Institutional Overlap in Global Governance and the Design of Intergovernmental Organizations

Students of international cooperation have acknowledged the importance of institutional overlap in shaping the design of international institutions (Copelovitch & Putnam, 2014; Duffield, 2003; Hofmann, 2011; Jupille et al., 2013; Westerwinter, 2022). Studies argue that states do not create new institutions in a vacuum but consider the extant landscape of global governance institutions (Abbott et al., 2016; Copelovitch & Putnam, 2014; Kalyanpur & Newman, 2017). They emphasize how pre-existing institutions can facilitate future cooperation by reducing uncertainty, generating trust, and providing institutional focal points, which inform states in their decisions about the design of new cooperation (Copelovitch & Putnam, 2014; Hofmann, 2011; Jupille et al., 2013). Our argument builds on this productive line of research. We start from the observation that, under conditions of increasing institutional density, states, when making decisions about how to design new IGOs, do not start from a clean slate but build on and respond to the existing institutional architecture of global governance. In such situations, the ideas, interests, and experiences of government officials are already aligned around some set of existing institutions which they consider when negotiating the creation of new IGOs (Jupille et al., 2013; Raustiala & Victor, 2004). In doing so, we assume, states act strategically, although boundedly rational in the sense that they face cognitive limitations and information constraints, and seek to maximize the benefits and minimize the costs of cooperation considering existing institutional context (Jupille et al., 2013).²

We define institutional overlap as the intersection of the governance tasks, memberships, and policy domains of two or more institutions (Hofmann, 2011; Urpelainen & Van de Graaf, 2015). The governance tasks of two institutions can be considered similar if they perform similar tasks, such as knowledge generation, standard-setting, or monitoring, for their member states. The memberships of two institutions intersect if they share some portion of their members. The policy domains of two IGOs overlap if they operate in a similar set of issue areas.

Institutional overlap can emerge from the co-existence of a growing number of international agreements and organizations and an array of informal forms of cooperation among states and non-state actors, including informal intergovernmental organizations (IIGOs) (Vabulas & Snidal, 2013), transgovernmental networks (TGNs) (Slaughter, 2004), and transnational public–private governance initiatives (TGIs) (Westerwinter, 2021). Investigating the consequences of the plethora of overlaps among these different types of global governance institutions is a daunting task. We begin this endeavor by focusing on the overlap among IGOs. IGOs form part of the institutional core of governing many issue areas of world politics (Jupille et al., 2013; Kahler, 2021). They act as focal points (Jupille et al., 2013), orchestrators (Abbott et al. 2015), and managers (Oberthür & Stokke, 2011) of institutional interactions and play important roles in the creation and design of new IGOs (Johnson and Urpelainen, 2014) and other types of institutional arrangements (Abbott & Snidal, 2009; Andonova, 2017). Thus, although overlap among IGOs captures only a portion of all institutional overlaps in global governance, it is an essential portion and consequential for the design of new organizations. In the conclusions, we reflect on how future research may explore overlap among a broader set of global governance institutions.

States look to existing overlapping organizations when deciding which set of institutional design features to adopt for a new IGO (Abbott et al., 2016; Grigorescu, 2010; Hofmann, 2011; Ovodenko & Keohane, 2012; Sommerer & Tallberg, 2019). States that seek to identify the most suitable design combination for a new organization examine the stock of design components and their combinations in pre-existing IGOs that perform similar tasks, in similar policy domains, for similar groups of member states and use these existing design combinations as institutional blueprints (Hofmann, 2011; Kalyanpur & Newman, 2017). In other words, they look at the designs of the “reference group” of overlapping pre-existing IGOs and model new organizations that are meant to fulfill similar tasks in similar policy domains based on these templates (Simmons & Elkins, 2004).

Why would states choose to imitate the designs of overlapping organizations when deciding how to structure new IGOs? Designing the institutional structures of IGOs and matching them to the problem that states seek to address is a time-consuming and challenging task (Abbott & Snidal, 1998; Koremenos et al., 2001; Koremenos 2016). Obtaining high-quality information about the problem under consideration and which combination of design elements are most likely to deliver effective governance and with what distributive consequences is costly and often remains partial even for resourceful governments (De Búrca, Keohane, and Sabel 2014; DiMaggio & Powell, 1983; Jupille et al., 2013; Ovodenko & Keohane, 2012). In other words, uncertainty about effective and otherwise desirable institutional solutions is pervasive among states that negotiate over IGOs based on incomplete information. This uncertainty is further enhanced under conditions of institutional complexity where multiple overlapping organizations are active in governing the same issue domain and interact with one another in various ways. In such situations, matching IGO designs to a given problem is especially challenging because it becomes harder to foresee how overlapping organizations will interact, how these interactions will shape governance outcomes, and how partnering states will respond to the new opportunities to forum-shop among organizations (Jervis, 1997; Orsini et al., 2020). Thus, the uncertainty that states face when designing new IGOs under institutional complexity contains elements of both uncertainty about the state of the world and uncertainty about other states’ behavior (Koremenos et al., 2001). This uncertainty reinforces the bounded rationality of states which makes modeling new IGOs based on overlapping existing organizations an attractive strategy for institutional designers.

Modeling the institutional design of new IGOs based on the features of existing organizations that perform similar tasks and engage with similar policy fields allows states to reduce the costs and uncertainty involved in selecting the features of new organizations (DiMaggio & Powell, 1983; Johnston, 2008; Simmons & Elkins, 2004). As aptly stated by DiMaggio and Powell (1983: 156): “in fields characterized by a high degree of uncertainty, new entrants, which could serve as sources of innovation and variation, will seek to overcome the liability of newness by imitating established practices within the field.” This imitation sets in motion a dynamic of design mimicking which over time leads to increasing homogenization of designs among overlapping IGOs (Biermann, 2008; DiMaggio & Powell, 1983). As a result, as the overlap of IGOs increases, their institutional designs converge toward a similar set of features as new IGOs are crafted drawing on the features of existing organizations (DiMaggio & Powell, 1983; Dingwerth & Pattberg, 2009). This relationship between overlap and design convergence can be expected to be particularly pronounced in uncertain and information poor environments (DiMaggio & Powell, 1983; Kalyanpur & Newman, 2017). High uncertainty may originate from a variety of sources, including the complexity of the problems that new IGOs are tasked to address, such as climate change or financial crises, (Búrca et al., 2014; Henning, 2017) and the ambiguity of their mandates (Kalyanpur & Newman, 2017).

Tendencies toward the convergence of institutional designs between newly created and existing organizations can be observed in many cases where two IGOs overlap in their governance tasks, policy domains, and memberships. For example, the African Union (AU) has been modeled with respect to some important institutional design features after the United Nations (UN) and the European Union (EU) (Brosig, 2011). “One can observe that the decision-making bodies in the UN, the EU, and the AU which deal with foreign and security policy (the UN Security Council, the EU Peace and Security Committee, and the AU Peace and Security Council) are structured similarly” (Brosig, 2011: 161). In the area of development, the newly-founded Asian Infrastructure Investment Bank follows the design patterns of legacy multilateral development banks, such as the Asian Development Bank and the World Bank (Ella, 2021). These examples illustrate that institutional designers, as our argument suggests, look regularly toward the structures of overlapping organizations when making decisions about the design features of new IGOs.

Design similarity through imitation can occur through different pathways. We identify two: direct and indirect contextual design (Biermann, 2008; Grigorescu, 2010; Ovodenko & Keohane, 2012; Simmons & Elkins, 2004; Sommerer & Tallberg, 2019). Under direct contextual design, the convergence of designs is facilitated by overlapping memberships whereby states that participate in both old and new IGOs share their experience with a certain combination of design features for the provision of specific governance tasks and the tackling of particular problems while crafting the new organization. Shared members constitute network links among IGOs (Böhmelt and Spilker 2016; Hafner-Burton et al., 2009) which are “characterized by multiple interaction lines (and) provide an ideal exchange structure for the flow of knowledge across institutions” (Biermann, 2008: 161). This overlap of memberships and the information exchange that can occur through these links acts as a direct transmission belt for design similarity. The more states a new and an existing organization share, the higher the likelihood that information about design features is shared and the architects of the new organization draw on the design components present in the overlapping pre-existing organization (Böhmelt and Spilker 2016; Gehring & Oberthür, 2009; Grigorescu, 2010; Ovodenko & Keohane, 2012). As a result, the new IGO will adopt a package of design features that resembles the design of the legacy IGO.

The creation of the compliance systems of the Montreal Protocol on Substances that Deplete the Ozone Layer (Montreal Protocol) and the Kyoto Protocol under the United Nations Framework Convention on Climate Change (UNFCCC) provides an illustration of design convergence through direct contextual design (Gehring & Oberthür, 2009). When the UNFCCC was negotiated, those states that were already involved in the Montreal Protocol had made positive experiences with its compliance procedure. They shared these experiences with the other countries participating in the UNFCCC negotiations which shaped their beliefs about the possibilities of this compliance mechanism and eventually facilitated the adoption of the same institutional design feature in the Kyoto Protocol (Oberthür, 2001). Thus, the design of the new institution was shaped by the direct flow of information through overlapping member states between the new and pre-existing cooperative effort.

Design convergence may also happen without direct interactions of the memberships of new and old organizations. Without being involved in both new and old IGOs, states may simply survey the designs of existing organizations that provide similar governance tasks in similar policy fields in search for institutional solutions for the problems they seek to address. They assess the available stock of combinations of design elements (Börzel & Risse, 2012; Gehring & Oberthür, 2009; Kalyanpur & Newman, 2017). Based on this stocktaking, they model new IGOs that set out to address similar tasks in similar domains as the legacy organization based on the older design templates (Biermann, 2008; DiMaggio & Powell, 1983; Hofmann, 2011; Ovodenko & Keohane, 2012). When choosing from the menu of possible design packages, states turn to the functional reference group of pre-existing organizations of the new IGO to indirectly derive information and decide how to tailor its formal structure toward the goals they want to achieve (Elkins and Simmons 2004; Sommerer & Tallberg, 2019).

Hofmann (2011) emphasizes the importance of indirect contextual design in the negotiations over the EU’s common security and defense policy (CSDP). Although the institutional structure of the CSDP was a matter of choice among the EU member states, pre-existing institutional structures in the security and crisis management domain were used as templates and informed the range of available options. “NATO represented the most appealing template as it was the most credible and successful security institution around and its membership overlapped significantly with that of the EU” (Hofmann, 2011: 108). As a result, the main bodies of the EU’s new institutional architecture—the Political and Security Committee, the European Union Military Committee, and the European Union Military Staff—“were modelled on their NATO counterparts” (Hofmann, 2011: 108). As this example suggests, direct and indirect contextual design are not mutually exclusive but can operate in parallel and complement each other. We explore both pathways of design convergence in our empirical analysis.

To summarize, we argue that if a new IGO performs similar governance tasks in similar policy fields for a similar group of member states as a pre-existing organization, the design of the new organization is more likely to be similar to the design of the existing organization. In contrast, IGOs that overlap only little or not at all should have less similar institutional features. This discussion yields the following hypothesis:

In practice, IGO design similarity under conditions of overlap is rarely perfect. States use the existing stock of design elements creatively and adapt it to the particular governance problem they wish to address as well as the interests and power of the actors involved in creating a new organization (Börzel & Risse, 2012; Gehring & Oberthür, 2009; Kalyanpur & Newman, 2017; Ovodenko & Keohane, 2012). This adaptation of design templates drawn from overlapping IGOs to the specific circumstances of a new IGO opens up room for design differentiation alongside similarity and allows for the avoidance of exact copies of existing organizations. This is illustrated by the similarity of international courts in regional integration systems, particularly in Latin America and Africa, which have emulated the EU’s European Court of Justice model, while adapting some of its features to protect the national sovereignty of member states (Alter, 2012).

States are likely to consider the performance of past institutional features when deciding on the design of new organizations. If, for example, states use the information contained in the institutional designs of pre-existing IGOs indirectly by looking at how similar the governance tasks they perform are compared to the new organization they intend to form, then rational states can be expected to pay more attention to the design of past IGOs if these IGOs are considered successful. If they failed, we would expect no effect or possibly even a negative effect as states actively try to avoid a design that performed badly in the past (Simmons & Elkins, 2004). The same holds for the situation in which states directly access information about the design of pre-existing organizations through overlapping memberships. States that are members of both old and new organizations are more likely to share their experiences with previous design combinations if these designs effectively served their intended purposes. One principal and highly visible indicator of the general success of a past institutional solution is its viability (Debre & Dijkstra, 2021; Eilstrup-Sangiovanni, 2020; Gray, 2018). While dead organizations may often be considered failures, IGOs that still exist at the time a new organization is created have managed to achieve at least basic organizational goals and ensured their survival. Thus, when drawing on the design templates of overlapping IGOs, the creators of new organizations can be expected to pay more attention to the designs of those legacy organizations that survived as opposed to those that faded away. We, therefore, hypothesize:

To test our hypothesis, we require a measure of IGO design similarity. We define the variable IGO design similarity as the similarity between the design vectors of a given new IGO and each of the IGOs that were created up until one year before its creation. Design vectors are five-dimensional sets of binary variables: 1) secretariat measures whether the founding document of an IGO establishes an independent secretariat; 2) monitoring captures the presence of an institutionalized monitoring mechanism; 3) enforcement accounts for an institutional mechanism to sanction non-compliant behavior; 4) dispute settlement records the presence of a procedure to settle disputes, including issues of interpretation; and 5) voting records the presence of codified procedures for decision-making. We coded these design features of IGOs based on primary sources, such as their founding treaties and public websites, as well as secondary sources, such as scholarly articles and reports.³

These design features are broad traits of IGOs. Naturally, within these categories more nuanced variation exists and is relevant for the operations of organizations. For example, IGOs may have different types of monitoring systems (Dai, 2002) or secretariats with more or less autonomy (Haftel & Thompson, 2006). Nevertheless, our five design variables capture important differences in the formal structures of IGOs that have been highlighted in previous studies (Abbott & Snidal, 1998; Downs et al., 1996; Koremenos et al., 2001; Smith, 2000). Other recent studies of IGO similarity also use binary design features as constituent elements of their similarity indices (Jetschke et al., 2021). 91.6 percent of the IGOs in our data have a secretariat.⁴ Dispute settlement, monitoring, and enforcement are less common, with 30.5, 24.0, and 16.9 percent of the IGOs in the data having these design features. 56.0 percent of IGOs codify decision-making procedures (Table A1).

To obtain dyadic design similarities, we compute the cosine similarity of the two five-dimensional design profiles for any pair of IGOs:where k is an index running through the five IGO design elements, and \(d(k)\) is an indicator function for whether an IGO has the design element k. The subscripts i indicate a given IGO at its founding moment, and j runs over all pre-existing IGOs. Cosine similarity ranges between zero (completely different designs) and one (completely identical designs). Thus, a higher value indicates that the design of a new IGO is largely inspired by the pre-existing organization. Ranging empirically from 0 to 1 in our sample, design similarity has a mean of 0.665 and a standard deviation of 0.229. An example of an IGO pair with low design similarity is the dyad between the Organization of Arab Petroleum Exporting Countries (OAPEC) and the Economic Community of West African States (ECOWAS) (0.447). Both have a secretariat, but only ECOWAS has all remaining design features while OPAEC has not. An example of a dyad with high design similarity are the Multilateral Investment Guarantee Agency (MIGA) and the International Fund for Agricultural Development (0.894). Both have a secretariat and stipulate provisions for decision-making, dispute settlement, and enforcement, but only MIGA has a monitoring mechanism.⁵

Figure 1 plots the distribution of design similarity over time across decades. In the total sample, IGO design similarity has remained relatively stable throughout the period from 1815 to 2014. Looking at the median, IGO design similarity was lower only for IGOs created in 1985–94. In terms of inter-quartile spread, the IGOs created during 1945–54, 1965–74, and 1985–94 were generally less similar to their predecessors than in other periods.

In the appendix, we explore the evolution of IGO design similarity for different subsets that of theoretical interest for students of world politics. We find similar patterns in the evolution of IGO design similarity across issue areas (Figure A1). We focus on trade and commerce, finance, development, and environment. While median IGO design similarities are stable over time for development and the environment, they have first increased in finance until the mid-1970s and declined thereafter, coinciding with the end of the Bretton Woods era. In trade, median design similarity has been high throughout, but was particularly high for trade IGOs established in 2005–14, coinciding with the stalemate at the WTO and the spread of (similar) alternatives. These aggregate trends conceal outliers with more limited overlap, such as in the area of finance where states have experimented with new combinations of institutional design features (Henning, 2017). In addition, we differentiate IGOs by geographical scope (Figure A2). Here we observe relatively stable median IGO design similarities and somewhat increasing diversity in the designs of regional organizations (left panel of Figure A2). These patterns are broadly in line with recent work on the similarity of regional IGOs (Jetschke et al., 2021). For global IGOs, we observe an increase in design similarities throughout most of our observation period, with a peak in the most recent period, indicating a relatively continuous convergence of design portfolios (right panel of Figure A2). Finally, we compare interventionist IGOs and non-interventionist IGOs (Boehmer et al., 2004). For non-interventionist IGOs, we find a gradual shift in the over-time distribution toward more similar designs, while no clear patterns are discernible for interventionist IGOs, which have a generally higher median IGO design similarity.

We introduce a new measure of institutional overlap in global governance, anchored in the regime complexity literature (Alter & Raustiala, 2018; Raustiala & Victor, 2004; Urpelainen and Van de Graaf 2015; Young, 1996). We conceptualize this measure as the overlap of IGO memberships, governance tasks, and issue areas.

The notion of overlap is central to our understanding of institutional complexity in global governance. In their seminal article, Raustiala and Victor (2004: 279) define a regime complex as “an array of partially overlapping and nonhierarchical institutions governing a particular issue area.” In another influential contribution, Orsini et al. (2013: 29) develop a definition of a regime complex as a “network of three or more international regimes that relate to a common subject matter; exhibit overlapping membership; and generate substantive, normative, or operative interactions recognized as potentially problematic whether or not they are managed effectively.” While these definitions differ in several ways, they share a common conceptual core: Institutional complexity captures a situation in which the governance authority of two or more institutions overlap.

This suggests two aspects that are constitutive of institutional overlap: Overlap in policy focus and overlap in membership. Overlap in policy focus means that two IGOs contribute to the governance of the same issue area. For example, the World Trade Organization (WTO) and the EU overlap in their focus on governing trade relations. Similarly, the UN and the AU are both active in the governance of international peace and security. Overlapping membership describes a situation in which the same states are members of two IGOs. For example, the EU and the OECD share large parts of their membership and so do the Organization for American States and the Andean Community.

A third feature that is fundamental for understanding institutional complexity in world politics is the overlap of the functions or governance tasks that IGOs perform. In their pioneering contributions, Young (1996) and Rosendal (2001) state that “overlap implies that the functional scope of one regime protrudes into the functional scope of others” (Rosendal, 2001: 96). More recent studies also highlight intersecting governance tasks, i.e. institutional functions, as a central dimension of overlap (Hofmann, 2011; Urpelainen and Van de Graaf 2015). In the regime complex literature, task overlap among the constituent elements of a complex is identified as one of the key characteristics that make the strategic environment of regime complexes distinct from situations governed by integrated regimes (Alter & Meunier, 2009; Alter & Raustiala, 2018; Raustiala & Victor, 2004). Yet, no existing quantitative measure of overlap in global governance incorporates this critical dimension of intersection between IGOs. For example, the Coalition for Rainforest Nations and the Cocoa Producers Alliance have similar portfolios of governance tasks. They both engage in standard-setting, standard implementation, and information collection and dissemination. Conversely, the Commission for Technical Cooperation in Africa South of the Sahara and the Common Fund for Commodities have entirely orthogonal task portfolios. While the former focuses on service provision and information collection and dissemination, the latter is a standard-setting and implementation IGO.

For any given pair of a new and a pre-existing IGO, our dyadic overlap measure is calculated in four steps. First, we compute the membership overlap between the new and old IGO as the share of common member states between them. Based on information on state memberships in IGOs from the COW IGO data (Pevehouse et al., 2021), we count the number of states that are members of both organizations and divide by the total number of (unique) states that are members in any of the two organizations. More formally, for any IGO dyad, we compute:where \({s}_{i}\left(n\right)\) and \({s}_{j}\left(n\right)\) are indicator functions for the membership of state n in IGO i and IGO j, respectively, and N is the number of states in the international system. \({M}_{ij}\) has a minimum of zero indicating two IGOs with completely disjoint sets of member states and one for two IGOs with identical memberships. We refer to this measure as membership overlap.

Second, governance task overlap refers to the cosine similarity of the profiles of non-mutually exclusive governance tasks of two IGOs. A governance task profile is a tuple of eight dummies, capturing whether the organization has a mandate respectively for 1) information-gathering, 2) agenda-setting, 3) service provision, 4) funding, 5) capacity-building, 6) standard-setting, 7) policy implementation, and 8) monitoring (Abbott & Snidal, 2009; Avant et al., 2010; Westerwinter, 2021).⁶ Formally, we compute this measure as follows:with k being the running index over the K governance tasks and \({g}_{i}(k)\) and \({g}_{j}(k)\) being indicator functions for whether IGOs i and j are active in governance task k, respectively. \({G}_{ij}\) ranges from zero (indicating orthogonal task portfolios) to one (identical portfolios). Importantly, while there may be some overlap between organizations that perform monitoring as a governance task and have a monitoring mechanism as an element of their institutional design, the two variables are conceptually and empirically distinct.⁷ IGOs that monitor implementation of a specific international rule or standard (monitoring as governance task) may or may not have an institutionalized monitoring mechanism for their member states in place (monitoring as design feature).⁸

Third, issue overlap is computed as the cosine similarity of the profiles of the issue areas in which two IGOs are active. An issue profile is a 9-tupel of binary issue areas, drawn from the COW IGO data (Table A3). Similar to other recent datasets on international institutions (Hooghe, Lenz, and Marks 2019; Koremenos 2016; Tallberg et al. 2014), we distinguish nine major issue areas in which IGOs can operate: Security, environment, health, human rights, development, trade and commerce, finance, social affairs, and technical affairs.⁹ Formally, the computation of issue area overlap \({I}_{ij}\) is similar to the formula for \({G}_{ij}\). As Table A3 in the online appendix shows, membership overlap, governance task overlap, and issue overlap are virtually orthogonal, with Pearson correlation coefficients ranging from -0.002 to 0.056. Using factor analysis, we confirm that the three dimensions do not have any common variation that would warrant combining them into a single factor (Table A4).

Finally, we define overlap between a newly created IGO i and a pre-existing IGO j as the product of their membership overlap, governance task overlap, and issue overlap. The rationale for this measure builds on the intuition that all three kinds of institutional overlaps must be present for overlap to exert an effect on the design of IGOs (Urpelainen and Van de Graaf 2015). Formally, we compute:

Overlap varies between 0 for IGOs that are completely unique in terms of their membership, governance task portfolio, and issue area portfolio compared to a given legacy IGO, and 1 for IGOs that have identical memberships, task portfolios, and issue area portfolios with this legacy organization. Values in between these two extremes capture degrees of overlap. In our sample, overlap has a mean of 0.010 and a standard deviation of 0.036. The average overlap of new IGOs with prior organizations is small because of its composite nature: limited overlap in either of the three dimensions causes overlap to be small, even if there is significant overlap in the other dimensions. For example, an IGO dyad with low overlap are the Montreal Protocol Multilateral Fund Secretariat and the Niger River Commission (overlap = 0.008). While both are environmental IGOs that overlap in some functions, their membership is relatively disjunct, in part because the Niger River Commission is a regional organization. An example of an IGO pair with high overlap are the WTO and the World Intellectual Property Organization (WIPO) (overlap = 0.448), given their near-universal membership, functional similarity, and issue area similarity. In this particular case, overlap was engineered through regime-shifting of intellectual property rights protection from the WIPO into the newly-created WTO, notably by the United States (Helfer, 2004). Table A5 in the online appendix shows the top-20 most overlapping pairs of new and pre-existing IGOs in our data.

We use overlap as main predictor to test our hypotheses on the effect of institutional overlap on IGO design similarity. We also disaggregate overlap into two components, examining the individual effects of membership overlap and functional overlap—the product of overlaps in governance tasks and issue areas.¹⁰ This disaggregation allows us to start exploring whether states use the information contained in the designs of past institutions in a more direct or indirect way. If institutional overlap is related to design convergence based on direct information transmission through shared members, we would expect a positive correlation between the membership overlap and the design similarity of new and pre-existing IGOs. If states use information from the designs of past IGOs more indirectly by surveying the design features of existing organizations that perform similar tasks in similar issue areas, this would suggest a positive relationship between functional overlap and design similarity. Table A6 in the online appendix provides an overview of our key measures along with descriptive statistics.

Figure 2 plots overlap averaged over the IGOs created in a given decade. With respect to all IGOs, we find that overlap first declines relatively fast between 1945–1964 before stabilizing in the mid-1960s. In 2005–14, IGO overlap has declined further. These patterns suggest that overall IGOs have become increasingly diversified.

In the appendix, we explore the evolution of IGO overlap among pertinent subsets of IGOs. Looking across issue areas, we find broadly the same characteristic pattern of declining average overlap, albeit starting from different absolute levels (Figure A4). Furthermore, overlap decreases considerably in two long periods in finance (1945–84 and 1985–2014), interrupted only by a sudden increase in overlap in between. In all issue areas, except for the environment, overlap is lowest in the most recent period. Distinguishing IGOs by geographical scope, we find the same decline of overlap for regional IGOs and global IGOs (Figure A5). For regional IGOs, the drop in overlap occurs in 1955–64, while for global IGOs, the drop occurs in 2005–14. Finally, for interventionist IGOs, overlap patterns do not follow the general pattern. Instead, we find an increase in overlap in 1995–2004, the most recent period with available data (Figure A6).

Together, these descriptive findings suggest that overlap among IGOs has declined over the past decades. As Figure A8 in the online appendix shows, this decline is not driven by the creation of new organizations with increasingly larger memberships, broader governance task portfolios, and larger policy scopes. This finding questions case study scholarship which portrays overlap among IGOs as increasing (Gehring & Faude, 2014; Gomez-Mera, 2015; Raustiala & Victor, 2004) but resonates with more recent studies which find that states may not always desire more overlap and that IGO overlap may in some issue areas have been more pronounced in the past than today (Eilstrup-Sangiovanni, 2022; Fioretos, 2021). Given that these case studies often focus on prominent organizations, such as the WTO (Gehring & Faude, 2014), NATO (Hofmann, 2009), the EU (Hofmann, 2011), or the International Monetary Fund (Henning, 2017), their finding of increasing institutional overlap is likely to be driven by the focus on prominent cases of major, interventionist organizations rather than reflecting a more general trend in the larger population of IGOs. This highlights the importance of systematic measurement and the use of data that includes different types of IGOs for advancing our understanding of institutional overlap in global governance.

Our substantive control variables are located at different levels of analysis and reflect commonly-discussed alternative accounts of institutional design in world politics.

At the dyad level, we always control for the number of shared issue areas between IGO i and IGO j. This captures the main rival explanation for IGO design similarity based on rational design theory according to which similar cooperation problems and the resulting functional demands lead to similar institutional designs (Koremenos et al., 2001). In additional robustness checks, we also measure the number of shared sub-issues of two IGOs which takes into account the possibility that cooperation problems are more likely to be similar within narrower sub-issue areas.¹¹ Data on IGO issue areas and sub-issue areas comes from the COW IGO data (Pevehouse et al., 2021). Furthermore, we control for whether an IGO i is based in the same region as a pre-existing IGO j (Sommerer & Tallberg, 2019). Regional identifiers are exclusive, so that a given organization is located in either Africa, Asia and Oceania, the Middle East, Europe and North America, Latin America, or operates globally.¹² We code organizations that have members from more than one region as global and those that have members from only one region as based in that particular region. In additional model permutations, we include the difference in the number of major powers—the G7 countries, China, and Russia—between IGO i and IGO j as well as the policy distance of the member states of both IGOs. The latter is computed as the difference in the average ideal point estimates among the member states of both organizations based on member state voting behavior in the UN General Assembly (Bailey et al., 2015).

At the IGO level, we distinguish between covariates for IGO i and IGO j, with IGO i being the newly created organization and IGO j being a pre-existing organization. For IGO i, unless we include fixed effects for IGO i, we control for the (logged) number of member states, the size of the state system, and dummies for issue areas and governance tasks. Fixed-effect estimation is our default choice because it captures any unobserved confounder that does not vary across IGO i. Moreover, we control for decennial period effects in all models, capturing common trends in IGO design.

We measure design similarity and control variables at the foundational year of an IGO, which in our view provides the cleanest test of our arguments. Nonetheless, this choice is not without its challenges. One (albeit unlikely) challenge arises from evolving IGO designs that respond to changes in overlap.¹³ The other challenge is measurement error in overlap to the extent that existing IGOs expanded their memberships, policy scope, or governance functions over time. We are willing to accept these potential measurement errors because institutional variables are relatively slow-moving. Table A7 in the online appendix presents summary statistics for all dyadic variables included in our main analysis.

Table 1 presents our first set of results where different columns represent models with different sets of controls. We find a robust positive correlation between overlap and IGO design similarity. In substantive terms, based on model 1 we find that an increase in overlap by one standard deviation is associated with an increase in design similarity by up to 0.004—equivalent to 1.92 percent of a standard deviation of that variable (p < 0.001). For a change from minimum to maximum overlap, this effect would be 0.118 (95%-CI: 0.068–0.168)—equivalent to 53.5 percent of a standard deviation (p < 0.001). To illustrate this substantive effect, consider the case of three trade organizations: Common Market for Eastern and Southern Africa (COMESA), East African Community (EAC), and World Trade Organization (WTO). Despite overlap of governance tasks and issue areas, their overlap with the EU is zero (for COMESA and EAC) or small (for the WTO), due to non-overlapping membership. If EAC, COMESA, and WTO were formed exclusively by EU member states, design similarity would be 0.04 index points higher (95%-CI: 0.02–0.06). This finding is consistent with our argument that in their decisions about the design of new IGOs, states use the design of overlapping existing organizations that serve similar groups of states and perform similar functions in similar policy fields as templates when selecting the features of new organizations.

In terms of our substantive control variables, we find a positive relationship between the number of shared issue areas between IGO i and IGO j and their design similarity. Since our key predictor of interest remains significant, this means that results are not driven by IGO pairs with many common issue areas. Being based in the same geographical region is not associated with design similarity. By contrast, we find some characteristics of the pre-existing IGO to be relevant: Design similarity is consistently higher with less recently established IGOs. Including additional dyadic covariates, we find that a greater difference in the number of major powers in both IGOs increases their design similarity. Specifically, if the new IGO is created primarily by major powers, while the legacy IGO is not, design similarity increases. Conversely, new IGOs created by lesser powers in response to powerful legacy IGOs are more dissimilar. Furthermore, a greater difference in the average ideal points of the members of both IGOs is not significantly associated with less similar designs.

To test whether designers consider the success of previous overlapping organizations when using them as blueprints, we divide the institutional context within which a new IGO is crafted in successful and unsuccessful organizations. For any given new IGO, a successful prior IGO is one that survived up until the year of its creation, whereas an unsuccessful prior IGO died before that. We use the survival of previously created IGOs as a fundamental and for IGO designers highly visible indicator of the general effectiveness of prior institutional solutions (Gray, 2018). Table 2 shows the determinants of IGO design similarity separately for the two sets of pre-existing organizations. We find a strong positive relationship between institutional overlap and design similarity in the subset of successful IGOs (p < 0.001). In contrast, the relationship between overlap with unsuccessful pre-existing organizations is at best marginally significant. In terms of effect sizes, an increase in overlap with successful pre-existing IGOs increases design similarity by at most 0.139 (p < 0.001) from minimum to maximum—or 61.6% of its standard deviation. Conversely, a similar increase in overlap with unsuccessful previous IGOs is related to an increase in design similarity by at most 0.072—a substantively smaller and statistically only marginally significant effect (p < 0.1). This implies that successful pre-existing IGOs provide an informative template for future design projects under institutional complexity, while there is no consistent relationship between overlap and design similarity with previous organizations that died.

In the appendix, we probe robustness of this finding using a coarser variant of the measure of unsuccessful IGOs. Following Vabulas and von Borzyskowski (2022), we consider as unsuccessful only those prior IGOs that died because they ceased operations, and not those prior IGOs that were replaced by other IGOs or that merged with other IGOs to form a new organization. Our main takeaways are unchanged: We find that overlap is significantly positively related with IGO design similarity in the sub-sample of successful IGOs, but unrelated in the sub-sample of unsuccessful IGOs (Table A8).

We now disaggregate overlap into two components—membership overlap (M_ij) and governance task − issue area or functional overlap (G_ij ⋅ I_ij)—which provides insights into how institutional complexity is related to IGO design convergence. Table 3 shows the results. We find a robust positive and statistically significant correlation between functional overlap and IGO design similarity. In substantive terms, an increase in functional overlap by one standard deviation increases design similarity by up to 0.012 (p < 0.001)—about 5.4 percent of its standard deviation. Membership overlap is also positively related to design similarity, but this result is not consistently significant across model specifications. This suggests that the convergence of institutional designs between overlapping new and pre-existing IGOs is possibly more driven by indirect rather than direct information transmission between a new cooperative effort and the context within which it is embedded.

Next, we examine the effect of overlap on IGO design similarity across issue areas, focusing on trade and commerce, environment, development, and finance. Tables A9 and A10 in the online appendix report the results. We see in Table A9 that overall overlap is positively related to design similarity across issue areas, although this relationship is most significant in trade and in development, followed by the environment. Table A10 further shows that functional overlap has a positive and significant association with design similarity across issue areas, while membership overlap is not consistently significant across issue areas. These findings provide additional evidence that design convergence between overlapping IGOs is occurring more through indirect information flows than direct information flows based on shared members.

To probe the robustness of our findings, we use three alternative operationalizations of overlap. First, we calculate overlap as the product of membership and governance task overlap conditioning on two IGOs having at least one issue area in common. This measure could be considered an alternative operationalization of the conceptualization of institutional overlap within regime complexes which focuses on shared functional domains and memberships for organizations that are involved in governing a particular issue area (Raustiala & Victor, 2004). We do not find any marked differences in our results when using this alternative measure (Table A11).¹⁵ Second, we replace our measure of membership overlap with the overlap of major powers and compute overlap as the product of major power overlap, issue area similarity, and governance task similarity of pre-existing and new IGOs.¹⁶ Again, our results do not qualitatively differ when using this alternative operationalization, although precision is lower than before (Table A12). Third, we are concerned that our results could be affected by the skewness of the overlap measure. Hence, we re-run the analysis after taking the natural logarithm of overlap (with an offset of one to avoid missing values). Our results are unaffected (Table A13).

To demonstrate the significance of our contribution of introducing governance task overlap, we show how conclusions drawn from focusing only on membership overlap and issue overlap may be misleading. To that end, we augment our baseline regression with both membership overlap and issue area overlap. Using three overlap measures, we find that only our comprehensive measure of overlap which incorporates all three dimensions is statistically significant in the hypothesized direction. While membership overlap has no relationship with IGO design similarity, the coefficient of issue area overlap is significantly negative (Table A14).

We also estimate our models using alternative operationalizations of our dependent variable. First, we drop codified decision-making procedures as design feature, because decisions about how to calibrate this design dimension may follow a different logic than for monitoring, enforcement, and dispute settlement (Hooghe & Marks, 2015; Koremenos & Betz, 2013; Koremenos et al., 2001). Re-running our analysis with the resulting similarity measure based on four items, we find qualitatively similar results (Table A15). Second, some may argue that the structure of IGO decision-making procedures—rather than whether such procedures are formalized at all—may be conceptually more pertinent. Hence, we source data on voting procedures in the main bodies of IGOs (Blake & Payton, 2015). We consider whether an IGO uses non-unanimous voting, which reflects greater pooling (Hooghe & Marks, 2015). Using this alternative operationalization of the decision-making dimension in our five-element design vector, we corroborate our earlier findings (Table A16). Third, we drop the design feature for monitoring, addressing the potential concern that monitoring as a design feature and monitoring as a governance function are conceptually too close. To be sure, both variables are operationalized differently and capture different aspects of IGOs. Nevertheless, we estimate our models excluding monitoring from our design similarity measure. We corroborate our core findings (Table A17). Finally, we use a different distance metric—the Jaccard index—to compute design similarities. Intuitively, the Jaccard index captures how many design features are identical between two IGOs, as a proportion of the number of design features. Here we consider both the common absence as well as the common presence of a design feature. The Jaccard index has been used in prior related work (Jetschke et al., 2021), and we find that it is highly correlated to ours. Hence, the results are qualitatively unaffected (Table A18).

Next, we address concerns that our results are driven by individual matching design features. To that end, we use five alternative dependent variables, capturing whether two IGOs both have adopted a given design feature. In each regression, we compare pairs of IGOs which both have the design feature to those pairs of IGOs in which only one IGO has the design feature. We find that the relationship between overlap and design similarity is significantly positive for all design features, although it is only marginally significant for secretariats (Table A19). In further analyses, we find significantly positive relationships between governance task overlap and matched design for all five design features (Table A20).

In the following robustness tests, we address methodological concerns. To begin, we replace ten-year period dummies with five-year period dummies, which is appropriate if common design trends change more frequently than every ten years (Table A21). We also use year-fixed effects instead of ten-year period dummies (Table A22). None of these alternative choices affect our core result.

Next, we probe whether our main results hold using imputed data (Honaker, King, and Blackwell 2011). With non-imputed data, loss of observations can be substantial. This makes our results less comparable across models and could induce bias, given that missingness in covariates is likely to be more prevalent in earlier years. Using imputed data on our control variables addresses these concerns (Lall, 2016). Our main takeaways are the same. We continue to find a strong relationship between institutional overlap and design similarity (Table A23). Furthermore, we find that this relationship holds strongly with respect to successful prior IGOs, but not unsuccessful prior IGOs (Table A24).

To confront concerns about selection bias, we estimate a Heckman-type model that estimates IGO creation in addition to IGO design. Following Reinsberg and Westerwinter (2021), we create a set of proposal IGOs based on the empirical distribution of member states, issue area profiles, and governance task profiles. This induces variation in the observability of IGO dyads. This analysis confirms all results from our single-stage models (Table A25).

In another set of robustness tests, we show that our argument holds against further alternative explanations of IGO design similarity. As these alternatives require contextual measures that do not vary across dyads for the same organization, we remove the fixed effects for IGO i and instead include additional indicator variables for issue areas and governance tasks for IGO i. As we cannot do justice to all these alternatives, our proxies are relatively crude, but the results are indicative. World polity theory would expect a global trend towards a single dominating organizational script, implying IGO design convergence across world regions (Boli & Thomas, 1997). To test these expectations, we add region dummies. Theories of organizational ecology expect the density of a particular type of organization in an issue area to affect the design of new organizations (Abbott et al., 2016; Freeman & Hannan, 1977; Morin, 2020). Hence, we add the logged number of prior IGOs in the same issue area profile, along with the logarithm of the squared number. In addition, gridlock theories would expect lower IGO design similarity as the heterogeneity of the states in the international system increases (Hale et al., 2013). Hence, we add the average coefficient of variation of the polity scores of all states in the foundational year of an organization. Finally, arguments about organizational progeny suggest that IGO bureaucracies can affect IGO design decisions and replicate design templates for new creations (Johnson & Urpelainen, 2014). To explicitly account for such possibility, we draw on new data indicating whether an organization is an emanation of an existing organization (Jud, Westerwinter, & Wright 2022). In all these tests, we continue to find a strongly robust positive relationship between overlap and IGO design similarity (Table A26).

Finally, we address potential inaccuracies in the COW IGO dataset. Based on new data that identify various potential shortcomings of the COW IGO data (Jud, Westerwinter, & Wright 2022), we exclude cases that do not meet the criteria highlighted in the COW IGO definition because they are not independent of other IGOs (type I), are subsidiary bodies or special agencies of other IGOs (type II), or are emanations of other IGOs (type III) (Pevehouse et al. 2020). We also exclude cases that raise questions regarding a possible merging of two previously separate organizations (type IV). These inaccuracies are problematic for our analysis because they potentially distort our empirical assessment of the relationships between some of the IGOs in the data which, in turn, may affect our overlap measure as well as our design similarity variable. In total, we identify 130 IGOs that could be problematic in one or more of these four ways and re-run our main models without the affected 34,938 IGO dyads. Table A27 provides an overview of the excluded cases. Table A28 shows that our findings are qualitatively unaffected when using this more restrictive sample.

Having established robustness of our main findings, we conduct additional analyses. Using split-sample analysis, we explore effect heterogeneity. First, we divide the sample into regional IGOs and inter-continental IGOs. The results are not different between inter-continental IGOs and regional IGOs (Table A29). Second, we probe effect heterogeneity with respect to whether IGOs have ambiguous mandates that cover more than a single policy domain. Organizational isomorphism research emphasizes that imitation of institutional designs of legacy organizations by new entrants is particularly pronounced in situations of ambiguity and uncertainty (DiMaggio & Powell, 1983; Kalyanpur & Newman, 2017). Table A30 shows that the substantive effect of overlap on design similarity is stronger for multi-issue IGOs. Third, when splitting the sample at the median number of governance tasks, we find that the positive relationship between overlap and design similarity is significant only for IGOs with an above-median number of governance tasks—for organizations with high levels of governance task ambiguity (Table A31). These sets of findings resonate with our argument that in uncertain and information poor environments the informative value of institutional overlap for designers of new IGOs is especially pronounced. Fourth, we examine whether the relationship between overlap and design similarity varies between IGOs that include major powers (G7 countries) and those that do not, without finding a sizeable difference (Table A32). Finally, we assess whether overlap is differently related to design similarity in different types of international systems, particularly focusing on the Cold War and post-Cold War periods (Table A33). We find the effect to be more significant during the Cold War.

These additional findings corroborate the results of our main analysis and identify institutional overlap as an important factor that shapes IGO design convergence across different types of IGOs, scales of cooperative endeavors, with more and less ambiguous mandates, IGOs with and without major power members, and historical periods.