Differentiated participation, uniform procedures: EU agencies in direct policy implementation

Abstract

European Union (EU) institutions have become increasingly involved in direct policy implementation in the member states, creating a new domain of differentiation in EU governance. What brings about such differentiation, and how does it vary across policy fields? Drawing on theories of differentiated integration, this article argues that differentiated implementation occurs at the intersection of postfunctional obstacles (politicisation) and functional pressures for joint implementation (interdependence). There are two identified dimensions of direct implementation, a territorial one referring to states’ participation in such activities, and a procedural one capturing the degree of uniformity in the guidelines for organising implementation. The resulting typology is applied to direct implementation activities (DIAs) conducted by EU agencies alongside national authorities. The qualitative analysis reveals that differentiated participation is a stable feature of DIAs in politicised fields, and although there is a tendency to create more uniform procedures over time and across policy fields, higher uniformity prevails under symmetric interdependence.

Keywords:

• European Union agencies
• differentiated integration
• EU policy implementation
• Europol
• European Asylum Support Office

In European Union (EU) studies, ‘differentiation’ is seen as a response to the widening and deepening of European integration over time (Schimmelfennig et al. 2015). As EU membership and competences expanded, there was a growing need to accommodate diversity and thus allow for flexibility in governance arrangements. Differentiation permits member states to participate at different speeds, in different formats, or not at all, in line with national preferences and capacities (Dyson and Sepos 2010). So far, the literature has identified two types of differentiation. When member states participate selectively in EU legal arrangements, this is understood as ‘differentiated integration’ and constitutes a well-established feature of EU governance in the post-Maastricht era (Holzinger and Schimmelfennig 2012). Conversely, when differentiation applies to the discretion enjoyed by member states under common EU legal arrangements, scholars have recently described the phenomenon as ‘differentiated implementation’ (Zhelyazkova et al. 2023). The crucial innovation of this literature lies in its move beyond compliance studies. Instead of focussing on infringement procedures or the legal transposition of EU rules (Börzel et al. 2012; Duina 1997; Falkner et al. 2005), research on differentiated implementation sets out to investigate the variegated implementation practices across member states (Zhelyazkova 2014; Zhelyazkova et al. 2016).

We contribute to this literature by arguing that it should not only examine the activities of national authorities (Brendler and Thomann 2023; Pollex and Ruffing 2023), but also account for the participation of EU institutions in direct policy implementation. In an increasing number of cases, EU policy implementation occurs with the active and direct involvement of EU institutions (Freudlsperger et al. 2022; Scholten 2017). This is an important development because it goes beyond the traditional paradigm of the EU as a ‘regulatory state’ (Majone 1994), focussed on the production of rules that require direct implementation by the member states. While this paradigm has been changing since the Maastricht Treaty (1993), it received little attention from implementation scholars. Mirroring competences that used to exist solely in competition policy, EU institutions have begun to lead or assist national authorities in direct implementation activities (DIAs) such as inspections of public and private premises, investigations, border patrols, the joint processing of asylum-seekers, returns of migrants, etc (Freudlsperger et al. 2022; Scholten 2017).

By default, the institutional context of DIAs makes them prone to differentiation. In fact, the involvement of EU institutions in direct policy implementation occurred in the context of ‘integration without supranationalisation’ (Bickerton et al. 2015), reflecting the general preference of member states towards institutional arrangements that preserved their control over further integration. A crucial aspect thereof was the gradual empowerment of de novo bodies such as EU agencies (Bickerton et al. 2015: 713). Unlike classic supranational institutions like the European Commission, agencies were designed with a shared governance structure that included both national and supranational actors (Busuioc 2012; Vos et al. 2018). Over time, some of these agencies have acquired a wide range of direct implementation powers. Examples can be found in border management (the European Border and Coast Guard Agency, known as Frontex), asylum (the European Union Agency for Asylum [EUAA], formerly EASO), police cooperation (Europol), or civil and criminal justice (Eurojust and the European Public Prosecutor’s Office [EPPO]). From the perspective of differentiation, the shared governance structure of EU agencies was expected to grant member states continued discretion over future decisions – and by extension create highly differentiated DIAs.

In this article, we explore how the differentiation of DIAs looks in practice. How do DIAs vary across fields and agencies, and what explains this variation? We focus on direct implementation by EU agencies as the most numerous, prominent, and yet understudied activities present in the field of EU direct implementation.¹ To complement studies on implementation outputs (the legal transpositions of EU rules) or outcomes (the practical application of EU policy at the domestic level), we study implementation at the level of throughputs, i.e. the procedures that govern the interactions between EU and national officials in the process of direct policy implementation. Theoretically, we borrow from the literature on differentiated integration to identify conditions for the emergence of differentiation across policy fields. First, we argue that highly politicised fields are prone to territorial differentiation because domestic opposition drives national authorities to opt out (de jure or de facto) of the support of EU institutions. Conversely, in policy areas where national actors are not constrained by domestic public opinion, there is less territorial differentiation in implementation. Second, among participating member states, we explain differentiated implementing procedures by considering the type of functional interdependence characterising different domains. Where interdependence is asymmetric and only some countries face strong functional pressures for EU-level cooperation, differentiated participation in DIAs should be complemented by procedural differentiation.

Empirically, we investigate throughput implementation as laid down in the rules and procedures for multi-level interactions. Specifically, as outlined in the introduction to this special issue (Zhelyazkova et al. 2023), we study varieties of differentiated policy implementation in reference to procedural rules of interaction agreed between national authorities and EU institutions. To this end, we gather original data about the mandates and operating procedures of DIAs involving EU agencies and map them according to the level of politicisation and interdependence of the fields in which they operate. We show how DIAs can be grouped into a four-pronged typology which includes (1) non-differentiation, (2) territorial differentiation, (3) procedural differentiation and (4) full differentiation. Subsequently, we illustrate these configurations of differentiated implementation using three ‘pathway cases’ (Gerring 2007) that allow us to capture different combinations of politicisation and interdependence in DIAs found in maritime safety, police and criminal justice cooperation, and migration and asylum. Our findings are twofold. First, territorially differentiated participation is a stable feature of DIAs in politicised fields. Second, although there is a tendency to create more uniform procedures over time and across policy fields, higher uniformity prevails in areas affecting member states equally (i.e. under symmetric interdependence).

The article proceeds as follows. First, we clarify the term differentiated implementation and explain how it connects to the increasing presence of DIAs performed by EU bodies. Second, we develop a typology of differentiated implementation drawing on the theoretical expectations of the related literature on differentiated integration. Third, we apply the typology to DIAs and the case of EU agencies. Fourth, we present three key examples in different areas of policy making to illustrate the differentiated implementation of DIAs in practice. The conclusion emphasises the importance of the topic and its future relevance in the EU context.

Policy implementation in the EU: the rise of direct implementation activities

For the purposes of this study, we consider two fundamental dimensions of EU policy implementation. First, given the EU context, there are different levels of governance involved in implementation. According to the Treaties, member states are responsible for implementing EU policy (Article 4(3) TEU). By contrast, EU institutions have competences to implement EU rules only in a limited number of fields like competition policy and the administration of the EU budget. At the same time, the infringement procedure allows EU institutions to check and ensure that member states implement EU law (Article 258 TFEU). This brings a second dimension of EU policy implementation to the fore, namely the distinction between direct and indirect policy implementation. Traditionally, member states implemented EU policies on their territories, at times after coordinating in intergovernmental regulatory networks (cf. van Kreij 2022); by comparison, EU institutions were only involved in indirect implementation through the infringement procedure. Legal scholars refer to these activities as ‘enforcement’, which encompasses the powers to monitor, investigate, and sanction compliance with EU law (Scholten 2017: 1350). We prefer the term implementation because we are interested in a broader set of DIAs, for which enforcement powers are limited: for instance, in border management or criminal justice, EU agencies assist member states in monitoring or investigations but do not have enforcement authority on their own.

In the last decade, several EU bodies have acquired the right to engage in DIAs within the member states. In contrast to indirect implementation, DIAs are distinctive because they presuppose that officials from EU institutions are physically present on the ground to carry out EU policy. DIAs are not (or are only partly) conducted from an institution’s headquarters; instead, they comprise ‘street-level’ interactions with national bureaucrats and private actors. Nowadays, there are various examples of EU institutions conducting DIAs (for an overview, see Freudlsperger et al. 2022). In financial supervision, the European Central Bank (ECB) and the European Securities and Markets Authority (ESMA) conduct inspections and investigations on the premises of private companies. In border management, Frontex and the EUAA can participate in border control operations, the processing of asylum applications, or the return of irregular migrants. In criminal justice, Europol, Eurojust, EPPO and OLAF are involved to different degrees in criminal investigations and prosecutions. The extent to which EU actors are involved in DIAs varies across fields. In some areas, there is a hierarchical relationship between EU bodies and national authorities, e.g. the ECB and ESMA. In other areas, EU agencies have a supporting role, meaning that they assist national authorities upon request, e.g. the EUAA (Scholten et al. 2017: 358–360).

Despite their pervasiveness in contemporary EU governance, DIAs have received little attention from the academic literature. Most studies on multi-level policy implementation in the EU focus on implementation outputs, i.e. the legal transposition of EU rules (Börzel et al. 2012; Falkner et al. 2005). This line of research seeks to explain member states’ differing performance in implementation, either by exceeding EU-prescribed standards (Fink and Ruffing 2017; Zaun 2018) or by lowering them down (Börzel et al. 2012; Zhelyazkova et al. 2016). Conversely, the more recent literature on (differentiated) multi-level implementation is interested in implementation outcomes, i.e. the practical application of EU rules on the national level and the discretion of member states to adapt the rules to their specific situation. In this context, it is widely accepted that EU legislation ‘is hardly implemented uniformly across member states’ (Andersen and Sitter 2006: 318), with practical deviations from rules being the norm rather than the exception (Zhelyazkova et al. 2016).

In this article, we widen the scope of existing research in two ways. First, we are interested in the drivers and patterns of differentiation in direct policy implementation. To this end, we adapt theories devised to explain differentiated integration in the EU and develop specific expectations for the occurrence and form of differentiated implementation across different policy fields. Second, we go beyond outputs and outcomes and focus on the throughput side of differentiated policy implementation. Throughputs are important because they refer to the process through which actors – in this case EU and national officials – conduct direct policy implementation in the EU. When throughput implementation is differentiated, this means that there is no one-size-fits-all approach to the interactions between the national and EU level in DIAs. Differentiation is an explicit expectation of the literature on ‘integration without supranationalisation’ (Bickerton et al. 2015): after all, DIAs allow member states to stay in control of further integration by exercising discretion on a case-by-case basis. The validity of this expectation is an empirical question that we explore in the article. Specifically, we investigate throughputs by studying and explaining the rules of procedures devised to govern multi-level interactions.

Drivers of differentiation and uniformity in direct policy implementation

Why does differentiation occur in direct policy implementation? To answer this question, we adapt theories explaining differentiated integration (DI) to the context of direct policy implementation. While the two phenomena are distinct, they capture variations of the same feature – namely the flexibility for member states to participate in different ways in EU governance. Moreover, the literature on DI is already advanced theoretically and can help inform research on the causes of differentiated policy implementation. For instance, Schimmelfennig et al. (2015) provide a general theory of differentiation by combining ideas from key strands of European integration theories. According to them, differentiation (or uniformity) is the result of the interaction between specific drivers of and obstacles to European integration. In line with the classic functionalist perspective, transnational interdependence is considered the main driver of integration. To put it simply, member states decide to participate in EU policy areas either to avoid negative externalities arising from unilateral action or to reap the economies of scale deriving from joint action. Moreover, interdependence can be symmetric, i.e. affecting all member states equally, or asymmetric, i.e. affecting a subset of member states significantly more than others. Conversely, following a postfunctionalist argument, the principal obstacle to European integration stems from politicisation. When an issue is highly politicised within EU member states, the result tends to be horizontal (or ‘territorial’) differentiation. Notwithstanding the functional benefits that states could gain from further integrating, they may decide to opt out of policy areas in case of ‘mass-level salience and contestation of European integration, the mobilisation of Eurosceptic public opinion by Eurosceptic parties, and opportunities to voice Eurosceptic opinions in national referendums’ (Schimmelfennig et al. 2015: 771). Politicisation tends to occur in areas that touch upon national sovereignty, such as foreign and security policy, taxation, monetary policy, or migration and police cooperation (Schimmelfennig et al. 2015; Schimmelfennig and Winzen 2020). In other words, the stronger a country’s attachment to national sovereignty, the more politicised European integration is domestically, and therefore the higher the demand for differentiation. Due to the lack of a European public sphere, issue-specific politicisation at the national level tends to have an asymmetric nature, affecting some countries whereas remaining absent in others.

In our view, the dynamics between functional drivers and postfunctional obstacles to European integration can also be applied to the domain of differentiated implementation. According to previous studies, higher national discretion in implementation is affected by several factors, especially the need for unanimity in the Council (Franchino 2007) and high public salience of an issue (Calvert et al. 1989). Both unanimity and salience tend to be features of sovereignty-sensitive areas. It is due to the sensitivity of these domains that EU laws often constitute ‘watered-down compromises with weak adjustment pressure on administrative practice’ (Zhelyazkova et al. 2016: 838). Politicisation thus regularly results in territorial differentiation, as states are keener to secure the possibility to refrain from participating in the implementation of specific policies.

Beyond the territorial differentiation of participation, the features of a given policy area also offer different incentives for the ensuing uniformity of implementation practices. We argue that such incentives are determined by the type and degree of sectoral interdependence. Functional pressures push towards the adoption of uniform implementation procedures to regulate member state behaviour and the interactions between EU and national actors. Procedural uniformity is functionally necessary in domains in which states want to avoid negative externalities from other states’ implementing actions or to reap the benefits of intergovernmental implementation efforts among groups of states. Furthermore, we distinguish two types of interdependence. Under asymmetric interdependence, only some member states face a functional necessity to implement a specific EU policy. For example, migration flows can affect member states differently (Thielemann 2018). In turn, under symmetric interdependence, all member states are equally compelled to implement a policy. For example, counter-terrorism intelligence is required by states to similar degrees, to the point that even a state with a rigid legal opt-out from police cooperation like Denmark wants to partake in these initiatives (Migliorati 2021). In sum, we consider asymmetric interdependence to work as a driver of procedural differentiation, whereas symmetric interdependence will push member states towards procedural uniformity.

We expect the interplay between politicisation and interdependence to unfold sequentially in time. Politicisation affects primarily the ability of individual countries to participate in policy implementation in the first place. Its influence is thus particularly visible at the time of initial delegation. Politicisation can also occur when a novel implementation instrument (e.g. an agency or a specific DIA) is first created. Conversely, symmetric interdependence creates incentives to create uniform procedures for policy implementation. Its effect, however, is likely to be drawn out over time. This expectation is in line with research on life cycles of interorganisational relations, which argues that a gradual standardisation and codification of rules for interactions is a typical reaction to problems of trust and commitment that arise in practical cooperation (Borry et al. 2018; Ring and van de Ven 1994). In the context of multi-level implementation, standardisation refers to the process leading to uniform procedures (Brunsson and Jacobsson 2002) and has two primary features. First, actors devise detailed rules of procedure, specifying who does what in direct policy implementation. Second, actors specify who contributes what, by defining their individual inputs in terms of equipment, personnel, and money (Dijkstra 2017). In sum, we expect symmetric interdependence to act as a driver for procedural uniformity over time.

Based on our expectations concerning the interplay of politicisation and interdependence, we construct four ideal-typical configurations of territorial and procedural differentiation in implementation. These are summarised in Table 1.

Table 1. Interdependence and politicisation in throughput implementation.

		Interdependence
		Asymmetric	Symmetric
Politicisation	High	Full differentiation Differentiated participation, differentiated procedures	Territorial differentiation Differentiated participation, uniform procedures
	Low	Procedural differentiation Uniform participation, differentiated procedures	Non-differentiation Uniform participation, uniform procedures

The four configurations create a typology of differentiation in throughput implementation, which has the following characteristics:

1. Non-differentiation: If a policy area is characterised by symmetric interdependence and low politicisation, all states will participate in policy implementation through uniform procedures.

2. Territorial differentiation: If a policy area is characterised by symmetric interdependence and high politicisation, policy implementation will be territorially differentiated but procedurally uniform.

3. Full differentiation: If a policy area is characterised by high asymmetric interdependence and high politicisation, implementation will be differentiated in terms of both participation and procedures.

4. Procedural differentiation: If a policy area is characterised by asymmetric interdependence and high politicisation, implementation will be procedurally differentiated but territorially uniform.

In the next section, we apply this typology of differentiation to the case of direct policy implementation carried out by means of DIAs. Subsequently, we delve further into the specificities of individual cases of differentiated implementation.

Mapping the differentiated implementation of DIAs

Our mapping exercise focuses on EU agencies because of their comparability to each other (as opposed to the Commission, for instance) as well as owing to their pervasiveness and lack of scholarly attention when it comes to DIAs. So far, the literature on EU agencies has prioritised issues of institutional design, accountability, or independence from the Commission and the member states (Busuioc 2012; Egeberg et al. 2015; Keleman 2002). By comparison, there is less research on how agencies operate in practice (Busuioc 2016; Groenleer et al. 2010) and participate in DIAs. Accordingly, we start by screening agencies’ mandates and selecting only those tasked with the direct implementation of EU policies through a physical presence on the ground in the member states. For the purposes of this article, we exclude agencies which cover some operational tasks but do not participate in DIAs, such as the European Union Agency for Law Enforcement Training (CEPOL) or the European Union Agency for the Operational Management of Large-Scale IT Systems (eu-LISA). In a nutshell, we only consider EU agencies that participate in direct policy implementation jointly with member states and have tasks in supporting the physical enforcement of compliance with EU rules by both public and private actors.

To group EU agencies along the two dimensions of politicisation and interdependence, we follow existing research and identify the inherent features of the policy areas under focus. In line with our theoretical framework, we consider two dimensions of differentiated policy implementation: (1) the involvement of all or only some member states in DIAs (uniform vs differentiated participation), and (2) the existence (or not) of standardised rules or guidelines for the organisation of DIAs (uniform vs differentiated procedures). The first dimension is consistent with classic DI theory and the idea that member states can opt in or out of specific policy areas, thus creating horizontal differentiation (Schimmelfennig et al. 2015). The second dimension, i.e. the existence of uniform procedures for the organisation of DIAs, is specific to direct policy implementation by EU agencies and the study of throughputs as intended in this article. This refers to the extent to which DIAs follow a set of formal, standardised rules of procedure agreed by member states in cooperation with EU agencies. For our purposes, the presence or absence of such a written ‘organisation manual’ marks the difference between uniform or differentiated procedures for DIAs. While it is conceivable to measure throughputs in a more fine-grained manner (e.g. by examining practical processes of implementation), we consider the presence of standardised guidelines as an initial, necessary step towards regulating the interactions between the national and the EU level.

To this end, we investigate the presence of differentiation in the design of DIAs as well as agencies’ rules of procedures for direct policy implementation. We check if agencies’ original legal frameworks include provisions on opt-outs and rules of procedure for carrying out DIAs, and whether this type of ‘hard law’ was amended over time. Moreover, we examine whether other kinds of rules were established outside of the main founding act through ‘soft law’ (recommendations, guidelines, best practices) or ad-hoc agreements (operating plans adopted on a case-by-case basis between an agency and a member state or a group of member states). The distinction between hard law and soft law is well-known in the EU setting and refers to the difference between binding and non-binding rules adopted at the EU level (Trubek and Trubek 2005). In this case, Commission implementing acts fall under the category of hard law, while soft law comprises Council recommendations or ‘best practices’ adopted by intergovernmental networks.

Table 2 plots the institutional configurations of joint implementation that we find among the nine operational agencies under investigation.

Table 2. Participation in and procedures for DIAs in nine EU agencies.

		Procedures for DIAs
		Differentiated	Uniform
Participation in DIAs	Differentiated	Full differentiation EUAA, Frontex	Territorial differentiation Europol/Eurojust, EPPO
	Uniform	Procedural differentiation n/a	Non-differentiation EASA, EFCA, EMSA, ESMA, ELA

We start the discussion with the category ‘non-differentiation’, which includes DIAs conducted with the participation of all member states and following uniforms rules of procedure. By and large, this category applies to agencies working in regulatory fields and responsible for the enforcement of internal market rules. Historically, such fields are characterised by symmetric interdependences deriving from the expansion of the internal market (Schimmelfennig et al. 2015). Moreover, they deal with policy areas characterised, on average, by low salience and low politicisation among national publics. For instance, the European Aviation Safety Authority (EASA) carries out inspections on aircrafts and airport facilities in the member states. The European Fisheries Control Agency (EFCA) checks compliance with EU legislation by inspecting fishing vessels. The European Maritime Safety Authority (EMSA) does the same in relation to, among others, port reception facilities, marine equipment, and passengers’ registrations on ships (Groenleer et al. 2010). Furthermore, in financial supervision, ESMA examines records and documentation, summons persons and conducts interviews, and inspects credit rating agencies or trade and securitisation repositories (Joosen and Zhelyazkova 2022). The recently established European Labour Authority (ELA) coordinates member states, upon request, in concerted and joint inspections to monitor compliance with EU labour standards (European Labour Authority 2020). What these agencies have in common is the ability to conduct direct inspections on all member states’ territories, report on their findings, and eventually take actions in case of breaches of EU law in all member states with no exception (if not directly then via the European Commission, see Scholten 2017: 1351). In terms of procedures, inspections are mandated and defined in ‘hard law’ (regulations, directives, decisions, and Commission implementing acts) and ‘soft law’ (internal guidelines of agencies).

Next, there is the category of ‘territorial differentiation’, which includes DIAs conducted by some (but not all member states) following uniform rules of procedure. This category applies to agencies operating in the fields of judicial and police cooperation in criminal matters. These fields are central to state sovereignty, particularly the monopoly over legitimate coercion in a territory (Weber 1946). Their integration is hence subject to sporadic politicisation at the national level (Liedlbauer 2021). A classic example are the Danish referendums on participating in these fields (Jacobsen 2015). Simultaneously, police and criminal justice cooperation are characterised by symmetric interdependence. In the absence of internal borders in the Schengen area, organised crime and terrorism affect all member states to a certain extent. Consequently, national authorities have a functional interest in the exchange of information and expertise and the joint investigation of cross-border crimes (Lemieux 2013). Since the mid-2000s, Eurojust and Europol support the organisation of Joint Investigation Teams (JITs) in criminal matters (Horvatits and de Buck 2007) in a highly intergovernmental fashion, allowing member states to participate on a case-by-case basis. In terms of rules of procedures, Europol and Eurojust use soft law. Elsewhere, the EPPO can conduct criminal investigations and act directly as the prosecuting authority before national criminal courts (Claes et al. 2021). Apart from Denmark and Ireland, which have a general opt-out from EU justice and home affairs), three other member states opted out of EPPO: Hungary, Poland, and Sweden (Council Regulation (EU) 2017/1939).

Finally, there is the category of ‘full differentiation’, which includes DIAs conducted in a few member states following mission-specific rules of procedures. Here we can find the agencies dealing with the processing of asylum applications (EUAA, formerly EASO) and border management (Frontex). These policy fields tend to be highly politicised (Grande et al. 2019) and are characterised by asymmetric interdependence, as states are affected differently by asylum applications and migration influxes (Thielemann 2018). The EUAA assists member states in the processing of asylum application (Tsourdi 2016), while Frontex can organise, together with the member states, joint return operations, rapid border interventions, and DIAs at sea, air, and land (Frontex 2022). These agencies are characterised by opt-outs (for instance, the United Kingdom – while still a member – and Ireland were not part of Frontex; Denmark is not part of EUAA); moreover, they can intervene on the territory of member states only upon invitation from national authorities who require operational support in various ways. This means that the two agencies must conclude ad-hoc agreements with individual member states on a case-by-case basis, in line with the needs of national authorities. Internal rules of procedure can be found in soft law (internal guidelines and codes of conduct) and are further specified in ad-hoc agreements depending on the mission to be pursued.

We currently find no cases of DIAs in the category ‘procedural differentiation’, which would be characterised by asymmetric interdependence and low politicisation. We can only speculate about the reasons for this absence. Asymmetric interdependence assumes that only some member states are affected by an issue and push for joint implementation instruments. However, if politicisation in these countries is low, there is, at least initially, no reason for procedural differentiation. Over time, ‘procedural differentiation’ could be populated by cases moving from other categories, as a case of full differentiation might become less politicised, or a case of non-differentiation might become subject to asymmetric pressures. Moreover, due to the novelty of the empirical phenomenon of joint implementation, we may have not yet been able to observe such sectoral dynamics.

In order to illustrate the typology of differentiated policy implementation in practice, in the next section we discuss three pathway cases of agencies which conduct DIAs and can be found in one of the categories of Table 2. Methodologically, pathway cases represent variations of crucial cases where the factors of interest and the outcome are already known and the goal is to elucidate causal mechanisms (Gerring 2007: 238). Since the mapping conducted above has already identified the interplay of two factors of interest (interdependence and politicisation) as well as the outcome (differentiated participation or procedures), the idea is to illustrate the process through which (non-)differentiation occurs. The cases are based on internal documents from EU agencies, secondary legislation, one interview conducted with an agency official (see section on EMSA), and previous studies in the field.

Patterns of differentiated implementation in three DIAs

Non-differentiation: inspections of the European Maritime Safety Agency (EMSA)

We first investigate DIAs consisting of inspections on member states’ territory to enforce the EU maritime acquis. The key agency here is EMSA, which clearly conducts its activities in an area of low politicisation. Indeed, the regulation and enforcement of pollution and safety standards in EU waters constitute a low-salience issue that is rarely, if ever, politicised at the national level (Pallis 2006). Simultaneously, EMSA operates in an area of symmetric interdependence, as the safety and pollution standards that it investigates apply to both maritime and landlocked states. EMSA inspections are a tool of incident prevention, and they aim to hold back any hazard that could occur in both maritime and river traffic. Hazards that apply to member states with seashores only, e.g. oil spills, are merely a subset of the risks that EMSA seeks to mitigate. Since all members are equally affected by at least some of the security hazards that EMSA investigates in its inspections, we regard it as operating in an environment of symmetric interdependence. Against this background, we find that EMSA is marked by a pattern of uniform participation and increasingly uniform implementing procedures.

Established in 2002, EMSA is responsible for ensuring a ‘high, uniform, and effective level of maritime safety and prevention of pollution by ships’ within the EU (Art 1, Regulation 1406/2002). EMSA’s involvement in DIAs revolves around inspections designed to monitor the port state control regime, assess the activities of so-called ‘classification societies’, and check the work of notified national bodies (Groenleer et al. 2010: 1219). During inspections, EMSA acts on behalf of the European Commission, verifying whether member states comply with the EU maritime acquis (EMSA 2022: 20). When EMSA discovers a lack of compliance or enforcement problems, it reports it to the Commission, which can then initiate infringement procedures. In the performance of inspections, national authorities play an auxiliary role: their job is to allow EMSA officials into the facilities to be inspected and provide relevant information. The possibility of infringement procedures puts pressure on national authorities, who are well aware of the connection between EMSA inspections and potential sanctions for non-compliance (Heims 2016: 889). Since 2007, EMSA has carried out 299 inspections of twelve different types, in line with the different standards of the maritime acquis to be checked (EMSA Document Request, 26 November 2021). On average, missions consist of 2 to 3 inspectors operating on member states’ territory for 4 to 6 days, depending on the number of facilities to be checked and the geographical area under consideration (interview with EMSA official, 26 November 2021).

Initially, there were no clear procedures for the organisation of inspections. Their standardisation occurred gradually via ‘soft law’, namely through internal guidelines. The first rules were approved by the Administrative Board in 2004. While the agency aimed for consistency in the organisation of inspections, the official policy remained generic, seeking ‘to address all general circumstances and situations where the agency may have to carry out visits to the member states in order to execute its tasks’ (EMSA 2004: 1). The document provides a template for the organisation of inspections in respect to the parties who should receive information about the planned visits, the nature of the inspections and the people who will carry it out, the duration, reporting requirements, confidentiality provisions, travel and accommodation, as well as overall costs (EMSA 2004: 3–4). This minimal workflow for inspections was updated and expanded in 2015. The new document introduces the concept of ‘visit cycles’, according to which all member states shall be visited within each cycle. Moreover, the rules specify that ‘the first cycle of visits should focus on the implementation of the relevant Union law in general while any subsequent cycles in the same area can concentrate on specific elements as highlighted by the horizontal analysis’ (EMSA 2015: 3).

Furthermore, the 2015 methodology specifies the role of the Commission as well as pre-visit steps to be taken by the agency (EMSA 2015). Importantly, the rules set uniform constraints on the member states by specifying that, ahead of country visits, national authorities should provide EMSA with any amendments to the national legislation or organisation and with any other documents that might be relevant to help EMSA prepare adequately for the visit and reduce the inspection to a minimum duration. Furthermore, EMSA may also send a specific pre-visit questionnaire to the member state concerning the areas concerned (p. 4). Moreover, the 2015 document includes guidelines for the execution of inspections and the post-inspection phase to assist member states in better implementation and capacity building.

The two documents show the gradual procedural standardisation in the execution of EMSA inspections with the support of national authorities. Coupled with the uniform cooperation of member states in EMSA inspections, the example makes for a clear case of ‘non-differentiation’ in direct EU policy implementation. This pattern is not present, however, in other policy fields, as shown in the next pages.

Territorial differentiation: Eurojust and Europol Joint Investigation Teams (JITs)

This section addresses a case of territorial differentiation, namely the JITs supported by Europol and Eurojust. JITs bring together police officers, prosecutors and judges in two or more member states for the purpose of conducting criminal investigations (Eurojust 2021a). The European integration of criminal justice cooperation is a sporadically politicised domain as it deals with a core feature of national sovereignty. From the Maastricht to the Lisbon Treaty, the area developed in an intergovernmental fashion, allowing member states to retain the lead of direct policy implementation within their territories. We also observe some formal territorial differentiation, with both Denmark and the UK (when still an EU member) enjoying an opt-out from the activities of Europol (Peers 2011). Additionally, figures provided by Eurojust show vastly varying degrees in individual member states’ participation in JITs (Eurojust 2021c). Procedurally, JITs have become increasingly uniform over time given the symmetric interdependence that characterises their implementation and the need to fight cross-border crime in all member states. Moreover, the need for procedural uniformity is a matter of functional necessity to safeguard the efficiency of this collaboration tool, as participants to JITs alternate on a case-by-case basis but their interactions – especially with EU agencies – follow a repeated pattern. Below we show how the gradual standardisation of implementation procedures came about.

JITs were established in 2000 as an intergovernmental initiative adopted by the Council in the framework of the EU’s third pillar.² In 2003, the JHA Council adopted a recommendation on a model agreement for setting up a JIT (Council Recommendation 2003/C 121/01). The agreement standardised the organisation of JITs by providing a template for the provisions that had to be agreed to launch such an operation (purpose, geographical scope, chain of command, range of activities, logistics, etc.). Despite their codification in EU soft law, JITs were barely used in the first years, also because member states lagged behind the transposition of Council decisions on JITs into national legislation (European Commission 2005). In response, the Council created a network of national experts on JITs and additionally involved Eurojust, Europol and OLAF in supporting operations. At the first meeting of the JITs Network on 23 November 2005, national experts emphasised in capital letters, using exclamation marks, that their interactions should be ‘AS INFORMAL AS POSSIBLE!’ and ‘AS UNBUREAUCRATIC AS POSSIBLE’ (General Secretariat of the Council 2005). Soon afterwards, Eurojust and Europol became involved in JITs and drafted a ‘Guide to EU member states’ legislation on JITs’ (Horvatits and de Buck 2007). The guide was supplemented by a manual in 2009, which provided additional guidelines on how to structure a JIT, which activities were possible under JITs, what support could be obtained from Europol and Eurojust (Europol/Eurojust 2009). The manual was updated periodically, with the latest available version dating from 2017 (General Secretariat of the Council 2017). The standardisation of procedures thus occurred over time, providing clear templates for national authorities to follow when proposing a JIT.

Moreover, the network of national experts also became more formalised over time: since 2011, the JITs Network Secretariat has been hosted by Eurojust; starting 2013, the Secretariat is also responsible for the management of Eurojust’s JITs funding programme. In addition to the JIT Practical Guide, the JITs Network Secretariat administers a restricted web platform that provides information on national legislation, how to draft JIT agreements, and what legal and practical difficulties can be expected when establishing and managing a JIT (Eurojust 2021b). At the same time, the JITs Network evaluates the activities of JITs to identify recurrent challenges and best practices (JITs Network/Eurojust 2020). The centrality of the JITs Network – as opposed to Eurojust officials – shows the remnants of intergovernmental institutionalisation, where cooperation is horizontal and dependent on the member states.

Standardisation occurred against the backdrop of a significant increase in the usage of JITs. According to data obtained from Eurojust, the agency supported financially a total of 708 JITs between 1 December 2009 and 30 September 2021 (Eurojust 2021c). The types of activities are diverse, including coordination/operational meetings, hearings/interviews, wiretappings, house searches, seizures, arrests, and general exchange of evidence. On average, a JIT comprises 25 people including the JIT leaders.

Overall, the example of JITs shows that while member states ensure the endurance of differentiated participation in politically sensitive areas, they also seek a certain degree of uniformity to make decentralised units collaborate in direct implementation in the first place, and to render that repeated collaboration more efficient. In such instances, we observe a recurrent pattern of territorial differentiation. Other DIAs follow a pattern of full differentiation, as discussed in the next section.

Full differentiation: EASO and the hotspot approach

The third and final example of differentiated implementation looks at a case of full differentiation, based on the case of DIAs involving the European Asylum Support Office (EASO).³ The agency is charged with supporting the implementation of the Common European Asylum System, a highly politicised domain and a bone of contention in virtually all European mass publics. Denmark, again, has secured an opt-out from the agency’s implementation activities. Other member states have repeatedly refrained from requesting and relying on EASO support when faced with mass inflows of refugees or crises at their external borders, among others Germany in 2015–16 or Poland in 2021. In addition to this pattern of territorial differentiation, EASO’s operational activities are also marked by a pattern of asymmetric interdependence. In general, migration inflows are unevenly distributed and affect mostly Schengen frontline states, while administrative capacities are more developed in some member states than in others (Zaun 2018). Consequently, the operational support offered by EASO tends to be provided in a case-by-case and ad-hoc manner, undermining the Commission’s and EASO’s various attempts at a cross-case standardisation of implementation procedures.

EASO was first established in the early 2010s and was originally envisaged as a decentralised tool of multilevel implementation, much like the JITs. According to Article 8 of Regulation 439/2010 (henceforth ‘EASO Founding Regulation’), the agency could provide limited operational support to member states in exceptional circumstances that put pressure on their asylum systems and reception facilities. Only if national authorities requested the help of EASO, the agency could send to the member state in question an ‘Asylum Support Team’ (AST), which included both interpreters and experts able to identify vulnerable groups among asylum-seekers (Comte 2010: 401). Before ASTs could intervene, the requesting member state and EASO had to agree on a detailed ‘operating plan’ comprising information about the situation on the ground, the mission’s details, and the responsibilities of EASO officers (Article 18, EASO Founding Regulation). The official website of the agency lists all operating plans concluded from 2011 onwards (EUAA 2022a). In the first years, the format of operating plans varied from country to country, but they all included the main headings mentioned in the Founding Regulation.

In the next years, an expansion of the practical scope of cooperation between EASO and member states occurred. The increased inflow of asylum-seekers at the EU’s external borders created a tension between EASO’s original mandate and the role of ASTs. Formally, EASO had competences to assist in the processing of asylum applications but not to take executive decisions at different stages of the administrative procedures (Tsourdi 2016). However, some operating plans, such as those for Bulgaria (2013), Italy (2015), Greece (2016) or Cyprus (2017), allowed EASO officials to register asylum applications, prepare files, conduct screening interviews, or draft decisions of first instance (Fernández-Rojo 2021: 76).

In the context of the 2015–16 refugee crisis, the EU established ‘hotspots’ for the registration of migrants and the joint processing of asylum applications by national authorities and EASO. The ‘hotspot approach’ was meant to help countries on the frontline of the refugee crisis ‘to address the migratory pressure in a swifter and more effective manner’ (European Commission 2015: 6). In practice, the tasks of ASTs and the number of people deployed increased, so much so that in 2018, EASO officers participated in the registration of 22,821 applications lodged in Italy, representing 41 percent of total cases that year (Fernández-Rojo 2021: 76). Within hotspots, EASO did more than ‘assist’ national authorities, as agency officials conducted screening interviews with asylum applicants and gave legal opinions on their admissibility and eligibility for the procedure (European Commission 2015). Critics have argued that such direct involvement in asylum procedures partly violates the agency’s mandate, which stipulates that it ‘shall have no direct or indirect powers in relation to the taking of decisions by member states’ asylum authorities on individual applications for international protection’ (EASO Founding Regulation, Preamble 14; for this argument, see European Center for Constitutional and Human Rights 2019).

The details of the hotspots approach were agreed through bilateral agreements between the member states in question and the European Commission. The initial idea was proposed in a Commission Communication in May 2015 and detailed in an ‘explanatory note’ sent by Commissioner Avramopoulos to Justice and Home Affairs Ministers in July (Neville et al. 2016: 26). The second document outlines how coordination on the ground was to take place and what kinds of operational support could be provided (European Commission 2015). Soon afterwards, several hotspots were established in Italy and Greece. The Italian Interior Ministry issued its own version of Standard Operating Procedures (SOPs) for hotspots, which served as a model for Greece as well (Ministry of the Interior 2016).

Over time, the repeated attempts at cross-case procedural standardisation have remained limited and non-binding. In 2016, EASO adopted an internal document that further standardised the agency-internal procedures for operational support and the role of the EASO coordinator (Fernández-Rojo 2021: 74). In 2017, the Commission proposed ‘best practices on the implementation of the hotspot approach’, which included a template for SOPs to be submitted as early as possible by the host member states, specifying the workflows to be applied and the tasks of different actors thereof (European Commission 2017). The document attempted to turn the former ad-hoc agreement for the Italian hotspots into a standardised model to be used by other member states. In its recent missions in Spain (2020), Lithuania (2021), Latvia (2021), Belgium (2021) and Romania (2022), however, the agency again relied on Operational Plans that were negotiated bilaterally and ad hoc with the respective member states. In sum, the original dynamic of territorial differentiation in DIAs involving EASO has been complemented by a differentiation of implementation procedures on a case-by-case basis, resulting in a pattern of full differentiation.

Discussion

The cases presented above deal with policy areas characterised by different levels of politicisation and interdependence. As expected in our theoretical framework, low levels of politicisation combined with symmetric interdependence are associated with uniform participation in DIAs and uniform procedures that apply to all member states. EMSA’s inspections provide a case in point, despite differences in the applicable laws (e.g. certain maritime legislation does not affect landlocked countries, while river safety rules do). In turn, high politicisation paired with symmetric interdependence, as in the case of police cooperation via JITs, produce differentiated participation and uniform procedures. Finally, a constellation of high politicisation and asymmetric interdependence brings about fully differentiated participation and procedures. In terms of causal mechanisms, all cases revealed a push towards gradual standardisation over time – either through ‘hard law’ or ‘soft law’ – lobbied by EU institutions involved in DIAs. While this is not always possible because of politicisation and asymmetric interdependence, the tendency of EU agencies towards uniformity in DIAs is clear.

Furthermore, the three cases provided a plausibility probe for our argument, illustrating a pattern that may also apply to the other agencies and institutions performing DIAs. For example, we can reasonably speculate that EASA and EFCA, two agencies that operate in regulatory fields characterised by low politicisation and symmetric interdependence, have developed uniform procedures over time. Simultaneously, we can extend our argument to future events. Where the political environment of a DIA changes either in terms of the type of interdependence or the level of politicisation, we may expect procedural transformation over time. For instance, if an agency that currently conducts non-differentiated DIAs acquires new direct implementation powers in politicised areas, some states may ask for territorial differentiation. As we are dealing with a relatively novel phenomenon, such gradual changes require long-term observations which are not yet available. At the same time, there is also a path-dependent component that we should not underestimate. Once a policy field is uniform and competences are transferred to the EU level, it is harder for member states to reclaim discretion, while ‘opting in’ a differentiated field seems the more likely option once politicisation subsides.

Conclusion

In this article, we expanded the purview of differentiated implementation studies to a new phenomenon in EU governance, namely direct implementation activities exercised by EU institutions together with national authorities. We argued that such activities have the potential to produce differentiated implementation because they have been institutionalised in a context of ‘integration without supranationalisation’ (Bickerton et al. 2015) in which member states sought to maintain discretion over future EU initiatives. Based on the literature on differentiated integration, we showed that postfunctional politicisation drives territorial differentiation in multi-level implementation. In turn, asymmetric patterns of functional interdependence create differentiation in implementation procedures. Our empirical evidence broadly supported these theoretical expectations. Interestingly, in almost all areas of DIAs, increasingly uniform implementation procedures are employed. This procedural standardisation is designed to overcome functional obstacles to vertical (between EU institutions and member states) and horizontal (between member states) cooperation. Only in areas of asymmetric interdependence, especially in the realm of border and migration policy, the drive towards procedural uniformity is countered by a general reliance on bilateral ad-hoc agreements between the respective member states and the EU-level. The implications of this empirical pattern are ambivalent. On the one hand, the enduring differentiation of implementation practices in politicised domains might relieve the pressure of the presence of EU actors on member states’ territory and offer the chance of adapting practices to local circumstances. On the other hand, given the inherent potential for dysfunctionality in joint implementation (Freudlsperger et al. 2022), exceedingly differentiated procedures may represent an obstacle to building a more viable system of multi-level implementation in the EU.

There are several ways in which future studies can engage with our findings. First, since we only examined EU agencies, it would be relevant to know whether the differentiation dynamics identified in this article also apply to other EU institutions such as the ECB in banking supervision. Additionally, since our study investigated only three types of DIAs, it would be interesting to extend the framework to other agencies involved in similar activities. Finally, one could shift the empirical study from throughputs (as laid down in the rules and procedures for multi-level interaction) to outcomes of direct policy implementation by using, for instance, fieldwork and in-depth interviews.

Overall, we believe there is great potential for further research on this topic. As recent developments show, DIAs have become an established asset in the EU’s toolbox to deal with new challenges and especially crisis situations. Two examples from the ongoing Ukraine crisis serve to illustrate this point. In March 2022, the EUAA signed an operational plan with Romania to deploy 120 officers to help with the registering and accommodating of displaced persons who had fled Ukraine (EUAA 2022b). Next, in April 2022, Europol, Eurojust and Frontex, jointly with the member states, launched ‘Operation Oscar’ to carry out investigations and seizures of assets owned by individuals and legal entities sanctioned by the EU in relation to the Russian invasion of Ukraine (Europol 2022). As these examples show, DIAs are here to stay. This makes it even more pressing for scholars to understand their functioning, efficiency and evolution over time.

Acknowledgements

We are grateful for suggestions received on earlier drafts of the manuscript from Frank Schimmelfennig, Tina Freyburg, Bernard Steunenberg, the special issue editors, as well as participants in dedicated panels at EUSA and ECPR SGEU 2022. Special thanks to the two anonymous reviewers for their helpful comments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

Adina Akbik

Adina Akbik is Assistant Professor of European Politics at Leiden University. Her research focuses on political accountability in EU economic governance, the enforcement powers of EU agencies, institutional roles in EU migration and home affairs, and European integration theory more generally. [[email protected]]

Christian Freudlsperger

Christian Freudlsperger is a postdoctoral researcher at ETH Zurich. His research interests lie at the intersection of European integration, comparative federalism, and political development. His book Trade Policy in Multilevel Government: Organising Openness was published with OUP in 2020. [[email protected]]

Marta Migliorati

Marta Migliorati is Lecturer of European Studies at the University of Malta. Her research focuses on EU institutional development and integration processes, legislative-executive relationships in EU multi-level governance and EU agencies. [[email protected]]

Notes

1 However, DIAs are not an exclusive feature of EU agencies, as institutions like the European Central Bank (ECB) and the European Anti-Fraud Office (OLAF) are also involved in this kind of activities (Scholten 2017).

2 The legal basis for JITs is a 2000 Convention on Mutual Assistance in Criminal Matters and a 2002 Framework Decision (2002/465/JHA) – both adopted by the Council of the EU.

3 As of January 2022, EASO became the European Union Agency for Asylum (EUAA).