Event-focused network analysis: a case study of anti-corruption networks

ABSTRACT

Research on diffusion and transfer increasingly relies on the concept of policy networks, but often in inductive, descriptive, and anecdotal ways. This article proposes a more robust method for the comparative analysis of policy networks, a method we term ‘event-focused network analysis’ (EFNA). The method assumes that networks are most clearly revealed in ‘events’ – conferences, meetings, workshops, etc. Databases of participants at these events provide the foundation for social network analysis of the networks of which they are part. The Organisation for Economic Co-operation and Development (OECD) has hundreds of such events annually that are connected to a myriad of policy issues, thus allowing cross-sectoral network comparisons. The article begins with a review and critique of current approaches to network analysis, explains the EFNA approach, and then applies it to anti-corruption networks centred in the OECD. The case study shows the promise of the method, particularly in being able to trace a wider range of actors than is typical, taking us beyond the ‘usual suspects’ in conventional transfer studies.

KEYWORDS:

• Policy networks
• policy transfer
• anti-corruption
• OECD
• social network analysis

Introduction

Networks are now recognized as primary channels of policy diffusion and transfer, particularly at the global level. How networks are defined and conceptualized in research on global policy networks, however, has primarily been metaphorical. In some cases, networks are introduced as a form of organization with distinctive features (e.g. informal, dynamic, non-hierarchical) that are contrasted with other modes of organization, such as markets or hierarchies (Kahler, 2009; Powell, 1990; Slaughter, 2004a, 2017); whereas, in other contexts, the network concept is adopted as means of analyzing relationships between actors that have diverse forms of organization and features (Borgatti, Mahra, Brass, & Labianca, 2009; Scott, 2017; Wasserman & Faust, 1994). Overall, research on global policy networks and their role in policy diffusion and transfer has been inductive, descriptive and – somewhat paradoxically – often focused on single organizations.

For the purposes of further strengthening our understanding of global policy networks in policy diffusion and transfer (Abbott, Genschel, Snidal, & Zangl, 2015a; Hadjiisky, Pal, & Walker, 2017; Legrand, 2015; Milhorance, 2019; Yi & Scholz, 2016), this article adopts a basic definition of a social network as consisting ‘of a finite set or sets of actors and the relation or relations defined on them’ (Wasserman & Faust, 1994, p. 20). We draw on the network concept as a means to analyze the relationships between actors involved at the global level in specific policy fields with the intention of advancing our understanding of the arenas in which policy diffusion and transfer take place. We propose a methodology that is potentially more systematic and empirical. We call this ‘event-focused network analysis’ (EFNA).

While the EFNA methodology is also inductive, it potentially can canvass a wider array of networks, capturing them ‘in action’ at those moments (events) when they connect around a policy issue. It proceeds on the basis of five assumptions:

1. There are N number of global policy networks organized around a policy issue or policy field.

2. We cannot know exactly how many networks constitute N, and we cannot know when and where these networks will ‘reveal’ themselves. In almost all cases, the ‘network’ is a post-hoc mapping of relationships.

3. While social networks can be traced in various ways (e.g. web links to each other are a proxy of connection and interaction; see McNutt and Pal (2011a) or authorship of policy reports see (Corbera, Calvet-Mir, Hughes, & Paterson, 2016; Spence, 2016)), they are most evident in ‘events’ that bring together their constituent members. An event, such as a conference, a workshop or other meeting, is a focal point of network activity (Aykut, 2017; Dumoulin & Saurugger, 2010).

4. The participants (organizations as well as individuals) at these events are the empirical expressions of the ‘network in action.’ There are other ways of tracing networks – e.g. the web connections among organizations, email traffic, social media connections – but events are self-evidently when network participants meet face-to-face, hence ‘in action.’

5. There are some peak, international organizations that have an almost omnibus policy orientation, and hence are magnets for a wide range of policy network events.

For this article (and the wider project of which it is a part), we have selected the Organisation for Economic Co-operation and Development (OECD) as an entry point to try to capture a range of policy sectors and their networks. The organization is well known as a global policy shop on almost every significant issue facing the planet (Carroll & Kellow, 2011; Kellow & Carroll, 2017; Mahon & McBride, 2008; Pal, 2012). It is a tireless venue for meetings, conferences, forums, workshops, and dialogues, bringing together both formal networks (ones that are constituted and self-defined), as well as informal ones (that crystallize around an issue because of a shared interest, but which do not have formal structures). From a network perspective, if can be viewed as a ‘seed node.’

This article uses the policy field of anti-corruption and bribery as a case study to test the EFNA methodology for two reasons. First, the OECD has been a strong and active participant in formulating and advancing global anti-corruption and bribery policies for over 20 years. This provides us with a strong case to assess the approach of using a ‘peak’ organization as an entry point to map a specific global policy network. Second, the policy area of anti-corruption and bribery is global in scope and transnational in character. It demands the attention and efforts of both Northern and Southern governments individually and collectively, but it often holds particular relevance for developing countries. This offers us an interesting opportunity to empirically examine the interactions between actors from different regions, consider how a global policy network may be affected by the relevance of the policy problem to the global South, and explore how the structure of the network facilitates the flow of information and knowledge in multiple directions (South-to-South, South-to-North, North-to-South and North-to-North).

The article has two main sections: (1) review of the approach to policy networks that has prevailed in the diffusion and transfer literature and (2) introduction and application of the EFNA using global anti-corruption and bribery events as a case study. Our methodological conclusion is that the EFNA approach is robust and revealing, and gives us a fresh understanding of policy diffusion and transfer.

Global policy networks: conceptual foundations and approaches

Policy networks: a genealogy

The network concept is often confusing because it has been developed and applied in different ways. This can, in part, be explained by its distinct origins and trajectories in the different social sciences, for example, in sociology with Moreno’s 1938 article on sociometry (Ward, Stovel, & Sacks, 2011) and Granovetter’s classic article on weak ties (Granovetter, 1973). In political science, one source was the emergence of interest group theory as a challenge to traditional institutionalism (Truman, 1951), which fed into Heclo’s prescient notion of ‘issue networks’ (Heclo, 1978) and eventually the idea that pressure groups were organized in ‘policy communities’ (Atkinson & Coleman, 1992; Coleman & Skogstad, 1990; Pross, 1986, 1992). The rise of policy studies as a subfield within the discipline encouraged analysis of policy processes that connected state as well as non-state actors (Marsh & Rhodes, 1992).

This work was focused on domestic politics, but the network concept had a distinct trajectory in international relations and eventually, global public policy studies. Diane Stone helped build the foundations for the study of global public policy, first, through reinforcing the importance and existence of knowledge networks (Stone, 2000, 2002), and then linking them directly to policy transfer and diffusion (Stone, 2004, 2008). By now, networks are a fixture in global public policy studies, and important as well in work on policy transfer (Hadjiisky et al., 2017), though possibly less so for policy diffusion, which may rely more on processes of osmosis than on deliberate transfer (Stone, 2012, p. 3). It should be noted that networks are not necessary for policy transfer, which might happen as a result of rational ‘borrowing’ (Weyland, 2006), learning, or lesson-drawing by decision-makers (Rose, 1991, 1993). However, even these cases benefit from the network concept as an alternate way of thinking about the policy process in that even single ‘actors’ like international organizations need to work through and with other actors to orchestrate policy fields (Abbott, Genschel, Snidal, & Zangl, 2015b). So, we can assume that networks are crucial in the transfer process. How then are networks conceived and studied, what methods are used to map their contours and their characteristics?

Conceptualizing global policy networks

Despite a variety of meanings, the network concept as it has been applied to analyze global policy systems does have distinctive underlying assumptions that, while not always shared among all global policy network proponents, are broadly held in common. We start with those and then build toward the EFNA.

To begin with, global policy networks help us capture something at the meso-level, and indeed are often portrayed as helping capture ‘sub-system’ dynamics. So, one assumption or observation is that global policy networks have a larger number of actors. A second one is that those actors consisted of both states and non-state actors. A third is that states cannot entirely dominate the process that they too have to bargain and coordinate. The network concept captures multiple and distinct (i.e. independent, but at the same time, interdependent) actors, a variable architecture of relationships, and fluid and loose policy processes. These features make networks a natural tool in analyzing global public policymaking, and transfer dynamics.

This trio of assumptions – actors, relationships, and processes – has been at the core of the application of the network concept in global public policy studies. In key early Stone (2002, 2004, 2008) introduced the concept of global public policy networks (GPPN), arguing that they captured those loosely coordinated constellations of governmental and non-governmental organizations dedicated to particular policy issues at the global level. Her interest was in the mobilization of knowledge and policy advice into these global processes, but the crucial innovation was to conceptualize the actors and processes in network terms. Stone’s work is rich with conceptual innovation, and while she offered a general definition of networks, she was also alive to distinctions among different types of global policy networks involving different actors: GPPN combines both state and non-state actors; transnational knowledge networks (consisting of distinct networks of think tanks, foundations, consultancies, but sometimes in combination or overlap); ‘global agora’ (‘global public space of fluid, dynamic, and intermeshed relations of politics, markets, culture, and society’ (Stone, 2008, p. 21); and transnational executive networks (both private sector and international civil servants) (Stone & Moloney, 2019). These of course join earlier conceptual innovations such as epistemic communities (Haas, 1992) and transnational advocacy coalitions/networks (Keck & Sikkink, 1998).

But if we look more closely, despite the detail in definition and in types of global policy networks, the core definition of what constitutes them is still fuzzy – essentially, the trio of assumptions about actors, relationships, and processes. In looking more closely at these three assumptions, we can discern five deeper aspects of global policy networks. These are reflected in more recent work on these networks (Paterson, Hoffman, Betsill, & Bernstein, 2014), but not always explicitly.

1. Ties Between Actors/Members: A global policy network implies the need for connectivity between actors (nodes). Simple contiguity and even some minimal level of interaction (people waiting at a bus stop) does not make a network. A group of actors only constitute a network when there are ‘ties’ (relationships or interactions) that link or connect them. Identifying these ties, measuring the strength (or weakness) of these ties and determining which variables contribute to the strength of a tie are critical components of advancing our understanding of global policy networks and their contribution to the policy diffusion and transfer.

2. Types of Actors/Members: The actors/members of a global policy network can be of the same ‘type’ or organizational form, or a mix of different types. So, we can conceive of networks of think tanks, or of foundations, or of businesses, or of advocacy groups, or of civil servants. Different types of actors may participate in a global policy network for different reasons. For example, a government may have the power to unilaterally make policy decisions related to anti-corruption; however, it may seek out relevant policy networks to inform its decisions or enhance the credibility and legitimacy of its decisions. In contrast, Transparency International has no policy making authority related to anti-corruption, but it does have information and expertise about this policy field. Its purpose for participating in a network might be to directly or indirectly influence policy through issue framing, agenda setting, policy options, or policy monitoring and evaluation. Global policy networks, of course, often combine different types of actors. There is a tendency in network theory to emphasize homophily – like attracts like – and while this is indeed an important insight, it should treated carefully (Ferguson, 2018, p. 46).

3. Power Relations of Actors/Members: Power dynamics exist between actors (or nodes) within a global policy network. Not all actors have the same resources and not all actors are as equally engaged or invested in network interactions (some may be more peripheral).

4. Type and Structure of Network: A global policy network’s structure can be shaped by many variables, including the type(s) of actors (nodes) in the network, relative power of nodes, the position of nodes in the network, and the strength and number of ties between nodes. Some global policy networks might be disaggregated or open, others more concentrated or closed. If we consider on-line social networks, many of them have extremely flat and fluid structures – anyone can sign-on, everyone has the same capacity to contribute, exchanges and interactions are unregulated and completely up to members. In contrast, the ‘Five Eyes’ intelligence alliance is also a network, but it has a very contained membership and formalized structure to manage interactions between members and the exchange of information. Each global policy network has boundary and entry characteristics.

5. Focus, Currency and Value: Global policy networks are defined in terms of some focus, usually framed as a ‘shared interest,’ and what we will call a ‘currency’ of exchange or interactions. The shared focus is easy enough to see in the context of global policy networks – it could be the issue/problem (political prisoners; internet regulation), or the field (human rights; telecommunications). To be an active network member, there has to be interaction, and that interaction must be some form of contribution or exchange, or what we will call the ‘currency’ of the network. In some cases, this can be literal – the financial contributions of donor agencies in some networks. In most other cases, it is the exchange of information.¹ We assume that the higher the value and credibility of the information, the more status, prestige and power that an actor may have in the network. Furthermore, sharing information and interacting with other actors in a network demands time, energy and resources; therefore, the perceived value and credibility of the network itself may have implications for the amount a member is willing to invest in maintaining ties in the network.

These five underlying aspects of the three core assumptions behind global policy networks (actors, relationships, processes) provide a more nuanced appreciation of network actors and dynamics, and also give a sense of the range and variety of global policy networks that might exist. They also have implications for the theme of policy diffusion and transfer being explored in this issue of Policy and Society. At this point, we return to the charge we levelled at the beginning of this article: research on global policy networks is largely inductive, descriptive, and often focused on single organizations.

The inductive and descriptive nature of most of the research is not itself a major sin, nor a methodological violation, but we do think it imposes unacknowledged limits. For example, few if any of the mainstream contributions in the field tackle any of the five points made above in any depth. The ‘variety’ in network structure is ascribed to the mix of state and non-state actors. Reflections on network structures (density of nodes, for example), are rare. There are relatively few references to any of the more rigorous and deductive work on ‘small worlds’ (Watts, 1999,2003). How then are global policy networks identified? It is hard to ‘see’ these networks, especially if we assume that they can be quite widely dispersed, both in space and in membership. Consequently, in the face of this challenge, analysts often begin with a pre-determined policy field/issue X, and then look for organizations or collections of organizations that are obviously focused on that policy, building from there (e.g. Mamudu, Cairney, and Studlar (2015)). A second strategy is looking to formal global policy networks that define themselves as such, and in relation to a policy area or issue. Slaughter’s work on transnational administrative networks relies on this approach (Slaughter, 2004a, 2004b), as does Legrand’s mapping of networks in the Anglosphere (Legrand, 2015; Legrand & Vas, 2014). Coupled with this emphasis on induction and description is what might appear as a (paradoxical) emphasis on single organizations, either because they are themselves networked (Stone) or because they have a leading role in a given networks (Abbott et al., 2015b; Tallberg, 2015).

More recently, social network analysis (SNA) has been introduced as a means to empirically examine specific policy networks. SNA uses graph theory to study the relationships between specific actors (‘nodes’) and the number and strength of ‘ties’ (or ‘edges’) that exist between these actors, which constitute a network. Graph theory enables visual representations of data, which allows researchers to uncover patterns that might otherwise go undetected (Wasserman & Faust, 1994, p. 94). SNA was initially applied in the social sciences by sociologists to study the relationships between people and the structures of social networks (Scott, 2013). Over time, other social sciences have adopted SNA as a tool to map the relationships between different actors, organizations, states and issues to study the structures, features and dynamics of different networks (Borgatti, Everett, & Johnson, 2013; Scott, 2017).

It provides useful analytical tools to organize and study relational data. SNA enhances our ability to empirically analyze the relationships and linkages between actors and the structures of networks rather than on specific actors and their actions (Wasserman & Faust, 1994, p. 19). SNA places the focus on the dynamics that exist in a network. SNA further lends itself to comparative analysis because it assumes ‘that different networks (and nodes within them) will have difference characteristics and that these variations account for differences in outcomes for the networks (or nodes)’ (Borgatti et al., 2009). One application of SNA has been recent work on ‘virtual policy networks’ (VPNs) through the study of weblinks, which signal mutual recognition of membership in a given community (in our case, a policy network) (Barabási, 2014). Mapping these hyperlinks has been used to portray, for example, policy networks around public management reform ((McNutt & Pal, 2011b), women, peace and security (Carpenter & Jose, 2012), and climate change (McNutt, 2008). For the purposes of this research project, we apply SNA to model the relationships of systems of actors interacting at the global level in particular policy fields. This article specifically examines the global policy network(s) engaged with international anti-corruption and bribery policies.

To summarize, the network concept is a valuable tool in thinking about global policy diffusion and transfer. As Stone emphasized, with global policy networks we have a sense of shared ideas or ‘global communities’, of distributed power, of polycentric exchanges of information. But if we stop here, we are still for the most part relying on the network concept as a metaphor, as a mental image. The global policy networks research landscape is still dominated by qualitative studies of self-identified networks or single, influential organizations. SNA methodologies, however, give us techniques to systematically and empirically analyze actors, relationships, connections, and varieties of network structures. How can we combine the best of these approaches to make an advance in the analysis of global policy networks, and get better purchase on policy transfer through these networks?

The EFNA method applied

OECD as an entry point

If we want to see as much of the constellation of global policy networks as possible, we need an entry point – not climate policy, health care, migration, anti-corruption, human rights or any other field just on its own, but potentially all of them at once. For these, we need a focal point or ‘peak’ organization that is simultaneously connected to a number of global policy networks. There are actually fewer of these than one might think, given the proliferation of international bodies. Almost every one of these bodies is policy-field specific. In our estimation, only four global organizations are sufficiently broad in their mandates to connect into multiple policy fields: the United Nations (UN), the World Bank, the OECD, and the G20. Of these four, the OECD seems to be the best candidate. The UN is the largest body, encompassing all states as members, but it is broad in ambit and highly politicized (outside of a few organs such as the World Health Organization). The World Bank is a development bank, and in the last 30 years has linked economic development to political and social development, and so has broadened its mandate and its work. Nonetheless, it retains a focus on poverty reduction among the poorest nations on earth. The G20 is focused on steering the global economy, and only recently (since 2012) has begun to institutionalize a policy advisory capacity through its engagement groups (Cooper & Thakur, 2013).

In our view, the OECD is a node in a wide number of global policy networks, allowing us to capture a wider set of networks that are all linked through the organization (Carroll & Kellow, 2011; Mahon & McBride, 2008; Woodward, 2009). The OECD is neither a bank, a regulatory agency, nor primarily a treaty-making body. Its members² are considered, in the Economist’s usual description, as ‘a club of mostly rich nations,’ but over the years, the OECD’s membership has continued to grow and, more importantly, it has relations with well over 100 other countries and a diverse array of IGOs, NGOs, research institutions and private sector organizations. In fact, it credibility and legitimacy as a ‘peak’ organization in global policy making depends on its ability to broaden the tent both in terms of the types of actors it engages and their geography.

The OECD Secretariat has 12 departments, nine of which are policy-oriented ‘directorates’ (development cooperation; economics; education and skills; employment, labour and social affairs; environment; financial and enterprise affairs; public governance; science, technology and innovation; trade and agriculture). Each of these directorates supports a key committee or committees. These can be further divided into working groups and other bodies, though some are quite small and meet only episodically. There are about 250 in all. The larger and more important ones meet once or twice a year, and with the directorate’s support, host global meetings of interested participants around broader policy issues. These types of events are OECD inspired, hosted or co-hosted, but not themselves OECD bodies or agencies. In this way, the OECD serves as an entry point or portal to global policy networks on almost every significant policy area.

Our approach is to harvest publicly available data from all events sponsored by the OECD through the nine policy-oriented directorates (and their associated committees and bodies) and events hosted by other organizations that the OECD profiles on the events page of its website (http://www.oecd.org/newsroom/upcomingevents/). The data collected about events include the event title, the policy issues, the policy themes, the date and location of the event, the event website, event host organization(s) and event partner organizations. In addition, data are collected on event participants that are identified in the event program or agenda. Data collected on participants include the participant’s name, organization, position and the role of the participant in the event.

With the support of the OECD Library and Archives Unit, publicly available data were collected about events hosted or advertised by the OECD from 1 January 2016 until 24 November 2017. From 1 October 2017 until 30 December 2018, we collected event data from the OECD events webpage. This provides us with a three-year longitudinal dataset of OECD hosted and supported events.

The above catalogue of events also produces a list of all organizations that have participated in these events as part of the formal program or agenda. Subsequently, data are collected on these organizations, including the organization’s name, website, country, mandate, organization type (government, intergovernmental organization, non-governmental organizations, private company, etc.), and primary scale of work (local, national, regional, international).

The EFNA dataset includes 388 events covering over 30 policy issues. This dataset offers opportunities to examine specific global policy networks, compare networks and examine the relationships between these networks. For the purposes of introducing the EFNA method, we will now examine a specific policy field, anti-corruption and bribery, to demonstrate its value to map global policy networks and explore its potential to enhance our understanding of policy transfer.

The OECD’s anti-corruption and bribery policy agenda

The ‘globalization’ of anti-corruption and bribery of foreign officials can be traced back to the Lockheed Aircraft scandals in the mid-1970s. Lockheed (like many other corporations) had routinely made ‘foreign payments’ (often tax deductible) to officials and politicians to ensure contracts. After Lockheed almost went bankrupt and was bailed out by the US government in 1971, a congressional investigation revealed that board members had secretly made ‘foreign payments’ to officials in West Germany, Italy, Japan and other countries in the 1950s and 1960s. One result of the outrage was the passage of the Foreign Corrupt Practices Act that made it a criminal offence for US citizens or companies to bribe government officials. This legislation was unique to the US, and so the American government had an interest in inducing its competitors to adopt similar law. It eventually worked through the OECD to achieve a global regime (Pal, 2012, p. 108–109), and as a result, the OECD has been a key site for the development of global anti-corruption policies ever since.

A key instrument in the OECD’s anti-corruption agenda is an international legal instrument, the 1997 OECD Convention on Combating Bribery of Foreign Public Officials in International Business Transactions. It is the only international instrument that focuses on the ‘supply side’ of bribery transactions, and at the time of writing had been signed by all 36 OECD members as well as eight non-members (Argentina, Brazil, Bulgaria, Colombia, Costa Rica, Lithuania, Russia, and South Africa). It uses peer review as its enforcement mechanism, and is overseen by the OECD Working Group on Bribery. The convention provides a definition of the offence of bribing public officials, and has provisions for sanctions, mutual legal assistance, enforcement, and extradition, among other things.

The OECD’s work related to anti-corruption is linked to three key business lines: anti-corruption and integrity in the public sector (led by the Directorate for Public Governance), tax and crime (led by the Centre for Tax Policy and Administration) and bribery in international business (led by the Directorate for Financial and Enterprise Affairs). The work of each of these directorates can be linked back to the Anti-Bribery Convention; however, they offer different expertise and alternative lenses to approach this policy field. These directorates increasingly work together and with other directorates to understand and tackle the multifaceted and interconnected issues that underpin the OECD’s anti-corruption agenda. In 2011, the OECD’s launched the CleanGovBiz Initiative specifically dedicated to ‘support[ing] governments, business and civil society to build integrity and fight corruption’ (http://www.oecd.org/cleangovbiz/about/).

The OECD is tireless in exploring new frontiers – for example, in the anti-corruption field, recent support for and recognition of ‘whistleblowers.’ It also routinely ensures that wider global policy communities (outside of governments and intergovernmental bodies) are included in the work of developing policy ideas. Its 2018 (March) Global Anti-Corruption and Integrity Forum, for example, had over 1,800 participants from 120 countries, and billed itself as ‘leading the policy debate on integrity and anti-corruption worldwide [that] uniquely brings together government, business and civil society leaders and experts on an equal footing, on the broadest possible range of issues related to integrity and anticorruption, such as trade, foreign bribery, infrastructure, local governance, sports, education, behavioural insights, revenue collection, competition, and development co-operation’ (OECD, 2018, p. 1).

The OECD anti-corruption policy network

As a ‘peak’ organization in the global system, the OECD is an active participant in global anti-corruption policy. We identified 17 public events hosted by the OECD that took place between January 2016 and December 2018 related to anti-corruption and bribery. Event and participant information was available and has been collected for 15 of these events.³ In total, these events included 463 unique participants and 393 organizations.

The data provide a rich picture of the composition of the OECD anti-corruption network. To begin, we examined the characteristics of the nodes (organizations). As might be expected given the OECD’s status as an IGO, government representatives make up the largest percentage (42%) of participants in these events (Figure 1). However, nine other types of organizations participated, including a strong showing by the IGOs (22%), NGOs (11%), private sector organizations (10%), academia (6%) and financial institutions (4%). These events covered a broad range of topics, including foreign bribery and whistleblower protection for public officials, corruption in customs and trade, competition and antitrust, and transparency in public and private business. These data serve to confirm the complexity and multidisciplinarity of anti-corruption as a policy field and demonstrates the OECD’s role as a global convener of diverse actors with the relevant expertise, experience and interests.

Figure 1. Types of organizations participating in OECD anti-corruption events.

Given this Special Issue’s particular attention to the dynamics of policy transfer within and between the global South and the global North, we examined where participating organizations were located. Figure 2 arranges organizations by the region where their office or head office is based. This analysis is quite telling, with a large majority of participating organizations based in Europe (61%) and notably small levels of participation from African (4%) and Arab regions (2%). Classifying the home countries of participating organizations by global North and global South provides an even more dramatic result with 80% of organizations participating in all OECD anti-corruption events based in the global North.

Figure 2. Regional locations of organizations participating in OECD events.

These findings substantiate criticisms that the OECD is a ‘closed club of rich countries’ and provide evidence that the anti-corruption policy network convened by the OECD remains dominated by the global North. However, the dataset also exposes two factors that may exaggerate the dominance of the global North in this network that deserve further investigation. First, initial analyses of global policy networks using the EFNA method indicate that participation levels at events are consistently higher by organizations that are located close to an event. Of the 15 OECD events that were analyzed, 13 took place in Europe (12 in France and 1 in the United Kingdom). The two remaining events took place in Georgia and Cameroon, and for these events a majority (51%) of the organizations that participated were based in the global South. This suggests that access to, and participation in, global policy events are influenced by where they are hosted. Secondly, a review of the data collected about the organizations in this dataset indicates that 58% of the participating organizations have a supranational focus in their work, which suggests that although an organization’s office may be in the global North, it may not necessarily seek to influence anti-corruption policies of the global North or seek to transfer policy ideas from the global North.

To enrich our understanding of the OECD anti-corruption policy network, we use SNA tools to map the relationships between these organizations and examine the structure, patterns and dynamics of the network. As a field of study, SNA has a substantial and diverse array of methods, applications, and techniques. In particular, SNA can be differentiated into descriptive and statistical methods. For the purposes of this article, we adopt descriptive methods to analyze the OECD anti-corruption policy network and develop and test the EFNA methodology. As the research project advances, we will look at opportunities to apply additional SNA techniques, including statistical SNA methods, to the global anti-corruption policy network(s) and other global policy networks identified through the OECD dataset.

To analyze and visualize the OECD anti-corruption policy network UCINET and Netdraw (Borgatti, 2002), two SNA software tools, are used. Nodes are organizations that participated in OECD events related to anti-corruption and the ties are organizations that participated in an event together. As a starting point, we were interested in examining the characteristics of the OECD anti-corruption policy network. Did the organizations in the dataset form a cohesive group or are there distinct sub-groups evident in the network? If there are subgroups within the network, what nodes connect the sub-groups in the network?

Figure 3 visualizes the entire OECD anti-corruption policy network. It highlights the characteristics of the nodes by using colour to represent the type of organization (e.g. state, IGO, NGO, etc.) and shapes to represent the geographic location of the organization (e.g north, south, multiple, etc). Figure 3 highlights that there are distinct sub-groups within the network. For example, there is a sub-group in the network that is dominated by states (green nodes) with a small number of IGOs (yellow nodes). These actors do not interact directly with other types of actors. Interestingly, this segment of the network has a balanced proportion of states from the global North (square nodes) and global South (circle nodes). Other sub-groups in the network have a greater diversity of types of organizations (represented by node colour) spread throughout the network and interacting with each other, but are dominated by organizations from the global North.

Figure 3. 1-mode graph of OECD anti-corruption network using Ucinet and Netdraw software: Node colour represents type of organization, node shape represents location of organizations and node size indicates centrality using the measure of betweennes.

Of course, as we have discussed conceptually, not all nodes in a network are equally important. What more can Figure 3 tells us about the characteristics of the OECD anti-corruption network that might inform our understanding of how this network enables (or doesn’t) the transfer or diffusion of policy and actors to influence global policy making in this field. First, we know that the presence of states in the network (although somewhat isolated from other subgroups in the network) signals a conduit for network members to directly or indirectly transfer and influence policy. Second, examining the overall network, we observe that the OECD anti-corruption network has a core-periphery structure (with a core-periphery fit of 0.83). This network structure consists of a central core of nodes which are cohesive and the remaining nodes in the network which connect only to the core and not to each other (Wasserman & Faust, 1994, p. 419). Networks with this structure are believed to function better than networks with multiple clique structures because the flow of information is centralized (Borgatti & Stephen, 2018, p. 185). Given the importance of the nodes in the core of the network, it useful to focus particular attention on their characteristic, including the types of organizations and the number of nodes because these nodes and their ties are critical to the connectivity of the network and the flow of information through the network. In Figure 3, we also applied a measure of node ‘betweenness’ to the analysis. This measure of centrality is important to understand the characteristics of a large, diffuse global policy network because those nodes with high betweenness scores (represented be the larger size of the nodes) are positioned to control the interactions and the flow of information and ideas between organizations throughout the network (Borgatti et al., 2009). These nodes can be understood as bridges that facilitate the connectivity of the network or gatekeepers that control the connections within the network. These organizations with high betweenness scores, have social capital and power to let people or ideas into the network, connect those with an interest in influencing policy to those with the power to make policy, or play a role in setting the overall agenda for the policy field. These central nodes facilitate the supply and demand of policy advice that is informed by different perspectives, experiences and expertise.

So who are the nodes at the core of the OECD anti-corruption policy network? Despite the large number and diversity of organizations that make up the network, Figure 3 shows us that the core of the network, using the centrality measure of betweenness, is made up of a limited number of organizations – the most central being (in order of importance):

1. OECD

2. Germany⁴

3. United States

4. France

5. United Kingdom

6. European Commission

7. Netherlands

8. World Bank

9. Transparency International

10. Norway

11. Switzerland

12. Singapore

13. Italy

14. Columbia

15. Honduras

16. Brazil

17. Canada

18. New Zealand

19. Nigeria

20. United Nations

It is important to note that the OECD’s centrality in the network is elevated because it is the host of all of the events and in many cases, it draws on in-house expertise. We also see that the core of the network is dominated by states from the global North, two additional IGOs are present (World Band and the United Nations) and Transparency International is the only NGO in the core.

The OECD network is a relatively large network to apply SNA methods and techniques. Its size and complexity can make it difficult to ‘see’ and analyze important aspects of the network’s structures and features. Various techniques have been developed to manage the challenges of analyzing large networks. Figure 4 is a sub-graph of the OECD network presented in Figure 3. It was produced using a k-core measure, which helps to identify interlinked core areas in a network. Figure 4 displays only those organizations that are linked to a minimum of five OECD events (a k-core 5 or greater). This figure also uses a 2-mode network, where black square nodes are the OECD events and coloured circle nodes are the organizations that participated in those events. What we observe is that this figure includes 13 of the 15 OECD events, but only 18 of the 393 organizations in the OECD anti-corruption dataset. This technique produces a visual that incorporates many of the organizations identified in the core of the full OECD network, but this figure allows us to focus in on the relationships between these core organizations and the relationship between these core organizations and specific OECD events. This figure again highlights the prominence of states in the network as well as a handful of well-known Intergovernmental organizations (including the OECD, United Nations, European Commission and the World Bank). We also observe that the only non-governmental organization, Transparency International, rests at the periphery of this core group, which at very least draws into question the meaningful integration of a broader suite of actors in the OECD anti-corruption policy network.

Figure 4. 2-mode sub-graph of OECD anti-corruption network with a k-core of 5 or greater using UCINET and Netdraw software: Black squares represent OECD events, coloured circle nodes represents organization type.

Global anti-corruption policy network(s) beyond the OECD

Is the peripheral nature of many actors in the OECD anti-corruption policy network unique to the OECD anti-corruption dataset? In further developing the EFNA methodology, we are conscious that the OECD is not the only organization doing work at the international-level related to anti-corruption and bribery policy. To expand our view of the global anti-corruption policy network(s), we conducted a preliminary analysis of the OECD events anti-corruption dataset, in particular the core organizations in the OECD network, to inform ‘second-order’ data collection and identify other global events in this policy field. For example, based on the data collected from OECD events, we know that the United Kingdom, United Nations and Transparency International (TI) are active participants in anti-corruption meetings hosted by the OECD, but these actors are also involved in other events around bribery, corruption, and transparency. Therefore, we identified additional events that United Kingdom, United Nations and TI profile on its websites and incorporated these events and the associated participant and organization data, into our database. We also sought the advice of an anti-corruption policy expert to expand and validate our list of global events.

Based on this second-order data collection, we identified 25 additional global anti-corruption events that took place between January 2016 and December 2018. Of these events, data were available and collected for 21. Combined with the OECD dataset, this global anti-corruption dataset includes 31 events and 989 organizations. This dataset is more than twice as large as the OECD anti-corruption dataset, which offers many opportunities to analyze the global anti-corruption policy network (or networks) that it produces; however, it also makes it even more challenging to effectively visualize the network in a way that meaningfully illuminates distinctive characteristics of the network’s (or networks’) structure and characteristics.

For the purposes of this article, we start by applying some of the same SNA descriptive methods and techniques that we applied to the OECD anti-corruption network in order to compare the OECD-specific network with the full global anti-corruption policy network that it forms part of. First examining the composition of the network, we find that the global anti-corruption policy network includes a much larger percentage of NGOs (30% vs 11%), but that governments hold a smaller but still substantial share of network (29% vs 42%). Most of the other types of organizations maintain similar percentages of the overall network with the exception of IGO whose involvement shrinks from 22% to 13% of the network composition. In terms of regional representation, there is no substantial change in the composition of the network, the North continues to dominate, while Southern organizations represent less than 25% of the network.

Examining the overall network, we also observe that the global anti-corruption policy network does form one cohesive network with a core-periphery structure (with a slightly weaker correlation of 0.74). Applying the centrality measure of betweenness to this larger network, the core of the network, includes (in order of centrality):

1. OECD*

2. Transparency International*

3. Germany*

4. World Bank*

5. United States*

6. Argentina

7. UNDP

8. United Nations*

9. France*

10. Ghana

11. United Kingdom*

12. Ukraine

13. Open Contracting Partnership

14. European Commission*

15. Netherlands*

16. Brazil*

17. European Bank for Reconstruction and Development

18. Canada*

19. Romania

20. World Wildlife Fund for Nature

This analysis provides us with unique insights about those actors that serve as bridges or gateways between the sub-groups within the network and reinforces the centrality of 12 of the organizations at the core of the OECD network (marked with *) using the measure of betweenness.

Finally, we undertake the same k-core analysis that we applied to the OECD anti-corruption network. Figure 5 presents the sub-graph that it produces. What we observe is that the core of the global anti-corruption network expands beyond the 13 OECD events to include 30 global anti-corruption events (represented by black square nodes) and the 18 core actors identified in the OECD network are joined by several others to include a total of 46 actors. Analyzing the core of the global anti-corruption policy network, we observe that the very core of this network involves many of the same states and IGOs at the core of the OECD network; however, this sub-graph introduces several new actors at the edges of this core group, including a more diverse collection of states skirting the densest part of the core and an interesting sub-group of IGOs and academia in the bottom left of the graph. This in no way diminishes the centrality of the states and IGOs that clearly continue to dominate this global policy field; however, it does suggest that a broader array of actors have opportunities to assume the responsibilities and behaviours of central nodes in this global policy network.

Figure 5. 2-mode sub-graph of global anti-corruption network with a k-core of 5 or greater using UCINET and Netdraw software: Black squares represent events, coloured circle nodes represents organization type.

Overall, it demonstrates the richness of this type of analysis to enhance our understanding of global policy networks as mechanisms for policy transfer.

Conclusion: implications for the study of policy transfer and diffusion

The term ‘network’ is now used almost universally in scholarship on policy transfer and diffusion. It captures the intuition and observation that policy ‘travels’ through channels and connections that have some institutional scaffolding, but are far from formal institutions themselves. The concept also helps capture the variety of actors in the transfer process, and the speed of information (models, policy ideas) flows among those actors. This article has argued that research on policy networks to date has primarily been inductive, descriptive and often focused on single, key organizations. We then presented our own, alternative EFNA-based analysis, with the illustration of anti-corruption networks. The work is in early stages, but we may draw some preliminary conclusions.

With respect to networks themselves, this approach (and drawing of course on SNA methods) will provide more rigour and clarity in defining the policy networks themselves. For example, some organization X may call itself a ‘global network’ and we might check its membership, or even its website, but the ‘network in action’ is revealed in the events it holds and participates in. Not every event will necessarily expose the full network each and every time, but over time as the catalogue of participants builds, it should show the real network in action, and not simply the organization’s self-presentation.

Another contribution of our approach – in this instance, the aspect that relies on the OECD as our entry point – is that we will be capturing, as part of one research effort, both an eclectic range of policy networks, but also possibly be able to see emergent networks of networks. As mentioned, we are cataloguing the events across nine OECD policy directorates. We have almost 400 cross-sectoral events in the database at this point (for the period 2016–2018), that is, events from policy areas as disparate as environment to education. Almost all of the work on policy networks tends to focus on one, single field, even if that field might be quite extensive (e.g. human rights). An organization like the OECD is so catholic in its mandate that it has events about almost everything that in one way or another impinge on its overarching mandate of contributing to global economic development. This will allow cross-sectional comparisons of networks from different policy fields, but it will also possibly suggest ways and mechanisms by which disparate networks are connected horizontally.

This suggests some important contributions in how we conceive of policy transfer and diffusion. We echo the argument that policy transfer and diffusion need to be seen in ‘polycentric’ terms (Paterson et al., 2014). What this means is an understanding of networks having many different types of actors, from government to NGO to international organizations. Our results above showed the mix of actors, with a significant role played by NGOs. Our approach of focusing on events also suggests something important about the transfer process: transfer does not occur in a linear progression, from a beginning to a clear end, but is an iteration of ideas and models, articulated through cycles of events, meetings, symposia, conferences, congresses, etc. From the perspective of a key actor in a diffusion network, like an OECD or a Transparency International, the task is not merely incremental, it is ‘micro-mental’, with an entire cycle of events dedicated to the wording of a single paragraph, the recognition of a ‘best practice.’ Each small victory can be easily dismissed, but veterans in the transfer theatre take each small, micro step – even the event itself, empty of concrete results – as the foundation for further iterations and further (small) advances. A focus on events in a policy field helps us see the iterations, the cycles, and the constant flows of transfer.

Finally, while we need to be cautious, the EFNA of the OECD anti-corruption network contributes to our understanding global policy networks and the ways in which they are dominated by IOs and other actors from the global North, and there are many other entry points or ‘seed nodes’ that might make a better choice depending on the global policy issue. First, we can see that the OECD network – a crucial one globally because of the importance of the OECD legal instruments – is indeed diverse and eclectic. Up to the end of the 1990s, the OECD, like most IOs, had an aversion to public participation, preferring backrooms and intergovernmental consensus. Since then it has fervently embraced global fora as venues of engagement for both global civil society and governments from around the world (OECD members or not). The events analysis shows regular and on-going meetings, and endless sequence of discussion, reports, reviews, and debates. But the events analysis also suggests that the core of the network is much more exclusive, and dominated by the global North. We see a similar pattern in the second-tier events, with an overlap in core members. It is quite possible that as we expand the research and begin comparing different policy fields and their respective networks, that we may find that some fields and networks are diverse and decentralized both in membership and in core, and that perhaps in others there may be a significant ‘southern’ presence. We may also find changes occurring in core and in membership over time, as networks adapt to reflect the shifting balance of power – and perhaps fragmentation – of the global policy order. Whatever the case may be, we believe that close and careful network analysis is the best path forward to truly mapping and understanding global policy networks.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Social Sciences and Humanities Research Council of Canada.

Notes on contributors

Leslie A. Pal

Leslie A. Pal is Founding Dean, College of Public Policy, Hamad Bin Khalifa University in Qatar, and Chancellor's Professor at Carleton University, Canada. He is currently the Vice-President of the International Public Policy Association, and has served on the Executive Committee of the International Political Science Association. His research focuses on policy advisory systems, transnational public administration, and governance in the MENA region. His most recent books are Global Governance and Muslim Organizations (2019), and Beyond Policy Analysis. 6th ed. (2020).

Jennifer Spence

Jennifer Spence is an Adjunct Research Professor with Carleton University's Department of Geography and Environmental Studies and a Fellow with the Centre for International Governance Innovation.  She is also the Executive Secretary of the Arctic Council’s Sustainable Development Working Group. Jennifer holds a Ph.D. in public policy from Carleton University’s School of Public Policy and Administration.

Notes

¹ Powell’s classic exploration of the network form of organization versus markets or hierarchies (firms) highlighted information flows as a key feature of networks: ‘What is it about networks that makes them more adaptive and well suited to coping with change? One of the key advantages of network arrangements is their ability to disseminate and interpret new information. Networks are based on complex communication channels’ (Powell, 1990, p. 325).

² Twenty countries signed the original convention that established the OECD in the early 1960s (Austria, Belgium, Canada, Denmark, France, Germany, Greece, Iceland, Ireland, Italy, Luxembourg, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, United Kingdom, United States). 16 additional countries have since become members (Australia, Chile, Czech Republic, Estonia, Finland, Hungary, Israel, Japan, Korea, Latvia, Lithuania, Mexico, New Zealand, Poland, Slovak Republic, Slovenia), The European Commission is a partner, not a member, therefore it does not have a vote. As of January 2018, the OECD was engaged in accession discussions with Columbia, Costa Rica, and Lithuania. It started discussions with the Russian Federation in 2007, but in 2014, after events in Ukraine, postponed them indefinitely.

³ The two OECD events that are not included in the dataset are: 1) a 28–29 March 2017 meeting of OECD Task Force on Countering Illicit Trade (TF-CIT) 2017 where no agenda or participant information was posted (http://www.oecd.org/gov/risk/task-force-on-countering-illicit-trade-meeting-2017.htm) and 2) the 7–8 November 2017 OECD Forum on Tax and Crime where no participant information was posted (http://www.oecd.org/ctp/crime/forum-on-tax-and-crime.htm).

⁴ For the purposes of the SNA, all national government departments and agencies were rolled up as one node.