Democratic memberships in international organizations: Sources of institutional design

Abstract

Domestic regime type has emerged a powerful explanation of multiple phenomena in world politics. This article extends this argument to the design of international organizations (IOs), where a profound development in recent decades is growing access for transnational actors (TNAs). While earlier research has shown that democracy in IO memberships helps to explain IO openness, we know little about the mechanisms that drive this effect. This article unpacks the relationship between democratic memberships and IO design by theorizing and assessing the impact of three different constellations of democracies on the openness of IOs. Empirically, we conduct a multivariate analysis of TNA access to 50 IOs from 1950 to 2010, combined with a case study of the Organization for Security and Co-operation in Europe. Our main findings are three-fold. First, democracy’s effect on openness is primarily a product of the combined weight of democracies within IOs and their resulting capacity to secure support for their polity preferences. Second, in contrast, we only find limited support for a specific influence of new democracies and democratic major powers on IO openness. Third, decision rules that allow for openness reforms to be adopted by a majority of member states facilitate and strengthen the influence of democracies, by reducing the ability of autocracies to block change. The findings have implications for our understanding of institutional design in global governance and democracy’s effects in world politics.

1 Introduction

A growing body of research privileges domestic regime type as an explanation of world politics. This explanation is most commonly associated with the democratic peace program (Russett 1993; Bueno de Mesquita et al. 1999). Yet, over the past 15 years, interest in regime type as an explanation has extended to a broad range of international political phenomena, including trade liberalization (Mansfield et al. 2000; Kono 2006), public-goods provision (Bueno de Mesquita et al. 2003; Bättig and Bernauer 2009), human rights protection (Moravcsik 2000; Simmons and Danner 2010), democracy promotion (Pevehouse 2005), international cooperation (Mansfield and Pevehouse 2008; Poast and Urpelainen 2013), and dispute settlement (Keohane et al. 2000; Davis 2012). These analyses are broadly consistent with a liberal approach to world politics through their emphasis on how democratic norms and institutions enable and constrain leaders and governments (Moravcsik 1997). Taken together, these contributions suggest that the nature of political regimes affects how states behave and organize in world politics. In this article, we extend and refine this argument as it applies to the design of international organizations (IOs).

One of the most profound shifts in the design of IOs in recent decades has been the expansion of institutional arrangements for the participation of transnational actors (TNAs), such as non-governmental organizations, philanthropic foundations, scientific communities, and multinational corporations (e.g., Raustiala 1997; O’Brien et al. 2000; Steffek et al. 2008; Jönsson and Tallberg 2010; Scholte 2011; Risse 2012; Tallberg et al. 2013). IOs increasingly consult with TNAs in the formulation of new policy, invite TNAs to make statements in decision-making bodies, involve TNAs in the implementation of policy, and engage TNAs as watchdogs of state compliance. This is a development that spans IOs in all world regions and policy areas, and that has been particularly intense since the end of the Cold War. According to one estimate, 56 % of IOs offered some form of access to TNAs in 1990, while more than 90 % did so in 2010 (Tallberg et al. 2013, 74). At the same time, already open IOs have become increasingly generous. For instance, the World Bank’s inclusion of civil society organizations in projects increased from 21 % in 1990 to 81 % in 2009 (World Bank 2014).

There is growing evidence that such TNA involvement matters for the legitimacy, effectiveness, and distributional outcomes of international cooperation (Risse 2012). Normative research highlights the contribution of TNA involvement to democratic legitimacy through its effects on transparency, participation, and accountability in IOs (e.g., Grant and Keohane 2005; Macdonald 2008; Steffek et al. 2008; Scholte 2011). Empirical research in the same area finds that civil society involvement positively affects public support for global governance (Bernauer and Gampfer 2013). Similarly, many studies establish that TNA involvement has positive consequences for the problem-solving capacity of IOs, by making IOs more efficient and effective in fulfilling their mandates (e.g., Raustiala 1997; Betsill and Corell 2008; Alter 2014).¹ However, research also suggests that TNA involvement can have distributional effects, since TNAs have interests too, and IO access provides them with additional channels to pursue these interests (Bouwen 2002; Sell and Prakash 2004; Betsill and Corell 2008; Klüver 2013).

Existing research points to three main explanations of the opening up of IOs: domestic democracy among IO member states, demand for the resources and services of TNAs, and diffusion of participatory governance as an organizational model (Raustiala 1997; Grigorescu 2007; Dingwerth and Pattberg 2009; Saurugger 2010; Steffek 2013; Tallberg et al. 2013). In this article, we focus in depth on the first of these factors, which has been shown to have a profound effect on the access of TNAs to IOs, but received no detailed analysis in earlier research. Existing analyses demonstrate that the average level of democracy in IO memberships constitutes the most powerful predictor of patterns in openness across organizations and over time (Grigorescu 2007; Tallberg et al. 2013). It is no coincidence that democratic clubs, such as the Council of Europe, were among the first IOs to open up; that the third wave of democratization preceded the growth in openness over the past two decades; and that IOs in world regions where democracy is strong offer most access to TNAs. Figure 1 offers an illustration of this relationship, highlighting how domestic democracy and IO openness have tended to co-vary over time.

Fig. 1

Domestic democracy and IO openness

The purpose of this article is to unpack this relationship. What drives democracy’s effect on the transnational design of IOs? Existing research has left this central question open, being content with establishing a relationship, while making no claims to distinguish between alternative causal pathways. Yet understanding the effects of domestic regimes on global governance is essential. By establishing how and why democracies shape IO openness, we can isolate the conditions under which states share authority with private actors, the logics that drive the institutional design of IOs, and the scope of democracy’s effects on international political outcomes.

We address democracy’s effect on IO openness by evaluating three distinct mechanisms, each linked to a specific constellation of democracies within IOs. The first mechanism suggests that openness originates from the polity preferences of democracies and these member states’ combined weight in IOs. The second mechanism stipulates that openness is the result of credible commitment motives among new democracies anxious to lock in domestic reforms, where TNAs can help to raise domestic and international audience costs from defection. The third mechanism attributes openness to the influence of democratic great powers with a disproportionate influence over the design of IOs. In addition, we analyze whether and how the institutional rules governing decision-making in IOs condition the influence of these alternative constellations of democracies on IO openness. Empirically, we assess and illustrate the explanatory power of these mechanisms through a nested mixed-method design (Lieberman 2005) that combines a multivariate analysis of TNA access to 50 IOs over the time period 1950–2010 with an in-depth model-testing case study of TNA access to the Organization for Security and Co-operation in Europe (OSCE).

We argue that democracy’s effect on openness primarily is a product of the combined weight of democracies within IOs, and their resulting capacity to secure support for their polity preferences. In our account, democracies and autocracies hold competing polity preferences, rooted in their respective domestic political systems. Governments in liberal democracies work to extend the constitutive principles of democracy to global governance. Openness toward civil society is one such constitutive feature of democracy, alongside accountability, participation, and rule of law. Since autocratic regimes do not share this commitment to democratic ideals, the likelihood of openness depends on the relative weight of democracies within IOs. The more IOs are dominated by democracies, the more likely it is that openness reforms will be adopted. The analysis shows that this relationship holds irrespective of the institutional rules governing decision-making within IOs. The strength of this relationship offers an important clue to the expansion of openness in global governance after the end of the Cold War: As the number of democracies in the international system increased through the third wave of democratization, this shifted the balance between autocracies and democracies within IOs, creating more favorable conditions for an extension of openness in global governance.

In contrast, we only find mixed evidence that new democracies and democratic major powers have been particularly influential in the expansion of openness, as theories of credible commitments and state power would lead us to expect. While the presence of new democracies or democratic major powers in an IO’s membership matters little on its own, it matters in conjunction with a largely democratic membership under conditions of majority rule. Hence, where institutional rules facilitate the adoption of openness reforms, having newly democratized states or major powers in the democratic membership helps to further expand openness in global governance.

The remainder of this article is organized in four parts. In the next section, we develop the three mechanisms linking democratic memberships and IO openness. We then proceed to the multivariate analysis, where we establish that the combined weight of democracies best accounts for IOs’ openness toward TNAs, and discuss the substantial significance of this effect. As a third step, we test the robustness of these findings by tracing how changes in the democratic density of the OSCE translated into openness reforms in the early 1990s. We conclude by briefly summarizing the findings and outlining the broader implications of the argument.

2 The argument: Domestic democracy and IO openness

Why would domestic democracy in IO memberships generate higher levels of openness toward transnational actors? We identify three distinct mechanisms, privileging polity preferences, credible commitments, and great power influence. Each mechanism is associated with a particular actor constellation: all democracies, new democracies, and democratic major powers.

First, the opening up of IOs may be driven by the polity preferences of the democratic component of an IO’s membership. This argument assumes that democracies and autocracies hold systematically different preferences on dimensions of IO design that tap into constitutive differences between democracies and autocracies. This logic is sometimes referred to as “liberal constructivism” (Risse-Kappen 1996) or “ideational liberalism” (Moravcsik 1997), as it derives the preferences that states promote internationally from domestic commitments to particular political values and institutions. It conventionally translates into the expectation that democracies seek to extend abroad the liberal ideals to which they adhere domestically.

In this view, openness toward civil society is a constitutive feature of democratic political systems, next to free and fair elections, freedom of expression, transparency, accountability, and rule of law (Tocqueville 1835). While democracies organize their relations with civil society in various ways, ranging from corporatist to pluralistic arrangements, interaction with autonomous societal actors is an integral aspect of all. Extending such democratic features to IOs is therefore not a radical step for democracies, since it involves applying the same procedural standards to all levels of political organization.

Existing research points to several examples of democratic states pursuing polity preferences derived from democracy as a system of government. Grigorescu (2007, 2010) finds that IOs dominated by democracies are more likely to institute mechanisms of transparency and accountability. Several scholars argue that the principle of rule of law leads democracies to push for and accept intrusive international dispute settlement (Kahler 2000; Keohane et al. 2000; Acharya and Johnston 2007). Related, Simmons (2009) finds that democracies are particularly willing to regulate rights protection internationally.

For autocracies, by contrast, openness toward TNAs is both foreign and dangerous. Processes of democratization have often been driven by civil society actors, mobilizing popular opposition, demanding regime change, and supporting alternative elites (Przeworski 1991; O’Donnell et al. 1996; Diamond 1999). Autocratic rulers are therefore anxious to maintain strict control over domestic civil society, either through repression or co-optation into the state apparatus (Bermeo and Nord 2000; Heurlin 2010). In this context, TNA access to IOs presents a channel whereby domestic opposition groups can bypass the control of the regime and join international allies in criticizing authoritarian policies (Keck and Sikkink 1998). Autocracies have every interest to cut such processes short and no interest to promote them. Examples of authoritarian attempts to restrict the international reach of domestic opposition groups include China’s curbs on human rights activists from Tibet, and Russia’s restrictions on Chechen civil society groups.

In this logic, the likelihood of democracies determining the institutional design of IOs is shaped by their relative weight in the membership, or what Pevehouse (2005: 46) refers to as the “democratic density” of an organization. It should be relatively easier for democracies to upload their polity preferences where they make up a larger share of the membership. By contrast, we can expect greater resistance in IOs where democracies make up a smaller share of the membership. Hence, we hypothesize:

• H1: The higher the democratic density of IO memberships, the greater the likelihood that IOs will be open.

Second, the opening up of IOs may be driven by credible-commitment motives among young democracies in IO memberships. This mechanism suggests that it is not the democratic component of an IO’s membership as a whole that matters (as above), but its newly democratized member states. These states may be old members that have transitioned to democracy, or new member states that join an IO upon democratization. The leaderships in these young democracies face particular incentives, since processes of consolidation take time, back-sliding remains a possibility, and government turn-over can derail reforms. To convince domestic and international audiences that democratic and policy reforms will stick, the new regimes attempt to “lock in” these reforms through political moves that will make defection costly or difficult.² One strand of this argument emphasizes how young regimes strive to lock in democratization internationally as a means of preventing future unravelling by non-democratic opponents (Mansfield and Pevehouse 2008). Another strand highlights how international commitments can help young democratic regimes to bolster economic or security-related reforms at home (Mansfield and Pevehouse 2006). In contrast to the first explanation, this mechanism assumes that there is no domestic consensus on political and policy reforms, and therefore a particular incentive for new democratic leaders is to tie their own hands and those of their successors.

Committing themselves to democratic principles of international organizations and treaties serves this purpose for new democracies. Pevehouse (2005) shows that membership in regional IOs, especially when conditional on domestic liberalization, can help new regimes to complete democratization by sending a strong signal that reforms will continue. Examples include the efforts of countries in Central and Eastern Europe to gain membership in the Council of Europe, the European Union, and the North Atlantic Treaty Organization after the fall of the Berlin wall. Similarly, Moravcsik (2000) establishes that the primary proponents of reciprocally binding human rights obligations in Europe in the post-World War period were the regimes of newly democratized states with an interest to stabilize the political status quo against non-democratic threats. By contrast, established democracies were tepid and selective in their support.

This mechanism is based on the premise that openness in IOs can serve the same instrumental purpose. By supporting an institutional design of IOs that invites TNAs to monitor state behavior, regimes of fragile democracies raise domestic and international audience costs from defection, and strengthen the credibility of their commitments to political and policy reforms. TNAs are useful as compliance watchdogs, since they often have privileged information about state behavior on the ground, and a willingness to act on it through shaming, complaints, and litigation. TNAs famously serve in this capacity in the monitoring of human rights worldwide, but also in areas such as environmental protection, trade, and democracy (Raustiala 1997; Simmons 2009; Kelley 2012). The costs from being exposed as a violator of international principles should be particularly high for young democracies, anxious to establish reputations as reliable members of the international community (Pevehouse 2005, 40; Milewicz and Elsig 2013).

TNA access may also provide information to domestic constituencies about whether their political leaders have fulfilled their promises. IOs “can convey information to voters about the behavior of their leaders, thus allowing voters to better judge their leaders” (Mansfield et al. 2002, 504). TNA access to IOs’ activities enables better distribution of this information, which domestic constituencies can use to punish new leaders if they were to renege on a promise. New democracies are more likely to lack the incentives and infrastructure to ensure adequate transparency domestically for constituents, increasing the value of TNA access as a channel of information on the credibility of commitments made by domestic leaders. We therefore hypothesize:

• H2: The higher the share of new democracies in IO memberships, the greater the likelihood that IOs will be open.

Third, democracy’s effects on IO openness may be attributable to the ideology and influence of a democratic major power. Unlike the first mechanism, this explanation rests on the premise that states have unequal influence over the design of IOs, based on differentials in political and economic resources. Major powers will exploit their greater capacity for influence by promoting institutional designs in conformity with their ideology. Other states accept these designs because they are compelled to do so by great powers. The expectation that relative power matters in design negotiations is well anchored in general theories of institutional design, as well as in a broad range of IR scholarship (Knight 1992; Mearsheimer 1994/95; Krasner 1991; Drezner 2007; Stone 2011).

In this vein, realists conventionally argue that democratic states employ liberal ideology to justify the pursuit of geopolitical interest. Donnelly (1986), for instance, submits that much of the explanation for the Inter-American Convention on Human Rights lies in the dominant power of the US, which exploited its hegemonic position to ensure the regime’s creation and support its operation. Waltz draws out the historical parallels of this logic: “Like some earlier great powers, we [the United States] can identify the presumed duty of the rich and powerful to help others with our own beliefs about what a better world would look like. England claimed to bear the white man’s burden; France spoke of her mission civilisatrice… For countries at the top, this is predictable behavior” (1979, 200).

The expectation that democratic great powers pursue their ideological preferences in international cooperation is often tested in empirical literature. Pevehouse (2005, 89–90) examines the possibility that successful promotion of democratization by regional IOs is not the product of the organizations per se, but reflects the influence of their most powerful members. Likewise, Moravcsik (2000) assesses the extent to which the origins of the European human rights regime are attributable to great powers extending their national ideals. However, in none of the cases does the evidence support this conjecture.

Extending this logic to IO openness, we would expect that democratic major powers advocate greater openness and are in a position to shape rules accordingly. These are states in which interaction with civil society actors is a constitutive component of domestic democratic ideology. In contrast, authoritarian major powers can be expected to resist such developments for the same reason. Where there is a single major power, the scope for this state to shape design decisions could be extensive. Where there are multiple major powers, the outcome should reflect the parties’ relative bargaining power. Thus, we hypothesize:

• H3: Where democratic major powers stand unchallenged by authoritarian major powers, IOs are more likely to be open.

Each of our three mechanisms privileges a particular constellation of democratic actors within IOs: all democracies, new democracies, and democratic major powers. The better represented these actor categories are within IOs, the more likely IOs are to be open. Since one actor category does not exclude another, it is conceivable that democracy’s influence on openness both reflects independent and combined effects of these constellations of democracies.

However, none of these mechanisms theorize the institutional setting within which democracies must secure support for their preferences. While some IOs require that all member states agree for openness reforms to be adopted, other IOs only require the support of a majority of member states (Blake and Payton 2014). The impact of institutional rules on decision-making is widely theorized in rational institutionalist scholarship (Scharpf 1997; Meunier 2000; Tsebelis 2002), and we expect such dynamics to be present here as well.

Since most IOs have been closed at the point of establishment, arriving at openness requires reform decisions by a sufficiently large group of member states. When such decisions only demand the backing of a single or qualified majority of member states, openness is relatively easier to achieve, since autocracies must gather a large blocking minority or even majority to stop reforms. In contrast, when decision-making is governed by consensus or unanimity, it is formally sufficient with only one or a few recalcitrant autocracies to stop reforms, given the veto power that all parties enjoy under this rule. In practice, the effect of institutional rules might be less dichotomous and more a question of distributing bargaining strength among the parties, thereby shaping the likelihood and terms of agreements. In this view, bargaining is a process that takes place in the shadow of decision rules, which condition the impact of actor constellations and preferences.

We expect this logic to pertain to the influence of all three categories of democracies in IO memberships. Everything else constant, it should be relatively easier for established democracies, new democracies, and democratic major powers to secure the support of the necessary number of member states when openness reforms can be adopted through majority voting. Hence, we hypothesize:

• H4: When decisions are taken by majority voting, democracies are more likely to be influential in institutional bargaining, and IOs more likely to be open.

3 Democracy and IO openness: Multivariate analysis

We now turn to a multivariate statistical analysis of the relationship between the democratic composition of IOs and TNA access under alternative decision rules. First, we introduce measurements of democracy in IO memberships and TNA access. In this context, we present a set of new indicators that capture the exact composition of democratic and autocratic IO membership. Second, we assess the explanatory power of the three mechanisms under different decision rules. Based on evidence from existing research, we start from the assumption that domestic democracy among IO member states is one of the main determinants of openness. While we control for alternative explanations, the main purpose of this analysis is therefore to empirically test which aspects of the democratic composition of IOs contribute to variation in openness. As part of this, we also present a series of robustness checks and look more closely into the most significant results.

3.1 Data and measurement

Beginning with the independent variables, we test Hypothesis 1 with a new measure of the democratic density of an IO that captures the share of democratic member states (Pevehouse 2005). For this measure, we combine information on state membership in IOs from the COW-IGO dataset with information on the democratic character of domestic regimes.³ For the latter, scholars have produced various measures and datasets over the past decades. While the Polity dataset is the most common, we use the Democracy-Dictatorship (DD) data with a binary measure of democracy (Cheibub et al. 2010). For our purposes, this measure has two main advantages compared to the alternatives. First, the DD data capture all countries between 1950 and 2008, including small states like Malta and Iceland that are not included in the Polity data. Having complete data on domestic regimes helps us to avoid a potential bias. Even if small states may not be the powerhouse of international politics, they still have a vote on the inclusion of TNAs, particularly in consensus-based IOs. Complete data on domestic regimes also allow us to assess whether there is a threshold share of democratic states—like a simple or two-third majority—that is necessary for democracies to shape the design of IOs through institutional bargaining, and to identify IOs with fully democratic or authoritarian memberships. Second, the percentage of democracies in an IO is a more meaningful measure than an average score of the member states’ level of democracy because an average does not offer any valuable clues on the distribution of democracies and autocracies. For example, if an IO’s membership has an average Polity IV score of 3, it could be composed of a combination of full democracies and authoritarian regimes—or it could be a club of anocracies. While we prefer the DD data for these reasons, our robustness checks include two variables based on the Polity IV scores of those member states of an IO for which this information is available: level of democracy and heterogeneity of regime. These measures are based on the mean and the standard deviation of the Polity IV scores, and capture the average level of democracy and the heterogeneity of regime types in an IO’s membership. In practice, we observe a high correlation between Polity IV and DD data (r = .96).⁴

We test Hypothesis 2 about the effect of credible commitment motives among new democracies with a variable that captures the share of young democracies among all member states of an IO. We calculate the percentage of member states—old and new—that have experienced a regime change from autocracy to democracy during the previous 4 years, which is the typical length of an electoral cycle. As with the first mechanism, we use the information on domestic regimes from the DD dataset to calculate the share of new democracies.

We test Hypothesis 3 on the influence of a major democratic power through a measurement that combines information on whether an IO has a major power in its membership with information on the domestic regimes of those powers.⁵ The result is a dummy for IOs that have at least one democratic major world or regional power, but no undemocratic major world or regional power that we can assume would veto access. In our robustness section, we control for the influence of the most powerful democracy—the United States—and test the reverse effect of a blockage of access reforms by non-democratic major powers.

Next to these variables, each capturing one mechanism and IO composition, we include two additional variables that measure changes to the composition of IOs. First, we include the variable democratization of old member states, which is a dummy indicating if at least one old member state has transitioned to democracy in the previous year. Second, we include the variable democratic accessions, which measures whether at least one democratic state has joined the IO in the previous year. Theoretically, these variables help us to disaggregate the effect of changes to the composition of IO memberships. Both types of events boost the democratic density of an IO (Hypothesis 1), but do they generate the same effect on IO openness? Likewise, both types of events contribute to boosting the share of young democracies (if the acceding democracy is newly democratized) (Hypothesis 2), but do they lead to the same effect on TNA access?

For information on voting rules (Hypothesis 4), we draw on a recent dataset that distinguishes between alternative voting rules for an IO’s supreme decision-making body (Blake and Payton 2014).⁶ Under unanimity, a single member state can block a proposal. Under majority rule, each vote has the same weight, and single member states cannot prevent the adoption of a decision. For most IOs in this category, the voting rule requires a simple majority.

Turning to the dependent variable, we measure IO openness on the basis of a new dataset on formal TNA access to 298 organizational bodies of 50 IOs over the 60-year period of 1950 to 2010.⁷ The IOs have been selected on the basis of a stratified random sample from a list of 182 international organizations, based on a revised and updated version of the Correlates of War IGO Dataset (Pevehouse et al. 2004).⁸ Our selection includes IOs from ten different policy domains and all world regions. We stratified the sample to ensure that the larger population of IOs was adequately represented. As a result, our sample includes major IOs, as well as lesser-known regional and specialized organizations. TNA access normally is regulated at the level of IO bodies, and previous research has shown that access varies significantly within IOs (Tallberg et al. 2013, 56, 68-70). Only few IOs open up all bodies at the same time. For example, the World Bank administration established formal access rules in 1981, whereas other World Bank bodies in our sample followed much later (Inspection Panel in 1993; Board of Governors in 1999) or never (Board of Directors). Even though our main independent variables do not vary within IOs, openness clearly does, as do many important control variables. For this reason, we operate with the IO body as the unit of analysis.

The information included in these data refers to formal rules of access, granted to TNAs through decisions either by the member states or the bureaucracy of an IO. Formal rules include provisions in treaties, protocols, and conventions, as well as regulations in rules of procedure and administrative decisions. The data have been collected on the basis of documents from archives and databases, and where necessary, through direct data requests to the relevant IOs.

Our dependent variable is a binary measure of openness that captures whether an IO body has an arrangement for TNA participation in place or not. This measure reveals that institutional arrangements for TNA access have become much more common over time. Between 1950 and 1989, mainly bodies of European, American, and global IOs were open. Pioneers were bodies such as the Economic and Social Committee of the United Nations (UN) and the European Court of Human Rights. After the end of the Cold War, the adoption of access arrangements took off dramatically, and this development spanned all issue areas and world regions. For 2010, 208 out of 294 bodies in our sample were open to TNAs.

In addition, we control for the influence of non-regime-based explanations. These variables are drawn from a recent study that examines a broad range of explanations for variation in TNA access (Tallberg et al. 2013).⁹ We assess the effects of a participatory discourse in global governance by an indicator based on references in scientific and non-scientific publications (Charnovitz 1997; Saurugger 2010). Moreover, we test whether the UN has functioned as a norm entrepreneur in the spread of TNA access through the indicator UN conferences (Clark et al. 1998; Willets 2000). We also include variables to test whether challenges by civil society actors make IOs more likely to open up, as a way of securing organizational legitimacy (O’Brien et al. 2000; Kissling and Steffek 2008). We measure media references to protests against IOs, estimate the effects of protests against similar IOs, and test whether contentiousness make IOs more likely to open up through the variable media coverage. We use three indicators to assess the rational functionalist argument that TNA access is influenced by the nature of each IO body’s governance tasks (Raustiala 1997; Tallberg et al. 2013). We expect a positive effect on access for IO bodies that engage in tasks that are technically complex, require local activities, and present significant non-compliance incentives. We measure the influence of sovereignty costs associated with TNA access through two variables, assuming that such costs are higher for decision-making bodies and bodies in the field of security (Hawkins et al. 2006; Bradley and Kelley 2008). To assess whether IOs with extensive resource deficits will have stronger incentives to involve TNAs, we include the variable IO budget (Liese 2010, 97). To evaluate whether the degree of political conflict among member states affects the provision of TNA access, we include the variable affinity of member states (Raustiala 1997, 731; Kahler 2005, 29). Finally, we assess whether TNA supply affects the extent to which TNAs are granted access to IOs (Reimann 2006; Steffek 2013).

3.2 Statistical analysis

The multivariate analysis lends strong support to the argument that democracy’s effect on IO openness is driven by democracies’ ability to secure support for their polity preferences. Our analysis shows that TNA access is shaped by the relative weight of democracies within IOs (Hypothesis 1). We also find that the presence of a major democratic power in the membership of an IO had a positive effect on IO openness during the Cold War, but not after 1990 (Hypothesis 3). In contrast, the analysis offers limited evidence in favor of a particular role for new democracies (Hypothesis 2). Last, we show that decision rules matter, with majority rule strengthening the effect of and interaction between different categories of democracies (Hypothesis 4).

As the dependent variable is binary, we use logistic regression to estimate the effects of democratic memberships on IO openness (Table 1). The dependent variable and most control variables vary at the level of IO bodies, so that we estimate the equation

$$ \begin{array}{c}\kern1em \mathrm{Logit}{\left(\mathrm{Access}\right)}_{\mathrm{i}\mathrm{x},\mathrm{t}}=\propto +{\beta}_1 Democratic\ densit{y}_{i,t-1}+{\beta}_2 Young\ democracie{s}_{i,t-1}+{\beta}_3 Democratic\kern1em \\ {}\kern1em powe{r}_{i,t-1}+\gamma {\mathrm{W}}_{\mathrm{i}\mathrm{x},\mathrm{t}-1}+\delta\ {\mathrm{Z}}_{\mathrm{i},\mathrm{t}-1}+\mathrm{t}+{\mathrm{t}}^2+{\mathrm{t}}^3+{\mathrm{e}}_{\mathrm{i},\mathrm{x}},\kern1em \end{array} $$

where Logit(Access) denotes the logit function of the likelihood that a body x in IO i provides TNA access, W is a vector for control variables that vary at the level of IO i’s body x, Z is a vector for control variables that vary at the level of IO i, t, t2 and t3 are our controls for time dependence, and ε _i,x is the error term. We cluster the data at the body level to account for potential dependence between units, employ robust standard errors in all models, and enter all time-variant independent variables with a lag of 1 year if no other specification is mentioned.¹⁰

Table 1 Logit regression of TNA access

Model 1 includes one variable for each mechanism, along with the non-regime type control variables from existing research on TNA access that help us to avoid omitted variable bias. We then look into interactions between the three mechanisms (Models 2–4) and re-estimate Models 1 to 4 for IOs with majority rule (Models 5–8) and IOs with consensus rule (Models 9–12). We evaluate the robustness of the results from Table 1 in four different ways (Table 2). First, we control for the time period, estimating the effects of the three mechanisms for the period before 1990 (Model 13) and after 1990 (Model 14). Second, we test if our results are influenced by a European norm by excluding IOs with a majority of European member states (Model 15). In addition, we assess if the results hold with alternative measures for the three mechanisms (Models 16–18). Finally, we control for alternative specifications of the model. We substitute control variables with IO dummies (Model 19), aggregate the data from IO body level to IO level (Model 20), and run a conditional fixed-effects logit (Model 21). We report the results in three steps, first describing direct effects, then interaction effects, and finally marginal effects.

Table 2 Logit regression of TNA access – Robustness checks

The analysis shows that the democratic density of an IO’s membership has a strongly positive and highly statistically significant effect on the likelihood that an IO body will provide access to TNAs. This finding holds across all models and also for different specifications (Model 19–21) and lends firm support to Hypothesis 1. The greater the overall share of democracies in an IO’s membership, the more likely that democratic polity preferences guide the design of IOs in terms of TNA access. We observe a strong positive effect for IOs with both majority and consensus rule (Models 5 and 9), which speaks in favor of the strength of this factor. We further find that democratic density had a significant effect on IO openness both during and after the Cold War (Models 13 and 14), and that this effect is not reducible to European dominance (Model 15). Finally, we establish that the effect of democratic density holds for the alternative measure based on the average Polity IV score for each IO as well, and that the heterogeneity of political regimes in an IO does not affect openness (Model 16). The latter finding suggests that it is not the mix of regime types in a membership, in terms of varying degrees of democratic consolidation, that matters, but whether member states are democracies or not.¹¹

We find mixed evidence for an effect of democratic major powers on IO openness (Hypothesis 3). While there is no effect for the full sample of IOs (Model 1), a division of IOs into majority and consensus organizations reveals that democratic major powers have a significant effect on openness in the former (Models 5 and 9). When democratic major powers only need to secure the support of a majority of member states, they have been more influential, endorsing Hypothesis 4 while challenging claims in existing literature that great powers are equally or more influential in consensus IOs (Stone 2011). Furthermore, we find that democratic major powers were consequential for IO openness during the Cold War, but have not been so in the post-Cold War era (Models 13 and 14). During the Cold War period of East–West tension, the support of major democracies, such as the US, France and the United Kingdom, was thus conducive to greater IO openness toward TNAs. We also assess whether the effect of a democratic major power on IO openness primarily is a consequence of US leadership (Model 17). While only two IOs in the sample have this composition, the result is positively statistically significant.

We find least support for the third mechanism, privileging credible commitment motives among new democracies. The share of young democracies is only weakly statistically significant after the Cold War (Model 14), and not in any other model assessing direct effects. This suggests that credible commitment motives among new democracies may have had a positive effect on TNA access after 1989, but cannot account for the large part of democracy’s effect on IO openness. Instead, the most important consequence of new democracies is probably their positive contribution to the overall democratic density of IOs (Hypothesis 1).

The results on the two measures of change in the democratic composition of IO membership help us to further interpret these findings (Model 18). The results show that recent democratization of old member states, but not the accession of democracies, increases the likelihood of TNA access. Democratization of old member states is highly statistically significant, while democratic accessions is not significant. This result suggests that there is something about the democratization of authoritarian member states that affects the likelihood of openness differently than democratic accessions, despite identical effects on democratic density.¹² We suspect that it might be the unlocking of access reforms previously blocked by the democratizing member state. Hence, democratization of existing member states may have a dual effect on IO openness: it removes their resistance to reforms already considered, and it boosts the strength of the democratic camp, thus improving the odds for future openness reforms.

Shifting to interaction effects, the results show that the three mechanisms combine in generating higher levels of openness. However, these positive and significant interaction effects only hold for IOs with majority rule, not consensus IOs, generating support for Hypothesis 4. While the presence of democratic major powers and young democracies had partial or no effects on IO openness on their own, they have significant and positive effects when interacted with democratic density (Models 6 and 7). Substantively, this means that different constellations of democracies have mutually reinforcing effects on TNA access to IOs, as long as openness reforms only require the support of a majority of member states. The already significant effect of a democratic membership on openness increases further with the number of young democracies and the presence of a democratic major power. This suggests that young and powerful democracies function as openness entrepreneurs in the broader group of democracies, but only are successful in this role when IOs are relatively democratic and ruled by majorities.

Several of our non-regime control variables display significant results as well. We find a strong positive and significant effect of non-compliance incentives on TNA access to bodies engaged in monitoring and enforcement for the full period and after 1990 (Model 14). Together with a positive and significant effect of local activity after 1990, this finding corroborates research that emphasizes functional reasons for private access to IOs after the end of the Cold War. The analysis also offers firm support for an effect of sovereignty costs. The variable decision-making has a significant negative effect on TNA access before 1990 (Model 13), while this effect disappears in the later period, as decision-making bodies, too, have begun to open up. In addition, there is a strong significant, negative effect for security across all time periods. The effect of IO budget is also significant, suggesting that IOs with larger resources are more likely to grant access. We also find a positive effect of TNA supply for the period after 1990, suggesting that the existence of a large NGO population has influenced the likelihood of access during the past two decades. We observe a positive effect of protests against IOs, and a negative effect for participatory discourse on TNA access before 1990. Finally, our results show that the effects of some control variables vary with decision rules. For instance, we see a negative effect of IO budget and a positive effect of UN conferences for majority-rule IOs (Model 4–6). The consistently positive effect of democratic density in both majority- and consensus-based IOs stands out even more in light of this variation.

We now focus in depth on the most significant finding by exploring the substantial significance of democratic density on IO openness over time and under different decision rules. Figure 2 displays the democratic density of the IOs in our sample over time, as well as separate measures of the share of IOs with a majority of democratic member states, only democratic member states, and no democratic member states. The democratic density was relatively high from 1950 to the mid-1960s, decreased abruptly in the late 1960s and early 1970s, before it began to grow again, eventually leading up to a share of 70.2 % in 2008. The democratic density at the end of the observation period only marginally exceeded the level at the beginning of the period. Hence, in order to fully understand the effect of democratic density, we need to disaggregate this variable. Figure 2 shows that the share of IOs with a majority of democratic members increased significantly between the late 1980s and 2000, while the shares of purely democratic or autocratic IOs remained fairly constant over time. This pattern suggests that the central change to the composition of IOs, captured by the variable democratic density, pertains to the control of IOs by a majority of democratic states.

Fig. 2

Democratic density of IOs in sample

Figure 3 goes one step further by demonstrating how the marginal effect of democratic density on IO openness varies in strength at different levels of democratic density. For this purpose, we plot the marginal effects of democratic density on our binary measure of TNA access for the full observation period (based on Model 1), for the periods before and after 1990 (based on Models 13 and 14), and for different majority and consensus rule IOs (based on Models 5 and 9). The figure for the full time period indicates that the strength of the marginal effect grows more or less continuously. Yet the slope indicates a stronger increase in the marginal effect between 30 and 60 % democratic density, suggesting that the command of a majority of member states is particularly important for democracies to influence IO openness. As we move from 0 to 100 % democratic density in an organization, the likelihood of that IO being open doubles.

Fig. 3

Marginal effects of democratic density on TNA access. Note: The gray lines display the margins of the 95 % confidence intervals

The figures for the two sub-periods offer a clearer picture of these effects. In the period 1950 to 1990, the marginal effect of democratic density on IO openness increased continuously and was particularly strong in organizations with a majority of democratic members. Hence, during the Cold War, the democratic composition of IOs mainly mattered when IOs were made up of at least 50 % democracies; at lower levels of democratic density, the effect of this variable was weaker. In the category of IOs with a democratic majority in 1980 (19), those with a democratic density over 95 % had a much higher likelihood of having at least one open body. The Organization for Economic Co-operation and Development (OECD) offers an over-time illustration: as the share of democratic member states grew from 86 to 100 % during the 1970s, the number of open bodies almost doubled.

After the end of the Cold War, the marginal effect of democratic density is stronger. This holds for all levels of democratic density, with the exception of IOs with more than 95 % democratic membership, where the effect is almost the same as before 1990. During this period, the effect of democracies is strong already at lower levels of democratic density. However, the effect does not increase as steeply with higher proportions of democracies in an IO’s membership as before 1990. Hence, after 1990, a smaller group of democracies in an IO is sufficient for the likelihood of access to increase substantially. While it still matters if IOs are populated by a majority of democracies, the effect is more even and it matters less if they were entirely composed of democracies. The 1990s were a period when many global IOs reached a level of more than 50 % democratic member states. This was paralleled by growing openness toward TNAs. The World Bank, for instance, went from a democratic density of 47 % in 1990 to 55 % in 2000, and simultaneously tripled the share of open bodies.

The figure depicting the marginal effects of democratic density in majority and consensus rule IOs underscores the result from the regression analysis in terms of clear positive effects in both types of IOs. However, as opposed to the regression analysis, the figure also points to a notable difference between majority and consensus IOs, supporting Hypothesis 4 on the importance of decision rules. While there is no difference in the marginal effect at lower levels of democratic density, and the direction of the effect is the same, the effect increases more steeply in majority IOs than in consensus IOs once democracies make up more than 50 % of the membership. This suggests that the threshold for agreeing on openness reforms matters: once democracies make up about half of the membership, the likelihood of a majority-rule IO being open increases significantly. Underscoring the same point, the figure also shows that the strength of the marginal effect is the same at a democratic density of 60 % in majority IOs and 100 % in consensus IOs. In other words, it takes an entirely democratic membership for democracy to have an equally strong effect on openness in consensus IOs as it has with 60 % democracies in majority IOs. When combined with the pattern showed in Fig. 2, these results demonstrate that majority relationships matter in two important ways for the expansion of openness in global governance: on the one hand, the most central change in the democratic composition of IOs over time is the growing share of IOs where democracies make up a majority of the membership; on the other hand, the effect on openness is particularly strong in majority-rule IOs where democracies make up at least 50 % of the membership.

4 Case study: Democracy and openness in the OSCE

The statistical analysis showed that the democratic composition of IOs primarily affects openness through democracies’ combined weight in these organizations, rather than the presence of young or powerful democracies. In this section, we complement the statistical analysis with a case study of democracy and openness in the OSCE, in accordance with principles of a nested mixed-method design (Lieberman 2005; see also Gerring 2007, 91–93). In a model-testing small-N analysis (SNA), the purpose is to supplement the large-N analysis (LNA) with in-depth study of an “on-the-line” case as a way of demonstrating the robustness of the causal argument. “As the goal is to complement the LNA, the use of SNA in nested analysis should aim to gain contextually based evidence that a particular causal model or theory actually ‘worked’ in the manner specified by the model” (Lieberman 2005, 442). Such a case study increases the confidence in the statistical finding and provides a check on spurious correlation.

The OSCE is an appropriate case for these purposes. It is drawn from the broader sample of 50 IOs and conforms well to the theoretically predicted and statistically established relationship between democratic density and openness. While founded with a mixed membership during the Cold War, the organization subsequently experienced a democratization of the membership that greatly increased its democratic density. This transformation coincided with an extensive expansion of openness to TNAs. Given this pattern, the function of the case study is to show how these factors are causally related as predicted.¹³

The OSCE was created during the détente period of the early 1970s as the Conference on Security and Cooperation in Europe (CSCE) to serve as a multilateral forum for dialogue and negotiation between East and West. Comprised of 35 states, the CSCE was divided between a group of democratic states, including the US, Canada and West European states, and a group of non-democracies, including the USSR and its aligned states. During the Cold War period, TNA access to the CSCE was extremely limited. The lack of openness applied equally to all policy fields that the CSCE addressed, which included political and security issues, economic and environmental issues, and humanitarian issues. While the Helsinki Final Act of 1975, the foundational document of the CSCE, stated that “institutions, organizations and persons have a relevant and positive role to play” (CSCE 1975), the only access to the CSCE was for “press and guests of delegations” to observe plenary opening and closing sessions.¹⁴

The lack of openness was not caused by an absence of proposals for more TNA involvement. Rather, throughout the 1970s and 80s, democratic states (Western bloc) often pushed for greater access for TNAs. Yet, non-democratic member states (Eastern bloc) blocked such proposals from being adopted. For example, in the negotiation of the Helsinki Final Act, Western states pushed for a more far-reaching agreement on the involvement of TNAs, while the Eastern bloc preferred the more vaguely worded provision eventually adopted (Bloed 1990, 10). Likewise, during the 1977–1978 Follow-up Meeting of Belgrade, Western states proposed to “strengthen the role of private citizens in the process of détente in respect of human rights and humanitarian issues. In contrast, the East European States stressed the intergovernmental nature of détente…” (Bloed 1990, 13). Yet another example is the 1985 Meeting of Experts on Respect for Human Rights in Ottawa, when Western diplomats tried to push for as many open sessions as possible, which Soviet diplomats opposed.¹⁵

The Eastern bloc’s resistance to TNA access was triggered by trepidation that access would enable opposition to the regimes to gain greater traction. The Eastern bloc had already found that the Helsinki Final Act contributed to the formation of independent NGOs in the Eastern bloc (Tudyka 2001; Brett 1993; Bloed 1990). Many of these groups invoked the Helsinki Final Act as a symbol of human rights in their resistance to the regimes (Bloed 1990, 12; Tudyka 2001, 465). As a result, Eastern states believed TNA access to the CSCE would embolden NGOs who were viewed as traitors, and thus blocked the proposals for greater openness put forward by Western democratic states. The difference of view on TNA access “led to heated debates and…delayed preparatory meetings, particularly in the period 1980 to 1986, at times placing the whole CSCE edifice in the balance” (Heraclides 1993, 119).

The absence of TNA access during the Cold War era was largely a consequence of the distribution of democratic and non-democratic states within the CSCE, given that democracies and non-democracies held differing preferences on openness. While democracies were proponents of access to the CSCE, they jointly lacked the combined weight to determine the institutional design under a rule of consensus. To illustrate, Fig. 4 depicts the balance between democratic and non-democratic states within the OSCE membership over four cross-sections of time. In 1980, the distribution of democracies versus non-democracies was relatively equally balanced. Since the CSCE operated by consensus, any concerted objection from non-democratic states could prevent the adoption of Western proposals.

Fig. 4

The democratic composition of the OSCE

The end of the Cold War brought a transformation of the access opportunities for TNAs. A wave of reforms in the 1990s made the organization much more open to TNAs, especially as the CSCE transformed into the OSCE and developed an organizational structure. The shift toward a more open OSCE was marked by the adoption of the 1990 Charter of Paris. The Charter affirmed the commitment to facilitate the activities of NGOs, other groups, and individuals “for the implementation of the CSCE commitments by the participating States,” and that these groups “must be involved in an appropriate way in the activities and new structures of the CSCE” (CSCE 1990a). At the Copenhagen Meeting of the Conference on the Human Dimension of the CSCE later that same year, the participating states agreed on new rules, which granted TNAs access to the conference premises, unimpeded contacts with state delegates, accreditation of media representatives, access to official conference documents, and opportunities to distribute relevant information to delegates (CSCE 1990b).

Further reforms in the 1990s included the creation of an NGO unit within the Office for Democratic Institutions and Human Rights (ODIHR) and permission for NGOs to speak at the first Human Dimension Implementation Meetings (HDIM) in 1993 (Brett 1996, 683). TNA access also expanded to the field of security. Field missions became accessible to TNAs in the mid-1990s through the “Platform for Co-operative Security,” which included mechanisms for regular exchange of information, appointment of liaisons, and development of common projects and training efforts (OSCE 1999, 44). By the end of the 1990s, the OSCE had become a relatively open IO, providing TNAs with some form of access to most of its organizational bodies and all policy domains.

This opening up of the CSCE/OSCE in the 1990s was driven by a transformation of the organization’s membership following the collapse of Communism. Democratization in the former Eastern bloc, and a splintering of the former Eastern bloc into independent states, meant that the CSCE became comprised of mostly democratic states. As Fig. 4 illustrates, the group of democratic states made up a vast majority of the membership in 1990, while the group of non-democracies dwindled to a very small portion. Overall, the 1990s was the period of greatest democratic density for the CSCE/OCSE, and this corresponded with the opening of the OSCE.

Collectively, democracies were now sufficiently strong to set the institutional agenda of the CSCE/OSCE. While old democracies remained as some of the most fervent supporters of TNA access, new democracies joined them in their preference for openness. For example, during the Helsinki Follow-Up meeting in 1992, two proposals for TNA access were put forward by Austria and the UK.¹⁶ Each of these proposals envisioned a more open OSCE and was co-sponsored by both strong democratic states and newly democratizing states. Austria’s proposal was co-sponsored by the former Czechoslovakia, Hungary, Finland, Norway, Sweden, Russia and Lithuania, while the UK’s proposal was co-sponsored by Germany, Italy, Spain, Belgium, Ireland, the Netherlands, and Georgia (Heraclides 1993, 122–123). In the end, the Helsinki meeting decided to extend to all CSCE meetings the previously established guidelines for NGO access to human dimension meetings, and to “instruct Directors of CSCE institutions and Executive Secretaries of CSCE meetings to designate an ‘NGO liaison person’ from among their staff” (CSCE 1992, 17–18).

The debate over NGOs was no longer about whether they should be granted access, but on what terms (Brett 1992, 22). Rhetoric of respect and appreciation for NGOs had supplanted the Soviet bloc’s earlier suspicion of NGOs. The shift in polity preferences among new democracies was facilitated by the arrival of a new set of leaders. Leaders from the former opposition movements were now often central figures in the new democratic governments. Prominent examples include Vaclav Havel in the Czech Republic and Lech Walesa in Poland. This shaped the perspectives of the new governments on TNA access. For instance, at the Helsinki meeting in 1992, the Armenian Foreign Minister explained that he was formerly an NGO representative who recognized the value of having a seat at the table and what it is like to try to have a voice in the process (Brett 1992, 22). Opposition movements had for a long time been strong advocates of principles of democratic rule at the national level, which fuelled their desire as new state officials to support openness in the OSCE.

The ability of democracies to push through reforms for more access came to a halt in the early 2000s, when the democratic density of the organization declined. As Fig. 4 illustrates, the portion of democratic states in 2000 was lower than in 1990. As some member states backslid in their transition to democracy (e.g., Russia, Belarus, Tajikistan, and Uzbekistan), they began to question TNA access again. For instance, backsliding states have at times sought to exclude specific NGOs from participation (Bloed 2007; Bakker 2007). A case in point is Russia’s attempt to block participation of the Russian-Chechen Friendship Society (Bloed 2007). Multiple examples point to TNA access being a source of conflict between East and West. Yet the democratic density of the OSCE has remained high enough to prevent retrenchment in TNA access.

To summarize, the OSCE illustrates well the causal process linking democratic density and openness in IOs. During the Cold War, the relative weight of democracies in the IO was too low for them to impose their polity preferences on recalcitrant autocracies. As the composition of the organization’s membership changed after 1989, and new democracies sided with old democracies, extensive openness reforms became possible.

5 Conclusion

IOs have become increasingly open to TNAs over recent decades. At the same time, significant variation remains across IOs in the extent to which they engage with TNAs. Existing research suggests that domestic democracy in the membership of IOs is the most powerful determinant of these patterns, but has refrained from exploring the mechanisms that could explain this effect. This article closes this gap by theorizing and empirically assessing how the democratic composition of IOs matters for openness, based on a combination of statistical analysis and a case study of the OSCE. Our conclusion briefly summarizes the findings of the analysis and expands on their implications.

We find that democracy’s effect on IO openness primarily is the product of the combined weight of all democracies in an IO’s membership and their resulting capacity to secure support for their polity preferences. Democratic member states have worked to extend the liberal democratic ideals they value domestically to the sphere of global governance. The likelihood of success has been shaped by the share of democracies in the overall membership. During the Cold War, openness reforms were difficult to push through, as many IOs were composed of democracies and autocracies in relatively equal numbers, and competition among political systems ran high. For democracy to have a strong effect on IO openness, it took a highly democratic membership. As a result, most of the forerunners among open IOs were democratic clubs, such as the Council of Europe, the Nordic Council, and the OECD. After the end of the Cold War, democracies have been considerably more successful. Not only do democracies make up a larger share of IOs nowadays; in addition, it takes less democratic density in an IO for democratic polity preferences to have a strong impact. After 1989, it has often been sufficient for democracies to gather a majority in order for them to successfully upload democratic standards of openness to IOs. As a result, we have seen the creation and expansion of openness arrangements in most IOs in the world, save those composed exclusively of authoritarian states, such as the Arab-Maghreb Union and the Organization of Arab Petroleum Exporting Countries. Our case study of the OSCE illustrates this development by tracing how democratization in Central and Eastern Europe strengthened the democratic camp within the IO at the expense of its authoritarian component, thereby paving the way for openness reforms.

These findings stand in contrast to the results for the two other mechanisms. While new democracies have been particularly eager to sign human rights treaties and to join IOs out of credible commitment motives, they have not played a profound role in the creation and expansion of IO openness, despite the expectation that TNA access is conducive to monitoring the domestic and international commitments of new democracies. This does not mean that new democracies do not matter; they do, but primarily through adding weight to the democratic camps of IOs, thereby increasing the likelihood that openness reforms can be pushed through. Likewise, the presence and influence of democratic major powers in IO memberships is only granted partial support in our analysis. While this factor was decisive during the Cold War, when US leadership pushed IOs toward greater openness, having a democratic state as dominant power has mattered little after 1989, when most of the expansion of TNA access has taken place.

Our analysis further shows that decision rules in IOs matter for the impact of democracies on openness. When we disaggregated our sample of IOs by decision rule, three additional findings emerged that all speak to the positive consequences of majority rule for democracies’ ability to open up global governance. First, the substantive effect of an IO’s democratic density on openness is considerably stronger where democracies only need to gather the support of a majority of the membership, as opposed to the entire membership. Second, democratic major powers positively affect the openness of majority-rule IOs, but not consensus-rule IOs. Third, where IOs are subject to majority decision-making, different constellations of democratic states—established democracies, new democracies, and democratic major powers—have a mutually reinforcing effect on IO openness.

By providing us with a better grasp of the mechanisms that underpin democracy’s effect on IO openness, these findings have two broader implications. First, we highlight a source of institutional design in global governance that so far has received less extensive attention. The dominant approaches to date are rational functionalism, with its emphasis on cooperation problems, and sociological institutionalism, with its emphasis on organizational culture and institutional isomorphism (Koremenos et al. 2001; Hawkins et al. 2006; Barnett and Finnemore 2004; Acharya and Johnston 2007). We argue that regime type constitutes a third important explanation of international institutional design. This notion is reinforced by findings that institutional arrangements for transparency, accountability, and dispute settlement, too, appear to be systematically shaped by the domestic political systems of member states (Grigorescu 2007, 2010; Keohane et al. 2000; Hooghe et al. 2013). An urgent task for future research is to identify the scope conditions for this explanation. We find it reasonable to hypothesize that democracy’s effects are restricted to—or at least particularly strong for—international design dimensions that tap into the constitutive differences between democratic and autocratic political systems.

Second, and by extension, we join those who highlight domestic regime type as an explanation of international political phenomena. While most closely associated with the democratic peace program, this liberal explanation is nowadays invoked to account for a growing range of international outcomes, where institutional design is only the most recent. The more common the finding that domestic regime type matters, the more important it becomes to identify the mechanisms that drive democracy’s effects. While others have moved down this path by exploring variation in, for instance, executive-legislative relations (Davis 2012), this article has shown the usefulness of distinguishing between mechanisms on the basis of variation in the composition of IOs.