Differentiated policy implementation in the European Union

Abstract This special issue analyses the patterns, causes and consequences of Differentiated Policy Implementation (DPI) in the European Union (EU). DPI is an umbrella term for the diversity in the presence and use of discretion during legal and practical policy implementation processes and outcomes in the EU. The emergent DPI research agenda emphasises differentiation in EU policy implementation beyond mere legal compliance, which is more widespread, and its role in the broader political and policy processes of EU multilevel governance. The contributions highlight anticipated implementation as one dimension of DPI, as well as legal and practical implementation. DPI serves as an alternative to differentiated integration (DI), accommodating heterogeneous national preferences, capacities and conditions, and feeding back into EU policy-making. The impact of DPI on the EU’s output legitimacy and effectiveness depends on scope conditions that require more scholarly attention.

Euroscepticism (Hungary) (Dinan, Nugent, and Paterson 2017), threaten the functioning of the EU system of governance.In fact, the wish to 'take back control' was an oft-cited reason why the British electorate voted to leave the EU.In response, there is renewed interest in allowing member states to opt in or out of EU policies, or commit to them at different speeds and possibly at different levels-a phenomenon called 'differentiated integration' (Hooghe and Marks 2001;Leuffen, Rittberger, and Schimmelfennig 2013).
This special issue contributes to discussions about differentiation, but moves beyond the study of the adoption of supranational policy.Instead, it analyses the differentiated ways in which European integration works in practice, during policy implementation (Bondarouk and Mastenbroek 2018).Even in the absence of differentiated integration (DI), EU member states have extensive discretion when implementing EU law.This leads to an immense diversity of legal and practical policy solutions in EU member states (Thomann 2019).For example, some member states go much further than what is minimally required by the EU to reduce air pollution (Bondarouk and Mastenbroek 2018;Gollata and Newig 2017).Such 'customised' implementation can contribute to improved practical implementation of EU policies (Zhelyazkova and Thomann 2022).At the same time, additional requirements can also create red tape and unnecessary burdens that hamper the competitiveness of producers in the single market ('gold-plating') (Voermans 2009).Leaving room for member states to make choices during the implementation of EU policies may lead to arrangements that are better adapted to domestic conditions but may also lead to fragmentation and suboptimal policy solutions at the EU level (Knill 2015).
Independently of its consequences, the central starting point of the research agenda on DPI is that differentiation in EU governance is not limited to differentiated participation in EU policy (where different policy regimes apply to different member states) or to compliance with EU law (Treib 2014;Zbiral, Princen, and Smekal 2023).Member states may choose different, yet equally EU law-compliant policies, and they may combine policies with different procedural rules or informal practices.In this special issue, we therefore aim to conceptualise and explore the diversity of policy implementation practices in the EU and to analyse the drivers of this diversity, as well as its consequences.
The question of DPI in the EU is of great relevance as we lack theoretical and empirical knowledge about this diversity, its causes, and its implications (Fink and Ruffing 2017;Princen 2022).Furthermore, issues of policy implementation are seldom linked to other stages of the EU policy cycle and broader questions of European integration and its legitimacy (Hooghe and Marks 2001;Schmidt 2013;Scholten and Scholten 2017;Zhelyazkova 2014).Nevertheless, the processes and outcomes of EU policy implementation are central to the accountability, transparency, and effectiveness of European governance and its problem-solving capacity (Schmidt and Wood 2019;Thomann 2019).On the one hand, it is expected that DPI facilitates the ability of the EU to solve domestic problems by enabling member states to adapt EU rules to local needs.On the other hand, high diversity in member states' practices could also decrease the ability of the EU to reap the benefits of European harmonisation.
To set the stage for the individual contributions, this introduction addresses four core questions: 1. How can DPI be conceptualised and how does this concept relate and add to the field of EU implementation studies? 2. What is the relationship between differentiated integration and DPI in the EU? 3. What patterns and causes of DPI can be observed?4. What are the consequences of DPI for the effectiveness and legitimacy of the European Union?
Our introduction proceeds as follows.
In order to answer the first research question, we define the term differentiated policy implementation and situate the emerging research agenda on DPI within the literature on EU implementation, differentiation, and related concepts.Based on this, we briefly summarise the contributions.In the remainder of the article, we discuss the second, third and fourth research questions in turn and what the special issue contributions add to them.For each, we conclude by outlining areas for future research.

Conceptualising differentiated policy implementation
At a basic level, DPI refers to diversity among EU member states in the implementation of EU law and policies.This definition seems to state the obvious: that the implementation of EU law and policies is not the same in all member states (see Knill 2015;Treib 2014).However, by placing patterns of variation at its core and linking them to the notion of 'differentiation' , this concept provides a specific angle to studying implementation in the EU.In this special issue, DPI is broadly defined as the diversity in the existence and use of discretion during legal and practical policy implementation in the EU, both in terms of implementation processes and outcomes.This definition deliberately includes both compliant and non-compliant practices.In doing so, DPI includes other concepts denoting the variety of implementation practices among EU member states, while explicitly linking them to the notion of differentiation in the EU.Two particular developments have stimulated the academic agenda on DPI: the need to move beyond the question of compliance, and the need to understand policy implementation as another forum for differentiation within the EU multilevel system.

DPI: moving beyond legal compliance
The increased focus on DPI has been fostered by the need to better understand policy outcomes beyond compliance and the policy processes that lead to these results (Heidbreder 2017;Thomann and Sager 2017).Until recently, EU implementation scholars have primarily focused on questions of member states' non-conformity with EU law as the key output of implementation processes.This focus can be traced back to international relations theories on state compliance.It stems both from the assumption that states' conformity with international law is the key to effective international problem-solving, and the observation that compliance is far from self-evident in an international system without a central enforcement institution (Simmons 1998).As a result, EU research tended to almost exclusively frame EU policy implementation as a problem of state compliance with international (including EU) law with limited interest in other variations (Knill 2015;Treib 2014).
The top-down focus on compliance reduces deviations from centrally decided provisions to a control problem.Thus, it neglects the central role of discretionary decisions of national implementing actors in policy success and bottom-up problem-solving (Thomann and Sager 2017).However, discretion is an inevitable reality of EU policy implementation: Europeanisation scholars have pointed out that many EU rules are characterised by an inherent ambiguity and compromises in need of interpretation (Schmidt 2008).EU legislation usually provides (extensive) discretion to national authorities to implement EU rules in a way that meets both EU requirements and reflects domestic interests and capacities (cf.Van den Brink 2017).For example, member states can make national rules either stricter or more lenient for different target groups (Thomann 2019).Furthermore, transposition and practical implementation is dominated by national politics, where member states interptet EU rules differently (Falkner et al. 2005).Consequently, domestic bureaucratic actors also vary in the way they implement EU rules.
To account for these realities, scholars are more recently moving 'beyond compliance' as a dependent variable (Schmidt 2008;Thomann and Sager 2017) and instead conceive of DPI in a normatively neutral way, as the variation in implementation practices between and within member states, in response to the heterogeneity of domestic conditions (Bondarouk and Mastenbroek 2018;Bondarouk, Liefferink, and Mastenbroek 2020;Princen 2022;Zhelyazkova 2022;Thomann 2019).This perspective understands EU policy implementation as a combination of various processes and outcomes related to incorporating and executing the diverse policies enacted at the EU level-including binding regulations and directives, as well as voluntary 'soft policies' , targets, and economic instruments (Blom-Hansen et al. 2022;Knill and Tosun 2012).In this context, DPI consists of several stages involving a range of different actors, institutions, and organisations (Knill and Tosun 2012;Knill 2015;Zhelyazkova and Thomann 2022).The primary focus is not on the (undoubtedly relevant) question of whether member states do as the EU tells them (compliance), but on the ways in which member states adjust, interpret, and fine-tune EU policies (differentiated implementation), and the implications this has for the wider political and policy process in the EU.
The EU implementation literature distinguishes between legal (transposition) and practical implementation of EU law (Thomann and Zhelyazkova forthcoming).Most EU implementation research focuses on the transposition of EU directives, partly because legal implementation (in particular the timeliness of transposition measures) is relatively easy to measure (Treib 2014).Recent literature addresses, in addition, the implementation of regulations, coined as supplementation, as some regulations either explicitly require implementation or are that vague that they are in need of specification (Blom-Hansen et al. 2022).Practical implementation refers to laws and policies being put into practice by the relevant administrative institutions.While most scholars analyse the implementation of binding EU acts, EU soft law instruments also produce important legal and practical effects.These, however, are less well researched (Falkner et al. 2005).
Even if member states achieve legal compliance, EU policies change during their subsequent implementation in national law and in practice.Beyond compliance with EU policies, more fine-grained, diverse patterns of both legal and practical national responses to EU law are widespread (Thomann 2019;Thomann and Zhelyazkova 2017).Domestic changes and adaptations are equally relevant in the implementation of EU directives, regulations, as well as soft law.Moreover, research on DPI stresses the importance of practical implementation (Versluis 2007;Zhelyazkova, Kaya, and Schrama 2016) and linking implementation with other stages of EU policy processes (Thomson 2010;Thomson, Torenvlied, and Arregui 2007;Zhelyazkova 2013Zhelyazkova , 2014)).

DPI as a form of differentiation
Although DPI as a phenomenon is as old as the EU itself, the research agenda has been reinvigorated in response to research on differentiated integration (DI) in the EU.Studies of DI have exclusively focused on explaining 'variations in the level and intensity of participation in European policy regimes' (Wallace 1998: 137;Majone 2009).This literature understands DI as an institutional strategy to overcome increased diversity in integration preferences and capacities of member states (Dyson and Sepos 2010: 5-6;Majone 2009: 221;Schimmelfennig and Winzen 2014).However, there is a lack of theoretical and empirical work regarding the relationship between different levels of participation in the EU and national policy outcomes (Zhelyazkova 2014).Börzel (2002: 194) has already highlighted how most studies of EU governance 'self-consciously concentrate on one side of the equation' .The result is a limited understanding of the role of implementation in the EU's broader policy and political process (Zgaga et al. forthcoming).Thus, Thomann and Sager (2017: 1401-1402) highlight that 'More research should scrutinise how EU policy-making interacts with implementation' .
Clearly, forms of differentiation do not only emerge at the EU level but are even more prominent in the national implementation of EU law and policies (Zbiral, Princen, and Smekal 2023).In fact, instances of DPI are more widespread than DI, as member states use their discretion beyond exemptions from treaty articles and secondary legislation (see, for instance, Thomann and Zhelyazkova 2017).For example, Thomson (2010) shows that member states exercise 'opposition through the back door' against EU policies when transposing them.Zgaga et al. (forthcoming) show that member states customise EU fiscal policy depending on, amongst other things, how much consensus there was amongst important national political actors to actually delegate fiscal power, a 'core state power' , to the EU.As Steunenberg (2019) argues, citizen preferences influence policy-as-implemented (as a form of policy responsiveness), which may therefore differ from policy-as-adopted.Overall, it is still surprising how little is known about the political role of DPI for member states.
As these elaborations show, DPI may act as an alternative to DI, where legislative and administrative authorities modify EU policies during implementation.In addition, the policy cycle may start over again, feeding experiences with implementation back into EU policy-making (Börzel 2002;Polman, van Eerd, and Zwaan 2022;Zhelyazkova 2014).The research agenda on DPI centrally seeks to understand this wider, but under-researched, role of differentiated implementation in the EU's policy and political processes.

DPI and related concepts
The notion of DPI serves as an umbrella term for a number of related concepts in the literature, many of which capture differentiation in legal transposition (Fink and Ruffing 2017).For example, Thomann (2019) conceptualises the diversity in domestic responses to EU law as 'customisation': the extent to which member states change EU rules when transposing them.These regulatory changes occur along two dimensions: density and restrictiveness.Customised density refers to the quantity of changes to the degree of regulatory penetration, complexity and internal differentiation (Knill, Schulze, and Tosun 2012).Customised restrictiveness captures how EU rules are adapted in the scope and extent of state intervention to influence target group behaviour (Schaffrin, Sewerin, and Seubert 2015).Such changes affect the obligations of national implementers, their level of leeway, or the generosity of the services or resources they provide (Bauer and Knill 2014: 33;Knill, Schulze, and Tosun 2012: 430).
Other related concepts include elements of policy-making and/or practical implementation.For example, Bugdahn (2005) refers to the process of adaptation of EU-wide rules to domestic circumstances as 'domestication' , while others speak of 'creative' compliance (Batory 2016).Bondarouk and Mastenbroek (2018: 17) study (subnational) 'implementation performance as the intensity of policy outputs undertaken by implementers in response to EU policy instruments -relative to the directive's objectives (vertical aspect) and to other implementers' outputs (horizontal comparison).While 'experimentalist governance' does not only focus on implementation but also on the type of policies adopted in the first place, we see it as a concept related to DPI, as it refers to a mode of governing in which policies develop over time in response to (local) experimentation (Sabel and Zeitlin 2008;Zeitlin and Rangoni 2023).

Summary of contributions
The special issue contributions address the conceptualisation of differentiated implementation, its empirical varieties, and its causes and consequences (see Table A1).They cover both 'classical' patterns of customisation and restrictiveness at the level of transposition (Brendler and Thomann 2023;Pircher, de la Porte, and Szelewa 2023;Zhelyazkova 2022), but also study DPI as implementation 'on paper' at the level of management plans (Mancheva, Pihlajamäki, and Keskinen 2023), as well as delays in implementation (Pollex and Ruffing 2023).The special issue includes studies of practical implementation activities (Thatcher and Garcia Quesada 2023), enforcement by EU agencies alongside national authorities (Akbik, Freudlsperger, and Migliorati 2023), network dynamics regarding EU policy implementation (Soares 2023), and implementation practices (Mizrahi-Borohovich, Newman, and Sivan-Sevilla 2023).Contributions also extend the notion of DPI not only to uses of discretion by member states, but also to the degree of discretion used by member states that could potentially lead to DPI.We call this 'anticipated implementation' .Princen et al. (2022) analyse the degree of flexibility granted in EU legislation to national implementation choices.Cabane and Lodge (2023) suggest that anticipated consequences of DPI play a central role in the design of transboundary crisis management responses.Princen et al. (2022) address the question of whether differentiated integration and flexibility serve different functions in EU legislation.Flexibility is defined as the room member states have to make choices during the implementation of EU legislation.Like differentiated integration, flexibility in implementation is a way of dealing with heterogeneity among member states.It allows for more tailored implementation and may also facilitate decision making at the EU level.

Anticipated implementation
Their study analyses the relationship between flexibility in implementation and differentiated integration in EU directives adopted between 2006 and 2015.The findings indicate that flexibility in implementation and DI tend to be used together in directives, but address different aspects of heterogeneity between member states.Cabane and Lodge (2023) analyse the development of transboundary crisis management regimes at the EU level.The authors expect that EU crisis management regimes are driven by EU institutions' and member states' anticipation of differentiated implementation.It is thus expected that member states will use their discretion during policy implementation, which might result in negative externalities for other member states' crisis management responses.Therefore, EU-level actors deliberately limit the scope of member states' discretion, leading to distinct modes of EU transboundary crisis management.

Transposition
In their contribution, Brendler and Thomann (2023) study the role of institutional misfit as an explanation for member states' customisation of three provisions of EU Directive 2009/28/EC on the promotion of renewable energies (RE): (1) national RE targets, (2) renewable electricity (RES-E) support requirements, and (3) biofuels support requirements across six countries: Austria, France, Germany, the Netherlands, Sweden and the United Kingdom.Differentiated implementation is defined as changes in the restrictiveness and density of the directive during the transposition of EU rules.The findings suggest that both literal transposition and extreme customisation of EU provisions are rare.The study finds that, contrary to the goodness-of-fit theory, customisation does not require high misfit.Instead, we may observe customised implementation, where high institutional fit is combined with high salience.More precisely, member states with a tradition of active state intervention and legalism in the renewable energy sector (high policy fit) proactively customised salient EU provisions to pursue more ambitious policies.These results suggest that differentiated policy implementation is distinct from compliance and requires further theorising (Mancheva, Pihlajamäki, and Keskinen 2023;Pircher, de la Porte, and Szelewa 2023;Zhelyazkova 2022).
In a similar vein, Pircher, de la Porte, and Szelewa ( 2023) study both the customisation of EU Social Policy and how domestic actors implement adopted rules in practice.Thus, differentiated implementation is broadly defined as diversity in legal and practical implementation.In particular, the study employs an actor-centered approach to analyse the implementation of the Work-Life-Balance Directive (WLBD) after its adoption at the EU level in three countries: Denmark, Germany, and Poland.The authors expect that economic costs as well as values (displayed in different gender equality regimes) motivate domestic actors to either push for change or to defend the status quo.The findings suggest that differentiated implementation is determined by both motivations (costs and/or values) and degree of mobilisation of relevant domestic actors.Moreover, the increased flexibility and openness of the WLBD's provisions to find a compromise at the EU level opened the door for customisation as well as greater variation in practical implementation.Pollex and Ruffing (2023) examine differentiated implementation of EU tobacco policy in four EU member states: Austria, Germany, Ireland and the United Kingdom.Their contribution uniquely combines theoretical approaches focusing on analyses of party and bureaucratic politics around the national implementation of both soft-and hard-law acts.The study emphasises the importance of integrating political and -often neglected -bureaucratic factors to better understand differences in implementation, explaining both 'gold-plating' and insufficient or improper implementation.In particular, the findings suggest that portfolio allocation, departmental struggles and partisan politics explain differential implementation in the four countries.Thus, EU policy implementation hinges on domestic aspects that go beyond policy misfit.Moreover, national regulatory styles, general positions on health and tobacco policy, and the specific history of tobacco policy in the country are important scope conditions in explaining differences in gold-plating.Zhelyazkova (2022) analyses the impact of compliance and DI of EU asylum policy on differentiated transposition in 27 EU member states.In particular, the study addresses the conditions under which compliance with more liberal EU asylum policies diminishes the autonomy of national governments to adopt more restrictive immigration rules (when these are in line with their policy preferences).The findings suggest that EU policies do not constrain governments to restrict immigration or limit the standards applied to asylum seekers when they attach high salience to their policy preferences.Similarly, differentiated integration (i.e.opt-outs from EU asylum policy) allow member states to adopt more restrictive policies when they disagree with the liberal EU rules.In other words, ideas about the restrictiveness of EU rules (at least in the context of EU asylum and immigration policy) are unfounded.Instead, differentiated implementation depends on the importance domestic decision-makers place on immigration policy.Mancheva, Pihlajamäki, and Keskinen (2023) study the consequences of differentiated implementation on the accountability of national collaborative institutions.They compare collaborative institutions in Finland and Sweden that were responsible for the transposition and implementation of two EU directives in water policy.Following existing research, differentiated implementation is broadly defined as member states' interpretation and adaptation of EU policies during the process of implementation (Fink and Ruffing 2017;Zhelyazkova and Thomann 2022).The authors specifically focus on 'customised restrictiveness' in the rules establishing collaborative institutions, or 'how domestic rules differ from the EU legislation in content' (Zhelyazkova and Thomann 2022).The contribution indirectly touches on the connection between establishing accountable processes for planning and decision making, and the legitimacy of EU policies.The results reveal that, when directives define only some, but not all, of the important rules and procedures needed for achieving accountable collaborative institutions, member states design the implementing governance structures according to their administrative traditions.This discretion leads to institutional variation, even when countries have relatively similar governance contexts.

Practical implementation
The article by Akbik, Freudlsperger, and Migliorati (2023) seeks to explain differentiated participation of EU agencies in joint direct implementation activities (DIA), conducted together with national authorities.Thus, the study does not only examine the involvement of national authorities, but also the participation of EU institutions in policy implementation.The authors specifically focus on inspections, investigations, border patrols, processing of asylum-seekers, and returns of migrants.They distinguish between two dimensions of differentiated implementation: territorial and procedural.Whereas territorial differentiation relates to member states' participation in joint implementation activities with EU agencies, procedural implementation relates to the degree of uniformity in the guidelines for organising DIAs.The findings suggest that territorially differentiated participation is a stable feature of DIAs in politicised fields.Despite tendencies to create more uniform procedures over time and across policy areas, high uniformity prevails mostly in areas affecting member states equally.The implications of this empirical pattern are ambivalent.On the one hand, the enduring differentiation of implementation practices in politicised domains reflects adapting practices to local circumstances.On the other hand, exceedingly differentiated procedures may be an obstacle to building a more viable system of multi-level implementation in the EU.Mizrahi-Borohovich et al. (2023) analyse how NGOs vary in their policy implementation patterns in relation to EU legislation on privacy issues.They specifically focus on GDPR-related complaints filed by 13 NGOs from 12 EU member states and the activities of each of these NGOs based on the activity they document on their websites.The study presents a novel typology to understand policy implementation engagement by NGOs and finds how NGOs converge towards a strategic transnational mode of implementaiton.The authors address the causes of policy enforcement in data protection law and, more precisely, de facto interactions between various stakeholders in the policy implementation process.It shows that, while GDPR implementation is still in its infancy, NGOs seem to engage in actions that reach beyond their own member states and file complaints with broad policy implications, while potentially neglecting local privacy issues in NGO jurisdictions.These efforts have increased the saliency of data protection issues both within and beyond national borders, shaping GDPR implementation in the EU through bottom-up decision making and actions.Thus, NGOs can complement top-down implementation gaps in the EU and could serve as a strategic resource for the EU's problem-solving capacity.An analysis of the EU's problem-solving capacity, then, should not only include interactions between member states and firms, but also how third-parties engage and shape the policy implementation process.
Thatcher and Garcia Quesada (2023) contribution focuses on outlining and explaining the consequences of differentiated implementation for further integration.Differentiated implementation is studied in the context of the EU labelling system for food and drink, based on its place of origin and production processes involved.The study shows that despite features such as diverse national interests and gastronationalism, differentiated implementation has resulted in renewed unified vertical integration.The article identifies three processes that contributed to further EU integration in this area: EU legislative requirements that encouraged the establishment of producer groups; European Court of Justice decisions; and free trade negotiations that altered EU rules.
The study by Soares (2023) addresses how European Administrative Networks (EANs) perceive their role in the implementation of EU policy.More precisely, the article analyses the exchanges of information, best practices and advice about implementation of EU environmental policy between national regulators participating in the Network of the Heads of European Environmental Protection Agencies (EPA Network).While the contribution does not specifically study outcomes or processes of differentiated implementation, national regulators contribute to differentiated implementation through their diverse expertise on EU environmental policy.Furthermore, even though EANs may be presented as horizontal mechanisms aiming to enhance EU problem-solving, the diverse interaction patterns point to existing differentiation across network members and the co-existence of diverse implementation practices.Finally, the findings also show that while EANs may be seen as important mechanisms to improve national implementation, participating member states have different perceptions about the impact of the network on national implementation.
This short line-up illustrates the different notions of differentiated implementation in this special issue and shows how empirically rich this phenomenon is.The remaining sections of this article elaborate on how the articles contribute to the three research questions, what the relationship between differentiated integration and DPI in the EU is, what patterns and causes of DPI can be observed, and what consequences we can find of DPI on the effectiveness and legitimacy of the EU.

Relationship of DPI with DI
Since DPI and DI are both forms of differentiation within the EU, the relationship between the two merits further attention.Both forms of differentiation can occur in various combinations.EU-level rules that apply to all member states (so, do not provide for DI) may offer a lot of room for DPI, or not.Likewise, an EU-level arrangement that includes several opt-outs (so, does provide for DI) may leave little or a lot of room for differences in implementation among the member states that it applies to.Historically, DPI precedes DI as an empirical phenomenon in the EU.As studies of DI have shown, the early decades of European integration were premised on a uniform process of integration across member states.The notion of using opt-outs to accommodate diversity among member states only arose in relation to enlargement in the 1970s (Dyson and Sepos 2010).The use of opt-outs has expanded since then, often related to new rounds of enlargement and treaty revisions (Leuffen, Rittberger, and Schimmelfennig 2013).
By contrast, DPI has always been part of the way the EU operated, both formally, through the use of directives and discretion in EU legislation, and informally, through differences in the actual implementation of EU policies between member states.Moreover, as mentioned earlier, empirically, DPI is much more widespread than DI.DI scholars seeking to quantify the empirical significance of opt-outs in the EU find opt-outs in about 43 per cent of EU treaty law, only a minority of which was actually used by member states (Schimmelfennig and Winzen 2014), and less than 20 per cent of EU secondary law was differentiated in the mid-2010s (Duttle et al. 2017).Conversely, a comprehensive study of policy implementation in two EU policy sectors and all member states have shown not only how widespread customisation is, but also that about 80 per cent of customised implementation goes along with legal compliance (Thomann and Zhelyazkova 2017).As we discuss below, additional (mostly case-study based) evidence on DPI suggests that literal, i.e. non-differentiated policy implementation, is a rare phenomenon in the EU.This is even more the case when moving from legal transposition to practical implementation of EU policies (Zhelyazkova, Kaya, and Schrama 2016).
Until recently, however, DPI was not seen as a deliberate strategy to deal with diversity and hence not linked to the broader notion of differentiation.This is surprising not least because DI in the EU is seen as a means to achieve responsiveness to member state preferences and idiosyncrasies (Winzen 2016).However, local citzen preferences influence policies-as-implemented at least as much as they influence policy-as-adopted and, therefore, policy implementation is a crucial means for national governments to be responsive to citizen preferences (Steunenberg 2019).Seeing DPI as a form of differentiation and exploring its potential as an alternative to DI is therefore worthwhile for several reasons.To begin with, under certain circumstances, DPI may serve as a more efficient response to heterogeneity of preferences than negotiating opt-outs at the EU level.For instance, DPI can help national governments avoid reputation costs for not joining international agreements or not being able to negotiate exemptions at a given point in time.Moreover, DPI is often more accepted than DI.While scholars contend that DI could lead to unequal treatment and exacerbate divergences across member states (Eriksen 2018;Kelemen 2021), DPI could also be an effective and more legitimate tool to fit EU objectives to national contexts (Merlo and Fasone 2021;Thomann 2019).
It is also unclear whether DI facilitates the implementation and compliance of EU policies, as early empirical findings remain mixed.In particular, Zhelyazkova (2014), who studied the link between differentiated integration and conformity with EU law, finds that member states with flexible arrangements at the EU level are less likely to comply with EU policies than fully participating member states: selective integration translates into selective compliance.Similarly, a recent study finds that DI increases the likelihood of non-compliance (Sczepanski and Börzel 2023).These findings raise doubts about whether DI is an effective instrument to resolve domestic policy problems.For instance, EU-level exemptions are typically negotiated within a limited period of time during the EU policy-making process.Instead, DPI provides for a more flexible response to changes in the preferences and capacities of national implementers over time.
Generally, only very few studies have explicitly looked at the link between policy adoption and policy implementation in the EU.Several studies find that member states have incentives to deviate from EU policies during the implementation stage if they were adopted without their support (Thomson 2010;Zhelyazkova 2013).These incentives translate into non-compliance during transposition, depending on how strongly the Commission monitors compliance.Recently, Zgaga et al. (forthcoming) focus on the link between EU integration and customised implementation of EU fiscal policy, a core state power.They find that member states exercised more subtle forms of 'opposition through the back door' , rather than outright non-compliance when EU decisions either did not reflect their preferences, or did not provide a credible deterrent against a minimalist approach to transposition.Another dynamic is when member states act as eager problem-solvers or signal compliance when 'uploading' their policy preferences at the EU level (Börzel 2002).
Some of the contributions to this special issue look at the link between DI and DPI.For example, Abkik et al. (2023) rely on the literature on DI and argue that, like DI, DPI occurs at the intersection between functional pressures for uniformity (interdependence) and post-functional pressures for differentiation (politicisation).Zhelyazkova (2022) directly and empirically analyses the link between DI and DPI, and finds that more flexible arrangements are conducive to DPI in EU migration policy.Princen et al. (2022) argue and show that discretion (which they call 'flexibility in implementation') and DI are responses to the same underlying issue of heterogeneity among member states.However, flexibility in implementation and DI are used to address different aspects of that heterogeneity.
Other contributions show that DPI can feed back into DI.Cabane and Lodge (2023) argue that anticipated DPI leads to distinct forms of DI at the level of crisis management.Finally, Thatcher and Garcia Quesada (2023) examine how differentiated national implementation of EU policies affects later European integration.In the case of geographical indications for food and drink in the EU, they find that differentiated implementation actually reinforced unified vertical integration, thus decreasing DI in the context of food policy.

Patterns and drivers of DPI in the EU
Research is still emerging on the patterns and drivers of DPI, and few studies have systematically explained variations in member states' responses to EU policies.In this section, we address the broader discussion and highlight the contributions of the special issue to the discussion.

Patterns of DPI
As mentioned previously, DPI, when defined as customised implementation, has been shown to be widespread and follow sector-specific logic of either adding stringency to EU minimal requirements, or conversely not going as far as the EU would allow (Thomann and Zhelyazkova 2017).However, most available evidence on DPI patterns is based on single or small-N analyses, which complicates gaining a systematic picture of DPI in the EU.Two contributions to the present special issue, however, move our quantitative empirical knowledge on DPI forward.Princen et al. (2022), focusing on flexibility in implementation as a form of anticipated DPI, analyse all individual EU directives adopted between 2006 and 2015.They show that, while opt-outs (DI) occurred in only about 20 per cent of the directives, virtually all directives contained some form of flexibility.When distinguishing between five types of flexibility, they find that elaboration discretion represents the most frequent type (about 40%), followed by reference to national law provisions (about 25%).Focusing on EU migration policy, Zhelyazkova (2022) looks at the differentiated restrictiveness of the transposition of 120 policy issues between 2006 and 2013.She finds that, on average, 24 per cent of these rules were implemented more restrictively than the EU rule.Additionally, Akbik, Freudlsperger, and Migliorati (2023) innovate the study of DPI by analysing the direct implementation activities of EU agencies, and developing a four-fold typology along the two dimensions of territorial and procedural unformity.Empirically, they find that differentiated participation is a stable feature of DIAs in politicised fields.However, they also find a tendency to create more uniform procedures over time and across policy fields.Thomann (2019), in her comparative study of customisation of food safety policies in five member states, finds that customisation often has the aim of either making EU policies work in practice ('problem-solving'), adapting them to local circumstances, or exercising 'opposition through the back door' (Thomson 2010).The contributions of the special issue expand our knowledge of the drivers of DPI by focusing on factors at both the EU and national levels.At the EU level, DI and discretion are expected to allow governments to follow their own preferences during the transposition process.At the national level, the policy-specific preferences of relevant domestic actors and the mismatch between national and EU rules are expected to affect DPI.Generally speaking, research on DPI has identified five main drivers of DPI (Thomann and Zhelyazkova, forthcoming).

Drivers of DPI
First, the literature focuses on the level of discretion granted by EU policies (Thomann and Zhelyazkova, forthcoming;Zbiral, Princen, and Smekal 2023), but also their coherence or ambiguity (Thomann 2019).If EU rules offer flexibility, then this influences the extent and direction of customisation (Thomann and Sager 2017).If EU rules offer limited flexibility to member states, this will arguably diminish the room for DPI without incurring infringement costs (Princen 2022).However, Katelouzou and Konstantinos (2021), who study the surprisingly literal and harmonised implementation of EU shareholder rights, also find that EU soft norms and stewardship practices can foster good practices in member states, by increasing market actors' familiarity with and preparedness for EU rules, and by offering mechanisms for innovative norm-generation.
Among the special issue contributions, Pircher, de la Porte, and Szelewa (2023) find that the flexibility and openness of EU provisions allowed for customisation of EU rules in national implementation and more variation in practical implementation.Princen et al. (2022) dig deeper into the kind of flexibility that matters for DPI, and distinguish between five types of discretion.Elaboration discretion means permission for member states to further specify a provision.Reference to national legal norms indicates the use of pre-existing national legal norms for the definition of concepts or the scope of a directive.Minimum harmonisation gives member states permission to adopt more stringent standards.Scope discretion prevails when member states can expand or restrict the categories of cases to which a provision applies.Finally, member states can have discretion in application on a case-by-case basis.These distinctions lay a useful basis for further and more systematic theorising of the precise ways in which discretion matters for DPI.
Second, DPI is a way for member states to adjust EU policies to their own policy preferences, regulatory systems, and problem-solving styles, i.e. align their domestic policy and institutional fit (Casula 2022;Katelouzou and Konstantinos 2021).For example, Thomann (2019) finds that countries customise EU food safety rules to make them fit their domestic sectoral interventionist styles, or their own national approaches to self-regulation that include non-state actors in implementing and enforcing EU policies.Aureli, Magnaghi, and Salvatori (2019) highlight the importance of 'goodness of fit' for the achievement of formal harmonisation of non-financial disclosure rules between the UK, France and Italy: DPI may allow member states to reduce adaptation costs (a finding shared by Skjaerseth and Rosendal 2023), or to embrace policy change.Borraz et al. (2022), focusing on practical implementation of food safety inspections, show that EU requirements were filtered through long-entrenched regulatory styles and modes of food business organisation, ultimately reinforcing pre-existing inspection practices.
In the current special issue, Brendler and Thomann (2023) also apply the insitutional misfit theory, developed to explain member state compliance, to explain the customisation of EU renewable energy policy.They find that EU policies' fit with national regulatory styles plays an unexpected role in DPI, depending on issue salience (Bondarouk, Liefferink, and Mastenbroek 2020): Misfit does not trigger customised implementation.Instead, when high institutional fit meets high salience, member states may issue substantively more ambitious policies than the EU requires.However, when high institutional fit meets low salience, member states have no impetus to customise EU rules.Pircher, de la Porte, and Szelewa (2023) focus on the importance of informal institutional fit, in terms of the fit of EU work-life balance policies with norms and values of gender equality.
Third, DPI is a way for member states to exercise responsiveness to domestic politics.Examples abound, and include populist politics triggering creative compliance (Batory 2016), organised domestic interests triggering overachievement of EU rules (Skjaerseth and Rosendal 2023), and bureaucratic politics acting as a driver of customisation (Logmani et al. 2017;Hundehege 2023).In this vein, (varying) levels of issue salience across member states and subnational units have been identified as an important driver of differences in implementation practices (Bondarouk, Liefferink, and Mastenbroek 2020;Versluis 2007).If EU rules grant some flexibility and few domestic institutional constraints prevail, then domestic resistance leads member states to extensively customise salient EU rules to make these rules fit the prevalent preference constellation (Thomann 2019).Skjaerseth and Rosendal (2023) report a dynamic of policy feedback due to municipial protests.
In this special issue, Abkik et al. (2023) find that politicisation is associated with different levels of participation of member states in direct implementation activities.Moreover, Pollex and Ruffing (2023) highlight the importance of support by political parties, but also that of competent ministries.In particular, they demonstrate that turf conflicts between ministries lead to minimal implementation, which is why the structure of portfolio allocation in EU member states is decisive for differentiated implementation.Zhelyazkova (2022) shows that governments with highly salient and restrictive immigration stances are more likely to find solutions to comply with EU rules in a way that fits their policy preferences.The study by Mizrahi-Borohovich, Newman, and Sivan-Sevilla (2023) offers a new perspective, showing that involving NGOs can decrease variation in DPI at the enforcement stage over time.
Fourth, DPI has functional drivers: for instance, strong shared functional interests can lead to coherent ways in which member states regulate more ambitiously than the minimal EU rules (Hartlapp 2014;Thomann 2019).Liefferink et al. (2021) stress the role of problem pressure for differentiated water policies; they find that high problem pressure tends to trigger the use of more authoritative policy instruments, while soft policies and incentives are used when problem pressure is lower.Thomann (2019) crucially finds that DPI is a means to successfully solve intractable policy problems when policies are characterised by high conflict and ambiguity: 'Here adaptive implementation allows for policies to be specified, modified, and revised, and for target groups to learn by doing, to compromise, and' own' the policy during the implementation process' (Thomann 2019: 209).Similarly, Fink and Ruffing (2017) show that functional and domestic political factors work together to lead to over-implementation.
Finally, (administrative) capacity is another driver of DPI, especially at the level of practical implementation (Bondarouk and Liefferink 2017;Lindstrom 2021).For example, Rossel, Unger, and Ferwerda (2022) find that corruption, government effectiveness, regulatory quality, tax morale, and tax administrative capacity help explain differences of legal and practical implementation of the EU's anti-money laundering directive.In the current special issue, Pircher, de la Porte, and Szelewa (2023) further find that not only the administrative, but also the economic costs of implementing an EU directive matter.

The consequences of DPI for the EU's effectiveness and legitimacy
Even more than the drivers of DPI, research on the consequences of DPI is in its infancy.Hence, we have a limited understanding of the impact of DPI on the effectiveness and legitimacy of the EU's multi-level system of governance.An exception is Thomann (2019) in her comparative case study of customised EU food safety regulations in five Western European countries.She finds scope conditions matter for when DPI contributes to both better and worse practical implementation outcomes.First, the degree of differentiation can be irrelevant for successful practical implementation when policy problems are tractable or domestic circumstances are otherwise ideal.Moreover, DPI is not a solution if the policies are incoherently designed.Second, customisation can improve problem-solving when policy problems are intractable or when domestic regulatory contexts challenge the prospect of successful implementation.However, sometimes, non-differentiation contributes to policy success, particularly if the EU rule is important and uncontested, and if domestic policy designs support its smooth implementation.More often, however, literal transposition explains why practical implementation fails (Thomann 2019: 198).She concludes that 'Customised implementation not only affects policy results (output legitimacy), it also provides procedural opportunities to involve societal actors and 'take back control' over distant EU policies (input and throughput legitimacy)' (Thomann 2019(Thomann : 2012)).
Another notable exception is a recent study by Zhelyazkova and Thomann (2022), which examined the effect of customisation on member states' practical implementation of EU directives.Their analysis shows that different dimensions of customisation have contrasting effects on practical implementation.While higher levels of density (without increased restrictiveness) impede member states' compliance in practice, customised restrictiveness is shown to facilitate practical implementation (Zhelyazkova and Thomann 2022).In other words, regulatory stringency puts pressure on implementers and target groups to make EU policies work (Knill, Schulze, and Tosun 2012;Thomann 2019).Far from watering down original policy intentions, customised restrictiveness reduces the ambiguity of EU rules by clarifying the obligations of national implementers and enforcement institutions.
DPI allows for the adaptation of EU policies to local circumstances and may thereby increase problem-solving capacities at the national level.If member states use discretion to tailor policies to their domestic needs, this might reduce the ambiguity of EU law and thereby strengthen goal attainment.In a recent study, Zhelyakova and Thomann (2022) demonstrated that increased restrictiveness does indeed contribute to the problem-solving capacity of national law.On the other hand, there is no guarantee that member states use DPI for attaining the goals enshrined in European law.As Batory (2016) demonstrated, governments may also use DPI to subvert EU policy-goals, which are not in line with their preferences.In addition, problem-solving at the national level can be at odds with problem-solving at the EU level, in particular if problem perceptions differ (Thomann 2019).In particular, if European cooperation aims to mitigate externalities of national policies or transboundary policy problems, tailoring EU law to domestic circumstances could impede the EU's problem-solving capacity.In this vein, Erkoreka and Blas (2023) find that DPI has proved inefficient in ensuring an equivalent level playing field of customs controls in the EU, with three substantial negative outcomes: economic and budgetary; problem-solving capacity; and output legitimacy.
The special issue moves beyond explanations of patterns of DPI to focus also on the implications of diversity in domestic policy outputs and outcomes.In particular, some contributions provide novel insights about the consequences of DPI for the legitimacy and effectiveness of EU policies.For example, Mancheva, Pihlajamäki, and Keskinen (2023) argue that the institutional variation in similar governance contexts could threaten the accountability and legitimacy of EU policies.Conversely, Zhelyazkova (2022) argues that differentiated transposition could allow governments to follow their own policy preferences in the area of asylum and immigration, especially when national interests diverge from EU objectives.Moreover, DPI could also feed back into EU decision-making processes (or is even anticipated in these processes) and push supranational policy-makers to limit the discretion of EU rules when they anticipate negative externalities from the increased diversity in policy outcomes (see Cabane and Lodge 2023) and could even lead to unified integration in the long run (see Thatcher and Garcia Quesada 2023).

Conclusions and ways ahead
The current special issue aims to advance the research agenda on DPI, offering several key insights and areas for future research.First, the current special issues offers new, deeper evidence about the empirical importance of DPI in the EU.Princen et al. (2022) fill an important gap by providing a comprehensive overview of DPI at the level of anticipated implementation, i.e. discretion as granted to member states.Future research should build on this to analyse the role of different types of discretion for DPI in more detail.However, we still lack a similar comprehensive database of DPI in both the legal and practical implementation context as exists for DI (Duttle et al. 2017;Schimmelfennig and Winzen 2014).
Moreover, the special issue contributions show that DPI has multifaceted links with DI.On the one hand, they add to previous literature showing that DI can lead to DPI, due to member state responsiveness to politicisation and national policy preferences.On the other hand, new findings suggest that DPI also feeds back into DI, both by triggering policy feedback through further policy-making, and by anticipation by regulatory actors.These findings have important implications for studying differentiation in the EU.First, they suggest that DI and DPI should be studied together rather than in isolation-in line with previous calls for linking the uploading and downloading stages of Europeanisation (Börzel 2002;Thomann and Sager 2017;Zgaga and Thomann forthcoming).Second, these results suggest that theories of EU differentiation could learn from policy feedback theory, as a promising avenue for future research (Béland, Campbell, and Weaver 2022).The empirical dynamics linking DI and DPI in the literature intuitively makes sense from a responsiveness perspective, but DPI scholarship should take the next step and theorise them more systematically as an underlying cycle of differentiation in the EU policy process.
Third, the contributions of the special issue add to our knowledge on the drivers of DPI.They confirm the importance of five factors: discretion, alignment with domestic institutions, policies and regulatory styles, responsiveness to domestic (including bureaucratic) politics, functional drivers, and administrative and economic capacity.Moreover, they advance current theorising on the role of discretion, goodness of fit, and civil society in DPI.Particularly, the contributions link DPI to compliance and institutional theories.However, a more comprehensive theoretical and empirical account of the drivers of DPI, which relates to broader theories of EU integration and Europeanisation, is still to be developed (Thomann 2019).In addition, research should focus also on factors systematically hampering DPI (such as crises) to provide comprehensive explanations of variation in DPI.Finally, the special issue contributions add more evidence to existing research that DPI has implications for the accountability and legitimacy of EU governance.It remains an empirical question whether differentiated implementation at the national level is at odds with or beneficial for the problem-solving capacity of the EU.Empirical research needs to explore, however, if there are systematic scope conditions under which DPI increases or decreases problem-solving capacities at the national and EU level.The current special issue presents clear evidence of feedback effects between DPI and other stages of the EU's political and policy processes.This, we hope, will motivate a new generation of researchers to take DPI seriously as an important avenue for differentiation within the EU.

Disclosure statement
No potential conflict of interest was reported by the author(s).