What Makes Economic Differentiation Effective? Insights from the EU Energy Sector, Banking Union and Third-Country Access to the Single Market

ABSTRACT Few studies so far have analysed the effectiveness of differentiation in EU policies. This is surprising given the importance and permanence of many differentiated arrangements, for example in EU economic policy. Insights from three studies on differentiation in the energy sector, the financial sector and third-country access to Single Market highlight the importance of institutional factors. EU economic differentiated arrangements tend to be more effective when: (i) there is a good ‘fit’ between the institutional design and the policy objectives; (ii) there are mechanisms to adapt them over time; and (iii) there are institutional provisions to prevent or mitigate negative side effects for the Union as a whole.

The issue of differentiation has traditionally been analysed through the lens of the 'grand theories' of European integration (see Schimmelfennig and Winzen 2019). Most research has focused on explaining the conditions under which member states opt for or against differentiation, or analyse its consequences for the whole EU, that is, its possible centripetal or centrifugal effects (see Kölliker 2001;Winzen and Schimmelfennig 2016;Schimmelfennig 2018). In contrast, there has been little academic interest in analysing whether existing differentiated arrangements work well. This lack of attention to the effectiveness of differentiation is surprising. Despite much discussion about centripetal and centrifugal effects, empirical reality shows that differentiated solutions have a tendency to persist over time once they are created. Even in cases where differentiation is only conceived as a temporary step and intended to progressively cover all of the Union (such as the Economic and Monetary Union [EMU]), in practice this rarely happens. Path dependence plays a crucial role in this regard. As Frank Schimmelfennig (2020, 8) explains: "If prior integration has put states on two different paths, sunk costs and endogenous interdependence may propel states onto divergent integration trajectories and increase the costs of changing paths". As a result, differentiated governance has become an established feature in many EU areas.
Against this backdrop, it is particularly pertinent to explore under what conditions differentiated policies are effective. This confronts us, however, with various conceptual and methodological challenges. First, while the notion of 'effectiveness' is widely used in various disciplines, it does not have any clear and commonly agreed meaning. Second, differentiated arrangements are very heterogeneous. The 'classic' literature (for example, Schimmelfennig and Winzen 2018) defines them as formal arrangements established within a reduced number of member states (internal differentiation) or allowing EU rules to be legally valid in certain non-member states (external differentiation). A recent strand of literature, however, expands the notion of differentiation to include more 'informal' modalities of differentiated cooperation within member states and even sub-national authority networks insofar as the latter allow for a non-homogeneous participation of sub-national authorities in the venues where EU policies are designed and implemented (Tortola and Couperus 2020). This broad definition of differentiation, which we endorse in this article, offers a more comprehensive vision of differentiated dynamics but also makes it more difficult to identify common factors affecting the effectiveness of differentiated integration.
A way of reducing this heterogeneity is by narrowing the analysis to differentiated initiatives developed in the same EU policy area. This is the approach taken by this article. In particular, we aim to compare the effectiveness of various differentiated arrangements currently in place in EU economic policy and formulate some tentative institutional hypotheses to explain variation in their performance. To do so, we draw on the empirical findings of three studies conducted in the framework of the Horizon 2020 EU IDEA research project analysing the effectiveness of differentiation in the energy sector (Franza et al. 2021), the financial sector (Mack 2020) and the provision of third-country access to the Single Market (Eisl 2020). 1 The advantage of considering these three studies is that they adopt a rather similar approach to the study of 'effectiveness', partly based on the conceptual framework proposed by Sandra Lavenex and Ivo Križić (2019).
The article is structured as follows. The next section reviews prior studies linking differentiation and policy effectiveness, and identifies a number of shortcomings in conceptualising effectiveness. The following section explains why a narrow goal-attainment approach is unfit to analyse the performance of differentiated arrangements and proposes a broader approach, which takes into account unintended effects and the performance of a policy over time. The third section presents the three studies under analysis, assessing the functioning of differentiated arrangements in the energy sector (Franza et al. 2021), the financial sector (Mack 2020) and the provision of access to the Single Market to third countries (Eisl 2020). For each study, we present the differentiated organisations or arrangements under analysis and the authors' findings regarding the factors influencing their effectiveness. The next section discusses the findings of the three studies and explores under what institutional conditions EU economic differentiated integration is effective. We argue that arrangements are more effective when: (i) there is a good 'fit' between their institutional design and objectives; (ii) there are institutional mechanisms to adapt the arrangements' objectives and instruments over time; and (iii) there are institutional provisions to mitigate or prevent negative side effects for the EU as a whole. This is followed by a conclusion, summarising the article, discussing its limitations and providing suggestions for future research.

The missing link between differentiation and policy effectiveness
To our knowledge, there are only very few studies that explicitly engage with the relationship between differentiated integration and policy effectiveness, conducted by Cristian Frommelt (2017) and Asya Zhelyazkova (2014). Frommelt analyses the functioning and effectiveness of the European Economic Area (EEA), 2 defining effectiveness as rule homogeneity, which is achieved "by consistent selection, timely and complete incorporation and correct application of EEA relevant EU legislation by the EEA EFTA states" 3 (Frommelt 2017, 7). To assess the conditions of effective external differentiated integration, the study distinguishes between policy-related factors, country-related factors and institutional factors. The analysis mainly focuses on testing country-related explanatory factors, such as the administrative capacity of the EEA EFTA states to transpose EU laws and their economic interdependence with the EU, and policy-related explanatory factors, such as the salience attached to a specific EU act as well as the institutional compatibility of an EU act with the EEA's two-pillar structure. However, Frommelt also identifies some institutional factors affecting the EEA's effectiveness. In particular, he contends that the more formal access to EU policy-shaping EEA EFTA states have (something that may differ depending on the policy area) and the more resources, autonomy and expertise to monitor the enforcement of EU rules the EEA EFTA institutions have, the higher the EEA's effectiveness (136-7). Zhelyazkova (2014) investigates the extent to which differentiation affects the national conformity with EU legislation. She compares the level of national compliance with ten EU directives adopted in the period 2001-2004 by member states that participate fully, selectively or that do not participate at all in EU initiatives in the field of asylum and immigration. Zhelyazkova finds that selective participants (opting-in countries) show more conformity with EU laws than countries not participating at all (opting-out countries). However, there remains a gap in terms of EU law incorporation between fully integrated member states and selective participants. Zhelyazkova (2014, 743) provides two tentative explanations to explain this gap. First, selectively participating countries are less sensitive to the cost of noncompliance (for example, reputational costs from the Commission's infringement procedures) than fully participating countries. Second, "domestic opposition to policy implementation on EU conformity is stronger for selective participants" (733). This is because national governments in fully participating countries could use Brussels as a scapegoat to ensure compliance with EU policies that are undesirable for some political actors, but a government in an opting-in member state cannot do so if it had the flexibility not to participate in the first place. These hypotheses point towards country-specific factors influencing policy effectiveness in differentiated integration arrangements.
The two studies focus on very different types of differentiated arrangements and identify different conditions of effective differentiated integration. What they have in common, however, is a similar understanding of 'policy effectiveness'. Both studies define effectiveness as the effective enforcement of EU-level rules at the national level. This definition works well for the EEA and the various arrangements on migration and asylum, but it is too narrow to encompass all the dimensions of today's differentiated integration. Many differentiated organisations have more open-ended and less legalistic goals, such as promoting the exchange of views, fostering mutual learning or coordinating national policies in a 'soft' manner. Even differentiated organisations whose primary goal is to enact and enforce common decisions also play an important role as venues to promote the exchange of views among participants with a view to building up a common vision on shared problems.

Beyond goal attainment: towards a more encompassing approach to policy effectiveness
Understanding policy effectiveness as the enforcement of commonly adopted decisions assumes that differentiated arrangements respond to commonly agreed, fixed and well-articulated objectives. This vision of policy effectiveness as goal attainment is dominant in the 'new policy design' literature (Peters et al. 2018;Capano et al. 2019;Bali et al. 2019) and inspires most experiment designs in public evaluation. It has the merit of simplicity and is very useful for accountability purposes. However, it is rooted in a rational, functionalist and largely depoliticised conception of the policy process that does not fit well with reality (Le Galès 2020). In particular, a goal-attainment assessment is difficult in cases when the goals are vaguely formulated, contradict each other, carry different meanings to different people or are too large and the necessary trade-offs between multiple goals are not indicated (Vedung 2013). As pointed out by B. Guy Peters et al. (2018), the assumption of a rational and linear policy process is also particularly unfit to the reality of transnational policies, as in the case of EU-level differentiated arrangements. In transnational arenas, the number and type of actors may be extremely diverse, thus giving ground to more goal ambiguity and goal conflict. In many cases, there is no 'ready-made' toolbox of policy instruments to respond to problems. Instrument selection is, therefore, more a process of invention and innovation than about selecting among available tools. Finally, the 'line of command' between designing and implementing global policies is longer, more complex and potentially filled with ambiguity in comparison to that associated with domestic policies.
Another problem with a strict goal-attainment approach is that it neglects the existence of second-order effects and unintentional consequences of public interventions. In this respect, some authors consider it necessary to adopt a broader approach when assessing the effectiveness of policies, an impact approach that takes account of actual effects rather than intended results . When analysing the effectiveness of differentiated arrangements, it is particularly important to be attentive to the potential unintended effects for the EU as a whole. Differentiated policies may create policy inconsistencies with other EU-27 policies or political tensions between 'ins' (countries participating in the arrangement) and 'outs' (non-participating member states). Conversely, they may favour further harmonisation at the EU level. Ellen Vos and Maria Weimer (2016, 35), for example, demonstrate that successful invocations of opt-outs in the fields of public health and environmental protection have, in some cases, "ultimately triggered a process of further harmonisation at the EU level aiming to remove the regulatory disparities among the Member States". 4 An even broader vision of 'effectiveness' equates it to problem-solving capacity. This approach is frequent in comparative politics and the literature on international regimes. Understood this way, effectiveness becomes a synonym of performance or policy success and assesses the ability of a policy or arrangement to address, mitigate or resolve a given issue that triggered its creation. This broad vision of effectiveness has some advantages. It allows the researcher to judge a policy or arrangement according to different dimensions. Dorina Baltag and Iulian Romanyshyn (2017), for example, use five indicators to assess the performance of the EU neighbourhood policy related to the input of the policy (relevance, cohesion), the output (effectiveness, impact) and its interaction with an external environment (resilience). Allan McConnell (2013) organises many of these indicators into a three-dimensional assessment matrix, with performance evaluated in programmatic (goal achievement), process (the extent to which the policy is organised and managed in a way to maintain sustainable political coalitions and stakeholders' support) and political terms (the level of public support to the policy).
Two important ideas stand out from these classifications of indicators and dimensions. The first is that any problem-solving assessment shall combine "more objectiveoriented achievements, with perceptions of whether such achievements deserve the label of 'success'" (McConnell 2013, 487). The corollary is that the assessment becomes inherently subjective. In effect, any problem-solving analysis is susceptible to "the eye of the beholder problem", meaning that its definition depends on who is conducting the evaluation (Gutner andThompson 2013, 59, cit. in Romanyshyn 2017, 50). A way of reducing the degree of subjectivity is to compare the performance against some objective points of reference or benchmarks. Drawing on Arild Underdal (2002), Lavenex and Križić (2019) propose three benchmarks against which to compare the effectiveness of differentiated arrangements: a status-quo scenario (or "no-regime scenario": Underdal [2002, 9]); a scenario in which the policy would have been developed "without differentiated membership but otherwise with the same institutional characteristics as observed for the differentiated institution" (Lavenex and Križić 2019, 11); and an "ideal solution scenario" that would reflect most experts' view on "what constitutes the maximum that a particular group of actors can accomplish" (Underdal 2002, 9).
A second aspect that characterises a problem-solving analysis is the temporal dimension. As noted by Mallory E. Compton et al. (2019), both goal attainment and public and political perceptions unfold over time. As a result, "for a policy to be classified as completely successful, this level of performance must be sustained even in the case of exogenous contextual changes" (Compton et al. 2019). This temporal dimension is very relevant to analyse the effectiveness of differentiated arrangements. Differentiated solutions are frequently 'learning-by-doing' solutions. They are initial EU-level attempts at building a common response in a given policy area. They are also more prone to suffer from changes in the composition of the group of participating countries, affecting in turn the policy preferences and political dynamics within the group. In this respect, it seems particularly pertinent to adopt a diachronic approach to assess whether a given differentiated policy or arrangement maintained its effectiveness over time and was able to adapt to different contexts.
Based on this discussion of different conceptions of effectiveness, in studying differentiation, we suggest moving beyond a goal-attainment definition of effectiveness towards a more encompassing notion that includes elements from both the impact and problem-solving approach to effectiveness. From the impact approach, we borrow the idea that actual effects, also for the EU as a whole, are crucial in evaluating the effectiveness of differentiated policies or arrangements. And from the problem-solving approach, we adopt the argument that effectiveness needs to be evaluated over time, taking into account changing political preferences and contexts.

The effectiveness of differentiation in EU economic policy: empirical insights
A common finding from the studies on differentiation and policy effectiveness is that the effectiveness of differentiated arrangements cannot be explained by a single factor but results from the interplay of various elements, which we can group as country-specific, policy-related and institutional features. This article aims to explore in particular the role of institutional factors on the performance of differentiated economic arrangements. To this end, we use insights from the three aforementioned empirical studies that analyse the functioning and effectiveness of various differentiated arrangements in the EU energy sector (Franza et al. 2021), the financial sector (Mack 2020) and in the provision of access to the Single Market to third countries (Eisl 2020). These three studies have adopted a more encompassing definition of policy effectiveness, going beyond a goal-attainment approach. They are mostly of an exploratory nature. Based on qualitative methods such as expert interviews and document analyses, they provide tentative explanatory factors -sometimes policy-specific -of the relative success of the differentiated integration arrangements under study.

EU energy policy
Luca Franza et al. (2021) assess the effectiveness of three differentiated arrangements operating in the EU energy sector: two inter-state arrangements (the Energy Community and the Pentalateral Energy Forum) as well as a network of sub-national authorities (the Covenant of Mayors). The stated goal of the paper is to assess the extent to which these various arrangements have contributed to the attainment of "the three key EU energy policy goals of affordability (or economic competitiveness), security of supply and sustainability (or decarbonisation)". They do so by collecting information about their performance "through semi-structured interviews with personnel of the three differentiated integration institutions analysed" (4) .
The Energy Community is a treaty-based, highly formalised international organisation that involves the EU and most of its eastern and south-eastern neighbours. Its goal is to create a pan-European energy market across borders by promoting the adoption of the energy acquis communautaire by non-EU member states. When assessing its functioning and results, the authors conclude that it has been effective in promoting the adoption of the EU energy acquis in neighbourhood countries. However, whereas implementation has been highly successful in some areas (for example, EU rules on unbundling in the gas sector, which pushed a major gas reform in Ukraine), progress has been more limited in areas such as anti-trust regulation or decarbonisation legislation. Besides, there is a significant time lag between the adoption of a new EU rule and its transposition by the contracting parties. Finally, Franza et al. note that the Energy Community has played a positive role as a venue for dispute resolution between investors and governments in the region.
The Pentalateral Energy Forum (PEF) is a 'soft' governance framework for regional cooperation comprising six EU member states (Austria, Belgium, France, Germany, Luxembourg and the Netherlands) plus Switzerland. It aims to promote bilateral cooperation between electricity operators and to facilitate the discussion on cross-border trade issues with a view to enhancing the integration of electricity and gas markets among participating countries. Franza et al. consider that the PEF played a key role in achieving north-western European market coupling. They mention several initiatives put forward by the Forum that have favoured market integration, such as the establishment of a central auction point for the transmission of electricity capacity called PRISMA or the creation of a regional service authority that monitors cross-border electricity flows. In addition, they highlight some positive side-effects of the PEF. In particular, "the market coupling model that was implemented in north-western Europe thanks to the PEF later became the standard for electricity market integration between Italy and Slovenia, in central-eastern Europe, in the Iberian Peninsula and in Scandinavia" (13). Besides, the PRISMA capacity booking platform developed by the Forum became standard throughout Europe. At the same time, they highlight the risk that the PEF, over time, may accentuate differences between participating and non-participating countries. This may be particularly a problem for countries sharing borders with central-eastern Europe and the Mediterranean region. According to the authors, "these countries should ensure that their work and progress within the PEF does not result in unbridgeable differences with other neighbouring countries" (14).
The Covenant of Mayors (CoM) is a network of European local governments voluntarily committed to cooperate to attain the EU's climate and energy goals. CoM signatories pledge to reduce their CO 2 emissions by at least 40 per cent, to increase their resilience to the impact of climate change and to take action to alleviate energy poverty. Franza et al. consider that the CoM provides a "pathway to function and cooperate in a nonhomogeneous flexible manner that advances EU integration and the pursuit of EU goals" (17). They argue that the fact that participation is entirely voluntary and that implementation is based on tailored plans without negative consequences for those not complying with them facilitated the capacity to attract both large and small towns or villages. They note, however, that local action plans developed under the CoM umbrella are more successful where there is good coordination between signatories of the Covenant and their national governments.
Overall, Franza et al. conclude that "[differentiated integration] arrangements with very different degrees of institutionalisation, binding nature of the rules, membership, mandate and so on can all contribute to effectiveness. There is no 'one size fits all' approach in differentiation" (18). In particular, their analysis points out the existence of two different rationales inspiring differentiation in the energy sector, having different institutional requirements. Regional cooperation schemes (such as the Energy Community) aim to enhance coordination among interdependent countries. They have a stronger focus on rule compliance and thus need binding rules and enforcement capacity. Coalitions of the willing (such as the Covenant of Mayors) "emerge from the ambition to incentivise a faster pursuit of EU energy objectives by countries or subnational actors that show higher-than-average levels of ambition". They are likely to be more inclusive and "tend to rely predominantly on soft power and soft law, i.e., attempts to 'lead by example' through persuasion rather than coercive means" (7).

Banking Union
Sebastian Mack (2020) analyses the effectiveness of the arrangements and institutions included in the Banking Union. He does so by collecting information from external experts and analysis of official documents and reports. As he argues, whereas the main purpose of the Banking Union project was to break the sovereign-bank nexus among Euro area countries, the Banking Union was expected to play an integration-deepening role not only within the euro area but also for the EU as a whole. It was essential that the Banking Union did not jeopardise the Single Market; and in fact it was designed to reinforce it and eventually create positive spillover effects in the European integration dynamics (6).
In line with this reasoning, Mack focuses his analysis on whether the creation of the Banking Union has been effective in promoting integration among its members while avoiding distortions in the Single Market.
The Banking Union is currently composed of two main pillars: the bank supervision pillar, based on the Single Supervisory Mechanism (SSM), and the bank resolution pillar, based on the Single Resolution Mechanism (SRM) which consists of the Single Resolution Board (SRB) and the Single Resolution Fund (SRF). The paper analyses the effectiveness of the two pillars separately. The SSM is an EU mechanism managed by the European Central Bank (ECB) that is in charge of supervising all large banks in the participating states of the Banking Union. Mack observes that financial institutions have become safer and sounder since the inception of the SSM: banks built up additional capital to absorb potential losses and improved their asset quality. However, the existence of a single supervisor has not sufficed to completely eliminate the bank-sovereign nexus. In his view, a central reason for the insufficient progress in this regard is that supervision is not as strict as it should be because there is no parallel strong mechanism to deal with failing banks (11). Inside the bank resolution pillar, the SRB is an EU-level authority whose task is to draw up plans to wind down failing banks in an orderly manner. Unlike the ECB within the SSM, the SRB does not have direct executive powers over banks and relies on national resolution authorities for the implementation of its decisions. In some circumstances, the SRB can decide on the use of the common fund (SRF) to save banks. Mack notes that the bank resolution pillar has been less successful than the supervisory pillar. Despite the existence of a common set of rules on how to handle bank failures and a new EU resolution authority (the SRB), "in the vast majority of bank failures since 2015, member states ingeniously exploited the loopholes in the legal framework and circumvented the application of bail-in or even the involvement of the SRB" (11).
Apart from analysing the effectiveness of the two pillars separately, Mack discusses the overall effectiveness of the Banking Union. He draws on Lavenex and Križić (2019) to compare the performance of the current Banking Union against a 'hypothetical best solution' that leading experts regard as attaining the desired goals. Since bank supervision and crisis management complement each other, Mack argues, the 'ideal solution' would be to establish a fully fledged bank resolution pillar, with an SRB empowered to directly resolve all large banks under ECB supervision within the SSM without having to rely on national interventions. Despite this modest success in breaking up the banksovereign nexus, Mack concludes that the Banking Union "was successful in containing the euro area sovereign debt crisis" (18). Finally, Mack observes that the Banking Union has been very successful in preserving the integrity of the Single Market. Not only the risk of a two-tier EU bank regulatory and supervisory system has not materialised, but the Banking Union has helped to further integrate the whole EU financial system. According to him, to a certain extent, this has been the result of informal coordination practices between the ECB experts (in charge of the SSM) and experts from the EUwide European Banking Authority (EBA), regularly meeting in different international bodies and policy networks. However, Mack also points out the existence of specific legal safeguards to take account of the interests of non-participating member states, thus preventing the Banking Union from causing distortions in the Single Market. First, both the SSM and SRM regulations contain unity-protection provisions requiring the ECB and the SRB to refrain from any "action, proposal or policy" that would, "directly or indirectly, discriminate against any Member State or group of Member States" (Art. 1 SSM Regulation), "nationality or place of business" (Art. 6 SRM Regulation). Second, the two bodies are obliged to respect the interests of member states participating and not participating in the Banking Union and shall carry out their work "with full regard and duty of care for the unity and integrity of the internal market" (Art. 1 SSM Regulation and Art. 6 SRM regulation). Third, national resolution authorities from non-participating member states take a seat in the SRB's plenary sessions, as well as in any SRB executive session where the crisis management concerns an institution that has subsidiaries or significant branches in non-participating member states. Finally, the EUwide authority in charge of developing common rules for the banking sector -the EBA -applies a "double majority" to ensure the balancing of interests between member states inside and outside the Banking Union (Gren 2014, 72). Major decisions are subjected to the qualified majority of all EU-27 member states but also require a simple majority from both participating and non-participating member states in the Banking Union (Art. 44 EBA Regulation).

Third-country access to the Single Market
Andreas Eisl (2020) analyses and compares the effectiveness of various external economic differentiation arrangements -that is, arrangements providing access to the Single Market to third countries. Empirically, Eisl draws on "interviews with representatives of EU and national institutions [. . .] involved in the negotiations and management of different forms of external economic differentiation". He also examined "the main treaties and agreements, negotiation mandates, decisions of joint committees and other types of written documentation" of the various differentiated arrangements (4).
The analysed differentiated arrangements include the EEA, the regime of EU-Swiss bilateral agreements, EU Customs Union agreements, Deep and Comprehensive Free Trade Areas (DCFTAs) and Stabilisation and Association Agreements (SAAs) with neighbouring countries, as well as two comprehensive free trade agreements with other developed countries. These different forms of external economic differentiation vary both in their breadth and depth, regarding the covered Single Market policies and the governance frameworks in place. The EEA is organised in a two-pillar structure through which Norway, Iceland and Liechtenstein are -with a few exceptions -full members of the European Single Market and also participate in a number of other EU policies. EEA member states have to 'dynamically' integrate the evolving EU acquis into their legislation, which is enforced by the EFTA Surveillance Authority and the EFTA Court. 5 While excluded from voting on Single Market rules, Norway, Iceland and Liechtenstein are nevertheless involved in the decision-shaping process. Similarly, the more than 100 different agreements governing the EU-Swiss economic relationship provide deep Single Market access for Switzerland. In contrast to the EEA, the complex EU-Swiss bilateral agreements are not governed by an overarching institutional architecture and -for the most part -do neither include dynamic adaptation to the evolving EU acquis nor dispute settlement mechanisms to resolve conflicts. 6 The EU Customs Unions with Turkey, Andorra and San Marino provide a more partial access to the Single Market to third countries. They allow for the tariff-free movement of goods between their members, while forbidding them to negotiate their own tariffs with other third countries. In addition, these agreements require the non-EU Customs Unions members to align certain technical goods regulations with the EU acquis and to adopt some level playing field obligations, for example, regarding state aid. In institutional terms, EU Customs Unions do not contain any strong dispute settlement mechanisms. DCFTAs with Ukraine, Georgia and Moldova, and SAAs with six Western Balkans countries 7 equally provide partial access to the Single Market. There is a large degree of variation between these agreements in terms of breadth and depth. Especially DCFTAs can even include dynamic adaptation to the EU acquis, but often only in very narrow policy areas. Finally, comprehensive free trade agreements that the EU has recently established -for examplewith Canada and Japan provide the most limited access to the Single Market to third countries. Nevertheless, they go beyond 'traditional' free trade agreements as they contain more ambitious level playing field obligations but generally lack the institutional means to enforce rule alignment.
All of these arrangements thus provide, to a varying degree, access to (parts of) the Single Market. As Eisl notes, however, it is difficult to analyse and compare their effectiveness as "some are expected to promote alignment to EU legislation in view of future accession, [while] others basically aim at intensifying trade relations" (9). The former objective applies, for example, to the EU Customs Union with Turkey (an EU accession candidate country), the various SAAs with countries from the Western Balkans (Croatia had an SAA with the EU prior to joining it in 2013, while Serbia, Montenegro, 5 These institutions mirror the European Commission and European Court of Justice on the EU's side. 6 With the so-called 'guillotine clause' whereby the EU could terminate several key agreements simultaneously in the case of non-alignment, the EU has nevertheless leverage to keep divergence in the application of Single Market rules limited over time. 7 Serbia, Bosnia & Herzegovina, Albania, North Macedonia, Montenegro, Kosovo. Albania and North Macedonia currently have accession candidate status) but also to the agreements with Switzerland, as most bilateral agreements were signed when a future EU accession was not yet 'off the table'. The latter objective is central to the EU Customs Unions with Andorra and San Marino, as well as the different DCFTAs and comprehensive free trade agreements. The EEA does not really have any of these two broader objectives but rather attempts to ensure the homogeneity of the existing quasi-complete integration in the Single Market over time (see Frommelt 2017, 4). Given these differences in institutional arrangements and objectives, Eisl attempts to identify a common denominator of effectiveness across all forms of third-country access to the Single Market. According to him, a central objective common to the various agreements is "to foster long-term economic integration and cooperation between the EU and third countries without endangering the functioning of the Single Market" (Eisl 2020, 9). He thus defines effectiveness as the capacity to achieve this double objective. The focus on a wide range of differentiated arrangements governing third-country access to the Single Market makes Eisl's empirical analysis inherently comparative, where effectiveness is assessed in relative rather than absolute terms.
Based on this view of effectiveness, Eisl identifies three recurrent weaknesses in many arrangements. First, he argues that the lack of procedures to ensure a dynamic alignment to the EU acquis (that is, procedures to continuously update an agreement to new EU legislation), such as in the case of the vast majority of EU-Swiss bilateral agreements, undermines the effectiveness of agreements in the long run. This results from the fact that "the legal foundations of the Single Market themselves are constantly evolving, which can lead to growing divergences and loss of homogeneity between the EU and the respective third country if no counteractions are taken" (10). In contrast, Eisl's analysis suggests that the EEA, with its various procedures and institutions for ensuring dynamic alignment, is better suited to be effective over time.
Second, Eisl points out that arrangements based on complex and fragmented governance frameworks and lacking mechanisms for uniform interpretation and implementation (via common interpretation by the European Court of Justice or a dispute settlement mechanism) "might lead to varying interpretations of Single Market rules across different policy fields, threatening the homogeneity of the market inside and across different forms of external economic differentiation" (Ibid.). The more than 100 EU-Swiss bilateral agreements as well as the DCFTA with Ukraine are mentioned as deficient examples in this respect.
Furthermore, Eisl notes that well-functioning arrangements should be based on a fair balance between the depth of granted access to the Single Market and the regulatory, institutional and financial obligations in return. Perceived disequilibria between rights and obligations in the access to the Single Market granted to third countries could, on the one hand, undermine the support for the Single Market among member states themselves (potentially leading to attempts to circumvent certain Single Market rules, to lobby for opt-outs or even to leave the EU in the hope of more favourable conditions). On the other hand, they can also destabilise the various existing agreements with third countries for access to the Single Market, with the EU or third countries attempting to renegotiate less favourable arrangements. In the latter case, however, the exact equilibria also depend "on broader economic and political objectives for the EU and the respective third countries, such as EU membership" (11). When the prospect of EU membership disappears, this can lead to a perceived disequilibrium between previously agreed rights and obligations. "If not addressed, these evolutions can potentially endanger long-term economic integration and cooperation" (Ibid.). Key examples of such a situation are Turkey's membership in the EU Customs Union as well as the EU-Swiss bilateral agreements. In the first case, Turkey's disadvantageous position in the EU Customs Union was originally counterbalanced by the prospect of EU accession, which has since basically vanished. As a response, the functioning of the agreement has been partially undermined by the Turkish side. In the second case, the uncertainty of a later EU accession by Switzerland allowed the Swiss side to negotiate a complex system of Single Market access which the EU increasingly deemed as cherry-picked, especially since the Swiss side officially abandoned its ambitions to become to an EU member later. Consequently, the EU has sought to renegotiate the EU-Swiss bilateral agreements towards a unified governance framework, a dynamic adaptation to the evolving EU acquis and a potent dispute settlement mechanism.
Finally, Eisl notes that disequilibria between rights and obligations in various degrees of third-country access to the Single Market can not only affect the functioning of a specific form of agreement but can also develop side-effects on the effectiveness of other agreements. Swiss and Norwegian observers, for example, were following the Brexit negotiations closely. According to Eisl, should the EU have granted the United Kingdom a deep preferential access to the Single Market, this would have undermined the support for existing arrangements, increasing the perception of a relative disequilibrium between rights and obligations. To address the risks of perceived absolute or relative disequilibria of rights and obligations between the EU and the participants of a specific arrangement regarding third-country access to the Single Market, this should include in-built procedures for renegotiating the objectives of the agreement in light of changing political circumstances.

The influence of institutions on the effectiveness of differentiated arrangements
The discussion of the three empirical studies on differentiated arrangements in the EU energy sector, Banking Union and third-country access to the Single Market highlighted the extent to which these arrangements are effective and which factors might influence their (lack of) effectiveness. Comparing the findings, especially institutional factors seem to matter in a consistent fashion across differentiated arrangements in the EU economic policy field. Focusing on these institutional factors to explain effectiveness, we can derive three tentative hypotheses, potentially generalisable to other areas of differentiated integration.

Fit between institutional design and policy objectives
A first tentative hypothesis that we can draw from our analysis, mainly based on the work of Franza et al. (2021), is that different differentiated arrangements may require specific types of institutional set-up, more institutionalised and binding in some cases and more informal and flexible in others. This qualifies the hypothesis formulated by Lavenex and Križić (2019) that considers differentiated arrangements with a strong regulatory dimension more likely to ensure effectiveness because of their capacity to ensure members' compliance with agreed rules. As Franza et al. show, a stringent institutional set-up, with binding rules and enforcement capacity, is crucial for treaty-based energy cooperation schemes such as the Energy Community, but is not necessary, and probably not well suited, for a bottom-up differentiation initiative such as the Covenant of Mayors, in which participants want to lead by example through ambitious policies and where broad diffusion of innovation is key. For this latter type of differentiated arrangements, the institutional design needs to be shaped in a way that allows actors to maximise innovation and provide fora in which newly acquired knowledge can be shared and discussed in an inclusive manner (Tortola and Couperus 2020). At the national level, Important Projects of Common European Interest (IPCEIs) are a recent form of differentiation in the EU economic policy field that also tends to move in this direction, providing exceptions from competition rules for participating member states and their companies to achieve common EU objectives such as decarbonisation (European Commission 2021). This highlights that particularly those differentiated integration arrangements whose objectives are experimentation and policy-learning can benefit from a less institutionalised and more inclusive approach. Where rule compliance is a key objective, instead, the capacity of actors is more important, which should lead to a more reduced set of participating countries, as evidenced by the studies on Banking Union and third-country access to the Single Market.

Institutional provisions to prevent or mitigate negative side effects
The second tentative hypothesis developed in this study is that differentiated arrangements tend to be more effective when they include institutional mechanisms to address potential negative side effects on the EU as a whole. In this regard, particularly the paper by Mack (2020) on the Banking Union provides useful insights. A common concern among member states about differentiated arrangements in economic policy is that they could undermine the integrity of the Single Market and run against the interests of nonparticipating EU countries. Mack's analysis, however, shows that this is not the case for the Banking Union, as its institutional set-up includes several legal safeguards to rule out discriminatory actions against specific countries or groups of countries, to ensure the inclusion of non-participating member states in relevant meetings and to require support from countries outside of the Banking Union for important decisions. Institutional provisions to prevent or mitigate negative effects may also be relevant for other types of internal differentiation arrangements. Franza et al. (2021, 17), for example, find that the design of the Pentalateral Energy Forum may threaten its effectiveness, as its integration dynamics tend to accentuate differences between participating and nonparticipating member states. Here, the institutional set-up of the Banking Union could provide a template for provisions that could help prevent or mitigate negative side effects of differentiation for the EU as a whole.

Adaptable governance frameworks
The third tentative hypothesis derived from our analysis is that the effectiveness of differentiated arrangements depends on their adaptability over time to avoid institutional drift 8 and to incorporate improvements based on a learning process through policy feedback (Compton et al. 2019, 124). The differentiated arrangements studied in Eisl's (2020) analysis on thirdcountry access to the Single Market are particularly suited for making this type of arguments, as many of them have existed for more than twenty years, while the Banking Union and the arrangements in the EU energy sector are more recent. In addition, the constantly evolving EU acquis, changes in regional and global economic conditions, and changing political preferences among member states and third countries regarding EU enlargement pose constant challenges to the effectiveness of the various differentiated arrangements granting access to the Single Market, providing a particularly rich background for studying how well they work over time. Agreements that do neither include mechanisms for the dynamic integration of the EU regulatory environment nor contain mechanisms to adapt their objectives and instruments to changing circumstances tend to be considered ineffective or can even be actively rendered ineffective by at least one of the agreement partners by allowing institutional drift to occur (Hacker 2005). Eisl's study highlighted this phenomenon especially for the case of the EU-Swiss bilateral agreements and the EU Customs Union with Turkey. The static institutional designs of these differentiated arrangements made it difficult to adapt them to changing political preferences regarding EU accession, rendering them ineffective to a certain extent. To improve policy effectiveness, the institutional frameworks of differentiated integration arrangements should thus include mechanisms that allow for flexibility and adaptation over time to deal with changing political or economic circumstances. For differentiated integration arrangements that are more focused on mutual learning or policy experimentation (such as the Covenant of Mayors), the adaptability to changing circumstances and political preferences is to a certain extent already built into their less institutionalised bottom-up approach to policymaking. Such flexible arrangements are more protected from institutional drift as actors preferring the status quo cannot block other participating actors from, for example, experimenting with policies.

Conclusion
This article started with the observation that there has been a lack of interest in studying the effectiveness of EU differentiated arrangements, a surprising fact given the permanent nature of most differentiated arrangements. It subsequently pointed out that the few existing works studying effectiveness in differentiation have, so far, applied a narrow approach to effectiveness as goal attainment. We discussed the shortcomings of such a definition in comparison to an impact or a problem-solving approach to effectiveness, highlighting the need to move towards a more encompassing definition. This definition takes into account actual effects of differentiation for the EU as a whole rather than focusing only on intended results, and pays attention to the effectiveness of differentiated arrangements over time, constantly challenged by changing economic conditions and political preferences. In the empirical part, we then studied the effectiveness of differentiated arrangements in EU economic policy based on three studies conducted in the framework of the EU IDEA research project. These papers share an approach to effectiveness going beyond goal attainment and include the EU energy sector, the Banking Union and the provision of access to the Single Market to third countries.
Based on our analysis of the various differentiated arrangements in these three policy areas, this article derived three tentative hypotheses regarding the influence of institutional factors on their effectiveness. First, the effectiveness of a specific institutional set-up depends on the concrete objectives of a differentiated arrangement. Effectiveness is not necessarily related to the degree of institutionalisation or formalisation. Instead, it crucially depends on how much the institutional design fits with the policy goals of the organisation. These goals can be the implementation of some common EU rules or the adoption and enforcement of joint decisions, but can also be the exchange of information and the promotion of mutual learning on how to address a common problem. Second, differentiated arrangements are more effective when they include specific provisions to prevent or mitigate potential negative side effects to non-participating member states. Third, the adaptability of governance frameworks is key to ensuring policy effectiveness over time. Institutional mechanisms that facilitate adjustment in the face of changing economic conditions and political preferences are better suited to avoid 'institutional drift'.
As this study is largely of an exploratory nature, the discussion of a few points of caution and potential caveats are in order. First, the presented hypotheses should be considered preliminary, as they were formulated in an inductive manner based on empirical materials collected in a limited number of case studies. Second, the different hypotheses might be more relevant for differentiated arrangements in specific policy areas and with specific objectives. For example, the role of adaptable governance frameworks in ensuring effectiveness over time might be more crucial for arrangements where rule compliance is important for achieving their policy objectives. Third, as our analysis has focused on explaining the effectiveness of already existing differentiated integration arrangements, it is more limited in understanding why specific institutional designs were chosen in the first place and whether the factors determining their initial institutional setup make it difficult to modify them in line with our proposed hypotheses. Further research should thus investigate the context conditions which allow implementing institutional frameworks that favour policy effectiveness in differentiated integration arrangements. Finally, whether our tentative hypotheses might be applicable to differentiated arrangements beyond the economic policy field remains an open question at this stage, even if recent research, for example by Giovanni Grevi et al. (2020, 3) in the area of EU foreign policy, seems to suggest this to be the case. Follow-up research to test these hypotheses on other differentiated arrangements in EU economic policy and in other policy fields will help to further refine the arguments developed in this article.
innovation programme under grant agreement no. 822622. This article, and the Special Issue as a whole, reflect only the views of the authors, and the European Commission is not responsible for any use that may be made of the information it contains.