Deference and Hierarchy in International Regime Complexes

Transcription

1 Deference and Hierarchy in International Regime Complexes August 18, 2016 Abstract How do states allocate governing authority across overlapping institutions? Global governance of many issue areas is increasingly fragmented among multiple international organizations (IOs). This fragmentation is believed to undermine cooperation as different institutions adopt conflicting rules. However, existing work overlooks the potential for inter-institutional coordination. This paper develops a novel theory of institutional deference: the selective conferral of authority by one IO to another. By granting another organization jurisdiction over a specific sub-issue, member states can mitigate rule conflict and facilitate a division of labor within the regime complex. I measure institutional deference using an original dataset of over 2,000 IO policy documents in the counterterrorism, intellectual property, and election monitoring regime complexes. Empirical tests support two key theoretical claims. First, institutional deference is indeed associated with a division of labor among institutions: IOs that defer to each other are more likely to focus their rule-making efforts on separate sub-issues. Second, deference is a strategic act that is shaped both by efficiency concerns and power politics. Statistical tests confirm that deference is used to efficiently pool resources among disparate organizations, and that IOs with weaker member states tend to defer to organizations with more powerful members. 1

2 1 Introduction In 2005, the United Nations Security Council (UNSC) adopted Resolution 1617, extending the Council s financial and travel sanctions against al-qaeda and affiliated terrorist groups. 1 The resolution was an opportunity for the Council to articulate a set of legally binding rules regarding state efforts to combat terrorist financing, 2 an increasingly important component of global counterterrorism efforts. But Security Council members passed on this opportunity. Instead of designing its own rules for counterterrorism finance, the UNSC opted to defer to those of another international organization, instructing member states to implement the comprehensive, international standards [of] the Financial Action Task Force (FATF). 3 The FATF is a relatively obscure, informal, and poorly staffed organization with no founding treaty or formal authority of international law. 4 The UN Security Council, by contrast, is a legally empowered international institution with abundant administrative and technical resources. By deferring to the FATF, the Security Council effectively made all UN member states subject to a set of rules created by a handful of states in an informal institution. Why did the Council choose to elevate FATF rules for terrorism finance instead of articulating its own in effect, conferring authority to a weaker institution? More broadly, why would any international organization (IO) offer jurisdiction to another rather than extending its own? In this paper, I explore the practice of institutional deference the conferral of authority by one IO to another and argue it is a strategy used by IO member states to manage and mitigate potential rule conflict. In many issue areas, contemporary global governance is fragmented among multiple international institutions that simultaneously attempt to regulate state behavior. The presence of multiple institutions governing a single issue, deemed 1 S/RES/1617 (2005) 2 The resolution was adopted under Chapter VII of the Charter of the United Nations, which gives the UNSC the authority, in the case of a threat to international peace, to decide what measures shall be taken to restore international peace and security (Charter of the United Nations, 1945). 3 Operative Paragraph 7. 4 The FATF maintains a small permanent staff and relies on Secretariat of the Organization for Economic Cooperation and Development (OECD) for much of its administrative and budgetary work (see http: // 2

3 an international regime complex in the literature, often leads to inefficient overlap and inconsistencies in international rules. To avoid these problems, international organizations coordinate rulemaking via institutional deference. Just as the UN Security Council recognized the authority of the FATF to regulate counterterrorism finance, IOs often selectively confer jurisdiction over a particular sub-issue to another institution. This allows institutions to avoid the adoption of conflicting rules and to capture efficiency gains through division of labor. Unlike the formal nesting of one IO in another, institutional deference does not constitute a permanent transfer of authority. On the contrary, conferring jurisdiction on another organization is often contingent, issue-specific, and reversible, particularly when the conferring organization such as the Security Council relative to the FATF has been granted superior formal authority by states. Yet institutional deference is a common and significant phenomenon that scholars of international relations should seek to explain. The use of institutional deference confounds the conventional wisdom about international regime complexes, which are often portrayed as poorly coordinated and harmful for cooperation. As Abbott et al. (2015) state, Typically, regime complex theory treats the co-existence of multiple governance actors with overlapping mandates as a pathology ( overlap or fragmentation ) that threatens governance effectiveness through redundancy, inconsistency, and conflict. 5 By highlighting deference as a mechanism for coordination among institutions, this paper challenges the prevailing view of regime complexity. 6 More broadly, I argue that rather than assuming that regime complexes are pathological for cooperation, scholars should reconsider how they judge the effectiveness of international institutions. In a single, unified regime this judgment is usually made on the basis of depth of cooperation, defined as the extent to which [the institution] requires states to depart from what they would have done in its absence. 7 In regime complexes, however, effective cooperation demands not only depth but also inter-institutional coordination. In other words, 5 Abbott et al. (2015, 7) 6 For exceptions to the generally pessimistic view of regime complexes, see Gehring and Faude (2014), Abbott et al. (2015), and Keohane and Victor (2011). 7 Downs et al. (1996, 383) 3

4 assessing cooperation in regime complexes requires investigating how institutions interact with each other in addition to the obligations they place on member states. The paper proceeds as follows. Part II shows why international organizations would engage in the practice of institutional deference, which serves as a functional solution to inefficient overlap and inconsistent rules among IOs. Part III describes in detail the use of deference in three issue-areas: counterterrorism, intellectual property, and election monitoring. Using a dataset of over 2000 IO policy documents, I demonstrate that deference is a frequent and recurring tool used by IOs. In addition, I use a novel application of topic modeling to show that deference matters i.e., that it is indeed associated with a division of labor among overlapping institutions. Finally, Part IV explains why some IOs are privileged over others in the granting of regulatory authority. I show that the direction and intensity of deference reflect two broad processes. First, deference is used to overcome cooperation problems by pooling resources among different organizations (e.g., partnering an IO with technical expertise with one that has binding legal authority). Second, deference is shaped by state power: IOs with weaker member states tend to defer to organizations with more powerful members. 2 Why Defer? Institutional Density and Rule Conflict During the last several decades, the proliferation of international organizations has altered the basic multilateral architecture that governs interstate cooperation in many issue areas. Issues such as trade and global health, which were once regulated by relatively integrated regimes, are now governed by a complex network of distinct institutions with partially overlapping mandates and memberships (Raustiala and Victor, 2004). The IOs that comprise a regime complex do not always act in concert. As Raustiala and Victor state, The rules in these elemental regimes functionally overlap, yet there is no agreed upon hierarchy for resolving conflicts between rules (Raustiala and Victor, 2004, 279). Independent rule-making by multiple IOs often results in a fragmented and potentially 4

5 conflicting set of international rules and standards. 8 In other words, a common feature of regime complexes is inconsistency in law or regulatory practice. Much of the literature on regime complexity focuses on the potential for rule conflict to undermine international cooperation. Inconsistent rules among IOs can raise a number of potential barriers to cooperation, including increasing uncertainty and raising transaction costs across regulatory jurisdictions. I briefly highlight two problems with inconsistent rules that have received the most attention in the literature: reduced compliance via regulatory arbitrage and inefficient provision of goods. Regulatory Arbitrage and Inefficient Duplication Inconsistent rules may reduce compliance when the same actors are subject to multiple regulatory jurisdictions. States can leverage loopholes and incongruencies in regulatory systems to circumvent costly rules a process known as regulatory arbitrage 9. If the compliance costs of conflicting regulations differ, states will prefer to recognize the authority of the lowest-cost regulator. In some cases, this dynamic can create a race to the bottom effect that empowers the weakest regulatory authorities. States may also leverage inconsistencies to escape compliance with burdensome requirements for example, by exploiting gaps in regulatory authority, or by claiming the absence of a clear global standard on a particular issue. 10 and standards. The net effect is a reduction in compliance with international rules, norms, Chinese and Russian use of the Shanghai Cooperation Organization provides a useful example of arbitrage in the counterterrorism regime complex. Several IOs, including the United Nations (UN) and the Organization for Security and Co-operation in Europe (OSCE), re- 8 The literature on regime complexity is abundant with examples: Raustiala and Victor describe legal inconsistencies in the regime complex for plant genetic resources, and further argue that legal conflict among overlapping rules...is a recurring and difficult challenge for regime architects. (Raustiala and Victor, 2004, 300) Similarly, Helfer finds institutions adopting a competing regulatory approach in the intellectual property regime complex,(helfer, 2009, 40), and Davis notes the potential for contradictory legal rulings among the set of institutions governing international trade.(davis, 2009, 25) 9 Riles (2014) 10 Riles notes that regulatory arbitrage depends on a rich ecosystem of diverse regimes and types of laws, which are not organized into any clear, coherent, hierarchical whole a close description of an international regime complex(riles, 2014, 65). 5

6 quire member states to ensure their national counterterorrism policies protect basic human rights, such as the right to freedom of expression and the right to seek asylum. 11 In 2004, the Shanghai Cooperation Organization, a regional organization comprised of Russia, China, and the Central Asian states, adopted a set of counterterrorism rules that conflicted with these basic principles. The SCO employs a very expansive definition of terrorism, explicitly conflating terrorist acts with the co-equal evils of separatism and extremism. It has also adopted controversial policies related to extradition and denial of asylum for individuals suspected of participating in these activities. 12 In practice, SCO members have leveraged these rules to escape compliance with international human rights obligations. 13 Unfortunately, this behavior is far from unique in international regime complexes. States and other targets of regulation often opportunistically select among multiple regulatory authorities (forum shopping) or empower new regulatory bodies (competitive regime creation) to avoid complying with international rules they do not like. 14 The second major problem associated with overlapping institutions is inefficient duplication of efforts. Duplication occurs through the promulgation of redundant rules, as well as efforts to implement rules such as capacity-building or information exchange. In many cases, this regulatory effort requires the expenditure of significant material and human resources. In the regime complex for intellectual property rights, for example, the articulation of rules can require lengthy negotiations and a high level of legal expertise. Duplication among mul- 11 For a discussion of UN rules regarding the protection of human rights while countering terrorism, see For OSCE rules, see download=true. 12 For a compendium of official SCO documents including its counterterrorism-related rules and standards, as well as a detailed discussion of how SCO policies relate to other international standards, see Human Rights in China (2011). 13 Notably, SCO members have signaled that the rules of the SCO, despite its informal nature, may receive precedence over globally recognized international legal commitments. In an interview with the NGO Human Rights in China, a high-ranking official in Kyrgyzstan indicated that the Kyrgyz governments decision to return Uzbeks fleeing the 2005 Andijan crackdowndespite the likelihood that they would be tortured or executed upon their returntook place after he had weighed the extradition requirements of the SCO treaties against the prohibitions of the Convention against Torture. In the face of these conflicting obligations, the Kyrgyz government had concluded that the SCO framework took precedence. 14 Many of the behaviors scholars have linked to regime complexes, including forum shopping, regimeshifting, and other behaviors associated with contested multilateralism, are manifestations of regulatory arbitrage. See Helfer (2004), Raustiala (2006), and Morse and Keohane (2014), respectively, for a discussion of forum shopping, regime-shifting, and contested multilateralism. 6

7 tiple institutions, such as the World Intellectual Property Organization (WIPO), the World Trade Organization (WTO), and the World Health Organization (WHO), generates inefficiency costs and limits the quantity of regulation IOs in a regime complex can provide. Scholars have previously noted how rule conflict among IOs can generate barriers to cooperation. However, existing work has failed to take seriously the argument of rational institutionalist theory, which argues that states create international institutions in order to generate gains from cooperation (Keohane, 1984). This theory implies that actors in world politics will recognize the costs associated with inconsistent rules and will take action to ensure that institutions are not undermined by regulatory arbitrage and inefficient duplication. 15 In the counterterrorism regime complex, policymakers are keenly aware of the dangers of rule conflict and have adopted inter-institutional coordination as a key policy goal. 16 States respond to regulatory arbitrage and inefficient duplication by directing IOs to mitigate these problems. 17 The result has been a significant but largely unexamined effort by IOs to coordinate rule-making and capacity building efforts. In the regime complex for intellectual property, for example, the three institutions referenced in the preceding paragraph have established a regular consultative process for strengthening their cooperation and practical coordination on issues around public health, intellectual property and trade. 18 Highlighting efforts to mitigate rule conflict among IOs begs the question of how regulatory inconsistencies emerged. If, as I argue, states have incentives to encourage coordination among IOs, why did they construct multiple institutions in the first place? A full treatment of regime complex emergence is beyond the scope of this paper, but in the majority of cases regime fragmentation occurs as a byproduct, rather than an intentional goal, of state behav- 15 Surprisingly few scholars have investigated efforts by IOs to coordinate their rules, norms, and standards, despite its clear connection to cooperation in the context of overlapping institutions. Biermann et al. (2009) s study of inter-organizational networking and Gehring and Faude (2014) s examination of institutional division of labor are notable exceptions. 16 Interview by the author with State Department counterterrorism official. August 27, U.S. Department of State. 17 I assume states are the primary actors that drive IO behavior. While I do not preclude the possibility that actors within the IO (e.g., secretariat officials) may have some agency in shaping institutional deference, it is the member states who create the institution and maintain ultimate authority over its activities. 18 WHO, WIPO, WTO Trilateral Cooperation on Public Health, IP and Trade, policy/en/global_health/trilateral_cooperation.html. 7

8 ior. Scholars have argued states proliferate institutions to overcome bureaucratic capture (Urpelainen and Van de Graaf, 2015), reconcile bargaining power asymmetries in existing IOs (Pratt, 2016), and provide strategic leverage to influence rule-making in the original institution (Helfer, 2004). Multiple regulations may also arise due to unexpected decisions made by international courts (Morse and Keohane, 2014). In each of these cases, states do not intentionally incur the costs associated with multiple institutions, though they are sometimes willing to pay those costs to achieve other means. As Gehring and Faude (2014) argue, even when states disagree about the specific content of institutions, they have a larger interest in ensuring IOs operate in a complementary fashion. Nevertheless, the process of regime complex emergence is undoubtedly endogenous to state behavior a fact I return to in the empirical section. 2.1 Institutional Deference: a Mechanism for Coordination Faced with the potential for regulatory arbitrage and duplication, member states of IOs have incentives to mitigate the problem of inconsistent rules. One remedy is full unification or harmonization of IO rules and standards. If the targets of regulation face a single set of rules, the costs of compliance are constant across regulatory authorities, and regulatory arbitrage ceases to be an attractive strategy. Harmonization also alleviates transaction costs, since rules do not differ across regulatory jurisdictions. This strategy, however, can be difficult to implement if not completely infeasible. 19 IOs in a regime complex typically have different member states, mandates, and decision-making mechanisms, usually with no formal hierarchy of authority. The likelihood of all states agreeing to a single, comprehensive set of rules is therefore often quite low; unless states have an unusually high level of interest convergence, they will generally be unwilling to formally and permanently delegate to a single, unified institution. 20 Institutional integration is especially difficult if multiple IOs already lay claim to governance of an issue area, since the elimination of existing institutions will be 19 See Riles (2014) for a discussion of the challenges and costs associated with harmonization. 20 As Downs et al. (1996) have argued, even if a single universal institution were feasible it would necessarily lack the depth of a more exclusive cooperative arrangement. 8

9 opposed by states and IO bureaucrats with vested interests. When harmonization proves unrealistic, regulatory bodies in other contexts have turned to an alternative approach: re-defining the scope of each body s regulatory authority to address potential overlaps in jurisdiction. Instead of agreeing on a single set of rules, this approach mitigates legal inconsistencies by defining how far each regulatory authority extend[s], and what should be done when these overlap. 21 In the realm of international law, this approach is known as the Conflict of Laws or Private International Law. 22 Its goal is to create a sufficiently coordinated set of laws, rules, and standards by assigning areas of contested jurisdiction to a single regulatory authority. This strategy is more feasible in practice than complete harmonization, since each act of delegation is limited in scope and does not require the elimination of any regulatory authority. A primary argument of this paper is that international organizations mimic this Conflict of Laws process. When the potential for rule conflict arises, one IO often engages in institutional deference, defined as the selective conferral of authority by one institution to another. Institutional deference represents a coordination strategy to allocate governing authority among IOs in the same regime complex. In the introduction, I described how UN Security Council members opted to defer to the FATF s rules for terrorist finance rather than articulating their own regulations. By doing so, the Council avoided setting multiple, potentially conflicting sets of rules. Other organizations involved in the governance of counterterrorism policy, including the Asia Pacific Economic Cooperation (APEC) forum, have similarly deferred to FATF rules. APEC committed in 2002 to supporting the FATF s eight Special Recommendations on terrorist financing and pledging to comply as quickly as possible with the recommendations. 23 In a similar act of deference, in 2013 the Organization for American States (OAS) pledged to help member states comply with the International Civil Aviation Organization (ICAO) standards and recommended practices. 24 The Organization for Security and Cooperation in Europe (OSCE) also adopted ICAO s rules, deciding in 21 Riles (2014, 66) 22 See Collins et al. (2006) for the definitive treatment of the Conflict of Laws and associated legal concepts. 23 APEC (2002) 24 OAS (2013) 9

10 2003 that all OSCE participating states should aim to comply fully with the recommended ICAO minimum security standards. 25 These are explicit attempts to confer authority on a specific sub-issue (in this case, terrorism finance and aviation security) to a single institution, despite a lack of any formal legal hierarchy. Other examples of institutional deference are less conspicuous, with one institution implicitly adopting the rules of another as it regulates state behavior. This is prominent among some institutions in the regime complex for election monitoring, where multiple IOs articulate rules for fair and open elections and assess compliance via election monitoring missions. In a 2008 electoral observation report, for example, the European Union (EU) evaluates the fairness of Nepal s Constituent Assembly elections based not only on compliance with EU standards, but with rules adopted by both the OSCE and the Council of Europe (COE). 26 Institutional deference has several features that make it attractive to states as they coordinate rule-making across IOs. It can be used to target specific issues where the potential for multiple rules is particularly costly. It can also be granted on a contingent basis and reversed if necessary. Both of these features distinguish deference from the alternative coordination strategies of regime unification or the formal nesting of one IO in another. Deference is also distinct from the related concepts of delegation (Hawkins et al., 2006) and orchestration (Abbott et al., 2015), which imply a well-defined pre-existing power relationship between a principal, who issues a contract for a specific activity, and an agent, who would not engage in that activity in the absence of the contract. The use of deference does not does not require any specific ex ante distribution of authority among organizations. When deference is granted by formal, legally empowered institutions (e.g., the UNSC), it confers legal authority on the rules developed by another IO. When it is granted by informal institutions (e.g., APEC), it bestows legitimacy and capacity-building resources on a different IO s rules. As the preceding examples illustrate, cooperation in international regime complexes is not necessarily condemned to suffer from arbitrage, noncompliance and duplication. In at least some instances, IOs coordinate in order to reduce or eliminate overlaps in governing au- 25 OSCE (2003) 26 EU (2008) 10

11 thority. In other cases, IOs in a regime complex will be unable to reach agreement regarding whose authority should be privileged when their jurisdictions overlap. But the challenges involved in institutional deference do not justify the widespread assumption that regime complexity is pathological for cooperation. Instead, assessments of cooperation in regime complexes should be informed by how successfully its constituent institutions are able to coordinate their rules and standards. In the next section, I measure patterns of institutional deference in three of the regime complexes discussed above: counterterrorism, intellectual property rights, and election monitoring. I use the text of IO policy documents to demonstrate the frequent use of deference among international institutions, and I compare patterns of deference in each regime complex. I then demonstrate that deference is indeed used to coordinate among IOs: organizations that defer to each other are more likely to focus their subsequent rule-making on separate issues. 3 Patterns of Deference How does institutional deference work in practice? In their analysis of international organizations governing plant genetic resources (PGR), Raustiala and Victor suggest coordination among institutions will not necessarily occur at the initial stage of rule formation. When IOs first begin to articulate rules governing a specific issue area, member states negotiate a broad set of rules that are then formally adopted by high level officials. During this process, member states may be uncertain about how the issue will evolve and what other international organizations will attempt regulate the issue in the future. Anticipating the potential for rule conflict, member states instruct their diplomats [to] first negotiate broad ex ante rules and then defer the task of working out detailed implications to the process of implementation (Raustiala and Victor, 2004)[280]. In other words, we should observe few efforts at inter-institutional coordination in the formal treaties that initially set an IO s mandate to govern an issue area. Institutional deference, if it indeed occurs, should be found in the subsequent set of IO policy documents produced in order to implement the obligations 11

12 included in those treaties. In contrast to the rarity of high-level treaties, IOs are constantly articulating policy documents containing specific rules, norms, and standards their member states are expected to follow. These policy documents are meant to implement broader goals set by the leaders of member states. They are often written in issue-specific subsidiary bodies within an IO, or they may be developed by the organization s secretariat. Policy documents take a variety of forms including resolutions, policy guidelines, best practices, mission reports, and codes of conduct and while they typically do not have the binding force of a formal treaty, these documents are useful for coordinating with other institutions on specific sub-issues. If IOs defer to other institutions, it should be visible in their corpus of policy documents. 3.1 Data Collection For the purposes of this study, an act qualifies as deference if it constitutes a documented conferral of authority approved either by member states themselves or by institutional actors empowered by member states to make decisions for the organization. To measure the occurrence of institutional deference, I collected a large sample of policy documents adopted by IOs in three regime complexes: election monitoring, counterterrorism, and intellectual property rights. I selected these particular regime complexes for two reasons. First, they represent a diverse variety of topics, from human rights to security to economic issues. Each issue area has its own set of cooperation problems that complicate the task of coordinating global governance. 27 Second, each regime complex has been the focus of previous scholarly attention. The existence of previous work helps to establish the conceptual bounds of each issue and limits subjectivity in determining which international organizations should be considered participants in each regime complex. To find relevant policy documents in these regime complexes, I first identified the set of multilateral IOs that participate in global governance of the issue. To be considered part of the regime complex, an IO must formally include the relevant issue (e.g., election mon- 27 Additional details on each regime complex, including summary statistics, are in the Appendix. 12

13 itoring) in its institutional mandate or actively regulate state behavior in the issue area. 28 Even with this criteria, there may be reasonable disagreement regarding which IOs govern a particular issue. A broad conception of intellectual property, for example, would encompass a wide range of related issues regulated by dozens of organizations. To narrow the field of candidate IOs, I drew on Kelley s 29 and Hyde s 30 extensive examination of the election monitoring regime complex; Helfer s work on the intellectual property regime complex; 31 and Rosand s study of the counterterrorism regime complex. 32 Table 1 displays the IOs included in the regime complexes for election monitoring, counterterrorism, and intellectual property rights, as well as the year of entry for each IO This excludes IOs that occasionally reference an issue area but cannot be said to participate in global governance of the issue. IOs which condemned terrorism immediately after the 9/11 attacks, for example, do not automatically become part of the counterterrorism regime complex. 29 Kelley (2009, 2012) 30 Hyde (2011) 31 Helfer (2004, 2009) 32 Rosand (2006) 33 An international organization can join a regime complex in two ways. First, a new IO may be constructed in order to govern interstate cooperation in the relevant issue area; in these cases, the year of entry corresponds to the creation of the IO. Second, an existing IO may expand its mandate to include the relevant issue area; the year of entry then corresponds to the year in which the issue area was formally added to the IOs mandate. For these IOs, the time of entry often coincides with the establishment of a subsidiary body to govern member state behavior in the issue area. For example, the OSCE added an Action against Terrorism Unit (ATU) in 2002 when it began to participate in the regulation of state counterterrorism behavior. 13

14 Table 1: International Organizations in the Election Monitoring, Counterterrorism, and Intellectual Property Regime Complexes Regime Acroynm Organization Year Added Election OAS Organization of American States 1962 Monitoring AU African Union 1989 CS Commonwealth Secretariat 1989 COE Council of Europe 1989 OSCE Org. for Security & Cooperation in Europe 1990 UN United Nations 1990 EU European Union 1993 ECOWAS Economic Community of West African States 1997 SADC Southern African Development Community 1997 CIS Commonwealth of Independent States 2001 Counter- SAARC South Asian Assn. for Regional Cooperation 1987 terrorism AU African Union 1999 OAS Organization of American States 1999 UNSC United Nations Security Council 1999 COE Council of Europe 2001 FATF Financial Action Task Force 2001 G8 Group of Eight 2001 ICAO International Civil Aviation Organization 2001 NATO North Atlantic Treaty Organization 2001 ARF ASEAN Regional Forum 2002 APEC Asia-Pacific Economic Cooperation 2002 OSCE Org. for Security & Cooperation in Europe 2002 EU European Union 2004 SCO Shanghai Cooperation Organization 2004 UN United Nations 2006 GCTF Global Counterterrorism Forum 2011 Intellectual WIPO World Intellectual Property Organization 1967 Property CBD Convention on Biological Diversity 1994 FAO Food and Agriculture Organization 1994 WTO World Trade Organization WHO World Health Organization 1996 UN United Nations 2000

15 For each year that a particular international organization is part of a regime complex, I collected all IO documents that articulate rules relevant to the issue area and are publicly available from the IO s website. 34 Examples of these documents include texts of resolutions, decisions, treaties, technical guidelines, codes of conduct, mission reports, and best practices. For the election monitoring, counterterrorism, and intellectual property regime complexes, this collection yields a total of over 2,000 policy documents. I then use statistical software to search through each document to see whether it makes reference to any other IO in the regime complex. These cross-institutional references are then hand coded on a 5-point scale, with 1 denoting a passing or irrelevant reference and 5 (deference) representing a clear conferral of authority to another IO. Missing references documents in which one IO fails to reference another are coded as zero. Further details of the coding scheme can be found in the appendix. The reference scores are then used to calculate a measure for institutional deference for each directed IO dyad in a given year. To measure institutional deference from IO A to IO B in year t, I calculate the proportion of IO A s policy documents from year t that include at least one instance of deference (a score of 5) to IO B. The resulting dataset of institutional deference includes 3,718 unique directed IO dyad-years. 34 Policy documents were retrieved from the public websites of each international organization. All IOs examined in this paper maintain websites where policy documents are stored. 15

16 3.2 Deference in Three Regime Complexes A review of the deference data reveal three notable trends. First, deference is a common practice among IOs in the same regime complex. The 32 institutions in three regime complexes examined in this paper engaged in over 1,250 unique instances of institutional deference in the time period under study. 35 Even if each particular occurrence of deference represents a minor conferral of authority, collectively they represent a significant redistribution of governance power in the regime complex. Overall, approximately 14% of IOs defer to at least one other organization in a given year. This rate differs slightly across issue areas (16.8% in intellectual property vs. 14.4% in counterterrorism and 11.9% in election monitoring) but is commonplace in all three regime complexes. Second, deference from one IO to another is not necessarily reciprocated. In theory, two IOs occupying the same regulatory space could strike a quid pro quo bargain, engaging in mutual deference to each other on separate sub-issues. In practice, however, patterns of institutional deference appear asymmetric: among pairs of IOs, one tends to dominate in terms of the total flow of deference. Figure 1 displays asymmetric deference between the UN Security Council and other institutions in the counterterrorism regime complex in the year As the graph illustrates, the UNSC is granted a high level of deference from many institutions but does little deferring in return. Only three IOs (ICAO, the FATF, and the Council of Europe) are recipients of institutional deference from the Council. 35 Temporal coverage of each regime complex begins the first year that more than one institution participates in governance of the issue. See the Appendix for additional details. 16

17 Deference (% of Policy Documents) Deference to UN General Assembly Deference from UN General Assembly G8 UNSC SAARC FATF OAS OSCE EU COE APEC ARF SCO Figure 1: Institutional Deference with UN Security Council: The figure shows the level of institutional deference granted to the UN Security in the year 2012 by other IOs in the counterterrorism regime complex regime (dark bars). Light bars represent deference granted by the UNSC. Third, asymmetric deference generates an informal hierarchy of authority among international institutions. When uneven patterns of deference persist over time, the jurisdiction of certain IOs expands while the authority of others contracts. This redistribution of governing power repositions organizations with no formal chain of authority into a clear hierarchical relationship. As noted earlier, the hierarchy arising from institutional deference is more informal and less deliberate than the practice of legally nesting a new IO within an existing institution. 36 Each instance of deference may be a relatively small conferral of authority on a specific sub-issue; it is only in the aggregate that a clear hierarchy becomes visible. Figure 2 demonstrates this hierarchy in each of the three regime complexes using a net- 36 Aggarwal (1998) 17

18 work graph. IOs are represented as nodes (circles) in the network, which are connected to each other via directed flows of institutional deference (lines). The size of each IO is proportional to the total amount of deference granted to the institution by other IOs in the regime complex. Line thickness represents the amount of deference granted from one IO to another. Intellectual Property Election Monitoring Counterterrorism CBD FAO OAS ECOWAS G8 ARF NATO SCO SAARC WTO WHO WIPO UN CIS OSCE COE EU UN AU SADC ComSec OSCE ICAO COE EU OAS UN UNSC FATF GCTF APEC AU Figure 2: Networks of Institutional Deference: Patterns of institutional deference in three regime complexes are visualized as network graphs. The thickness of each edge (gray lines) is proportional to the level of deference between organizations. Nodes (circles) are sized relative to the total amount of deference granted to the IO. The figure demonstrates that the degree of hierarchy differs across each regime complex. Counterterrorism institutions display a clear hierarchical structure, with deference flowing overwhelmingly to the UN General Assembly and UN Security Council. The election monitoring regime complex is comparatively flat, with at least some deference granted to each organization and some asymmetry in the flow of deference. The intellectual property regime complex has the flattest structure, with deference distributed horizontally among institutions. Section IV will examine the determinants of deference flows in greater detail, but a 18

19 few broad trends are immediately evident in the figure. First, IOs with the power to set universal, binding rules receive more deference than regionally-defined institutions or those that lack the force of international law. In particular, the UN and UN Security Council loom large in each regime complex. Second, the extent to which deference flows are hierarchical differs dramatically across issue areas. Counterterrorism is an issue area where divergent rules are particularly costly; noncompliance by a subset of states threatens the viability of the entire regime. 37 It is also an issue where states largely agree on the basic goals of the regime. These features facilitate the emergence of a few powerful institutions as clear focal points for coordination in the regime. In both election monitoring and intellectual property, however, clubs of states can sustain cooperative outcomes even if others follow a different set of rules. The pressure for coordination and thus the structure of institutional deference patterns are therefore attenuated Division of Labor Now that deference has been established as a practice that generates systematic patterns of relations among IOs, an important question remains: does it matter for how IOs actually govern? To demonstrate that deference is used as a tool to mitigate overlapping rules, I examine the division of labor among counterterrorism institutions. This issue area is wellsuited for examining whether deference is associated with a division of labor among IOs. Contemporary cooperation on counterterrorism includes a broad range of sub-issues, includ- 37 In other words, because terrorists can organize and raise funds in one country to attack another, noncompliance generates very high negative externalities in the counterterrorism issue area. 38 I thank an anonymous reviewer for these valuable insights. 19

20 ing transportation security, terrorist financing, criminal justice, and immigration. Global governance of the issue has accordingly been clustered by sub-issue, with IOs elaborating separate rules for terrorist financing, intelligence sharing, and aviation security. This governance structure allows us to examine the extent to which IOs in the counterterrorism regime complex divide regulatory effort. Do IOs tend to focus on the same substantive areas, with each institution articulating its own rules for aviation security, criminal justice, and other issues? Or do some IOs, as hypothesized, use institutional deference to resolve jurisdictional overlaps and focus their rule-making efforts on separate topics? The counterterrorism regime complex evolved largely in response to the 9/11 attacks. While many terrorism-related issues were subject to regulation in international regimes prior to 9/11, 39 the tendency to group such issues under the rubric of counterterrorism rose markedly after The regime complex that took shape in the years includes institutions within the United Nations (UN) system, technical IOs focused on particular subissues, and counterterrorism bodies nested within regional institutions. A detailed narrative of the regime complex and its constituent institutions can be found in the appendix. To measure division of labor among counterterrorism institutions, I use a structural topic model to estimate the relative attention IOs devote to particular sub-issues each year. A topic model is a statistical tool for estimating latent themes, or topics, in a body of text. 40 Given a large number of documents, a topic model can inductively recover the primary top- 39 For example, the International Civil Aviation Organization (ICAO), which regulates the safety, security, and efficiency of international air travel, was established in 1944 ( icao-in-brief.aspx). The Financial Action Task Force (FATF) was established in 1989 to facilitate cooperation on measures to combat money laundering ( 40 In a topic model, topics are probability distributions over the set of terms that appear in the documents. See Blei (2012) for an introduction to topic modeling methodology and an overview of existing models. 20

21 ics discussed in the text, identify the words most closely associated with each topic, and estimate the proportion of each document devoted to each topic. 41 I estimate a 10-topic structural topic model on the 677 policy documents produced by IOs in the regime complex from Many of the ten topics identified by the model are closely associated with functional sub-issues that arise in the regulation of states counterterrorism policies. For example, one topic features the terms money, launder, and fund among its most frequent words, clearly signaling the topic of terrorism finance. Other topics focus on preventing nuclear terrorism (signified by words such as nuclear, materi[al] and safet[y] ), criminalizing terrorist acts ( crime, investig[ate], court ), and protecting human rights ( right, human, protect ). Table A2 in the appendix displays the ten topics along with the words that appear most frequently in each topic. Given the results of the topic model, it is straightforward to calculate the total attention each IO devotes to each functional sub-issue in a given year. This measure of regulatory attention can be used to approximate a division of labor among IOs. Deference generates a division of labor only if IOs subsequently differ in their rule-making priorities: IO A chooses to confer authority over an sub-issue to IO B and instead focuses on separate matters. If this occurs, we should see the regulatory priorities (and thus the topical focus) of two IOs diverge as they engage in deference with each other. 41 Structural topic models improve on this basic technique by allowing for the incorporation of document metadata, such as authorship and year (Roberts et al., 2014). 42 As in most other topic models, the structural topic model requires the analyst to specify ex ante the number of topics to be estimated. I estimated a variety of models with 5-20 topics and selected the 10-topic model because it scored highest on the key dimensions of exclusivity and semantic coherence (see Roberts et al. (2014). Author IO and year are included as covariates in the model to allow the distribution of topics to vary by organization and year. 21

22 Figure 3 shows how the distribution of topics can provide evidence of a division of labor. It compares the OSCE s regulatory efforts in 2008 with two other IOs, the EU and the SCO. There are several instances of deference between the OSCE and EU in the years preceding 2008 and none between the OSCE and SCO. We should therefore expect the OSCE and EU to divide labor by focusing their rule-making efforts on separate issues, while the OSCE and SCO will have more overlap in their regulatory focus. The figure confirms this expectation. The distribution of topics in the policy documents of the OSCE and SCO are very similar (left panel). Contrast this pair with the OSCE and the EU (right panel), which have taken a more complementary approach to governance of counterterrorism: the OSCE tends to emphasize human rights, while the EU focuses on transportation and nuclear security. Topic Distribution Topics Criminal Justice Terrorism Finance Transport & Nuclear Human Rights Other OSCE SCO OSCE EU Figure 3: Division of Labor with Organization for Security and Cooperation in Europe (OSCE), 2008 : The figure displays the distribution of rule-making effort among counterterrorism sub-issues (as measured by the structural topic model) for the OSCE in 2008, compared to the Shanghai Cooperation Organization (SCO) and the European Union (EU) in the same year. 22

23 To quantify the division of labor between all pairs of IOs, I construct an annual measure of the total difference in regulatory focus between each IO dyad. This measure calculates the total absolute difference between the two topic distributions, summed across each topic. It increases as two IGOs prioritize different sub-issues, and decreases as they devote regulatory effort to the same sub-issues. In the figure above, the OSCE and SCO have a division of labor score of.57, while the OSCE and EU score almost twice as high (1.05). The measure is bounded between 0 and 2, and the sample average is Data in the counterterrorism regime complex are consistent with the use of deference as a coordinating tool. IO dyad-years in which two IOs engage in institutional deference have, on average, a division of labor score of 1.19; IO pairs that do not defer to each other average only The difference is statistically significant (p-value < 0.01), suggesting IOs that coordinate via institutional deference devote their governance efforts to different sub-issues. More important than levels of division of labor is how two IOs change over time once deference occurs. Table 2 displays the results of a linear regression model estimating the effect of institutional deference on the change in division of labor between two IOs. The unit of observation is the directed IO dyad-year. Deference, the independent variable, is measured as the proportion of IO A s policy documents that contain an instance of institutional deference to IO B. I control for several potential confounders that may influence both the use of deference and the division of labor between organizations: whether the IO pair includes an institution with binding legal authority, the number of overlapping member states, and 23

24 similarity in the foreign policy preferences of member states, 43 I also control for state power asymmetries among IOs, since Gehring and Faude (2014) argue a balanced distribution of power should generate greater division of labor. 44 The construction of the dependent variable the change in division of labor from the previous year also mitigates confounding variables specific to each pair of IOs. Standard errors are clustered by directed IO dyad. Dependent variable: Division of Labor (1) (2) (3) Counterterrorism Counterterrorism All Issue Areas Deference (0.022) (0.030) (0.028) Binding IO (0.012) (0.060) (0.135) Power Asymmetry (0.004) (0.034) (0.026) UN Ideal Point Similarity (0.018) (0.240) (0.126) Membership Overlap (0.0003) (0.017) (0.001) IO Dyad Fixed Effects Observations 1,177 1, Table 2: Effect of Deference on Division of Labor. Results of linear models estimating the effect of deference on division of regulatory labor. Coefficient estimates are displayed with standard errors in parentheses. p<0.1; p<0.05; p< I measure foreign policy preferences using the ideal point estimates provided by Voeten and Merdzanovic (2009), which are constructed using UN voting records. The variable represents the average ideal point distance between members of IO A and IO B, multiplied by This variable is constructed by taking the absolute difference in the number of great powers (as recognized by the Correlates of War dataset) in IO A and IO B. 24