Differentiation and De-Differentiation in EU Border Controls, Asylum and Police Cooperation

ABSTRACT The leading policy objective in EU differentiation underlying border controls, asylum and police cooperation has been to achieve the abolition of internal border controls to create a borderless European single market. Germany has been the main proponent kickstarting and maintaining this agenda through differentiation. For roughly two decades, differentiation has proved effective in abolishing internal border controls, integrating the related cooperation into EU structures, enlisting the cooperation of non-EU member states and producing joint policy outputs on asylum, external borders and police affairs. Yet, growing external migration challenges have undermined the effectiveness and legitimacy of existing arrangements, ushering in disintegration tendencies.

The article makes four contributions. First, regarding the triggers of DI, we shift the perspective from the opponents to the proponents of European solutions and emphasise the driving role of Germany as a weak Euro-centred hegemon promoting differentiation to achieve integration. We define "Euro-centred" as seeking common standards at the European level. This shift departs from common explanations accounting for differentiation through the role of "comparatively Eurosceptic countries opting out quasipermanently from the deepening of integration in areas of high politics" (Schimmelfennig and Winzen 2014, 368; see also Schimmelfennig and Winzen 2020, 6-7).
Second, we clarify how these triggers of differentiation have left their marks on the governance structures of JHA. Compromises on the rules, forms of commitments, degree of participation and the strategic outreach towards non-EU states were essential to develop this cooperation, allow the centripetal effects to play out and progressively include partners that would otherwise favour different forms of international organisation. After reaching a critical number of participants, European institutions could take responsibility for the arrangement, gradually expanding and harmonising the scheme.
Third, has DI performed effectively in terms of policy output, outcome and impact (Lavenex and Križić 2019, 12f.)? We find that in the medium term differentiation has overcome, step by step, important obstacles to the abolition of internal border controls and has been resilient enough to result in "de-differentiation", which we define as the gradual homogenisation of rules and legal commitments and the extension of participation to all EU members. In the longer run, however, DI has fared less well. After the influx from Syria in 2015, Germany has sought to stabilise the system by taking in a large share of asylum seekers. Yet, it has not succeeded in overcoming opposition to a relocation scheme promoting a more balanced distribution of asylum seekers and refugees across Europe. Member states have subsequently shown increasing levels of non-compliance with EU obligations, also leading to the partial reintroduction of border controls (De Somer et al. 2020). External differentiation towards Turkey, Libya or the Western Balkans has brought short-term relief but has not solved the existing imbalances among EU countries while at times highlighting the EU's vulnerability towards external partners (Okyay et al. 2020). Thus, we show, fourth, that these developments have undermined the legitimacy of existing arrangements (Lavenex and Križić 2019, 18).
In what follows, this article discusses first the triggers, second the governance and third the effectiveness and legitimacy of DI in asylum, border controls and police cooperation.

Triggers of differentiation
The impetus for differentiation in border controls, asylum and police matters originates in deeper economic integration objectives promoted most forcefully by the German government in the early 1980s: creating a vast economic space without internal borders in Europe. As few member states of the European Community (EC) could accept this move, integration should progress via differentiation: starting with a limited group of states and forging compromises with reluctant counterparts, ultimately creating a centripetal dynamic towards de-differentiation.
The proposal to abolish internal border controls came in response to the fall in international trade and cross-border investment within Europe following the second oil shock of [1979][1980]. Business associations, the German government, together with the Benelux governments, border region associations and, finally, the European Commission pointed to the economic costs of persisting barriers between the national markets of EC member states (Bigo 1996;Comte 2018a, 144). In 1984, the German president of the Permanent Conference of Chambers of Commerce and Industry of the Community, Herbert Pattberg, considered that "the administrative obstacles at borders [meant] a harmful waste of time and money that it [was advisable] to eliminate as soon as possible". 2 Even though the movement for the abolition of internal border controls was transnational, Germany, as a leading export nation, was the most active in promoting change and deepening European integration in this area. The French government did not share this agenda. It had actually stepped up border controls throughout the 1970s and early 1980s to enforce its increasingly restrictive migration policies (Comte 2018b). When, in July 1982, the Commission and Benelux countries supported loosening controls, the trend in France was for even more controls (Comte and Lavenex 2021). A framework with only West Germany and the Benelux countries led France to change position, when it signed the June 1985 Schengen Agreement with these other countries. 3 Negotiations within this framework on "compensatory measures for the safeguarding of internal security" (Lavenex 2018.) would give France leverage to have Italy and Spain increase external border controls and take in more asylum seekers. These negotiations were less a case of functionalist spillover from one field of integration to another (Niemann 2006) than a deliberate attempt by France, but also West Germany and the Benelux countries, to influence the migration policies of other European countries. The denunciation of 'Europe passoire' (Sieve Europe) and 'asylum-shopping' dramatised the risks of unauthorised border crossings and spread the fear that immigrants could exploit the absence of border controls and the different asylum standards within Europe to lodge multiple asylum applications (Bigo 1996;Lavenex 2001, 862;Comte 2020, 6-7).
When the first negotiations to implement the Schengen project occurred in the late 1980s, immigration from third countries was at historically high levels. The main French concerns were about immigration from the south. In March 1989, the French Minister of the Interior, Pierre Joxe, asked Prime Minister Michel Rocard: "What if tomorrow a serious political and social crisis in a Maghreb country brought to us waves of asylum seekers for completely justified reasons, both political and economic? [. . .] We must [. . .] avoid solutions that would leave us helpless in the event of an acute crisis". 4 In Schengen negotiations, Joxe asked for the obligation for third-country nationals to declare themselves at the border as soon as they entered French territory. 5 He also wanted to have "mixed brigades at external borders, to mutually ensure the quality of the controls carried 2 Archives centrales du Conseil de l'Union européenne, Brussels, Liste Rouge 1842, Telex No. 074, 20 March 1984 out on behalf of all". 6 In the face of this French hesitation to give up controls, Germany had to insist that France complete the negotiations to implement the Schengen Agreement. 7 For France, the possibility to force Mediterranean member states to adopt stricter immigration policies was a key advantage of the Schengen framework. Paris was able to obtain an agreement on stringent external border controls, later to be imposed on Mediterranean member states if they wanted to join the borderless European market. 8 Besides fearing losing control over third-country immigration, France also feared fiercer German competition if it abolished internal border controls. To alleviate French fears, Germany not only gave France a pre-eminent role in the definition of external border controls, asylum and police cooperation, it also acquiesced to the long-term French request for European monetary integration, which was decisive in leading France to side with Germany on Schengen matters (Comte 2018a, 146-8).
After Germany, France, the Netherlands, Belgium and Luxembourg signed the Schengen Implementation Convention on 19 June 1990, Italy, Spain, Portugal and Greece were wary of taking up the obligation to step up external border controls and be responsible for examining more claims of asylum seekers. However, the vast borderless single market created strong centripetal effects. Their consent was won through this simple issue-linkage: to have the five Schengen states abolish controls at common borders, Italy, Spain, Portugal and Greece had to join the Schengen Implementation Convention, which entailed strengthening their external border controls, tightening immigration regulations and, with Article 30, accepting the first-entry principle in allocating asylum seekers across members (Lavenex 2018(Lavenex , 1202. 9 It would nevertheless take eight years after the Schengen Convention was signed before France would abolish systematic controls at the Franco-Italian border. Even then, the French police would continue controls that were mobile and supposedly 'random' -in practice based on racial profiling -within 20 kilometres of the border (Casella Colombeau 2020, 2261-2).
Neither compensations nor issue-linkages were able to overcome the opposition by Denmark and the United Kingdom (UK). In the UK, reluctance to abolish border controls originated from the fact that Great Britain is an island and controls at its ports and airports were effective in stopping "the movement of drugs, of terrorists and of illegal immigrants", as Prime Minister Margaret Thatcher put it in 1988 (Thatcher 1988;Comte 2018a, 150-1). For Denmark, besides the control of immigration, a particular area of concern was that the German police could enter Danish territory uncontrolled without border checks. During the Maastricht negotiations in 1990 and 1991, Denmark and the UK were the leading opponents that "blocked German plans for the full Communitarisation of immigration and asylum policy" (Adler-Nissen 2014, 116). 6  Reluctance in those two countries was not only related to issues specific to border controls, it also matched a broader attitude towards European integration. According to the calculations of Frank Schimmelfennig and Thomas Winzen (2014, 366; see also 2020, 6-7), Denmark and the UK accounted for nearly half of differentiated arrangements in the EU in 2014. The UK never abolished border checks and eventually left the EU in 2020. Denmark did abolish border checks but then restored them at its border with Germany spectacularly in 2011 -before removing them once again (Adler-Nissen 2014, 134). Ireland, which has no land border with EU members other than the UK, had to follow British steps to preserve their common travel area and kept border controls with the rest of the EU (Schimmelfennig and Winzen 2014, 367).
In short, differentiation in JHA affairs aimed to abolish internal border controls to deepen the internal market while ensuring compensatory measures at external borders and in asylum and police matters. Germany was the driver of this agenda, soon joined by France, whose reluctance diminished due to German economic concessions and joint interests in controlling immigration. Given their pre-existing open border arrangement and pro-European stance, the Benelux countries were also on board. Differentiation would then serve for those like-minded Euro-centred states to proceed among themselves, gradually deepen their cooperation and extend it to other partners. Having explained what triggered differentiation in JHA matters, we now turn to the governance of DI and the role of EU institutions -the latter gradually replacing direct German interventions.

Standards and legal instruments
The member states resorted to an array of differentiated arrangements when they proceeded to the abolition of border controls and the compensatory measures regarding external border controls, asylum and police matters. The Dublin Convention, signed on 15 June 1990 by all EC member states outside of the Community framework, included the same provisions about asylum as in the Schengen Implementation Convention, which the five Schengen members were to sign four days later. Determining only one member state that should examine an asylum claim implied that member states should mutually recognise each other's asylum decisions. This solution avoided harmonising their heterogeneous asylum legislations (Lavenex 2018. This flexibility was necessary to include the most reluctant members. Given the wide differences across countries (Adler-Nissen 2009, 73), however, this solution was not sustainable and, under the Treaty of Amsterdam signed in October 1997, the member states adopted directives to establish minimum asylum standards. Despite the expansion of these directives under the Treaty of Lisbon, EU asylum law still concedes a wide margin of discretion to the member states, so that differentiation remains extensive. For example, adjusted for the composition of asylum seekers, refugee recognition rates have varied from 25 per cent to 70 per cent across countries (Leerkes 2015). Also, an asylum seeker could take up employment immediately upon applying for asylum in Sweden, whereas they had to wait nine months in France. Social security benefits for asylum seekers have been lower in Denmark than in other countries (Den Heijer et al. 2016, 609, 614).
Even though the Dublin Convention was part of the compensatory measures for the abolition of internal border controls, the UK could join it without abolishing border checks. The reason the UK joined was because the criteria to allocate applicants meant it would receive further options to return them to the country of their first entry in the EU (Asderaki and Markozani 2019). 10 Cooperation on abolishing internal border checks also included differentiation in standards to overcome French concerns. Following the first and second Schengen agreements, a compensatory measure for reluctant members -later formalised in the Schengen Borders Code of 2006, itself subsequently revised -allowed states to restore border controls under specific conditions. These included, first, foreseeable events, such as sports events; second, serious threats to internal security; and third, from 2013, deficiencies in the control of the external border of the Schengen area. Controls could not exceed six months in the first case, two months in the second and two years in the third (De Somer et al. 2020, 3, 8). Next to these variations in standards, some states could conclude special agreements allowing certain border practices. For instance, with the Chambéry bilateral agreement, signed in October 1997, France traded its abolition of systematic checks at the Italian border for Italy's commitment to readmit persons who had transited through its territory (Casella Colombeau 2020, 2262).
Besides differentiation in standards, flexibility among the member states also led to differentiation in legal instruments. After British and Danish opposition prevented Germany from communitarising the compensatory measures for the abolition of border controls in the Treaty of Maastricht, the member states created a pillar structure in that treaty, with a third pillar devoted to "Justice and Home Affairs". In contrast to the "Community" pillar, the JHA pillar was intergovernmental, with little involvement of EU institutions, which was a way to reassure those member states reluctant to abolish their controls. France was confident that it could more easily exit such intergovernmental agreements and restore controls if it considered that other members were not respecting them (Comte 2018a, 148). Denmark eventually joined Schengen cooperation on the abolition of internal border checks as the other members of the Nordic Passport Union joined the EU or the Schengen area. It mattered to Denmark to preserve this union, and flexibility in legal instruments helped ensure Danish participation (Adler-Nissen 2014, 118). As member states extended Community procedures in this area with the Treaty of Amsterdam, they accepted that intergovernmental procedures would still apply to Denmark, with a "Protocol on the position on Denmark" annexed to the Treaty, exempting Denmark from Community procedures. Therefore, this arrangement stipulated that different member states would be subject to different types of legal instruments (Adler-Nissen 2009, 75;Peers 2017, 257).

Opt-outs
The dominant form of differentiation of participation occurred when the member states integrated while allowing some of them to opt-out of new arrangements. Even though opt-outs are the most emblematic aspect of EU differentiation, in border controls, asylum and police matters, they have often resulted from attempts to de-differentiate previous differentiated arrangements (Tekin 2012, 27). In 1997, all EU member states had signed the Schengen Convention, except the UK and Ireland. Labour's victory in the 1997 UK general election helped Schengen members incorporate their arrangement into the European treaties with the Treaty of Amsterdam, provided that the UK and Ireland could receive a formal opt-out from the Schengen acquis (Adler-Nissen 2009, 68). The Protocol on the position of the United Kingdom and Ireland specified that both countries could request to participate in Schengen measures on a case-by-case basis if they wished, subject to unanimous approval of the other participating states. 11 Likewise, as previously mentioned, Denmark received an opt-out and could continue to cooperate in this area on an intergovernmental basis, while Community law applied to other states. Denmark used this opt-out to diverge further from other member states regarding asylum standards in the following years (Comte 2020, 11).
In a move towards de-differentiation, the Treaty of Lisbon, signed in December 2007, incorporated into EU law all border controls, asylum and police cooperation. All the powers of the Commission and the EU Court of Justice would apply accordingly. This was, however, the context of further differentiation in participation. Denmark extended its opt-out to the entire area (Adler-Nissen 2009, 75). Likewise, the British and Irish optouts included this whole area, with an opt-back-in clause on a case-by-case basis. 12 Article 4 of Protocol No 19 annexed to the Treaty of Lisbon allowed Ireland and the UK to request at any time "to take part in some or all of the provisions of the Schengen acquis". 13 The UK informed the European Council of its decision to exercise a complete opt-out in July 2013, before the area was scheduled to come fully under the European Court of Justice's jurisdiction on 1 December 2014. 14

External differentiation
To move towards integration, the Euro-centred member states compromised not only on standards, types of legal instruments and internal membership, but also by taking in non-EU members when this could help integrate. They included Central and Northern European countries into the Schengen area and involved third countries at the eastern and southern periphery to expand the EU system of migration management.
In 1996, to have Denmark, Sweden and Finland abolish their controls, the other member states accepted Norway and Iceland, which had been part of the Nordic Passport Union without internal borders along with the other three since 1952, into the Schengen area. Likewise, in 2008 and 2011, the member states integrated Switzerland and Liechtenstein, two countries that were surrounded by the EU but did not want to join it entirely, into the Schengen area (Lavenex 2006a;Peers 2017, 255 EU's richest regions, they were a major communication node in the heart of Europe. Their borders could, therefore, not become external borders with stringent controls without seriously hampering the internal European market. External differentiation has also extended to countries located along migratory routes. Their enlisting in the fight against irregular immigration has facilitated the absence of border checks within the EU. Flexible cooperation with neighbouring countries of transit for migrants to the EU has developed in parallel to internal cooperation since the early 1990s (Lavenex 1999;2006b). While part and parcel of the EU's "Global Approach to Migration and Mobility", external cooperation has also been driven by single EU member states, leveraging on their respective diplomatic links. 15 The growing weight given to the association of non-EU countries with the EU system of migration governance implies participation by non-members in this policy field through regulatory commitments, organisational involvement or a combination of both. External cooperation arrangements, therefore, constitute instances of external differentiation, which we define as extra-EU actors' selective participation in EU policies "from the perspective of regulatory commitment and organisational participation" (Lavenex and Križić 2019, 8; see also Okyay et al. 2020, 4).
In the 1990s, Germany was a driving force extending JHA cooperation to the candidate countries of Central and Eastern Europe. In 2006, the Austrian Presidency of the EU Council supported the "Police Cooperation Convention for Southeast Europe" on border controls with Albania, Bosnia and Herzegovina, Moldova, Montenegro, North Macedonia and Serbia. Spain developed ties with Morocco, and the EU later co-funded the "Seahorse Atlantic" network of border surveillance, promoted by Spain and involving Morocco and other Western African countries. The EU also endorsed agreements between Italy and Libya on external border controls and funded programmes that integrated Libya in managing the EU's outer border. The EU-Turkey statement issued on German initiative in March 2016 enlisted Turkey in the management of the EU's external borders, with rapid and decisive effects on flows (Okyay et al. 2020, 6-10). This external dimension also appears in the web of cooperation arrangements concluded by the EU's agency for external border controls -the European Border and Coast Guard Agency, also known as Frontex -with third countries .
In short, an array of differentiated standards, legal commitments and participation levels across member and non-member states has created a participation pathway channelling the centripetal effect of cooperation for more Eurosceptic member states and allowed reaching out to strategically important non-EU countries. In the next section, we will evaluate the effectiveness and legitimacy of this differentiation.

Effectiveness and legitimacy of differentiation
Effectiveness entails generating policy outputs whose outcome is to solve policy problems (Lavenex and Križić 2019, 10). In JHA, the basic measure of effectiveness in light of EU leading policy objectives is the capacity of DI to ensure a sustained absence of internal border controls in the Single Market. This objective itself entails achieving gradual dedifferentiation and developing common measures on asylum, border controls and police cooperation. Legitimacy hinges on the conformity to the preferences of representative institutions and overarching normative commitments (input legitimacy), the continued control of the policy development by representative institutions (throughput legitimacy) and the capacity to produce effective solutions (output legitimacy) (Lavenex and Križić 2019, 18f.; Tallberg and Zürn 2019). There is a dilemma in EU differentiated integration in JHA between the demand for effectiveness -which entails maintaining the conditions for the absence of internal borders -and legitimacy -which entails conformity with the EU's commitment to human rights, rule of law and an equitable relationship between internal and external partners (Lavenex 2018).

De-differentiation and policy outputs
The case of JHA combines enduring differentiation with important instances of dedifferentiation. Initial strictly intergovernmental agreements gradually came under the responsibility of the Commission, the European Parliament and the Court. The participating states signalled their willingness to de-differentiate early. The first Schengen Agreement exempted all EC nationals from border checks. The EC Council of Ministers was also connected to the 1990 Dublin Convention: it hosted the preparatory negotiations, and its Secretariat could fulfil certain tasks related to the Convention. The Treaty of Maastricht made asylum and immigration a "matter of common interest" (Comte 2020, 9). The transfer of responsibility to European institutions occurred in two steps, with the treaties of Amsterdam and Lisbon.
In taking charge, European institutions delivered new policy outputs regarding harmonisation, the creation of new agencies and instruments to manage the area at the EU level. In 1995, the member states established the European Union Agency for Law Enforcement Cooperation, better known as Europol. After incorporating asylum cooperation into the Community pillar with the Treaty of Amsterdam, the European Council in Tampere gave the impetus for harmonising the member states' asylum policies. With two regulations, in December 2000 and February 2002, the Council created the European dactylographic system (Eurodac). In 2002, EU member states agreed on a European Arrest Warrant to prosecute criminals across borders and guarantee the continued absence of border checks inside the EU. In 2003 and 2004, the Council adopted directives laying down minimum standards for the reception of asylum seekers, the conduct of asylum procedures and the status of refugees. In 2004, the member states created Frontex. In 2008, they partly harmonised their return procedures and standards for irregular immigrants with the EU Return Directive. In 2011, they created the European Asylum Support Office to reduce the divergence in recognition rates across countries. To meet asylum seekers' needs more homogeneously, they created the European Refugee Fund, later called the Asylum, Migration and Integration Fund. 16 While retaining a degree of flexibility for domestic implementation, all those steps facilitated and prepared the full integration of this policy area at the EU level.
Another crucial aspect of de-differentiation was the extension of participation from the five initial Schengen members to most EU members. As previously mentioned, issuelinkage between the abolition of internal border controls and cooperation on external border controls meant that the Mediterranean member states, initially excluded from the 16 Interview with an Italian expert, 3 April 2020.
Schengen cooperation, eventually joined it -even if this implied accepting border controls, asylum and police cooperation that placed a heavy burden on them (Comte 2020, 8). Likewise, the strategic inclusion of this policy area in the Treaty of Amsterdam and of significant parts in the Community pillar, governed by qualified majority voting, forced the new member states from the 2004 and 2007 enlargements to accept this cooperation as part of the acquis communautaire without opt-out provisions (Adler-Nissen 2014, 123;De Somer et al. 2020, 7).
In parallel, the Schengen Information System (SIS) acted as a club good to generate centripetal effects and bring in countries that had negotiated an opt-out: Denmark, Ireland and the UK. Providing valuable data for security and law enforcement, the SIS is excludable and has increasing returns as new members join, making non-participation increasingly costly (House of Lords 2007; Lavenex and Križić 2019, 15). In March 1999, the UK applied to participate in several areas of the Schengen acquis, including police and judicial cooperation (Council of the European Union 2000 and 2004). Ireland followed the UK's move (Council of the European Union 2002). By the time the Treaty of Lisbon entered into force, the UK had opted into most civil law measures, asylum measures and measures concerning illegal migration (Adler-Nissen 2009, 69). To exercise the opt-out right under this treaty, the UK could only opt out of all legislation, but immediately exercised the right to opt back into 35 EU measures it had previously agreed to, accepting the Commission's enforcement powers and the Court's jurisdiction over them. Examples of measures in which the UK participated include the European Arrest Warrant framework, the SIS, the Prüm Decisions on cross-border cooperation against crime and the EU Passenger Name Record. The UK even fought unsuccessfully in front of the European Court of Justice to participate in Frontex despite not abolishing its internal border checks. Such cooperation had become crucial to control immigration at British borders (Asderaki and Markozani 2019). Even though not taking part in the core of Schengenthe abolition of internal border checks -Ireland and the UK finally opted into most of the compensatory measures to tackle irregular migration (De Somer et al. 2020, 6-7).
Successful de-differentiation led to a high level of effectiveness, with most EU members giving up their systematic internal border controls. Even though differentiation in standards and external differentiation remained extensive, for a short period after the Treaty of Lisbon entered into force, there was an illusion that differentiation in JHA had become limited to a few opt-outs. This situation explains why there was a tendency among scholars to focus on a couple of opponents to integration to account for differentiation. Such a situation, however, did not last long.

Integration stalemate and legitimacy crisis
Notwithstanding its initial success, the effectiveness of DI has encountered limits in border controls, asylum and police cooperation. The lack of compliance, divisions among members, the increasing difficulty of reforms and the ambiguous impact of common policies have spurred a sense of crisis over both the effectiveness and legitimacy of cooperation.
The first major challenge came with the Arab revolutions of 2011. The escalation in the number of migrants from Tunisia to Italy caused a first crack in the Schengen system (Casella Colombeau 2020, 2265-8). On 5 April 2011, the Italian government of Silvio Berlusconi issued temporary residence permits to all the citizens of North African countries who had arrived since 1 January (Italian Ministry of the Interior 2011). These permits allowed them to travel freely in the Schengen area for up to three months if they could demonstrate sufficient resources. In reaction, the French police first increased its staff at the French-Italian border (French Ministry of the Interior 2011) and, on 17 April, closed the border, shutting down all traffic. This episode heralded a decade of more frequent reintroductions of border controls in the Schengen area, undermining effectiveness. The Franco-Italian affair triggered, in 2013, the reform of the Schengen Borders Code, which allowed member states to reintroduce internal border controls in the event of "serious deficiencies in the carrying out of external border controls" (European Parliament and Council 2013) -thereby increasing the flexibility member states could use at internal borders, at odds with the plans of de-differentiation. This augmented flexibility soon came to bear in the next crisis when member states again closed their borders facing an influx of refugees and asylum seekers in 2015-2016 (Lavenex 2018). Such influx occurred as the military situation in both Libya and Syria was deteriorating.
The incapacity to find political agreement to reform the EU asylum system, which the Commission and Parliament have supported since 2015, is another indicator of the limits of current arrangements. The negotiations on harmonising asylum standards have been unable to proceed beyond a certain point (Den Heijer et al. 2016, 609). Yet, the Dublin system has put unsustainable pressure on the countries of first entry in the EU and has led to the collapse of asylum systems in Greece and Malta, and partly also in Italy and Spain. As early as 2011, the European Court of Human Rights and, soon afterwards, the European Court of Justice ruled that asylum seekers found escaping Greece should not be sent back under the Dublin procedure, given the deplorable conditions in which they lived in that country and the lack of access to a functioning asylum system (ECtHR 2011).
Until that point, Germany had taken a step back, hoping the normal working of EU treaties and institutions would achieve de-differentiation and integration. Yet, the stalemate in which EU institutions found themselves in the migration crisis of 2015-2016 led the weak hegemon to attempt to stabilise the system. It tried to do so in three ways: first, in a unilateral move, taking in more than a million migrants; second, supporting relocation quotas for asylum seekers in the attempt to fix the flawed Common European Asylum System; and, finally, in consequence of sustained opposition from member states, reverting to external differentiation, sponsoring a migration deal with Turkey in March 2016. By suspending the Dublin rules for Syrian refugees, Chancellor Angela Merkel hoped to alleviate the situation in the overburdened entry points in Greece and Italy. Yet, the move alienated other member states that feared a potential pull-effect on other migrants and asylum seekers.
In parallel, Germany intensified efforts to develop a relocation scheme in the EU that would distribute asylum seekers among member states based on a quota system. This attempt faced the resistance of the traditionally reluctant northern partners -Denmark, the UK and Ireland -but also the Visegrad countries of Central Europe -Poland, the Czech Republic, Slovakia and Hungary. The former used their opt-outs to avoid the scheme (Den Heijer et al. 2016, 614). The latter, with no such opt-out option, used voice by voting against. As they were outvoted, the Council adopted two relocation decisions concerning a total of 160,000 asylum seekers then in Greece and Italy on 14 and 22 September 2015 (Council of the European Union 2015a; 2015b). Although formally bound by the decisions, the Visegrad countries did not comply. In April 2020, the European Court of Justice ruled that, by refusing to comply with the temporary mechanism for the relocation of applicants for international protection, Poland, Hungary and the Czech Republic had failed to fulfil their obligations under EU law (CJEU 2020). These proceedings showed the difficulty to achieve less differentiation and more stringent common solutions. Member states used voice or exit strategies to avoid cooperation (Goldner Lang 2020).
Non-compliance has not been limited to Visegrad countries. As of 23 September 2015, the European Commission had launched 40 infringement procedures against 19 member states for failing to implement EU asylum legislation (Den Heijer et al. 2016, 625). The Commission took these procedures a step further, and overall, by the end of 2016, there were 138 pending procedures in migration and home affairs, half of them on asylum cases (Goldner Lang 2020). In July 2017, the Grand Chamber of the Court of Justice confirmed, in cases involving Slovenia and Austria, that the creation and facilitation of the Western Balkans route violated Schengen and Dublin rules (CJEU 2017a;2017b).
The situation became even more problematic as Chancellor Merkel came under pressure from within. Faced with a significant inflow of asylum seekers and other migrants, Bavarian Finance Minister Markus Söder called to restore controls at German borders (Der Spiegel 2015). In early September 2015, German Interior Minister Thomas de Maizière suggested reintroducing border controls temporarily (Bond et al. 2015). The German police started intensively checking incoming flights from Greece to prevent unauthorised arrivals. To avoid being turned into a dead-end, Austria reintroduced checks at its southern borders on 16 September. These decisions at the centre of the system triggered a domino effect, leading all countries on the migrants' route to barricade themselves. On 11 November, Slovenia started building a razor-wire fence at its border with Croatia. On 12 November, Sweden ordered border controls at ports in the south and west of the country and announced that it could also extend border checks at its land and air borders (De Somer et al. 2020, 10). On 13 November, France declared a state of emergency following deadly terrorist attacks in Paris: as some perpetrators had used the migrants' route to enter the country and others had come from Belgium, France reintroduced controls at all its borders. The French police could act as if French borders were external EU borders and could issue non-admission decisions, pushing irregular migrants back into neighbouring countries, such as Italy (Casella Colombeau 2020, 2269. Slovakia, Hungary, Norway, Denmark and Belgium also reintroduced border controls. Even though these controls concerned third-country nationals only and not the free movement of goods, capital and services, the internal market was also at stake. Peripheral countries dependent on the export of goods, including perishable agricultural commodities, towards more prosperous core European countries voiced serious concerns about this situation. Eastern European states denounced controls that hit their goods transport sector (Than and Nasralla 2016). Southern European states -Cyprus, Greece, Italy, Malta and Spain (2020) -considered that: Reintroduction of internal border controls must revert to an extraordinary, proportionated and limited in time last resource. The image of traditional controls in our internal borders makes us go back decades in the European project and could jeopardise trust among [member states]. To reinforce security in our territory, less coercive and more efficient measures are possible.
Nevertheless, controls persisted at the borders of Germany, Austria, France, Norway, Sweden and Denmark, including, therefore, core European countries and other countries of immigration. After having reached the time limit of two years for those controls, they resorted to other articles of the Schengen Borders Code to open new time limits (European Commission 2021; Carrera et al. 2018). The global pandemic's outbreak in March 2020 led 18 of the 26 Schengen member states to reintroduce border checks straightaway (De Somer et al. 2020, 12;Rausis and Hoffmeyer-Zlotnik 2021).
To counteract increased differentiation at internal borders, Germany intensified the search for external solutions. In March 2016, Chancellor Merkel negotiated the aforementioned EU-Turkey statement, committing Turkey to retain Syrian refugees heading towards the EU. Austria sponsored cooperation with the Western Balkans, and Italy, France and Spain invested in cooperation with southern Mediterranean neighbours. The New Partnership Framework adopted by the Commission and Council in June 2016 reinforced this external differentiation (European Commission 2016).
Whereas in the short term, these deals helped reduce pressure at external borders, this temporary relaxation came at a price. The EU and its member states formalised their dependence on cooperation with other governments, which can exploit the EU's vulnerability. Moreover, the EU and its member states not only became complicit in human rights abuses in third countries, but they also downgraded their protection and rule-oflaw standards (Okyay et al. 2020). Therefore, these means to achieve EU goals have weakened the input legitimacy of the scheme. Furthermore, the intermediary situation between differentiation and de-differentiation has created uncertainty over the control by representative institutions: the scheme is still too differentiated to be under the control of the European Parliament, but it has reached such a level of de-differentiation that national parliaments and even governments no longer have full control over what they accept. The proceedings against Visegrad countries illustrate this latter problem. This situation has affected throughput legitimacy.
These concerns over both effectiveness and legitimacy ushered in further disintegration. The blow came from the traditionally most reluctant partners. Denmark and the UK took even more distance from the scheme. In the Treaty of Lisbon, the member states had tried to invite Denmark to de-differentiate and had offered Denmark the option to convert its full opt-out into a flexible opt-in on a case-by-case basis, like the UK's and Ireland's. By 2015, EU legislative developments had made clear that if Denmark did not implement this option, it would have to leave Europol -losing access to its databases. The Danish government accordingly organised a referendum to transform Denmark's complete opt-out into an opt-out with case-by-case opt-in. Yet, on 3 December 2015, in the middle of the migration crisis, a majority of Danish voters rejected this option, taking the risk to have to leave Europol. Eventually, the Danish government and the EU agreed on a deal maintaining cooperation with Europol that was signed in April 2017 -two days before Denmark would have been cut off from the agency (Denmark and Europol 2017). Under this agreement, the Danish police lost the capacity to access Europol databases directly like the other member states, but it could still rely on Danish liaison officers stationed at Europol -a procedure that takes more time and is, therefore, less effective than previous Danish membership ( Mortera-Martínez et al. 2021, 9).
As far as the UK is concerned, in June 2016, British voters decided to exit the EU entirely. The move came for various reasons, but the apparent loss of control over migration during the migration crisis favoured the 'Leave' vote. The hope that the EU record of differentiated cooperation would help mitigate the effects of Brexit was dashed with the EU-UK Trade and Cooperation Agreement signed on 30 December 2020, which was applied from 1 January 2021 (EU-UK 2020). The UK exited the Dublin system, cooperation on migration flows in the Mediterranean and police cooperation in Europol (Comte 2021). The Agreement envisages that the UK and Europol should keep exchanging information and personal data, but the details of this cooperation remain vague, at the discretion of both parties (Mortera-Martínez et al. 2021, 14).

Conclusion
Differentiated integration in border controls, asylum and police matters has evolved at the nexus of three dynamics: first, the strategic interest and investment of a weak hegemon, Germany, gaining the support of the Benelux countries and, at a greater cost, France; second, these countries' capacity to attract other members through centripetal effects by issue-linkages and network effects; and, third, the partial externalisation of the burden of migration control through external differentiation. In terms of governance, an array of differentiated instruments, including the differentiation of legal commitments, flexibility in standards, cooperation outside EU law and opportunities for optouts and case-by-case opt-ins, allowed cooperation on border controls, asylum and police matters to progressively enter EU treaties and then to extend Community procedures, develop various policy outputs and increase participation.
However, since 2011 and, above all, 2015, with the Arab revolutions and the civil wars in Libya and Syria, the EU's periphery entered a period of political turmoil. The reintroduction of internal border checks by member states and some member states' open rejection of Council decisions and Court rulings undermined the effectiveness of common policies. Germany re-emerged as a weak hegemon mobilising its resources to save the Schengen area by opening up its borders to asylum seekers who could not be sent back, while sponsoring flexible arrangements with key external partners to stem the inflow. Nevertheless, Germany itself had to partially withdraw from this regime by repeatedly enforcing controls at its borders -leading several other members to do the same and the most reluctant partners to take even more distance. Uncertainty over the control by representative institutions, attempts to halt internal disintegration through the watering down of human rights standards at external borders and the EU's vulnerability to cooperation with external actors have all gradually undermined the legitimacy of existing arrangements.
In short, differentiation in JHA emerged when a weak hegemon was able to enlist only a few partners for its Euro-centred agenda in the face of widespread adverse policy preferences. Strategically selected and compensated initial partners served to create a core exerting centripetal effects. De-differentiation followed, with the inclusion of more members, the homogenisation of standards and some transfer of responsibility to European institutions, enhancing effectiveness for Euro-centred objectives. However, changing circumstances increased the costs of these objectives, while external solutions and the dispersal of political responsibility over different layers undermined legitimacy. When costs rose, cooperation could only subsist through more differentiation, thus reversing the dynamic of de-differentiation.
Presently, it remains unclear how, in the current state of division among member states, the situation could eventually stabilise and de-differentiation resume. Also, if new crises occur, in which risks increase, differentiation may not suffice to prevent disintegration.