Greening global governance: INGO secretariats and environmental mainstreaming of IOs, 1950 to 2017

Abstract

The last decades have seen a remarkable expansion in the number of International Organizations (IOs) that have mainstreamed environmental issues into their policy scope—in many cases due to the pressure of civil society. We hypothesize that International Non-Governmental Organizations (INGOs), whose headquarters are in proximity to the headquarters of IOs, are more likely to affect IOs' expansion into the environmental domain. We test this explanation by utilizing a novel dataset on the strength of environmental global civil society in proximity to the headquarters of 76 IOs between 1950 and 2017. Three findings stand out. First, the more environmental INGOs have their secretariat in proximity to the headquarter of an IO, the more likely the IO mainstreams environmental policy. Second, proximate INGOs’ contribution increases when they can rely on domestically focused NGOs in member states. Third, a pathway case reveals that proximate INGOs played an essential role in inside lobbying, outside lobbying and information provision during the campaign to mainstream environmental issues at the World Bank. However, their efforts relied to a substantial extent on the work of local NGOs on the ground.

Introduction

The last decades have seen a remarkable expansion of International Organizations (IOs) into the environmental domain. While no IOs considered environmental policy as an integral component of their work in the 1950s and 60s, more than half of (our sample of 76) IOs have mainstreamed environmental issues since then. Environmental mainstreaming is part of a broader phenomenon, where IOs have increased their policy scope substantially over time (Hooghe et al., 2019; Tallberg et al., 2020; Johnson, 2015; Dalal-Clayton & Bass, 2009; da Conceição-Heldt & Dörfler, 2021).

A rich literature has developed that explains such environmental mainstreaming. Some scholars have posited the importance of the interests of member states (Nielson & Tierney, 2003), the initiatives of IO bureaucracies (Hall, 2016; Luken, 2009; Pollack & Hafner-Burton, 2010) or diffusion between IOs (Ovodenko & Keohane, 2012). However, most attention has been paid to the importance of environmental International Non-Governmental Organizations (INGOs) in specific cases of IO expansion (Wade, 1997; Nelson, 1997; Park, 2005a; Keck & Sikkink, 1998). Civil society actors have been crucial in driving IO engagement in other domains, like gender inequality (Hafner-Burton & Pollack, 2002; Joachim, 2003), water as a human right (Reiners, 2022), or women, peace and security (Shepherd, 2008). Scholars have greatly contributed to our understanding of the mechanisms through which INGOs can press IOs to expand their policy scope. Nevertheless, the focus on single case studies means that we lack an understanding of the extent of INGOs’ influence on IO scope expansion over many IOs and years. Consequently, we ask: To what extent do environmental INGOs affect the environmental mainstreaming of IOs?

Our paper makes three contributions: first, we develop an argument on the role of INGOs for the environmental mainstreaming of IOs. We advance this argument in two steps. First, based on discussions in the literature on non-state influence in IOs, we argue that INGOs can drive change at IOs mainly through inside lobbying, outside lobbying and information provision. We further qualify that INGOs can affect change better when their secretariats are located in geographical proximity to IO headquarters. In a second step, relying on the literature on advocacy coalitions, we maintain that these INGOs will be even more influential when they can draw on domestic NGOs in member states that supply them with relevant information about national contexts.

Second, we test these hypotheses across 76 IOs active between 1950 and 2017. This large-n comparative approach allows us to understand the overall importance of proximate INGOs for explaining the environmental mainstreaming of IOs. Our quantitative approach also allows us to ascertain whether INGO proximity remains a critical factor when controlling for other kinds of civil society—like domestic NGOs and INGO members. The quantitative analysis yields robust support for our expectation that proximate INGOs have played an important role in IO environmental mainstreaming. Furthermore, the association seems to have been magnified by relying on a strong environmental civil society within the member states of IOs.

Third, we show how INGO proximity can affect INGOs’ ability to influence IOs to mainstream environmental issues through a pathway case—the campaign of environmental civil society to “green” the World Bank. We demonstrate that US-based INGOs shaped the campaign’s influence through three pathways: (a) US-based INGOs were able to leverage their contacts with member state representatives and World Bank staff to lobby them directly (inside lobbying). (b) they could draw on their links to prominent US newspapers to ramp up the public pressure on the World Bank (outside lobbying). (c) US-based INGOs could establish themselves as experts for World Bank projects’ environmental impact in project areas, allowing INGOs to speak with authority on the issue (information provision). However, in each of the three mechanisms, they relied extensively on their contacts with NGOs based in recipient countries that provided them with information on matters on the ground and moral authority in US congressional hearings. The case also shows how proximity affects inequality of stakeholder influence in global governance. Proximate INGOs were able to dictate the focus of the campaign over the interests of locally-based NGOs. Therefore, our findings have important implications for debates over the participation of stakeholders in IOs (Bäckstrand, 2006; Tallberg et al., 2013), the location of organizations and its impacts on access, influence, effectiveness, and equity (Johnson, 2014; Ivanova, 2021, 2007).

The article proceeds in four steps. First, we outline our theoretical explanation for IO scope expansion into the environmental domain—the proximity of environmental INGOs secretariats to IO headquarters. Second, we introduce our dataset, research design, operationalization, and measurement of our variables. Third, we explore the empirical explanatory potential of our variable of interest. Fourth, we probe the importance of proximate INGOs through the pathway case of civil society’s importance in driving environmental mainstreaming at the World Bank. Finally, we summarize the main findings, discuss the limitations and outline the broader implications of our findings for the literature on the location and its impacts on access and influence as well as for the comparative IO literature.

The impact of proximate INGOs on environmental mainstreaming of IOs

We define environmental mainstreaming as “the integration of environmental objectives into non-environmental sectors” (Nunan et al., 2012: 263; also Dalal-Clayton & Bass, 2009). The concept refers to instances in which IOs that are not primarily focused on environmental policy include environmental issues in their policy scope (Hooghe et al., 2019: 53, also Koremenos, 2016: 44). We argue that INGOs¹, whose secretariats work in proximity to IOs, drove this process. We establish our argument in reference to the existing theory on civil society impact on IOs. Approaches on the influence of actors vary among two dimensions: how influence is yielded and who influences. We discuss both dimensions in turn.

The literature has provided comprehensive answers on how NGOs can drive change at IOs (Tallberg et al., 2018; Steffek, 2013; Park, 2005a; Grigorescu, 2007; Johnson, 2016; Dür & Mateo, 2012). Such impact is said to materialize for three primary reasons: inside lobbying by NGOs, outside lobbying by NGOs and informational requirements of IOs (Tallberg et al., 2018). First, inside lobbying refers to advocating for policy change directly with decision-makers (Dür & Mateo, 2013). For instance, Park (2005a) examines how transnational advocacy networks reconstituted the identity of the International Finance Corporation (IFC) of the World Bank stressing the Pangue Dam project in Chile as an example of “problem projects”. She found that NGOs were successful in environmental mainstreaming IFC through socialization via direct NGO-IFC interactions and via indirect network-member state-IFC interactions.

Second, when NGOs mobilize public opinion, they engage in outside lobbying (Dür & Mateo, 2013). By building public support for their cause, NGOs can incentivize decision-makers to pursue policy change (Keck & Sikkink, 1998). Outside lobbying includes campaigning, protesting, rallying, inducing boycotts or civil disobedience to ramp up the pressure on decision-makers (Gulbrandsen & Andresen, 2004). For instance, Rietig (2016) explores how and under which conditions government representatives paid attention to NGO contributions during the post − Kyoto Protocol negotiations. She demonstrates that public opinion mobilization through large-scale demonstrations over a sustained period can open up the political leeway to increase climate ambitions and incentivize governments to reconsider their positions.

Third, NGOs can generate and provide specialized, issue-specific or locally-generated information of great value for the work of IOs (Tallberg et al., 2018: 215–217). NGOs are highly specialized in an issue area relevant to their cause. As a result, they are key sources of knowledge used by IOs in decision-making. The information they provide can illuminate policy options and the consequences of different policy choices. NGOs can also elucidate the views of an IO’s societal stakeholders and may serve as vital conduits between IOs and civil society (Tallberg et al., 2018: 216–217). For instance, Betsill and Corell (2008) show that NGOs yield influence by helping decision-makers navigate the highly complex and technical nature of many environmental issues.

While the reasons for NGOs’ influence are well explored, we know much less about which NGOs can affect change at IOs (‘who influences’ dimension). We argue that INGOs gain in their ability to affect change at IOs when their secretariats are located in geographical proximity to IO headquarters. The argument is based on debates on lobbying, NGOs’ domestic impact and IO diffusion.

Studies on the effects of lobbying in the EU have underlined the relevance of spatial access of interest groups for their ability to influence (Hermansson, 2016; Egdell & Thomson 1999; Biliouri, 1999). For example, Klüver (2010: 179) claims: “a physical presence in Brussels has proved to be useful“ for interest groups lobbying European institutions. Hermansson (2016) argues that being in Brussels allows interest groups to build the necessary expertise to affect EU policy. Egdell and Thompson (1999: 128) argue that proximity to the EU headquarters allows interest groups to gain an advantage “partly for convenience and flexibility in arranging and attending meetings, and partly in terms of the casual contacts that occur at third-party events”.

Moreover, proximity has also been discussed in the literature on NGOs’ domestic impact. For example, Murdie and Bhasin (2011) show that the presence of human rights NGOs in a country increases the likelihood of protest. Similar effects also materialize in proximity to NGOs secretariats across state borders (Bell et al., 2012). Furthermore, differences of the environmental and human security impact of NGOs depend on their ability to engage freely in public debate (Murdie, 2014; Pacheco-Vega & Murdie, 2021). These findings imply that the domestic environment in which INGOs operate matters for both their reach and ability to affect change (Stroup & Murdie, 2012). INGOs themselves seem to be aware of these factors when deciding where to place their secretariats. They try to minimize the distance to actors they seek to affect while allowing for operations to run smoothly at the same time (Barry et al., 2015). Similar arguments have been made in the literature on IO diffusion and connectivity. As Tallberg and Sommerer (2019: 404) argue: co-location “(…) entails greater opportunities for informal interaction” (Sommerer & Tallberg, 2019: 407).

Based on these insights, we argue that environmental INGOs proximate to IOs will be better able to affect change in their policy scope than those further away from them. Many IO headquarters cluster in cities like New York, Washington DC, Geneva, Brussels, London, Vienna, Paris, or Bonn while other IOs are based in countries where few other IOs have their headquarters (Grigorescu, 2010: 877). The 76 IOs in our dataset are located in 49 cities and 38 countries. If INGOs secretariats are located in close proximity, they have advantages in inside lobbying, outside lobbying and information provision.

First, inside lobbying requires direct access to decision-makers. INGOs have more readily available informal channels of influence as they interact with representatives from IOs headquartered in the same city (Grigorescu, 2010: 877). As Sommerer and Tallberg (2019: 407) argue for the proximity of IOs: “As IO staff meet and discuss in social and political arenas, experiences and norms travel more easily”.

Second, INGOs can try to campaign, protest, rally, induce boycotts or civil disobedience to ramp up the pressure on and shame decision-makers from the outside (Gulbrandsen & Andresen, 2004). IO officials and member state representatives often live and work around IO headquarters. These cities are often characterized by media bubbles of officials, interest groups and journalists that discuss the specific issues the IO is dealing with. Proximate INGO secretariats have better access to these media bubbles than secretariats working far away.

Third, proximate secretariats can also give INGOs an advantage in information provision. They can establish themselves as go-to experts for IO staff and member state representatives that seek out information. Furthermore, they need fewer resources to appear in IO meetings and consultations than INGOs located further away. Therefore, proximity likely is a considerable advantage for the ability of INGOs to lobby decision-makers directly, to gain better access to media bubbles surrounding IOs and to establish themselves as information providers. This leads us to the following hypothesis:

• H1: The more environmental INGOs with secretariats are in proximity to a given IO, the more likely the IO mainstreams environmental issues in its policy scope.

We further posit that the influence of proximate INGOs interacts with domestically focused NGOs in member states because they can form advocacy coalitions. While INGOs in proximity to IO secretariats have superior access to IOs, they often lack crucial information from the member states of IOs. INGOs need to build advocacy coalitions with NGOs focused on member states’ national context to obtain such information. An advocacy coalition comprises “actors working internationally on an issue, who are bound together by shared values, a common discourse, and dense exchanges of information and services” with the goal “to change the behavior of states and of international organizations” (Keck & Sikkink, 1998: 2). Prominent examples are advocacy coalitions in the field of disarmament (Price, 1998), human rights (Keck & Sikkink, 1998), World Bank environmental reform (Park, 2005b) or negotiating the International Criminal Court (Deitelhoff, 2009). Coalitions or networks of NGOs allow for individual groups to collaborate, join forces to build momentum and wield more influence than what would be possible for any individual NGO (Tallberg et al., 2018: 218–219). In this context, “advocacy networks often reach beyond policy change to advocate and instigate changes in the institutional and principled basis of international interactions” (Keck & Sikkink, 1998: 2). The argument that coalitions can advocate more effectively than NGOs in isolation is also well established in the domestic and EU interest group literature (Baumgartner, 2009). For instance, Mahoney (2007) argues that domestic interest groups join ad hoc coalitions to signal to decision-makers that a policy position has broad support and to more efficiently use their resources. Klüver (2013: 18–19) reasons that “grouping of interest groups into lobbying coalitions is the decisive point (…) [as] the likelihood that interest groups can influence the policy-making process increase with the aggregated amount of information, citizen support, and economic power provided by their lobbying coalitions”.

INGOs can draw on NGOs in member states to obtain necessary information. Keck and Sikkink (1998: 16) make this argument by introducing the concept of “information politics”, i.e. “the ability to quickly and credibly generate politically usable information and move it to where it will have the most impact”. At the heart of the INGO-NGO relationship is the exchange of information (Keck & Sikkink, 1998). National NGOs know the work of IOs in their country context. INGOs in proximity to the IO can leverage their location to translate such information in their interactions with member state representatives and IO officials. For instance, Nelson (1997) discusses the interaction of Washington-based INGOs with NGOs from the Global South that lobby the World Bank and finds that better communication has helped improve the effectiveness of INGO actions. Hence, we can expect that INGOs forming an advocacy coalition with national NGOs are particularly likely to influence the environmental mainstreaming of IOs. This leads us to the following hypothesis:

• H2: The association of proximate environmental INGOs with environmental mainstreaming of IOs increases with the number of domestic environmental NGOs among member states.

Research design

To ascertain the association of INGO proximity with IO environmental mainstreaming, we estimate several linear probability models that regress a binary variable indicating whether an IO has mainstreamed environmental issues into its policy scope on the number of INGO secretariats in proximity to the IOs headquarter. The dataset has a panel structure and the unit of analysis is the IO-year. All models include IO fixed effects to control for heterogeneity among IOs and (in some specifications) year fixed effects and/or IO-specific time trends to control for common shocks and the substantial expansion of transnational environmentalism over time (Hale, 2020; Falkner & Buzan, 2019). We use the Huber-White correction to account for heteroscedasticity introduced by using OLS for binary response variables. We employ linear probability models because conditional logit models would require excluding all IOs that never expand or have always been engaged with environmental policy. Doing so would omit theoretically relevant variation from the analysis. However, we run conditional logit, Poisson pseudo, two-stage-least-squares regression, a Cox proportional hazards model and a probit model with the Heckman correction as robustness checks (see below). In the following, we discuss the dependent, independent and control variables we employ in our models.

Dependent variable

For our analysis of policy scope expansion, we draw on the Measure of International Authority (MIA) dataset (Hooghe et al., 2019).² The MIA dataset includes policy areas of 77 IOs that were active between 1950 and 2017. In their dataset, Hooghe et al. (2019) choose IOs that fulfill the following criteria: a distinct physical location or website, a formal structure, at least 30 permanent staff, a written constitution, and a decision body that meets at least once a year. The sample is not representative of the larger body of IOs, which somewhat limits the generalizations that can be made from the sample. Nevertheless, we believe that the dataset is useful as it covers a variety of global, regional and multi-regional IOs that are active in different policy areas and is the most comprehensive dataset that includes data on IO policy scope. We exclude the Global Environment Facility because it is a core environmental IO (Ingram et al., 2005). That way, we arrive at the 76 IOs in our sample. Table A1 (Appendix) displays the headquarter locations of the IOs covered in the dataset.

Our primary dependent variable measures environmental mainstreaming as a binary variable. The MIA dataset codes whether IOs fulfill one of eight criteria.³ Our environmental mainstreaming variable takes the value of one if any of these criteria are fulfilled and of zero if none are fulfilled. Environmental mainstreaming relates to the inclusion of environmental concerns into the work of an IO. The binary variable has the advantage that we do not need to make a qualitative judgment on which engagement with environmental issues is meaningful. Such an approach is useful for an initial analysis of the broad patterns of environmental mainstreaming across many diverse IOs. Nevertheless, this choice cannot capture the qualitative differences and considerable diversity in the extent to which IOs engage with environmental issues.⁴ Figure 1 displays the number of IOs in the dataset and the number of IOs that have mainstreamed environmental issues. It shows that the general trend is towards the environmental mainstreaming of IOs.

Fig. 1

Number of IOs in the dataset (grey line) and number of IOs that have mainstreamed environmental policy (black line)

Independent variable

Our main independent variable is a count of environmental INGOs with their headquarters in the same country as a given IO. The literature on INGO strength has used two approaches to measure the presence of INGOs. Most studies utilize INGO membership data to operationalize INGO presence (Pacheco-Vega & Murdie, 2021; Longhofer et al., 2016; Baccini et al., 2022). However, others have used the secretariat location (Murdie & Bhasin, 2011; Barry et al., 2015). Our argument focuses directly on interactions between the secretariats of IOs and INGOs. Therefore, we employ the secretariat measure. Data was coded in a two-step process based on the Yearbook of International Associations of the Union of International Associations (Union of International Associations, 2021). The UIA is the most common source used in the literature on INGOs (Murdie, 2014; Pacheco-Vega & Murdie, 2021; Longhofer et al., 2016; Longhofer & Schofer, 2010). First, we identified INGOs working on environmental issues by extracting those INGOs with the subject code “environment” in the yearbook. Second, we coded the location of secretariats based on their address listed. A given INGO was included in the count variable for each year it was active after its founding. One caveat with yearbook data is that it is not a census of all INGOs active in all countries. Therefore, the data cannot be taken as an exact count of the relevant INGOs in a country. However, the literature on the influence of NGOs has shown that the data helps to sort countries according to the strength of their respective (environmental) civil societies (Longhofer & Schofer, 2010; Longhofer et al., 2016).

Control variables

We employ several control variables to ensure that correlations are not driven by heterogeneity among IOs that is unrelated to our theoretical explanations. We control for variations on the IO-level, member-state level, and INGO-level that could threaten proper inference from our models. First, we include IO-level variables that have been discussed in the literature on IO scope expansion (Hooghe et al., 2019). Hence, we control for differences in pooling and delegation. Pooling refers to joint decision-making among member states and delegation to the grant of competencies to the secretariat (Hooghe & Marks, 2015). Data on pooling and delegation stems from Hooghe et al. (2017). The third IO-level control variable is the diffusion among member states of IOs with headquarters in the same country (Sommerer & Tallberg, 2019). Diffusion among IOs has been a prominent explanation in the literature on IO scope expansion (Tallberg et al., 2020). The main threat to inference from diffusion processes for our models is due to the co-location of IOs that might affect both the environmental mainstreaming of other IOs and the strength of proximate environmental INGOs. Hence, we try to ensure that our results are not driven by diffusion among proximate IOs rather than INGOs by including a spatial lag of environmental mainstreaming, where connectivity is defined by co-location in the same country.

Additionally, our data shows a secular and monotonic increase in mainstreaming over time. While reversals are theoretically possible because IOs can reform and exclude certain policy areas from their policy scope, the non-occurrence of reversals might be problematic for theoretical and methodological reasons. Theoretically, IOs that have mainstreamed environmental policy may be influenced by path dependency and, hence, find it more difficult to abandon a focus on environmental mainstreaming. Methodologically, the lack of reversals may lead to issues of pseudo-replication. To account for both arguments, we use IO-specific time trends in some specifications and present robustness checks using a cross-sectional model, an alternative dependent variable counting the number of environmental features and control for a lagged dependent variable (see Section 5 below).

Second, we control for four variables operationalizing arguments on the heterogeneity among member states. First, we include a measure of the preferences of member states for global environmental governance to ensure that common exposure to environmental INGOs and IOs does not drive the results (Panke, 2020). Here, we take a given country’s membership in environmental IOs for each year based on Ingram et al. (2005). Second, we control for the average level of democracy of the member states because democracies are more likely to participate both in environmental agreements and IOs and also comply more often with environmental agreements (Neumayer, 2002). Data on democracy is taken from the Varieties of Democracy Project (Coppedge et al., 2021). Third, we include a measure of the average level of GDP per capita of member states because member states’ income may influence both the costs that member states can bear and whether they can afford to invest in the mitigation of environmental degradation. This variable is also used in the national environmental performance literature (Bernauer & Böhmelt, 2013). Data on GDP comes from the Penn World Tables (Feenstra et al., 2015). Finally, we follow common practice in the literature on the institutional design of IOs and control for preference heterogeneity because reforms of IOs among heterogeneous principals are more complicated than reforms among homogenous principals (Graham, 2014). The indicator is based on the distance to the median UNGA ideal point (Bailey et al., 2017).

Finally, we incorporate additional control variables to ascertain whether results are based on proximate INGOs and not—likely correlated—proximate national NGOs or INGO membership measures. Of course, these results need to be interpreted cautiously due to multicollinearity. Therefore, we include these variables only in a second step to ensure that multicollinearity does not threaten inference overall. We display information on the sources (Table A2) and coverage of our main covariates in the appendix (Table A3).

A comparative analysis of the importance of proximate INGO secretariats in environmental mainstreaming of IOs

Table 1 displays the results from the estimations of five models focusing on the role of proximate INGOs in environmental mainstreaming of IOs. All regressions include IO fixed effects. In Model 1, we only incorporate our variable of interest. In Model 2, we further control for the two institutional design features of IOs (pooling and delegation) and alternative explanations for IO environmental mainstreaming (member state preferences, diffusion, income, democracy, and member state preference heterogeneity). We also employ IO-specific time trends in Model 3 by interacting each fixed effect with the year of interest to account for the potential issue of pseudo-replication. Model 4 includes year fixed effects to account for common shocks (like environmental summits). Finally, Model 5 presents the most stringent specification because it incorporates IO fixed effects, year fixed effects and IO-specific time trends.

Table 1 Regression models of proximate INGO influence

The results imply that proximate INGOs seem to play a critical role in the environmental mainstreaming of IOs. The coefficient is positive and statistically significant at the most demanding conventional level (p < 0.001) in all models displayed in Table 1. This is strong evidence that INGOs with proximate secretariats are associated with the environmental mainstreaming of IOs. The coefficients are unstandardized, which means that they imply the conditional increase in the probability that a given IO has mainstreamed environmental issues with each additional INGO with a proximate secretariat.

How strong is the association of proximate environmental INGOs with environmental mainstreaming of IOs according to these models? An additional INGO increases the likelihood of environmental mainstreaming by around 0.2%. Since our models include IO fixed effects, one can only grasp the magnitude of the association in relation to within-unit changes in environmental INGOs. For example, our data records an increase of 197 environmental INGOs for the World Bank between 1950 and 2017. Such an increase would imply that, on average, the likelihood for environmental mainstreaming would increase by 45% from 1950 to 2017 based on the coefficient reported in Model 4. The World Bank is among the organizations faced with the strongest increase in proximate environmental INGOs in the data. The average within-IO increase of environmental INGO over the period of interest is 29. Such an expansion of environmental INGOs would be associated with a more modest increase in the likelihood of environmental mainstreaming by, on average, 5.8%. Hence, the association between proximate environmental INGOs and environmental mainstreaming is moderate but substantively important.

To decrease the chance that we are over-interpreting a spurious correlation, we now present evidence from a placebo check. The rationale behind a placebo check is to compare the results from an analysis where we expect a theoretical relationship with the results from an analysis where we do not expect such a relationship that relies on data generated through a similar process. To do so, we run simple linear probability models with IO fixed effects, control variables and the proximate INGO variable. However, we swap the dependent variable for each model to focus on different policy areas within the IOs’ policy scope. For instance, we look at the association of proximate environmental INGO secretariats with an IO having security policy within its policy scope. Figure 2 displays the results from ten models. The first regression focuses on environmental mainstreaming. Subsequent models concentrate on nine additional policy areas (agriculture, education, finance, foreign policy, health, human rights, justice, security, trade) where we do not expect an influence of environmental INGOs. While environmental INGOs are related to environmental mainstreaming, they do not have a positive and statistically significant association with any of the other nine policy areas. The results increase our confidence that the correlations are meaningful, and a relationship between proximate INGO secretariats and environmental mainstreaming can be inferred from the data.

Fig. 2

Placebo check of the impact of proximate environmental INGOs on IO policy scope expansion in ten policy areas (95% confidence intervals). Note: The figure is based on ten linear probability models (Model 6 to Model 15) with IO fixed effects and Huber-White standard errors that regress the different policy scope variables on proximate INGO secretariats, diffusion, member state preferences, delegation, pooling, average income of member states (log), average democracy among member states, member state heterogeneity. Visualizations use the plotplain package (Bischof 2017)

Exposure of IOs to different kinds of civil society are likely correlated. However, our argument focused on one specific type of exposure: connectivity to proximate environmental INGOs. To ensure that other types of exposure do not drive results, we now re-estimate model 4 but control for five additional ways in which IOs can be exposed to environmental civil society. Data on environmental civil society can essentially vary on three dimensions: focus, location, and exposure type. First, focus refers to the target of the work of civil society. They can be either (a) domestically focused NGOs or (b) INGOs focused on transnational issues. Second, location refers to where civil society is sited. For our purposes, it can be located either (a) in proximity to the IO headquarters or (b) in the member states of an IO. Third, exposure type refers to which part of civil society IOs are exposed to. These are either (a) the secretariat of NGOs/INGOs or (b) the members of civil society organizations. Our argument posited the importance of the impact of INGOs (focus) whose secretariats (exposure type) are located in proximity to an IO headquarter (location). Table 2 displays the different kinds of civil society variables. Our variable of interest is presented in italics, and the correlations with our primary variable of interest are shown in brackets. The INGO membership and the NGO secretariat/membership variables have been kindly provided by Longhofer et al. (2016).⁵ Since we measure civil society strength on a country-level, we do not differentiate between the secretariats and membership of domestically-focused NGOs as they are in the same country.

Table 2 Dimensions of exposure to civil society

Table 3 displays the results for alterations of Model 4 that further include each kind of exposure to civil society. Of course, these models face multicollinearity issues as the different kinds of civil society variables are correlated to a substantial extent (see Table 2). The vif values of our core variable of interest, proximate INGO secretariats, are higher than the critical value of 10 in four of the five models. However, in Model 18 and 20 they are only slightly higher. Nevertheless, the presence of multicollinearity urges caution for the interpretation of the results.

Table 3 Proximate INGO influence controlling for other kinds of civil society exposure

Despite these cautionary comments, the results indicate that proximate INGO secretariats appear substantially more important than any of the other five kinds of IO exposure to environmental civil society. The coefficient for proximate INGO secretariats is statistically significant (p < 0.001) and positive in all five models. When holding proximate INGO secretariats constant, the additional information included in the other variables does not seem to be positively related to environmental mainstreaming of IOs⁶. Overall, the results further underline the relevance of our main variable of interest. Indeed, the results seem to be driven by the relationship between proximate INGO secretariats and environmental mainstreaming rather than any other kind of civil society exposure. Together, the results presented so far indicate that we can reject the null hypothesis that proximate INGO secretariats are unrelated to IO environmental mainstreaming (H1).

Do these INGOs rely, partly, on networks with domestic NGOs for this influence as implied in H2? We re-estimate Model 4 while interacting proximate INGO secretariats with domestic NGO secretariats to evaluate this question. Figure 3 displays the linear predictions of the interaction holding INGOs at their mean and increasing the average number of domestic NGOs among member states from 0 to 30 domestic NGOs. The figure shows that the null hypothesis can be rejected. There seems to be an interaction between the strength of proximate INGOs and the strength of domestic NGOs in member states. The association of proximate INGO secretariats with environmental mainstreaming almost triples when linked to an average of thirty compared to zero domestic NGOs among an IOs membership. These findings imply that proximate INGOs advocacy efforts to mainstream environmental issues seem to have relied substantially on networks with domestic environmental NGOs among the membership of IOs.

Fig. 3

Interaction between proximate INGO secretariats and MS NGOs. Note: The figure is based on a linear probability model with IO fixed effects, year fixed effects and Huber-White standard errors that regresses environmental mainstreaming on proximate INGO secretariats, average NGOs among member states, their interaction, diffusion, member state preferences, delegation, pooling, average income of member states (log), average democracy among member states, and member state heterogeneity (Model 21). Visualizations use the plotplain package (Bischof, 2017)

Robustness checks

We estimate several robustness checks to minimize the possibility of faulty inference. We test the robustness towards using an alternative dependent variable, independent variables and model specifications.

First, readers may be concerned about reverse causality. Our research design relies on well-specified models that need to control for all relevant observed confounders. However, unobserved factors might still threaten inference. Environmental INGOs may anticipate that IOs mainstream environmental policy and establish themselves in proximity to ensure that they can affect this process. We address this concern through several robustness checks. Initially, we re-estimate models using three-year (Table A4) and five-year (Table A5) lags of our environmental INGO variable because anticipation is much more difficult over longer time frames. Then, we utilize an instrumental variable approach to account for these concerns (Table A6). To be valid, an instrumental variable needs to fulfill two criteria: it should predict the presence of environmental INGOs (relevance) and should only affect IO environmental mainstreaming through its influence on the presence of environmental INGOs (excludability). We use a variable measuring entry and exit restrictions on civil society from the Varieties of Democracy Project (Coppedge et al., 2021) as an instrument. Exit and entry restrictions on civil society clearly affect whether INGOs can establish themselves in a given country. A one-unit increase in these restrictions increases the number of environmental INGOs by 10 and explains around 11% of the variation in environmental INGOs in a simple binary regression. At the same time, the instrument is plausibly excludable because these restrictions have no clear theoretical pathway to environmental mainstreaming of IOs, conditional on covariates like the level of development and democracy of the host country.

Second, one may be concerned that using a binary dependent variable focusing simply on the presence of mainstreaming could affect the results. Therefore, we re-estimate the models using a count of the number of areas where an IO has introduced environmental policy as the dependent variable (Hooghe et al., 2019). We also use pseudo-maximum likelihood models that are more appropriate for count data (Table A7). Instead of comparable estimation techniques for count data we employ the Poisson-pseudo maximum likelihood models due to their superior performance in cases with over-dispersion and many fixed effects (Santos Silva & Tenreyro, 2011).

Third, we utilize two alternative independent variables. To ensure that our results hold when using a different definition of proximity, we use an alternative measure of proximate INGO secretariats that counts the number of INGOs located in the same city as an IO (Table A8) as well as a measure based on a 500 km driving distance between the cities where IOs and INGOs are located (Table A9). Additionally, to safeguard against outliers biasing our results, we use a log (+ 1) count of proximate INGOs (Table A10).

Fourth, we use several additional specifications. The main models include IO fixed effects and, hence, utilize variation in the environmental mainstreaming of the same IO over time. In the Appendix, we also report results without IO fixed effects that show that the number of environmental INGOs is associated with between-IO differences in environmental mainstreaming (Table A11).

Additionally, some readers may question whether the results can be explained by peculiarities around where IOs are located. Specifically, many IOs are located in certain Western countries that also show a solid civil society presence. Hence, we re-estimate our models, excluding respectively one of the five countries with the most IO headquarters (USA, UK, France, Belgium, Switzerland) to ensure that the results are not simply driven by including or excluding one particularly influential headquarter location (Table A12). We also estimate sub-samples for task-specific and general-purpose IOs, respectively (Table A13) and control for IO engagement in different policy areas (Table A14). Additionally, we estimate a conditional logit model to ensure that results are not substantially different when using non-linear models (Table A15).

Another possible concern could be that environmental mainstreaming follows a two-step process. Focusing solely on the environmental policy of IOs could induce bias because models could not account for the fact that member states do not always decide to reform IOs. In a robustness check, we model this two-stage process explicitly using the Heckman correction (Heckman, 1976). To correctly fit the Heckman model, we need to include an instrument that predicts the first stage but not the second stage. To identify an instrument, we rely on two arguments from the literature on IO decision-making: (a) authors have argued that member states’ heterogeneity increases the difficulty to adopt a decision and increases the likelihood of gridlock (Copelovitch, 2010; Sommerer et al., 2021); (b) the literature has emphasized that increasing the number of member states makes gridlock more likely (Sommerer et al., 2021). We use the interaction of both variables as our instrument and estimate a Heckman probit model (Table A16).

We also conduct several additional tests to ensure that results are not driven by pseudo-replication. To this end, we re-estimate models including a lagged dependent variable (Table A17), use a Cox proportional hazards model to study environmental mainstreaming as IOs become older (Table A18), and several cross-sectional regressions including the first difference of proximate INGOs in the last year before a given IO mainstreamed environmental issues as the main independent variable (Table A19).

Finally, in our main models, we control for co-location as a means of diffusion between IOs. However, the literature highlights additional channels of diffusion, such as similar functional reference groups, or broader geographical dynamics, like being co-located in the same region. We account for these alternative channels in a robustness check by including corresponding control variables (Table A20).

The results, which we showcase in the Appendix, are robust to most of these variants of the model specification. However, our primary independent variable fails to attain statistical significance in four specifications: the Poisson pseudo-maximum likelihood with year fixed effects and IO-specific time trends (Table A7, Model 34); one of the models using an alternative independent variable which measures INGOs in a 500 km radius (Table A9, Model 40); the conditional logit with year fixed effects (Table A15, Model 60); and the Cox proportional hazards model (Table A18, Model 65).

Two of these regressions, Model 34 and 60, are maximum likelihood estimations. Since these regressions include IO fixed effects, the maximum likelihood estimators drop observations that do not vary over time—those IOs that never expand into environmental policy. However, they are informative because IOs could have plausibly expanded into environmental policy. Our main variable of interest is statistically significant in the comparable OLS estimations (Model 4 and 32), which do not drop these IOs. Hence, sample restrictions are one possible interpretation of the null findings in the two specifications discussed. In addition, we are also not very concerned that an alternative measure of our independent variable, based on a 500 km radius around IO headquarters, fails to attain statistical significance in Model 40 because the results are consistent when using all other variants of our independent variable.

Finally, the hazard ratio for our main variable of interest is positive but fails to attain statistical significance in the robustness check using the Cox proportional hazards model (Table A18, Model 65). We prefer the linear probability models over the Cox regression for three reasons. First, the Cox regression builds a risk set for IOs at the same age and compares them with each other. Our main hypothesis focuses on the growing intensity of pressure to mainstream associated with increases in the presence of proximate INGOs. In other words, we are interested in changes in the presence of proximate INGOs rather than the overall number of proximate INGOs. Hence, a linear probability model is closer to our argument than the Cox regression. Second, the risk set includes IOs of the same age in the dataset. This procedure means that we compare IOs that, for example, are two years old in the data irrespective of when they were founded. The model focusing on age compares, for example, the WTO in 1997 with other IOs two years after their founding, like the FAO in 1952. The comparison might be problematic, as environmental issues had a different standing in the 1990s than in the 1950s. Furthermore, one-third of the IOs were founded before 1950 (when the dataset starts) and we do not have data for either the independent or many of the control variables before this date. Third, while reversals of environmental mainstreaming do not occur in the data, they are theoretically possible and meaningful. Pressure by proximate INGOs could make abandoning environmental mainstreaming much more difficult and, therefore, should be accounted for in the data.

Despite the results when using the four discussed specifications, we are confident in the conclusions drawn in this article because all other models show robust results in line with the main arguments made.

Probing the mechanisms: The role of US-based INGOs in the coalition to mainstream environmental issues within the World Bank

To probe the plausibility of our argument and the causal mechanisms behind the associations presented above, we study the role of US-based INGOs in the coalition to mainstream environmental issues within the World Bank. The case selection literature recommends selecting a pathway case that is most “useful for elucidating causal mechanisms rather than verifying or falsifying general propositions (which are already more or less apparent from the cross-case evidence)” (Gerring, 2007: 241). For our purpose, a pathway case is one where civil society has played a central role in mainstreaming environmental issues in an IO. We use the case of the World Bank for which the crucial role of INGOs in mainstreaming the environment within the World Bank is well established: “external pressure has overwhelmingly, although not exclusively, influenced the Bank” (Park, 2010: 182, also Wade, 1997; Park, 2005a; Gutner, 2005; Cleary, 1995; Nelson, 1997). We utilize the manifold case studies that traced the World Bank’s environmental mainstreaming to highlight the particular relevance of proximate INGOs in inside lobbying, outside lobbying and information provision (H1). We also show that these efforts relied on advocacy coalitions with NGOs in recipient countries (H2).

The World Bank is a noteworthy example of an IO that underwent sweeping reform towards mainstreaming the environment. The Bank’s environmental initiative began with establishing the Office of Environmental and Health Affairs in 1970 (Gutner, 2005: 17). Until the mid-1980s, NGOs largely ignored the Bank’s environmental record (Wade, 1997: 623–624) and internally, staff had broad discretion regarding how much attention they paid to environmental aspects (Wade, 1997: 611). However, coinciding with a lobbying campaign by US-based INGOs, the Bank has performed “a paradigmatic shift from ‘frontier economics’ before 1987, to ‘environmental protection’ up to the early 1990s, and on to the more comprehensive ‘environmental management’ after that (…), accompanied by changes in staffing, organization, and procedures” (Wade, 1997: 730). The World Bank created a separate Environment Department in 1987, installed safeguard policies in 1989, empowered an Inspection Panel in 1993 alongside a public information policy in 1994 (Gutner, 2005: 17–20; Nielson & Tierney, 2003: 266).⁷ While the Bank had just three environmental specialists in 1983, their number increased to more than 300 in 2000 (Wade, 1997: 611–612; Nielson & Tierney, 2003: 264). Ultimately, the World Bank fully integrated environmental mainstreaming into its policy scope. By 2001, the World Bank had developed an environmental strategy (World Bank, 2001), committed substantial resources to environmental lending and had several environmental committees and working groups (Hooghe et al., 2019).

Environmental mainstreaming resulted from fierce campaigning initiated by three US-based INGOs, the Natural Resources Defence Council, the Environmental Policy Institute and the National Wildlife Federation, in 1983 (Wade, 1997: 657). One factor for the US-based INGOs to select the World Bank as a target of their campaign was that the World Bank “happened to be right next door” (Wade, 1997: 658). US-based organizations benefitted from their effective representation in Washington and long-standing lobbying experience. As a result, INGOs based in Washington, DC, have led the campaign (Nelson, 1997: 432). These organizations were joined by others, including the Sierra Club, the Environment Defense Fund and European NGOs, and later linked to groups from the project countries (Park, 2005b: 123).

INGOs’ success in pushing for environmental mainstreaming was based on visible campaigns focusing on the environmental and social impact of World Bank-funded ‘problem projects’ (Park, 2005b). US-based INGOs used the World Bank funded Polonoroeste project in Brazil “as their trampoline for demanding changes in Bank policy” as the project represented the “quintessential example of its pursuit of misguided development strategies” (Wade, 2016: 217). Other problem projects included the Sardar Sarovar Dam in India (Nelson, 1997), the Transmigration and the Kedung Ombo projects in Indonesia (Cleary, 1995), or the Pangue Dam in Chile (Park, 2005a).

INGOs played a crucial role in the environmental mainstreaming of the World Bank through inside lobbying, outside lobbying, and information provision. For each of these three mechanisms, proximate INGOs played an outsized role in the campaign. First, INGOs in proximity to the World Bank headquarters used inside lobbying to advocate for greening Bank operations. At the beginning of the campaign, INGOs attempted to directly lobby the Bank, which, however, ignored the INGOs as a “passing irritation” (Wade, 1997: 656–657). Representatives of the three leading INGOs met with a senior Bank representative for the first time in 1984, while the Bank played down their concerns (Wade, 1997: 661–662). On Polonoroeste, the US-based INGOs coordinated a letter campaign, signed by 32 NGOs from eleven countries to World Bank president Clausen in October 1984, calling for immediately suspending disbursements and demanding that the Bank undertook “concrete measures (…) to improve the ecological design and review of its projects” (Wade, 1997: 662–663). A coalition of INGOs also tried to persuade the IFC via letters, petitions and direct meetings to stop the Pangue Dam project and for an accountability mechanism to be established (Park, 2005a).

However, as direct inside lobbying was insufficient to stop environmentally devastating projects and achieved little policy change, INGOs shifted their strategy towards indirect inside lobbying by approaching donor state representatives (Nielson & Tierney, 2003: 257). The US Congress was particularly suited for INGO lobbying because the Bank received sizeable US taxpayer contributions. Congress appropriates funds annually, which allowed INGOs to lobby for attaching environmental conditions (Wade, 1997: 658). Using Polonoroeste as an image, the US-based INGOs appealed to Congress to intervene. From 1983 to 1987, Congress held over twenty hearings on development banks’ performance in various subcommittees (Gutner, 2005: 17). The timing was perfect for US-based INGOs as they could build support for when the US negotiated an IDA replenishment alongside a capital increase for the Bank. By 1986, the campaign redirected pressure towards the US Treasury to lobby the Bank to improve its environmental performance indirectly. In 1986, the US Executive Director for the first time voted against a power sector loan for Brazil on environmental grounds. Other major donor countries, including Germany and Japan, were adding further environmental reform pressure (Park, 2005b: 125).

Second, US-based INGOs were also crucial to outside lobbying. A central component of the campaign was public opinion mobilization through generating press coverage in donor country newspapers. While this included European newspapers, the vast majority centered on US newspapers like the Washington Post and the New York Times (Nielson & Tierney, 2003). As public attention rose, media organizations actively looked for critical views to the official World Bank statements, often obtained from US-based INGOs (Wade, 1997: 727). On Polonoroeste, the New York Times featured the 1984 INGO letter to the World Bank president (Wade, 1997: 663). In 1985, one of the leading INGO campaigners, co-authored an article in the environmental Defenders magazine, which generated thirty to forty daily protest letters (Wade, 1997: 665). In 1987, leaders of seven DC-based INGOs authored an article in Science demanding to withhold US contributions to IDA unless the Bank reduced its environmental impact (Wade, 1997: 672). The US-based INGOs used a similar strategy to problematize the Sardar Sarovar project in India. They orchestrated 250 organizations from thirty-seven countries to sign a full-page letter in the New York Times, the Washington Post, as well as the London-based Financial Times, demanding that the World Bank withdrew from the project (Nelson, 1997: 428). US-based INGOs persuaded members of Congress to hold public hearings before adopting legislation on World Bank appropriations. At these hearings, INGOs and representatives of indigenous rights organizations could scandalize the environmental and social impact of Bank-funded projects (Wade, 1997: 659). By inviting local indigenous leaders, US-based INGOs presented persuasive testimony contradicting World Bank claims about project benefits to local people (Wade, 1997: 673).

Third, US-based INGOs played a crucial role in information provision on environmental issues. Initially, the Bank “strongly discouraged the staff from having even informal contact” with both US-based and NGOs from recipient countries (Wade, 2016: 223–224). However, since the 1990s, the World Bank has sought out INGO expertise in community development, participation, and social sectors (Nelson, 1997: 434–435). INGOs relied on coalitions with NGOs in recipient countries to serve as transmission belts for information on local contexts. US-based INGOs could assemble evidence that the projects harmed rather than benefitted local communities (Wade, 1997: 673). INGOs could partner with and represent local NGOs and offer unique perspectives from project-affected people, which were excluded from global policy-making (Cleary, 1995: 10). Local information proved beneficial in interactions with World Bank staff and Executive Directors, for instance, when an NGO coalition informed the US Executive Director on the detrimental impact of the Pangue Dam based on local reporting (Park, 2005a: 104). INGO provided information that led to policy changes at IFC have improved the effectiveness of IFC projects (Park, 2005a: 113).

While the role of proximity to the World Bank was crucial in each of the three mechanisms, US-based INGOs extensively relied on building advocacy coalitions with NGOs from recipient countries (Park, 2005b: 123). One notable example was the Sadar Sarovar campaign, which formed an “alliance of determined villagers, local activists, and international groups” organized simultaneously both from the project area in India and World Bank headquarters (Park, 2005b: 125). In those advocacy coalitions, US-based INGOs coordinated the campaigns, identified other groups to join the networks, prepared actions, congressional hearings, press conferences, and information provision activities. On inside lobbying, US-based INGOs strategically amplified the voice of local organizations, for instance, when they brought a leading Brazilian environmentalist to testify before the US Congress (Wade, 1997: 655–656, 673). Regarding outside lobbying, US-based INGOs used images and local information provided and distributed through advocacy networks to feed their media campaigns in US-based newspapers. On information provision, INGOs informed decision-makers by drawing on local NGOs’ materials during the Polonoroeste and the Sardar Sarovar campaigns. INGOs used this locally generated evidence to inform World Bank staff, Congress members or the Executive Directors of the detrimental effect of Bank projects (Wade, 1997: 695–697, 707). The advocacy network on the Kedung Ombo project in Indonesia also adopted such a twin-track advocacy strategy (Cleary, 1995: 19), as did the advocacy network on the IFC Pangue Dam project (Park, 2005a: 104).

The case of the campaign to mainstream environmental issues also illustrates the problems that come with the special access of proximate INGOs often located in Western countries. North-South advocacy coalitions have formed around specific programs but may differ on the goals of a campaign. For the US-based INGOs, the main goal was to achieve World Bank environmental reform while projects creating images of devastation were most beneficial for their campaigns. INGOs saw lobbying the US Congress to withhold funding as their best opportunity to create leverage, even if this meant that funding to the World Bank was cut or delayed significantly. Local NGOs, however, were interested in ensuring that the projects themselves did neither endanger the environment nor indigenous populations, rather than questioning World Bank development funding per se. In the Kedung Ombo and the Scott Paper cases in Indonesia, local NGOs noted that a less confrontational approach by INGOs would have benefitted local people more, while INGOs present only an incomplete version of these views in Washington, DC (Cleary, 1995: 25,32–33; Nelson, 1997: 433).

In sum, the case illustrates the pathway from proximate INGOs to World Bank environmental mainstreaming and the amplifying impact of INGO-NGO advocacy coalitions. US-based INGOs were able to leverage their contacts with World Bank staff and member state representatives. They could also draw on their relationships with US newspapers to ramp up the public pressure on the World Bank. US-based INGOs established themselves as information providers for World Bank projects’ environmental impact in project areas, allowing INGOs to speak with authority on the issue. In each of the mechanisms, they relied extensively on building advocacy coalitions with NGOs based in recipient countries. The case also shows how proximity affects inequality of stakeholder influence in global governance.

Conclusions

IOs have increasingly mainstreamed environmental issues into their policy scope. This article has offered a large-n comparative empirical assessment of the role environmental INGOs played in this development and probed the theoretical mechanisms with the campaign of US-based environmental INGOs for mainstreaming the environment at the World Bank as a pathway case.

Three findings stand out: first, we show through a battery of statistical tests that environmental INGOs whose secretariats are in geographical proximity to IO headquarters are associated with the environmental mainstreaming of IOs. The results also hold when accounting for alternative kinds of IO exposure to environmental civil society. Second, INGOs seem to rely on a strong presence of domestically focused environmental NGOs to provide relevant information from their member states. If INGOs can rely on such domestic NGOs, the likelihood that an IO mainstreams environmental issues expands substantially. Third, the case study of the campaign to mainstream environmental issues in the World Bank illustrates that INGOs close to IO headquarters could increase the campaign’s ability to influence environmental mainstreaming through inside lobbying, outside lobbying and information provision. However, they relied largely on domestic NGOs from recipient countries that provided them with the necessary information and moral authority for their campaigns.

Our analyses have two main limitations that should be kept in mind when interpreting our results. First, the measures we employed for the strength of the environmental civil society in different countries do not permit disaggregating these INGOs’ activities and qualities. Some INGOs may have no interest in engaging with IOs, and there is variation in the importance of certain INGOs over others. Future research could unpack these dynamics by collecting more fine-grained data on civil society’s engagement with a large number of IOs. Second, we relied on a non-representative sample of “important” IOs (Hooghe et al., 2017) and data on formal institutional mainstreaming. The comparative IO literature could gain by ensuring the representativeness of the influential datasets and collecting representative samples of IOs which allow for broader generalization. Furthermore, the literature on scope expansion has highlighted differences between formal engagement and informal behavior of IOs (Tallberg et al., 2020). By studying other dependent variables, like decision-making or budget allocation, more insights could be gained on the impact of environmental civil society on the environmental mainstreaming of IOs. Newly available data on climate adaptation, one crucial dimension of environmental mainstreaming, could help better understand the differential engagement of IOs with environmental issues (Kural et al., 2021). Such analysis is necessary to better understand the quality of IO environmental mainstreaming.

The findings speak to two broader debates in international relations: first, the demonstrated differences between the influence of civil society actors imply that there is a spatial component in their ability to affect policy change. This points to an important yet largely neglected element: the location of IOs and its impacts on access, influence, effectiveness, and equity (Johnson, 2014; Ivanova, 2021, 2007; Dairon & Badache, 2021). The World Bank case study illustrates that differences in access may have important distributional consequences. NGOs in recipient countries had to rely on US-based INGOs to get their voices heard. This led to a marginalization of their interests during the campaign. INGOs pushed for cutting funding for IDA despite the objections from the NGOs in recipient countries. This problem seems particularly pronounced for IOs with a focus that spans multiple regions. Of the 36 IOs with a global or multi-regional focus in the dataset, only three are located in non-OECD countries. The International Seabed Authority is the sole multi-regional or global IO not located in a high-income country (Jamaica). This implies that key decisions, such as the choice for the location of an IO agreed upon many years ago, can exacerbate the inequality among the different NGOs that make up global civil society. Ivanova (2021, 2007), focusing on the genesis of UNEP, notes that these location decisions are highly political and the emergence of a few centers of political power with high IO concentration is not by chance. Location, in turn, impacts the IOs capacity and connectivity. The literature on civil society influence on IOs could benefit from making location a more important variable in discourse and analysis.

Second, large-n comparative IO research may benefit from incorporating variables that measure the preferences of actors other than member states and IO bureaucracies (Debre & Dijkstra, 2021; Tallberg et al., 2020; Hooghe et al., 2019). While we have a good understanding of civil society actors’ formal access, we know much less about the degree to which actors can leverage such access for policy change (Tallberg et al., 2013). Our argument on the informal influence of proximate INGOs could prove useful for scholars studying IOs from a comparative perspective. Comparative IO research could further extend our understanding of the decisions of the large population of IOs by considering the preferences of a greater number of actors that seek to affect the decisions of IOs and by discussing the consequences of the physical location of IOs more systematically.