When Do Withdrawal Threats Achieve Reform in International Organizations?

Conventional wisdom is that international organizations (IOs) have weakened (Singh and Woolcock 2022, this special collection).¹ NATO’s future has been disparaged by president Trump; the World Trade Organization has been limping along (Broz and Bowen 2022, this special collection); and the World Health Organization floundered in handling the COVID-19 pandemic (Carnegie and Carson 2022, this special collection). In each of these cases, states have threatened to withdraw from the IO in question unless it reformed. While these recent cases are perhaps well known, withdrawal threats related to IO reform are not new. For example, in 2011 the United Kingdom “showed UNESCO the ‘yellow-card’” because “a review found it was wasting money and failing to show concrete outcomes” (McGee 2011, 10; Broz and Bowen 2022). In 1985 Peru threatened to withdraw from the International Monetary Fund (IMF) if it could not reach a decision on “reform of the monetary system and the distribution of world liquidity in a more fair manner” (Brooke 1985).

Yet we lack a broader understanding about these threats to withdraw and the reforms they purport to achieve. As scholars and policymakers think about reforming multilateralism, this article addresses the following questions: How often are states’ threats to withdraw from IOs associated with demands for IO reform? And under what conditions are these ultimatums successful? In other words, when can threatening to exit be a catalyst for IO reform? Examining the conditions under which the threat of withdrawal is effective at reforming IOs can help us understand when overhauls happen—and can also reveal other reasons (unrelated to reform) for which states might threaten to withdraw.

We argue, leaning on Hirschman’s (1970) framework of “exit, voice, and loyalty,” that public threats to exit can be an explicit strategy for voice—but only under certain circumstances. Building on international relations theory, we contend that threatening to withdraw is more likely to result in reform under two conditions: when reform demands are made by powerful states and when they are limited instead of broad.

We test this argument on an original dataset of threats to withdraw from all international organizations and states since 1980. For each IO withdrawal threat, we parse media reports, IO archives, and other material to categorize the potentially leaving state’s demands. We also study what happens to the IO in the aftermath of these pressures to understand when withdrawal threats are a catalyst for reform and when they are not.

The empirical analysis supports our arguments and reveals other important findings. First, IO withdrawal threats are not always related to demands for IO reform—in fact, they are related less than half the time—even if the media often links withdrawal threats to the organizations’ weaknesses. Our analysis shows that IO withdrawal threats are instead often linked to geopolitical conflict between states (Patrick 2020) (including dissatisfaction with other states’ policies or their membership status in the IO) or states trying to undermine the IO’s legitimacy (a face-saving strategy to preempt their own punishment) without making calls for IO reform.

Second, our analysis suggests that few states are successful in achieving IO reforms by dangling the stick of exit. Only a third of threats that are coupled with reform demands are completely successful. Nonetheless, results show that powerful states are more likely to generate IO reform, likely because other member states know that these leaders have credible outside options and are important to maintaining the IO. The United States and the United Kingdom are the most frequent to threaten withdrawal. Their exit would deprive the organization of substantial funds, legitimacy, and technical know-how. This underscores another way that powerful states can channel their influence through IOs (see, for example, Dreher, Sturm, and Vreeland 2009; Kilby 2009; Kuziemko and Werker 2006; Stone 2011). Weaker states have less power to evoke reform with this strategy. IO reforms, therefore, may not tilt in weaker states’ favor. This contrasts with alternative explanations arguing that state power may make IO reform demands less likely because the IO may be perceived as kowtowing to powerful member states.

Threatening withdrawal is also more likely to result in reform when the state makes limited calls for reform, including those related to the IO budget, administrative affairs, and operations. Threatening to leave is unlikely to push member states to address broader changes in the IO’s mandate or mission.

States often express dissatisfaction about the effectiveness, efficiency, or outcomes of international organizations. At times, such dissatisfaction has led states to push for IO reforms (O. R. Young 1994; Gutner and Thompson 2010; Tallberg et al. 2016; Lipscy 2017). States have called for IO reforms starting shortly after their creation (Eilstrup-Sangiovanni and Verdier 2021), during external shocks including the aftermath of wars (Debre and Dijkstra 2021; Eilstrup-Sangiovanni 2020), and to reflect shifts in the balance of power (Vabulas and Snidal 2020).

Nonetheless, reforming international organizations can be difficult. Path dependence and status quo bias often mean that IOs tend to be “sticky” (Pierson 1996) and substantial IO overhauls are rare. When IO reform does occur, it is often attributed to large political shocks (punctuated equilibria) (Bennett and Elman 2006; Hall and Taylor 1996; Krasner 1976; Page 2006; Wallander 2000; H. P. Young 1996). IOs also reform in more incremental ways through a process of adaptation and learning (Haas 2018); diffusion of policies and norm cascades might also lead to change. Jupille, Mattli, and Snidal (2013) argue that policymakers are boundedly rational and “satisfice” when reforming, using familiar solutions first and then moving to “selection, change, and creation.” Smaller changes can also be provoked by powerful states, but the distributional outcomes depend on IO performance (Carnegie and Clark 2019).

Another important way to think through member states’ reform options involves Hirschman’s (1970) seminal “exit, voice, and loyalty” framework.² This model outlines that disgruntled actors can speak out or exit, and it can be applied to the case of international organizations. First, dissatisfied states may voice their frustration and call for IO reform. For example, state leaders have made negative statements at the United Nations General Assembly (UNGA) about the WTO, the World Bank, and the IMF (Kentikelenis and Voeten 2020), or voiced their frustration directly in respective organizations. Second, dissatisfied states can exit the organization or threaten to do so. Lipscy (2015) shows that implicit exit threats can catalyze IO reforms based on the availability of states’ outside options. The World Bank, for example, is quicker to reform than the IMF because the World Bank faces more competition from the network of development-focused IOs. We contribute to this work by examining an understudied tactic for IO reform: explicit withdrawal threats.

While Hirschman (1970) outlines that opportunities for voice and exit can influence each other, we hone in on instances when member states merge two of these strategies: when states explicitly use the threat of exit to voice their demands for reform. Threatening to withdraw to catalyze reform is a rather bruising and perhaps coercive strategy, but it addresses an important lacuna in the literature on IO reform more broadly.

Dissatisfied states can try to reform IOs by presenting other member states with a choice—undertake reforms and keep the threatening state in the IO, or keep the IO as is but lose the threatening state. We define a threat to withdraw as occurring when the state³ explicitly and publicly voices that it is considering ending its membership in the IO.⁴

Beyond the threat, a member state can actually withdraw from the IO by voluntarily removing itself from all contractual obligations and legally terminating its membership.⁵ While we are beginning to understand when states withdraw from IOs (von Borzyskowski and Vabulas 2019; Davis and Pratt 2020; Daßler and Heinkelmann-Wild 2021), we know little about the lead-up to this—and how exit threats or posturing affect IO reforms. Threatening to withdraw might be a final sign of displeasure, a last push for change after previous efforts were unsuccessful. A threat to withdraw might be a final chance at bargaining inside the IO.

We define an IO reform demand as a request made by an IO member state that is aimed at changing the procedures, outcomes, or nature of the organization. It can be broad (relating to the IO’s mandate, mission, or leadership rules) or limited (relating to voting rules, financial contributions, administration, or operations). By procedures, we mean rule changes such as the way in which decisions are made, members’ voting rights, or financial contributions. By outcomes, we mean potential changes to the deliverables that the IO achieves through its work and programming. By nature, we refer to potential changes in the IO’s scope or mandate. In sum, demands for reform refer to calls for overhauls regarding the way in which the institution works. Reform-based threats to withdraw therefore must include an explicit “if-the-IO-does (not)-change… then exit” logic.

We caution that calls for IO reform are not necessarily in the interest of the organization or most of its member states, even if “reform” often has a positive connotation. States and IO bureaucrats may have suboptimal ideas for change or desire reform for personal advancement or vainglorious reasons (e.g., to leave an institutional legacy). In other words, while challenges in international cooperation are generating many legitimate calls for IO reforms, it is important to remember that resisting “reform” may sometimes be beneficial to the IO (Starr 2019).

There are many examples of states using the threat of withdrawal as a bargaining chip to push reform. For instance, in 1985 Japan was the second-biggest contributor to UNESCO after the Soviet Union threatened to withdraw from the organization because “progress on reform was insufficient.” Japan explicitly warned that “it would reconsider its membership unless more reforms were made by the end of the year” (M’Bow 1985). In 1991 Thailand threatened to withdraw from the International Natural Rubber Organization if “it did nothing to give rubber a bounce” (The Independent 1991). In 1992 Cyprus threatened to withdraw from the Non-Aligned Movement because “if the movement does not change, it won’t survive and we won’t stay in it” (The Times 1992). And as early as 2012, the United Kingdom said that it “should use the threat of possible withdrawal [from the European Union] as a weapon in negotiations” (Guardian 2012).

The “we’ll-leave-unless-you-change” threat can fulfill several purposes and speak to several audiences: to rally other states for collective reform (increasing bargaining power and pressure on the organization or other member states), but also as a tactic to gain domestic electoral advantage or sway legislators to ratify changes. For example, “President Trump announced his intention to quickly withdraw the United States from the North American Free Trade Agreement (NAFTA), a move intended to force House Democrats to enact a revised version of the pact” (Thrush 2018). Threatening to withdraw can change the status quo ante: in the case of NAFTA, the possible exit forced Congress to compare the revised pact to no pact at all, rather than assess their satisfaction with the old pact. In addition, “some saw the threat as posturing by Trump to gain leverage over Mexico and Canada as he trie[d] to negotiate changes to the deal” (Associated Press 2017).

While states sometimes threaten to withdraw unless reforms are made, other times states do the opposite: they threaten to leave if reforms are made because they do not want to see the IO change. For example, Brazil threatened to leave the International Coffee Organization in 1986 if the organization lifted quotas for coffee sales as this would have devastated Brazil financially following a year of low production and drought. These examples show that threatening to withdraw is not always linked to calls for reform, and that even when it is, these reform calls are not always successful.

So how often is threatening to withdraw IO membership used as a strategy to push for IO reform? And when does it generate reform? Leaning on international relations theory and research on power and negotiation, we argue that withdrawal threats are more likely to result in reform under two conditions: when they are made by powerful states and when the reform demands are limited rather than broad.

First, we argue that withdrawal threats by powerful states are more likely to result in reforms because of these states’ greater financial power, technical know-how, and political leadership, which translates into a higher threat credibility and thus more bargaining power for reform. While powerful states are not always necessary for cooperation to continue (Keohane 1984), some research suggests that losing a powerful state may make the IO worse off because powerful states can be key in helping IOs overcome collective action challenges (Stone 2013). Foundational work by Olson (1989) shows the importance of a privileged group in coalescing collective action. Due to the outsized contributions of a powerful state, its absence is more likely to cripple the privileged group necessary for cooperation (Milewicz and Snidal 2016). A different line of work on hegemonic stability theory also shows that powerful states can be key to IO sustainability; without them, IOs are more likely to flounder.⁶ Powerful states are also more likely to have outside options, making their threat to leave more credible (Gruber 2000). Thus, remaining member states may rally to reform the IO to keep the powerful state in (and the IO afloat). However, there is a tension: even powerful states cannot threaten withdrawal to improve bargaining outcomes all the time, as they would likely lose credibility and detrimentally affect diplomatic relations with remaining member states.

Powerful states’ threats may also be more likely to be successful due to their financial power, as states with larger economic resources or issue-area capabilities are often responsible for shouldering larger shares of funding for an organization (Bayram and Graham 2015, 2017; Broz 2008). This could make the state more easily dissatisfied (because why pay for an organization that does not yield sufficient benefits) and more likely to use their contribution power as a bargaining chip; their potential departure could be a greater threat to the organization. A state’s financial power can manifest itself in wealth or other forms of importance in the particular issue area. For example, in commodity organizations, member states with large resource endowments (e.g., oil, coffee) can hold power in directing the IO to change. Losing a state that controls one-third of the commodity might cripple the institution, so the IO might be coerced to reform. Financial power can also relate to growing economic potential.

Beyond finance, powerful states also bring knowledge and expertise in the issue area (Nelson 2017; Clark and Dolan 2021); their absence could mean a void of both experience and capabilities. Furthermore, powerful states often bring soft power to the institution; their departure may reduce IO legitimacy and authority (Tallberg and Zürn 2019). Remaining member states may therefore be more likely to work on reforms to keep the state in. Research on powerful states’ influence in IOs underscores their role in political leadership (even if it is through “horse-trading”) (Kuziemko and Werker 2006). Powerful states can grease the wheels of IO operations by speeding up loan disbursement, adjusting the stringency of conditions attached to assistance, and determining the magnitude of foreign aid (Dreher, Sturm, and Vreeland 2009; Kilby 2009; Stone 2011). We argue that less powerful states do not have the same power to evoke reform with membership ultimatums. Weaker states need to rely on other strategies besides unilaterally threatening to withdraw to reshape multilateralism. IO changes, therefore, may not tilt in weaker states’ favor. This leads to our first hypothesis:

In contrast to our argument, one could also imagine the alternative: that public demands by powerful states may make reform demands less likely to succeed. In this line of thinking, IOs that reform in the face of ultimatums by powerful states could damage their credibility because the IO may appear to be kowtowing to its most powerful members. This could challenge the putatively egalitarian nature of organizations. In other words, an alternative argument is that powerful states’ reform demands pose greater legitimacy problems to IOs and are thus less likely to succeed.

Second, and in addition to state power, we argue that threatening withdrawal is also more likely to result in successful reform when the state makes limited reform calls (such as for incremental change) rather than broad calls. We argue that limited (technocratic) reforms such as those related to the IO budget, administrative affairs (such as voting formulas), and operations are more likely to be adopted than grand calls for change (including reforms related to the IO’s mandate or mission), which would require a more fundamental rethinking of the institution. For example, proposing to revise discrete formulas is more likely to gain traction than claiming the IO is going beyond its founding goals or that the IO’s overall approach needs overhauling, which does not create a directly actionable pathway for reform. Broader debates about the nature, mandate, or mission of an IO are unlikely to lead to reform in the IO because normative biases can lock in. Therefore, we argue that grand critiques of IOs are perhaps less likely to be achieved with a withdrawal threat. Moreover, incremental change is easier to achieve through collective action and also more palatable for reform-hesitant states than broad changes. This leads to the second hypothesis:

To test these hypotheses about member states’ threats to withdraw and how these threats are linked to (demands for and implementation of) IO reform, we collected an original dataset on withdrawal threats pertaining to all states and IOs since 1980. Our dataset is at the member state-IO-year level. If a state threatens multiple times in the same year (e.g., the United States’ threat to leave NATO multiple times in 2018), we record it once. To be included in the dataset, the threat to leave must have been credible—that is, it must have been issued by the country’s government representative/executive.⁷ In line with our definition of withdrawal threats above, we focus on threats to leave an IO (not just parts of an IO).⁸ We record a withdrawal threat when it is made; we do not include actual withdrawals.

We source data on withdrawal threats from media sources because this is a common means of withdrawal announcement that is available to all states and IOs and addresses all target audiences. Focusing on media accounts of IO withdrawal threats helps move beyond individual IOs. To date, it has been difficult to study what leaders say at IOs in a cross-IO manner because there is no uniform documentation of meeting minutes or voting patterns across all IOs.⁹ Thus, studies often focus on one or a few IOs to keep consistent internal IO sources (speeches or meeting minutes, where available). The advantage of using media sources is that we can see when states use the public “bully pulpit,” which is accessible to all, to exact gains (reforms) within an IO by threatening to leave. Since threats to withdraw from IOs are often public by design (because they involve muscle flexing aimed at increasing a country’s bargaining power), the media is likely to pick up many (if not all) IO withdrawal threats. If a threat to withdraw remains private, it cannot elicit pressure from the public, and attention from other countries may be muted.

To compile the dataset of withdrawal threats, we searched LexisNexis News, a prominent media database, for historical newspaper articles. For search terms, we included each of the 534 IO names (and abbreviations), representing each active organization in the COW IGO dataset (Pevehouse et al. 2020) and the words “threat*” and “withdr*”. For each news article, we then used human coding and filtering (to check relevance and avoid duplicates). Since we are interested in threats of membership withdrawal, we focused on threatening withdrawal of state membership from the IO and excluded other kinds of withdrawal, such as that of funds or delegations. We trained research assistants and oversaw the process with regular quality checks.

We caution that the data are imperfect, so our results are suggestive of the plausibility of our arguments and the dynamics of IO reform, but not conclusive. We outline several data limitations in the coding section below and in the appendix. We assume that most actual membership withdrawals are preceded by withdrawal threats—yet in many cases, we were not able to document these preceding threats. In addition, we suspect that the data-generating process is subject to selection issues. Dissatisfied states may threaten withdrawal based on the expected likelihood of reform. In that case, many possible exit threats may not be made. One example is potential threats in IOs that are simply unlikely to reform; this would mean that the threats we do see are addressing softer targets and thus threat effectiveness may appear stronger than it is. Another example includes potential threats by less powerful or low-income states—while perhaps wanting an IO to reform, these states may find that their voice carries less weight, so weaker states may be less likely to threaten withdrawal in pursuit of reform. One future research agenda is thus a more comprehensive cataloging of withdrawal threats in various outlets and through interviews with IO staff and state representatives, even if that is for a subset of organizations.

Our search resulted in a new dataset of 130 withdrawal threats from 1980 to May 2022 across all countries and organizations. For each withdrawal threat, we code several variables: the IO name, country name, year of threat, who made the threat, reason for withdrawal threat, whether reform demands were mentioned, and at least one media resource documenting the announcement.

Figures 1 through 3 illustrate some basic patterns in the withdrawal threat data. Figure 1 shows that withdrawal threats have been fairly persistent over time, with a slight uptick after 2005 (but also note that IOs and IO memberships have increased during this time frame as well) (von Borzyskowski and Vabulas 2022, 340–41). Some of the clustering around and just after 2010 is explained by member states adopting tighter budgets after the 2008 financial crisis, which sometimes demanded a reckoning in spending.

Figure 2 lists the states that most frequently threaten withdrawal. In line with the logic outlined above, more powerful states globally (the United States) and regionally (the United Kingdom, Japan, Nigeria, and Brazil) have threatened IO withdrawal more often. Yet there are also some surprises, such as Thailand and Libya (which has threatened multiple times to leave the Arab League). These are not conventionally seen as powerful countries in terms of GDP but may have IO or issue-specific power.

Finally, Figure 3 lists the IOs that are most frequently subject to withdrawal threats. This is simply for transparency of the nature of the data—we do not develop specific expectations here about which organizations should be more subject to withdrawal threats or more likely to reform after such threats, but future work could advance theory and testing on this front. As may be expected, though, organizations with more members are more likely to experience withdrawal threats because the pool of (potentially dissatisfied) states is larger. The list is dominated by (near) global organizations (including the ICC, UNESCO, WTO, FAO, WHO, UN) along with some regional and task-specific organizations (such as the International Whaling Commission/IWC, Arab League, NATO, and the Council of Europe).

We code reasons for a state’s withdrawal inductively based on our research and reading of the cases. Figure 4 shows the reasons for how states have justified their withdrawal threats. States justify the majority of withdrawal threats based on dissatisfaction with the organization. Examples in this category include Ecuador’s threat in 1992 to withdraw from OPEC because it did not think that OPEC’s quota system was working in its favor and Australia’s threat to leave the European Bank for Reconstruction and Development in 2008 because it doubted the IO’s importance (until the global financial crisis unfolded).¹⁰ States also threaten to withdraw based on internal politics in the IO, including Venezuela’s threat to leave the World Bank in 2007 because it rejected the financial institution’s conditions on poorer and less powerful states.

Other justifications for withdrawal threats relate to the IO overstepping its mandate or infringing on state sovereignty (e.g., Sudan’s 2006 threat to withdraw from the African Union if the AU allowed foreign intervention in peace talks for the Darfur conflict)¹¹ and financial considerations or overall budget allocation (e.g., Austria’s 2009 threat to withdraw from the European Organization for Nuclear Research/CERN because it consumed too much of its international budget). Justifications also include disagreements regarding which state should take on IO leadership positions (e.g., Lithuania’s threat to leave INTERPOL in 2018 due to the potential election of Russia’s president, creating fears that Moscow would use the role to target political opponents)¹² or even membership status of another state (e.g., Pakistan’s threat to withdraw from the Organization of Islamic Cooperation in 2003 if it admitted India).¹³

So far, we have shown that withdrawal threats have been a persistent phenomenon in world politics, have been issued by various countries against many organizations, and have been justified mainly based on dissatisfaction with the organization’s policies. We now turn to variable coding and hypothesis testing.

In our dataset of withdrawal threats, we measure reform demand as a binary variable, coded 1 when the state linked the withdrawal threat to a demand for IO reform, and 0 otherwise. We acknowledge a shortcoming of our data collection strategy in this regard, as not all states may make this demand explicit in their withdrawal threats and that even if they do, media reports may not comprehensively report on these demands. In both cases, it means we would undercount reform requests, heightening our observation that withdrawal threats demanding reform are a significant part of how states try to renegotiate multilateralism.

We code the ultimate dependent variable, reform success, as a categorical variable of none, partial, or full (0, 0.5, 1). To mitigate coder bias, the outcome variable was coded by a different coder than the independent variables, once a joint understanding of the list of requested reforms was reached. One example of a demand in the modal case—a threat with a reform demand that was not met—is Senegal’s 2010 threat to leave the ICC if the ICC prosecuted Sudan’s leader, Omar al-Bashir. The ICC did not stop the indictment or prosecution of Bashir and did not meet three other requests associated with the exit threat. Coding reform success can be somewhat subjective because changes can take time to debate and implement. For example, one of the associated requests of Senegal (then in a coalition with other states and expressed through the AU) was to amend the ICC’s founding statute (Rome Statute, Articles 27 and 98) to allow immunity for sitting heads of state. This was outright rejected. Another of Senegal’s requests, to amend Article 16 to allow the UN General Assembly to defer cases to the ICC, was first delegated to an ICC working group which later rejected it. For our analysis, this could result in an underestimate of how successful withdrawal threats are in pushing IO reforms, especially for the more recent cases in our dataset.

An example of a demand that was partially met occurred in 1987 when the United Kingdom threatened to withdraw from CERN, calling for greater efficiency within CERN and a cut in British payments to no more than £35 million a year. The threat also called for a thinning of CERN staff. CERN agreed to cut annual UK contributions to £45 million, which the United Kingdom accepted, even though the amount was higher than the original demand; the United Kingdom, feeling appeased, subsequently removed the threat.

A case of full success on reform demands is the 2018 US withdrawal threat from the Universal Postal Union (UPU)—a UN body that has regulated international mail service for more than a century. In issuing the withdrawal announcement, the White House said “the UPU enabled foreign postal services to take advantage of cheap shipments to the US, creating an unfair cost advantage over US companies”¹⁴ and that it would “seek to renegotiate the terms of the UPU rules.” Going further, White House spokeswoman Sarah Sanders stated that “If negotiations are successful, the administration is prepared to rescind the notice of withdrawal and remain in the UPU.”¹⁵ On September 25, 2019, delegates struck a deal to allow the United States to start setting its own postal fees and to allow other countries to start phasing in higher rates. Delegates from 140 countries stood to applaud Bishar Hussein, the head of the UPU, when the new formula passed, and he called the outcome “the most remarkable day in the history of the union.”¹⁶

For our first independent variable, we code powerful state based on economic power. Member state’s GDP is a useful proxy for power given that it applies to all states, IOs, and years (whereas other measures of power are less generalizable). We code powerful state as 1 for states that rank first or in the top 5 percent of largest GDP among all member states in a given IO-year.¹⁷ This allows powerful states to be time varying, and also allows for the possibility that more than one state in a given IO can exhibit power (for example, both France and Germany in the European Union). An example of a less powerful state is Senegal in the ICC in 2010, as the ICC then had 113 member states, and its richest 5 percent were dominated by European countries (Germany, France, United Kingdom), Japan, and Brazil. Instead, Senegal ranked sixty-third in terms of power in the ICC in 2010.

We code limited reform demands as those related to the IO budget, administrative affairs (like voting formulas), and operations. In contrast, broad reform demands are coded as those relating to the IO’s mandate, mission, or rules for leadership selection. For example, the request to change the Rome Statute to allow immunity for sitting heads of state would change the founding charter and chip away at the fundamental mission of the ICC. Another reform request was to change the Rome Statute to allow the UNGA (instead of the UN Security Council, by changing Article 16) to defer cases under certain circumstances. We thus code this as a broad demand, relating to the IO mandate and founding document. We note that if there is bias in this coding, it likely goes against hypothesis 2: we are more likely to undercount limited demands because these are less widely publicized. If we find that limited demands are more successful, that result is likely a conservative estimate in these data.

Since the number of cases of withdrawal threats associated with reform demands is too small for a robust multivariate (regression) analysis,¹⁸ we use primarily bivariate analysis. Specifically, we use chi-square tests of independence to evaluate whether our independent variables are significantly related to reform success. For each of the two arguments above, we test the null hypothesis that there is no significant association between the independent variable and reform success. A significant p-value on the chi-square test indicates that we can reject the null hypothesis of no association (rejecting that the two variables are independent of each other), meaning that there is a statistically significant association between the two measures.

Starting with our first question of when withdrawal threats are used as a strategy to push reform, we find that of the 130 threatened withdrawals since 1980, 52 threats (40 percent) include calls for reforms in order for the state to remain in the organization. This is significant: threatening to withdraw is an important strategy in pushing IOs to reform. Nonetheless, we highlight that this is less than half of the withdrawal threats, also showcasing that withdrawal threats are connected to other motivations beyond pushing for reform.

Withdrawal threats without reform demands are often linked to IO legitimacy or geopolitical conflict. Sometimes states despise IOs outright and want to leave without calls for IO change. One example is ‘face-saving’ withdrawal threats. For example, Russia’s threat to leave the Council of Europe (CoE) in 2018 was not related to a call for reform. Instead, Russia wanted to challenge the CoE’s legitimacy and cast it as a biased Western institution, lacking credibility, and to signal that Russia was perfectly fine not being a part of it. Similarly, Zimbabwe threatened to withdraw from the Southern African Development Community (SADC) in 2013 if they demanded that Zimbabwe should hold free and fair elections. In doing so, Robert Mugabe attempted to deflect attention away from his democratic backsliding by blaming the fifteen-member southern African regional bloc for bad policies. Recognizing that membership suspension was imminent, he said, “If SADC decides to do stupid things, let it be known that we can withdraw from SADC.”¹⁹ The regional bloc had pressed Mugabe to delay elections to allow time for electoral reforms and a smaller military role in politics.

However, threats that are not linked to reform demands may still ruffle remaining states because they recognize that the institution could lose legitimacy should the threatening state leave. In the case of the CoE in 2014, for example, many members did not want Russia to withdraw because no longer being part of the CoE’s European Court of Human Rights could let Russia off the hook in terms of its human rights violations.²⁰ Russia’s threat to withdraw might rally remaining members to keep Russia as a member to influence the country to change as part of membership criteria.²¹ More recently, when Russia invaded Ukraine in 2022, we have seen similar dynamics, with Russia being suspended from the CoE (but subject to its Court for human rights violations until December 2022), then withdrawing (to save face) and announcing its intention to denounce the European Convention on Human Rights, and then being expelled by the CoE.

Another group of withdrawal threats unrelated to IO reform is triggered by geopolitical conflict (Davis and Pratt 2020) and internal politics with other states in the IO. For example, the United States threatened to leave the Australia-New Zealand-US Treaty Organization (ANZUS) in 1985 if New Zealand introduced legislation banning visits by ships that could be carrying nuclear weapons. This threat was therefore intended to push one state to change its stance, rather than to push for IO reform. Similarly, Georgia’s threat to leave the Commonwealth of Independent States (CIS) in 1999 was in direct response to the Russian State Duma’s decision to send its deputies to attend the presidential elections in Abkhazia, an unfriendly step in relations with Georgia. Another example of withdrawal threats unrelated to reform relates to geopolitical conflict over IO leadership positions. For example, in 1987 Belgium, the Netherlands, West Germany, and Denmark threatened to withdraw from UNESCO should Amadou-Mahtar M’Bow (from Senegal) win a new six-year term. They accused him of being too ideological, inefficient, and supportive of limiting press freedom, and of spending too much money at the headquarters and not enough in the field.²²

Also related to geopolitics, many withdrawal threats are triggered by the membership of other states; a state may use the threat of exit to express displeasure at another state’s joining. This occurred when Nigeria threatened to leave the Economic Community of West African States (ECOWAS) in 2017 if Morocco was allowed to join.²³ Similarly, the US Congress has laws stipulating that it will exit certain IOs if they recognize Palestine. For example, in 1989 the United States warned the Food and Agriculture Organization (FAO) that it might leave if FAO extended any further recognition to the Palestinian Liberation Organization (PLO).²⁴ Together, these examples show that states use the threat of IO withdrawal to achieve many political goals; reform represents a substantial share of these, but it is not always the threatening state’s aim.

Moving on to our second question: when are reform demands backed by withdrawal threats successful? Our data show that 35 percent of reform demands are met, sometimes forestalling the state’s actual exit. Another 27 percent are partially successful, meaning that some demands are met and others are not. However, 38 percent of states’ threats to withdraw do not lead to reform.

We find suggestive support for H1: powerful states are more likely to achieve reform demands after threatening to withdraw than less powerful states. Among powerful states, half of their demands are met, while only 19 percent of weaker states’ demands are met. The p-value associated with the chi-square statistic is statistically significant (0.034). We illustrate this relationship in figure 6.

The United Kingdom, a powerful state in many IOs, has been successful in pushing for reforms in several IOs—for example, when it threatened to withdraw from UNIDO in 1997.²⁵ It also achieved partial success in enacting reforms after it threatened to withdraw from the Food and Agriculture Organization (FAO) and the International Organization for Migration (IOM).²⁶ In 2011 a domestic review of UK contributions to IOs found several institutions wanting in terms of value for money, efficiency, and outcomes. The United Kingdom put them on notice to reform urgently or risk losing the UK’s support. The directors general committed to cost savings and increased transparency. Moreover, the General Conference agreed on resolutions regarding IO reviews, including some by external assessors. These same reforms were not picked up after similar reports in 2010 that were not accompanied by the United Kingdom’s threat to withdraw.

Brazil’s power in certain issue areas also highlights the relative nature of power in IOs. For example, Brazil’s threat in 1986 to withdraw from the International Coffee Organization (ICfO) showed the importance of its power because it accounted for 30 percent of global coffee production and thus would have crippled the institution in its absence. Brazil’s threat to exit yielded full success: this helped it retain its share of the export quota and avoid renegotiation of the agreement (which arguably would have favored remaining members). Similarly, Brazil’s threat in 2007 to withdraw from the IMF was fully successful: it led to a new agreement on credit and loans, effective in 2011, called the New Arrangements to Borrow (NAB). As part of the NAB effort, Brazil, Russia, India, and China gained veto power in the organization. The IMF also increased voting shares and basic voting share protection in a proposal put forward in the 2008 Annual Report which led to 135 countries seeing a total increase in vote shares of 5.4 percentage points. Brazil was one of the countries with the highest increase. Peru, without the same power as Brazil, was unsuccessful in using a similar withdrawal threat to push for the same sorts of reforms in 1985.

Beyond a single state’s power, the cases also point to the power of coalitions and gaining allies’ support. Weak states in particular may feel more empowered to threaten withdrawal after other states have threatened before them, creating a bandwagon or strength-in-numbers effect. These threat coalitions are often informal and fluid in nature, arising organically as issues emerge. On the one hand, the Brazil-IMF case shows that getting like-minded states on board is important to coalesce bargaining power and successfully push for change. Moreover, the United Kingdom’s threat to withdraw from UNIDO was echoed by Australia, the United States, and Germany within a short time frame, helping to mount support for the threatened-exit-as-voice campaign for reform. Denmark’s in-depth review cemented the needed changes. But trying to gain a whole group’s support for the ultimatum can also backfire. Many African Union states pushed for all members to threaten to leave the International Criminal Court unless it reformed, but this strategy revealed fractures in how these countries wanted the institution to change. Djibouti, Senegal, and Comoros were holdouts, perhaps giving more support for retaining the status quo.

It is also worth noting that while powerful states are more successful at using withdrawal threats to push for successful reform, power alone may not fully determine exit options. For example, the credibility of the United States turning to alternative IOs (underlying the power of a withdrawal threat) may be challenged because the existing order is US led and US skewed, making the idea of a more beneficial IO setup seem weak. In addition, some withdrawal threats are a result of great-power rivalries, which may complicate the causal mechanism of powerful states threatening exit. The role of power is also not always clear in some near-universal IOs where key powerful states are not members (e.g., the International Criminal Court). Future research can examine other factors that moderate the link between power and credible withdrawal threats and their success.

Our data also support H2: limited reform threats were more likely to be successful. Almost half (47 percent) of all limited demands were successful, while only 18 percent of broad demands were fully successful. This difference is even starker if we also consider partial successes: 83 percent of limited demands but only 32 percent of broad demands were successful to some degree. The p-value on the chi-square statistic is highly significant (p<0.001).

Namibia’s threat in 1996 to withdraw from the Southern African Customs Union (SACU) accused South Africa of unfair trading practices, claiming that SACU favored South Africa over other members such as Botswana, Lesotho, Namibia, and Swaziland. Moreover, Namibia argued that South Africa used SACU to prevent members from making trade deals with nonmembers, which would unfairly restrict Namibia. Namibia’s detailed proposals led to successful wholesale renegotiations of SACU on terms that made South Africa and the other member states more equal.

On the other hand, threats to leave IOs attached to broad demands for changes in an IO’s ideology usually do not result in change. The US and UK threats to leave UNESCO, saying that it was “going beyond its founding goals and attempting too much to dictate global policies,” fell on deaf ears because these were regarded as ideological musings amid Cold War debates.²⁷ Sometimes those threats result in actual withdrawals, where the member state moves on without the IO.

Our findings on withdrawal threats to stimulate reform and the correlates of their success provide important implications for policymakers as we think about how to strengthen global governance. We offer three warnings of potential misinterpretation of states’ threats to withdraw unless an IO reforms, followed by three policy suggestions. First, states (especially less powerful states) may be dissatisfied with an IO but reluctant to threaten withdrawal if the IO does not reform. Diplomats should not conclude that just because moderately or less powerful states have not threatened to withdraw, they are not dissatisfied with the IO. Because they have fewer go-it-alone or alternative options and are perhaps more reliant on collective action within the IO, it is less likely that these states will flex their membership muscles by threatening to withdraw, knowing it is probably a futile endeavor. In other words, less powerful states may be severely dissatisfied with an IO’s policies and/or performance even if they do not try to use “threatened exit as voice.” Instead, their dissatisfaction with an IO can lead to retreat in support for the IO (short of threatened exit), which can lead to diminishing legitimacy in the institution overall. On the flip side, powerful states are able to exacerbate the inequalities of IOs by using withdrawal threats more readily, and thus policymakers need to carefully consider how weaker states express dissatisfaction and can facilitate change (Ferreira 2022, this special collection).

Second, and equally important, diplomats should not conclude that just because states have threatened to withdraw unless the IO reforms, these reform demands are generated by specific dissatisfaction with the organization. Our data show that withdrawal threats are sometimes performative signals targeted at other states or domestic audiences. Governments often blame IOs to buck-pass or scapegoat. Indeed, European Parliament members have recognized “the risk of political blackmailing through the means of exit threats.”²⁸ Overreacting to an IO withdrawal threat—particularly one that is not related to reform—can be as bad as underreacting, because the exiting state may gain an advantage in international negotiations where it is not warranted.

Third, states’ threats to withdraw from IOs present a mixed bag; in some cases, these threats do result in actual withdrawals and a loss to the IO. At other times, threats to withdraw can sound the alarm for meaningful IO reform, and thus may sometimes help IO resilience and vitality.

Turning to policy implications: member states should ensure that bureaucratic hurdles are in place between a threatened IO withdrawal and its effective date. This includes incorporating a significant waiting period for withdrawal in the IO treaty, approval by the national legislature if a state wants to walk away, and mandating that IO dues be paid during the waiting period. These hurdles can make it easier for a threatened or announced withdrawal to become void if governments change. Rules like this kept the United States in the World Health Organization in 2021 when the US presidency transitioned from Trump to Biden and the waiting period following the United States’ withdrawal threat had not yet lapsed. Moreover, a clear waiting period can provide important focal points that can encourage negotiations among remaining member states regarding reform during the waiting period (instead of letting a state simply walk away).

Further, diplomats should pay particular attention when several states threaten withdrawal from the same organization. Multiple threats could signal that the IO faces potential collapse unless reform takes place (von Borzyskowski and Vabulas 2022). In other words, threats by multiple states warn how direly needed the reforms may be. If several states actually leave the organization—even if they are less powerful states—this may facilitate the death of smaller IOs (von Borzyskowski and Vabulas 2022). Policymakers should therefore be particularly wary when multiple states threaten to exit at the same time.

Finally, by and large, member states should disincentivize other states’ withdrawal threats because this may help remaining members achieve reform outcomes that are in line with all member states rather than giving powerful states an advantage at the negotiating table. Moreover, working with outlier states to stay in the IO rather than actually withdrawing may allow countries to continue influencing those outlier states’ policies rather than letting them go rogue. For example, North Korea’s nuclear weapons ambitions might have been better thwarted had the country not withdrawn from the International Atomic Energy Agency in 1994.

Future research should examine both exit threats and reform demands using sources at IO headquarters, such as interviews with IO staff and state representatives, and more extensive secondary case notes. While our findings provide support for both arguments, more work is needed to ascertain the dynamics at play and the relative weight of factors with a larger sample size. Future work should also examine how the strategy of “threatened exit as voice” compares to other methods for reform. If withdrawal threats are less effective and more costly than other strategies for IO reform, institutional change may be better handled while states remain in the organization. Participating in multilateral dialogue and diplomatic brainstorming rather than issuing ultimatums might result in more resilient IO setups. In Johnson’s (2020) words, “instead of starving a poorly performing IO, it might be better to feed it.”

More broadly, policymakers need to consider how the threat of withdrawal-unless-the-IO-reforms may be changing over time. The twenty-first century has created an explosion of new organizations and thus, states that want IO reforms may now be more likely to stay in but complement their dissatisfaction by using overlapping or alternative IOs (without having to take on a costly exit). Additionally, the threat of a powerful state leaving an IO may have different leverage over time as other states (e.g., China) rise in power. On one hand, remaining members may not worry much if the United States threatens to exit an IO if China (or another state) picks up the leadership role, allowing the IO to continue its mission. On the other hand, remaining member states might consider a potential US void as more detrimental now, knowing that a different power (e.g., China) with very different policy preferences could take the IO’s helm.

As we rethink multilateralism—and the strategies that states use to push for reform—it is important to consider the conditions under which states use the ultimate bargaining chip of threatening to withdraw their membership to achieve IO reform. Threatening to withdraw, despite much criticism and some cause for concern, may not sound the death knell for multilateralism. Powerful states often get some of the reforms they advocate, which can improve the IO going forward. And when they do not, they may exit and work unilaterally to promote change, returning to the cooperating table later. But this study shows that threatening to exit an IO as a strategy to push reform is not a panacea and certainly has a mixed record of success.

Dr. Inken von Borzyskowski is associate professor of global policy and international relations at University College London, where she also codirects the MSc International Public Policy program and serves as the Global Governance Institute’s Director for International Security. She received her PhD in political science at the University of Wisconsin-Madison; she also spent a year at Duke. Previously she received an MA at Free University Berlin and worked at the Social Science Research Center. She has been a research fellow at the European University Institute in Florence, at the International Foundation for Electoral Systems (IFES) in Washington, DC, and a consultant at the US Institute of Peace (USIP). Her research focuses on the domestic politics of international relations with an emphasis on international organizations and their effect on domestic conflict and elections. Specifically, her research falls into three areas: international democracy assistance; the causes and consequences of election violence; and international organizations’ membership politics (withdrawals and suspensions). Her research is published or forthcoming in the British Journal of Political Science, Journal of Conflict Resolution, Journal of Peace Research, International Studies Quarterly, and the Review of International Organizations. Her book “The Credibility Challenge: How Democracy Aid Influences Election Violence” was published by Cornell University Press. Her work has been supported by the Philip Leverhulme Prize, the British Academy Mid-Career Fellowship, the US Institute of Peace, and the Swedish government’s Folke Bernadotte Academy. Dr. Felicity Vabulas is an associate professor of international studies at Pepperdine University, where she is faculty advisor for the Model United Nations team and leader of the Women in International Studies Speaker Series. Vabulas earned her PhD and MPP in public policy from the University of Chicago. She earned her BS in business administration from the University of Illinois. Her research focuses on the political economy of international organizations and foreign lobbying. She has three ongoing research projects. Her first project examines when and why states exit international organizations (IOs) and the implications of these exits. This set of papers investigates both suspensions as a multilateral diplomatic sanctioning tool as well as unilateral IO withdrawals (like Brexit). She is currently working on her book project, Exit from IOs: Bargaining and Change in International Organizations (coauthored with Inken von Borzyskowski). Her second project studies states’ increasing use of informal international organizations such as the G7 and G20 rather than traditional IOs with treaties and permanent secretariats. Her third research project examines how foreign lobbying affects US foreign policy, including how it alters human rights, trade, and foreign aid allocations. Vabulas’s research includes work published in the Review of International Organizations, International Studies Quarterly, the Journal of Peace Research, Global Perspectives, Global Policy, and multiple edited volume chapters.

The authors affirm that there are no competing interests in the publication of this article.