Measuring Transnational Public-Private Governance Initiatives in World Politics: A New Dataset

Transcription

1 Measuring Transnational Public-Private Governance Initiatives in World Politics: A New Dataset Oliver Westerwinter Department of Political Science, University of St. Gallen April 11, 2017 Word count: 10,176 (without online appendix) Abstract This paper summarizes the Transnational Public-Private Governance Initiatives in World Politics Data. The data contains detailed information on the scope, participants, and institutional design of 468 transnational public-private governance initiatives (TGIs) in which states and/or formal intergovernmental organizations cooperate with business and civil society actors to govern global problems. I describe the sample generation and discuss coding rules. I also map the development of TGIs at the initiative and state level of analysis, and show how the new data contributes to enhancing researchers understanding of transnational public-private governance. The paper concludes by identifying puzzles and avenues for future research on transnational public-private governance initiatives in world politics. Keywords: transnational public-private governance initiatives, institutional design, international institutions, international cooperation. This research is funded by the Swiss Network for International Studies. I would like to thank the excellent research assistance of Christian Andres, Ruslan Aybazov, Tino Good, Stefano Jud, Rosie Keller, Laura Leibundgut, Giulia Parini, Dominik Schneeberger, Johannes Schultz, and Keto Schumacher in preparing this paper. I also thank Ken Abbott, Liliana Andonova, Michael Barnett, Tom Biersteker, Tom Hale, Virginia Haufler, Dirk Lehmkuhl, Miles Kahler, Erasmus Kersting, Christopher Kilby, Barb Koremenos, Lisa Martin, Katja Michaelowa, Christine Neuhold, Joost Pauwelyn, Jon Pevehouse, Kal Raustiala, Bernhard Reinsberg, Duncan Snidal, Jonas Tallberg, and Felicity Vabulas for helpful suggestions. I especially thank Sharlene Westerwinter for sharing her thoughts and for always supporting me. Müller-Friedbergstr. 8, 9000 St. Gallen, Switzerland, oliver.westerwinter@unisg.ch, web: oliverwesterwinter.com.

2 Introduction Transnational public-private governance initiatives (TGIs) are an important element of contemporary world politics. They are organizations in which states and/or formal intergovernmental organizations (FIGOs) cooperate with business and civil society actors to govern global problems. In contrast to traditional state-based regulation, in TGIs private actors are not only the objects of governing, but are involved at the center stages of the governance process including decisionmaking, monitoring, and enforcement (Reinicke and Deng, 2000; Börzel and Risse, 2005; Abbott and Snidal, 2009; Avant, Finnemore and Sell, 2010). These hybrid institutional arrangements are the most recent development in the emergence of private authority in global governance (Hall and Biersteker, 2002). 1 They demonstrate that the authority to govern in some issue areas of world politics is moving from national governments and FIGOs to a diversity of non-state actors (Abbott and Snidal, 2009; Lake, 2010; Avant, Finnemore and Sell, 2010; Büthe and Mattli, 2011; Green, 2014; Avant and Westerwinter, 2016). As such, TGIs challenge the traditional statist understanding of world politics which conceives of states as the only actors in the international system with decision-making authority (Waltz, 1979; Krasner, 1983; Keohane, 1984). They show that the state is far from the only game in town, and may no longer be the most important game in town (Abbott and Snidal, 2009, p. 87). 2 TGIs have been flourishing since the late-1990s. Today, they govern a broad range of global policy domains including environmental protection, human rights, health, and even the sensitive areas of trade, finance, and security (Börzel and Risse, 2005, pp ; Abbott and Snidal, 2009, p. 55). Some TGIs are among the most important governors in their issue area. In the security domain, the International Code of Conduct for Private Security Service Providers Association that brings together states, private security companies, and non-governmental organizations (NGOs) is 1 Early statements on the growing importance of non-state actors in world politics include Keohane and Nye (1971), Rosenau and Czempiel (1992), Risse-Kappen (1995), and Cutler, Haufler and Porter (1999). 2 This trend remains partial and uneven, however. Rather than a retreat of the state (Strange, 1996), what we observe is more accurately described as shifting patterns of authority which accumulate to a complex patchwork of global governance in which public-private and private-private institutions supplement and overlap with unilateral, bilateral, and multilateral arrangements among states (Abbott and Snidal, 2009; Avant, Finnemore and Sell, 2010; Abbott, Green and Keohane, 2016; Kahler, 2016). 2

3 the central, and to date, only effective institutional platform for regulating private security service providers at the global level (Avant, 2016). Similarly, the global governance of conflict diamonds critically depends on the Kimberley Process in which states, the diamond industry, and NGOs set, monitor, and enforce standards for the import and export of rough diamonds (Haufler, 2015). Finally, in the health domain, TGIs, such as the Global Alliance for Vaccines and Immunization and the Global Fund to Fight AIDS, Tuberculosis, and Malaria, are conerstones of the institutional architecture of global governance that dwarf the relevance and financial capabilities of the World Health Organization and other FIGOs (Hanrieder, 2015). 3 As these examples indicate, TGIs are a central part of global governance in nearly all issue areas of world politics. They often possess the authority to make and enforce decisions that change the behavior of governments, FIGOs, and multinational corporations, and affect the lives of millions of people. Thus, understanding the emergence, design, and effects of TGIs has become critical to fully understanding the complex institutional landscape and authority structures of contemporary global governance (Lake, 2010; Avant, Finnemore and Sell, 2010; Kahler, 2016; Barnett, Pevehouse and Raustiala, 2016; Krisch, 2017). The bulk of the literature on transnational public-private governance initiatives is based on single or small-n case studies (Reinicke and Deng, 2000; Dingwerth, 2007; Bartley, 2007; Vogel, 2009; Liese and Beisheim, 2011). 4 While large-n analyses and datasets are common in the study of traditional forms of global governance, such as intergovernmental organizations (Wallace and Singer, 1970; Jacobson, Reisinger and Mathers, 1986; Shanks, Jacobson and Kaplan, 1996; Pevehouse, Nordstrom and Warnke, 2004; Tallberg et al., 2013; Blake and Payton, 2015; Hooghe and Marks, 2015) and international agreements (Smith, 2000; Denemark and Hoffmann, 2008; Baccini, Dür and Elsig, 2015; Koremenos, 2013, 2016), 5 they are rare in research on transnational publicprivate governance. One of the first large-n studies of transnational public-private governance is Andonova and Levy (2003). Based on a dataset with 231 public-private partnerships that emerged 3 See also Institute for Health Metrics and Evaluation ( accessed: ). 4 For a more comprehensive although slightly outdated overview, see Schäferhoff, Campe and Kaan (2009). 5 See also the International Environmental Agreements Database Project led by Ronald Mitchell ( accessed: ). 3

4 after the 2002 UN World Summit on Sustainable Development, the authors examine variation in state participation in partnerships. Using the same data, Andonova (2014) analyzes the factors that underlie state participation more systematically. Similarly, Pattberg et al. (2012) examine 330 global sustainability partnerships and investigate the factors that shape their emergence, effects, and legitimacy. Focusing on the climate change area, Andonova, Hale and Roger (2016) use a dataset of 71 transnational governance initiatives that address problems related to climate change and analyze why states decide to participate in these initiatives. Finally, Abbott and Hale (2014) map patterns of empirical variation in terms of issue areas and types of actor involvement across 223 global solution networks. While these analyses constitute important steps in the systematic study of transnational publicprivate governance, they have difficulties capturing variation in governance initiatives across issue areas because they focus on a single or narrow set of policy domains. They also lack the information needed to examine differences in the institutional structures of transnational governance initiatives, such as monitoring, enforcement, and dispute settlement. This makes addressing questions regarding the emergence, development, and design of transnational governance initiatives from a comparative perspective challenging. In this paper, I present the Transnational Public-Private Governance Initiatives in World Politics Data. Expanding on previous data collection efforts, this dataset provides information on TGIs across issue areas. It includes data on the policy issues addressed by governance initiatives, state and non-state participation, as well as a broad range of institutional design elements. It extends existing datasets and is original in three ways: 1) the scope of the analysis; 2) the levels and units of analysis; and 3) the type of data included. A first innovative feature of the data is its broad scope. Evidence that spans issue areas and time periods is rare in research on transnational public-private governance initiatives. This makes the comparative analysis of the emergence and development of TGIs difficult. To address this challenge, we constructed a dataset with a broad representation of TGIs, covering organizations of varying size and institutional design in a multitude of issue areas. This allows researchers to go 4

5 beyond existing studies that focus on TGIs in a specific issue area and examine variation across TGIs in a broader context. It also provides the basis for testing hypotheses about the influence of policy domain on TGI creation and participation and reduces the selection bias in empirical analyses that can emerge from a narrow issue area focus. A second characteristic of the dataset is its compatibility with different levels and units of analysis. The main level of analysis, and the level at which the bulk of the data was collected, is the level of the governance initiative. At this level, our data allows to unveil empirical patterns of distribution of TGIs across issue areas, institutional design features, and participation of states and other actors. We can also test hypotheses about the driving forces underlying TGI emergence. Furthermore, our data collection protocol included the gathering of information on which states, FIGOs, business actors, and civil society organizations participate in which TGIs. With respect to states, this information permits the creation of country-year data where, based on TGI founding dates, researchers can examine which states participate in the creation of TGIs in a given year. This information provides the basis for analyzing patterns of state participation over time and across issue domains and testing hypotheses about why some states participate in TGIs while others abstain. Finally, our data allows to construct state, FIGO, business, and civil society organization by TGI matrices for each year in our data. At this network level of analysis, patterns of connection among actors (based on joint TGI participations) and TGIs (based on shared participants) can be detected and their evolution examined. A final original characteristic of the dataset is its focus on a broad set of institutional design features of TGIs. The institutional design characteristics in our data include the specification of behavioral obligations for participants, the presence of a secretariat, as well as the existence of monitoring, enforcement, dispute settlement, and decision-making mechanisms. While the list of institutional design elements covered by our data is not exhaustive, our data is the first that allows researchers to describe and analyze empirical patterns of variation in the specificities of TGI design in such detail. This opens up room for exploring the heterogeneity among TGIs and testing hypotheses about what drives institutional design differences across initiatives as well as 5

6 what role institutional design plays in shaping states and other actors decisions to participate in TGIs. The purpose of this paper is not to analyze a particular theoretical argument about the emergence, development, or effects of TGIs, but to introduce the Transnational Public-Private Governance Initiatives in World Politics Data. I start out by discussing our definition of a TGI and how it was used to generate the sample. Second, I describe the definitions and coding procedures for the variables that measure TGI scope, participation, and institutional design. Third, I present descriptive statistics at the initiative and state level of analysis to show some of the empirical patterns in the data. 6 I also re-examine a recent finding about state participation in transnational governance initiatives to show how the new data can be used to enrich empirical tests of established theoretical arguments. I conclude with a discussion of puzzles and avenues for future research on transnational public-private governance initiatives in world politics. The dataset No undisputed definition of the universe of TGIs in world politics exists and it is unclear how many organizations this universe of cases contains. While there is consensus that the number of transnational public-private governance arrangements has risen dramatically over the past decades, numbers range across sources. Kaul (2006, p. 219), for example, shows that the number of transnational public-private partnerships has increased from 50 in the mid-1980s to at least 400 in the 2000s. Similarly, Abbott and Hale (2014) document the existence of 223 global solution networks of varying mixtures of states, business, NGOs, and other actors. Focusing on environmental governance, Andonova (2010, p. 25) reports the creation of over 400 so-called type II partnerships in the aftermath of the 2002 UN World Summit on Sustainable Development and the initiation of more than 150 collaborations between UN agencies, states, and non-state actors between 1998 and 2008 under the auspices of the UN Fund for International Partnerships. Finally, Widerberg and Stripple (2016) show that even within a single issue area, such as climate change, the number of existing 6 I leave the exploration of the network-level information in the data to future research. 6

7 governance arrangements reported varies considerably across databases. This ambiguity is at least partially due to the lack of a widely shared definition of transnational public-private governance arrangements (Börzel and Risse, 2005; Abbott and Snidal, 2009). Therefore, building on existing research, we develop our own definition of TGIs and use it to specify the criteria for selecting organizations into our sample. We define transnational public-private governance initiatives as institutions that 1) involve at least one state and/or FIGO, 7 one business actor, and one civil society organization; 2) perform tasks that are related to governing global problems; and 3) are institutionalized to the extent that they create a basis for shared expectations about behavior and are observable. First, we focus on institutions that are of a multi-stakeholder nature and bring together actors from the public sector, the private for-profit sector, as well as the private non-profit sector. In other words, our definition of a TGI zooms in on the center of the governance triangle proposed by Abbott and Snidal (2009). State actors are governments, government agencies, or representatives of government agencies. Institutions in which the only public participant is a local government actor, such as a municipality or a city, are not included in our sample. FIGOs may participate in a TGI either through their main secretariat or their organizational branches. Business actors encompass firms, business associations, and business foundations. Civil society actors may be NGOs, NGO coalitions, or universities and research institutes. Second, TGIs are built to fulfill a task or set of tasks that is related to providing governance at the global level in a broad sense. Governance tasks that TGIs may be concerned with include the creation of rules and standards that govern the behavior of states, corporations, and other actors, the implementation of rules and standards, the financing of projects, as well as the facilitation of information exhange and networking related to a global problem. Other governance tasks that may be part of the activities of TGIs are agenda setting as well as the monitoring, enforcement, and 7 FIGOs are traditional formal intergovernmental organizations, such as the United Nations, the World Bank, or the World Trade Organization. We define FIGOs in accordance with the Correlates of War project which considers an international institution an intergovernmental organization if it has at least three member states, holds regular plenary meetings at least once every ten years, and possesses a permanent secretariat and corresponding headquarter (Pevehouse, Nordstrom and Warnke, 2004). 7

8 adjudication of rules and standards (Abbott and Snidal, 2009; Avant, Finnemore and Sell, 2010). In other words, TGIs are explicitly focused on contributing to governing global problems or the provision of global public or collective goods. As a consequence, our sample does not include, for example, information sharing platforms without a focus on a particular governance problem. Third, TGIs are characterized by a minimum level of institutionalization that generates stable shared expectations about the behavior of the actors involved. This minimum level of institutional structure is critical for distinguishing TGIs from a range of instances of cooperation at the global level that are of a less regular nature. For example, once off, ad-hoc meetings of public and private actors are not captured by our definition of a TGI. Our definition also does not include projectbased collaboration between states and/or FIGOs, business, and NGOs. The minimum level of institutionalization of TGIs is also important from a practical research perspective. TGIs are often characterized by a lower level of institutional formalization than other forms of global governance, such as FIGOs. While mapping the empirical variation in the level of institutional formalization of TGIs is part of the purpose of the dataset, there is a limit to the extent of the informality of global governance institutions that researchers are able to observe. One of the hallmarks of informal institutions is that they are difficult to trace empirically because they leave little or no publicly accessible paper trail (Christiansen and Neuhold, 2012; Koremenos, 2013). At the lower end of the continuum of institutional formalization, governance efforts become difficult to observe and many global governance institutions of a highly informal character may never be observed by any researcher. To address this problem and to avoid bias toward more institutionalized forms of governance in our dataset, we theoretically exclude from our definition of a TGI, governance efforts that feature very low or no institutionalization. All institutions that meet these three requirements are part of the population of TGIs in world politics as understood here and therefore targeted by our data collection. An example of a TGI that governs issues related to global health is the Global Fund to Fight AIDS, Tuberculosis and Malaria (GF) (Liese and Beisheim, 2011). Based on the participation of states, the private sector, and NGOs, the GF garners, manages, and disburses resources to fight HIV/AIDS, Tuberculosis, and 8

9 Malaria. Founded in 2002, the GF has a governing board and its work is supported by the Global Fund Secretariat, the Global Fund s Office of the Inspector General, as well as other institutional structures. 8 In the security area, the International Code of Conduct for Private Security Service Providers Association (ICoCA) was founded in 2013 (Avant, 2016). Building on the International Code of Conduct for Private Security Service Providers process that started in 2009, the ICoCA brings together states, private security companies, and a range of civil society organizations that collaborate to create and monitor standards for private security service providers and how states make use of their services. The ICoCA meets on an annual basis in the form of its general assembly, has its own secretariat, and working bodies focused on different substantive aspects of the governance of private security providers. 9 The dataset presented in this paper is a sample of the population of TGIs in world politics. The creation of our sample took place between 2015 and To identify TGIs, we applied our definition of a TGI to seven source databases that were identified as containing information about transnational governance arrangements. These databases include the Global Solution Network (GSN) database, 10 the United Nations Environment Programme (UNEP) Climate Initiatives Platform, 11 and the UN World Summit on Sustainable Development (WSSD) partnerships database. 12 We also searched all organizations that are recorded as networks in the Yearbook of International Organizations. 13 Further, we applied our TGI criteria to the organizations documented in the databases of Andonova, Hale and Roger (2016) as well as ongoing research projects at the University of Zurich and Duke University. 14 Finally, we searched websites of TGIs that were included in our sample for references and links to identify additional organizations that meet our definition. Thus, in total we used eight sources to identify cases for our dataset. In a first step, a team of researchers worked with a portion of the GSN database and tested our 8 See accessed: See accessed: See accessed: See accessed: See accessed: See accessed: I thank Katja Michaelowa and her team at the University of Zurich as well as Ben Collins and Suzanne Katzenstein (Collins and Katzenstein, 2016) and their team at Duke University for sharing their databases. 9

10 Figure 1: TGI cases identified in source databases TGI cases identified in source databases Cases 50 0 GSN Andonova et al. UNEP Duke Zurich Yearbook UN WSSD Note: The case identification stages are shown in the chronological order in which they were undertaken. TGI identification criteria and their empirical operationalizations. Each researcher independently identified TGIs in this first portion of the GSN database. To identify cases, we used websites of organizations, primary documents, such as founding charters and mission statements, government and NGO reports, as well as secondary sources, such as academic papers. Whenever possible, we triangulated sources to increase the validity of coding decisions. Once the first set of case identifications was finished, results were compared and the initial experiences used to revise the coding guidelines and harmonize coding practices. After this identification of a first set of TGIs, researchers continued individually to work through the remainder of the GSN database and the other source databases to identify additional cases. In total, we included 468 TGIs in our dataset (for a full list of the TGIs in the sample, see the online appendix). At the time of the creation of our sample, the GSN database consisted of 762 organizations of which we identified 151 as TGIs. The UN WSSD partnerhip list and the UNEP Climate Initiatives Platform included 187 and 184 organizations and we identified 40 and 23 TGIs in these databases respectively. A total of 2,625 organizations were identified as networks in the 10

11 Yearbook of International Organizations. Using our selection criteria, we found 96 TGIs among these organizations. The database of Andonova, Hale and Roger (2016) contains 119 cases, while the databases of the research projects at the University of Zurich and Duke University involve 684 and 84 cases respectively. Across these three sources, we identified 18, 39, and 15 TGIs. Finally, we found an additional 86 TGIs through our search of the websites of organizations that were identified as TGIs in our data. Figure 1 provides an overview of the sample creation. For each TGI in our data, we collected basic information, such as its full name, founding date, and termination date (if the organization does not exist anymore). The temporal domain of our data starts with the founding of the first TGI in 1885 and ends in The growth in the number of TGIs over time is unevenly distributed temporally. The first organization in our data is the International Association of Public Transport founded in 1885 and the most recent one is the European Technology & Innovation Platform Photovoltaics created in TGIs started to grow dramatically in numbers beginning in the 1980s and particularly in the 1990s and 2000s. While most of the TGIs in our data continue to operate today, a small number of 33 (7 percent) do not exist anymore. Measuring TGI scope, participation and institutional design To advance the understanding of transnational public-private governance initiatives, we measure the scope, participation, and institutional design of the TGIs in our dataset. We build on previous data collection efforts for identifying issue areas and expand those issue area categorizations to capture additional variation (Tallberg et al., 2013; Abbott and Hale, 2014; Pevehouse et al., 2015; Koremenos, 2016). We measure TGI policy scope based on a fine-grained catalogue of 40 subissue areas ranging from education, migration, and human rights to monetary policy, trade, finance, and defense. For each sub-issue area, we create an indicator variable that captures whether a TGI is concerned with this policy sub-issue or not. The empirical assessment of whether a particular TGI deals with a specific sub-issue is based on a list of key words related to this issue. For example, the coding of a TGI as being concerned with finance issues is based on the key words bank, 11

12 fund, financial, finance, financial system, and capital. If one or several of these key words are present in the founding documents, mission statement, or other materials produced by or about a TGI, it is coded as active in the finance area. The list of key words for identifying TGI sub-issue areas was generated by an iterative deductive-inductive procedure. We started with a first set of key words for each sub-issue area identified based on existing theoretical and empirical research on transnational governance arrangements and international institutions more generally. This provisional list of key words was then used to code a first set of cases. During this first coding round, additional sub-issue areas and key words were identified and existing sub-issue areas and key words further specified. The resulting revised list of sub-issues and key words was then used to re-code the initial set of cases as well as the remainder of our sample. The 40 sub-issue areas are not mutually exclusive. One TGI can be concerned with more than one sub-issue area. For the TGIs in our dataset, it is common to operate in more than one sub-issue area. The minimum number of issues addressed by TGIs in our sample is 1 and the maximum is 13. Twelve percent of the organizations in our sample have a single issue focus. By contrast, about 77 percent address between 2 and 5 issues and about 10 percent deal with 6 and more issues. For example, the Communities and Small-Scale Mining Initiative deals with a broad range of 9 issues including commerce, labor, health, and technology, while the Roll Back Malaria initiative focuses exclusively on health-related problems. To allow for an analysis of the policy scope of TGIs in comparison to FIGOs, we aggregate our sub-issue areas into a set of 9 broader issue areas; namely, security, trade and commerce, finance, development, social affairs, technical, human rights, health, and environment. The issue area data for FIGOs comes from the most recent version of the Correlates of War (COW) intergovernmental organization (IGO) data (Pevehouse et al., 2015). 15 For the TGIs, the aggregated issue area category development, for example, consists of the sub-issues development in general, food and agriculture, and regional development. For the FIGOs, organizations that are concerned with food and agriculture as well as development in general, are coded as operating in the development area. 15 I thank Jon Pevehouse for sharing the data prior to its public release. 12

13 Table 1 provides an overview of the issue area aggregation that we apply to make our TGI data comparable with the COW IGO data. Table 1: Issue aggregation TGIs and FIGOs Issue area Sub-issue area TGI Sub-issue area FIGO Security Peace and stability, soft security, private security, defense, conflict, counter-terrorism, money laundering, piracy, nuclear proliferation, proliferation Defense Trade & commerce Trade, commerce Trade, commerce Finance Bank and finance, monetary policy Bank, monetary policy Development Development general, food and Food and agriculture, development agriculture, regional development general Social affairs Tourism, labor, legal, crime, communication, Culture, labor, legal, education, transparency, culture, transport, communication, sport, education, society, transport, women, tourism women, sport, migration Technical Technical, secience and technology Technical, science and technology Human rights Human rights Human rights Health Health Health Enviroment Environment general, climate Environment general, energy change, energy In addition to policy scope, the data includes information about the participation of different types of actors in TGIs. Specifically, we measure state, FIGO, business, and civil society participation for each TGI in our data. We measure participation as involvement of an actor in a TGI in a broad sense, which may or may not include voting rights. For measuring state participation, we start from the COW project list of states that are part of the international system and record whether a state occurs as participant of a TGI. Non-COW states or state-like entities (e.g. Guam, Palau, Cook Islands) are not included. Our data includes both information on how many states participate in a TGI as well as which states are participants of which TGIs. We also measure how many FIGOs participate in the TGIs in our data. We count sub-bodies of larger FIGOs (e.g. sub-agencies of the UN or EU) as individual FIGOs. In cases where the EU participates in a TGI, we clarify whether it is the EU as an international organization and/or 13

14 its member states that participate and adjust the counting of FIGOs and state participants based on this assessment. We again collect information on the number of FIGOs participating in each TGI as well as detailed information on which FIGO participates in which TGI. In addition to states and FIGOs, we include information on the participation of business and civil society actors in the TGIs in our data. For business actors, we include counts as well as detailed participation information for firms, business associations, and foundations. For civil society organizations, we include information on the number of NGOs and NGO coalitions as well as the number of universities and research institutes involved in a TGI. We also collect information on which civil society organization is part of which TGI. As for states and FIGOs, we report current or most recent available participation information. To measure the institutional design of TGIs, we assess whether the organizations in our dataset have a founding document, specify behavioral obligations for their participants, are precise in terms of their policy scope, have a secretariat, have a forum for regular meetings of participants, define monitoring mechanisms, have enforcement procedures, have dispute settlement mechanisms, and specify procedures for making decisions. We measure these institutional design features of TGIs using indicator variables that are coded 1 if a design element is present as part of the institutional structure of a TGI, and 0 if it is absent. Table 2 summarizes the 10 institutional design variables in our data. Measuring design dimensions of TGIs dichotomously provides a rough proxy for capturing institutional structures. An ordinal (e.g. low, moderate, high) or continuous measure that captures for each design element not only its presence, but also the extent to which it is present (e.g. no monitoring versus self-monitoring versus peer monitoring versus independent third-party auditing), would be preferable. However, measuring institutional design intensities validly and reliably across a large number of TGIs and across researchers is a challenging task. Dichotomous measures are less error prone and likely to produce measurements that are reproducible across coders and over time. Thus, our dichotomous measures of the institutional design features of TGIs provide sufficient detail to capture theoretically different categories of TGI design and highlight interesting 14

15 Table 2: Institutional design variables Variable Description Coding Document TGI has a founding document {0,1} Obligation TGI creates behavioral obligations for its participants {0,1} Precision TGI addresses a single policy issue {0,1} Secretariat TGI has a secretariat {0,1} Secretariat independent TGI has an independent secretariat {0,1} Forum TGI has a representative body that meets regularly and {0,1} can make decisions Monitoring TGI has institutional procedures to monitor the behavior of its participants {0,1} En f orcement TGI has institutional procedures to address noncompliant {0,1} behavior of its participants Dispute settlement TGI has institutional procedures to settle disputes among {0,1} its participants Decision-making TGI specifies decision makig procedures {0,1} empirical variation, while at the same time facilitate valid coding and minimize measurement error by focusing on clear differences across TGIs (Mitchell and Rothmann, 2006, pp. 9-10). The variable Document captures whether a TGI has a founding document. A founding document can be an international treaty, charter, convention, declaration, memorandum of understanding, bylaws, or other documents that constitute a TGI. The variable captures whether such a document exists, in which case it is coded 1, or not, in which case it is coded 0. If no founding document can be found, but there are sources that refer to the existence of such a document (e.g. reports, academic papers), the variable is coded 1. Obligation captures whether a TGI creates rules and commitments for the behavior of all or some of its participants (Abbott et al., 2000). Obligations may or may not be legally binding and are meant to be understood broadly. Specifically, the variable Obligation is coded 1 if a TGI has one or several of the following characteristics: 1) the participants of the TGI are expected to behave in line with a specified set of rules, principles, recommendations, or guidelines created by the organization; 2) the participants of the TGI have specified duties, tasks, or other obligations that they are expected to fulfill; 3) the participants of the TGI are expected to satisfy or respect a specified standard or other criteria to become or remain participants of the organization; and 4) the 15

16 participants of the TGI are expected to execute or implement decisions of the organization. If none of these four criteria is met by a TGI, Obligation is coded 0. We also measure how precise the policy focus of the operations of a TGI is. For this purpose, the variable Precision captures whether or not a TGI is a single-issue organization. Institutions that cover only one issue area are coded 1, all others are coded 0. The underlying idea is that addressing multiple issues at the same time opens up room for interpretation and discretion for participants as well as the organization itself, and leads to less precision (Hafner-Burton, Mansfield and Pevehouse, 2016). The variable Secretariat captures whether a TGI has any kind of secretariat. It refers to an institutional body whose purpose it is to assist the institution as a whole or its participants by sharing and circulating information, providing technical assistance, preparing meetings, or similar activities. This secretariat must be a part of the institutional structure of a TGI, but does not need to be part of the organization. It can also be provided by another organization, such as a FIGO. If an institutional body that provides some form of administrative support to the participants of the TGI exists, then Secretariat is coded 1, otherwise 0. We further specify the nature of a TGI s secretariat by coding whether or not a secretariat is independent. The variable Secretariat independent is coded 1 if a TGI has a secretariat or secretariat-like structure as part of its own institutional framework (e.g. the Kimberly Process where the secretariat belongs to the organization itself). If there is no evidence for such an independent secretariat, the variable is coded 0. For example, TGIs are coded 0 on this variable, if they either have no secretariat or if there is a secretariat that is provided by another international institution (e.g. the Inter-American Competitiveness Network has a secretariat that is hosted by the Organization of American States). The variable Forum captures whether a TGI has a broadly representative decision-making body, such as a general assembly, a (bi)annual plenary session, or a steering committee (Pevehouse et al., 2015). The variable is coded 1 if three conditions cumulatively hold: 1) the forum is representative regarding the TGI s participants; 2) the forum meets at least on a biannual basis; and 3) the forum 16

17 has decision-making power. If one of these elements is absent, the variable is coded 0. Theoretically, the underlying idea is to capture institutionally structured interactions that occur with a minimum level of regularity. The type of representation, as well as the nature of the decisions the forum is able to make, are not taken into account. The variable Monitoring captures whether a TGI has a monitoring mechanism. It covers any institutional means to observe participants activities with respect to the rules and obligations specified by a TGI (Mitchell, 1994; Downs, Rocke and Barsoom, 1996; Dai, 2002; Raustiala, 2005; Koremenos, 2016). The variable is coded 1 if there is a (working)group, task force, committee, or other institutional body whose purpose it is to collect information about the behavior of the participants of a TGI. If there is no evidence for any form of institutionalized monitoring in a TGI, the variable is coded 0. The monitoring can be mandatory or voluntary and conducted by an external auditor as long as it is mentioned as part of a TGI s institutional structure. As for all other design variables in our data, this variable captures institutional design only, not institutional activities or outputs. Thus, whether or not monitoring mechanisms that are in place are actually used or how they are used, is not relevant for the coding of this variable. The variable En f orcement captures whether a TGI has an enforcement mechanism. The variable is coded 1 if there is an institutional body in a TGI whose purpose it is to enforce participants compliance with the rules and standards of the initiative (Mitchell, 1994; Downs, Rocke and Barsoom, 1996; Raustiala, 2005; Dai, 2007; Koremenos, 2016). This body can take the form of a (working)group, a committee, a general assembly, or the chair of a TGI, among others. If there is no evidence for any form of institutionalized enforcement in TGI, the variable is coded 0. If the organization encourages or authorizes enforcement and sanctioning of non-compliant behavior by its participants, but does not itself engage in enforcement, En f orcement is coded 0. The variable Dispute settlement captures whether an institutional body exists in a TGI whose purpose is to settle disputes, differences, or misunderstandings among participants (Smith, 2000; Koremenos, 2007, 2016). These disputes can be about the implementation of rules or other commitments under the organization. They may also arise in the context of the internal operations of 17

18 a TGI. Dispute settlement is coded 1, if dispute settlement is delegated to a body or procedure as part of the institutional structure of a TGI. If no institutionalized dispute settlement exists in a TGI, the variable is coded 0. If lacking its own dispute settlement mechanism, a TGI may delegate it to another international institution (e.g. an international court). The variable Decision-making captures whether procedures for decision-making are specified in a TGI. It is coded 1 if such rules are explicitly specified as part of the institutional structure of an initiative, and 0 otherwise. The variable captures decision-making procedures in general, not only voting in a narrower sense. If a TGI, for example, uses a consensus procedure and does not take votes in a narrow sense, then, as long as consensus decision-making is specified as decisionmaking procedure, Decision-making is coded 1. We used a coding approach of sequentially coding each variable for all cases to code TGI issue areas, participation, and institutional design (Mitchell and Rothmann, 2006). Coders were randomly assigned to variables and portions of our sample and then coded one or few variables at the same time across all cases assigned to them. Several sources were used to collect data on TGI scope, participation, and institutional design. Much of the data was obtained from websites of organizations. Further, we consulted primary sources, such as founding documents, mission statements, and other output generated by the governance initiatives in our sample. In addition to documents produced by TGIs, oftentimes we found reports and other information issued by states, FIGOs, and NGOs that allowed us to measure one or several of the TGI features of interest. Finally, particularly for more prominent and older initiatives, we were able to use secondary sources, such as academic papers and books, to obtain information. Mapping TGIs in world politics I begin the overview of the data with several observations at the level of the population of TGIs from 1885 until 2014 as captured in our sample. 16 The left-hand panel of figure 2 presents the I drop the years 2015 and 2016 since there are only five organizations in total in our data that were founded after 18

19 net cumulative total numbers of TGIs based on initiatives founding and termination dates. Up until the 1980s and 1990s, the growth rate of TGIs in the international system has been low. TGIs were not absent from the scene of world politics, but constituted in terms of numbers a marginal phenomenon. During the 1990s, however, we observe a dramatic shift in the speed of TGI growth. Specifically, from a moderate growth rate in the periods before, we now have an exponential growth with growth rates of 173 percent between 1990 and 2000 and 153 percent between 2000 and Figure 2: TGI growth and issue areas Net cumulative growth, TGIs across issue areas, Count 200 Percent Development Environment Finance Health Human rights Security Social affairs Technical Trade & commerce While the growth of TGIs in the past decades is striking, it is not universal. In some issue areas we observe more TGIs than in others. The right-hand panel of figure 2 shows that in 2014, we have 64 percent of the TGIs in our data dealing with social problems, 40 percent addressing development questions, and 50 percent governing environmental issues of various sorts including climate change and sustainable energy. Health, technical, and human rights TGIs are less prominent with 25, 18, and 6 percent respectively. The emergence of TGIs as an element of global governance is not restricted to the areas of low politics. Twenty-four percent of the TGIs in our data operate in the trade and commerce area, 5 percent are active in the finance domain, and 4 percent govern security issues. Importantly, the latter finding is not exclusively driven by TGIs that address issues 19

20 Figure 3: TGI growth across issue areas TGI growth, Development Environment Finance Health Human rights Security 200 Count Social affairs Technical Trade & commerce that could be referred to as soft security, but also by organizations that address issues, such as counter terrorism, piracy, money laundering, and nuclear proliferation. This underlines that transnational public-private governance is becoming a relevant mode of global security governance (Avant and Westerwinter, 2016). In sum, TGIs are not distributed equally across issue areas and are increasingly important in both low and high politics. The growth trajectories of TGIs also vary across issue areas. We see in figure 3 that the creation of TGIs in the development, environment, and social affairs areas can be traced back to the 1970s and earlier. In all three issue areas, we observe a slow growth in the 1970s and 1980s followed by a dramatic increase in growth rates in the 1990s and particularly 2000s. For TGIs active in the health, technical, and trade and commerce areas, we observe similar growth patterns, although here the development started in the 1990s with exponential growth rates only occurring in the 2000s. In the finance, security, and surprisingly the human rights areas, the growth of TGIs has remained moderate so far, both in terms of extent and speed. In addition to growth trajectories and variation across issue areas, our data allows for examin- 20

21 Table 3: Average number of TGI participants over time Year States Business Civil society FIGOs Total ing differences in the participation in TGIs, both over time and across issue areas. Table 3 shows the average number of state, business, civil society actors, and FIGOs that participate in the TGIs in our sample over time. As we can see in the last column of the table, the average total size of the TGIs in our data varies from 65 participants for the TGIs that were founded in the period to about 210 participants for the initiatives that were created between 1995 and Furthermore, business and civil society are on average the most prominent groups of TGI participants, independent of the time period. State participation occurs, compared to business and civil society participation, at a moderate level ranging from on average 6 state participants per TGI in to about 12 state participants on average in FIGOs are the least prominent group of TGI participants. Additional variation becomes visible if we examine participation in TGIs across issue areas. Table 4 reports average participation numbers across issue areas in Starting again with the average total number of participants, we can see that total participant numbers range from 78 participants on average in the security domain to more than 600 participants on average in the human rights area. Note that while human rights TGIs are small in number compared to TGIs in other issue areas, they are the largest in terms of average number of participants and particularly in terms of business and civil society participants. Further, as we see from column three of table 4, the business participants drive the high average total numbers of participants in human rights, trade and commerce, and environment TGIs. With the exception of the health area, companies, business associations, and foundations participate in TGIs on average more frequently than civil society organizations, states, and FIGOs. State and especially FIGO participation is low across 21

22 issue areas. Table 4: Average number of TGI participants across issue areas, 2014 Issue States Business Civil society FIGOs Total Development Environment Finance Health Human rights Security Social affairs Technical Trade & commerce Our data also allows for exploring differences across TGIs in terms of their institutional design. Table 5 maps these differences across issue areas in We can see that institutional design features, such as forum, secretariat, document, obligation, and decision-making are relatively common among the TGIs in our data. While about 36 percent of the initiatives in our data provide for regular meetings of participants that can make decisions, 52 and 29 percent have a secretariat and a founding document. About 46 percent of the TGIs in our data specify behavioral obligations for their participants and articulate expectations toward their behavior. The explicit specification of decision-making procedures is also relatively common with about 32 percent of the initiatives in our data. The variation across issue areas with respect to forum, secretariat, document, and decision-making is limited with the exception of the security and human rights areas. In the security area, TGIs tend to have a secretariat less frequently compared to other issue areas and the overall sample, while they are more likely to establish a forum, have a founding document, and specify their decision-making procedures. In the human rights area, we observe a low percentage of TGIs with a forum, secretariat, founding document, and specified decision-making procedures. By contrast, institutional structures that provide for monitoring, enforcement, and dispute settlement are less frequent with 16, 11, and 8 percent of the initiatives in our data respectively. This could be an indicator that as with many formal intergovernmental organizations, cooperation that takes the form of TGIs lacks depth and therefore does not require institutional structures that allow for the detection and punishment of non-compliant behavior of participants (Downs, Rocke and 22