A Very European Way Out: Polity Maintenance and the Design of Article 50

Abstract

Multilevel polities do not typically facilitate secession, so why did the European Union adopt Article 50? Revisiting formative debates from the 2003 Convention on the Future of Europe, we combine archival research with an original dataset of delegate debates over two levels: the existence and procedural operation of an exit article. This reveals essential new detail on the genealogy of Article 50. We locate this institutional innovation within a Rokkanian–Hirschmanian theoretical framework which treats exit closure as necessary for loyalty and resilience. Further refining this ‘polity’ perspective, we find many participants showed awareness of the potentially disruptive implications of an exit article. Yet, given extant tensions around ‘ever closer union’, a Eurocentric procedural design prevailed as a safety valve, granting EU authorities default control over any exit process. This European logic of ‘controlled opening' offers a potential blueprint for other integrating compound polities and international organizations facing backlashes from member states.

Brexit brought a little-studied aspect of the European Union's constitution into focus: Article 50 (hereafter A50), the mechanism granting member states formal exit rights. This treaty provision surfaced in 2002, during the European Convention on the Future of Europe, the latter's task ‘to consider the key issues arising for the Union's future development and try to identify the possible responses’, particularly in view of forthcoming eastern enlargement (Fischer 2000: 2). The Convention produced a draft treaty, which was a key part of the Constitution for Europe in 2005. The retention of this withdrawal clause in the revived 2009 Lisbon Treaty prompts a puzzle: why would a draft aimed at binding the EU ‘ever closer’ explicitly grant exit rights?

Federalist literature suggests ex ante secession rights reduce incentives for compromise and encourage blackmail (Sunstein 1991: 634). For some economists, an absence of exit clauses is a ‘commitment device to increase … stability’ (Bordignon and Brusco 2001: 1812), while others argue that the risk of withdrawal jeopardizes sincere cooperation (Bolton and Roland 1997; Richardson and Stähler 2019). Inspired by Albert Hirschman's (1972, 1974) ‘exit, voice and loyalty’ framework, the polity-building literature – on which this article will draw – similarly concludes that external closure (not granting members a formal right to leave) is a precondition for internal structuring (creating authority structures and mechanisms for voicing grievances and potential reforms) and loyalty formation, which is in turn necessary for any nascent polity to endure (Bartolini 2005; Ferrera 2005; Rokkan 1974).

Against this backdrop, introducing an exit article appears counterproductive. Yet, with hindsight, we know pessimistic expectations were overblown. There is no evidence that A50 catalysed Euroscepticism or actively inspired the Brexit vote; indeed, the EU may even have become more cohesive after the UK's withdrawal (Ferrera et al. 2023). Can it be assumed that convention delegates were in fact prescient? Even at the time of the Convention, they were cognizant of rising Euroscepticism and its frequent casting of the EU as a supranational prison. Instead of opening the stable without considering a bolting horse, delegates might have calculated that unfastening the lock would release tension, calming soi-disant stablemates, while reassuring prospective central eastern entrants that they were not entering all-too-familiar supranational captivity (Huysmans 2019).

Examining the procedure in new detail, this article argues that A50 was designed, refined and adopted purposely to safeguard the EU polity at a critical juncture. Although A50 defies Rokkanian–Hirschmanian expectations, a modified polity-building perspective provides explanatory tools. We argue that the exit option was a bold response to the developmental challenge of EU consolidation and expansion without provoking sudden disruptions to mature member states. The integration project was launched to safeguard peace and generate prosperity for participating countries. The main challenge for this new meta-polity was and remains holding together without generating (perceptions of) harm and persistent resentment from its constituent units. When encountering this type of reaction, polity maintenance may require counterintuitive steps, such as divorce procedures. An orderly exit might be used to re-establish political calm or prevent disruption. While the European polity case is sometimes considered sui generis, with international politics increasingly characterized by nationalist threats to international governing organizations, this principle is parsimonious enough to have potentially wider comparative applications.¹

Combining archival accounts of the Article's Praesidium drafting process and a novel dataset of delegate interventions during the Convention's wider debate plenary phases allows for a comprehensive appreciation of agenda-setting and influence at key junctures. We reconstruct and analyse actors' interventions preceding and during the Convention, by breaking the debate over A50 down into two levels – existence and terms – and analysing both sources, we show the withdrawal clause was not negligence, but purposeful maintenance – a compromise avoiding a potential impasse or the twin perils of ‘Hotel California’ (members checking out cannot leave) and liberi tutti (unfettered exit).² The next section locates the novelty of our own empirical and theoretical approach. The third outlines the research strategy. Empirical results follow: an archival genealogy of A50, accompanied by an original dataset revealing delegates' positions and justifications on both levels. The conclusion summarizes findings and their wider implications.

Theoretical background: balancing closure and structuring

State of the art

The introduction of A50 is puzzling for many analysts. Following Hirschman, some fear that it undermines loyalty, with an explicit exit door reducing incentives for voicing complaints or suggesting repairs (Closa 2016). In the compound European polity where the voice mechanism is weaker than in national polities, A50 could weaken any transnational allegiance (encouraging criticism, partial exits), undermining governance (Sunstein 1991). Voice erosion can also turn into ‘demands for degradation of the Union’ (Closa 2017: 213). Consequently, Europhiles might see A50 as lose-lose: making exit more likely while debasing voice (Harbo 2008).

A50 scholarship has been chiefly concerned with such normative arguments, rather than empirical analysis of its existence and design. To our knowledge, only Martijn Huysmans (2019) pursues the latter, arguing that prospective states exhibiting marked differences from current members have incentives to seek exit assurances, especially if they expect to hold minority opinions frequently. A50 therefore placates the demands of ‘heterogeneous’ Central and Eastern European accession states, plus other wealthy, Eurosceptic outliers (the UK and Denmark).

However, tracing binary support–opposition positions leads Huysmans (2019: 170–171) to puzzling outliers: some delegates from European institutions and core countries favoured A50, while some accession delegates opposed it. Our empirical innovation traces the A50 debate across two levels: (1) whether an exit clause should exist; (2) if it does exist, on what terms? This second level penetrates the puzzle, addressing whether the exit process is controlled by either the departing state or the EU institutions and remaining members. This two-phased appreciation allows for more fine-grained coding of positions, capturing logics underlying A50's design. This can be explicitly illustrated by some theoretical expectations, to which we now turn.

Theoretical background

As it builds on national polities, the EU polity formation faces a ‘master tension’: weakening the closure conditions of constituent states without disrupting their internal structuring patterns (voice channels, binding authority and loyalty), while establishing a degree of pan-EU external closure and promoting the Union's own internal structuring (Ferrera et al. 2023). The hybrid federal–confederal nature of the EU makes its closure properties ambiguous. Enlargements are under the control of supranational authorities, though new accessions must be ratified by member state parliaments. Exit, however, remained a grey area, unacknowledged by treaties. Since 1969, withdrawal could theoretically be invoked under certain conditions using international law, per the Vienna Convention. However, the actual feasibility of this option was unclear (Gatti 2017).

Between 1992 and 2001, the EU witnessed three treaty revisions (Maastricht, Amsterdam and Nice) and a fourth enlargement wave (Austria, Sweden, Finland). This period is said to have ushered in the era of ‘constraining dissensus’ (Hooghe and Marks 2007). In many member states there were visible symptoms of integration fatigue and growing Euroscepticism. Some incumbents felt the EU was deepening too fast, while other candidates from Central and Eastern Europe feared accession required an irrevocable commitment soon after regaining national autonomy. The early 2000s were thus a critical juncture during which EU-building had to address yet another instantiation of the master tension, standing ‘at a crossroads, a defining moment in its existence’ (European Council 2001). Soon after the inaugural meeting of the Convention in February 2002, proposals started to circulate, including a novel withdrawal clause. What motivated its proponents?

To formulate expectations, we revisit Rokkan–Hirschman. This comprises a macro-framework of state-building dynamics that Stein Rokkan developed, resting on the micro-foundations of Albert Hirschman's exit-voice-loyalty scheme. Rokkan's general assumption was that to consolidate a new autonomous and internally structured political entity (acting ‘on a polity logic’), state-building elites must demarcate territory (Flora et al. 1999). The foreclosure of exit options for territorial actors ‘domesticated’ their strategies, orienting them towards central elites while eliciting the formation of alliances and organizations, in this way encouraging voice and the negotiation of political compromises. Closure thus serves a key structuring function. Under certain conditions, however, it can itself become a target of contention, both for outsiders seeking entry and insiders intent on exit. When the internal voice for exit becomes overriding, what are the options for leaders acting on a developmental polity logic?

Table 1 summarizes the primary logical alternatives: retaining closure or allowing for opening. Both options can in principle contribute to polity maintenance, ensuring relative membership stability and limiting or actively discouraging Eurosceptic exit contagion; or polity disruption, potentially facilitating a revanchist spiral whereby Eurosceptic forces across states are emboldened. It must also be noted that exit can happen in different ways. Rokkan argued, in fact, that what matters is not closure per se, but the specific design and control of closure conditions (Flora et al. 1999: 100–104). Accordingly, if a withdrawal option is granted, there are two possible modes of organizing exit, depending on who controls exit terms. For the EU context, stylized options are: departure under the control of the EU itself (pro-EU) or through a negotiation between equals, which would privilege the position of the lone departing member state vis-à-vis all remaining members and institutions (pro-state).

Table 1. Challenges to Closure: Logical Responses and their Potential Polity Effects

The main features characterizing the ‘crossroads’ moment of the early 2000s can be summarized as follows (Dimitrakopoulos and Kassim 2004; Laursen 2008). First, there were tensions between member states broadly interested in pursuing ‘ever closer union’ and those opposing this process. Second, Euroscepticism was rising in member states. Third, some accession countries harboured post-Soviet fears of deadlock after entry. Pro-EU elites faced delicate dilemmas. How to reconcile opposing integration preferences? How to deepen the EU despite member resistance, to widen it without frightening candidates? Should exit be constitutionally facilitated at all?

Based on the EU political context in the late 1990s, Table 2 summarizes stylized expectations about the aggregate preferences of various convention participants, albeit with ideological exceptions in each segment. Delegates from the UK and Denmark can be expected to have advocated exit options and a mode of exit based on negotiations among equals (Huysmans 2019). These two countries had always been rather lukewarm towards integration, being the two most significant critics of aspects of Maastricht, and with both seeing Euroscepticism on the rise in the second half of the 1990s. Delegates from other member states can be expected to have supported closure or, at most, opening under the control of the EU. Candidate countries can be expected to have supported the introduction of an exit clause, and probably an EU-controlled type of withdrawal (e.g. in order to restrain the risk of nostalgic radicalizations).

Table 2. Expected Preferences of Convention Delegates by State Origin

While always potentially operative, the polity logic remains dormant during normal phases, driven by ordinary contests about policies. At critical junctures, however, the polity logic becomes dominant and, with it, the issue of boundaries. In the following sections, we gauge the extent to which the choice constellation we have derived from the Rokkan–Hirschman model played out during the Convention. To what extent were architects' and delegates' designs congruent with expectations? Our research strategy describes how these questions are addressed.

Research strategy

Our explanatory strategy is based on ‘retrospective causation’ (Kalberg 1994). The European Convention was a bifurcation point for basic EU boundary configuration (Table 1): the latter could remain closed or be opened. The second option would originate a further bifurcation, regarding the mode of exit: controlled by the EU authorities or negotiated between equal partners. Retaining closure was an objective possibility, given typical path dependence dynamics and lock-in effects. With hindsight, we know however that the option to withdraw was adopted, and in a pro-EU manner. Our explanatory account must therefore identify the factors that pushed towards the inclusion of exit – considering the possibility that its pro-EU design was key in securing the first-level acceptance of an exit article. We focus on agent-level observable behaviours and stated motivations on the assumption that other, more distal and structural factors (including heterogeneity à la Huysmans) operate to the extent that they actually shape – directly or indirectly – individual motives for action (Davidson 2001).

Empirical analysis first reconstructs the debate concerning the inclusion of an exit clause into the treaties. Why and how did the basic bifurcation come about, with the possibility of exit reaching the agenda? For this, we draw on archival sources, press coverage and secondary literature. In a second step we systematically map deliberations concerning the exit clause during the Convention proper. What positions did actors take vis-à-vis the basic bifurcation? We provide data on both positions and stated justifications for these positions, where available, from two sources, available online: (1) amendment forms submitted by delegates to the Praesidium concerning the draft article; (2) verbatim transcripts of delegate speeches at the Convention's plenary session on 24–25 April 2003, when the exit clause was debated. We hand-coded each intervention along several dimensions. This yielded a total of 190 coded interventions, from 47 unique amendments and 42 speeches, delivered by 123 unique delegates. Some were solo initiatives, but most were co-signed by multiple likeminded delegates spanning institutions or countries. Further description of coding and the dataset is included in the Supplementary Material as an online appendix.

Analysis: Why was an exit right included in the EU treaties?

Proposal origins

In 2001, Franco-German and Anglo-Swedish pairs of leaders sketched out opposing federal and intergovernmental visions of the EU (cf. Blair and Persson 2000; Fischer 2000). In line with this debate, the 2001 Laeken Declaration called for ‘a democratic and globally engaged union’, without creating ‘a European superstate’ (European Council 2001). The Rome Treaty had not envisaged withdrawal, and exit from international organizations had been implicitly regulated by the United Nations' 1969 Vienna Convention on the Law of the Treaties. However, it remained unclear if the latter could apply to the modern EU, creating the possibility for disorderly and disruptive secessions (De Waele 2015).

The Laeken Declaration established a European Convention to discuss a possible EU constitution, with former French President Valery Giscard d'Estaing (2003) as president; Giuliano Amato and Jean-Luc Deahene the vice-presidents. The Convention comprised 28 representatives of heads of state or government, 56 representatives of national parliaments and 18 representatives from European institutions. The aim was to give ‘recommendations if consensus [was] achieved’ on how to further the EU polity.³ In practice, Giscard d'Estaing assumed a strong agenda-setting position. He refused to define ‘consensus’ but did set tight deadlines for amendments, maintaining a free hand in crucial decisions (Bruton 2004; Tsebelis and Proksch 2007). The Convention's working method was based on a soft legislature–executive type interplay between the plenary, where all convention members participated, and the work of the Praesidium, which would ‘listen’ to plenary debates, received individual contributions and created coherent drafts.

Exit article design was a five-step process: first, representatives in the Praesidium provided inputs on why A50 merited constitutional recognition. The president, based on the Praesidium's opinions, decided whether to include the article in the Treaty architecture. An amendment phase was then initiated, succeeded by an open plenary debate that allowed participation from all convention delegates. The Praesidium would then proceed to revise the draft document, incorporating the consensus reached during the plenary discussion and taking into consideration any proposed amendments. A smaller, second round of amendments would then ensue, followed by another plenary debate, whereby only three speeches concerned the exit clause. Finally, the Praesidium would be tasked with composing the definitive text of the article.

The withdrawal clause formally surfaced in autumn 2002 within four ‘free initiatives’ submitted to the Praesidium by convention delegates (Table 3). First, Andrew Duff, a UK Liberal Democrat from the European Parliament, drafted the possibility for ‘member states to withdraw from the union on terms to be agreed with the union’ (European Convention 2002a, 2002c). The Eurosceptic Daily Telegraph (2002) had suggested unregulated exit would be easier and more democratic, since member-state governments have electoral mandates. Duff countered that unilateral exit would guarantee ‘maximum disruption for everyone concerned’ (The Times 2002). A Euro-federalist, Duff couched his withdrawal clause in polity maintenance terms (see Table 1): allowing the departure of recalcitrant members facilitated restabilization, possibly unlocking progress.

Table 3. Timeline of Withdrawal Clause Proposals

Note: / indicates no clear stated preference. Adoption and signing processes carried over from the final Praesidium draft without amendment or scope for further debate and justification.

Alain Lamassoure, French delegate from the European Parliament, Union for a Popular Movement, followed. Lamassoure suggested confederal, federal and ‘mixed’ polity models were all ‘inadvisable’. A fourth, ‘Community’, arrangement was preferable, based on a single executive with a president democratically appointed by both Parliament and Council. This would envisage withdrawal rights, which would be subject to ‘strict and deterrent conditions’, but would ‘remove the sting from the criticism levelled by those who reject the concept of the “superstate”’ (Lamassoure 2002: 12). Lamassoure argued closure would favour voice for exit, corroding loyalty (Table 1, cell 4). A withdrawal right would neutralize Eurosceptic ‘superstate prison’ arguments. In sum, Lamassoure took a clear position on the mode of exit, reflecting polity maintenance concerns related to possible opportunistic misuses.

Three other proposals contained clauses with modes of exit. Peter Hain's (UK government, Labour) proposal countered the ‘deterrent conditions’ advanced by Lamassoure and Duff, claiming that ‘a Member State does not need “permission” to withdraw from the Union’ (European Convention 2002b: 47). This was the expressly pro-state draft. Conversely, the proposal by French parliamentarian and constitutional expert Robert Badinter (Socialist Party) strengthened the role of EU authority by introducing European Court of Justice arbitration into any withdrawal process. The Penelope draft of Romano Prodi's Commission, presented in December 2002, allowed withdrawal ‘when the Constitution or its additional acts are revised’, and only ‘after a period of two years’ (Lamoureux 2006). The Commission motivated the withdrawal clause with polity building at critical junctures: it would allow other states to overcome integration vetoes by recalcitrant members, facilitating equal-status negotiations between the EU and the departing member state.

Elmar Brok's (Germany, Christian Democratic Union, European Parliament) draft did not contain the withdrawal clause because it would undermine ‘mutual obligations of solidarity’. Moreover, Brok (2002) argued that if a clause was to be included, then ‘also the possibility of expulsion of a member would have to be considered’. His preference for continued closure was motivated by the goal of safeguarding the solidarity foundations of the EU and the stability of the single currency (see Table 1, cell 1).

These initial proposals expressed diverging goals and concerns about both the ‘if’ and the ‘how’ of an exit clause. They disagreed on the possible effects of regulated exits: for some it would help advance the EU by cornering Eurosceptics; for others it would provide reluctant member states with disruptive veto powers; for others still, it was instrumental to deliver differentiated integration. Against this contested backdrop, Giscard d'Estaing included a clause (Article 46) in the Praesidium draft (28 October 2002) containing the ‘architecture’ of the future Constitution. After Brexit, Giscard d'Estaing explained his choice: ‘in the early 2000s, there was a campaign by the American press saying that the EU was a prison …. I said to myself that it was necessary … to provide for a possibility of leaving under legal, diplomatic condition’ (Le Parisien 2019). This version is also espoused by other relevant Praesidium actors, such as Duff and Lamassoure (Duff 2016; Le Journal du Dimanche 2016). However, as Table 3 summarizes, despite these differences and with the exception of the proposal from UK government delegates Peter Hain (Labour) and Alan Dashwood (a constitutional scholar), the general theme of their arguments was that a clause should exist with a pro-EU design, following the logic of ‘opening for maintenance’ in Table 1. Finally, this reconstruction is underscored by what the procedurally powerful President Giscard d'Estaing himself argued when the draft first circulated: ‘if a country, following a democratic consultation, wants to leave the Union, there is no reason to force it to stay. The EU will never be a prison. But if a country decides to leave, it will be necessary to make a fair assessment of the conditions of its departure’ (Agence Europe 2002). By then, this was the ‘default’ position that was taken into the plenary debates.

Positions of convention delegates

Trends in positions

We now turn to our dataset based on the amendments and speeches of the delegates. After the Chair took the decision to include an exit clause in the draft treaty, the Praesidium published draft terms in a document dated 2 April 2003 (CONV 648/03). Delegates reacted by annotating the article text. Some amendments were signed by individuals, but more frequently they were co-signed by multiple participants, revealing long-standing coalitions and alliances, with a view of prompting the Praesidium to modify the original proposal. The Convention Secretariat summarized the main points of the delegates' amendments in a document dated 14 April 2003 (CONV 672/03).

For written amendments, delegates took positions at two levels. First, the fundamental question of whether an exit provision should be included. The ‘Delete versus Amend’ bars of Figure 1 group interventions by delegate state and institutional affiliation for all interventions (amendments and speeches). Figure 1 shows 37 delegates proposing full deletion, coming disproportionately from Greece, the Netherlands, Austria and Portugal. These breakdowns accord broadly with expectations (Table 2), including no Danish or British delegates. However, France and nine post-socialist accession states' delegates did so, contradicting any assumption that the latter might be uniformly supportive of an exit clause. The French position might be accounted for by the contemporaneous success of domestic Eurosceptics. However, the nature of the French interventions will shortly reveal more about delegates' intent (cf. Figure 2). Joschka Fischer (Green) is the single German government delegate proposing deletion.

Figure 1. All Amendments – Retain vs Delete

Figure 2. Retain Amendments – Pro-EU vs Pro-State

All nine accession deleters were affiliated to an active centre-right European People's Party (EPP) delegation, which submitted a joint proposal comprising 22 co-signatories, drawn from 15 member and candidate states (see Table A1 in the Supplementary Material for a partisan breakdown based on European Parliament affiliations and ideological traditions). This group includes eight national parliamentarians and the Latvian government delegate, Roberts Zile. Despite this prominent accession presence, its rationale prioritizes concerns over a potential erosion of solidarity and a reversal of ‘ever closer union’. Overall, despite potential associations with greater levels of Eurosceptic nationalism in some quarters than their social democratic rivals, centre-right party-affiliated delegates are responsible for 23/37 (62%) of total deletion requests, an outsized contribution even when accounting for centre-right delegates being the plurality overall.

Acceptance of an exit article present in the remaining 82.5% of all amendments should not be taken as being synonymous with pro-exit Euroscepticism. Recall, second-level positioning refers to whether its design should strengthen or moderate the rights and obligations of the EU and remaining states vis-à-vis the departing state. There was intense debate and multiple amendments in both directions, albeit not equally split. A clear majority (86% of amendments) emerge for pro-EU positions. Figure 2 shows the distribution of amendments grouped by this divide, summarized qualitatively in Table A2 in the Supplementary Material.

In terms of geographical and institutional dispersions, pro-EU and pro-state factions break down in a similar fashion to the first-level debate: Danish and British; Italian delegates are more divided, with the latter notably more pro-state; while Austria, Belgium, Germany, France, Greece and even the Nordic states of Sweden and Finland skew more heavily pro-EU. Curious, and challenging Huysmans's (2019) heterogeneity theory, is the trend among accession states: a general acceptance of an exit clause but on pro-EU terms. Accession delegates seek almost exclusively to empower the EU at the expense of exiting states, although the Slovak and Turkish governments dissent. This is again driven predominantly by the co-signing group of typically around 20 EPP affiliates and members who promoted quite mild pro-EU qualifiers. Combined with our analysis of justifications below, this suggests that candidate states were not motivated by the notion of plausibly exiting, but instead wanted a basic guarantee of the right, while still showing ex ante concern for polity maintenance.

Along partisan and institutional lines, the distribution skews pro-EU across all groups, even among left and right partisans and especially national parliamentarians. This indicates that a broad consensus emerged among delegates that the article should exist, but that it also might be used to control the means of exit and strengthen the procedural hand of EU institutions and remaining states.

Intent of interventions

More qualitative detail about the nature of the amendments further reveals this. A particular pro-EU focus was lent to the first of three clauses, which addressed basic exit rights. The primary goal appeared to align exit rights to treaty or constitutional moments, allowing a member to leave only if they failed to accept onward integration. Table A2 shows 20 signatories to this notion, including the French and Belgian governments. Other amendments sought assurances that leaving states were bound by the provisions of the entire article. Pro-state interventions were instead primarily focused on the second clause, which addressed the institutional handling of the exit procedure (see Table A3 in the Supplementary Material). With hindsight, it is clear from the UK's experience that the Commission was strongly empowered to coordinate Brexit negotiations, with the UK locked out of Council meetings. At the Convention, Eurosceptic contributors such as David Heathcoat-Amory (UK Parliament, Conservative) and Jens-Peter Bonde (Denmark, MEP, June Movement) jointly proposed keeping departing states present during other states' talks on exit terms, but like pro-European interventions in the first paragraph, they failed to secure changes to the original Praesidium draft.

The final clause addressed timing and duration, with pro-EU forces seeking to erase the two-year limit so that negotiations could run indefinitely and to impose time limits for re-entry (Lamassoure proposed a 20-year moratorium). Meanwhile, pro-state delegates such as Bonde and Heathcoat-Amory sought to reduce the limit to one year, terminating membership more swiftly, if an agreement was absent. This was also proposed by Pál Vastagh, from the Hungarian parliamentary Socialists, whose only other amendment was to suggest a pro-EU ban on re-entry for five years. This implies a somewhat complicated scenario whereby the same amendment could have been motivated by both pro-EU and pro-state sentiments (hastening exit being seen as positive for the leaving state but also the EU). Unfortunately, not every amendment was explicitly justified (see below), and debates over the time limit should not be interpreted as indicative of one tendency being dominant over the other here.

The pro-state group also sought to ensure that the Council operated by unanimity rather than qualified majority voting, to guarantee individual states could veto any agreement. This amendment was posted by an Irish government representative, Dirk Roche. In the end, none of these amendments was adopted but a single substantial edit was made from the original Praesidium draft: the Council could decide to extend the two-year deadline.⁴ This clause was famously deployed three times during Brexit negotiations. Full details on amendments are outlined in Tables A2 and A3 in the Supplementary Material.

Justifications

Having mapped the positions of the participants, we now focus on their motivations, where stated. Given that the exit clause was included in the initial draft text, this pushed opponents to frame their arguments in negative terms, focusing on the perils of opening rather than the positives of closure (Table 1). Table 4 presents these arguments. For each contribution we coded up to two arguments, adding up to a total of 164 arguments recorded for written amendments and 65 for speeches. While almost all parliamentary speeches contained codable arguments, under half of annotated amendments were justified. Silence was similarly distributed across the different participants and positions, though those opposing the inclusion of the exit clause were somewhat more inclined to explain why.

Table 4. Delegate Arguments and Codes

Note: Codes in italic.

In the written amendments, arguments focusing on positive versus negative implications of including an exit clause tilted towards the former, but the opposite is true for the parliamentary debate. The relative salience of the arguments also differed between the amendments and the debate (Figures 3 and 4). In the amendments, considerations related to the nature of the EU polity were pre-eminent. The main principled argument against the inclusion of the exit clause was that it was antithetical to the nature of the EU, and that it would compromise the project of ever-closer union (Undermines polity building). Gijs de Vries (Dutch government, People's Party for Freedom and Democracy) opposed the exit clause in the plenary debate on such grounds:

an exit clause does not belong in this Constitution because it would change the character of the Union … into a mere union of States that you can leave at any time. There is a crucial difference between the International Postal Union and the European Union. The European Union is not just a Union of States; it is also a Union of peoples. To reconcile these peoples and to unite them in a common political union is the essence of European integration. We are not just creating a common market, an economic organization that you can join or leave at will; we are creating a political union.

Figure 3. Amendment Arguments, by Category

Figure 4. Speech Arguments, by Category

Among the most prevalent arguments in favour of opening concerned implications for the EU's long-term development. This sometimes cast the exit clause as part of a broader project to build a more flexible, resilient polity, including proposals for coupling (or substituting) the exit clause with the creation of a looser kind of association with the EU. For example, some delegates advocated for new ‘associate membership’ status or that any right of withdrawal must be complemented by expulsion powers. As noted, others argued that granting exit rights would unshackle those intransigent member states who did not subscribe to the direction of integration to exit, to the benefit of all parties (Off-ramp).

In both the amendments and the parliamentary debate some delegates warned that the inclusion of the exit clause would have destabilizing effects on the EU (Threat to stability), or that it was unnecessary given the Vienna Convention. Despite expressing reservations, however, delegates did not always oppose the inclusion of the exit clause entirely, but rather stressed the importance of maximizing EU control over the process given potential risks.

In the debate, arguments relating to EU polity building were complemented with more political considerations: threats from Euroscepticism and the need for democratic legitimation. The preoccupation with Eurosceptic tendencies harkens back to the motivations of the exit clause's architects. Here, once again, participants were split between a majority arguing that the exit clause would neutralize Eurosceptics, and a minority arguing that it would further empower them (Disarms Eurosceptics vs Enables Eurosceptics in Figures 3 and 4). Many of those in the pro-integration camp embraced the exit clause to respond to Eurosceptic tendencies back home. Over-represented among them were delegates from Denmark and the UK. In their view, an exit clause would disarm Eurosceptics, proving that membership is entirely voluntary. Others were concerned that an exit clause would simply encourage Eurosceptics, becoming a self-fulfilling prophecy. As Jürgen Meyer (German parliament, Social Democratic Party) explained:

The exit clause in Article 46 is controversial. I propose deletion here too. In my opinion, … every member of the Union is actually able to leave. … I believe that what is in any case actually and legally possible should not be offered on a silver platter to Eurosceptics who want to leave, and who then create constant unrest in their national parliaments and in the public sphere whenever any difficult decisions are proposed by Brussels, by saying: We won't put up with that, let's quit!

The handful of Eurosceptic delegates supported the exit clause and justified their positions mainly by appeals to the EU's (lack of) democratic legitimacy. Bonde said:

Now the time has come when those who want it can propose this ‘wanted’ federation to their people. … Only referenda will really engage our citizens in the European debate and make them aware of the constitutional process. Let us build the People's Europe by asking the people of Europe first.

This argument (Legitimation) accounted for a third of justifications among those who took a pro-state position, but it was also advanced by Europhiles. Several contributors from both member and accession states supported an exit right as evidence of the EU polity being a union of democracies resting on popular consent. This finding is at odds with the Rokkan–Hirschman model, which generally assumes structuring will occur in response to closure, neglecting the fact that – at least in the era of mass democracy – authority cannot simply be claimed and enforced through boundary closure, but must be earned in the eyes of citizens/voters (Ferrera et al. 2023). Like arguments related to Euroscepticism, this perspective was almost entirely absent from the amendments, but took centre stage in the debate session. On the one hand, in a context of dialogue and deliberation more political considerations were likely to have a greater resonance and, on the other hand, the discussion was dynamic and delegates with different national and political backgrounds were able to challenge one another's position. As Liene Liepina, a Latvian parliamentarian from New Era (centre-right) explained:

The article is necessary, as we have heard so often here, for one reason: it underlines the freedom of each country to join the Union as well as to leave it. In Latvia we say: Do we really want to go from one Union to another? Of course, there is a huge difference between a people deciding in a referendum to join a union and a so-called representative of an occupied country deciding this. Nevertheless, for us the freedom and independence so recently acquired is of very great importance.

In this sense, the inclusion of the exit clause facilitated enlargement by placating concerns that may have arisen in former socialist countries about the EU being a Soviet-like prison – a consideration that did not explicitly emerge during initial drafting. This consideration, however, was implicitly recognized in the constitution draft which followed the plenary debate. In the draft, the Praesidium commented that A50 ‘is an important political signal to anyone inclined to argue that the Union is a rigid entity which it is impossible to leave’ (European Convention 2003: 134). A breakdown of justifications by member-state groups (Table 5) shows a strong concern with democratic legitimation in Central and Eastern European candidate states, especially the Baltics (unlike the non-post-communist candidates Cyprus, Malta, Turkey),⁵ but also in states with strong Eurosceptic influences (the UK and Denmark). In the aggregate, however, the arguments advanced by delegates from candidate and member states were very similar, with post-communist countries emphasizing proportionately slightly more aspects of legitimation and polity flexibility, and existing member states showing relatively more concern with polity-building and arguing the article was unnecessary. The UK–Denmark pair stand out, with delegates not advancing any arguments against the exit clause. In conclusion, both delegates from countries with strong Eurosceptic tendencies and delegates from candidate states supported the inclusion of the exit clause but for somewhat different reasons, the latter acknowledging much more clearly the trade-off between closure and opening.

Table 5. All Intervention Arguments, by Country Group and Category (% of All Arguments Advanced by Country Group)

Note: Contains all delegates, speeches and text amendments, irrespective of institutional affiliation.

CEE accession: Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia, Slovenia.

Other accession: Cyprus, Malta, Turkey.

Other member states: Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden.

Summary of article evolution

Before concluding, it is instructive to briefly reflect on the changes to the article proper over time. Table A5 in the Supplementary Material reports the text of the withdrawal clause as formulated in the Convention draft (Article 46) and shows subsequent additions made through the Convention (Article 59), the Intergovernmental Conference establishing a Constitution for Europe (Article I-60), and finally the completed text contained in the Lisbon Treaty (A50). Amendments are largely cosmetic, and the article appears ‘frozen’ after 2004, with no meaningful change through 2009. This indicates that the overall tenor of the debate did not shift the exit clause architects' original intent significantly, validating their essential pragmatic pro-EU motives as described previously. We see this motivating logic prevailing in both sections of our empirical data: archival tracing of versions and the plenary debate dataset. More substantive additions clarified the departing state being excluded from Council talks concerning it, which was debated at the plenary; the possibility to extend the two-year negotiating period by unanimity only; and the need to reapply for membership should the withdrawing state wish to rejoin in the future. Their purpose was to prevent the withdrawing state from delaying negotiations or triggering exit for opportunistic or instrumental reasons (EPRS 2020). Following the rejection of the Constitutional Treaty by French and Dutch referendums in 2005,⁶ a new intergovernmental conference was convened in 2007, which led to the adoption of the Treaty of Lisbon (entering force 1 December 2009). The exit article was carried over into this treaty to little fanfare, and did not feature among contemporary debates and controversies over the ratification of Lisbon itself (Ziller 2019).

Conclusion

Why would a multilevel polity like the EU introduce an exit clause if its aim is to ensure ‘ever closer union’, and what does this reveal about international cooperation more broadly? Findings indicate that a rounded ‘polity logic' motivated EU members to introduce a device that should stabilize the union by demonstrating its openness while granting it control over exit procedures in case of a rupture. Empirical findings align with a modified polity argument in several respects. First, having traced the genealogy of A50, we show how constitutional drafters deployed polity-based justifications. These concerned disarming Eurosceptics and ensuring that integration could proceed without any potential disrupters. We then looked at the positions of delegates within the Convention and found that support for A50 came in two variants: pro-state or pro-EU, the latter dominating. The Vienna Convention offered no institutional roadmap for negotiations, and did not foreground EU institutions (Gatti 2017: 169). This could have encouraged divide-and-conquer strategies that would have empowered exiting states and created coordination problems. Some of those reluctant to support A50 dropped their opposition because of such fears.

Analysis of justifications provided in amendment and debate stages by delegates support and refine the polity logic of A50. The main pro-clause arguments were the creation of a flexible, resilient polity and an ‘off-ramp’ in case of unwanted deeper integration. Overall, national positions were broadly in line with our expectations, albeit with a few informative surprises. When they were not silent, candidate states were frequently supportive of A50, combined with a pro-EU exit mode: in principle they did not want a novel status of irreversible captivity, but at the same time they took accession as a serious commitment, to be protected from contingent and/or opportunistic nationalist backlashes.

Theoretically, this article sheds light on blind spots of the heterogeneity argument (Alesina and Spolaore 2005; Huysmans 2019). Core member states – including France, Sweden, Belgium and Finland – did in fact support A50. Explaining this position introduces a second-level debate over the exit process: some actors supported exit rights precisely because they thought it strengthened the EU's hand, should a member decide to leave. Findings reveal that delegates from candidate states – including Poland, Slovakia, Bulgaria, Turkey, Romania and Hungary – frequently supported similarly conditional opening and even sometimes full closure. Here, pro-state amendments were championed chiefly by British–Danish initiatives. Additionally, historical reconstruction illuminates a key detail: the exit clause entered the agenda via the (discreet) initiative of British, French and German political figures. In the end, the inclusion of A50 in the Praesidium draft was decided by key Europhile leaders from core member states, most notably Giscard d'Estaing.

Second, our perspective moves beyond static assumptions of economo-federalism, according to which secession clauses threaten stable economies of scale. The EU is not a fully fledged federation but a sui generis compound polity (Ferrera et al. 2023). Its constituent units are mature nation states that consider trade-offs en route to one of multiple possible endpoints. The need to safeguard development at any given stage while facilitating further polity building rationalizes allowance for exit. The withdrawal of a member state challenging polity foundations and membership terms can release tension, refocusing voice towards policy rather than polity debates. A single departure need not create a domino effect. From a Eurocentric perspective, to minimize disruption and uncertainty, institutional power dynamics should privilege the extant polity authority in any withdrawal process. We have shown that logic was operative and was carried into A50, in particular via the terms empowering the Council and locking the exiting state out of discussions. Scholars have since noted this playing out in practice during the Brexit negotiations (Craig 2017; Vollaard 2018).

Finally, the salience of democratic legitimacy motivations invites consideration of a polity-building mechanism that Rokkan neglected: in a mass democratic context, the boundary configuration of a novel multilevel polity invariably becomes subject to democratic stresses. External closure remains an important condition for internal structuring and the exercise of authority, but under certain circumstances, political contestation may come to challenge territorial and membership boundaries. For a polity-in-the-making, this type of contestation activates what we have earlier called the ‘master tension’ between supranational structuring and national de-structuring.

This story is in certain senses peculiarly European, an output of the union's own developmental trajectory and tensions. However, the EU is but the leading example of interstate cooperation subject to politicization and exit threats (De Vries et al. 2021). In this climate, there is no reason why the same responses from the Convention coalescing around ‘ordered exit’ cannot be applied to other international agreements and bodies facing potentially destabilizing exit threats. We argue that this European logic of controlled opening offers a potential blueprint for compound polities and international organizations facing backlashes from member states. Studies in international organization (Koremenos and Nau 2010) and federalism (Huysmans and Crombez 2020) underline that exit can be a potent device if it increases welfare for both parties, reducing the potential for contagious conflict, and if it is expressly designed with the protection of the greater union in mind.