows the results for the 'multi-speed Europe' model. We expected that member states with a higher European identity (greater size, greater wealth) would view multi-speed DI more positively than more national (smaller, less wealthy) MS (H4). 21 Seven of the ten countries with the highest share of citizens identifying as exclusively national are new, small, and poor member states. The remaining three countries are Greece, Austria, and Ireland. Robert Schuman Centre for Advanced Studies Working Papers

Figure 3 Positions on 'Multi-speed Europe' by member state size, wealth, identity
Starting with size, we find some indications that smaller countries view the 'multi-speed Europe' model more positively than larger MSagainst our expectation. This trend would be stronger still if Germany's neutral position were in line with the negative positions of France and Italy. Germany views 'multispeed Europe' more positively than 'multi-end Europe' because it offers a more inclusive EU vision (Nagel 2020). As discussed above, France favours a more exclusive (and deeper) European core. In Italy, the 'multi-speed Europe' model is viewed negatively due to fears that Italyas one of the 'founding countries' of the EUmay be drifting into a more peripheral position due to decreasing state capacity compared to other original members. At the same time, Italy is more favourable of the 'multiend Europe' model, which is seen as more compatible with Italy's ambition to remain one of the core countries of the EU (Volpi 2021).As such, Italy's position on the DI models reflects a perceived decline of state capacity combined with the ambition to remain an influential core state.
However, small countries are also split on the issue of 'multi-speed Europe'. While Bulgaria and Romania hold very negative positions, Croatia, Finland, Latvia, and Luxembourg have favourable positions. While the former group is structurally coherent, the latter group is quite diverse in terms of wealth and identity. Bulgaria and Romania reject the 'multi-speed model' due to their fear of perpetual 'second class' membership (Badulescu 2021;Markova 2020). The four small countries that view 'multispeed Europe' positively see it is as a way to pursue deeper integration.
Concerning wealth, we find some evidence that wealthier member states are more positive about the 'multi-speed Europe' model. Three countries at the opposite extremes of the wealth spectrum fit particularly well. On the one hand, the extremely wealthy Luxembourg views the 'multi-speed Europe' model as a positive avenue for deeper integration. On the other hand, the less wealthy member states of Romania and Bulgaria view it very negatively. In Bulgaria, "differentiated integration models such as two-speed / multiple speeds Europe are perceived by Bulgarian politicians as a threat to equal participation or fertile ground for growing discrepancies. In the rhetoric of national politicians, 'core Еurope' is used in parallel with 'periphery', triggering negative associations with second-class membership in a union of more developed and powerful countries." (Markova 2020: 14).
However, several countries contrast with our expectations. Latvia is an example of a less wealthy MS, which views 'multi-speed Europe' positively. The explanation can be found in the fact that the country's assessment of DI has improved considerably with the successful accession to the eurozone and the Schengen Area. Due to the accession experience, Latvia increasingly considers itself part of the European core rather than the periphery (Bukovskis, Palkova, and Varna 2021 less wealthy countries (Latvia, Bulgaria, Romania), it appears that the different accession experience (continued exclusion for Romania and Bulgaria, but deeper integration for Latvia) may explain the different positions despite the structural similarities of these countries. On the other end of the spectrum, the wealthy member states of Sweden and Ireland do not conform to our expectations. In Sweden, multispeed Europe is seen negatively as Swedish governments considered it to be detrimental to the cohesion with the Union. Swedish governments preferred initiatives to include all member states and feared, in particular, that a political deepening within the eurozone could lead to a "widening rift" with noneurozone countries (Herolf 2021). In Ireland, the 'multi-speed model' (as well as the 'multi-end model') is viewed negatively due to similar fears, combining limited appetite for more integration with the fear that further integration in the European core will negatively affect Ireland (Telle 2020).
Lastly, we expected that member states with a smaller share of citizens identifying as 'exclusively national' would view the 'multi-speed Europe' model more positively. While we find some evidence for this to be the case, our previous findings of the 'multi-end Europe' model provide reasons for caution. It is again Luxembourg, Romania, and Bulgaria which fit our expectations best. However, Spain has a small share of citizens identifying as exclusively national. Nevertheless, the 'multi-speed Europe' model is viewed as slightly negative here. We find an explanation for this outcome "in the idea that Spain does not want to be in the 'second-tier' of DI, fearing being left behind by more committed members" (Leon 2021: 12). This resonates with the Italian case, even though a much larger share of Italian citizens identify themselves as exclusively national. By contrast, both Latvia and Finland have more citizens identifying as 'exclusively national' than the average member-state. However, in these countries, national identity is arguably not in opposition with European identity, but rather an expression of the historical relationship of these countries to their Russian neighbour.
Again, we contrast most-likely and least-likely countries concerning their positive position on the 'multi-speed Europe' model. The most-likely cases combine a big size with high wealth and a belowaverage share of citizens identifying as 'exclusively national'. Germany (neutral) and Spain (negative) fit the description best. The least-likely cases are small, less wealthy, and more exclusively national. Bulgaria (very negative) fit this description best. The comparison provides limited support for our hypothesis. While Bulgaria's position fits our expectation, Germany and Spain are less positive than expected. In Germany, the constraining factor appears to be its 'inclusive vision' of the EU. In Spain, the negative assessment of 'multi-speed Europe' is related to the fear of drifting from the core towards the European periphery 22 .

Policy differentiation
We now move from the polity-level to the policy-level. Specifically, we look at three policies that pertain to different core state powers and for which differences in size, wealth, and identity are expected to matter. As discussed above, size should matter for member states' positions on defence policy, with smaller states seeking to benefit from pooled resources (H5). Wealth should matter with regard to fiscal policy, with wealthier states preferring higher standards (H6). Identity should matter for questions of self-determination and the rule of law, with member states with higher shares of citizens who identify as exclusively national rejecting EU policies (H7). Also, we expect that states will endorse DI if they believe to benefit from it. 22 When presenting an early draft of this report at the EGPP seminar series at the European University Institute in February 2021, we received the suggestion to consider the effect of the duration of EU membership on DI positions and to investigate whether we see a difference in the assessment of DI between the most and the least integrated countries. We looked into these questions and the results of our analyses can be found in the appendices 7 and 8. In short, our results appear to show that neither membership duration nor positionality in the EU are strongly related to the DI preferences of the member states. Robert Schuman Centre for Advanced Studies Working Papers 4.2.1 Defence policy First, we look at member states' positions on integration in defence policy. We do so by focussing on PESCO, which all EU member statesexcept Denmark and Maltahave joined. Agreement on PESCO was reached in 2017, and first defence projects were launched in 2018. The fact that almost all EU member states have joined PESCO would suggest that size does not play a decisive role here. Denmark's non-participation is explained by its opt-out from the Area of Freedom, Security and Justice. Malta is concerned that PESCO may violate a constitutional neutrality clause.
However, five other EU member states have a non-aligned stance in defence policy, meaning that they are not members of military alliances, such as NATO. Besides Malta, these non-aligned countries include Austria, Ireland, Finland, Sweden, and Cyprus. Accordingly, reluctance towards European defence cooperation should be high in these countries. However, all of them have joined PESCO and have taken a favourable position on PESCO. In Ireland, the 2017 rolling-party program of the governing Fine Gael party stated that the party "support[s] Irish involvement in PESCO to combat modern security threats like terrorism, cybercrime, drug and human trafficking provided it does not compromise our military neutrality" (cited in Telle 2020: 19). In Finland, there was some initial opposition to PESCO by the opposition party 'Left Alliance'. Remarkably, however, the Finnish government supported the launch of PESCO (Heinikoski 2020). Finally, Sweden announced its support to PESCO late, not because of general reluctance towards defence cooperation, but because it preferred a more inclusive approach (Herolf 2021). The participation of these 'least-likely cases' may be ascribed to a "German push for inclusivity, which prevailed over a French desire for a higher level of ambition" (Blockmans and Crosson 2019: 4) 23 . Inclusivity was achieved by re-conceptualizing PESCO as a non-binding framework for capacity building, rather than a more ambitious defence cooperation as initially envisioned in the Lisbon Treaty 24 .
The situation is inverted for the small and medium-sized EU member states of Central and Eastern Europe. Due to security concerns, the Central and Eastern European member states are interested in European defence cooperation. However, their limited capacities conflict with more ambitious designs. The examples of Estonia and Croatia illustrate that PESCO was seen here also as a way to move closer to the 'European core'. As such, Estonian Prime Minister Jüri Ratas (Centre Party) stated in May 2017 in the Estonian Parliament: "I think that during the last 13 years that we have been part of the European Union Estonia has always been among the core countries that wish to move on with Europe and to move on fast and which wish to keep the European Union. Talking about Estonia's latest decisions, it can be mentioned that last week the government decided to join the European Prosecutor and the week before that Estonia decided to join PESCO to enhance defence cooperation. I think this is the example in the best way that Estonia has given with its everyday activities and its internal politics" (cited in Reiljan 2020: 16) Similarly, Croatian Prime Minister Andrej Plenković (CDU) said that joining Pesco was "just one of the examples in which Croatia sends a strong message of wishing to be among those members that pursue deeper integration and closer cooperation" (cited in Beroš and Gnip 2020). Moreover, the Croatian government argued that joining PESCO would help to "strengthen and modernise the country's military force" (cited in Beroš and Gnip 2020) 25 . 23 Arguably, the wording of the Treaty of Lisbon would have allowed a more ambitious format. Thus, the establishment of Pesco as a framework for voluntary project-based defense cooperation can be seen as a 'watering down'. The fact that two member states opted-out of this watered-down version of Pesco may indicate that differentiation was a last resort after other strategies had been insufficient in overcoming the distance between the preferences of the member states. 24 Many thanks to Bruno de Witt for pointing out that the framers of the Lisbon Treaty had envisioned a more ambitious version of structured defence cooperation that what eventually was realized in PESCO. 25 It should be mentioned that the opposition cautioned that PESCO may impinge on Croatia's military autonomy and strain relations to the US. In sum, we can distinguish two groups of small states. On the one hand, the group of non-aligned small states was initially hesitant towards PESCO. Their hesitance had two reasons: First, PESCO represented a deviation from their traditional stance in defence policy. Second, especially in Sweden reluctance towards PESCO was rooted in concerns about European cohesion, preferring an inclusive solution. On the other hand, the small states of Central and Eastern welcomed PESCO to integrate deeper, benefit from the security provided by pooled European resources, and upgrade their own military capacity. Here, capacity concerns (wealth) appeared to matter more than size for positions on PESCO. Arguable, the re-conceptualization of PESCO as a more inclusive and less ambitious initiative accommodates the preferences of both groups.

Fiscal Policy
Next, we look at fiscal policy, where we expect wealthier countries to display a more positive favourable on the Fiscal Compact because it sets stricter regulatory standards for the eurozone. In particular, the Fiscal Compact sets out a balanced budget rule and defines corrective mechanisms in case of rule violations. The Fiscal Compact is part of the broader Treaty on Stability, Coordination and Governance in the Economic and Monetary Union (TSCG). It was signed in March 2012 in reaction to the eurozone crisis. While all 27 current member states have signed the TSCG, the fiscal compact applies only to eurozone countries and three voluntary adopters: Bulgaria, Romania, and Denmark. This leaves the third group of countries, which are not bound by the Fiscal Compact, including Croatia, Czechia, Hungary, Poland, and Sweden. Each of these group contains wealthier and less wealthy member states. We look at wealthy and poor countries in each group.
We begin with the wealthier countries of the eurozone. The German government saw the Fiscal Compact positivelyas a way to set higher standards for fiscal and economic policy within the monetary union. By contrast, François Hollande's 2012 presidential programme "pledged that ratification by France of the EU Fiscal Compact should be conditioned on an additional 'Growth Pact'" (Michel 2020: 11). Hence, France was not favourable of a Fiscal Compact which they believe had a strict emphasis on setting more rigorous fiscal standards. While Germany and France disagreed on the substance of the Two factors complicated the Irish situation. First, the Fiscal Compact required a popular referendum due to its constitutional implications. Second, at the time, Ireland received financial assistance from the EU and needed to join the Fiscal Compact to qualify for the European Stability Mechanism. The government (Fine Gael and Labour) and the largest opposition party (Fianna Fail) supported the Fiscal Compact. However, the then-Vice Prime Minister Eamon Gilmore (Labour) cautioned in the lower house of the Irish Parliament (Dáil Éireann, 10/07/2012) that "any future arrangements must avoid fragmentation of the Union and that maintaining the integrity of the single market is a matter of the highest importance" (cited in Telle 2020: 16). In this context, Seán Ó Fearghaíl of the largest opposition party, Fianna Fail, explained that his party "considers the United Kingdom's decision to step outside this [TSCG] treaty with grave concern" (Dáil Éireann, 29/2/12) (cited in Telle 2020: 17). By contrast, Senator David Cullinane of the left-leaning opposition party Sinn Fein vigorously opposed the Fiscal Compact, arguing that the "treaty is intergovernmental and we can simply opt out […]. Britain has opted out […]" (Seanad, 24/4/12) (cited in Telle 2020: 17).
Looking at the less wealthy eurozone states, we find relatively little resistance to higher regulatory standards. Instead, debates focused on the issues of solidarity and responsibility. In 2012, Estonia had already adopted the euro, and Latvia was eager to do so as well. Due to the accession process, two countries had exercised fiscal discipline in the past. As such, Latvia ratified the Fiscal Compact in 2012 "without extensive debate in the Parliament, due to the fact that the country had been introducing severe austerity measures already since 2009 and adopted its own Fiscal Discipline Law mirroring the Fiscal Compact requirements" (Bukovskis, Palkova, and Varna 2021: 15) . The situation was similar in Estonia. Here, however, the European Stability Mechanism sparked a debate about why a poor country like Estonia should help to bail out wealthier countries, such as Greece (Reiljan 2020). A similar discussion took place in Slovakia. Vice Prime Minister Miroslav Lajčák made it clear that being in the eurozone involves responsibilities: "We shall not speculate and think that we can choose the raisins from the cake and not carry any burden. We shall not expect anyone to bake the cake for us. We got into the core EU, into the European integration core and now is the time for us to act accordingly. […] We are in the most exclusive club and we want to stay there. We will not waste energy on ideas about how many speeds there will be or how the future EU might be. We want to have the highest speed" (Miroslav Lajčák, 19.9.2012, cited in Janková 2021. Similarly, Portuguese Prime Minister Pedro Passos Coelho (Partido Social Democrata) argued that the Fiscal Compact "represents our refusal to repeat the mistakes of the past". In saying so, he was not exclusively referring to other EU member states, but also to Portugal's own mistakes. He stressed that "[w]e must not forget that in less than 35 years we have had to ask for external financial support three times" (Silva 2021: 19).
In the group of voluntary participants in the Fiscal Compact, the motivations in Romania and Bulgaria echo those of Estonia and Latvia. As such, Romanian President Traian Băsescu justified participation in the Fiscal Compact in terms of the euro accession process: Bulgarian Minister of Foreign Affairs, Nikolay Mladenov, highlighted that his country was supporting fiscal discipline. However, this should not involve making additional financial commitments to IMF or European programmes. The opposition questioned the governmental strategy of participating in the Fiscal Compact as a vehicle for deeper integration, as leader of the main opposition party, Sergei Stanishev (BSP, 27/01/2012) argued in Parliament: "our status is not clear at all, when we will be invited there, how long we will be present, whether our voice can be heard in this new, imaginary format at all." (cited in Markova 2020: 23) Finally, the analysis turns to the group of member states which have signed the TSCG but are not bound by its fiscal and economic provisions. Sweden is not in the eurozone but has a preference for "strict rules for handling public finances" (Herolf 2021:19). As such, it signed onto the governance provisions of the TSCG in order to get a seat at the table for shaping European fiscal and economic regulations. At the same time, Sweden is not bound by the fiscal and economic provisions of the TSCG. The Fiscal Compact was seen in a different light in the non-euro states of Central and Eastern Europe. In Hungary, Richárd Hörcsik (Fidesz, Head of the European Affairs Committee, 13/02/2012) criticized in the Hungarian Parliament that "the agreement, unfortunately, created another dividing line towards a multi-speed Europe, which we definitely wanted to avoid" (cited in Kyriazi 2021: 12). The Minister for Public Administration and Justice, Tibor Navracsics (Fidesz, 27/02/2012), stated in the Parliament that "we are not willing to put up with being second-class members of the European Union, we are not willing to put up with Hungarians being a second-class nation in Europe. We must fight and the government also aims to ensure that all members, all nations, are equal within the European Union. Let there be no double standard, judge us by the same standard" (cited in Kyriazi 2021: 13) Similarly, the Polish centre-government of Prime Minister Donald Tusk took the position, in 2012, that excluding non-euro states from deeper integration might split the EU into two speeds. The government also worried that it would have no say in shaping eurozone rules, even though these rules would apply to Poland once it acceded to the eurozone (Walecka and Gagatek 2021). The Czech Republic was one of only two countries that did not sign-on to the TSCG in 2012. However, the Czech position was characterized by a lot of volatility, reflecting its ambiguous positions on euro accession. In 2014, a government program listed eurozone accession as well as participation in the Fiscal Compact as policy objectives. However, the 2018 program stated that the Czech government was not pursuing this agenda anymore, as it would require involvement in the European Stability Mechanism (Hlavik and Smekal 2020).
In sum, we find widespread support for stricter fiscal rules. The main dividing line seems to run not so much between rich and poor states, but between states inside the eurozone, states seeking to join the eurozone, and those who are reluctant to joining the euro. The member-states within the monetary union agree on the necessity of stricter rules in eurozone governance. Still, some would have preferred to combine the regulatory side with a 'growth agenda'. In non-eurozone countries, the aspiration to adopt the euro is an important dividing line. Those non-eurozone countries, which seek the common currency view the Fiscal Compact as a steppingstone towards deeper integration into the European core. However, even those non-eurozone states which do not seek to adopt the common currency are critical of the growing gap between the eurozone core and the non-eurozone periphery.

Rule of law
Finally, we explore the effect of identity on member states' positions on the European Public Prosecutors Office (EPPO), established in 2017 under enhanced cooperation. Ireland, Denmark, Sweden, Poland, and Hungary do not participate in EPPO. This out-group (as well as the in-group) is diverse in terms of the share of citizens identifying exclusively as national. Do the out-group countries justify nonparticipation in terms of self-determination?
Ireland is not a member of EPPO due to the country's opt-out from cooperation in the Area of Freedom, Security and Justice. Independent member of Parliament Finian McGrath expressed concern in 2013 that "the European public prosecutor's office is being used in a broader political game. We appear to be moving towards the formation of a super-state.
[…] The establishment of a European public prosecutor's office could be one more piece in that jigsaw" (cited in Telle 2020: 23). Nevertheless, four years later, Irish Prime Minister Enda Kenny (Fine Gael, 21/03/2017) stressed in Parliament that his government "took a decision to allow for the European public prosecutor's office to be established under enhanced co-operation" (Telle 2020)albeit without Irish participation. Sweden initially saw EPPO as "bringing little added value" but later on decided that it would not block enhanced cooperation if other member states wished to proceed (Herolf 2021: 18).
The Hungarian opposition to EPPO is rooted in a broad party consensus that "prosecution [is] a matter of national sovereignty" (Kyriazi 2021: 19). Similarly, the Polish opposition to EPPO is rooted in sovereignty concerns, as EPPO is seen as interfering with Poland's own national prosecutor's office. Stressing the risk of negative externalities for non-participating member states, Marcin Warchoł made a case for mobilizing a blocking minority against EPPO in the Parliamentary Committee on EU Affairs (20/04/2017): "We do not intend to enter into the enhanced cooperation, which is already legally exaggerated as the basis for the establishment of the European Public Prosecutor's Office, but that does not mean that we will be passive in our discussions on the establishment of this institution. It is crucial for us to indicate that Poland, as a country not participating in this cooperation, will not bear any financial consequences related to the establishment of the European Public Prosecutor's Office. The situation is that the question of a blocking minority is open. Our position most optimally takes into account the interests of our country." (cited in Walecka and Gagatek 2021: 22) In sum, the out-group is not characterized by an above-average share of nationally minded citizens. Nevertheless, sovereignty arguments play an important role in rejecting EPPO. Besides sovereignty, two additional motivations were given: these are pre-existing opt-outs in the case of Ireland and the Swedish expectation that EPPO would be an ineffective institution. Robert Schuman Centre for Advanced Studies Working Papers

DI mechanisms
This section centres on two mechanisms through which EU member states can realize differentiated integration. Opt-outs from common rules allow member states to preserve their status quo while other member states pursue deeper integration. Enhanced cooperation allows a limited group of member states to pursue more integration while other member states maintain their status quo. Overall, we expect that member states view 'enhanced cooperation' more positively than 'opt-outs' (H8) because of its expected consequences on the nature and functioning of the EU (i.e. more integration, more effectiveness, more legitimacy). Figure 4 shows how member states assess 'opt-outs' (Panel A) 26 and 'enhanced cooperation' (Panel B) 27 . A slight majority of the analysed member states views 'opt-outs' as negative or very negative. Importantly, only two EU member states appear to view opt-outs positively. Concerning 'enhanced cooperation', most member states have a positive or very positive position on enhanced cooperation, except Ireland, which has a negative position. When comparing the two DI mechanisms, we see that opt-outs are viewed much more negatively (2.38) than enhanced cooperation (3.77) by member states on average.

Panel A Panel B
Next, we look more closely at how the expected impact of the DI mechanisms relates to member states' positions on these mechanisms. We anticipate that member states with more positive expectations about the consequences of DI assess the mechanisms more positively (H9). Figure 5 shows a positive association throughout. Member states which evaluate the impact of 'opt-outs' (Panel A) and 'enhanced 26 No data for Estonia, Finland, Italy, Latvia, the Netherlands, and Portugal. 'Opt-outs' do not have a direct translation in all the member state languages, which made it more difficult to identify and assess references to this DI mechanism. 27 No data for Denmark. cooperation' (Panel B) positively also assess the mechanisms themselves positively. This appears to be the case for effectiveness as well as legitimacy and holds at the EU level and on the member state level.

Experience with informal voluntary opt-outs and discriminatory opt-outs
Not all opt-outs are alike. Schimmelfennig and Winzen (2014) distinguish between voluntary and discriminatory opt-outs. Concerning the latter, we expect that member states subject to discriminatory opt-outs view the DI mechanism of 'opt-outs' generally negatively (H10). Several member states are currently subject to discriminatory opt-outs. This group can be further subdivided into countries comfortable with their discriminatory opt-out and countries which seek to overcome their discriminatory opt-out. Sweden, Poland, Czechia, and Hungary belong in the first group, as they are arguably not "I don't think it's necessary for every country to be part of the eurozone. Whether or not it will be in Hungary's national interest at some time in the future is something we must continually examine and assess, and we may adopt such a decision if it is in our interest to do so." (cited in Kyriazi 2021: 17) As such, the position of these countries may better be described as informal voluntary opt-outs, as they still have a treaty obligation to join the EMU.
The situation is different for four other countries, which are excluded from common policies due to concerns that they would be incapable of implementing them properly. As such, Bulgaria, Croatia, and Romania are neither member of the Schengen Area nor the EMU. Cyprus is not a member of the Schengen Area. This group of countries views opt-outs much more negatively. Bulgaria and Croatia hold a negative view, while Romania has a very negative view of opt-outs. Cyprus, which is subject to just one discriminatory opt-out, has a neutral position. Both in Romania and Bulgaria, accession to the eurozone and the Schengen Area is perceived as 'rights' derived from EU membership. In both countries, acceding to the two regimes of integration are top policy priorities of all governments since the two countries joined the EU in 2007. Therefore, the continued exclusion from the euro and Schengen are seen as a 'symbol of discrimination' (Badulescu 2021;Markova 2020).

Size, wealth, identity
This section examines the relationship between the structural factors of member state size, wealth, and identity and member state assessments of the two DI mechanisms. To begin with, we expected that member states with higher national identity (greater size, greater wealth) are more likely to view 'optouts' positively (H11).

Figure 6 Positions on 'Opt-outs' by member state size, wealth, identity
Regarding size, we find that the three largest economies are somewhat critical of the 'opt-out' mechanism -against our expectation. In Germany, the negative assessment of 'opt-outs' is rooted in a hard stance towards British cherry-picking before and during the Brexit negotiations (Nagel 2020). Similarly, in France, the negative assessment of 'opt-outs' were associated closely with the United Kingdom to the extent that "the key phrase 'opt-out' was literally transcribed in the records on several occasions to insist on its association with the UK (English words are otherwise usually translated in the  (Michel 2020: 6). Here, too, the association of 'opt-outs' and the UK goes back to the time before the Brexit referendum. Explaining the difference between the Lisbon Treaty and the failed Constitutional Treaty, former French president Valéry Giscard D'Estaing explained that "[i]t is noteworthy that 50% of the proposed changes correspond to concessions made to the British, and that requests for exemptions and opt-outs are dramatically on the rise, chiefly coming from the UK, Poland and to a lesser degree Ireland" (Michel 2020: 12). Concerning France itself, then health minister Xavier Bertrand stated in June 2008 that "regarding opt-outs, it is evident that France will not seek them" (cited in Michel 2020: 12). This chimes with the Italian experience, where the negative assessment of 'opt-outs' is based on the Italian ambition to remain in the core of the EU. Therefore, "there has never really been discussion in Italy about not subscribing to or unsubscribing from any of the agreements" (Volpi 2021: 15).
In terms of wealth, we find some indications that member states with higher GDP/capita levels view opt-outs more positively, but only if excluding the extremely wealthy Luxembourg. Also, the results appear more mixed when zooming in on the extremes of wealth. Looking at the bottom of the distribution, we find that Romania and Bulgaria hold negative views about opt-outs. Poland and Hungary, only slightly wealthier, view 'opt-outs' much more positively. The difference between these countries appears to be integration experience. Romania and Bulgaria are still subject to discriminatory opt-outs from the eurozone and the Schengen Area. By contrast, Hungary and Poland have acceded to the Schengen Area and pursue an informal policy of voluntarily opting out from the euro as this is seen as serving the national interest.
At the other extreme of wealth, we find Luxembourg and Ireland. Luxembourg has a negative view of 'opt-outs', which are often seen as associated with the United Kingdom. In this regard, Luxembourg pursued an active policy to end the British opt-out of the working time directive (Michel 2021). The situation is more complicated in Ireland, which has opt-ins for Schengen and the AFSJ. In both cases, the opt-ins are justified as a pragmatic measure to ensure close alignment with the UK, which is seen as extremely important in economic and legal terms. In April 2008, Senator Déirdre de Búrca, (Green Party) stated in the Joint Committee on EU Affairs: "that it is anomalous that Ireland is not part of the Schengen arrangement. This is because of the common travel area we have with the UK and it poses difficulties for Ireland. For the same reason, under the Lisbon treaty, we sought an opt-out in the area of justice and home affairs because we are tied into the UK in a way that makes it difficult for us to take a stance on our own without considering that country's position." (cited in Telle 2020: 14) This view is echoed by the former Minister of Justice, Equality and Law Reform, Brian Lenihan (Fianna Fail, 03/04/2008), who argued that "it would have been a big ask for Ireland to participate in qualified majority voting on legal questions with the United Kingdom outside the room and Ireland left in the invidious position of being the only effective common law jurisdiction left, with the other 26 Member States" (cited in Telle 2020: 13) For identity, there is some evidence that member states with a higher share of citizens who identify themselves as exclusively national perceive 'opt-outs' more positively. At the same time, Bulgaria and Romania are countries with high percentages of citizens identifying themselves as only national, but with opposing views about the 'opt-out' mechanism.
Lastly, we contrast most-likely and least-likely countries concerning positive positions on the 'optout' mechanism. The most-likely cases combine a big size with high wealth and an above-average share of citizens identifying as 'exclusively national'. France (negative) fits the description best. The leastlikely cases are small, less wealthy, and less exclusively national. Croatia (negative), Slovakia (negative), and Hungary (positive) fit this description best. The results are mixed. The negative French position on the 'opt-out' mechanism can be attributed to the ambitious and exclusivist French vision of the EU, which envisages deeper (but not necessarily uniform) integration. While Croatia and Slovakia fit our expectations, Hungary does not. We see an explanation for Hungary's position in its informal voluntary opt-out from the monetary union.
Next, we turn from the 'opt-out' mechanism to the mechanisms of 'enhanced cooperation'. We expect that member states with lower national identity (smaller size, smaller wealth) are more likely to positively view 'enhanced cooperation' (H12).

Figure 7 Positions on 'enhanced cooperation' by member state size, wealth, identity
Beginning with size, we note that the five largest member states -Germany, France, Italy, Spain, and the Netherlands -view 'enhanced cooperation' positively. The German position on enhanced cooperation appears to be determined by two opposing considerations. On the one hand, 'enhanced cooperation' is seen as a way to move integration forward. On the other hand, 'enhanced cooperation' risks undermining the unity of the EU. This tension is well captured by a statement of Chancellor Angela Merkel (CDU, 24/04/2008) made in the German Parliament about the wider applicability of 'enhanced cooperation' in the Lisbon Treaty: "The [Lisbon] Treaty also offers the possibility to act in concert, given that an agreement with everyone cannot be found. For this reason, the enhanced cooperation instrument was created. All the Member States have to agree to enable the use of this instrument by a group of Member States. However, I want to make it clear that it must not become the norm to use enhanced cooperation concerning all important issues. On the contrary, we must put our efforts into trying to establish common positions." (cited in Nagel 2020: 10) The eurozone crisis put a test to this position, as functional and political pressures towards deeper integration inside the eurozone threatened to increase the distance between a deepening European core and those member states which had not (yet) adopted the euro. Chancellor Angela Merkel (CDU, 18/10/2012) underlined that she did not want a "two-class society" but added that "we cannot say that we will not solve these problems because not everyone has joined in yet" (cited in Nagel 2020: 14).
The French preference for a more ambitious and more exclusive version of integration is reflected in its stance on 'enhanced cooperation'. Similar to Germany, 'enhanced cooperation' is perceived as a way to deepen integration. In particular, 'enhanced integration' was seen as a way to deepen integration in defence policy, foreign policy, and economic policy (Michel 2020). However, in contrast to Germany, inclusiveness was seen as less of a constraint on the French side. Especially under President Macron (since 2017), France pursued a more ambitious (and exclusive) political vision for the EU, while Germany maintained a more inclusive (and less ambitious) idea of 'regulatory convergence'. In Spain, 'enhanced cooperation' is strongly associated with the Spanish ambition to achieve deeper integration. This is clearly expressed by Prime Minister Rajoy in Parliament when discussing the Commission's five scenarios for the future of Europe in 2017: "There will be people who may like solution one, two or four best. The ones I propose […] are five and three 28 ." Rajoy also stated that he prefers to call scenario three "enhanced co-operation" (cited in Leon 2021: 17) As such, Spain supports most initiatives of 'enhanced cooperation', including the Financial Transaction Tax. One important exception is the Spanish position on the European Unitary Patent, which Spain rejects due to perceived linguistic discrimination.
Turning from big to small member states, the Irish position on 'enhanced cooperation' stands out as particularly negative. This has several reasons. First, the demand for more integration in Ireland is limited, especially concerning taxation, defence, and law enforcement. Against this background, the mainstreaming of the 'enhanced cooperation' mechanisms in the Nice Treaty has been viewed very negatively in Ireland, and the first Irish referendum on the Nice Treaty has failed. As such, 'enhanced cooperation' was seen as a way for other member states to circumvent Irish vetoes in policy areas in which Ireland did not seek deeper integration. Second, in the runup to the Irish Lisbon referendum, 'enhanced cooperation' was seen as a mechanism by which the other member states could potentially proceed with further integration without Ireland. The then Minister of State for European Affairs, Dick Roche (Fianna Fail, 17/04/2008), summarized the concerns as follows: "The final question was whether a group of countries could go ahead. There are enhanced cooperation provisions in the existing treaties, so there is always an opening for enhanced co-operation.
[…]. Almost ten years have been invested in this project [of the Lisbon Treaty].
[M]y biggest concern about a 'No' vote is […] that countries in Europe will become disenchanted with the project and that a two-speed Europe will emerge." (cited in Telle 2020: 14) This negative assessment of the 'enhanced cooperation' mechanism is also reflected in the fact that Ireland is critical of (and does not participate in) most initiatives of 'enhanced cooperation' (i.e. Rome III, Matrimonial Property Rights, Financial Transaction Tax). Paul Murphy (Solidarity, 05/11/2019), member of the Select Committee on Finance, Public Expenditure and Reform, detected a general pattern in the Irish government's attitude towards 'enhanced cooperation': "[The government] says that it is not necessarily against the idea but that everyone would have to do it at the same time. This is just an excuse to lag behind developments and to try to maintain our tax haven status. The government will always find an excuse. When enhanced co-operation is taking place, for example, the government is very pointedly not involved even though it is something that is happening on a cross-Member State basis. The government then shifts the goalposts and says that it must happen all across the EU but it is not really interested in that happening either" (cited in Telle 2020: 21) Nevertheless, there is some evidence that the Irish position has evolved somewhat after the UK membership referendum. Discussing the five scenarios for the future of Europe in 2017, Irish Prime Minister Enda Kenny (Fine Gael, 08/03/2017) affirmed the Irish ambition to remain at the centre of the integration process: "Enhanced co-operation has been included in the treaties for this reason [to allow for a multi-speed Europe] and the provisions can be used to allow a group of states to move ahead if they wish with co-operation without all members supporting it. We want to stay in the vanguard, right up at the front. If it transpires that there are two speeds operating in Europe, we want to be at the higher speed and central to it" (cited in Telle 2020: 17) In sum, smaller member states do not appear to be more positively inclined towards 'enhanced cooperation'. The five largest member states view it as a means to achieve more integration. Concerning Ireland, size seemed to matter, but not in the expected way. Instead, we find that limited overall demand 28 The five scenarios are: Scenarios One 'Carrying On'; Scenario Two 'Nothing but the Single Market; Scenario Three 'Those who want more do more'; Scenario Four 'Doing less more efficiently'; Scenario Five 'Doing much more together'. for more integration resulted in a negative assessment of 'enhanced cooperation' in Ireland. After Brexit, a more positive discourse evolved, reflecting changing structural conditions, such as the perceived need for a closer alignment with the EU27.
Regarding the effect of wealth on member states' positions on 'enhanced cooperation', we find some positive evidenceespecially if excluding Luxembourg. Two countries at the opposite end of the wealth spectrum are very much in line with our expectations: Ireland and Romania. For Ireland, the then Minister for Finance, Michael Noonan (Fine Gael, 14/11/2012) provides pragmatic economic reasons to justify the Irish non-participation in 'enhanced cooperation' on the Financial Transaction Tax: "[W]e are not participating in the enhanced co-operation mechanism in which other countries are participating because we do not want to create a situation where the 33,000 jobs in the financial services sector in Ireland would be put at risk of transfer to a more benign tax regime in London. That is the only issue involved. We do not have a problem in principle with a financial transaction tax" (cited in Telle 2020: 16) Romania's assessment of 'enhanced cooperation' is very positive, as it is perceived as a mechanism to increase Romania's influence at the European level. Hence, great emphasis is put on the inclusiveness of 'enhanced cooperation', as President Klaus Iohannis (PNL, 25/03/2017) made clear in a press conference at the informal 2017 Rome summit: "I would like to emphasise that the mention in the final text of the Rome Declaration of the concept of enhanced co-operation is made strictly within the limits of the provisions of the current Treaty on European Union, with the aim of all states acting together. In addition, the text of the Declaration states that the door remains open for Member States wishing to join these forms of enhanced cooperation later." (cited in Badulescu 2021: 20) In sum, we find the most evidence for our wealth-related expectations for Ireland and Romania. However, the Irish example also shows that greater wealth does not necessarily lead to demand for stricter regulatory standards -especially when positive externalities are high. The Romanian case indicates that poorer member states view 'enhanced cooperation' as a way to increase their European influence. Overall, however, we find few differences between the positions of wealthy and poor member states. Stances on 'enhanced cooperation' seem to be determined by the demand for integration. But demand for integration does not appear to be determined by member state wealth.
In terms of the effect of national identity, we find little evidence that member states with a smaller share of citizens identifying as exclusively national have more favourable positions on 'enhanced cooperation'. As such, there is a large cluster of countries with positive assessments of 'enhanced cooperation' and above-average shares of nationally minded citizens. Romania is an excellent example of this cluster. At the same time, two countries closely conform to our expectations: Spain and Ireland. In both countries, the share of citizens identifying as exclusively national seems to correlate with demand for more integration. In Spain, the more widespread demand is reflected in a positive assessment of 'enhanced cooperation'. In Ireland, limited demand for more integration is registered in a negative assessment of 'enhanced cooperation'.
Finally, we contrast most-likely and least-likely countries concerning positive positions on the 'enhanced cooperation' mechanism. The most-likely cases combine small size with low wealth and a below-average share of citizens identifying as 'exclusively national'. Croatia (positive), Slovakia (neutral), and Hungary (positive) fit the description best. The least-likely cases are big, wealthy, and more exclusively national. France (positive) and Italy (positive) fit this description best. The results are mixed, as all of these states view enhanced cooperation rather positively. This supports the view that there is widespread agreement that DI is permissible when it leads to more integration. Robert Schuman Centre for Advanced Studies Working Papers

General
We now turn to two more general expectations about how DI positions change over time and about the role of party ideology.

Change over time
First, we expected that assessments of DI models and mechanisms would improve throughout the period under investigation (2008-2020) as the poly-crisis had rendered the need for more flexibility in EU crisis responses more immediate and obvious (H13). The present section first looks at DI models, finding slight improvements for both. Subsequently, it focuses at DI mechanisms, finding growing acceptance for 'enhanced cooperation', but not for 'opt-outs'. Figure 8 shows how member states' positions about the DI models 'multi-end Europe' 29 and 'multispeed Europe' 30 have changed during the time interval. Panel A shows that six of 22 analysed member states improved their assessment of the DI model 'multi-end Europe', while the remaining states did not change their position. Panel B shows that evaluations of the 'multi-speed model' improved in ten member states, remained unchanged in thirteen member states, and deteriorated slightly in three member states. Overall, we find limited support for the expectation that the acceptance of DI is improving over time, albeit with many countries with no change.
Looking at the 'multi-end Europe' model first, the positions of structurally very different countries seem to have improved. On the one hand, we find a group of smaller and less wealthy countries, including Estonia, Latvia, and Greece. On the other hand, we find France and Italy, both larger and wealthier than the EU average. The qualitative data reveals that the improved assessments are all related to the ambition of entering or remaining inside the group of core countries. Contrary to what one might have expected, for none of these countries, 'multi-end Europe' means that they seek less integration for themselves. We do not find differences between wealthy and poor, small and large countries. However, France appears to be the only country that endorses the 'multi-end Europe' model to actively pursue deeper integration. The other five countries are either concerned with catching up with the core (Estonia, Latvia) or worried about falling outside (Ireland, Italy, Greece).
European University Institute 31

Panel A Panel B
Among this last group of countries, the 'multi-speed Europe' model appears to be associated with the danger of being relegated to second-tier membership. Hence, it does not seem accidental that the assessment of the 'multi-speed Europe' model has deteriorated over time in Greece and Italy. Among those countries which have a favourable position on the 'multi-speed Europe' model, we can single out Demark and Hungary from the rest. Here, assessments have improved due to the perception that a Europe of multiple speeds would allow a postponement of certain integration steps. In all other countries, the evaluation improved because a multi-speed Europe was seen as a way to move integration forward. In Ireland, this integrationist position was a relatively recent development and strongly affected by Brexit. Overall, assessments of DI models have improved over time. However, improvements were limited to a small and diverse group of countries.
Panel A in Figure 9 shows that the assessment of the DI mechanism 'opt-out' deteriorated slightly in four of the nineteen analysed countries 31 . At the same time, it improved in Denmark, albeit from a negative position. Panel B indicates that the assessment of the DI mechanism 'enhanced cooperation' improved in thirteen of twenty-five analysed member states 32 , while the remaining states did not change their evaluation. Robert Schuman Centre for Advanced Studies Working Papers

Panel A Panel B
Denmark is the only country improving its position on 'opt-outs'. This change is related to the 2019 General elections when the Social Democratic Party around Mette Frederiksen formed a left-leaning coalition. While previous Danish governments had repeatedly stated their intention of ending the Danish opt-out in the AFSJ, the new government did not pursue this aim any longer. This shift had occurred after a 2015 referendum in which the majority of voters confirmed the existing Danish opt-out. Several ministers and MPs of the Social Democratic party had argued against the Danish opt-out before the referendum. Hence, the more positive assessment of 'opt-outs' by the new Danish government seems to represent a re-alignment of the government with voter preferences. The deteriorating review of 'optouts' in France, Germany, and Spain are related to the Brexit experience. Their perception is that the British 'opt-outs' had been a mistake since they did not prevent Brexit but weakened the integrity of the EU. In Croatia, the assessment of 'opt-outs' has deteriorated due to an increasingly negative perception of the discriminatory exclusion from the eurozone andespeciallythe Schengen Area.
In contrast to the mixed-to-negative development in assessing the 'opt-out' mechanism, we see a clear improvement in the assessment of 'enhanced cooperation'. For the new (small, less wealthy, more national) member states, 'enhanced cooperation' is increasingly seen as an avenue towards becoming a core state in the EU. This perspective applies similarly to Italy, where enhanced cooperation is seen as a mechanism to maintain core state status. In Germany, Belgium, and the Netherlands, the assessment of 'enhanced cooperation' has improved as it became increasingly seen as a way to overcome deadlock. The Belgian Vice Prime Minister and Minister of Foreign Affairs, Didier Rynders (15/02/2017) stressed in the foreign relations committee that even after Brexit "there are still divergences within the 27" and therefore "what we cannot do immediately in 27, we can do in a narrower circle, for instance in the 33 The labels of Figure 9 should read 'deteriorated strongly ', 'deteriorated', 'unchanged', 'improved', 'improved strongly' as in Figure 8. This will be updated in a future version of the report. eurozone basing ourselves on the tools that the Treaties recognise, such as enhanced cooperation. We will indeed continue to work along this path […]." (cited in Cescon 2020: 13)

Party ideology
Finally, given that European main-stream parties tend to be pro-European, we expected to find that party ideology is not central to member states' assessments of DI (H14). To see whether party ideology impacts DI positions, we have investigated whether a change in a state's position on DI corresponded to a change in the composition of the government. The results are presented in Figure 10 34 .

Figure 10
Effect of party ideology on member states' DI positions Figure 10 suggests that a change in government did not affect the overall assessment of DI in most member states. This means that party ideology does not play a central role in member states' positions on DI. However, there are a few exceptions that deserve a closer look to better understand the role of party ideology in shaping how Croatia, Czechia, Cyprus, Greece and Poland view DI.
In the case of Croatia, the Social Democratic Party (SDP)-led government coalition fully supported Croatia's accession to the Schengen area in 2012, hoping to join shortly after the country acceded to the EU in 2013. However, over the 2017-2020 period, the main political parties from Croatia have changed their position regarding the country's accession to Schengen. Thus, along with the shift of power from SDP to the Croatian Christian Democratic Union (CDU)-led governing coalition, Croatia's accession to Schengen was seen as complementary to the country's national interest, most probably because of the security issues raised by the migration crisis from 2015. Nevertheless, the CDU PM Andrej Plenković correlated Schengen with the need to avoid multi-speed DI in the future as that would be detrimental to Croatia's position in the EU: 34 No data for Belgium, Italy. Robert Schuman Centre for Advanced Studies Working Papers "The eurozone and Schengen area are the best known, but there are many other aspects of enhanced cooperation based on the Treaty of EU, where part of Member States decided to delve deeper into integration. Because of that, the thesis I've proposed in Croatia's name was, considering that multispeed Europe is now a fact that we all have to acknowledge, that the manner of cooperation in the future has to be inclusive. In other words, we have to avoid an amalgam in which a few powerful, wealthy and typically founding states will pursue closer cooperation, while the rest will remain at the margins with a decreased influence on the formation of the European project." (cited in Beroš and Gnip 2020: 12) In Czechia, a change of power from soft Eurosceptic Civic Democrat Prime Ministers to more pro-European Social Democrat PM Bohuslav Sobotka (2014-2017) triggered a slight shift in Czechia's position on DI. On the one hand, PM Nečas (2010-2013) expressed opposition towards the multi-speed model of DI. However, he argued in favour of more debate regarding the country's accession to the Eurozone, which he perceived as being the core of Europe, and expressed support for enhanced cooperation. On the other hand, PM Sobotka strongly opposed multi-speed Europe as something that would split Europe between the centre and the periphery (Hlavik and Smekal 2020).
In Denmark, as discussed above, the 2019 elections appear to have resulted in a re-alignment of the government's position with the more integration-sceptic public opinion in the country. As a result, the current Danish government under Prime Minister Mette Frederiksen does not seek to end the Danish opt-outs any longer, as previous governments had done (Madsen 2020).
In Greece, following the economic and financial crisis, the change in power from the New Democracy (ND) -Panhellenic Socialist Movement (PASOK) coalition government to the Coalition of the Radical Left (SYRIZA)-The Independent Greeks (ANEL) coalition government led to a slight change in Greece's position regarding the DI model that would best fit Greece's interests. Thus, before the crisis Greece supported the 'multi-end Europe' model of DI that would allow Greece to be at the core of Europe. After the crisis, however, the SYRIZA-led government adopted a position that was more in favour of a united Europe rather than a divided one, fearing that Greece would be left out of the core of Europe. Advocating for more European integration and unity, the ND Prime Minister Mistotakis also rejected the multi-speed model in 2019 during his first speech as PM: "We do not believe in a multi-speed Europe. This only leads to the rise of populism and nationalism. We believe in a Europe that goes together, that applies common rules of law and of course a Europe that gives prospects for its enlargement, especially in our neighbourhood, in the Western Balkans." (cited in Tellidou 2021: 17) The position of Polish governments on DI has been shaped, on the one hand, by the country's perceived 'national interest' when engaging on enhanced cooperation in the EU, and, on the other hand, by the position that Poland would have in the EU following Poland's participation or non-participation in certain instances of DI. While the Law and Justice (PiS) government emphasised the importance of the 'national interest' as a main factor that influences the country's position regarding the 'enhanced cooperation' mechanism of DI, other governments expressed more concern towards the possibility of Poland's retreat outside the core of the EU if the country decided not to join the Eurozone (Walecka and Gagatek 2021).
To summarise, we can see that the ideological differences have had relatively slight influences over the positions of these five countries, and most often, they have revolved around the stance that the country should have in the context of the emergence of a multi-speed or a multi-end Europe.

Summary and discussion
In this section, we are going to take into review the main findings in light of the theoretical expectations that guided this paper. We will also discuss how our findings may advance differentiation theory.
Hypothesis 4 stated that member states with a higher European identity (or greater size, or greater wealth) should view multi-speed DI more positively than member states with a more national identity, smaller size, and lower wealth. We found some indications that smaller countries view multi-speed DI more positively than bigger states. However, smaller states are not a coherent group in this matter. Instead, they are split between positive and negative assessments. We find a similar pattern concerning wealth, where richer member states view 'multi-speed Europe' more positively than poorer member states. At the extremes, Luxembourg, as well as Romania and Bulgaria, fit this pattern. About the latter, 'multi-speed Europe' is strongly rejected, as Romania and Bulgaria perceive it as a potential source of discrimination and division in the EU. Latvia and Ireland contradict our expectations most clearly. In Latvia, the 'multi-speed model' has come to be viewed positively since the country has adopted the common currency. In Ireland, 'multi-speed Europe' is rejected for the same reason as 'multi-end Europe' the fear of losing international leverage. Identity also seems to be associated with positions on multispeed DI. 'Multi-speed Europe' is viewed positively in Luxembourg, the member state with the lowest share of citizens identifying as 'exclusively national'. At the same time, it is viewed very negatively in Bulgaria and Romania, two member states with high shares of citizens identifying as exclusively Robert Schuman Centre for Advanced Studies Working Papers national. However, Spain, Finland, and Latvia do not fit our expectations. In Spain, a country with a low share of exclusively national citizens, 'multi-speed Europe' is seen negatively because it is associated with a drift toward the European periphery. Finland and Latvia have high shares of citizens identifying as exclusively national, but also assess 'multi-speed Europe' positivelypresumably because national identity here does not contradict European identity but is directed toward the historical experience with their Russian neighbour.
Overall, our results regarding the effect of size, wealth, and identity on the position of member states on DI models are mixed. The qualitative analysis revealed that member states' positions are not predominantly determined by their structural characteristics (size, wealth, identity), but also by differing ideas about the inclusiveness of the EU and the member state's desired positionality within the EU. Regarding inclusiveness, France and Germany exemplify two different visions about the EU. France pursues an ambitious agenda of deepening the European Union. Differentiation is seen as the price which needs to be paid to achieve this objective.
The 'multi-end Europe' model is seen as suitable for realizing this 'exclusivist vision' of Europe. This goes against the expectation that integration-seeking member states should favour the 'multi-speed model' because it involves lower risks and transaction costs in the long term. The French case illustrated, however, that integration-seeking member states may be willing to accept more fragmentation as a price to pay for further integration. Germany pursues a more inclusive agenda, which seeks to keep member states with more limited capacity engaged in the integration process. As a result, Germany favours the 'multi-speed Europe' model.
Regarding the process, 'multi-speed Europe' can be thought of as a two-way street. Countries that worry about their influence in the EU appear to view the 'multi-speed Europe' model more negatively (i.e. Spain, Italy, Bulgaria, Ireland, Sweden). By contrast, countries that currently perceive themselves as part of the 'core' rather than of the 'periphery' of Europe view multi-speed Europe more positively (i.e. Latvia, Croatia) or neutral (Slovakia, Estonia).

Policy differentiation
The analysis then turned from the polity to the policy level. The hypotheses testing states' positions on specific instances of DI did not receive strong support. For hypothesis 5, we found little evidence for the realist expectation that smaller member states had more positive positions on PESCO. Indeed, the small and mid-sized Central and Eastern European member states did not have more positive positions on PESCO than big Western member states. While the majority of countries from the 'least-likely' group of small and neutral states (Austria, Finland, Ireland, Malta, Sweden) joined PESCO, this was arguably achieved by successfully watering down the more ambitious French vision of PESCOnot because of active demand for a stronger European defence policy. We found some evidence that wealthier states were concerned about the cost and effectiveness of PESCO. In contrast, less wealthy member states viewed PESCO as a potential means to improve their military capabilities.
For hypothesis 6, our findings point to the fact that there is widespread support for stricter fiscal rules, even among less wealthy members of the eurozone. In this case, the dividing line seems to be drawn not between rich and poor states, but between members of the eurozone, states that want to join and states that are reluctant to join the euro. Moreover, it seems that the debates have focused more on the issues of responsibility and solidarity rather than on stricter rules for the monetary union.
Finally, we did not find strong evidence for hypothesis 7: it seems that the group of states that did not join the EPPO is not characterised by an above-average share of nationally minded citizens. Nevertheless, sovereignty concerns were important in member states rejecting the EPPO (Hungary). Another reason was the concern that the EPPO would be an ineffective institution (Sweden).
Overall, we do not find strong evidence for the association between member states' size and their stance on security integration, member states' wealth and demand for stricter regulatory standards.
Similarly, we found no evidence for the associated between member states' identity and demand for more sovereignty. Nevertheless, the demand for policy differentiation is rooted in diverse member states' capacities and preferences. What our analysis does seem to suggest is that the demand for policy differentiation is reduced by the 'shadow of polity differentiation'. The cases of PESCO and the Fiscal Compact show that even reluctant states (or those which are not strictly obliged to participate) prefer to be on board when there is a credible threat of being left behind.

DI mechanisms
We conceptualized 'enhanced cooperation' and 'opt-outs' as mechanisms through which DI can be realized. Hypothesis 8 stated that MS would have more positive views about the former because it allows for the realization of mutual gains through integration. Indeed, we found that a majority of the member states have a positive or very positive position on enhanced cooperation, but a much less favourable assessment of 'opt-outs'. The difference in the evaluation between the two DI mechanisms points to the interpretation that member states view DI primarily as a means to advance integration, not to obtain exceptions from EU rules.
With hypothesis 9, we turned to the question of why member states assess DI mechanisms more or less positively. Hypothesis 9 was concerned with the expected effect of 'enhanced cooperation' and 'opt-outs' on polity effectiveness and legitimacy. We found that expectations matter: states that perceive the impact of opt-outs and enhanced cooperation positively also assess these DI mechanisms positively. We found that the expected impact on effectiveness and legitimacy matter at the EU and member state level.
Hypothesis 10 stated that states currently subject to discriminatory opt-outs should view 'opt-outs' more negatively than other member states. It must be highlighted that here we have two categories of states. On the one hand, we have the states that are somewhat positive regarding these 'discriminatory opt-outs' as it currently suits them not to opt-in the EMU, although they still have a treaty obligation in this regard (Sweden, Poland, Czechia, Hungary). On the other hand, we have member states seeking to overcome their discriminatory opt-outs (Bulgaria, Croatia, Cyprus, Romania), which view opt-outs generally negatively and perceive their exclusion from Schengen or the EMU as a symbol of discrimination. Therefore, a conceptual distinction between discriminatory opt-outs and informal voluntary opt-outs seems appropriate.
Lastly, we turned to expectations about how the structural factors of size, wealth, and identity impact member states' positions about DI mechanisms. Hypothesis 11 stated that member states with a higher national identity (or greater size, or greater wealth) are more likely to view 'opt-outs' positively. In terms of size, we found that the three largest economies of the EU (i.e. France, Germany and Italy) are rather critical of the opt-out mechanism, showcasing a strong stance against British cherry-picking and an attachment to the optimal position as part of the core of the EU. As to wealth, on average opt-outs are opposed by poorer and richer states alike. On the one hand, poorer states oppose opt-outs as discriminatory, while richer member states are generally against 'preferential treatments' that would allow the UK to cherry-pick on certain policies. Regarding identity, we find some evidence that MS with a higher share of citizens identifying as exclusively national view 'opt-outs' more positively. Luxembourg, Spain, and Germany, and the Netherlands have the lowest share of nationally-identifying citizens and view 'opt-outs' negatively to very negatively. At the same time, some countries with a higher national identity (i.e. Bulgaria and Romania) are also strongly against the opt-out mechanism. For their turn, those with lower levels of nationally-identifying citizens (Poland and Hungary) hold positive views on 'opt-outs'.
Hypothesis 12 stated that member states with lower national identity (or smaller size, or smaller wealth) are more likely to view 'enhanced cooperation' positively. Our findings do not indicate a significant role of size, wealth or identity in influencing states' positions concerning enhanced cooperation. Smaller states do not appear to be more favourable towards enhanced cooperation. Indeed, Robert Schuman Centre for Advanced Studies Working Papers the five largest member states support this DI mechanism as a means to achieve more integration. When looking at states' wealth and identity, we find very few differences between the positions of wealthy or poorer member states or between those of states with a stronger national identity or European identity. Thus, positions on enhanced cooperation seem to be determined by the demand for more integration, but the latter does not seem to be closely associated with states' size, wealth or identity.

General expectations
Finally, when testing our general hypotheses, one was partially invalidated (H13) while the other was validated (H14). Hypothesis 13 stated that assessments of DI models and DI mechanisms improve over time. While many states did not change their position regarding DI models over time, we found limited support for the expectation that states' positions have improved over time. The few improvements found can be explained by the states' ambition to enter the core EU or their fear of remaining outside the core EU countries. As for DI mechanisms, it seems that the opt-out mechanism has been assessed increasingly negatively over time, while the enhanced cooperation mechanism saw a clear improvement in the eyes of the member states. This can be explained by the fact that enhanced cooperation has been increasingly seen as a way to become part of the core EU and as a way to overcome deadlocks of integration.
Finally, hypothesis 14 held that party ideology is not central to member states' assessments of DI. According to our findings, out of a total of 24 member states analysed for this hypothesis, in 19 cases, party ideology did not play a central role in member states' positions on DI. In the remaining 5 cases, party ideology has had rather minor influences over member states' position on DI. Most often they have revolved around the place that the country should have in the context of the emergence of a multispeed or a multi-end Europe.

Alternative explanations
Our analysis results suggest that member states' size, wealth, and identity are not necessarily dominant in how member states assess differentiated integration. The analysis has shown that expectations about the impact of DI on the EU and on the MS may be a better predictor. However, if these expectations are not determined by structural factors, such as size, wealth, and identity, what does explain them? Here we look at two alternative explanations.

Balancing policy preferences and polity preferences
As an alternative to the structural factors suggested by the literature on DI, we have also looked at how 'expected effects' of differentiation on polity efficiency and legitimacy affect how member states assess DI. We found some evidence that expectations about the expected impact of DI matter for how DI mechanisms are evaluated. While it is not surprising that the expected effects of a policy instrument influence how the instrument itself is perceived, the analysis points to a more substantial point: Assessments of DI are not merely a function of the structural characteristics of the member states. Instead, ideological and strategic considerations about the long-term effects of policy preferences on the nature and functioning of the EU polity appear to matter in how member states assess DI. In this view, policy differentiation takes place in the 'shadow of polity differentiation'.
The 'shadow of polity differentiation' can be understood as the anticipated long-term effects of policy differentiation on the nature and functioning of the European Union. Our analysis has shown that EU member states view both models of polity differentiation -'multi-end Europe' and 'multi-speed Europe'negatively on average. At the same time, individual member states have assessed either one or the other of the two models positively. Hence, the 'shadow of policy differentiation' may have a negative as well as a positive effect on how member states think about policy DI and DI mechanisms. Our analysis suggests that two factors determine such assessments. 39 The first factor concerns the visions member states have of the future nature of the EU 35 . To put it simply, the choice is between a 'deep Union' and an 'inclusive Union'. To states which seek a 'deep Union', DI can appear as a pragmatic way to overcome deadlock. Especially under President Emmanuel Macron, France has moved closer to the 'exclusivist' vision of the EU. Here, the 'multi-end Europe' model is seen positively because it represents a way to realize deeper integration among a more exclusive club of core member states. By contrast, states that prioritize inclusiveness tend to view DI as a potential long-term risk to the integrity of the EU. Germany comes close to this position. Accordingly, German governments see DI unfavourably; it is acceptable only if it is limited in time, as in the 'multi-speed Europe' model.
The second factor concerns expectations about the positionality a member state is likely to assume in the EU. The central choice here is between the preservation of national independence and stronger influence at the European level. When a member state values sovereignty, DI is likely to be seen positively as it allows a member state to opt-out of further integration. However, the long-term consequences of this policy may be a diminishing influence at the European level. For example, the analysis has shown that Irish politicians worry about the effect of the Irish opt-outs on the island's leverage at the European level. Member states which are concerned about their influence at the European level view DI more critically. The analysis revealed that such concerns are especially relevant in small countries (i.e. Ireland, Slovakia, Latvia), new member states (i.e. Romania, Bulgaria, Croatia), and declining founding states (i.e. Italy). In sum, the fear of being left behind or the ambition to be part of the core is thus one of the main reasons for why states oppose DI on a conceptual level (multi-speed and multi-end EU).

Low salience & strategic ambiguity
Despite these observations, the 'shadow of polity differentiation' is not a hard constraint. To begin with, due to the short-term time horizon of electoral politics, governments may not sufficiently take the longterm consequences of differentiation into account. We expect that the 'shadow of polity differentiation' matters most in older member states with mainstream parties which are deeply invested in the process of European integration. In addition, it should matter where challenger parties with a strong (positive or negative) European platform gain political influence (i.e. Macron's 'En Marche!' in France).
At the same time, DI is not a high salience issue for the member states 36 . Hence, political parties very rarely run on DI, and there is little popular pressure on governments to formulate a clear position on DI. Instead, many member states tend to have rather ambiguous positions on DI. This enables member states to pursue DI at the policy level while also engaging in a discourse which rejects polity differentiation and endorses European unity. Indeed, a country with a strong positive stance on DI (i.e. France or the UK) might face allegations that it is jeopardizing the unity of the EU. Simultaneously, a strongly negative position on DI might be interpreted as an attempt to block further integration. In this context, the best strategy for many members state may be strategic ambiguity regarding polity DI, while engaging in policy DI on a case-by-case basis.
Finally, we see some evidence for the strategic use of differentiation. We have repeatedly touched on the notion that DI leads to "second class membership". Interestingly, we find this discourse, especially (but not only) in countries which voluntarily engage in differentiation. Examples are those countries informally 'opting out' of the common currency, particularly Hungary. Arguably, the 35 Member states may value different visions of the EU because of the intrinsic characteristics of a vision (i.e. democratic legitimacy, regulatory efficiency) and/or due to utilitarian assessments about how they may serve the national interest. Questions about the overall design of the EU should matter to all member states, but especially to those with the greatest capacity to actively shape the overall contours of the EU, i.e. the bigger member states. 36 For country-specific levles of salience, consult the InDivEU member state reports: http://indiveu.eui.eu/integrateddatabase/ Robert Schuman Centre for Advanced Studies Working Papers discourse of "second class membership" is being used to gain political leverage by blaming the EU for the negative consequences of a voluntary national practice.

Conclusions
This report has set out to identify how EU member states position themselves concerning differentiated integration in the EU and which factors influence their positions. Thus, unlike previous studies, this paper analysed the integration preferences of member states and not the actual outcomes of differentiation in terms of EU rules and policies. Our ambition was to chart out the contours of a novel research agenda into member state preferences about differentiated integration. For this purpose, we have developed an integrated conceptualisation of DI which unpacks the concept of DI twofold: on the one hand, we distinguish between polity DI and policy DI, and, on the other hand, we separate DI mechanisms between opt-outs and enhanced cooperation. We formulated and tested a series of theoryderived hypotheses, drawing on qualitative and quantitative data collected from all 27 member states via document analysis.
Our findings suggest that the positions which member states take on differentiated integration are not strongly correlated with the structural characteristics of the member states. However, expectations about the impact of DI on polity legitimacy and effectiveness appear to be a good indicator for how member states assess DI. Overall, member states appear to have a preference for moving forward together (i.e. uniform integration), whenever possible, rather than advancing in multi-speeds that would lead to an ever-complex multi-end EU. When uniform integration is not possible, states support DI as a tool that can help overcome deadlocks of integration and allow member states to join a certain policy when they have the necessary capacity and domestic political support for it. Differentiation, thus, becomes acceptable if it does not negatively impact the effectiveness and legitimacy of the EU or the member states. In short, DI remains a tool of last resort when uniform integration is not viable, but some member states have an intense interest in more integration.
In addition to these theory-derived factors, out our rich qualitative data allowed us to inductively identify several additional potential determinants of how member states think about DI. First, we found that the fear of being left in Europe's periphery is a strong driver of support for more integration, be it among old or new, rich or poor, big or small member states. This agrees with our finding that member states view the DI mechanism of 'enhanced cooperation' more favourably than the 'opt-out' mechanism. Second, this fear seems to originate either in past integration experiences (i.e. Romania and Bulgaria) or in limited domestic demand for more integration (i.e. Ireland). Third, different visions of the EU polity matter for how member states assess DI. In simplified terms, we can distinguish between two visions: an ambitious core EU and a less ambitious but more inclusive EU. Fourth, the foregoing points suggest that the long-term effects of differentiation on the EU are a soft constraint on policy differentiation. We call this the 'shadow of polity differentiation'. Finally, the fact that many member states have rather neutral positions on DI might be related to the low salience of the issue for national party competition, which allows member states to maintain a position of 'strategic ambiguity'.
Our study certainly has limitations. To begin with, we have 'cast the net' of our analysis extremely widely. As such, we have formulated an unusually long list of hypotheses. Our ambition was not to test each hypothesis in-depth but rather to chart some of the terrain which future research may fruitfully explore. Another limitation concerned the nature of our survey data, which did not allow us to explore in detail how party ideology mattered or how positions have changed over time. Nevertheless, we hope that the analysis was able to show that there is merit in distinguishing between polity differentiation, policy differentiation, and mechanisms of differentiation.
Our findings point to several avenues for future research. First, our conceptualization could be fruitfully employed by both qualitative and quantitative studies to tease the intricacies of differentiated integration. Second, while our study focused on structural factors and the expected impact of DI, other factors should be explored. To mention only a few: centralist vs. federalist traditions of public administration, domestic party competition, duration of EU membership. Third, the limitations of our data render our findings preliminary. Hence, novel data could shed additional light on the factors that determine differentiated integration evaluations by the member states, especially at the policy level. Fourth, linking our findings back to the literature on the actually existing degree of differentiation in the EU, future research may explore whether there is a gap between the 'expressed preferences' of EU member states and what they achieve in terms of outcomes. Finally, while we have explored how expectations about the impact of DI shape DI preferences, this could be done more systematicallyby spelling out how different DI models might impact EU decision-making. Robert