More Than Choosing a City
Choosing the seat of an EU body, office or agency means much more than selecting a city. It affects an entire region, which may benefit from the economic activity generated by the agency – in terms of flights, restaurants and accommodation, but also schools, services and the property market – as well as the entire epistemic and professional community surrounding it. That community needs the agency to be located somewhere that is well connected and easy to reach, and that offers the services on which the sector depends.
Choosing a seat, then, does not merely give a particular city a European dimension. It also means laying the foundations for the agency to operate efficiently and to fulfil, as effectively as possible, the objectives for which it was created.
Earlier seat-selection procedures have not always provided models worth emulating. The purpose of this post, however, is not to dwell on the past, but to examine what followed. Taking as its starting point the procedure for relocating the European Medicines Agency (EMA), and the principles laid down by the Court of Justice (CJEU or Court) when reviewing its legality, this post examines how the EMA affair shaped subsequent procedures for selecting the seats of EU agencies (including the recent selection of Lille as the seat of the new EU Customs Authority), and how far those procedures have succeeded in remedying the shortcomings of their predecessors.
The Dutch Vivaldi
On 20 November 2017, 27 EU Member States gathered to choose the new seat of EMA, which was to be relocated from London because of Brexit.
Based on a procedure agreed a few months earlier in the margins of a European Council meeting, interested Member States had to submit an offer, which was to be publicly assessed by the Commission against criteria largely similar to those laid down in the 2012 Common approach. Voting could then run up to three rounds. A bid needed the first-choice support of at least 14 Member States to win in the first round, or 14 votes to win in the second. Failing that, the leading bids proceeded to a third and final round. At that stage, however, any tie was to be settled by… drawing lots.
That is precisely what happened: since Slovakia had abstained, Milan and Amsterdam finished level with 13 votes each, and the tie had to be broken by lot. Milan, which had led in both preliminary rounds without reaching the threshold required for selection, offered EMA the Pirellone: an emblem of the city, designed by Gio Ponti and built at the height of Italy’s dolce vita. Amsterdam, which drew level with Milan only in the final round, proposed a phased relocation. EMA would initially move into temporary premises, before transferring to a new, purpose-built permanent headquarters which, with some irony, was to be called the Vivaldi building. Fate favoured Vivaldi, and EMA therefore went to Amsterdam.
Paradoxically enough, the abovementioned procedure prompted the Commission to table a legislative proposal once the new seat had been selected. The proposal would amend EMA’s founding regulation to include an express reference to that seat. Technically, the amendment was unnecessary: the regulation then in force did not state that EMA’s seat was in London and therefore did not need to be amended in the first place.
The rationale may have been to lend an otherwise intergovernmental procedure a measure of democratic legitimacy. However, this resulted in little more than a formality to keep up democratic appearances. The choice of the new seat had already been made and announced to citizens and stakeholders, and the legislature was merely being asked to ratify it and write it into law. The Parliament was all sound and fury, until it was time to fall into line. As evidence of its determination to uphold the democratic principle, it managed only to have a declaration annexed to the resolution adopting the regulation, condemning the fact that it had been presented with a fait accompli.
Milan, followed somewhat obtorto collo by Italy, challenged both the decision of 20 November 2017 adopted by the Member States (the EMA I joined cases) and the regulation amending EMA’s founding act, which ratified that decision (the EMA II joined cases).
Yet the questions of why a choice of such importance was left to the drawing of lots and whether the outcome of that draw was truly the best one for EMA are now water under the bridge and hardly worth dredging up. And for all its flaws, the procedure put in place for EMA’s relocation still represents an improvement compared to those previously used to select the seats of EU agencies, that is, purely intergovernmental decision-making processes subject to opaque political trade-offs between Member States or worse. Indeed, the story goes that, during the negotiations over the seat of the European Food Safety Authority (EFSA), one of the arguments advanced in favour of Parma over Helsinki was that ‘the Finns do not even know what prosciutto is’ – and Parma was indeed ultimately selected.
The EMA / ELA Principles
What deserves attention instead is what happened next, both at the institutional level and in the proceedings before the Court of Justice.
While the actions brought by the Municipality of Milan and Italy were still pending, a very similar procedure – though without the absurd rule requiring ties to be settled by lot – was used to select the seat of the then forthcoming European Labour Authority (ELA). This time, however, Parliament made a wholly unexpected break with its earlier acquiescence. In a sudden assertion of its institutional prerogatives, it brought its own action challenging the Member States’ decision, arguing that it had been completely sidelined.
The EMA I, EMA II and ELA cases were joined solely for the oral phase but were, de facto, treated by the Court as raising a single legal issue. Is the choice of an EU agency’s seat a separate act – distinct from the decision establishing the agency – and can thus be reserved to the Member State governments? Or is it instead a matter for the EU legislature, to be negotiated in the same way as any other element pertaining to the legislative act establishing a new EU agency?
In the judgments (here EMA I, here EMA II, here ELA), the Court held that the seats of EU bodies, offices and agencies must be determined by the EU legislature, through the open and transparent legislative procedure prescribed by the legal basis used to establish the agency itself, rather than by the Member States. Article 341 TFEU therefore applies to the choice of the seat of EU institutions and is not to be used by analogy when it comes to the seats of EU agencies. The Court did not go so far as to spell out precisely what characteristics the procedure must possess: once it had held that the Member States’ decision was not amenable to judicial review and could not influence the legislative process, that procedure was obviously no longer the subject matter of the action.
Nevertheless, the Court used clear and forceful language to articulate the principles towards which the procedure must strive. The decision on the location of the seat of agency is – and the term deserves to be emphasised – ‘consubstantial with the decision on its establishment’. This means that the procedure is made of the same substance, and is of the same nature, as the decision establishing the agency itself. The procedure must therefore comply with all the rules and principles stemming from the legal basis on which the agency is established and respect the legislative nature of the procedure provided for by that legal basis. These concern, inter alia, the right of initiative, co-decision powers, voting majorities in the Council and Parliament, transparency and access to documents, and effective judicial review.
Selecting EUCA’s Seat: Theory and Practice
Considering the principles affirmed by the Court in the EMA and ELA cases nearly a year before, and the Commission’s consistent determination to safeguard its right of initiative, the approach taken in its proposal to establish the European Union Customs Authority (EUCA) may come as a surprise. In fact, as regards the selection of the new agency’s seat, the Commission ‘considered it appropriate to leave open the name of the hosting city in its proposal’, merely recalling the four selection criteria laid down in the 2012 Common approach and expressing its willingness to ‘assist in [their] evaluation’ and ‘actively cooperate with the co-legislators on the selection of the seat’.
In October 2025, the co-legislators first agreed on an ‘application process’: each Member State could submit one application, to be assessed by the Commission against the four abovementioned criteria, which were themselves further detailed in technical specifications also laid down in the application process. Both the applications and the assessments were to be made public. Then, in February 2026, after the Commission had assessed the nine applications received, and the European Parliament had organised a hearing with the candidate cities, the three institutions agreed on a procedure for selecting the agency’s seat. The procedure was formalised in a ‘Common Understanding’, broadly similar to the one concluded in December 2023 regarding the selection of the seat of the Anti-Money Laundering Authority (AMLA).
The Common Understanding provided that the Parliament and the Council would first assess the applications in parallel and then separately identify their two preferred candidates. If one candidate appeared on both institutions’ lists, it would be deemed selected; otherwise, joint rounds of voting by secret ballot would be held. The co-legislators would each hold 27 votes (and ballots), to be allocated within the two delegations according to their internal arrangements. The rounds of voting would be subject to different thresholds: first, an absolute majority in both delegations (14+14); second, a three-quarters majority of the votes cast (41 votes out of 54), and third, a two-thirds majority of the votes cast (36 votes out of 54). If no candidate secured the required majority after the third round had been repeated three times, the candidate receiving a simple majority of the votes cast in the final ballot would be selected. The outcome would then be ‘reflected in the legislative text to be formally adopted…in accordance with the ordinary legislative procedure’ at a trilogue convened immediately thereafter.
On 25 March 2026, the Parliament and the Council separately identified their two preferred candidates before convening for an interinstitutional meeting at the Council’s premises. A joint vote followed, and Lille’s selection was announced that afternoon, but neither institution disclosed the number of voting rounds held, the number and identity of the other shortlisted cities, the internal procedures through which those cities had been shortlisted, or how the voting rights were allocated among the delegations.
Nevertheless, the quite unusual doorstep statement made by the European Parliament’s rapporteur and the Permanent Representative of Cyprus, which then held the Council Presidency, following the interinstitutional meeting indicates that both the Parliament and the Council selected Rome and Lille (see also here, here and here), with the Parliament reaching that outcome after seven internal ballots. It also suggests that both finalists had broad support in both institutions and that Lille prevailed in the second repetition of the third voting round with 36 votes out of 54.
The outcome was promptly inserted into the draft Regulation at a trilogue held the following day, at which the co-legislators concluded negotiations on the EU customs reform. The resulting Regulation was subsequently adopted at first reading and entered into force on 20 September 2026.
EUCA Seat Selection: All’s Well That Ends Well?
The procedure used for selecting EUCA’s seat differs significantly from those followed for the seats of EMA and ELA. Most notably, of course, it involves the Parliament and does not resort to a draw. However, it also displays significant similarities to the innovative procedure established in 2017 for EMA and essentially reproduced for ELA. The question, therefore, is whether the procedure used for selecting EUCA’s seat fully complies with the legal implications arising from the EMA and ELA cases. Or, put it differently, whether the EU institutions have truly learned their lesson from the EMA affair.
In this regard, the EUCA procedure raises at least three legal issues, as well as a broader political concern. The first legal issue concerns the role of the Commission. Under Article 293 TFEU, read in light of Article 17(1) TEU, the Commission holds the power of legislative initiative and a correlative right to withdraw a legislative proposal, both designed to promote the Union’s general interest (see Council v Commission, paras 70-74). As regards EUCA’s seat, the Commission relinquished both the first word on its location and any opportunity to influence the last word. Had the seat instead been negotiated within an ordinary trilogue, the Commission could have played a more active role, for example in ensuring that adherence to the selection criteria prevailed over purely political considerations. After all, while the deliberative part of the procedure for choosing EUCA’s seat closely resembles that of the Conciliation Committee, the Commission participates in that committee pursuant to Article 294(11) TFEU.
A second legal issue concerns compliance with Treaty procedures, in light of the ECJ case law strictly prohibiting ‘secondary legal bases’ (see Parliament v Council, para 47). The most sensitive aspect in this regard concerns the alterations made to the voting system. When selecting the seats of EUCA (and AMLA), the Council delegation departed from the qualified-majority rule provided for in Article 16(4) TEU and appeared to have voted on a “one-member-one-vote” basis. The Parliament effectively acted as the direct-democratic counterweight to a decision-making process that was otherwise purely intergovernmental. However, allocating 27 votes to each co-legislator, while giving each Member State a single vote within the Council delegation, compounded the effects of Parliament’s degressively proportional composition, thereby increasing the relative weight of Member States’ governments vis-à-vis Parliament. This is all the more significant when one considers that the choice of seat is then “reflected into the legislative text” as an external element that is no longer open to negotiation as part of the ordinary legislative procedure.
A third legal issue is linked to compliance with the principles of transparency and openness enshrined in the Treaties. Pursuant to Article 15(2) TFEU, the European Parliament shall meet in public, as shall the Council when considering and voting on a draft legislative act. Although legislative acts are effectively negotiated through trilogues, which are held in camera, the General Court has, in principle, recognised a right of access to the so-called four-column tables that track these negotiations (see De Capitani, paras 68-112). Moreover, trilogues are based on negotiating mandates that are generally public and must ultimately be confirmed through public votes, while both co-legislators report on the progress of the interinstitutional negotiations. By contrast, although the decision on the seat is an integral part of the legislative process, neither the votes cast to select the seat of EUCA (or AMLA), nor the manner in which voting rights were allocated within the co-legislators were made public.
Finally, from a broader political perspective, it must be noted that, while a selection procedure like the one envisaged for EUCA appears to ensure that the two co-legislators act “on an equal footing” (see Common Understanding, para 4), it risks diminishing Parliament’s effective decision-making weight. Indeed, the Council might more easily rally behind a single candidate, possibly as part of a broader compromise aimed at ensuring geographical balance in the long-term allocation of agency seats. Reaching an equally cohesive position may prove more difficult within Parliament, where preferences are fragmented not only along political lines but also according to national affiliations. This is all the more likely in the case of a decision with significant political and economic implications, such as the location of an agency (see, in the same vein, regarding AMLA seat, Chamon). Political groups may therefore be internally divided over different candidate cities, while members belonging to different groups may support the same city on national grounds. Moreover, maintaining long-term agreements on the allocation of agency seats may prove more difficult in Parliament, as its political majorities are necessarily transitory.
In this respect, the voting rules established for EUCA arguably represent an improvement over those used for AMLA, under which Parliament and the Council were each allocated 27 votes, but only a simple majority of the votes cast in the joint ballot was required (i.e., potentially, the Council votes +1). However, they do not eliminate the problem entirely, as illustrated by the seven rounds of internal voting reportedly required for Parliament to shortlist its two preferred candidates.
EUCA Seat Selection Procedure: Blueprint or Cautionary Tale?
Despite the concerns raised above, the procedure for selecting EUCA’s seat had the merit of confirming and consolidating the practice developed since 2017: enabling a swift decision on the location of the seat prior to the conclusion of the legislative process establishing the agency and somewhat tying that choice to the objective selection criteria set out in the 2012 Common Approach. This approach allowed to safeguard the operational capability and effectiveness of the agency being established (at least, according to the Commission; see, however, the study requested by the BUDG Committee regarding the limitations of those criteria).
However, as highlighted in the preceding section this practice does not appear to give full effect to the constitutional principles upheld by the ECJ in EMA and ELA. In addition, the process risks tilting the delicate balance of the legislative process in favour of the Council, to the detriment of the Commission and the Parliament. Particularly from the perspective of an institution such as the Parliament – which has made the strategic exercise of its historically limited powers the cornerstone of its own development – this seems a (near) missed opportunity. Indeed, as noted above, formal parity between the co-legislators does not necessarily place Parliament on the “equal footing” with the Council that it has sought since 2019.
In light of the foregoing, it will be interesting to see whether, given its undoubted effectiveness in producing a rapid outcome, this voting procedure might be extended to other dossiers. For instance, it is worth noting that, around the same time as the joint vote on EUCA’s seat, the Council proposed that the co-legislators hold a “joint vote” to break the deadlock over the appointment of the new European Data Protection Supervisor, thus seemingly echoing the procedure used to select EUCA’s seat. It is also significant (and somewhat reassuring for Parliament, in light of the above) that the LIBE Committee’s coordinators rejected that proposal.
One cannot even rule out that the Council might be tempted to propose using this procedure again for other appointments – such as that of the EPPO’s European Chief Prosecutor or other top jobs in EU bodies, offices and agencies to be filled by common accord of the co-legislators – or for other strategic decisions within the legislative process, such as those depending on a prior Commission assessment of different options. If that were to happen, however, its flaws should be rectified, in particular by ensuring meaningful Commission involvement in decision-making, respect for the qualified-majority voting requirement applicable to the Council, and full compliance with the principles of transparency and openness throughout the process. Otherwise, the victory secured by the Parliament and the Commission in EMA and ELA might prove to be a Pyrrhic one.
Short bios
Jacopo Alberti is Associate Professor of EU Law at the University of Ferrara’s Department of Law. He holds a PhD in EU Law from the University of Milan and was a Junior Academic Visitor at Oxford. His research focuses on EU agencies, soft law, judicial protection, administrative adjudication and EU financial regulation. He leads the Jean Monnet Chair ‘Verba volant, sed imperant? The Legal Challenges of EU Communication’ and, previously, led the Jean Monnet Module ‘EU Specialized Judicial Protection’. He sits on the steering committee of the Review of European Litigation and of Eurojust and was National Rapporteur at the 2016 FIDE Conference. He is a lawyer at the Milan Bar.
Carlo Tovo is Tenure-track Researcher in EU Law at the University of Bologna. He holds a PhD in European Law and Public Law from the Universities of Bologna and Strasbourg (2015) and has been Postdoctoral Research Fellow in EU Law at the University of Bologna (2015-2018). He worked as a Local Parliamentary Assistant to a Member of the European Parliament (2010-2014) and served at the Court of Justice of the European Union, as a Legal Administrator (2018-2019) and Legal Secretary (référendaire) (2020-2024).