Orchestrating Global Governance: From Empirical Findings to Theoretical Implications

Transcription

1 Orchestrating Global Governance: From Empirical Findings to Theoretical Implications Kenneth W. Abbott, Arizona State University Philipp Genschel, Jacobs University Bremen Duncan Snidal, University of Oxford Bernhard Zangl, Ludwig-Maximilians-University Munich Abstract This chapter is the conclusion to an edited volume on orchestration, a mode of governance widely used by international organizations (IGOs) and other governance actors, but rarely identified or analyzed. IGOs engage in orchestration when they enlist public or private intermediary actors on a voluntary basis, by providing them ideational and material support, to address target actors in pursuit of IGO governance goals. Orchestration is thus both indirect and soft. In the framework paper for the volume, we develop six hypotheses as to the conditions under which governance actors, and IGOs specifically, are likely to rely on orchestration. Twelve chapters then explore IGO orchestration in diverse issue areas, providing empirical evidence on our hypotheses. Here we assess that evidence. Our hypotheses provide valuable explanations of many cases of orchestration (and non-orchestration). But the cases suggest important modifications. Most significantly, we find that states often approve of or initiate IGO orchestration, even when they are its targets. Where states rather than IGOs initiate orchestration, certain factors we hypothesized would foster IGO orchestration notably goal divergence among member states actually discourage it.

2 Introduction We started this project with a puzzle and a hunch. The puzzle was whether and how international intergovernmental organizations (IGOs) can have an impact on international governance even though they lack the authority and resources normally seen as necessary for successful governance. The hunch was that this could be understood by investigating IGOs use of soft governance techniques to mobilize third parties to assist in achieving IGO goals. In the introductory chapter of this volume, Orchestration: Global Governance through Intermediaries, we theorize our initial hunch, both in general terms and for the special circumstances of IOs. We introduce the concept of orchestration, a mode of governance that is soft and indirect; orchestration thus stands in contrast to modes of governance that are direct and/or hard, including hierarchy, collaboration and delegation. In orchestration, one actor, the Orchestrator, enlists the voluntary assistance of a second actor, the Intermediary, to govern a third actor, the Target, in line with the orchestrator s goals. We refer to this as the O-I-T model of orchestration. We focus in this volume on the special case in which the orchestrator is an IGO. The introductory chapter introduces six hypotheses regarding the conditions under which governance actors, and IGOs specifically, are likely to engage in orchestration. To evaluate our theory, we invited leading and emerging scholars of international governance working across a wide range of issue areas to explore whether and how IGOs employ orchestration, and to provide empirical evidence on our theoretical hypotheses. In this concluding chapter we bring together and evaluate their empirical findings.

3 Our theory contrasts orchestration to hard modes of governance, including the traditional view of domestic government as hierarchical and direct, as well as the contemporary view of international governance as hard but indirect, based on delegation from states to IGOs (Hawkins et al. 2006). While delegation provides a valuable perspective, the empirical chapters demonstrate that IGO orchestration captures an even wider range of activity. Indeed, orchestration matters across many more and more diverse issue areas and situations than we originally anticipated. The empirical chapters also testify to the explanatory power of orchestration theory. The six hypotheses developed in the introductory chapter provide valuable explanations of many cases of orchestration (and non-orchestration). However, the empirical chapters also suggest important modifications and extensions: certain hypotheses operate in more complex and nuanced ways than we expected. Perhaps most significantly, the case studies reveal that orchestration is not only a top-down technique initiated by an IGO or other orchestrator; it often operates bottomup, with intermediaries and even targets actively seeking orchestration. We were particularly surprised by the extent to which states initiate IGO orchestration even when they themselves are its ultimate targets. Contrary to our expectation that states would generally oppose orchestration out of concern for strengthening IGO autonomy, states frequently prefer it to alternative governance modes (and to no governance at all). We argue that states embrace and even initiate orchestration when it allows them to achieve a satisfactory level of governance at low sovereignty costs (compared to delegation) by granting only limited authority to IOs. This has significant implications for our explanatory theory: where states rather than IGOs initiate orchestration, some

4 factors that we hypothesized as discouraging orchestration actually foster it. In particular, while disagreement ( goal divergence ) among member states facilitates IGOinitiated orchestration (as we expected), it hampers state-initiated orchestration. The empirical chapters further suggest that IGO orchestration involves more varied intermediaries than expected. IGOs not only enlist non-state intermediaries such as NGOs, business organizations, public-private partnerships and transgovernmental networks, but to a surprising extent also orchestrate other IOs. Orchestration thus proves to be a significant technique for managing regime complexes, a primary motive for stateinitiated IGO Orchestration. In this chapter we summarize this volume s empirical findings on the extent and patterns of IGO orchestration (section 1); assess the explanatory power of our hypotheses in light of the case study evidence (section 2); and discuss the implications of orchestration for understanding global governance and governance more broadly (section 3). 1 Mapping IGO Orchestration: Extent, Patterns and Variations Orchestration is a new analytical concept. To assess its relevance, we asked the authors of the empirical chapters to identify cases of IGO orchestration (or non-orchestration) and to map the patterns of orchestration in their areas of expertise. How prevalent is orchestration? How important is it? How did it evolve? What are its dominant features? What actors are involved? The empirical chapters cover a wide spectrum of policy fields, including the environment, health, human rights, trade, security, atomic energy, social standards,

5 economic regulation and finance. This provides substantial variation for assessing the relevance of IGO orchestration across issue areas. Variation is further enhanced because many chapters compare cases of IGO orchestration with similar cases in which IGOs utilize other governance modes. Taken together, the chapters consider a total of 38 separate cases of IGO governance, of which 27 involve IGO orchestration; 11 are nonorchestration cases. Table 1 summarizes the main features of the 27 orchestration cases. 1 TABLE 1 HERE Orchestration is a valuable analytical concept First and foremost, the empirical chapters demonstrate the analytic value of the concept of orchestration. The concept proved to be sufficiently coherent and precise that the authors could readily distinguish orchestration from other modes of governance, such as hierarchy, collaboration and delegation (Gerring 2001). Moreover, the orchestration concept enabled case study experts to more precisely analyze governance activities of which they had been aware, but for which they lacked a theoretical framework. Finally, while there remains some ambiguity at the margins (e.g., where orchestration shades into delegation, as in Graham and Thompson, in this volume), most cases fall naturally either within the category of orchestration or outside it. Combined with the ubiquity, importance and variety of IGO orchestration, this makes us confident that the O-I-T model captures a distinctive and important mode of governance. The remainder of this section discusses the broad implications of our descriptive evidence on orchestration while the next section explores our causal hypotheses regarding orchestration.

6 IGO Orchestration is widespread and significant The cases summarized in Table 1 demonstrate that IGO orchestration is widespread, varied and significant. Each empirical chapter reports at least one case of IGO orchestration, and most report more. To be sure, this high incidence partly reflects selection bias: because we were concerned with establishing the significance of orchestration as a governance mode, we explicitly asked our authors to look for cases of orchestration. Yet it is nevertheless striking that orchestration appears across such a broad range of policy fields. In addition, many cases deal with high-stakes issues such as enforcement of Security Council sanctions (Haufler, in this volume), implementation of EU competition policy (Blauberger and Rittberger, in this volume) and allocation of international assistance (Graham and Thompson, in this volume). In short, the chapters provide strong evidence that IGO orchestration is both ubiquitous and significant. IGO Orchestration is long-standing Although we initially believed orchestration to be a relatively new phenomenon, the chapters show it to be a long-standing governance technique. As Tine Hanrieder s chapter (in this volume) demonstrates, for example, WHO has been orchestrating since its inception after World War II. An even more striking example is the ILO. Founded after World War I with a corporatist structure involving business, labor and governments, the ILO was implicitly designed as a major site of orchestration. The irony, as Lucio Baccaro points out, is that this very structure now leaves the ILO politically unable to orchestrate other non-state intermediaries, such as NGOs and private-public partnerships, because doing so might threaten the privileged position of trade unions and employers associations (Baccaro, in this volume). Rather than being in the vanguard of orchestration, then, the ILO turns out to be a striking laggard.

7 IGO orchestrators vary widely Our initial expectation was that the primary users of orchestration would be those IGOs with limited regulatory authority or operational capacity. The cases reveal, however, that a much wider variety of IGOs employ orchestration, as shown in the Orchestrator column of Table 1. Orchestrators include not only weak IGOs such as the FATF (Findley et al., in this volume) and CITES, but also strong organizations including the EU Commission and Security Council. Some orchestrators are formal and highly institutionalized, e.g., the WTO and ILO, but others are informal, e.g., the G20, or weakly institutionalized, e.g., the UN Global Compact. In short, orchestration is employed by IGOs of widely differing character. 1 IGO Orchestrators pursue a wide range of governance goals IGOs employ orchestration for varied purposes, as shown in the Functions column of Table 1. First, IGOs use orchestration for transnational rule-making and standard-setting. The EU Commission enlisted IASB to draft accounting standards for European business (Mattli and Seddon, in this volume), while UNEP and UNGC established PRI to develop standards for socially responsible investment (Van der Lugt and Dingwerth, in this volume). Second, IGOs use orchestration for rule implementation. The Office of the High Commissioner on Human Rights (OHCHR) actively promotes NGO participation in monitoring state compliance with human rights commitments, while the CITES secretariat collaborates with non-state actors to police international rules (Dai, in this volume; Tallberg, in this volume) against trade in endangered species. Finally, IGOs use orchestration to provide public goods. GEF orchestrates climate adaptation financing through implementing agencies including UNDP, UNEP and the 1 Because of our focus on IGOs, we do not include one case discussed in an empirical chapter in which the orchestrator is a nongovernmental organization (the IFAC in Mattli and Seddon, in this volume). As this case suggests, because orchestration is a general form of governance, a diverse range of orchestrators, including states and private actors, operates in other settings.

8 World Bank (Graham and Thompson, in this volume), and WHO relies on orchestration to control epidemic diseases (Hanrieder, in this volume). Importantly, orchestration serves not only the policy purposes of IGOs, but also their institutional interests. For example, IGOs employ orchestration to expand into new areas. UNEP used orchestration to gain a foothold in the field of corporate social responsibility (Van der Lugt and Dingwerth, in this volume), while the EU Commission employed it to expand its environmental activities (Mattli and Seddon, in this volume). One possible explanation for this phenomenon is that orchestration allows IGOs to test the waters and gain prominence in new areas without major institutional investments. Orchestration also helps IGOs develop expertise and legitimacy, thus enhancing their autonomy from state oversight. For instance, WHO bypassed member state vetoes on disease surveillance by orchestrating internet-based sources of epidemiological information (Hanrieder, in this volume). In sum, as theorized in the introductory chapter, IGOs may use orchestration as an instrument of agency slack, but may also use it to discipline principal slack when states renege on commitments. IGO Orchestration engages diverse intermediaries IGO orchestrators work with a broad range of intermediary organizations, as shown in the Intermediary column of Table 1. Unsurprisingly, NGOs are among the most common intermediaries. For example, the field offices of UN OHCHR collaborate with local human rights NGOs to monitor state behavior (Tallberg, in this volume). IGOs also engage transnational partnerships as WHO does with GAVI (Hanrieder, in this volume) and transgovernmental networks, as the European Commission does with the European Competition Network (ECN) (Blauberger and Rittberger, in this volume).

9 More surprisingly, IGOs regularly enlist other IGOs as intermediaries. Our expectation was that IGOs would use orchestration to carry out their own mandates, but in fact this often involves orchestrating how other IGOs carry out theirs. The most prominent example is the G20, which was established not to replace the many IGOs addressing global economic policy but to orchestrate them (Viola, in this volume). GEF administers its adaptation funding almost exclusively through IGOs possessing complementary areas of expertise, including UNDP, UNEP, the World Bank and the African Development Bank (AfDP) (Graham and Thompson, in this volume). More generally, while regime complexes are typically viewed as structures of horizontal coordination and contestation among overlapping regimes (Raustiala and Victor 2004), IGO orchestration often adds an element of soft vertical steering. Importantly, IGOs not only enlist existing intermediaries, they often create new organizations, or re-shape existing ones, in order to orchestrate them. For example, GEF deliberately attempted to re-shape AfDP s policy priorities to align more closely with the requirements of adaptation finance (Graham and Thompson, in this volume); the European Commission created ECN in order to orchestrate it. Other purpose-built intermediaries include PRI (Van der Lugt and Dingwerth, in this volume) and the Kimberley Process Certification System (KPCS) (Haufler, in this volume). Intermediaries often welcome Orchestration Because the goals of orchestrator and intermediary are highly correlated, both benefit from orchestration. The benefits to intermediaries extend not only to their substantive goals, but also to their organizational interests. Walter Mattli and Jack Seddon (in this volume) report, for example, that orchestration by the EU Commission conferred much-needed legitimacy upon the IASB, boosting its prestige as a global regulator and triggering a sharp increase in the demand of

10 its rules and services by governments around the world. The Codex Alimentarius Commission s rise from obscurity similarly began with its endorsement by WTO (Elsig, in this volume). In short, orchestration may lead to mutual empowerment (Tallberg, in this volume) of an IGO and its intermediaries. For that reason, potential intermediaries often welcome orchestration, and even actively seek to be orchestrated. For example, as Manfred Elsig explains, the ISO persistently lobbied the WTO to endorse it (as it had the Codex Alimentarius), in order to improve its competitive position vis-à-vis other international standards organizations. The WTO rebuffed ISO because some member states feared an explicit endorsement would make ISO even more dominant in the market for product standards. IGO Orchestration addresses varied targets We expected IGOs to orchestrate for two broad purposes: to bypass states (by directly regulating or providing collective goods to diverse private actors) and to manage states (by creating and implementing rules for state behavior). The empirical chapters bear out these expectations, as shown in the Targets column of Table 1 and in the organization of the chapters of this book. Managing states orchestration addresses both wealthy and developing states as targets in variety of settings; we discuss this further below. Even more notable is the diversity of private targets addressed through bypassing states orchestration. In the empirical chapters, private targets range from incorporation services and law firms in the FATF case (Findley et al., in this volume), to financial asset owners, managers and service providers in the PRI case (Van der Lugt and Dingwerth, in this volume), and to patients in international vaccination campaigns (Hanrieder, in this volume).

11 States often welcome and even initiate IGO Orchestration Perhaps our most surprising empirical finding is that states often welcome and even initiate IGO orchestration, even when they are the intended targets that is, in managing states orchestration. As Jonas Tallberg (in this volume) observes states actually may prefer, rather than fear, being managed by relatively autonomous IOs. Our original expectation was that IGOs would generally initiate orchestration against state interests in order to enhance their own autonomy and resources; states would generally oppose it because it undermines their control. As the Initiator column of Table 1 indicates, some cases do fit that description: for example, WHO used orchestration to avoid state restrictions in disease surveillance. With the G20 and GEF, however, states established relatively weak IGOs and instructed them to orchestrate; this allowed states to pursue their governance goals (perhaps imperfectly) while delegating only limited authority and incurring minimal sovereignty and material costs. GEF, for example, has only some 100 staff, of whom six work on adaptation funds; the G20 has no permanent staff whatsoever. In short, while orchestration may increase the effectiveness of IOs, helping states resolve cooperation problems, it also entails relatively low sovereignty costs for states, allowing them to keep IGOs underpowered. We also failed to fully anticipate the ability of states to structure orchestration so as to maintain control over IOs. In particular, the empirical chapters show how states can curtail IGO autonomy by limiting permissible intermediaries: either overriding IGO links with preferred intermediaries or specifying acceptable intermediaries in advance. In the WTO technical standards case noted above, for example, member states blocked WTO endorsement of ISO, preferring endorsement of a broader list of international standards organizations (Elsig, in his volume); in the EU telecommunications case, member states created and shaped BEREC as an

12 intermediary in ways that restricted the Commission s options (Blauberger and Rittberger, in this volume). IGO Orchestration involves a range of techniques The five general orchestration techniques highlighted in the introductory chapter (convening, agenda setting, assistance, endorsement and coordination) are widely used in the cases, often in combination, as shown in the Techniques column of Table 1. Yet each technique is most often employed for specific purposes. Convening is used to create purpose-built intermediaries, as in the establishment of PRI by UNEP and the UNGC (Van der Lugt and Dingwerth, in this volume) and the creation of ECN by the European Commission (Blauberger and Rittberger, in this volume). Assistance plays a special role in supporting rule implementation. For example, international human rights bodies finance training for local NGOs and provide information to prospective litigants on how to file complaints. The weaker and more fragile the intermediary, the more important is material assistance. Endorsement is particularly important for rule-making intermediaries. Examples include General Assembly resolutions supporting the KPCS (Haufler, in this volume), WTO endorsement of international food standards (Elsig, in this volume), and FATF endorsement of the private certification body ACAMS (Findley et al., in this volume). Finally, Agenda Setting and Coordination, while used in many contexts, are especially important for orchestrating regime complexes, as in G20 coordination of international financial institutions (Viola, in this volume). IGO Orchestration is dynamic While the introductory chapter presents orchestration in a comparative statics framework, the empirical chapters trace how it unfolds dynamically. The cases show how orchestration interacts over time with other governance modes. In some cases, other modes prepare the ground for

13 orchestration. For example, deficiencies in hard, direct governance may trigger a turn to orchestration. As Michael Blauberger and Berthold Rittberger (in this volume) explain, the unwieldiness of the EU s hierarchical competition regulation system led to its replacement by an orchestrated system, with the Commission working through the ECN. Xinyuan Dai (in this volume) highlights cases in which deficiencies of hierarchical implementation led to the orchestration of complementary monitoring by NGOs. In other cases, orchestration prepares the ground for alternative governance arrangements. Successful orchestration may increase the orchestrator s expertise, resources and legitimacy to the point where it no longer needs intermediaries. The infectious diseases case in Tine Hanrieder s chapter (in this volume) shows how its success in orchestrating a response to SARS empowered WHO. Yet even successful orchestration can have contrary effects. As Erin Graham and Alex Thompson (in this volume) explain, GEF s success in orchestrating adaptation funding attracted new competitors, disrupting intermediary relationships and eroding orchestration. Finally, Virginia Haufler (in this volume) shows how orchestration can trigger conflict among the parties that constitute an intermediary, undermining its ability to continue in that role. 2 Explaining IGO Orchestration: Assessing the Hypotheses Under what conditions do IGOs engage in orchestration? To address this question, we asked our authors to evaluate the six hypotheses developed in the introductory chapter in light of their case study evidence. Overall the hypotheses fare well: they provide good baseline explanations for IGO governance choices. But the case studies also reveal unanticipated causal complexities. Most importantly, they highlight a significant distinction between IGO-initiated and stateinitiated orchestration. Our hypotheses implicitly assume IGO-initiated orchestration, and most

14 hypotheses work well for such cases. Many hypotheses also work for state-initiated orchestration, but some most importantly the goal divergence hypothesis do not. We highlight those differences below. The empirical chapters also reveal some interaction effects between hypotheses. That is not surprising, as we initially limited our propositions largely to simple bivariate relations. We did this not because we believe the world is so simple, but because of the limits to what one can theorize ex nihilo, with limited empirical input. We hoped that investigating simple hypotheses would guide the identification of more nuanced interrelations within the cases. And in fact the case studies demonstrate clearly that some hypothesized causal relations are intertwined. We examine specific interaction effects below. Before reviewing the empirical findings in detail, we briefly address the issue of selection bias. Qualitative research almost always involves some form of selection bias. In this project, it lies not so much with the choice of policy fields the chapters cover a broad range as with the selection of cases within those fields, because we instructed authors to seek out cases of IGO orchestration. While some chapters do compare cases of IGO orchestration with similar cases in which IGOs utilize other governance modes, orchestration cases are still likely to be oversampled. Thus, 27 of the 38 cases in our sample involve IGO orchestration, whereas in only 11 do IGOs instead engage in governance through hierarchy, collaboration or delegation. For projects like this one, however, the costs of avoiding this form of selection bias outweigh the benefits (Collier and Mahoney 1996). In new research areas, where hypotheses are unsettled and often underspecified, close attention to context is required to distinguish between disconfirming evidence and evidence of alternative causal relationships. Here, for example, an unbiased but shallow analysis of the goal divergence hypothesis would have produced numerous

15 false negatives. The in-depth analysis in the empirical chapters, by contrast, reveals that IGOinitiated and state-initiated orchestration follow different causal logics, only one of which is captured by the hypothesis. Hence, what we lose in representativeness we more than gain in attentiveness to context. Importantly, the chapters still offer considerable variance on the dependent variable. They deal with cases where IGOs decided against orchestration, as in NAFTA s decision not to orchestrate compliance constituencies along EU lines (Tallberg, in this volume); where IGOs tried to orchestrate but failed, as in the ILO s proposed social label convention (Baccaro, in this volume); where IGOs orchestrate but with a very light touch, as in FATF s endorsement of private certification bodies (Findley et al., in this volume); where IGOs use orchestration as one governance instrument among others, as in the EU (Blauberger and Rittberger, in this volume; Mattli and Seddon, in this volume); and where IGOs do little else but orchestrate, as in the case of the G20 (Viola, in this volume). In the remainder of this section, we review the evidence on each hypothesis in turn. Table 2 summarizes the values of the variables for all hypotheses and all cases. 1 TABLE 2 HERE A. Orchestrator capabilities hypothesis We hypothesized that: governance actors are more likely to orchestrate when they lack certain capabilities needed to achieve their goals through other governance modes. The empirical chapters offer broad support for the capabilities hypothesis, showing that IGOs orchestrate to

16 harness capabilities they do not possess in sufficient quality and/or quantity on their own. Specifically, in the 34 cases we consider in this section, 2 IGOs lacking significant governance capabilities employed orchestration in 25 out of 30 possible cases (83%), while those judged to have adequate capabilities did not orchestrate in any of 4 possible cases, where they were able to go it alone. Thus the capabilities hypothesis points to what is probably the single most important determinant of IGO orchestration. 3 For example, WHO has considerable regulatory authority and legitimacy, but lacks operational capabilities for research, disease surveillance and public health assistance; as a result, it relies on intermediaries including UNDP, UNICEF and ProMED to provide such capabilities (Hanrieder, in this volume). After several futile attempts to harmonize European accounting rules, the European Commission turned to IASB to provide sound rules, because the Commission lacked the expertise IASB possessed (Mattli and Seddon, in this volume). UNEP compensated for its lack of legitimacy in socially responsible investment by orchestrating creation of PRI (Van der Lugt and Dingwerth, in this volume). In all these settings, IGO decisions to orchestrate were driven by the need for complementary capabilities. The empirical chapters also reveal certain more nuanced relationships between orchestrator capabilities and orchestration. First, capabilities must be evaluated not in absolute terms, but relative to the specific problem at hand and to the IGO s goals. Even notionally powerful IGOs such as the EU or Security Council may lack sufficient capabilities to tackle tricky problems and achieve demanding goals such as the surveillance of market imperfections across Europe 2 As noted in Table 2, note 4, we do not include the four cases discussed in Dai (in this volume) as part of our numerical analysis. 3 These figures include both IGO-initiated and State-initiated orchestration, a distinction that becomes important below. For the capabilities hypothesis, however, the identity of the initiator is not central, as in either case the choice to orchestrate is driven by the need to supplement IGO capabilities. The same point applies to other hypotheses, below except where initiation is explicitly discussed.

17 (Blauberger and Rittberger, in this volume) or the implementation of targeted sanctions in Africa (Haufler, in this volume). Second, orchestration allows IGOs to expand their authority beyond the limits of their initial capability endowments. It can be an instrument of IGO self-empowerment, attractive to organizations seeking greater influence and autonomy. By successfully orchestrating, IGOs can enhance their expertise, legitimacy and prominence, perhaps enough to pursue more demanding governance modes, or to expand into areas they were not previously capable of addressing, as UNEP did with corporate social responsibility (Van der Lugt and Dingwerth, in this volume). Third, an IGO s capability deficit is often not inadvertent but a consequence of deliberate design by member states. As noted above, states are generally reluctant to delegate the capabilities necessary for hierarchical governance, as this could make IGOs too powerful and independent. Instead, states can (partially) achieve their goals at lower sovereignty costs by limiting IGO capabilities in ways that force (yet still enable) IGOs to rely on third parties. In the European telecommunication sector (Blauberger and Rittberger, in this volume), for example, the Commission sought authority for hard, direct regulation. Because member states (and for different reasons the European Parliament) would not agree to grant such powers, the Commission was instead left to pursue an orchestrated solution. Lora Viola (in this volume) makes a similar case regarding the G20: confronted with a financial crisis but wary of creating a powerful new organization, major states constituted the G20 with just enough power to orchestrate existing financial IOs, but no power whatsoever to go it alone. Thus, while orchestration offers a means of IGO self-empowerment, it also allows states to limit IGO authority, thereby encouraging states to allow or even initiate IGO orchestration.

18 B. Intermediary availability hypothesis We hypothesized that: governance actors are more likely to orchestrate when intermediaries with correlated goals and complementary capabilities are available. The case studies provide overall support for the availability hypothesis: IGOs often orchestrate by working through intermediaries with correlated goals and complementary capabilities. To be sure, in some cases 7 out of 30 possible cases IGOs did not orchestrate despite the availability of appropriate intermediaries. Yet in most cases of intermediary availability 23 out of 30 (77%) IGOs did orchestrate. Where appropriate intermediaries were not immediately available, in contrast, orchestration occurred in only 2 of 4 cases (50%). For example, CITES can rely heavily on orchestration for monitoring because an intermediary with the interest and ability to detect non-compliance is readily available: TRAFFIC, the joint wildlife trade monitoring program of the World Wide Fund for Nature (WWF) and the World Conservation Union (IUCN). The ozone regime, by contrast, is less able to orchestrate compliance monitoring for lack of technically capable intermediaries able to measure national ozone emissions (Tallberg, in this volume; Dai, in this volume). How do we explain the two cases in which orchestration occurred with no appropriate intermediary available (WHO health aid I and WHO infectious disease I)? Hanrieder s analysis (in this volume) shows the limits of Table 2 s dichotomous coding of intermediary availability. Although the field of intermediaries was thin, WHO s focality allowed it to support the creation of networks such as the Global Public Health Intelligence Network (GPHIN) to bypass states in releasing timely health reports. If we had employed a finer coding of the variable, other cases of intermediary availability may well have been coded instead as ability to create new intermediaries.

19 More generally, lack of preexisting intermediaries need not preclude orchestration: IGOs can often facilitate the emergence of new intermediaries tailored to their needs, as the WHO example above illustrates. Other examples include the European Commission creating ECN (Blauberger and Rittberger, in this volume), the Security Council facilitating establishment of KPCS (Haufler, in this volume), and UNEP and the UN Global Compact jointly sponsoring PRI (Van der Lugt and Dingwerth, in this volume). While the introductory chapter foresaw this possibility, the case studies highlight its significance. Diverse factors affect the feasibility of intermediary creation. In some policy areas, the absence of intermediaries is a binding constraint. For instance, as noted above, it is difficult to create non-state intermediaries to monitor ozone compliance because specific technical capabilities are needed; it is equally difficult in arms control regimes because security concerns lead states to refuse access for non-state monitors. In other areas public sentiment matters: emerging public concern for corporate social responsibility facilitated creation of PRI by creating an incentive for institutional investors to participate (Van der Lugt and Dingwerth, in this volume). Existing institutional structures also matter: the longstanding informal network of European competition experts facilitated establishment of ECN (Blauberger and Rittberger, in this volume). Thus the availability hypothesis should be broadened beyond the existence of intermediaries to address the conditions that facilitate or hamper creation of new intermediaries by IGOs. Finally, the number of available intermediaries can have a strong impact on the content and distributive consequences of orchestration. An orchestrator able to choose among multiple (possibly competing) intermediaries will find orchestration more attractive, because it can better shape the outcome to its own preferences, at lower cost; with fewer potential intermediaries, the

20 orchestrator has less choice and may have to offer the chosen intermediary a larger voice in the outcome. Walter Mattli and Jack Seddon (in this volume) illustrate these factors by contrasting the EU Commission s orchestration of European regulation in two fields. In accountancy, the Commission depended on a single, well-entrenched intermediary, IASB; it consequently had few means to influence the relationship. In environmental regulation, in contrast, the Commission could forum-shop in a rapidly growing pool of potential intermediaries, putting it more firmly in the driver s seat. Despite this, potential intermediaries will often find orchestration attractive even when they are numerous. First, their individual incentives are to be chosen, so as to better attain their organizational and substantive goals. Second, intermediaries with overlapping missions may see their effectiveness undermined by a lack of coordination, and so welcome IGO orchestration that creates greater order. Lora Viola (in this volume) argues, for example, that the proliferation of IGOs in financial regulation created an urgent demand among them for orchestration, leading states to create the G20. Regime complexes feature an abundance of actors that would benefit from coordination. C. Orchestrator Focality Hypothesis We hypothesized that: governance actors are more likely to orchestrate when they are focal within the relevant issue area. The case study evidence does not support the focality hypothesis. While IGO focality is associated with orchestration in 21 out of 30 cases (70%), the 4 cases with non-focal IGOs all involved orchestration as well (100%). On that evidence, IGO focality does not significantly increase the probability of orchestration but actually appears to impede it. However, this result should not be generalized without further study: even if focality does not have a strong impact, it is difficult to see why it should impede orchestration. Here, looking

21 at the underlying processes may be more instructive than the correlational data. In several case studies, recognition of an IGO as the natural leader of a policy field facilitates orchestration, especially by increasing the orchestrator s authority and attractiveness for potential intermediaries. For example, WHO successfully orchestrates in areas such as disease surveillance where it is focal partly through its ability to mobilize new intermediaries but is less successful in areas such as health assistance (Hanrieder, in this volume) where it competes for primacy with the World Bank and other IGOs. Similarly, GEF s focality for adaptation finance initially facilitated orchestration, while the recent erosion of its focality is undermining its orchestration (Graham and Thompson, in this volume). One explanation for orchestration by non-focal IGOs relates to dynamics: IGOs with limited focality may enhance their focality through orchestration. While orchestration by such organizations may initially be costly or have limited impact, modest success can increase their prominence and legitimacy, reducing the cost and increasing the impact of future efforts. For example, WHO s success in orchestrating responses to SARS increased its focality and authority for disease surveillance. In contrast, the perceived failure of its response to AIDS fatally undermined its natural leadership on that disease. Even successful orchestration can undermine focality: GEF s orchestration of adaptation funding drew powerful competitors, including the World Bank and UNDP, which increasingly address climate issues, eroding GEF s focality (Graham and Thompson, in this volume). An explanation for non-orchestration by focal IGOs is that focality often accompanies other significant capabilities, making orchestration unnecessary. For example, the European Court of Human Rights is so visible, prominent and competent that it receives all the private complaints it can handle; it has no reason to orchestrate compliance constituencies. UN human

22 rights bodies, in contrast, have much lower visibility, and therefore orchestrate NGOs and potential complainants to encourage complaints (Tallberg, in this volume). The UN bodies orchestrate precisely because they lack focality, while the European Court does not orchestrate because it is already focal. Dynamically, moreover, successful orchestration might enhance an IGO s capabilities to the point where it can pursue stronger modes of governance. Focality is also relevant to the distributive impacts of orchestration, although its impact depends on oversight arrangements. On the one hand, focality strengthens IGOs not only vis-àvis their intermediaries, as discussed above, but also vis-à-vis their member states. Focal IGOs enjoy greater autonomy because states are more dependent on them than if they could forumshop among multiple IGOs (Mattli and Seddon, in this volume; Hanrieder, in this volume). On the other hand, focality entails visibility that may prompt state oversight, whereas lack of focality can bestow a degree of anonymity. UNEP, for example, could organize PRI under the radar of states because the environmental ministries it engages with were not sensitive to its involvement in investment regulation (Van der Lugt and Dingwerth, in this volume). A more prominent IGO might not have had similar latitude. We noted above that careful contextual analysis can reveal new causal relationships; the interactions of focality with dynamics, capabilities and oversight are prime examples. Based on these findings, the empirical chapters identify sufficiently significant and interesting impacts of focality that it should remain part of the orchestration research program despite the weak correlational support across the case studies. D. Goal Divergence Hypothesis We hypothesized that: IGOs are more likely to orchestrate when there is a divergence of goals among their member states and/or between member states and the IGO. However, the impact of

23 goal divergence is more nuanced than anticipated, due primarily to the distinction between IGOinitiated and state-initiated orchestration. Overall, in 16 out of 23 cases (70%) with divergent member state interests, IGOs employ orchestration; yet in 9 out of 11 cases (82%) with more homogeneous state interests, orchestration still takes place. 4 Under the hypothesis as formulated, then, goal divergence actually appears to decrease orchestration. In fact, however, as Table 3 shows, the hypothesis holds more strongly for IGO-initiated orchestration (15 of 17 cases or 88% entail high goal divergence), on which the hypothesis implicitly focused, than for state-initiated orchestration (1 of 8 cases or 13%). Thus, while goal divergence does not explain the absence or presence of orchestration very well, it is a powerful predictor of the initiator of orchestration. With high goal divergence we are likely to see IGOs initiate orchestration, whereas with low goal divergence we are likely to see states initiate orchestration. TABLE 3 HERE High goal divergence among states increases the likelihood of IGO-initiated orchestration in two ways. First, it hampers intergovernmental delegation of authority for hard, direct governance to IGOs ex ante, contributing to IGO capabilities deficits; this makes it more likely that IGOs will turn to intermediaries for help. Dai s analysis (in this volume) of IGO monitoring in four issue areas demonstrates this point well. 5 UNEP also provides a clear example: because diverging state preferences often preclude agreement on binding international environmental 4 Unsurprisingly, most cases of non-orchestration (7 of 9 or 78%) occur in situations of high goal divergence. 5 As noted above, Dai s cases are not included in the numerical analysis above.

24 law, UNEP relies on soft governance modes including orchestration to fill the governance gap (Van der Lugt and Dingwerth, in this volume). Divergent state interests also help explain why the G20 is limited to orchestration rather than more direct or harder forms of governance (Viola, in this volume). Second, high goal divergence hampers intergovernmental agreement to override IGOinitiated orchestration ex post. Once orchestration is underway, even powerful states may be unable to block it as long as there is sufficient support for the IGO s actions from other states. For instance, once WHO had successfully orchestrated a response to SARS, it was able to withstand strenuous objections to its role from both China and Canada. Indeed, member states actually strengthened WHO s orchestration capacity though the 2005 International Health Regulations (Hanrieder, in this volume). The effects of low goal divergence are more ambiguous. On the one hand, low goal divergence undermines the prospects of IGO-initiated orchestration, since states can readily block such actions. On the other hand, low goal divergence enhances the prospects of stateinitiated orchestration by facilitating intergovernmental agreement on an orchestration mandate for IGOs. Thus, broad intergovernmental agreement on the need for common SPS standards combined with a general consensus that the WTO was not the appropriate platform for developing such standards paved the way for WTO orchestration of the Codex Alimentarius Commission (Elsig, in this volume). To consider the effect of goal divergence fully, however, we must examine it in interaction with state oversight. E. State Oversight Hypothesis We hypothesized that: IGOs are more likely to engage in orchestration when their member states have weak institutional control mechanisms. The cases do not support the oversight hypothesis.

25 Where member states exercise loose control over the relevant activities, IGOs employ orchestration in 17 out of 22 cases (77%). Where states exercise tight control, IGOs orchestrate in 8 out of 12 cases (75%). On this evidence, then, state oversight has little impact on IGO orchestration. However, the true underlying patterns become dramatically clearer when we look at state oversight and goal divergence together, as shown in Table 4 below. While evidence for the goal divergence and state oversight hypotheses separately is mixed, together these hypotheses have strong explanatory power, indicating a clear interaction effect. The vast majority of IGOinitiated orchestration cases take place under conditions of both high goal divergence and loose oversight (15/17=88%). In these cases, goal divergence means that states would not grant IGOs authority for hard governance but loose oversight makes it difficult for states to limit IGOinitiated orchestration. In contrast IGO-initiated orchestration is infrequent when member states have low goal divergence (only 2 cases); there are no cases of orchestration when high goal divergence is combined with tight oversight. TABLE 4 HERE State-initiated orchestration is different. As Table 4 shows, seven cases of state-initiated orchestration involve low goal divergence with tight oversight. Here orchestration is easy: states can both agree on policy objectives and rein in IGOs that deviate from them. In such circumstances, states may well prefer orchestration to more intrusive forms of governance. For example, tight state oversight in the WTO eliminates any real possibility for IGO-initiated

26 orchestration; orchestration must be initiated by member states (Elsig, in this volume). As a consequence, orchestration occurs in precisely those areas where goal divergence is least pronounced, such as training and technical assistance; it does not appear in areas of conflict, such as fisheries subsidies. Similarly, G20 orchestration depends on goal alignment among member states; without it, even orchestration of powerful institutions such as the IMF and BIS would be impossible. However, state consensus is not strong enough to enable agreement on hard regulation (Viola, in this volume). The surprising case is the one involving GEF (Graham and Thompson, in this volume), in which states with high goal divergence initiate orchestration despite relatively loose oversight (in the upper-right cell of Table 4). Although we might normally dismiss a single case as an exception, we believe the GEF case deserves special attention, as it arguably works against our finding. 6 States designed GEF with insufficient capabilities, so that it would have to orchestrate other IGOs. The Conference of the Parties was intended to provide state oversight, but in practice GEF has become relatively autonomous. Yet states have failed to exercise direct control in large part because they are quite fond of GEF s policies (Graham and Thompson, in this volume). This may therefore be the exception that proves the rule: IGOs can exercise autonomy under state-initiated orchestration so long as they do what states want. One way to interpret these results is in terms of the difference between ex post and ex ante oversight. Where states exercise only ex post oversight, an IGO is free to initiate orchestration; 6 In addition, that case may have unusual weight, as we coded five episodes of GEF orchestration (each with a different IGO intermediary, but otherwise identical in terms of our variables) as one case to avoid artificially inflating our results. See Table 1, note 3. We checked whether including multiple subcases from other empirical chapters might have overly influenced the results shown in Table 4. Viola s G20 cases accounted for 3 of the 7 stateinitiated cases with tight oversight and low goal convergence; Hanrieder s WHO cases accounted for 4 of 16 IGOinitiated cases with loose oversight and high goal divergence; and Mattli and Seddon s EU cases accounted for 3 of 16 IGO-initiated cases with loose oversight and high goal divergence. While these clustered subcases are clearly influential, none of them drive the results.

27 goal divergence then gives it greater latitude, because states cannot readily overturn its actions. But where states exercise ex ante oversight, an IGO is very limited in its ability to initiate orchestration. If states have divergent goals, orchestration is unlikely; if they have convergent goals, an IGO may orchestrate in line with those goals, or states themselves may initiate orchestration as an alternative to greater delegation. Jonas Tallberg s comparison of the EU and the NAFTA trade cases (in this volume) nicely illustrates this interaction. In both organizations, the Commission or Secretariat monitors states compliance with trade commitments. Yet due to member state goal divergence, ex ante oversight in NAFTA prevented the Secretariat from mobilizing private actors to provide information about non-compliance. The Commission, by contrast, was able to foster private monitoring of member states non-compliance despite members goal divergence: many of its activities are subject only to ex post control. The Commission publicized its ability to utilize private non-compliance information and made the infringement procedure more accessible to potential complainants by means of a help desk and a simplified complaints form. What had already been a system of monitoring with extensive private participation effectively became complaint-based (Tallberg 2013: 257). Where member states exercise ex ante control, goal convergence is necessary for orchestration to be possible; here, however, states rather than IGOs typically initiate orchestration. Goal convergence would also facilitate hard, direct governance, but states often prefer orchestration because it entails lower sovereignty costs. In establishing the G20, for example, states regarded orchestration as a relatively effective but non-intrusive mode of governance (Viola, in this volume). Member states could have mandated the IMF to regulate financial institutions, but that would have significantly increased its authority. Instead, by