Democracy, Autocracy, and the Design of International Organizations

Abstract

Extensive research expects systematic differences in the design of international organizations (IOs) based on the regime composition of their memberships. Yet, so far, empirical analyses have found limited support for this expectation. This article resolves this puzzle by providing a new understanding of how the regime composition of IOs shapes their institutional design. Theoretically, it argues that this relationship is moderated by a critical but overlooked factor: the governance purpose of IOs, as expressed in the distinction between general-purpose and task-specific organizations. Empirically, it provides a comprehensive analysis of how changes in regime composition have affected institutional design in forty IOs from 1950 to 2019. The findings show that the regime composition of IOs indeed is related to their institutional design, but only in general-purpose organizations, which present democracies and autocracies with more divergent design incentives than task-specific organizations. The article suggests that democracy, autocracy, and international cooperation are linked in more complex and contingent ways than grasped in previous research.

Existen investigaciones exhaustivas, las cuales prevén diferencias sistemáticas en el diseño de las organizaciones internacionales (OOII) en función de la composición del régimen de sus miembros. Sin embargo, hasta ahora, los análisis empíricos llevados a cabo solo han podido ofrecer un nivel de apoyo limitado para esta expectativa. Este artículo resuelve este enigma proporcionando una nueva comprensión en materia de cómo la composición del régimen de las OOII da forma a su diseño institucional. De forma teórica, el artículo argumenta que esta relación está moderada por un factor crítico que tiende a pasarse por alto: el propósito en materia de gobernanza de las OOII, tal como se expresa en la distinción entre las organizaciones de propósito general y las organizaciones con tareas específicas. Desde el punto de vista empírico, el artículo proporciona un análisis exhaustivo con respecto a cómo los cambios en la composición del régimen han afectado al diseño institucional en 40 OOII desde 1950 hasta 2019. Las conclusiones demuestran que la composición del régimen de las OOII está relacionada con su diseño institucional, pero que esto solo ocurre en el caso de las organizaciones de propósito general, que presentan incentivos de diseño más divergentes para las democracias y las autocracias que las organizaciones con tareas específicas. El artículo sugiere que la democracia, la autocracia y la cooperación internacional están vinculadas de maneras más complejas y contingentes que las que se han identificado en las investigaciones anteriores.

D'après un nombre important de travaux de recherche, il devrait y avoir systématiquement des différences dans la conception des organisations internationales (OI) selon la composition de leur régime d'adhésion. Pourtant, jusqu'ici, les analyses empiriques n'ont apporté qu'un soutien modeste à cette attente. Cet article vient résoudre cette énigme en proposant une nouvelle compréhension de la façon dont la composition du régime d'une OI façonne sa conception institutionnelle. Sur le plan théorique, il affirme que cette relation est modérée par un facteur décisif mais négligé : la finalité de gouvernance des OI, qui s'exprime dans la distinction entre les organisations à objectif général et celles dédiées à une mission. Sur le plan empirique, il fournit une analyse exhaustive des effets des changements de la composition d'un régime sur la conception institutionnelle de 40 OI entre 1950 et 2019. D'après les conclusions, la composition du régime des OI est bel et bien liée à leur conception institutionnelle, mais seulement dans le cas des organisations à objectif général, qui proposent des incitations de conception qui diffèrent davantage aux démocraties et autocraties que les organisations dédiées à une mission. L'article suggère que la démocratie, l'autocratie et la coopération internationale entretiennent des liens plus complexes et fortuits que les recherches antérieures ne l'avaient compris.

Introduction

One of the most consistent findings in the study of world politics pertains to the relationship between regime type and international cooperation. Across a broad range of areas, research has identified systematic differences between democratic and autocratic states in their approach to cooperation. For instance, democracies have been shown to be more likely to liberalize international trade (Mansfield et al. 2000), commit to human rights treaties (Simmons and Danner 2010), engage in democracy promotion (Pevehouse 2005), provide global public goods (Bättig and Bernauer 2009), promote accountable cooperation (Grigorescu 2015), solve conflicts through dispute settlement (Davis 2012), and stay out of war with one other (Russett 1993).

Building on this record, a new wave of research has developed plausible expectations about the influence of regime type on the design of international organizations (IOs; e.g., Tallberg et al. 2016; Hooghe et al. 2019; Obydenkova and Libman 2019; Schimmelfennig et al. 2020; Ginsburg 2021; Debre and Sommerer 2022). The core assumption is that democracies are more willing than autocracies to accept institutional designs that relax state control over IOs. As a result, IOs with more democratic memberships—that is, higher democratic density—should have designs involving more delegation of authority to supranational bodies, more pooling of authority through majority decision-making, and more relinquishing of authority through access for transnational actors (TNAs). A comparison often invoked to support this expectation is the contrast between the democratic and supranational European Union (EU) and the autocratic and intergovernmental Shanghai Cooperation Organization (SCO).

Yet, when testing these expectations empirically, the same studies have overwhelmingly found limited support for their conjectures. Almost irrespective of measure, sample, and time period, the democratic density of IOs appears to be of little importance for the design of these organizations. How can we explain this puzzling divergence between well-founded expectations and empirical findings? Are there truly no or few differences between democracies and autocracies in how they design IOs? Or has previous research simply failed to capture more profound differences because of how it has approached the issue?

This article helps to resolve this puzzle. We argue that the regime composition and institutional design of IOs are linked in more complex ways than theorized and tested in previous research. Instead of expecting an unconditional association between democratic density and IO design, we submit that this relationship is moderated by a critical but overlooked factor: the governance purpose of an IO, as expressed in the distinction between general-purpose and task-specific organizations (Lenz et al. 2015; Hooghe et al. 2019). General-purpose IOs have open-ended contracts that allow them to address any problem that arises in a given community of states, while task-specific IOs have closed-ended contracts that aim to address a given problem among all states affected by it. Previous analyses have treated all IOs alike—theoretically as well as empirically—and therefore failed to capture how the impact of democratic density depends on the governance purpose of IOs. By contrast, we advance an account in two steps that is sensitive to the differences between general-purpose and task-specific organizations.

First, we develop a novel theoretical argument about the moderating impact of governance purpose on the relationship between democratic density and IO design. We argue that the regime composition of an IO’s membership matters more extensively for institutional design in general-purpose IOs compared to task-specific IOs. While task-specific governance presents democracies and autocracies with more similar benefits and costs from designs that relax state control, general-purpose governance leads to more divergent incentives vis-à-vis such designs. Under general-purpose conditions, designs involving delegation, pooling, and access are more likely to generate dynamic developments that take cooperation in other directions than originally envisaged. Autocracies are more fearful of such developments than democracies, since expansion of cooperation into new policy areas could threaten autocratic leaders’ hold on power by compromising their ability to reward selectorates and repress populations.

Second, we test this expectation through a comprehensive empirical analysis of the relationship between democratic density and institutional design in forty IOs from 1950 to 2019. Our analysis is based on a stratified sample of organizations, comprising general-purpose and task-specific IOs in different world regions and issue areas. Our estimation strategy identifies whether and how changes in the democratic density of IO memberships are related to changes in institutional design, holding organizational contexts constant. The extended time period enables us to study this relationship over the course of several periods of democratization and autocratization, including the most recent episode of democratic backsliding (Lührmann and Lindberg 2019). The analysis encompasses three key dimensions of institutional design—delegation, pooling, and access—based on a combination of novel and existing data. These dimensions have been at the forefront of the debate and together capture how IOs may relax state control through supranational and transnational designs (Tallberg et al. 2014; Hooghe et al. 2019).

Our findings demonstrate that the relationship between democratic density and IO design is more conditional than previously understood. In line with our expectation, governance purpose shapes whether and how regime composition matters for institutional design. In general-purpose IOs, where contracts are open and cooperation dynamic, democratic density tends to be positively associated with pooling and access, while estimates for delegation are positive but less precise. Conversely, in task-specific IOs, where contracts are closed and cooperation static, democratic density has no relationship to institutional design. The basic governance purpose of IOs thus moderates the impact of democratic density, which helps to explain why strong theoretical expectations on a direct positive relationship to institutional design are not borne out in recent studies. These findings are robust to a broad range of alternative specifications, including an extension to a larger sample of seventy-four IOs (Hooghe et al. 2019).

This article has several important implications. First, it shows that democracy, autocracy, and international cooperation are linked in more contingent ways than previous literature leads us to expect. While earlier research has identified independent effects of regime type on a range of cooperative outcomes (e.g., Mansfield et al. 2000; Pevehouse 2005; Bättig and Bernauer 2009; Simmons and Danner 2010; Grigorescu 2015), the design of IOs appears to be governed by more complex dynamics. Second, this article indicates that scholarship on institutional design has underestimated the importance of regime type for the ways in which states organize cooperation. Conditional on the governance purpose of IOs, democracies and autocracies make different choices on delegation, pooling, and access that cannot be captured by theories treating states as like units (e.g., Abbott and Snidal 2000; Hawkins et al. 2006; Koremenos 2016; Hooghe et al. 2019). Third, this article contributes to a growing literature on autocratization and IOs (e.g., Debre 2021; Cottiero and Haggard 2023; Meyerrose and Nooruddin 2023; Winzen 2023; Cottiero et al. 2024) by suggesting that current processes of autocratization in world politics will have a differential impact on international cooperation. While autocratization is likely to have negative effects on delegation, pooling, and access in general-purpose IOs, the implications may be less far-reaching or non-existent in task-specific IOs.

Puzzle: Regime Composition and Institutional Design

Recent years have seen a wave of new research on the relationship between regime type and IO design. These studies have typically focused on one or several of three key dimensions of institutional design: delegation of authority to supranational bodies in IOs, such as secretariats, courts, and assemblies; pooling of authority through majority voting in interstate decision-making; and relinquishing of authority through access for TNAs, such as non-governmental organizations, businesses, and scientific experts. Together, these three dimensions capture how states may cede control over IOs through supranational and transnational designs at the expense of pure intergovernmental cooperation (Tallberg et al. 2014; Koremenos 2016; Hooghe et al. 2017; Zürn et al. 2021; Sommerer et al. 2022). The core assumption in this body of research is that democracies are more willing than autocracies to relax state control over IOs. These studies therefore expect a positive relationship between the democratic density of IO memberships and the degree of delegation, pooling, and access in these organizations. If this expectation is correct, it could explain variation in design across IOs with different membership compositions, but also key developments in design over time as the democratic density of IOs has waxed and waned.

This widespread expectation extends a powerful tradition of research privileging the role of regime type in explaining outcomes in international cooperation. Previous studies in this tradition have shown that democracies are more likely than autocracies to join IOs (Simmons and Danner 2009), liberalize international trade (Mansfield et al. 2000), commit to human rights treaties (Simmons 2009), engage in democracy promotion (Pevehouse 2005), provide global public goods (Bättig and Bernauer 2009), adopt liberal international norms (Tallberg et al. 2020), foster accountable cooperation (Grigorescu 2015), solve conflicts through dispute settlement (Davis 2012), and stay out of war with one another (Russett 1993). Other studies in the same tradition suggest that states undergoing democratization processes are more likely to found IOs (Poast and Urpelainen 2013), join IOs (Mansfield and Pevehouse 2006), and commit to human rights treaties (Moravcsik 2000). A strong and consistent relationship between regime type and IO design would conform with this impressive explanatory record.

Yet, when turning from theoretical conjectures to empirical tests, recent studies of institutional design find little systematic support for their expectations. Almost across the board, they find no or limited evidence in favor of democratic density as an explanation of IO design. Organizations with a higher share of democracies in their membership do not appear, overall, to involve higher levels of delegation, pooling, and access. A key part of the explanation, we will argue, is another similarity shared by these studies: namely, an underappreciation of how this relationship may play out differently in general-purpose and task-specific IOs.

Ginsburg (2021) contributes a theory of the differences between liberal and authoritarian international law. He argues that democratic leaders cooperate to generate benefits for the electorate at large and seek to bind their successors, while autocratic leaders cooperate to secure private benefits and to stay in power. These differences should translate into democratic leaders favoring delegation to supranational bodies and pooling through majority voting, while autocratic leaders would resist both features and focus on preserving sovereignty and flexibility. However, when Ginsburg (2021, 97–101) examines this argument empirically in the context of delegation of judicial power in seventy-eight IOs, he finds contradictory evidence: While IOs with higher democratic density are more likely to have dispute-resolution provisions, they are less likely to establish international courts.

Daugirdas and Ginsburg (2022) expand on this argument in a subsequent paper, where they suggest that delegation and pooling pose risks that are unacceptable to authoritarian governments. However, when studying the founding treaties of about 400 IOs in a variety of world regions and issue areas, they find no support for their conjectures: “contrary to expectations, authoritarian international organizations are, in general, quite similar to the rest in the way that they are legally structured. The secretariats of authoritarian organizations have similar protections for their independence and autonomy. And the extent of majority decision making in intergovernmental organs is also quite similar across authoritarian and democratic organizations” (Daugirdas and Ginsburg 2022, 3).

Hooghe et al. (2019) provide an in-depth account of the drivers of delegation and pooling in IOs. Their core argument is that these features are driven by a combination of normative affinities between member states and the scope of IO policy portfolios. However, they also test for an effect of the democratic density of IOs “on the hypothesis that democratic rulers are less fearful of supranational authority than are authoritarian rulers” (Hooghe et al. 2019, 97). Their empirical analysis of delegation and pooling in seventy-six IOs indicates limited support for this expectation (Hooghe et al. 2019, ch. 6). They find no support for a relationship between democratic density and delegation, but note that democratic density is weakly significant in their model for explaining pooling.

Related studies explore the potential effects of democratic density on delegation in more specific empirical domains: parliamentary assemblies (Schimmelfennig et al. 2020), dispute-settlement bodies (Haftel 2013), and regional organizations (Lenz et al. 2023). These studies produce analogous results. Schimmelfennig et al. (2020) find no statistically significant association between democratic density and the existence or empowerment of parliamentary assemblies in IOs. Haftel (2013, 409) concludes in an analysis of twenty-eight regional economic IOs that “regime type is not an important determinant for regional institutionalization.” Lenz et al. (2023) observe no direct effect of democracy in IO memberships on the degree of delegation in regional organizations.

Turning to access, Tallberg et al. (2014, 2016) expect democracy in IO memberships to drive the inclusion of TNAs in IO policymaking. They theorize three mechanisms for this effect: the democratic density of memberships, the influence of new democracies, and the presence of democratic major powers. Similarly, Lall (2023) tests for an effect of democratic density on TNA access while privileging other theoretical explanations. These two studies arrive at contradictory evidence. While Tallberg et al. (2014, 2016) find democratic density to be a key driver of TNA access to fifty IOs from 1950 to 2010, Lall (2023) does not find a positive relationship in an analysis of fifty-two institutions from 1960 to 2018.

Concentrating specifically on regional IOs composed of authoritarian member states, Obydenkova and Libman (2019) expect such IOs to differ systematically from regional IOs with democratic memberships. A key difference, they argue, is that authoritarian leaders prefer to remain in control of these IOs themselves, which means limited delegation and pooling. Yet the empirical analysis finds few significant differences between democratic and autocratic regional IOs (Obydenkova and Libman 2019, 89–91). While autocratic regional IOs are more likely to have a forum that brings together the heads of state, consistent with an intergovernmental logic, they are not less likely to have supranational commissions, secretariats, and assemblies, contrary to the authors’ expectation.

Debre and Sommerer (2022) are particularly interested in the consequences of recent autocratization processes for the institutional design of IOs. They posit that states undergoing autocratization episodes have particular preferences with respect to institutional design, as leaders in these countries wish to avoid authoritative IOs that can limit their ability to manipulate elections while wanting to maintain a veneer of democratic legitimacy. This logic leads to the expectation that autocratizing states should oppose delegation, pooling, access, and broad policy portfolios in IOs. However, when examining the impact of democratic backsliding in fifteen IOs, they find little evidence of an effect on these dimensions of institutional design: “contrary to our initial expectation, increasing member state autocratization does not seem to coincide with a decrease in institutional design dimensions” (Debre and Sommerer 2022, 5).

In sum, while there are strong and reasonable theoretical expectations that IOs with higher levels of democratic density should involve more extensive delegation, pooling, and access, recent empirical research finds no or mixed evidence to this effect. What can account for this puzzling divergence between expectations and findings? Existing studies provide no explanation for these weak results but contribute a number of plausible interpretations. Some scholars suggest that these non-findings might be a result of diffusion, as autocratic IOs have begun to adopt designs from democratic IOs in an effort to legitimize themselves (Schimmelfennig et al. 2020; Daugirdas and Ginsburg 2022). Others speculate that these results might be due to high thresholds for institutional reform in IOs, making it difficult for processes of democratization or autocratization in memberships to translate into changes in design (Debre and Sommerer 2022). A third interpretation suggests that democracies in fact are more like autocracies, since democratic leaders, too, fear losing sovereignty, because of a risk of domestic populist backlash, and therefore resist expansions of delegation, pooling, and access (Lenz et al. 2023).

Argument: Governance Purpose as a Moderating Factor

We present a novel argument about the relationship between regime type and IO design that helps to resolve the puzzling non-finding in existing research. Instead of expecting an unconditional association, we argue that this relationship is moderated by a hitherto overlooked factor—the governance purpose of IOs—as expressed in the distinction between general-purpose and task-specific organizations.

General-Purpose versus Task-Specific Governance

The distinction between general-purpose and task-specific IOs grows out of research by Hooghe, Lenz, and Marks (Lenz et al. 2015; Hooghe et al. 2019, ch. 4). Conceptually, this distinction intends to capture two ideal types with respect to the basic governance purpose of IOs. While general-purpose IOs aim to address any problem that arises in a given community of states, task-specific IOs aim to address a given problem among all states affected by it. General-purpose and task-specific IOs are thus different in key respects (Hooghe et al. 2019, 47–50).

First, and most importantly, the two types of organizations have different degrees of contractual openness. While general-purpose IOs operate with open-ended contracts to enable cooperation on future issues that might arise for a given community of states, task-specific IOs have closed-ended contracts to restrict cooperation to the issue in focus.¹ Contractual openness refers to a treaty’s degree of specificity with respect to the purpose of cooperation. When contracts are open, treaties present purposes that are open-ended with respect to future areas of cooperation. An example is the Treaty on European Union, which speaks of cooperation as an open process toward “an ever closer union among the peoples of Europe” (Article 1). When contracts are closed, treaties specify purposes in ways that restrict cooperation to the attainment of distinct policy goals. An example is the North American Free Trade Agreement (NAFTA), which states that the purpose of cooperation is to “eliminate barriers to trade in, and facilitate the cross-border movement of, goods and services between the territories of the Parties” (Article 102).

Second, the two types of organizations have different issue scopes. While general-purpose IOs have a broad issue scope, since they may handle any problem that emerges in a given community of states and are open to further expansion of their policy remit, task-specific IOs have a narrow issue scope, since they concentrate on resolving specific problems and have closed off the option of policy expansion. General-purpose IOs thus have comprehensive and dynamic policy portfolios, whereas task-specific IOs have delimited and static policy portfolios.

Third, the two types of organizations have different membership sizes. While general-purpose IOs are formed by given communities of states and thus selective in their membership, task-specific IOs are oriented toward all states affected by a given problem and thus comprehensive in their membership. Whereas general-purpose IOs are exclusionary almost by definition, task-specific IOs are non-exclusionary, since they are organized by issue and open to the possibility of universal memberships.

While the conceptual distinction between general-purpose and task-specific organizations simplifies empirical realities, it maps reasonably well onto the universe of IOs in world politics, which tends to have a bimodal distribution (Hooghe et al. 2019, ch. 4). On one side are general-purpose IOs, such as the African Union (AU), Association of Southeast Asian Nations (ASEAN), and EU, which tend to have open contracts, broad policy portfolios, and restricted (often regional) memberships. On the other side are task-specific IOs, such as the International Atomic Energy Agency, International Civil Aviation Organization, and World Health Organization, which tend to have closed contracts, narrow policy portfolios, and comprehensive (often global) memberships. But the universe of IOs also features several exceptions to this pattern, such as the United Nations (UN), which is a general-purpose IO that combines an open contract and broad policy portfolio with a universal membership, and NAFTA, which was a task-specific IO combining a closed contract and narrow policy remit with a small regional membership.

Furthermore, this distinction between general-purpose and task-specific governance has shown to have explanatory leverage. Several studies treat the governance purpose of IOs as an exogenous constitutive factor and then assess its impact on institutional design. Schimmelfennig et al. (2020), for instance, find that IOs with a general-purpose orientation are considerably more likely to establish international parliamentary assemblies. Likewise, Lenz et al. (2023) conclude that general-purpose IOs are more likely to see an endogenous expansion of delegated authority over time.

We build on this distinction between general-purpose and task-specific IOs to theorize how governance purpose impacts the relationship between an IO’s democratic density and institutional design. Like earlier research, we treat the governance purpose of an IO as an exogenous constitutive feature that is distinct from the level of democracy in an IO’s membership and analytically prior to subsequent decisions on delegation, pooling, and access.

Both democracies and autocracies engage in cooperation through general-purpose and task-specific IOs. Indeed, as we will later show, the average level of democratic density in these two types of organizations is about the same. The task-specific category includes IOs with a high proportion of democratic members, such as the European Free Trade Association, but also IOs with predominantly autocratic members, such as the Organization of Arab Petroleum Exporting Countries. Similarly, the general-purpose category includes IOs with a large share of democratic states, such as the Nordic Council, but also IOs with a higher proportion of autocracies, such as the SCO. However, depending on whether IOs have a general-purpose or task-specific orientation, we expect the relationship between democratic density and institutional design to play out differently.

Preferences of Democracies and Autocracies

In order to understand this moderating effect of governance purpose, we need to consider the preferences of democracies and autocracies on delegation, pooling, and access. For these purposes, we build on rational choice institutionalism to identify general benefits and costs from institutional design and on regime type explanations to identify specific benefits and costs to democracies and autocracies.

On the benefit side, moving away from a strict intergovernmental design through an extension of delegation, pooling, and access has functional advantages that facilitate cooperation for all states involved (Abbott and Snidal 2000; Koremenos et al. 2001; Voeten 2019). Delegating authority to supranational bodies helps to generate policy information, raise decision-making efficiency, and strengthen the credibility of commitments (Hawkins et al. 2006; Bradley and Kelley 2008). Pooling authority through majority voting in interstate bodies helps to improve the efficiency of decision-making by lowering institutional hurdles to agreement (Blake and Payton 2015; Hooghe and Marks 2015). Providing access for TNAs helps to solicit expert information, outsource field implementation, and facilitate compliance monitoring (Raustiala 1997; Tallberg et al. 2014).

On the cost side, the same moves away from a strict intergovernmental design come with certain general drawbacks, making all states hesitant about extending delegation, pooling, and access. The unifying concern is sovereignty costs: perceived losses in sovereignty resulting from shifts to designs that reduce each individual state’s autonomy and control (Abbott and Snidal 2000). Delegation reduces state control by conferring authority on independent supranational agents (Hawkins et al. 2006; Bradley and Kelley 2008). Pooling impinges on the autonomy of states by enabling majority decision that do not require the backing of all parties (Blake and Payton 2015; Hooghe and Marks 2015). Access invites TNAs into the policy-making process of IOs and reduces states’ control over outcomes (Tallberg et al. 2014; Koremenos 2016).

Next to these general benefits and costs, states confront particular incentives related to their regime type. While research suggests that the differences between democracies and autocracies in terms of audience constraints should not be exaggerated (Hyde and Saunders 2020), and that both democracies and autocracies come in multiple forms (Geddes 1999; Coppedge et al. 2020), a simplified dichotomy is helpful to illuminate differences in preferences between democratic and autocratic leaders as they consider the prospect of international cooperation. We join others in assuming that autocratic and democratic leaders tend to be driven by different preferences in international cooperation (Poast and Urpelainen 2013; Ginsburg 2021; Cottiero et al. 2024) while recognizing the diversity of both groups (Debre and Sommerer 2022).²

Leaders in democracies tend to be concerned with reelection. They face audience costs if they go against domestic opinion, and therefore tend to be responsive to general publics (Fearon 1994; Bueno de Mesquita et al. 2003). This motivates democratic leaders to maximize gains from cooperation that can benefit the median voter (Putnam 1988; Cottiero and Haggard 2023). In addition, democratic leaders tend to support international policies that work to uphold and spread democracy as a system of government, such as human rights, the rule of law, and democracy promotion (Pevehouse 2005; Donno 2013). While sharing the same core orientation, democratic leaders may vary in the strength of these preferences. For instance, leaders in new democracies may have stronger incentives to support authoritative, democracy-promoting IOs as a way of committing the state beyond the lifetime of the current government (Hafner-Burton et al. 2015), while leaders in backsliding democracies may have relatively weaker incentives as they fear costly scrutiny (Debre and Sommerer 2022).

By contrast, leaders in autocracies tend to be concerned with survival and therefore responsive to the narrow selectorate that appointed them, while also anxious to keep domestic publics under control (Bueno de Mesquita et al. 2003; Geddes et al. 2018). This motivates autocratic leaders internationally to secure private goods that can keep selectorates happy, such as profitable contracts; to support policies that facilitate repression of domestic publics, such as internal security; and to resist features that could threaten their regime, such as democracy promotion and human rights (Poast and Urpelainen 2013; Debre 2021; Ginsburg 2021, ch. 1; Cottiero and Haggard 2023). While sharing the same basic orientation, autocratic leaders, too, may vary in the strength of these preferences. For instance, leaders in new or fragile autocracies, whose control is less consolidated, may be more fearful of authoritative IOs, empowered civil society organizations, and democracy promotion than leaders of full autocracies (Debre and Sommerer 2022).

Governance Purpose, Regime Type, and Institutional Design

Consider, now, how these similarities and differences in preferences among democracies and autocracies may lead decisions on institutional design to play out differently depending on an IO’s governance purpose. While we expect the regime composition of IOs to matter little for delegation, pooling, and access in task-specific organizations, we anticipate that it has larger effects in general-purpose organizations.

In task-specific IOs, delegation, pooling, and access generate functional benefits and sovereignty costs that are appreciated and loathed by democracies and autocracies alike. In addition, states face few regime-specific concerns when extending these designs in this type of organization. Since task-specific IOs have closed contracts that restrict cooperation to clearly specified areas and prevent dynamic policy developments, they pose less risk to autocracies. Autocratic leaders can use delegation, pooling, and access to secure private goods for their selectorate, expand cooperation in areas that benefit the regime, and safely stay away from areas that risk undermining the regime’s hold on power. The risk that cooperation would spiral out of control in ways that could threaten the survival of the regime is limited.

By contrast, in general-purpose IOs, delegation, pooling, and access not only generate functional benefits and sovereignty costs, but also increase uncertainty about the future direction of cooperation in ways that are differently appreciated by democracies and autocracies. When these design features are combined with open-ended contracts, they produce circumstances that are particularly conducive for dynamic policy developments. When contracts are open, gaps in state control resulting from delegation, pooling, and access may be exploited by supranational bodies, state majorities, and TNAs to push cooperation in directions not originally envisaged by member states (Hawkins et al. 2006; Lenz et al. 2023). We expect autocracies to be more fearful of such dynamic developments in cooperation than democracies.

For democratic leaders, reduced control over cooperation and unexpected policy developments are unfortunate, but seldom a cause of regime breakdown and often associated with benefits as well. Few political leaders appreciate relaxing control over future policy developments. However, the short-term concerns of democratic leaders—hoping to secure re-election but ultimately uncertain about future governments—often lead them to discount the long-term consequences of institutional design decisions (Pierson 1996, 135–6). Moreover, democratic leaders usually have fewer reasons to fear expansions in policy scope, partly because such expansions typically respond to demands for problem-solving, and partly because democracy as a system of rule is not threatened by cooperation in any particular area.

For autocratic leaders, reduced control over cooperation is a threat of a greater magnitude. It could make it more difficult to focus cooperation on areas that facilitate domestic repression, such as police cooperation and anti-terrorism measures (Debre 2021; Cottiero and Haggard 2023). It could also make it more difficult to secure the private goods needed to satisfy selectorates (Poast and Urpelainen 2013; Ginsburg 2021). Losing command of cooperation thus jeopardizes critical benefits from IOs and puts the basic contract between authoritarian leaders and their supporters in peril. An illustrative example is the EU’s introduction of rule-of-law conditionality to its funding and the subsequent decision to freeze funds to Hungary. EU funding has been an important means by which Viktor Orbán’s government satisfies supporters (Kelemen 2020, 490–1), but this cooperation benefit was reduced after the European Commission (delegation) triggered rule-of-law conditionality and the Council agreed by a qualified majority (pooling) to suspend funding (Kelemen 2025, 352).

By the same token, reduced control means that autocratic leaders could find it more difficult to keep cooperation away from policy areas that endanger their hold on power. Such domains include human rights and democracy promotion, which pose an immediate threat to autocratic regimes; cooperation on environmental and social issues, which provide a platform for opposition activists; and principles such as responsibility to protect, which offer a rationale for external interference in domestic affairs. A notable example of how autocratic states use intergovernmental institutional designs to prevent expansions into such policy areas is ASEAN, where democratic Indonesia’s push for human rights and democracy promotion has been circumscribed by the combination of a predominantly autocratic membership and an IO design emphasizing consensus and non-interference (Emmers 2014, 556–7).

Taken together, we therefore assume that democracies and autocracies hold more divergent preferences on institutional design in general-purpose IOs than in task-specific IOs. This leads us to expect that the regime composition of IOs will matter more in general-purpose organizations than in task-specific organizations. When states come together to take collective decisions on institutional design in general-purpose IOs, a more democratic membership should be more prone to decide in favor of more delegation, pooling, and access than a less democratic membership. By contrast, when states take collective decisions on institutional design in task-specific IOs, the regime composition of the membership should matter less for decisions on delegation, pooling, and access.

We will evaluate this expectation by focusing exclusively on changes in democratic density within IOs rather than differences in democratic density across IOs. This strategy serves several purposes. First, it allows us to better identify the impact of an IO’s regime composition on institutional design, since we hold the IO context constant and can disregard any confounding factors that only vary across IOs. Second, it helps to bracket the process whereby state preferences are aggregated into collective decisions on institutional design. While governance arrangements vary extensively across IOs (Blake and Payton 2015), with implications for how proportions of democratic and autocratic states can influence design decisions (Debre and Sommerer 2022), our exclusive focus on changes within an IO controls for cross-IO differences in the mode of preference aggregation. Third, it directs our attention to changes over time in the democratic density of an IO. Such changes can result from four types of events: democratization or autocratization among member states and entries or exits of member states. At its root, our argument is thus a dynamic account of how changes in the regime composition of memberships impact institutional design differently in general-purpose and task-specific IOs. We hypothesize:

 

Research Design

We evaluate our expectation in a sample of forty IOs for the time period 1950–2019. The selection of IOs and time period was guided by two concerns. First, we sought to ensure a representative sample in terms of general-purpose and task-specific IOs and global and regional IOs. To this end, we relied on a stratified sampling strategy where we selected IOs based on issue scope (general-purpose versus task-specific) and world region (global versus regional), adjusted based on overlaps between existing datasets (Hooghe et al. 2017; Sommerer and Tallberg 2017). By stratifying the sample, we seek to avoid biases that may result from focusing only on certain types of IOs. For instance, it may be the case that regional IOs or general-purpose IOs are associated with particular levels of institutional design and democratic density, which are unrepresentative of IOs in general. Our final sample includes seventeen general-purpose and twenty-three task-specific IOs, out of which twenty-one IOs have a global membership and nineteen IOs have member states drawn from specific world regions.³ Second, we sought to ensure that the sample covers a longer time period than existing research: 1950–2019. To this end, we rely on recent updates of existing datasets, and also updated existing data forward and backward in time where necessary. The selected time period ensures that our analyses cover several waves of democratization and autocratization, including recent instances of democratic backsliding.

To measure institutional design features of IOs, we rely on a combination of existing and new data on the level of delegation, pooling, and access in our forty IOs. These three dimensions together provide a comprehensive picture of IO design and capture the extent to which IOs are governed intergovernmentally or have moved toward more supranational (delegation and pooling) or transnational (access) forms of governance.

We measure delegation and pooling using the most recent version of the Measuring International Authority (MIA) dataset (Hooghe et al. 2017, 2019).⁴  Delegation captures “the allocation of authoritative competences to non-state bodies in an IO’s decision-making process” (Hooghe et al. 2017, 107–8) across agenda setting, final decision-making, and dispute settlement. Pooling captures “the extent to which member states share authority through collective decision making” (Hooghe et al. 2017, 113) in agenda setting and final decision-making, taking voting rules, bindingness, and requirements for ratification into account. Both pooling and delegation are measured using indices ranging from 0 to 1.

We measure TNA access to IOs using an updated and expanded version of the Transaccess dataset (Sommerer and Tallberg 2017). Access captures “the institutional mechanisms whereby TNAs may take part in the policy process of an IO” (Sommerer and Tallberg 2017, 248) and is measured at the level of IO bodies along the four dimensions of depth, range, permanence, and codification. The depth of access captures whether TNAs are deeply involved in the activities of an IO body, or if access is shallow in nature. The range of access captures the breadth of TNAs entitled to participate. Finally, permanence and codification capture the degree to which access is regularized and revocable, respectively (Sommerer and Tallberg 2017, 253–4). The index ranges from 0 to a theoretical maximum of 12. We updated Transaccess in accordance with the data collection instructions and codebook from the original project. The update includes an extension forward in time from 2010 to 2019 for several IOs, as well as the addition of three new IOs to the dataset. In our analyses, we capture access at the IO level by calculating its average across all bodies in a given IO year.

Our dependent variables vary both across and within IOs. Across IOs, some organizations, like the International Criminal Court, exhibit high levels on all institutional design variables, whereas other organizations, such as the Arab Maghreb Union, score comparatively low on all dimensions. Still others combine high levels on one or two design features with lower levels on other design variables. One example is the Economic and Monetary Community of Central African States, which combines comparatively high levels of delegation with lower levels of pooling and access. Within IOs, the most prominent development is an expansion over time on most or all of the three design dimensions, exemplified by IOs such as the Pacific Islands Forum and the Commonwealth. However, we also find several instances of decreases in delegation, pooling, or access, for example, in the United Nations Educational, Scientific and Cultural Organization (delegation), the UN (pooling), and the International Labour Organization (ILO; access).

We measure our main independent variable, the democratic density of IO memberships, using the average level of democracy among an IO’s member states. To this end, we combine data on IO memberships from the Correlates of War dataset on intergovernmental organizations (COW-IGO; Pevehouse et al. 2020) with V-Dem data on states’ level of electoral democracy (Teorell et al. 2019; Coppedge et al. 2022a; Pemstein et al. 2022).⁵ The electoral democracy index seeks to capture “the core value of making rulers responsive to citizens, achieved through electoral competition for the electorate’s approval” (Coppedge et al. 2022b, 43), by combining indicators measuring levels of suffrage, free and fair elections, freedom of expression, freedom of association, and the degree to which elections affect the executive’s composition. The index is commonly used in contemporary research on regime type and IOs (e.g., Cottiero and Haggard 2023; Meyerrose and Nooruddin 2023), and its emphasis on elections aligns with our theoretical focus on how variation in (s)electorates shapes regime preferences about international cooperation. Since the COW-IGO dataset only includes IO memberships up to 2014, we expanded the membership data forward in time to 2019 for our forty IOs.

Developments in democratic density vary across IOs. In most IOs, the tendency is for democratic density to increase over time. This is true for IOs such as the Organization of American States (OAS), the AU, and the World Tourism Organization (UNWTO). Nevertheless, we also witness several instances of overall decreases in democratic density, for example, in the European Association of National Metrology Institutes and the Bank for International Settlements since the mid-1990s. Most changes in democratic density are due to democratization or autocratization in the membership (69 percent) while remaining changes result from states joining or exiting an IO (31 percent).⁶ In some IOs, democratic density remains broadly stable over time, as in the World Trade Organization.

To capture our main conditioning factor, the governance purpose of IOs, we draw on the MIA dataset, which distinguishes between IOs that have an open-ended contract (general-purpose IOs) and IOs with a clearly specified contract (task-specific IOs) (Lenz et al. 2015; Hooghe et al. 2019, 49–50). The open-ended contracts of general-purpose IOs seek to attain broad-ranging cooperation among governments under weakly specified conditions and focus on the process rather than the outcome (Hooghe et al. 2019, 136). This is the case in IOs such as the EU, the UN, and ASEAN. In contrast, the closed-ended contracts of task-specific IOs seek to achieve a fixed objective in a given policy area under clearly specified conditions (Hooghe et al. 2019, 136). This is the case in IOs such as the North Atlantic Treaty Organization, the International Whaling Commission, and the Organization for Security and Cooperation in Europe. Since governance purpose concerns the basic set-up of an IO, it is generally stable over time. The only IOs in our data that change from one category to the other are the Intergovernmental Authority on Development/Intergovernmental Authority on Drought and Development (IGAD/IGADD), which transitioned from task-specific to general-purpose in 1996 when IGAD succeeded IGADD, and the Caribbean Community/Caribbean Free Trade Association (CARICOM/CARIFTA), which underwent a similar transition when CARICOM succeeded CARIFTA in 1973.

Results

What is the relationship between the democratic density and institutional design of IOs? In this section, we first map this relationship over time and then proceed to test our hypothesis about the moderating effect of governance purpose through multivariate statistical analyses.

Figure 1 shows the relationship between democratic density and institutional design across all IOs, general-purpose IOs, and task-specific IOs. Two key patterns stand out. First, the development in democratic density is similar across general-purpose and task-specific IOs. In both types of IOs, average democratic density has increased from around 0.45 in 1975 to 0.59 in 2019.⁷ Both general-purpose and task-specific IOs have also witnessed similar trajectories in their overtime developments, with an initial steady increase in democratic density in the 1970s and 1980s, followed by a sharp increase in the early 1990s, a plateau from the mid-1990s, and a continuation of that plateau or even a slight decrease in the 2010s. This pattern indicates that any difference between general-purpose and task-specific IOs in terms of institutional design is not due to differences between these two forms of IOs in terms of their respective levels of democratic density.

Second, there are notable differences between general-purpose and task-specific IOs with regard to the descriptive relationship between democratic density and institutional design. In general-purpose IOs, the rise in democratic density has been matched by a corresponding increase in delegation, pooling, and access. By contrast, task-specific IOs exhibit a comparatively stable level of delegation through the entire period, a small decrease in pooling, and a slightly weaker increase in access, despite similar developments in democratic density. This pattern thus provides provisional evidence consistent with our expectation that governance purpose moderates the relationship between democratic density and institutional design.

Are these diverging patterns between general-purpose and task-specific IOs coincidental or systematic? We proceed to test our hypothesis in a set of linear regression analyses while controlling for potential confounding factors affecting or coinciding with both democratic density and institutional design. Although this strategy cannot offer direct evidence of a causal relationship between democratic density and IO design, it has been the preferred empirical approach in existing research given the absence of exogenous variation in the democratic density of IOs (e.g., Tallberg et al. 2016; Hooghe et al. 2019). In combination with a set of robustness checks in which we subject our models to alternative specifications, this strategy allows us to move toward isolating the relationship between democratic density and institutional design.

First, we control for the regime heterogeneity of IO member states (cf. Tallberg et al. 2016). It may be the case that the difference in regime type among IO member states, rather than the level of democracy, produces variation in institutional design. The variable captures the standard deviation of the level of electoral democracy among an IO’s member states. Second, we control for the membership size of an IO (Pevehouse et al. 2020). Membership size has been shown to be associated with less intergovernmental institutional designs (Hooghe and Marks 2015), and if larger or smaller IOs also systematically have more democratic members, this might bias our estimates of democratic density. In line with existing research (Hooghe and Marks 2015; Hooghe et al. 2019), we take the logarithm of this variable to account for non-constant marginal effects. Third, we control for the affluence of IO memberships using the average gross domestic product (GDP) per capita of their member states and the GDP heterogeneity of IO memberships using the standard deviation of member states’ GDP (Feenstra 2015).⁸ These controls seek to ensure that our results are not driven by wealth differences between IO memberships. Fourth, we follow Hooghe et al. (2019, 98) and include a year count to control for common linear trends among IOs.⁹ Fifth and finally, we include IO fixed effects to control for potential time-invariant confounders. The inclusion of fixed effects implies that we restrict our analysis to identifying the relationship between democratic density and institutional design within IOs, whereas all between-IO variation is absorbed by the fixed effects. We lag all independent and control variables by 3 years, since changes in our dependent variables typically require member states to adopt new institutional rules for IOs—in many instances through constitutional revisions, which often are controversial and take time to negotiate. To account for potential serial correlation as a consequence of the panel structure of our data, we cluster standard errors at the level of IOs (Cameron and Miller 2015).

Table 1 reports the results of the regression models. Models 1–3 report the direct relationship between democratic density and delegation, pooling, and access. Models 4–6 mirror models 1–3, but also include an interaction term that captures how governance purpose conditions the relationship between democratic density and institutional design.

The regression models yield two important results. First, contrary to common expectations but in line with recent findings (e.g., Hooghe et al. 2019; Ginsburg 2021), we find no direct relationship between democratic density and institutional design. The coefficient for democratic density is not statistically significant in any of the models 1–3. These findings suggest that democratic improvements (retrenchments) in an IO’s membership are not, in general, associated with higher (lower) levels of delegation, pooling, or access. Second, there are statistically significant differences between general-purpose IOs and task-specific IOs in terms of the strength of the relationship between democratic density and institutional design. Across all three design dimensions, the relationship is significantly stronger in general-purpose IOs than in task-specific IOs. These findings suggest that democratic improvements (retrenchments) in an IO’s membership have different implications depending on the governance purpose of the organization.

To illustrate the conditioning role of governance purpose, Figure 2 plots the relationship between democratic density and institutional design separately for general-purpose and task-specific IOs. We show the marginal effect of a 0.1 increase in V-Dem’s electoral democracy index, since changes of this magnitude are a common threshold to identify episodes of democratization (Wilson et al. 2020) and autocratization (Lührmann and Lindberg 2019) in individual countries.

In general-purpose IOs, the marginal effect of democratic density on pooling and access is positive and statistically significant. For pooling, a 0.1 unit increase in the democratic density of an IO’s membership is associated with a 0.05 unit increase in the level of pooling. This is in the same order of magnitude as a shift from unanimous to supermajority decision-making in both the drafting and the final, binding decision on an IO’s budget. For access, a 0.1 unit increase in the democratic density of an IO’s membership is associated with a 0.14 unit increase in access. This corresponds to the introduction of an administrative rule that gives all TNAs access to one of five IO policy-making bodies, including a regular right to present statements at the body’s meetings. Taken together, the marginal effects suggest that democratic improvements (retrenchments) in a general-purpose IO’s membership are associated with substantively important increases (decreases) in levels of pooling and access.

The positive relationship between democratic density and institutional design in general-purpose IOs is illustrated by the OAS, where a democratization of the membership in the 1980s was followed by reforms in the 1990s that expanded both pooling and access. These revisions increased pooling by formalizing the bindingness of budgetary decision-making and by introducing supermajority voting for the suspension of member states whose democratic governments have been overthrown (Hooghe et al. 2017, 375–7), and they strengthened access by introducing several new modalities by which TNAs could participate in bodies like the Inter-American Council for Integral Development and the Inter-American Commission on Human Rights.

The marginal effect of democratic density on delegation in general-purpose IOs is similarly positive, but does not reach statistical significance at the 95 percent level. When taken together with the significant coefficient on the interaction term in Table 1, this result indicates mixed support for our expectation: While we can confidently conclude that there is a difference between general-purpose IOs and task-specific IOs in terms of the relationship between democratic density and delegation, we cannot with the same level of confidence say that this difference is due to a positive relationship in general-purpose IOs and a null or negative relationship in task-specific IOs. One potential explanation of this comparatively weaker marginal effect is that delegation as an institutional design feature is less closely associated with democratic governance than pooling and access. Whereas pooling and access connect to democratic ideals and practices like majority voting and civil society involvement (cf. Grigorescu 2015), delegation does not do so to the same extent.

In contrast to general-purpose IOs, task-specific IOs do not show the same positive relationship between democratic density and institutional design. The marginal effect of democratic density on delegation, pooling, and access is statistically insignificant. The Food and Agriculture Organization and the ILO provide illustrative examples of task-specific IOs where increased democratic density has not been followed by substantial reforms to institutional design. In the 1980s and 1990s, the average level of democracy in the memberships of these IOs increased substantially. Yet, in both IOs, the levels of delegation and pooling remained stable for the entire period following these increases, whereas the level of access only increased marginally.

In sum, our analyses generate two principal findings. First, there is no general positive relationship between more democratic memberships in IOs and less intergovernmental institutional designs. Consistent with our initial puzzle, we do not find a significant association between IOs’ democratic density and any of our three institutional design dimensions. Second, the relationship between democratic density and institutional design is conditional on the governance purpose of an IO. In line with our core expectation, the level of democracy in an IO’s membership consistently matters more in general-purpose IOs than in task-specific IOs. The evidence for this result is stronger for pooling and access than for delegation.

Robustness Checks and Extension

We proceed to estimate a set of models with alternative specifications to explore the strength and scope of our results. We do this by first conducting a series of robustness checks to assess the consistency of our findings, and then provide an extension by applying our models to a larger sample of IOs commonly used in previous research.¹⁰

First, we estimate models with alternative indicators for democratic density (Online Appendix Tables A3 and A4; Online Appendix Figure A2). While we rely on an index of electoral democracy in our main analyses, we recognize that there are other ways of conceptualizing democracy. In the alternative models, we construct the democratic density variable using indices for liberal, participatory, deliberative (Coppedge et al. 2015, 2022b), and egalitarian democracy (Coppedge et al. 2015, 2022b; Sigman and Lindberg 2015), as well as indicators that build on categorical understandings of democracy and capture the share of democratic IO members.¹¹ We also estimate models where we calculate democratic density using the polity2 index from Polity (Marshall and Gurr 2020) (Online Appendix Table A5; Online Appendix Figure A3). Finally, we draw on the Episodes of Regime Transformation dataset (Edgell et al. 2024) to estimate models in which the main independent variable is, respectively, the share of democratizing states, autocratizing states, stable democracies, and stable autocracies among an IO’s membership (Online Appendix Tables A6 and A7; Online Appendix Figures A4 and A5).

Second, we estimate models with alternative lags (Online Appendix Tables A8 and A9; Online Appendix Figure A6). Our main models lag all independent variables by 3 years, so we include models with a shorter lag of 1 year and longer lags of 5 and 7 years. All lags test alternative temporal relationships between democratic density and institutional design, but the models with longer lags have the additional benefit of also assessing potential endogeneity issues.

Third, we estimate models controlling for time trends in alternative ways (Online Appendix Tables A10 and A11; Online Appendix Figure A7). Our main models controlled for a linear time trend, but trends may also be non-linear. We first estimate alternative models using a post-Cold War dummy, cubic polynomials of time, and year-fixed effects. We then proceed to estimate models with separate time trends for general-purpose and task-specific IOs. The different institutional developments we observe in general-purpose and task-specific IOs may be an inherent difference between IOs of different types, rather than a differing response to changes in democratic density. Finally, we estimate a pooled ordinary least squares model to make use of variation not just within but also between IOs (Online Appendix Table A12; Online Appendix Figure A8).

Fourth, we estimate separate models for general-purpose IOs and task-specific IOs (Online Appendix Table A13). This is an alternative way of estimating the moderating role of governance purpose by splitting the sample instead of introducing an interaction term.

Fifth, we estimate models with alternative versions of our main sample (Online Appendix Tables A14 and A15; Online Appendix Figure A9). We test the sensitivity of our dependent variables by estimating models without all IO bodies where coders indicated a greater level of difficulty in the coding (access) and by estimating models without the two IOs not in the original version of the MIA dataset (delegation and pooling). We also test the sensitivity of our independent variables by dropping IOs where many member states have missing data on democracy or GDP per capita.

Sixth, we estimate models controlling for the extent to which IOs pool decision-making on constitutional reform (Online Appendix Table A16; Online Appendix Figure A10). Our theoretical argument focuses on the preferences of democratic and autocratic states, and thereby brackets the process whereby those preferences are aggregated. We account for that process by including a control for pooling on constitutional reform, since reforms to the institutional design of IOs often require treaty changes.

Seventh, we estimate models on the sample of task-specific IOs exclusively and interact democratic density with an indicator for whether an IO is active in policy areas that autocratic states may find particularly intrusive (Online Appendix Table A17). This robustness check addresses the potential suspicion that task-specific IOs that are active in sensitive policy areas could exhibit a similar dynamic to the one we find for general-purpose IOs. We rely on policy-area data from MIA (Hooghe et al. 2019) and operationalize intrusive IOs in three different ways: IOs active in policy areas that are core state powers, IOs active in human rights, and IOs with human rights as a core policy area.

Eighth and finally, as an extension of our analyses, we estimate our models for delegation and pooling using the full MIA dataset of seventy-four IOs between 1950 and 2010 (Hooghe et al. 2019) (Online Appendix Table A18; Online Appendix Figure A11).¹² For this sample, existing research has not been able to establish a direct relationship between democratic density and delegation or pooling (Hooghe et al. 2019, 98), contributing to our initial puzzle. To ensure that our argument about governance purpose helps to resolve this puzzle and that our findings are not just a consequence of our particular sample, we extend our models to the MIA sample and introduce an interaction term between democratic density and governance purpose.

Taken together, the alternative models strengthen confidence in our core results. Importantly, the coefficient on the interaction term between governance purpose and democratic density is positive and statistically significant across all but four of sixty-two alternative models. This indicates that there is a consistent positive difference in the association between democratic density and institutional design between general-purpose and task-specific IOs. The marginal effects of democratic density on institutional design, conditional on governance purpose, are also broadly in line with our main specifications, and reveal a positive relationship in general-purpose IOs and a null effect in task-specific IOs. Our main analyses and robustness checks combine to indicate that this result is most robust for pooling, slightly less robust for access, and least robust for delegation.

Conclusion

The dominant expectation in existing research posits that IOs’ regime composition should affect their institutional design in terms of delegation, pooling, and access. Yet empirical analyses have provided little support for this plausible conjecture. This article has sought to resolve this puzzle by providing a new understanding of how regime composition matters for the design of IOs. Theoretically, we have developed a novel argument about the moderating effect of governance purpose on the relationship between democratic density and institutional design, distinguishing between general-purpose and task-specific IOs. Empirically, we have evaluated this expectation through a comprehensive analysis of delegation, pooling, and access in forty IOs from 1950 to 2019, estimating whether and how changes in democratic density are related to changes in institutional design.

Our findings show that the relationship between democratic density and institutional design is conditional on the governance purpose of IOs. Consistent with our expectation, changes to the regime composition of IO memberships are more consequential for the institutional design of general-purpose IOs than for task-specific IOs. In general-purpose IOs, where contracts are open-ended and cooperation dynamic, democratic density is positively related to pooling and access, while estimates for delegation are positive but less precise. By contrast, in task-specific IOs, where contracts are closed-ended and cooperation static, democratic density is not related to institutional design. This finding helps to reconcile dominant theoretical expectations and contradictory empirical findings in earlier research by identifying the contingency of the impact of democratic density on IO design.

This article makes several important contributions to extant knowledge. First, it expands our understanding of how regime type matters for international cooperation, adding to a rich research tradition (e.g., Mansfield et al. 2000; Pevehouse 2005; Mansfield and Pevehouse 2006; Bättig and Bernauer 2009; Simmons and Danner 2010; Poast and Urpelainen 2013; Grigorescu 2015; Tallberg et al. 2016; Ginsburg 2021). While previous studies tend to assume an independent effect of regime type, we show how the relationship to institutional design is much more contingent than typically expected. Democracies are not generally more willing than autocracies to cede authority in international cooperation. But when the governance purpose of IOs renders the future of cooperation more uncertain, democracies and autocracies tend to diverge in their approaches to institutional design.

Second, this article contributes to the vibrant literature on institutional design in global governance (e.g., Abbott and Snidal 2000; Koremenos et al. 2001; Hawkins et al. 2006; Tallberg et al. 2014; Koremenos 2016; Hooghe et al. 2017; Voeten 2019; Zürn et al. 2021; Barnett et al. 2022). While this field tends to pay limited attention to the regime composition of IOs as a source of design compared to other explanations, we demonstrate that states are not like actors, with implications for delegation, pooling, and access in IOs. Conditional on the governance purpose of IOs, democratic and autocratic memberships make different choices about institutional design. Unless research on institutional design integrates the democratic density of IOs in its analyses, it risks missing a key source of transformation in an age of autocratization.

Third, on this note, this article can help us to understand when and how international cooperation might change as a result of autocratization. While most scholarship on regime type and international cooperation was produced at a time when liberal democracy was at its peak, the key development over the past decade has been its decline around the world (Lührmann and Lindberg 2019; V-Dem 2023). A growing body of research attempts to identify the consequences of autocratization for international cooperation (e.g., Debre 2022; Cottiero and Haggard 2023; Meyerrose and Nooruddin 2023; Winzen 2023; Cottiero et al. 2024). This article suggests that autocratization will have differential effects on IOs. While we can expect autocratization to negatively impact delegation, pooling, and access in general-purpose IOs, the implications should be less profound in task-specific IOs.

Funding

We acknowledge the generous funding provided by the Swedish Research Council (2021-01047), the European Research Council (101097437), and the Knut and Alice Wallenberg Foundation (KAW 2023-0360). This research was partially funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.

Author Biography

Jonas Tallberg is a Professor at the Department of Political Science, Stockholm University. His research focuses on issues of design, effectiveness, legitimacy, and power in global and regional governance.

Carl Vikberg is a researcher at the Department of Political Science, Stockholm University. His research focuses on international organizations, non-state actors, and the impact of regime type in global governance.

Notes

Authors’ note: Previous versions of this paper were presented at the Annual Meeting of the American Political Science Association, Los Angeles, August 31–September 2, 2023; the Annual Meeting of the Swedish Political Science Association, Gothenburg, October 4–6, 2023; the workshop “Non-Liberal States and Global Governance,” The Hague, December 13–15, 2023; and at talks organized by Lund University, Stockholm University, and Temple University. We are grateful to Christina Cottiero, Maria Debre, Lisa Dellmuth, Antoaneta Dimitrova, Andreas Duit, Daniela Donno, Alexander Dukalskis, Orfeo Fioretos, Felicitas Fritzsche, Johannes Geith, Alexandra Guisinger, Stephan Haggard, Gisela Hirschmann, Alexander Katsaitis, Faradj Koliev, Magnus Lundgren, Marina Nord, Ludvig Norman, Mark Pollack, Christina Schneider, Kilian Spandler, Jessica Stanton, Karin Sundström, Fredrik Söderbaum, Jan Teorell, Felicity Vabulas, Erik Voeten, Adam Ziegfeld, Ogi Zugic, and the three anonymous reviewers for ISQ for very useful comments. We also thank Thomas Sommerer for fruitful cooperation on data collection and Julia Feine and Ronek Jäär for excellent research assistance. The data underlying this article are available at the ISQ Dataverse at https://dataverse.harvard.edu/dataverse/isq.

Footnotes

References