External drivers of EU differentiated cooperation: How change in the nuclear nonproliferation regime affects member states alignment

ABSTRACT

Since its establishment, the Common Foreign and Security Policy (CFSP) has strived to increase convergence among EU member states. Yet, convergence remains elusive and scholars have started to explain the emergence of differentiated cooperation resulting from multiple internal EU crises. We posit that the convergence in the EU member states with respect to nuclear weapons has been fundamentally altered by the humanitarian turn to nuclear disarmament. This has led to a crystallization of differentiated subgroups among the member states, whose membership coincides with that of informal groupings active in the broader nuclear nonproliferation regime. Combining quantitative data on resolution sponsorship at the Non-Proliferation Treaty review process and voting at the UN General Assembly, we show that significant change in the international nuclear nonproliferation regime led to differentiated cooperation within the CFSP, resulting in two cohesive subgroups of member states.

KEYWORDS:

• Differentiated cooperation
• CFSP
• differentiation
• NPT
• nuclear weapons
• TPNW

Since its establishment, the Common Foreign and Security Policy (CFSP) has pursued the aspiration of fostering alignment among the foreign policies of member states of the European Union (EU). Yet, despite some progress toward this objective, the ability of member states to act together has stagnated or even declined, especially in international fora like the United Nations (UN) (Dee & Smith, 2017; Smith, 2017). The study of the CFSP has increasingly endeavored to accommodate dynamics that depart from the expectation of progressive, linear integration in which all member states participate (Siddi et al., 2022). As part of these efforts, researchers have shown a growing interest in borrowing the recent notion of “differentiation,” developed in the more mature subfield of EU governance. Distinct from the previously-established concept of “differentiated integration,” where a majority of member states launches an integration project in the expectation that the rest will follow at a later stage, “differentiation” refers instead to the emergence of quasi-permanent groups with divergent preferences in a given policy field (Bátora & Fossum, 2020).

To be applied to the CFSP, the notion of differentiation must take into account its specificities as a policy framework (Amadio Viceré & Sus, 2023). The fact that the CFSP remains largely member states-driven sets it apart from other EU policy fields that have experienced some integration into supranational governance and/or have become subject to formalized flexibility mechanisms (Fabbrini, 2021). The novel concept of “differentiated cooperation” reflects the nature of the CFSP as a non-integrated regime governed by unanimity where informality flourishes, as detailed in the introduction to this special issue (Amadio Viceré & Sus, 2023).

The concept of differentiation, in the distinct shape of differentiated cooperation, finds increasing application in EU foreign policy (Siddi et al., 2022). However, because the CFSP differs from other fields of EU governance in that it lacks a supranational authority to hold it together and is embedded in broader global regimes, the CFSP can be assumed as vulnerable to challenges as supranational policy fields—if not more. Does differentiated cooperation in the CFSP follow similar drivers observed in EU internal governance? With its emphasis on homegrown crises, the notion of differentiation tends to neglect the possibility that alterations originating in the global regime may disrupt the achieved level of integration. Because of its external linkages and embeddedness in global frameworks, we expect the CFSP to be as vulnerable to regime-wide crises as to internal challenges. Thus, in this article, we set out to explore whether differentiated cooperation in the CFSP can result from crises external to the EU, rather than from internal challenges at the root of differentiation in internal governance.

In order to investigate whether differentiated cooperation is driven by significant external changes in international regimes, we examine the case of EU alignment in the nuclear non-proliferation regime. We select this case for several reasons. Firstly, coordination on nuclear non-proliferation and disarmament is traditionally challenging for the CFSP due to diverse attitudes among EU member states. While France and the United Kingdom (UK) are nuclear-armed, Sweden, Ireland, and Austria have long championed nuclear disarmament (Jonter & Rosengren, 2014; Portela, 2004; Portela & Raube, 2012).¹ Most remaining EU members belong to the North Atlantic Treaty Organization (NATO), which embraces nuclear deterrence (Nuti, 2021). EU voting behavior at the United Nations General Assembly (UNGA) therefore unsurprisingly reveals that nuclear weapons constitute one of the most divisive topics among EU members (Burmester & Jankowski, 2018; Jørgensen, 2009; Luif, 2014). Secondly, since the nuclear non-proliferation regime has undergone significant change recently, it can help us test if regime-wide challenges lead to differentiated cooperation in the CFSP. Thirdly, the overlapping membership of some EU members in pre-existing, informal groupings within the regime context facilitated differentiation. The presence of informal groupings that bring together some but not all EU members with like-minded non-EU states is absent from other international regimes. We identify the development of the Humanitarian Initiative (HI) for nuclear disarmament which unfolded between 2013 and 2015 and the ensuing Treaty on the Prohibition of Nuclear Weapons (TPNW) of 2017 as a significant alteration in the regime. We examine data from two key fora, UNGA First Committee on Disarmament and the Review Process of the Nuclear Non-Proliferation Treaty (NPT), organized in quinquennial Review Conferences (RevCons), covering the period that saw the formation of the HI and led up to the signing of the TPNW.

We find that CFSP action in nuclear nonproliferation and disarmament gave rise to differentiated cooperation following the development of the HI and the ensuing TPNW. A picture of well-delineated, cohesive groups emerged reflecting a pattern of differentiated coordination, a situation that differs from the previous, fuzzier pattern, and from one of segmentation that might have emerged in its stead. This shows that differentiation in the CFSP can originate from regime-wide alterations, and need not respond exclusively to the homegrown crises that drive differentiation in internal EU governance. Our findings contribute to EU governance scholarship by identifying change in multilateral regimes as a driver of differentiated cooperation in the CFSP. In addition, our exploration contributes to the broader study on the EU’s role in international regimes, especially in the UN setting. Standard works on UN fora analyze EU cohesion and the outcomes of the EU internal coordination (Blavoukos & Bourantonis, 2017; Jørgensen, 2009; Oberthür, 2011; Panke, 2014), often focusing on their effects (see also da Conceição-Heldt & Meunier, 2014). Instead, we look beyond the cohesion and actorness to leverage the novel concept of differentiated cooperation. By considering indicators other than voting behavior, this article complements methodologically extant works exploring voting patterns of EU member states (Burmester & Jankowski, 2018; Luif, 2014; Panke, 2017b; Panke et al., 2019).

Differentiation in EU foreign and security policy

The Treaty on European Union (TEU) signed at Maastricht in 1992 introduced modalities for differentiated integration, defined as legally valid rules that exempt or exclude individual member states from specific rights or obligations of EU membership (Schimmelfennig et al., 2021). Differentiated integration was made possible through various mechanisms in order to accommodate diverse “stages” of integration, allowing members ready to integrate more closely in a particular domain to form an avant-garde group, obviating the need to wait for others, which would join the integration project later. The eruption of multiple crises shaking the EU over the past decade, such as the sovereign debt crisis, the refugee crisis and the British withdrawal from the EU (Webber, 2019), changed this understanding. Their combination, labeled as “polycrisis” (Bátora & Fossum, 2020), spurred the continent-wide proliferation of populist political parties which question, and purport to backtrack from, the achievements of European integration. The concept of differentiation entails a scenario of multi-speed Europe but relinquishes the temporary limitation foreseen under differentiated integration: Different levels of integration are conceived as permanent rather than temporary statuses.

The notion of differentiated integration accompanied the CFSP from its emergence in the Treaty of Maastricht, which left the foreign policy domain in the intergovernmental realm (Forster & Wallace, 1996). Yet, flexibility mechanisms remained scarcely utilized until the Lisbon Treaty, illustrating that member states valued the CFSP as a vehicle to “speak with a single voice” (da Conceição-Heldt & Meunier, 2014) and saw any formalized division within the CFSP as undermining its raison d’être (Kielmansegg, 2017). Has the move from differentiated integration to differentiation occurred in the CFSP in the same way as EU internal governance? While differentiation in the EU has mostly been considered in the context of formal integration (Kielmansegg, 2017; Schimmelfennig et al., 2021), it is increasingly present in the CFSP. Similar to EU governance, differentiation in the CFSP helps accommodate varying preferences of member states in foreign policy (Rieker, 2021), replacing Maastricht’s original aspiration of fostering integration by means of intergovernmental coordination (Genschel & Jachtenfuchs, 2014). Yet, the notion of differentiation requires some conceptual adjustment to fit the CFSP context. Characterized by the absence of supranational governance, the CFSP rarely experiences formalized differentiation (Siddi et al., 2022). Instead, differentiation in this policy realm takes the form of differentiated cooperation, where separate subgroups are recognizable but not formalized. The concept of differentiated cooperation thus captures the informal nature of cooperation between EU member states in diverse settings, as detailed in the introduction to this special issue (Amadio Viceré & Sus, 2023). Because of its intergovernmental nature and unanimity requirement, the CFSP has been singled out as especially vulnerable to populist pressures (Leruth, Gänzle, & Trondal, 2019; Müller, Pomorska, & Tonra, 2021; Webber, 2019).

However, does the emergence of differentiated cooperation in the CFSP follow the same drivers as differentiation in EU internal governance? Despite the growing recognition that EU foreign policy is becoming increasingly differentiated, it has not yet been determined whether the CFSP’s differentiated cooperation is brought about by drivers identical to those responsible for differentiation in internal governance. The present study purports to fill that gap by investigating whether a significant change affecting the regime in which the EU operates can foster a pattern of differentiated co-operation. For the purpose of this research, the notions of “regime-wide change” and differentiated co-operation require definition first. We understand that a regime-wide change occurs when an international regime experiences a policy development that aims at modifying fundamental rules which govern the regime. Despite its absence of formalization, we consider that the phenomenon of differentiated cooperation prevails when at least two subgroups of at least two EU members each align their positions among themselves but not with members of other subgroups, and this alignment remains constant over time.

In order to explore whether regime-wide change can drive differentiated cooperation to the CFSP, we examine EU alignment in a field where a regime-wide alteration is observable: the nuclear nonproliferation regime. This realm of the CFSP is selected because this domain experienced significant change over the past decade. The traditional intra-European divide over nuclear disarmament was exacerbated by the emergence of HI, which developed into a binding treaty—the TPNW—introducing a veritable turning point in the nuclear nonproliferation regime (Gibbons, 2018; Portela, 2021). In addition, in the NPT RevCon framework, several EU member states coordinate with other groupings outside the EU, simultaneously to their participation in the CFSP. Overlapping memberships between the EU and other informal groupings in the NPT allows us to examine the impact that a regime-wide alteration has on CFSP alignment versus non-CFSP alignment.

The EU in the nuclear nonproliferation regime

The EU coordinates its role at the NPT Review Conferences as well as at the annual sessions of UNGA, whose First Committee deals with “Disarmament and International Security.” Coordination in the NPT framework started in the run-up to the 1995 review, once France joined the NPT (Onderco, 2021; Pouponneau & Mérand, 2017). Ahead of the RevCon, the Council Working Party on Nuclear Proliferation (CONOP) routinely agrees statements and priorities, invariably pledging to “help build a consensus” and to “strengthen the international nuclear nonproliferation regime by promoting the successful outcome of the conference” (Council of the European Union, 2005). At NPT RevCons, the EU submits common working papers as a group. In addition to acting as the EU block, member states cooperate with non-EU members via their individual participation in various groupings (see below). By contrast, coordination at UNGA is geared towards the framing of a common stance on the resolutions that are voted upon at the end of each yearly session.

In the nonproliferation regime, the effort to frame a common line in the EU is hampered by the diversity of member state attitudes towards nuclear deterrence and disarmament. After the EU started acting jointly in nuclear nonproliferation fora, NPT membership gave some uniformity to a “patchwork” of nuclear attitudes towards nuclear deterrence. France remains, after the British withdrawal, the only nuclear weapons state inside the EU. Twenty-one out of the current 27 EU member states are NATO allies. Belgium, Germany, Italy, and the Netherlands host nuclear weapons on their territory while the remaining 17 are covered by NATO’s nuclear “umbrella,” or extended deterrence. Of the six EU members that remain outside NATO, nuclear-free Austria, Finland, Ireland, and Sweden are traditional champions of nuclear disarmament.² Cyprus and Malta, until 2004 part of the pro-disarmament Non-aligned Movement (NAM), share this goal but are less active in advocacy. Notwithstanding these discrepancies, CFSP coordination managed to foster convergence among member states positions. The expectation that coordination would increase convergence over time (Müller, 2020) was confirmed by the early success of the EU’s campaign in support of the 1995 NPT extension. The EU’s internal diversity was described as a “laboratory of consensus” that could help foster agreement in the NPT framework (Grand, 2000).

Meanwhile, diverging approaches towards nuclear weapons surfaced more clearly outside the outputs of the CFSP. Alongside the EU, multiple groupings operate in the NPT context: traditional clusters like the 120-strong NAM, or regional groups like the Arab League. Traditional blocs coexist with yet other like-minded formations, mostly of a cross-regional or cross-factional character. In parallel to CFSP coordination, several EU members coordinate with other groupings. Previous review cycles saw the action of the Netherlands and Germany under “NATO-5,” a now extinct coalition of umbrella countries. Currently, several—albeit not all—EU member states are active members of alternate groupings. The “Nuclear Weapon States” (NWS) assembled in the P-5 format includes France alongside former EU member the UK. The Netherlands, Germany and Poland formed the “Nonproliferation and Disarmament Initiative” (NPDI). The pro-disarmament “New Agenda Coalition” (NAC) was launched with Ireland, Slovenia, and Sweden, out of which only Ireland remains. Seven EU member states cosponsored Austria’s statement highlighting the humanitarian consequences of the nuclear weapon use at the 2015 NPT RevCon, along with 152 other countries (Kurz, 2015).³ Figure 1 shows the overlapping membership of EU members in various groupings.

Figure 1. EU member states membership in select NPT groupings (as of 2015). Note: Countries co-sponsoring the 2015 Joint Statement on the Humanitarian Consequences of Nuclear Weapons are marked in italics. Note that membership of NAC varied over the course of the period under study (see main text).

Gradually, the EU’s position at the NPT shifted from a “cooperation of European states” to “European cooperation” (Dee, 2012). Successive CFSP acts adopted in preparation for the sessions illustrate an incremental evolution: they grew longer practically at every RevCon (Portela, 2021). Notwithstanding this quantitative increment, a substantive analysis of EU coordination identifies the opposite trend: consensus peaked in 1995 and 2000. While 2010 saw an unprecedented volume of EU-sponsored working papers, the EU Common Position remained the “lowest common denominator” (Müller, 2020). At UNGA, certain resolutions on nuclear weapons attracted universal EU support consistently (Panke, 2014). Nevertheless, an overall trend towards greater convergence in the voting patterns of member states is marred by persistent divisions over nuclear weapons issues (Burmester & Jankowski, 2018; Luif, 2014).

A major setback in EU alignment occurred with the emergence of the HI, a series of state-convened conferences on the humanitarian impact of nuclear weapons (Kmentt, 2015), which culminated in the adoption of the TPNW in 2017. Austria was indeed one of the sponsors of the HI, in which it exercised a leadership role. By the entry into force of the new treaty in 2021, three EU members were full parties: Austria, Malta, and Ireland. This divide cut across EU membership, as nuclear-armed France and (at the time) the UK found themselves on opposite sides of the spectrum. The umbrella countries, comprising the vast majority of EU members, opposed the TPNW as incompatible with NATO’s extended deterrence. Finally, non-NATO members like Finland and Sweden, still in the run up to their accession application at the time, feared that TPNW accession would limit their ability to cooperate with NATO (Onderco & Farrés Jiménez, 2021). NATO’s official position decries the TPNW as an attempt to stigmatize nuclear weapons possession rather than contribute to nuclear disarmament (Gibbons, 2018; Ritchie, 2019). While the emergence of the TPNW can be seen as contributing to the nuclear nonproliferation and disarmament regime, it also increases the complexity of the regime and the competitiveness within it by establishing an alternative forum (Baldus et al., 2021).

This EU-internal polarization was visible as it hindered coordination at the 2015 NPT RevCon. Council Conclusions agreed in anticipation of the conference evidenced disagreement, oddly noting “ongoing discussions on the consequences of nuclear weapons, in the course of which different views are being expressed, including at an international conference organized by Austria, in which not all EU member states participated” (Council of the European Union, 2015, p. 4). In view of the split, the idea that the EU can be a laboratory of consensus lost validity.

Conceptualization and hypotheses

Our first hypothesis is that the HI and the conclusion of the TPNW altered EU alignment, leading to differentiated cooperation. We consider that a pattern of differentiated cooperation prevails if at least two collective positions crystallize among member states resulting in the establishment of subgroups of stable nature (see above). In the absence of formalized alignment within EU members, such phenomenon can only be detected empirically.

H1: EU convergence was fundamentally altered by the Humanitarian Initiative and the TPNW, after which EU member states positioning on nuclear weapons crystallized into differentiated subgroups.

If the hypothesis is confirmed and we find a pattern of differentiated cooperation, the notion that differentiated cooperation can result from regime-wide alterations will be validated.

Yet, the fact that a “newcomer” to the regime such as the TPNW disrupted alignment among EU member states does not explain why the situation after the regime-wide alteration became one of differentiated cooperation. The regime-wide alteration could have resulted in alternative outcomes such as segmentation—i.e., the formation of multiple sub-groups of small membership—or even atomization, featuring sub-groups of very small membership alongside stand-alone states. The fact that the disruption created by the TPNW on alignment among EU member states should take the form of differentiated cooperation is not a given. Yet, this outcome was promoted by the concurrent membership of some EU members in groupings with diverging orientation. This led those members to align closer to their partners in external groupings. In other words, EU alignment weakened while alignment with alternative groupings strengthened. The disruption of the EU’s common line and the resort to alternate groupings are two sides of the same phenomenon.

The switch from alignment within the EU to another grouping dovetails with the idea that states use focal institutions whenever satisfactory, in line with the literature on institutional choice (Jupille et al., 2013). States choose institutional platforms that are the most suitable to the issue at hand, given that different institutions have different resource endowments (Brosig, 2017). Because of their concurrent membership in various (formal or informal) institutions, member states are able to select which of them fits their goals. The pursuit of goals through alternative institutions does not necessarily entail spurring European institutions; rather, it means that member states activate a different institutional platform to pursue cooperation.

The global nonproliferation regime offers many such opportunities. Member states use various informal institutions to pursue goals related to nuclear nonproliferation, such as Nuclear Suppliers Group or Wassenaar Arrangement. These groupings entail a broader membership than the EU. However, the member states can also find (informal or formal) institutional platforms to pursue goals which others do not share, such as the push for the humanitarian norms-based approach to nuclear disarmament, where Austria and Ireland teamed up with numerous non-European countries (Gibbons, 2018). Thus, the concept of differentiated cooperation captures adequately those situations where EU member states make individual policy choices which are aligned with those of other but not all fellow member states in order to cooperate (in a differentiated fashion) in a context of informality, i.e., in the absence of a coordinating mechanism. Our hypothesis is consistent with the expectation of differentiated cooperation which conceives of membership in formal and informal international institutions as a root of differentiation (Amadio Viceré & Sus, 2023; Rieker, 2021).

H2: The differentiated subgroups that emerged in the EU following the Humanitarian Initiative and the TPNW coincide with overlapping membership in alternate groupings operating in the nonproliferation regime.

Methodology

To evaluate our hypotheses, we employ data from two global fora: the First Committee of UNGA, which constitutes one of the security fora where regional organizations are most vocal (Panke, 2017a), and the NPT quinquennial RevCons. We chose this data for their value as indicators of alignment that cover the period under study: from 2000 to 2020 for UNGA and from 2000 to 2015 for NPT. The selected timeframe covers the period leading up the HI/TPNW turning point and its immediate follow up, that is 2013–2017. The period under study starts in 2000, which marked the beginning of a renewed interest in the disarmament agenda in the nonproliferation regime, which first took the form of the more conservative roadmap enshrined in the “13 Practical Steps Towards Nuclear Disarmament” adopted at the 2000 NPT review conference (UN, 2000, art. V) before it evolved into the outright legally-binding ban of the TPNW.

Both sources express two measurable manifestations of convergence, voting and co-sponsorship: UNGA votes annually on over a dozen (and sometimes as many as two dozen) resolutions related to nuclear weapons. We use records from voting on these resolutions (Voeten et al., 2018, updated until 2020) to estimate voting convergence between individual EU member states over time. For NPT RevCons, we look at the co-sponsorship of working papers and proposals. We use different indicators for each forum because each of them operates differently. Since no voting takes place at the NPT Review Process, we focus on co-sponsorships. By contrast, nuclear-related resolutions within the UNGA are overwhelmingly sponsored by one single country, which prevents the use of co-sponsorship as an indicator. Instead, UNGA resolutions are voted on, providing a classical indicator.

Measuring voting convergence

We look at two measures of voting. First, we compare national ideal points over time. The measurement of ideal points originates from the analysis of legislative behavior and permits to locate legislators’ stances on an axis, based on a large number of votes. The relative position of individual legislators on the axis is determined by their likelihood to vote similarly. The more likely two actors are to vote similarly, the closer they are to one another. By definition, ideal points are scale-free, and estimate a position of a country in a policy space on one particular dimension. In this test, we apply the measurement of ideal points to voting by states instead of legislators, following Bailey et al. (2017). We use ideal points because of their superior ability to discriminate between divisive and consensual resolutions, outperforming other measures of similarity of state preferences in UNGA Bailey et al. (2017). We leverage the fact that the theme of each resolution receives a code (Voeten et al., 2021). Voeten (2020) provides issue-specific ideal points for the period up until 2019. We therefore use these issue-specific (nuclear) ideal points as the first source.

Second, we take the voting data from Voeten et al. (2021), select nuclear weapon-related resolutions, and then code recurrent resolutions, that is resolutions which are tabled repeatedly. Voting on recurrent resolutions illustrates the development of member states preferences over time. For each repeated resolution, we calculate an agreement index used by Hix et al. (2005) to measure unity. Agreement values vary from zero to one, where 1 stands for complete unity, and 0 stands for perfect division in equal numbers between Yeses, Nos, and Abstentions. The higher the agreement value, the higher the convergence of EU member states in their voting on a particular resolution. We calculate these scores for the period up until the 75th UNGA session that started in September 2020.

Measuring co-sponsorship

To study co-sponsorship, we consider working papers and proposals that member states present within the NPT RevCons. We coded co-sponsors of all resolutions tabled at the 2000, 2005, 2010, and 2015 NPT RevCons—the 2020 edition was postponed until August 2022 due to the Covid pandemic. We then calculate a dyadic co-sponsorship score (Alemán et al., 2009) which captures the number of co-sponsored proposals/resolutions with a particular member state B as a share of that member state A’s total proposals. Therefore, co-sponsorship is a directed-dyad-year specific, following the standard approach in the study of co-sponsorship in legislative politics (Louwerse & Otjes, 2015). We also measure the EU co-sponsorship score, which captures the number of EU-sponsored proposals as a share of member state A’s total proposals, which is country-year specific. Both co-sponsorship and EU co-sponsorship are measured on a scale from 0 to 1, where 0 stands for the absence of joint sponsorship, and 1 means all resolutions of member state A were co-sponsored with a member state B, or were submitted by the EU.

Analysis: Subgroups reconfigured

UNGA voting

We first examine the ideal points of the EU member states on nuclear weapons resolutions and the voting on repeated resolutions. We consider all members of the EU by 2020, including those countries that acceded after 2004. Figure 2 provides an overview of the ideal points of EU members since year 2000. The two nuclear powers in the EU are plotted on the top, together with the US as a non-EU NATO member. The six EU members outside NATO—i.e., Austria, Cyprus, Finland, Ireland, Malta, and Sweden—are visible in the dotted line below. Reviewing the figure, we observe that, twenty years ago, EU member states were divided into three major groups: European NWS, the thick mainstream composed of umbrella countries, and disarmament advocates. Over time, the European NWS move closer to the rest of the EU. On the other hand, the non-NATO members of the EU move away from the EU mainstream.⁴ Interestingly, Finland is located halfway between the five EU members outside NATO listed above and the NATO countries, revealing that the country’s nuclear disarmament policy was distinct from the group of the five most proactive disarmament advocates.

Figure 2. Ideal points of EU members since 2000. Legend: solid line represents members of EU and NATO, dotted line represents members of EU but not NATO, red line represents members of NATO but not EU.

To assess the “spread” of the data, and to ascertain whether that spread changes over time, we look at the kurtosis of the data. Kurtosis is a measure of “flatness” of the data—measuring how long the tails of the distribution are. Kurtosis around zero means a normal distribution, negative kurtosis indicates positive peakedness of the distribution (closer around the mean with shorter tails); whereas positive kurtosis indicates increased flatness of distribution (longer tails). Assessed on an annual basis, the kurtosis of the data varies over time, as displayed in Figure 3. However, it shows that the peak polarization (measured by the highest kurtosis) actually took place around 2010, indicating that the differences between states were highest around that time. Kurtosis has been in decline since, and to a greater extent for EU member states than for NATO countries, indicating that the EU countries are even more closely aligned than NATO counties. Figure 3 hence demonstrates growing similarity among most member states in recent years.

Figure 3. Kurtosis of ideal point data over time. Source: own calculation based on Voeten (2020).

Taken together, voting patterns show that, while overall polarization subsides, two stable groups of states with different preferences emerge. Looking at the voting on recurrent resolutions can give us some idea about the origins of such differentiation. We consider all eight resolutions on which there were at least two votes between 2000 and 2020.

Figure 4 shows the agreement index for each of these resolutions. Agreement drops precipitously in some resolutions, like the “Accelerating Disarmament Commitments” resolution, which enjoyed broad EU consensus until 2014. In 2015, the resolution added language delegitimizing nuclear weapons (United Nations, 2015) and in 2016, it welcomed the work of the Open-Ended Working Group which led to the negotiation of the TPNW (United Nations, 2016), a process ignored by the vast majority of EU member states (Nielsen & Hanson, 2014). Since 2015, only Austria, Cyprus, Ireland, Malta, and Sweden vote in favor of this resolution. Another case in point is the resolution on a “Nuclear-weapon-free Southern Hemisphere.” This resolution used to enjoy the backing of all EU members except France and the UK, but most EU members switched to abstention after the insertion of language welcoming the adoption of the TPNW in 2017 (United Nations, 2017).

Figure 4. Agreement Index for repeated nuclear resolutions. Source: own calculation based on Voeten (2020).

Close inspection of the topics of resolutions on which EU countries were divided demonstrate that “defections” in recent years are related to the rise of the HI, which culminated in the adoption of the TPNW in 2017. Thus, the divergence among the EU member states from 2015 onwards reflects the shift in the UNGA nuclear arms control agenda brought about by the HI and TPNW. This shift occurred exactly over nuclear disarmament, the issue that most divides EU members (Smetana, 2016), and was spearheaded by some EU member states, particularly Austria, a leader of the HI, whose ideal point moved the furthest in the period under study (Table 1).

Table 1. Nuclear disarmament resolutions at UNGA.

Resolution title	Resolution code
Follow-up on ICJ Legality and Threat of Nuclear Weapons	A/RES/75/66 (2020)
Accelerating Disarmament Commitments	A/RES/75/65 (2020)
2013 High-level Meeting Follow Up (Nuclear Disarmament)	A/RES/75/45 (2020)
Risk of Nuclear Proliferation in the Middle East	A/RES/75/84 (2020)
Nuclear-weapon-free Southern Hemisphere	A/RES/74/48 (2019)
Taking Forward Multilateral Nuclear Commitments	A/RES/72/31 (2017)
Treaty on the Prohibition of Nuclear Weapons	A/RES/75/40 (2020)
Humanitarian Pledge	A/RES/71/47 (2016)
Humanitarian Consequences of Nuclear Weapons Use	A/RES/75/39 (2020)

Note: most recent year included in the dataset in parentheses.

NPT RevCons

We now move to the analysis of co-sponsorship at the NPT RevCons. Table 2 displays co-sponsorship of working papers and proposals at the NPT RevCons, while Figure 5 shows the values of the two variables for all EU member states. In both tables, we add values for co-sponsorship with non-EU NATO countries for the sake of comparison.

Figure 5. Co-sponsorship of proposals at NPT RevCons among EU members.

Table 2. Co-sponsorship of proposals & working papers among EU members at NPT Revcons.

Year	Variable	Mean	Min	P10	P25	Median	P90	Max
2000	EU co-sponsorship	0.51	0.24	0.25	0.31	0.56	0.71	0.83
2000	Co-sponsorship with other EU countries	0.41	0.04	0.24	0.30	0.43	0.59	0.59
2000	Co-sponsorship with non-EU NATO countries	0.21	0.00	0.08	0.10	0.17	0.42	0.42
2005	EU co-sponsorship	0.81	0.45	0.53	0.77	0.91	1.00	1.00
2005	Co-sponsorship with other EU countries	0.83	0.52	0.59	0.76	0.91	0.94	1.00
2005	Co-sponsorship with non-EU NATO countries	0.23	0.15	0.15	0.19	0.20	0.50	0.54
2010	EU co-sponsorship	0.58	0.25	0.25	0.29	0.62	0.80	1.00
2010	Co-sponsorship with other EU countries	0.75	0.35	0.44	0.50	0.79	0.96	1.00
2010	Co-sponsorship with non-EU NATO countries	0.12	0.00	0.03	0.11	0.12	0.21	0.22
2015	EU co-sponsorship	0.53	0.30	0.33	0.43	0.55	0.75	0.75
2015	Co-sponsorship with other EU countries	0.82	0.48	0.56	0.70	0.88	0.99	0.99
2015	Co-sponsorship with non-EU NATO countries	0.17	0.10	0.11	0.15	0.19	0.22	0.25

P_x denotes the x^th percentile rank.

With very few exceptions, the majority of proposals handed in by most EU members within the setting of NPT RevCons are submitted with fellow EU members. These are not necessarily submitted on behalf of the EU, but they do include other EU members, whether through ad hoc coalitions or other groupings.⁵ The NPT RevCon in 2005 was exceptional because the EU submitted a peak of working papers as a group: A total of 10, about twice as many as during other RevCons. Given that the median number of co-sponsored working papers is between five and seven per conference per country, doubling the number of EU-sponsored papers meant that in 2005, the median number of co-sponsored working papers was 11 per member state. For over half of EU member states, EU-sponsored working papers are over half of what they propose at the conference; and for numerous member states, EU-sponsored working papers are the only ones they co-sponsor. Lastly, all EU member states without exception submit more papers with fellow EU members than with non-EU NATO allies. Cooperation between member states is much more intense than cooperation with other NATO allies, including the United States. By and large, EU members cooperate very actively with each other, while they cooperate much less with their NATO allies. At the same time, we also detect indications that point to differentiation: For all member states, co-sponsorship with other EU member states individually remained higher than formal EU co-sponsorship. Indeed, according to Figure 4, the differential between sponsorship with all-EU versus only some EU members grows at the 2015 RevCon compared with the 2010 edition in most countries under study. In those few countries where it does not grow, it remains constant, but there is no single instance where it decreases. This suggests the presence of intense cooperation, but in an increasingly differentiated format.

Discussion: A cleavage redux

While the HI and the conclusion of the TPNW disrupted EU cohesion (Smetana, 2016), neither the depth of such facture nor its evolution over time had been measured yet. We find that a degree of convergence in the stance of EU member states towards nuclear weapons prevailed prior to 2015. Our analysis demonstrates that, prior to the HI and NPTW, the picture was one of stable cooperation among EU members, both as a block and in the shape of bilateral links. Notwithstanding the persistence of divisions among member states over the traditionally controversial question of nuclear disarmament (Burmester & Jankowski, 2018; Dee, 2012; Smetana, 2016), remarkable convergence prevailed. Most EU member states were stable in their cooperation with other EU members, displaying a stagnant share of EU-sponsored resolutions.

Our hypothesis set out to confirm that the HI and the TPNW fundamentally altered EU convergence, after which EU member states positioning on nuclear weapons crystalized into differentiated subgroups. To corroborate a scenario of differentiation, we expected stable ideal points with member states divided into subgroups of stable membership in terms of votes on recurrent resolutions, co-sponsorship with other EU member states in the same subgroup and statements diverging from other subgroups of EU member states. Indeed, we find that the HI and TPNW transformed the picture, bringing about a scenario of differentiated cooperation. We observe convergence taking place between two out of three sets of states: NWS and umbrella countries, resulting in the formation of a voluminous mainstream subgroup. By contrast, the contrary is true for the third set of disarmament advocates, whose increasingly pronounced segregation from the rest of the EU indicates that two separate subgroups crystalized. This phenomenon is supported by evidence from voting on recurrent resolutions.

The consolidation of two separate subgroups holding opposing views on disarmament and unprepared to follow the lead of the other subgroup agrees with the notion of differentiation, which lacks the teleological dynamism of differentiated integration. Instead, it expects the persistence of neatly delineated subgroups of stable membership intent on maintaining their position. Remarkably, every EU member is part of a subgroup within the EU: the mainstream subgroup or the pro-disarmament subgroup—none of them stands alone or fluctuates heavily. In the post-2015 era, the delimitation of subgroup membership, in the absence of defectors, undecided members or Einzelgänger excellently fits the notion of differentiated cooperation. Ironically, the crystallization of two fairly cohesive subgroups among EU member states evidences the sort of convergence that CFSP mechanisms aspires to: the differential from the optimum is that, rather than one homogeneous group, two subgroups came about.

In sum, the analysis confirms the HI and the ensuing TPNW as the turning point accounting for the emergence of differentiated cooperation, in line with our first hypothesis. The departure of disarmament advocates from the mainstream exacerbated after the 2015 fracture. Shifts in voting patterns perceptibly respond to significant change in the nuclear nonproliferation regime, visible in the introduction of new language to existing resolutions. At the same time, the EU member states continue to cooperate with one another extensively, proven by continued high levels of co-sponsorship of papers among members of the same subgroup. The fact that abundant co-sponsoring with other EU members persists, although not necessarily on behalf of the EU, is another piece of evidence confirming the existence of differentiated cooperation in the EU foreign policy on nuclear weapons.

Our second hypothesis expected that the differentiated subgroups that emerged in the EU following the HI and the TPNW correspond to overlapping membership in alternative groupings. This assumption is confirmed as well. There is an evident coincidence between the NPDI and NWS members and the composition of the mainstream subgroup, where they are accompanied by the remaining umbrella countries, all of which lack concurrent memberships elsewhere. Similarly, the second subgroup includes the disarmament advocates, notably NAC member Ireland and HI leader Austria.

Yet, while all NSW and umbrella countries are in the mainstream subgroup, the correspondence between NPT groupings and the pro-disarmament subgroup is imperfect, which introduces some nuance. Firstly, Austria is not a member of NAC—despite Vienna’s manifest alignment with its positions—nor of any other grouping in the NPT framework. Still, Austria does not fit the profile of Einzelgänger on account of its leadership of the HI. Indeed, although it has not yet been treated as such, the HI could be considered an NPT grouping on its own right, which would situate Austria in the lead of a large followship. Sweden, an erstwhile NAC member, is closely aligned with the pro-disarmament cluster despite its abandonment of the grouping. Secondly, traditional disarmament advocate Finland is the only country that approximates an intermediary position in Figure 2, albeit it admittedly remains closer to the mainstream than to the disarmament subgroup.

Although the absence of Finland from the pro-disarmament subgroup does not contradict our hypothesis since it is not a NAC member, it suggests that NATO is the “elephant in the room” accounting for the membership of EU countries in each of the subgroups identified. Indeed, the heightened risk perception caused by the 2014 annexation of Crimea among the Nordic countries promoted a closer alignment with NATO (Michel & Pesu, 2019), eventually culminating in the Finnish and Swedish accession applications after the Russian invasion of Ukraine in 2022 (NATO, 2022). Albeit NATO does not constitute per se a grouping in the NPT framework, membership in the Atlantic Alliance remains a powerful determinant of parties’ positioning in the regime. No current or aspiring ally can be found in the small pro-disarmament subgroup in the period under investigation. Conversely, the accession of ten new member from Central and Eastern Europe to the EU in 2004 and 2007 did not alter the patterns identified, given that all new arrivals were NATO members that integrated swiftly into the mainstream subgroup. The incompatibility between a pro-disarmament stance and Alliance membership is evident in the withdrawal of Slovenia from the NAC while it was a candidate for NATO accession, and it surfaced, once again, with Finland’s and Sweden’s recent candidacies.

The notion of differentiation, alongside various concepts that capture an attempt at unraveling integration, locates the drivers of differentiation in the polycrisis the EU experienced over the past decade. Looking at the CFSP with the help of the novel concept of differentiated cooperation, we posit that change in international regimes in which the EU is embedded constitutes a driver too. The fact that coordination coexists both among EU member states and between these and external actors supports differentiated cooperation in the CFSP. The evolution of sponsorship and voting in key arenas of the nuclear nonproliferation regime show that differentiated cooperation is associated with informal coordination with groupings partly located outside the EU. A major innovation in the nuclear nonproliferation regime significantly transformed the picture of EU alignment. Our findings contribute to European foreign policy scholarship by identifying change in multilateral regimes as a driver of differentiated cooperation in the CFSP. Foreign policy differentiation must not result from homegrown crises but may arise from the broader dynamics at play in the nuclear nonproliferation regime. Thus, the exclusive emphasis on internal crises overlooks the significance of regime-wide factors.

The question is where all this leaves the CFSP. Our findings put the significance of the CFSP into perspective. On the one hand, they evidence the fragility of CFSP coordination despite the existence of an apparently robust record, as a powerful regime-wide alteration can still disrupt it. The implication for the future course of the CFSP is that this policy framework is confirmed as vulnerable to significant change in the broader regimes in which it operates. Yet, the prospect for the CFSP is not as bleak as this finding might suggest. On the other hand, the evolution detected testifies to the sustained effort made by EU capitals to find sufficient common ground to frame an EU position despite notable discrepancies, evidenced not only in the generation of CFSP outputs, but even in the intensity of EU member-to-member collaboration, unparalleled in their contacts with groupings outside the EU. The Council Conclusions on the 10th NPT RevCon demonstrate that the EU member states can find plenty of common ground, and can “agree to disagree” where that common ground is lacking (Council of the European Union, 2021).⁶ In other words, the CFSP as a coordination reflex continues unabated even after the HI/TPNW “shock.”

Conclusion

Our analysis shows that differentiation can be found in CFSP coordination in the non-proliferation field, and that this pattern of differentiation does not emanate from an internal crisis instigated by Eurosceptic actors intent on undermining the integration process from within, but from a significant change in the international regime. Equally, the analysis demonstrates the usefulness of the notion of differentiated cooperation, in whose absence we would face difficulties in grasping the informal character of alignment in the CFSP. Our findings make a novel contribution to research on differentiation in foreign policy. Having explored EU differentiation in the framework of the nonproliferation regime, we have established that drivers for differentiation in the CFSP need not coincide with drivers for differentiation identified in EU internal governance; rather, they can emanate from the regime rather than from within the EU. Yet, due to its focus on one single regime, it is outside the purview of our analysis to establish the extent to which the same can be said of EU alignment in other international regimes. We leave to further research to establish the mix between internal and regime-wide drivers of differentiation in other arenas and the identification of determinants that account for variation. Moving forward, a potential avenue for future work consists in examining whether significant changes in other global regimes affect differentiated cooperation between the EU member states. Having detected the emergence of differentiation in nuclear nonproliferation, a global regime subject to increasing regime complexity and forum shopping (Baldus et al., 2021), it would be reasonable to expect similar dynamics in any global regime characterized by similar features (Alter & Meunier, 2009).

In addition, our study generates refreshing insights about CFSP alignment. Notwithstanding scholarly insistence on the state-centered nature of security fora (Panke, 2017a) and on the divisive character of nuclear weapons (Luif, 2014), the reflex of EU coordination in the field proved robust relative to alignment with other actors like non-EU NATO allies. The fact that EU members maintained a common line up until the HI/TPNW and NATO’s resolute opposition to it split them into advocates and detractors underlines member states commitment to the CFSP. Moreover, the crystallization of differentiated cooperation did not spell out the end of the engagement with the CFSP, which has not discontinued the practice of coordination and formulation of common statements even after a fracture of the magnitude of that caused by the TPNW. This confirms, once again, the EU as an exceptionally active regional actor in international fora (Panke, 2017a; Panke et al., 2019). The launch on the “Stockholm Initiative for Nuclear Disarmament” in February 2020 (Ministry of Foreign Affairs of Sweden, 2020), which brings together the entire EU with some extra-European partners, testifies to EU members’ readiness to rebuild some European consensus in the nonproliferation regime.

Lastly, the transformation of European security following the Russian launch of the war in Ukraine in February 2022 draws attention to the future evolution of differentiation in the nuclear weapons domain. The members of the small pro-disarmament subgroup, in particular Finland and Sweden, will be worth watching. While neither Helsinki nor Stockholm contemplated NATO membershipduring the period under study, their impending accession to NATO opens the question as to whether their alignment behavior in the CFSP will be affected or otherwise.

Acknowledgements

An earlier version of this paper was presented at the authors’ workshop at the EUI in Florence in October 2021. We thank the organizers of the workshop and the editors of this special issue, Maria Giulia Amadio Viceré and Monika Sus, as well as Benjamin Martill and Stephanie Hofmann, for helpful comments. We are also grateful to the CSP editors and two outstanding reviewers for their helpful suggestions which greatly benefited this article. We thank Marin Lucic and Femke Verburg for diligent research assistance.

Disclosure statement

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

Additional information

Funding

This project was supported from Charles University Research Centre program under Grant UNCE/HUM/028 (Faculty of Social Sciences / Peace Research Center Prague).

Notes on contributors

Michal Onderco

Michal Onderco is Professor of International Relations at Erasmus University Rotterdam and affiliate at Peace Research Center Prague. He studies international security, with focus on nuclear politics and on domestic politics of foreign policy. He authored Networked Nonproliferation (Stanford UP, 2021) and Iran’s Nuclear Program and the Global South (Palgrave, 2016), as well as papers which appeared in International Studies Quarterly, European Union Politics, Journal of Common Market Studies, Contemporary Security Policy, European Journal of Political Research, Cooperation & Conflict, The Nonproliferation Review (and elsewhere). In 2018–2019, he was a Junior Faculty Fellow at Stanford University’s Center for International Security and Cooperation.

Clara Portela

Clara Portela is Konrad Adenauer Visiting Scholar at Carleton University’s Centre for European Studies in Ottawa, Canada, and a professor at University of Valencia in Spain. She previously served as a professor at Singapore Management University and as a research fellow with the European Union Institute for Security Studies (EUISS) in Paris, where she dealt with arms control and disarmament issues. She holds a PhD from the European University Institute in Florence and an MA from the Free University of Berlin. She co-edited the volume “The EU and the Non-Proliferation of Nuclear Weapons" (2015), among various publications on EU foreign policy. She is the recipient of the THESEUS Award for Promising Research on European Integration.

Notes

1 The UK was an EU member from 1973 to 2020.

2 Finland and Sweden did not apply for NATO membership until 2022. Since our data cover the period until 2020 (for UNGA) and 2015 (for NPT) only, these countries are not considered as NATO aspirants for the purposes of this study.

3 Given that different metrics can be used to indicate which countries are supporters of the Humanitarian Initiative, we chose the co-sponsorship of the statement delivered by Austria at the 2015 RevCon (Kurz, 2015) as an accurate indicator of support, given that the project had achieved maturity at that juncture.

4 The data shows one-off variation in Greece’s voting in 2010, which was mainly due to some erroneous votes, and which is systematically irrelevant.

5 Some EU members joined various nonproliferation informal arrangements – such as the Nuclear Suppliers Group or the Zangger Committee – only at the time of EU accession. The EU regularly calls on others to join such regimes (EEAS, 2018).

6 Council Conclusions omit any mention to the TPNW, referring instead to the “humanitarian impact” of nuclear weapons and to their “very severe consequences.”