Expanding, Complementing, or Substituting Multilateralism? EU Preferential Trade Agreements in the Migration Regime Complex

Intense pressure for international solutions and weak support for multilateral cooperation have led the EU to increas‐ ingly rely on its strongest foreign policy tool in the pursuit of migration policy goals: preferential trade agreements (PTAs). Starting from the fragmentary architecture of the migration regime complex we examine how the relevant content of the EU PTAs relates to multilateral institutions. Depending on the constellation of policy objectives, EU competence, and inter‐ national interdependence, we propose a set of hypotheses regarding the conditions under which EU bilateral outreach via PTAs expands, complements, or substitutes international norms. Based on an original dataset of migration provisions in all EU PTAs signed between 1960 and 2020, we find that the migration policy content in EU PTAs expands or complements the objectives of multilateral institutions only to a very limited extent. Instead, the predominant constellation is one of substitution in which the EU uses its PTAs to promote migration policy objectives that depart from those of existing multi‐ lateral institutions.


Introduction
Since the crisis of the Common European Asylum System (CEAS) in 2015, international migration governance has moved from a side aspect to a key priority for EU external action and a serious challenge to European integration overall. Yet, in dealing with migration flows, the EU depends on international cooperation. This is not an easy task: Contrary to other transnational flows, such as goods or capital, states have hitherto opposed the creation of strong international institutions dealing with migration, and what exists amounts at best to a fragmented and multi-layered migration regime complex (Betts, 2011;Lahav & Lavenex, 2012). The EU, in particular the European Commission, has been keen on developing a presence in relevant international organizations such as the UNHCR and the IOM (Beqiraj et al., 2019). However, internal divides within the EU and limited competence to act have compromised EU engagement in these fora. This became evident during the negotiations for the UN Global Compact for Migration adopted in 2018. The EU was a driving force behind this initiative; however, when it concluded, several member states ended up not signing it (Melin, 2021). Against this background of high political salience and failed multilateral engagement, this article investigates how far the EU has sought preferential trade agreements (PTAs) as alternative venues for advancing migration policy objectives and how these objectives relate to the migration regime complex. Our comparative analysis of all EU PTAs signed between 1960 and 2020 addresses the following research question: How far do migration provisions in EU PTAs expand, complement, or substitute multilateral institutions in the migration regime complex, and under what conditions?
The relevance of this focus emerges from the multifaceted and fragmented nature of the multi-layered migration regime complex and the fragile balance between different, partly contradictory, policy objectives and levels of governance. While multilateral institutions promote a rights-based approach and open select opportunities for economic mobility, the question of migration control is left to the sphere of state sovereignty or, as in the case of the EU, regional initiatives (Dauvergne, 2014;Geddes et al., 2020). The migration policy content in EU PTAs can therefore either sustain the rights-based focus of multilateral provisions by expanding or complementing them, or the EU can use its PTAs to promote self-serving migration control objectives, in which case PTAs constitute a substitute through which the EU can "venue shop" for its preferred objectives. Since migration control can come at the expense of migrant rights, such substitution may create tensions with multilateral institutions both in terms of substance-the balance between different substantive policy objectives-and in terms of levels-the balance between multilateral and regional institutions of the multi-layered regime complex.
After introducing the multidimensional migration regime complex with a focus on migrant mobility, migrant rights, and migration control we highlight the respective limits of EU leadership and zoom in on the role of PTAs. Drawing on the regime complexity literature, we conceptualise three types of the multi-layer institutional interplay between the migration policy content of PTAs and multilateral norms: expansion, complement, or substitution. Depending on the constellation of policy objectives, EU competence, and international interdependence, we then introduce an analytical framework explaining the conditions under which we may observe EU venue-shopping leading to either regime expansion, complement, or substitution. We examine our hypotheses by mobilising two original datasets on EU migration provisions in PTAs and ratification of international agreements on migration. We conclude with some reflections on the implications for the EU's international role as a migration policy actor.

The EU in the Multi-Layered Migration Regime Complex
International migration governance takes the architecture of a regime complex in which relevant provisions figure in different institutions that are partly overlapping and partly nested, but where none is focal (Alter, 2022;Betts, 2009;Eilstrup-Sangiovanni & Westerwinter, 2021;Jupille & Snidal, 2005). Unlike other areas of international relations such as trade or the environment, where states tend to agree on key objectives such as opening markets or saving the planet but disagree on the means, cooperation on international migration is fragmented across several partly contradictory objectives. Different institutions approach migration from different perspectives such as protecting the rights of displaced individuals fleeing violence or suffering exploitation, facilitating the allocation of labour in international markets, enforcing territorial borders, or enhancing distributive justice. These distinct objectives have been codified to different extents in different layers of international cooperation: multilateral, regional, and bilateral (Lahav & Lavenex, 2012). The multi-layered migration regime complex is illustrated in Figure 1. The shaded circles indicate the confines of the trade regime which constitute the focus of our analysis of EU venue-shopping in the multi-layered regime complex.

Fragmented Multilateralism: Migrant Admission, Rights, and Control
The multiple dimensions of migration policy-admitting migrants, protecting their rights, and controlling territorial borders-lead to different sub-constellations in the international migration regime complex.
Regarding the admission of migrants, multilateralism remains fairly limited. Decisions concerning who is allowed to enter and stay remain in the sovereignty of the nation-states, with only very few exceptions. The main human rights exceptions are the rule of nonrefoulement enshrined in the 1951 Refugee Convention and various human rights treaties, which prohibit the return of refugees and migrants to places where they would fear for their life or liberty, and the right to family reunification preserved in human rights law (Chetail, 2019). While there is no international regime regulating the entry of labour migrants, strictly circumscribed provisions facilitating the temporary mobility of business persons (mainly managers and specialists in multinational corporations) have been negotiated in the framework of the liberalization of trade in services. These provisions figure in the WTO's 1995 General Agreement on Trade in Services (GATS; Lavenex, 2006). Beyond these exceptions, states have omitted multilateral commitments constraining their sovereignty on the admission of migrants and have favoured cooperation at the regional and bilateral levels.
Rather than addressing the conditions under which states shall admit non-nationals on their territory, multilateral treaties and institutions have focused on the rights of migrants who have been admitted onto the territory as a subset of human rights. The 1951 Refugee Convention and its 1967 Protocol lay down the grounds for granting asylum and codify the rights of recognised refugees in the host country.  Figure 1. The multi-layered migration regime complex. Notes: The circles indicate the boundaries of individual regimes attached to a specific set of institutions; the shaded circles indicate the reach of EU trade competence in terms of levels (multilateral, EU intra-regional, EU bilateral) and instruments (linking PTAs and other types of migration-specific bilateral instruments).
received a small number of ratifications-and none from the Global North. In recent years, finally, the Office of the High Commissioner for Human Rights has become more involved in the promotion of general human rights to address the needs of migrants. The issue of migration control is the least codified migration policy objective at the multilateral level and the one that faces the strongest asymmetry of interest between countries of the Global North and countries of the Global South. This includes measures to fight irregular migration or to encourage the re-admission of irregular migrants by their countries of origin. Multilateral conventions address irregular migration only in a very indirect manner in the international travel regimei.e., concerning visa regulations and necessary travel documents as well as the anti-trafficking-smuggling regime, which targets organized crime exploiting (irregular) migrants. Border control, return, re-admission as well as practical cooperation on deterrence have remained in the competence of sovereign states or have been addressed via regional and bilateral cooperation. Corresponding unilateral, bilateral, or plurilateral policies are in tension with multilateral norms on the rights of refugees and migrants more generally (Carrera et al., 2019). Given opposing interests between countries of origin, transit, and destination of migrants (Ellermann, 2008) and considering the human rights focus of existing multilateral institutions, cooperation to fight irregular migration or promote return and re-admission is unlikely to take shape at the multilateral level (Lahav & Lavenex, 2012;Money & Lockhart, 2018).
A special position in this context is occupied by the IOM (next to the ILO and UNHCR). Established as a logistical organization charged with the repatriation of displaced persons after World War II, the IOM has been upgraded to an UN-related organization in 2018 and has expanded its activities beyond the realm of repatriation, return, and re-admission. Lacking a base treaty and being nearly exclusively financed by earmarked voluntary funds, its activities mostly reflect donors' priorities (Pécoud, 2020) and, therefore, can not be easily attributed to a specific section of the international migration regime complex. Finally, and beyond the treaties and organizations previously mentioned, migration governance has punctually been addressed in international bodies of law primarily concerned with different issues such as the Law of the Sea (i.e., the duty to rescue), international travel, and in the 2000 Protocols to Prevent, Suppress and Punish Trafficking in Persons and Against the Smuggling of Migrants by Land, Sea, and Air signed in the framework of the UN Convention Against Transnational Organized Crime.

EU Embedment in the Migration Regime Complex
The EU has developed common instruments in most aspects of this fragmented migration regime complex, yet its competencies remain limited and are shared with the member states. Beyond its internal system of free movement, the EU has developed the strongest competence in the fields of migration control, followed by common asylum policies, and two directives regulating the right to family reunification and the right of long-term third-country nationals living in the member states. The EU's competence over the admission of economic migrants from third countries has remained most limited, with fragmented policies for certain types of labour migration and clear limits codified in Art. 79(5) of the Treaty on the European Union (Geddes et al., 2020).
Paradoxically, it is in this latter area of economic immigration, in which the EU possesses the weakest level of internal competence, that it enjoys formal international actorness-albeit in a very circumscribed realm linked to trade in services. This happens in the WTO, where the EU, as a full member, has negotiated commitments regarding the admission of business persons in the framework of the GATS on behalf of the member states. This competence also applies to EU bilateral trade agreements. Thus, even though migration remains a peripheral issue in trade policy negotiations, the EU's exclusive competence on trade and its market power turn the trade regime into an attractive venue for EU migration diplomacy.
When it comes to migrant and refugee rights, the EU has hitherto not developed into a tangible collective actor in relevant multilateral institutions. EU cooperation on refugees was born out of the necessity to harmonise standards across member states given the dismantling of internal border controls in the Schengen area, but the adopted directives still leave broad discretion to the member states. The EU itself is not a party to international migration conventions and organizations but participates as an observer (Beqiraj et al., 2019). In the refugee regime, the EU participates in the UNHCR's governing bodies (UN General Assembly and ECOSOC) by way of UNGA Resolution 65/276. This upgraded its observer status to allow the EU to speak and make interventions on behalf of the member states. This also applies to the UNHCR Executive Committee. Yet member states retain a distinct voice of their own (Beqiraj et al., 2019). Although figuring among the main donors of the organization, the EU has not developed into a tangible actor within UNHCR. In contrast, UNHCR's influence on the evolving EU refugee policy has been formalised under the Strategic Partnership Agreement signed in 2005, which grants the UN far-reaching consultative rights in the EU's CEAS, including representation in the EU's external borders and asylum agencies.
The limited EU actorness in multilateral institutions also makes it vulnerable to internal divisions, further weakening its influence. The process leading up to the most recent multilateral migration policy initiative, the 2018 UN Global Compact on Migration, is illustrative. The EU was a major force behind this initiative, together with the Obama Administration (Ferris & Donato, 2019), and started negotiating as a bloc via the EU delegation to the UN. After only a few weeks, Hungary stepped out of the common position and made contradictory statements. From then on, the EU only spoke "on behalf of 27 member states" (Melin, 2021, p. 300). This common position gradually fell apart with only 14 member states approving the UN Global Compact on Migration without reservations: five approving it with an explanatory note stressing their national sovereignty, five others abstaining, one no-show, and three members opposing.
Migration control is the field of migration policy where the EU has developed the strongest internal competence while multilateral cooperation is least developed. EU cooperation in asylum and migration matters has emerged in conjuncture with the abolition of internal border controls under the 1985 Schengen Agreement and has focused on "compensatory measures" sustaining member states' capacity to counter irregular immigration from the start (Lavenex, 2018(Lavenex, , p. 1201. As a result, visa policies and measures applying to the control of the external borders, including via the EU borders agency Frontex, are the most integrated area of EU policy today. This cooperation agenda developed an external dimension early on and now encompasses a wide web of migration control arrangements with third countries of transit and origin of migrants. International organizations have come to play a distinct role as subcontractors in the external dimension of the EU's migration control policy via project funding (Lavenex, 2016;Spijkerboer, 2021). This constellation is particularly developed under the 2012 Framework Agreement which foresees a "strategic partnership" (IOM, 2012) between the IOM and the EU's migration, development, and humanitarian policy divisions.
Having established the fragmentary multilateral migration institutions and the evolving EU migration policy competence in broad lines, we now turn to the theoretical framework guiding our analysis of the institutional interplay between EU PTAs and multilateral settings in the multi-layered migration regime complex.

Theoretical Framework: Trade Agreements as Expansion, Complement, or Substitute to Multilateral Norms
While acknowledging the multi-layered nature of international regime complexes, the literature on international regime complexity has tended to focus on constellations of institutional or normative interplay from a more static perspective. As such, the literature distinguishes nested, overlapping, or parallel regimes (Alter & Meunier, 2009); scale, diversity, and density (Eilstrup-Sangiovanni & Westerwinter, 2021); or, more generally, hierarchy versus differentiation (Henning & Pratt, 2021). Taking a dynamic perspective, other scholars have highlighted strategic action within regime complexes such as hostage-taking and brokering (Hofmann, 2018) and forum or venue-shopping (Alter & Meunier, 2009;Jupille et al., 2013). Strategic venue-shopping occurs "where actors select the international venues based on where they are best able to promote specific policy preferences" (Alter & Meunier, 2009, p. 16). Such a move can involve establishing issue linkage with another field where the respective actors enjoy greater influence (Aggarwal, 1998).
In this article, we are particularly interested in how far the EU engages in venue-shopping when promoting migration policy objectives in its PTAs, and which effects this has on the migration regime complex. From this perspective, we distinguish three constellations of institutional interplay. Regime expansion occurs when changes in one sub-regime-here EU PTAs-enlarge the scope of existing multilateral rules. Regime complementarity denotes a parallelism between multilateral and sub-regimes, where the sub-regime reproduces multilateral rules and promotes these in new contexts. Finally, regime substitution takes place when the sub-regime develops norms or rules that do not figure into overarching multilateral settings.
Scholarship on EU actorness highlights several factors that make EU PTAs an interesting candidate for studying venue-shopping and institutional interplay in regime complexes. The brief overview of EU migration policy above has shown that while the EU has few tools (apart from funding) to shape migration policies at the multilateral level, the trade regime stands out as a partly overlapping regime in which the EU enjoys strong actorness. EU actorness implies that member states are encouraged to "speak with one voice" (Bretherton & Vogler, 2013, p. 381;da Conceição-Heldt & Meunier, 2014, p. 962; see also Eeckhout, 2011), thus allowing for common positions. In addition, PTAs can leverage the EU's market power (Damro, 2012) and offer issue linkages, thereby helping to overcome interest asymmetries that hamper cooperation with countries of origin and transit of migrants in multilateral institutions. What is more, the bilateral character of PTAs allows an adaptation of venue-shopping strategies according to the partner country. In all three areas of migration governance (admission, rights, and control) PTAs allow the EU to engage in strategic venue-shopping to further its migration policy objectives.
Regarding migrant admission, EU competence in trade matters allows the Commission to negotiate bilateral commitments on labour mobility falling under the scope of GATS as part of its PTAs, thereby extending the status quo under multilateral labour mobility norms. The EU can however also leverage this market power (Damro, 2012) in PTA negotiations to promote migrant rights and migration control. The com-mercial nature of PTAs is particularly suited to address interest asymmetries via issue linkage, that is, offering economic concessions facilitating mutually beneficial arrangements in areas where actors otherwise disagree (Axelrod & Keohane, 1985). EU competence and market power make PTAs an attractive venue for the EU to shop, especially for those policy priorities which lack multilateral support-i.e., migration control. Such a focus on EU internal priorities echoes the EU's explicit reorientation as a strategic trade policy actor (Cremona, 2017;European Commission, 2015). Such strategic venue-shopping is also alluded to in EU migration policy documents when saying that "the full range of policies and EU external relations instruments have to be brought to bear" to achieve migration cooperation goals (European Commission, 2016, p. 6).
Depending on the type of migration provision included in PTAs, different interplay constellations may emerge within the international migration regime complex regarding relevant multilateral institutions (see Table 1). In the following, we discuss how PTAs as a bilateral venue relate to the multilateral migration regime complex on the three dimensions of admission, rights, and control. Adopting a venue-shopping perspective, we propose hypotheses addressing the scope conditions under which migration provisions in EU PTAs are likely to expand, complement, or substitute multilateral norms.
The most straightforward objective to be sought in a trade policy instrument such as PTAs should be trade facilitation-including mobility in the context of trade in services. The inclusion of service-related labour mobility in the WTO/GATS and EU exclusive competence over commercial policies makes PTAs a privileged venue for widening international cooperation on desired forms of economic migration. These bilateral or plurilateral instruments have the advantage to allow for both the expansion and deepening of commitments in the GATS and for tailoring commitments to the respective trade partners. Given the trade-related base of these provisions, their focus on highly skilled managers and executives, mainly moving within multinational corporations, and European countries' general reluctance towards supranational commitments on the admission of economic migrants, we expect mobility provisions to be more frequent and expand the multilateral status quo, especially, in PTAs with close trade partners. Therefore, our first hypothesis is: For migrant rights and migration control, the inclusion of provisions in PTAs could follow the example of other "non-trade" issues that have found an entry into trade agreements such as environmental protection or labour rights (Lechner, 2019;Milewicz et al., 2018;Raess & Sari, 2018). Regarding migrant rights, we would expect the EU, as a normative power (Manners, 2002), to use its commercial relations to compensate for its limited actorness in international organizations and to promote overarching multilateral norms, such as those contained in migrant rights, including international refugee law. In this way, PTA provisions on migrant rights would complement existing multilateral norms in this field. Since the complementary effect of migrant rights provisions in EU PTAs is greater when the partner country has not ratified relevant multilateral conventions, we should, from a venue-shopping perspective, expect such provisions to be more frequent in PTAs with such countries than in other EU PTAs. Therefore, our second hypothesis reads as such: H2: Migrant rights provisions complement multilateral migrant rights, especially, in PTAs with countries that have not ratified relevant multilateral conventions.
Finally, in the area of migration control, where multilateral rules are absent and EU interests are strongest, PTAs can serve as a substitute for multilateral rules by enforcing deterrence and re-admission in the bilateral setting. In this way, the EU can use its PTAs as a source of leverage and issue linkage inciting the cooperation of countries of transit and origin of migrants where multilateral initiatives would fail. The strategic interest in control provisions should be particularly high towards countries from which the EU faces significant migration pressure in the form of asylum seekers and that are located on the migration routes towards the EU. That is, EU PTAs with such countries should contain migration control provisions more frequently than other EU PTAs. Therefore, our third venue-shopping hypothesis is: H3: Migration control provisions substitute multilateral rules, especially, in PTAs with countries from which the EU faces significant migration pressure.

Data and Methodology
The core of our analysis is based on a novel dataset of migration provisions in trade agreements (Lavenex et al., in press) that covers 109 bilateral and plurilateral trade agreements signed by the EU (among them are 105 bilateral agreements). The dataset provides detailed coding on migration content covering mobility, control, and rights (see the Supplementary File for details and valid-ity check). Mobility provisions facilitate the temporary mobility of specified categories of people such as independent professionals, business visitors, intra-company transferees, and contractual service suppliers (covered by the GATS), as well as other specialists, investors, installers, trainees, or non-business people (such as tourists, students, and researchers). These provisions abolish immigration barriers such as economic needs tests, quantitative limits, or skill requirements and facilitate visa procedures. Migrant rights provisions include commitments to general anti-discrimination clauses and specific economic and social rights, such as equal access to social security, the right to transfer social insurance capital, or access to the labour market for refugees. Control provisions include commitments concerning irregular migration and the re-admission of unauthorized migrants. We create dummy variables for the three types of migration provisions that measure whether the type of provision is included in a PTA as well as a continuous variable of the number of provisions of a particular type included in a PTA.
Our hypotheses mobilize some independent variables. The strength of trade connections (H1) is measured as the share of the bilateral trade volume of the signing parties' GDP, using expanded IMF trade data (Gleditsch, 2002) and GDP statistics from the World Bank. The trade volume can either be calculated based on export figures-Free on Board (FOB)or on import figures-Cost, Insurance, and Fright (CIF). We use the FOB data for our base models and use the CIF data for a robustness check (no difference in the results). For migrant rights (H2), we code the number of migration-related UN conventions a partner country has signed at the moment of concluding each PTA. These are the ILO Migration for Employment Convention (1949), the Refugee Convention (1951), the Convention Relating to the Status of Stateless Persons (1954), the Additional Protocol to Refugee Convention (1967), the ILO Migrant Workers Convention (1975), and the International Convention on the Protection of the Rights of All Migrant Workers and Members of Their Families (1990). Migration pressure in terms of asylum seekers is measured with Eurostat data, using the total number of asylum applications from the partner country in the six EU countries receiving the most applications throughout the study (H3). The trade and asylum variables are lagged by one year and log-transformed to adjust for their skewed distribution. In addition, we include a migration route dummy that captures whether a partner country lies along a main migratory route towards the EU, as identified by Frontex (for an overview of the routes see https://frontex.europa.eu/what-we-do/ migratory-map). Finally, we use the GDP differential between the EU and the partner country as a proxy to control for power asymmetry (GDP per capita of the partner country in percentage of the GDP per capita of the EU). Based on these various data sources, we present descriptive analyses and regression models to test the theoretical predictions outlined above. In the next section, we present the results with the complete model outputs presented in the Supplementary File.

EU Preferential Trade Agreements in the Migration Regime Complex
The analysis of the migration policy content in EU PTAs shows that these trade instruments have indeed become an important element of the multi-layered migration regime complex. The comparison between EU PTAs and all PTAs concluded by other non-EU countries worldwide, in the period of analysis, shows that the EU is unique in introducing all three types of migration provisions to a similar extent, whereas other countries use PTAs mainly to facilitate business mobility (see Figure 2). Put differently, the EU PTAs link up with more dimensions of the migration regime complex than PTAs by other countries.

Mobility Provisions: Venue-Shopping as Selective GATS Expansion
The first mobility provisions figure already in the EU's early association agreements with Greece (1961), Turkey (1963, Morocco (1969), and Tunisia (1969; see Figure 3). These provisions were derived from the rules of the single market and were independent of the EU's migration policy that evolved only from 1992 onwards. Interestingly, agreements after 1970 no longer include free movement provisions, which corresponds to the turn towards restrictive immigration policies at that time. A new, GATS-related type of mobility provision re-emerges and proliferates from 1990 onwards. These PTA commitments often go well beyond the EU's obliga-tions under the GATS, e.g., by including more categories of persons and granting extended periods of stay, visa facilitations, or recognition of qualifications. The EU has also expanded its competence to negotiate mobility provisions via its 2014 Intra-Corporate Transferees Directive that deepens its internal commitments in the matter. In sum, in the case of mobility provisions, EU PTAs expand the scope of the multilateral regime.
Taking a venue-shopping perspective, we hypothesized in H1 that this expansive effect of PTA provisions compared to the multilateral trade regime should concentrate on major trade partners. That is, provisions facilitating labour mobility should be more frequent in PTAs with countries with which the EU enjoys strong trade connections. Figure 4 shows the expected positive association between mobility provisions in PTAs and the EU's trade interdependence with the partner countries. The effect is stronger for the dichotomous variable suggesting that close trade relations primarily increase the likelihood of mobility provisions in a PTA (Figure 4, left side) but less the depth of these provisions in terms of the sum of mobility provisions per PTA (Figure 4, right side). The coefficients remain largely unaltered when controlling for power asymmetry between the EU and the partner country. This result provides tentative support for our hypothesis H1 in that closer trade connections motivate the inclusion of mobility provisions as an extension of multilateral commitments.

Migrant Rights Provisions: A Complement to Multilateral Conventions?
The longitudinal data in Figure 5 shows, similarly to early mobility norms, that migrant rights provisions have  Compared to these international conventions, however, EU PTAs contain only select and fairly general provisions.
References to the 1951 Refugee Convention and refugee rights are even rarer. Therefore, the provisions in EU PTAs are only a weak complement to existing multilateral institutions. What is more, not all PTAs include such provisions (see Figure 2), which calls for an explanation.  From a venue-shopping perspective, we proposed that the EU should seek complementarity to multilateral commitments when negotiating migrant rights provisions in PTAs. Hypothesis H2 hence predicts that the EU seeks to include rights provisions in PTAs to compensate for partner countries' lack of commitments under pertinent international conventions as a means to reinforce multilateral institutions. However, the empirical assessment shows that countries which have not signed relevant conventions are not more likely to have migrant rights provisions in their PTAs with the EUboth for the dichotomous and continuous variable (see Figure 6). If anything, the relationship is inverse because PTAs with partner countries who have signed one or more conventions are more likely to include rights provisions. However, the effect is statistically significant only for the continuous dependent variable. This result does not change when controlling for the GDP differential between the EU and the partner country. Thus, our theoretical expectation that the EU might include rights provisions as a complement to UN conventions cannot be corroborated.

Migration Control: Substitute to Contested Multilateralism?
Migration control provisions appear later than rights and mobility provisions. However, PTAs containing migration control provisions were concluded before the emergence of EU competence on the matter, first in the 1984 Lomé Convention with the African, Caribbean, and Pacific countries. This indicates that PTAs were used as an external migration policy tool before the EU could officially engage in such policies. At the same time, there is no multilateral convention or regime addressing migration control. The inclusion of such clauses and their proliferation from the second half of the 1990s onwards thus situates PTAs as a substitute for a gap in the multilateral regime complex.
Our third venue-shopping hypothesis proposed that control clauses in PTAs are a substitute for missing multilateral provisions on the matter. They should concentrate on PTAs with countries from which the EU faces significant migration pressure. We test this expectation using two proxies for migration pressure: the number of asylum-seekers and a dummy variable of whether a country lies on a major migration route toward the EU.
The results show that indeed PTAs with such countries are significantly more likely to contain control provisions (see Figure 8). This relationship also holds when we control for the level of trade interdependence and the GDP differential between the EU and the PTA partner country. The EU thus seeks to substitute the lack of multilateral norms, via the inclusion of migration control provisions in PTAs, primarily with countries from which it also faces asylum inflows.

Conclusion
International migration has become a core concern of EU policy-makers. The fragmentariness of the international migration regime complex, interest asymmetries between countries, and the EU's shared competence in the matter pose limits to the EU's attempts at establishing itself as an international migration policy actor. This coincidence of high demand and low opportunities for multilateral action in established international migration fora has driven EU migration diplomacy towards bilateral venues, where its institutional competence, leverage, and issue linkages bear higher prospects for strategic and tailored action. Our analysis of the evolving EU migration policy with the migration policy content of all EU PTAs signed between 1960 and 2020 corroborates that the EU engages in venue shopping via trade agreements and shows that this has ambiguous effects on the international migration regime complex. The EU makes much broader use of PTAs for migration policy purposes than other countries. Next to provisions facilitating labour mobility, the EU is practically unique in including migration control commitments and stands out by its frequent inclusion of migrant rights provisions. On the one hand, this underscores the priority that migration enjoys in the EU's policy agenda. On the other hand, the fact that we find these provisions in PTAs substantiates the preeminence of commercial instruments in the EU's foreign policy toolbox.
From the perspective of the international regime complex, this evolution is not without caveats. The mobilisation of trade venues for migration policy purposes can both sustain or constrain multilateral solutions. In this article, we present a conceptualisation of EU actorness in regime complexes that distinguishes between three forms of institutional interplay: expansion, complementarity, and substitution. Whereas the first two constellations sustain multilateral institutions, the third one can create tensions within the multi-layered regime complex when these regional substitutes run against the normative orientation of existing multilateral institutions.
Differentiating between the substantive interplay of migration provisions at the multilateral level and in PTAs (i.e., regulating mobility, rights, and control), EU competence in these matters, and the constellation of interdependence with PTA partners, we hypothesized the conditions under which EU PTAs expand, complement, or substitute multilateral migration rules. Our findings show that the EU expands multilateral institutions where its policy priorities converge and where it enjoys competence as an international actor. This applies to provisions facilitating the mobility and admission of business migrants with close trade partners where the relationship is seen as economically beneficial. Mobility provisions in EU PTAs with close trade partners go well beyond what the EU has committed to under the multilateral setting of the GATS. Paradoxically, the area where EU venueshopping has led to the strongest expansion of international migration commitments is labour migration where EU internal migration policy competence is most contested. This empowerment is also reflected in the adoption of an internal directive harmonizing the admission of this specific type of migrant, namely intra-corporate transferees, in 2014. Even though the scope of migrants benefiting from this trade-related mobility agenda is limited, in institutional terms, the case illustrates in a salient manner how international regime complexity opens avenues for strategic venue-shopping and norm expansion where this would otherwise fail.
An opportunity for complementarity opens up regarding migrant rights, where the multilateral norms are strongest but where the EU enjoys only weak actorness in relevant international organizations. In this constellation, the use of trade policy instruments could be an alternative way to promote multilateral norms in light of limited opportunities for influence in corresponding multilateral fora. Our analysis shows indeed that the EU has included migrant rights in its PTAs early on. However, these provisions have remained very limited in scope, reflecting only a fraction of what is covered by relevant international conventions, and have not expanded over time. Furthermore, and contrary to what a strategic venue-shopping perspective maximising complementarity would suggest, these provisions do not target countries that have not signed the relevant international conventions. Therefore, their inclusion in EU PTAs can be seen as a weak and patchy complement to multilateral institutions at best, showing a rather low level of ambition.
The constellation of institutional interplay for which we find the strongest evidence of strategic venueshopping is regime substitution. This applies to the field of migration control, which has so far remained unregulated at the multilateral level and is where the EU enjoys the strongest internal competence. The objective to protect the EU's external borders and fight irregular migration has been at the top of the EU's developing migration acquis. While multilateral cooperation on deterrence and re-admission faces normative and strategic obstacles, including the human rights focus of international law and interest asymmetries between countries, the bilateral setup of PTA negotiations provides a venue in which the EU can capitalise on its market power and mitigate interest asymmetries via issue linkage. Our analysis corroborates the EU's unique role in promoting migration control cooperation via its PTAs. In line with this venue-shopping perspective, the use of PTAs as a substitute for the lack of corresponding multilateral norms concentrates on countries along migration routes and where large numbers of asylum seekers originate. While these findings are robust, detailed analyses of EU external migration policy towards countries of origin and transit of migrants also point to the limits of the EU's effectiveness in attempting to conclude such deals . This only shows that this cooperation is not without contention and stands in contrast to the rights-oriented focus of multilateral migration fora.
In more conceptual terms, our analysis provides an innovative framework for studying EU actorness and impact in regime complexes via the use of its strongest foreign policy tool, PTAs. Depending on how coherent EU policy priorities are with existing multilateral instruments, how much competence the EU enjoys vis-à-vis its member states in the matter, and the constellation of interdependence with third-countries, the EU can engage in strategic venue-shopping, thereby altering the architecture and contents of multi-layered regime complexes. Whereas our focus here was on PTAs, our analytical framework can also be mobilised to examine EU external action under the premise of regime complexity concerning other bilateral or plurilateral policy instruments. Next to conceptualizing EU strategic venue-shopping, our analysis contributes to the broader literature on international regime complexes by distinguishing three constellations of institutional interplay in a multi-level perspective: extension, complement, and substitution, as well as their scope conditions.
Beyond putting EU bilateral foreign policies in a broader normative context, this regime complexity approach to EU external action has allowed us to disclose strategic and structural features of EU external action that would otherwise remain concealed. As such, we showed that by shifting the action to venues where it enjoys stronger clout, the EU can expand international norms even when its internal competencies are limited. Conversely, we found that EU bilateral outreach is not necessarily in harmony with its vocation toward multilateralism. Thus, the expansion of mobility norms beyond the GATS remains selective and concentrated on business migrants from main trade partners. At the same time, migrant rights provisions, which are at the heart of multilateral conventions, are not a target in PTAs. Instead, the EU is strategically using its PTAs to foster cooperation where multilateralism (be it because of normative concerns or interest asymmetries) fails. Even without turning into a bedrock of multilateral institutions, EU bilateral action may thus have a lasting impact on the multi-layered migration regime complex and alter its substantive focus from below.