Do Alliance Abandonment and Coercion Increase Support for Nuclear Weapons? An Elite Survey in NATO

Abstract

Alliances are central to U.S. nuclear nonproliferation efforts. Many scholars suggest that actions which might undermine allies’ faith in or dependence on the United States will in turn increase their interest in nuclear weapons. Yet this prediction has received only limited empirical attention. This research note tests whether U.S. signals of abandonment and efforts to encourage allied burden-sharing using threats of abandonment increase support for nuclear weapons acquisition, using a survey of foreign policy elites from sixteen European NATO members. The findings suggest U.S. threats to abandon allies unless they increase defense spending modestly increase support for nuclear weapons, but not as much as unconditional signals of diminished commitment such as a withdrawal of U.S. troops. The findings suggest that while U.S. signals of abandonment may undermine nonproliferation, there is less evidence that U.S. coercion has the same effect, or that there is a direct tradeoff between burden-sharing and nonproliferation.

Las alianzas son fundamentales para los esfuerzos de no proliferación nuclear por parte de EE. UU. Muchos académicos sugieren que aquellas acciones que podrían debilitar la fe o la dependencia que tienen los aliados en Estados Unidos aumentarán, a su vez, su interés en las armas nucleares. Sin embargo, esta predicción solo ha recibido una atención empírica limitada. Esta nota de investigación pretende averiguar si las señales de abandono de EE. UU. y los esfuerzos por fomentar el reparto de la carga de los aliados mediante amenazas de abandono aumentan el apoyo a la adquisición de armas nucleares. Para ello, utilizamos una encuesta a las élites de la política exterior de dieciséis miembros europeos de la OTAN. Nuestras conclusiones sugieren que las amenazas por parte de EE. UU. de abandonar a sus aliados a menos que aumenten el gasto en defensa aumentan ligeramente el apoyo a las armas nucleares, pero no tanto como lo hacen las señales incondicionales de una disminución del compromiso, tales como la retirada de las tropas estadounidenses. Las conclusiones también sugieren que, si bien las señales de abandono por parte de EE. UU. pueden debilitar la no proliferación, existe menos evidencia de que la coerción por parte de EE. UU. tenga el mismo efecto, o de que exista una compensación directa entre la distribución de la carga y la no proliferación.

Les alliances sont déterminantes dans les efforts américains de non-prolifération nucléaire. Nombre de chercheurs suggèrent que les mesures qui pourraient nuire à la foi des alliés dans les États-Unis, ou à leur dépendance envers ce pays, augmenteront par conséquent leur intérêt pour les armes nucléaires. Pourtant, cette prédiction n'a que peu attiré d'attention empirique. Cette note de recherche évalue si les signaux américains d'abandon ou leurs efforts d'encouragement du partage de fardeau des alliés par des menaces d'abandon accentuent le soutien à l'acquisition de l'arme nucléaire, au moyen d'un sondage des élites de politique étrangère de seize pays européens membres de l'OTAN. D'après les résultats, les menaces américaines d'abandon des alliés à moins qu'ils n'augmentent leur budget de défense renforcent modérément le soutien aux armes nucléaires, mais bien moins que les signaux inconditionnels de diminution de l'engagement, comme le retrait de troupes américaines. Les conclusions indiquent que bien que les signaux américains d'abandon puissent nuire à la non-prolifération, il est moins clair que la coercition des États-Unis s'accompagne du même effet, ou qu'il existe une relation directe entre partage du fardeau et non-prolifération.

Introduction

Discouraging the spread of nuclear weapons has been a core US foreign policy objective since the 1960s (Gavin 2015). A central pillar of US nuclear nonproliferation efforts is the United States’ network of military alliances. Many observers argue that US assurances of protection can reduce allies’ incentives to seek nuclear weapons (Bleek and Lorber 2014; Reiter 2014; Lanoszka 2018), while others likewise argue that alliances offer Washington coercive leverage to discourage nuclear proliferation by allowing the United States to wield threats of abandonment if allies do pursue nuclear weapons (Gerzhoy 2015; Monteiro and Debs 2017).

Its commitment to nonproliferation in its alliances, however, may come at a cost: reassuring allies of US protection may also discourage allies from investing in their own defense (burden-sharing) (Posen 2014). Conversely, encouraging burden-sharing, especially by signaling that the United States might be willing to abandon an ally, may in turn encourage them to acquire nuclear weapons (Gerzhoy 2015; Lanoszka 2015). But in part because the United States is strategic in its efforts to encourage burden-sharing and mindful of the risk that doing so might cause allies to pursue their interests in a way it opposes (Blankenship 2023), the link between US signals of abandonment, coercion, and efforts to solicit burden-sharing, on the one hand, and allies’ preferences for nuclear weapons, on the other, is difficult to establish in a systematic way.

This research note, then, studies whether US signals of abandonment and coercive efforts to solicit alliance burden-sharing increase allies’ desire for nuclear weapons using an elite survey of foreign policy officials and experts across sixteen NATO members. The survey asks these elites to rate both their own interest in their country acquiring nuclear weapons and their government’s likelihood of pursuing nuclear weapons when faced with four hypothetical scenarios—one in which the United States threatened to abandon the country unless it spent more on defense, one in which the country decided doubled its defense spending over five years, and two in which the United States signaled its unconditional intention to abandon the ally, either by outright declaring it would not defend the ally or by announcing that it was withdrawing all US forces from the ally’s territory. In doing so, I can distinguish the effects of directly signaling abandonment, conditionally threatening abandonment, and the actual act of burden-sharing.

Several results stand out. First, elites reported the highest level of support for pursuing nuclear weapons when faced with unconditional signals of abandonment, and somewhat higher levels of support when faced with US demands for burden-sharing and a scenario in which their country’s defense spending doubled. Compared to a baseline level of support of 1.9 (out of 5), respondent support for nuclear weapons increased to 2.35 when presented with unconditional signals of US abandonment, 2.18 when presented with a conditional threat of abandonment unless the country increased defense spending, and 2.06 in a scenario where their country doubled defense spending without any US pressure. This represents a 24, 15, and 8 percent increase in support relative to the baseline, respectively. Second, respondents indicated that their governments would be more likely to pursue nuclear weapons when faced with unconditional signals of US abandonment.

Third, most elites indicated that the most important reason why their government was not currently pursuing nuclear weapons was the alliance with the United States. Those that did also indicated that their governments would be more likely to pursue nuclear weapons when faced with unconditional signals of abandonment, but not necessarily when faced with American demands for burden-sharing using a conditional threat of abandonment.

Taken together, the findings suggest that while there may be a tradeoff between sending signals of abandonment, issuing coercive threats, and soliciting burden-sharing, on the one hand, and discouraging proliferation, on the other, effect sizes are modest overall. Moreover, effect sizes are smaller when US pressure is conditional—that is, when it relies on the threat to abandon an ally unless it complies—as opposed to when the United States signals its willingness to abandon an ally unconditionally.

This research note contributes to scholarship on nuclear proliferation and alliance politics in three ways. First, while some scholars argue that the pursuit of alliance burden-sharing can interfere with the United States’ ability to discourage allied proliferation, this proposition has received limited empirical attention, and efforts to establish this link are hampered by the non-random application of US pressure (Gerzhoy 2015; Blankenship 2023). The findings modestly support this hypothesis with the caveat that the United States can mitigate allies’ interest in nuclear weapons by using conditional threats of abandonment rather than signaling an unconditional willingness to abandon them. Second, the article contributes to a wave of recent literature that uses surveys to explore how US behavior shapes support for nuclearization among allied audiences (Ko 2019; Sukin 2020). While this literature has focused on the effects of positive inducements like assurances, this article instead explores the effects of coercion and signals of abandonment. Finally, this article builds on existing literature on alliance burden-sharing, which to date has explored the conditions under which the United States is willing and able to encourage allied burden-sharing (Olson and Zeckhauser 1966; Fang and Ramsay 2010; Blankenship 2023). The findings suggest that while the pursuit of burden-sharing may tradeoff with pursuing other alliance goals, the magnitude of that tradeoff depends in part on the method used to solicit burden-sharing.

Nuclear Nonproliferation and the Risks of Encouraging Burden-Sharing

Alliances allow their members to pool resources against common threats (Walt 1987). As such, determining the division of costs and tasks—that is, burden-sharing—is a central task in any alliance. However, existing literature suggests that burden-sharing is often contentious, as states may try to ensure that the costs of deterring and defeating common adversaries fall disproportionately on their partners rather than themselves. This tendency to “buckpass” is an inherent challenge in any alliance, though its intensity may vary depending on the perceived severity of the threat and likelihood that partners are willing and able to handle the threat on their own (Waltz 1979). As a result, burden-sharing is subject to constant negotiation, with partners using threats of punishment, promises of reward, appeals to shared obligations and fairness, and “naming and shaming” (Becker et al. 2023; Blankenship 2023; Becker 2024; Blankenship  2024).

The challenges to successfully encouraging burden-sharing are two-fold. The first is that one’s partners might believe they can free ride without consequences, either because they are confident that their ally will protect them or because they do not perceive a high level of threat. This is often considered to be an especially intractable problem in US alliances, for several reasons. First is the size disparity between the United States and its allies. When faced with a significant threat to the alliance, larger members’ contributions have the greatest potential to be decisive to the outcome of conflict. As a result, larger allies may have more difficulty passing the buck to weaker allies. Weaker allies, by contrast, may elect to free ride in the knowledge that their contributions are less likely to matter (Olson and Zeckhauser 1966). Second, the United States goes to great lengths to signal its willingness and ability to defend allies, including through the US overseas troop presence, which may weaken allies’ incentives to burden-share (Posen 2014).

A growing body of research, however, suggests that size disparity and the US troop presence abroad are not insurmountable barriers to encourage burden-sharing. The evidence for a relationship between allies’ defense spending, on the one hand, and the size of their economies and the US troop presence on their territory, on the other, is mixed at best.¹ This is in large part because the United States is in some cases able to wield threats to abandon allies unless they assume more responsibility for their own defense, which are particularly effective when allies are predisposed to fear abandonment or face a high level of external threat. US troop levels do not necessarily undermine these threats’ effectiveness, as those troop levels are subject to change and the United States in many cases wields threats of troop withdrawals (Fang and Ramsay 2010; Blankenship 2021, 2023).

The second general challenge to encouraging burden-sharing is that doing so is not impossible but rather potentially counterproductive. Particularly in unequal relationships between great powers (patrons) and weaker allies (protégés), alliances often include an asymmetric exchange of goods, with one partner offering security to the other in exchange for giving up some of its foreign policy autonomy (Morrow 1991). This can include offering specific policy concessions like the protégé hosting the patron’s military bases, but often entail a more general agreement to follow the patron’s preferences on foreign policy. One such preference that has played an especially prominent role in US alliances is that American allies refrain from acquiring nuclear weapons. American opposition to allied proliferation has historically stemmed from several sources, including a broader desire to reinforce the nonproliferation norm and a more narrow desire to keep allies dependent on US protection (Kroenig 2010; Gavin 2015; Miller 2018).

Encouraging alliance burden-sharing, then, risks undercutting the United States’ end of the bargain and undermining allies’ willingness to uphold their own—including by seeking nuclear weapons (Lanoszka 2015; Blankenship 2023). Asking allies to become more capable of defending themselves can increase these risks not only by reducing the value that the alliance holds for them, since they are now providing for more of their own security, but also by empowering them and reducing their dependence on US protection. Put simply, allies that become more capable of defending themselves also become more capable of going their separate ways. In particular, encouraging burden-sharing by stoking allies’ fears of abandonment may be especially likely to increase their interest in nuclear weapons as a hedge against being abandoned (Lanoszka 2018).

Although existing literature argues that encouraging burden-sharing may come with the risk of encouraging nuclear proliferation, there has been relatively little empirical study of this tradeoff's prevalence. Beyond the work of Gerzhoy (2015), Lanoszka (2018), and Blankenship (2023), who describe the proliferation-burden-sharing tradeoff in select cases during the Cold War, systematic evidence is limited on whether encouraging burden-sharing, including through threats of abandonment, does in fact increase allies’ desire for nuclear weapons.

Linking Burden-Sharing Pressure to Nuclear Proliferation

One can identify at least two mechanisms by which encouraging allies to assume more responsibility for their own defense might increase their desire to have nuclear weapons. One is substitution. When allies become, or are encouraged to become, more capable of defending themselves, they are more likely to ignore US preferences and do so on their own terms. Nuclear weapons may become an attractive substitute as the costs of conventional defense rise (Bell 2021). South Korea, for example, pursued nuclear weapons during the 1970s as a complement to its conventional military buildup, while some NATO members sought access to nuclear weapons during the 1950s and 1960s due to the perceived cost and futility of a conventional defense from a Soviet invasion (Trachtenberg 1999). Similarly, Britain used its nuclear arsenal to shore up its alliance commitments while reducing its conventional forces (Bell 2021, chap. 2).

Another is fear of abandonment. Existing research suggests that US efforts to encourage burden-sharing tend to be more successful when allies fear being abandoned (Fang and Ramsay 2010). This fear can stem either from factors outside of their patron’s control, such as when it faces domestic pressure to withdraw forces, or from direct patron signals that it might be willing to reduce or withhold support. The latter, in turn, can take the form of either conditional threats to withhold or reduce support unless the ally invests more in defense, or unconditional signals that the United States will do so regardless of the ally’s behavior (Blankenship 2023, 2024). Yet at the same time, fear of abandonment is also likely to increase allies’ desire for nuclear weapons (Reiter 2014; Lanoszka 2018).

Each mechanism has distinct observable implications. The substitution mechanism suggests it is the act of burden-sharing or being asked to burden-share that is likely to increase interest in nuclear weapons, whereas the fear of abandonment mechanisms suggests it is how the United States seeks burden-sharing that matters most. Thus, in the case of the latter, one might expect that a patron could reduce allies’ interest in pursuing nuclear weapons by softening its pressure. For example, if a patron threatens to abandon an ally, it could soften that threat by signaling that it will not abandon the ally if it increases its contributions (Blankenship 2023, 2024).

   

At the same time, however, there are also good reasons why allies might be reluctant to consider pursuing nuclear weapons even if the United States demands greater burden-sharing or signals its willingness to abandon them. One reason is fear of preventive attack, as allies may fear that others will attack them before they acquire nuclear weapons (Fuhrmann and Kreps 2010; Whitlark 2021). Another is the power of the nonproliferation regime. The evidence suggests that the Treaty on the Non-Proliferation of Nuclear Weapons has been effective in curtailing the spread of nuclear weapons since 1970, with only a handful of countries acquiring or even seriously pursuing nuclear weapons since that date (Fuhrmann and Lupu 2016). Scholars have attributed this success to several factors, including a desire to comply with the norm of nonproliferation (Rublee 2009), as well as fear of economic sanctions (Solingen 2007; Miller 2014). The transparency brought about by International Atomic Energy Agency inspections, in turn, raises all of these risks because states have greater chance of being caught (Kaplow 2022). Finally, for some countries, producing the necessary amount of fissile material and delivery systems may be prohibitively difficult, while for others, acquiring nuclear weapons may not be consistent with their national identity and self-conception (Hymans 2006).

   

Methodology

To test these hypotheses, I fielded a survey of NATO-Europe foreign policy elites across sixteen countries.² Following Busby et al. (2020, 119), this study defines foreign policy elites as “the community of individuals with professional experience or expertise related to foreign policy and international affairs.” The data were collected through an anonymous, five-minute survey circulated between November 2023 and March 2024 to foreign and defense policy experts at think tanks as well as officials in the Ministries of Foreign Affairs and Defense in each country.³ The total sample size is 110, with 19 percent of respondents being current government officials and an additional 31 percent having previously served in government.

In recent years, scholars have frequently turned to surveys to understand the conditions under which actors may consider nuclear use or acquisition (Ko 2019; Sukin 2020; Dill, Sagan, and Valentino 2022). Doing so offers several advantages. First, analysts often lack the real-world data needed to evaluate many important questions, as since 1945 there have been no cases of nuclear use in wartime and very few cases of nuclear weapons pursuit and acquisition, and nuclear decision-making is inherently difficult to observe in any case (Jervis 2021, 131; Bell 2023, 171–172; Gavin 2024, 176–178). Moreover, when it comes to alliances and nuclear proliferation, strategic interaction between the United States and its allies poses problems for causal inference even in the cases we can observe. Because the United States has historically tried to prevent proliferation among allies using security assurances, cases where it has withheld those assurances or wielded coercion are likely non-random in ways that create threats to inference—for example, they may be those where it is overdetermined that allies would want nuclear weapons, or cases in which the United States expected the risk of proliferation to be low. Surveys are thus a useful means to surmount the inherent uncertainty of studying the politics of nuclear weapons.

Additionally, using an elite sample not only allows me to collect data on the preferences of individuals who have an outsized influence on policy—either directly by holding government positions or indirectly through expertise—but also to draw on their knowledge of their governments’ preferences. Thus, the survey asked respondents both about their individual preferences for acquiring nuclear weapons and about their evaluations of their governments’ likelihood of seeking nuclear weapons.

The primary empirical tests are comparisons of mean levels of support for acquiring nuclear weapons across several hypothetical scenarios. These comparisons are split between the individual-level—respondents’ own preferences—and the government-level—respondents’ expectations about their governments’ behavior. After each scenario, respondents were asked to indicate both how much they themselves would support their country acquiring nuclear weapons, as well as how likely it was that their government would pursue nuclear weapons.

Individual-Level Preferences

For the individual-level comparisons, the survey asked respondents to rate their own baseline level of support for their country acquiring nuclear weapons on a five-point scale that ranged from “Strongly oppose” to “Strongly support.”⁴ Afterward, each respondent was presented with four hypothetical, randomly ordered scenarios indicating either that (1) “the U.S. President announced that the United States would not defend your country”; (2) “the U.S. President announced that the United States was withdrawing all U.S. forces deployed in your country”⁵; (3) “the U.S. President announced that the United States would not defend your country unless your country increased its defense spending”; or (4) their country had “decided to double its defense spending over the next five years.” The first two conditions allow me to gauge respondent support for nuclear weapons in the presence of unconditional signals of US abandonment, while the latter two allow me to gauge support in the presence of either US pressure to spend more using a conditional threat of abandonment or an actual government decision to substantially ramp up defense spending. T-tests were used to assess the statistical significance of the differences in means across scenarios.

To test H1a, an additional set of analyses reported in the supplementary appendix explores the correlates of support for nuclear weapons—and in particular its correlation with preferences over defense spending, which are measured on a five-point scale from “Decrease significantly” to “Increase significantly.”⁶ I do so using both simple pairwise correlations, as well as multivariate regression analysis. The regression analysis includes several control variables: (1) respondents’ beliefs about how safe their country is from foreign attack and their country’s capital city distance to Moscow, which proxy for external threat and may be correlated with defense spending and nuclear weapons preferences; (2) their beliefs about the utility of force (“militarism” versus “antimilitarism”), which some scholars expect to drive views about military arming (Berger 1998; Kagan 2004; Börzel and Risse 2009; Deni 2021); (3) defense spending as a percentage of GDP⁷; and (4) demographic characteristics such as gender, age, and whether the respondent had ever or was currently serving in government.⁸

Government Preferences

As for respondents’ assessments of their governments’ preferences, because of the potentially sensitive nature of asking respondents about their own government’s current intentions to pursue nuclear weapons, I did not ask respondents a baseline question. Instead, I asked respondents which of the following eight reasons explained why their country was not currently pursuing nuclear weapons: lack of a major foreign threat; the alliance with the United States; too difficult technically; that doing so was not consistent with international norms and international law; that doing so was not consistent with the country’s national identity; risk of economic sanctions; risk of preventive attack; and risk that pursuing nuclear weapons would cause the country to be abandoned by the United States.⁹ Respondents could select all reasons that applied, and were separately asked to indicate which reason was the most important.

As a result of the lack of a baseline question, the analysis focuses on between-scenario comparisons, rather than comparisons between the scenarios and a baseline, again using t-tests to assess the statistical significance of the differences in means.

Findings

I present the results in two parts. First, I describe individual-level preferences for nuclear weapons, including mean levels of support and how respondents rated their own support for nuclear weapons in each of the scenarios. Second, I discuss factors respondents pointed to as explanations for why their government does not currently have nuclear weapons ambitions, and describe how respondents rated their governments’ likelihood of pursuing nuclear weapons in the scenarios.

Individual-Level Preferences

Across the sample, support for nuclear weapons is generally quite low (see Table A2 and Figures A1 and A2 in the supplementary appendix). In the baseline, mean support is 1.88 out of 5, with 76 percent of respondents indicating that they strongly or weakly opposed acquiring nuclear weapons. This general opposition holds across countries, where only one (Lithuania) had respondents whose average support reached 3 (neutral).

Figure 1 shows the differences in mean support for acquiring nuclear weapons across the four scenarios relative to the baseline. Support is higher across all four scenarios than in the baseline, which is consistent with the expectations of H1b, H1c, and H2, but inconsistent with those of H3. Moreover, support was highest in the Refusal to Defend and Troop Withdrawal scenarios, and lowest in the 2x Spending scenario, which has a difference in means relative to the baseline that only barely reaches statistical significance. Tables A3 and A4 in the supplementary appendix likewise suggest that defense spending preferences are only modestly predictive of nuclear weapons support, offering at best weak support for Hypothesis 1a.¹⁰

These results are most consistent with Hypothesis 2, as they suggest that it is fear of abandonment that drives allied interest in nuclear more so than conditional coercion or the act of having to burden-share.¹¹ Notably, however, support for acquiring nuclear weapons was still relatively low across all scenarios, with support in the Refusal to Defend condition being only about 0.5 points higher than in the baseline.

Government Preferences

Turning to how respondents rated their governments’ likelihood of pursuing nuclear weapons, because there is no baseline condition here, I instead compare the four scenarios to each other. Figure 2 indicates that respondents reported that their countries would be more likely to pursue nuclear weapons in the Refusal to Defend and Troop Withdrawal treatments compared to the Threat of Abandonment and Spending Increase conditions. Much like the individual-level results, this suggests that it is fear of abandonment, more so than coercion and burden-sharing per se, that drives interest in nuclear weapons, which is more consistent with Hypothesis 2 than H1 or H3.

To provide additional insight into the causal mechanisms, Figure 3 shows the reasons that were cited by respondents as to why their government is not currently pursuing nuclear weapons. By far, the most cited reasons were: (1) the alliance with the United States; (2) the constraints of international norms; and (3) the constraints of national identity. Moreover, nearly 50 percent of respondents said that alliance with the United States was the single most important reason their government was not pursuing nuclear weapons. Figure 4, in turn, shows that respondents who pointed to the importance of the US alliance indicated a higher likelihood their governments would seek nuclear weapons in the Refusal to Defend and Troop Withdrawal scenarios, as well as the Conditional Threat and 2x Spending Scenarios, though the effects are not statistically significant at the p < 0.05 level.¹² This suggests that in cases where the US alliance is more important, signals of abandonment and to a lesser extent allied self-reliance and conditional threats are more likely to increase nuclear weapons interest.

Discussion

On the whole, the results provide qualified support for the proposition that signals of abandonment, alliance coercion, and allied burden-sharing would increase support for acquiring nuclear weapons among elites in US allied countries. Respondents’ support for nuclearization was highest when faced with scenarios where the United States gave unconditional signals of abandonment—namely, indications that it would not defend their country or would withdraw US troops regardless of the country’s behavior. Support was lower when presented with a scenario where the United States threatened to abandon their country unless it increased defense spending, though still higher than in the baseline. Respondents answered similarly when asked about their government’s likely behavior in each scenario, rather than their own preferences.

The evidence is weaker (though not absent) that greater burden-sharing on its own is associated with greater demand for nuclear weapons. This is likely because the impetus for pursuing self-reliance by any means necessary is far greater when presented with actual signals of abandonment, compared to simply increasing defense spending without any such signals. But it is also possible that the relationship between burden-sharing and demand for nuclear weapons is weak because greater investments in defense do not necessarily imply a higher risk of abandonment and are not necessarily substitutable for nuclear weapons. This is especially likely to be the case when allies specialize in complementary weaponry and equipment. By making themselves more capable of fighting together and less capable of fighting apart, burden-sharing may enhance rather than reduce alliance cohesion (Morrow 1994; Gannon 2025). While it is difficult to tell the extent to which respondents were thinking about specialization since the survey only asked about defense spending, it is possible that at least some of them were, which could partially explain why the relationship between defense spending and interest in nuclear weapons is weaker.

Several caveats are in order when drawing inferences from these results. The first is the nature of the sample. For one, its small size means that there is greater uncertainty about effect sizes due to the limited statistical power and renders comparisons between or within countries all but impossible. Moreover, the sample cannot claim to be representative of European foreign policy elites, and one should thus not draw sweeping conclusions about the broader population. Both limitations are perennial challenges in surveys using elite samples (Kertzer and Renshon 2022).

Second, setting aside the size and representativeness of the sample, there may be features of the strategic environment that limit the findings’ generalizability. The first is the timing of the survey, which occurred within two years of Russia’s full-scale invasion of Ukraine in February 2022. The context of the ongoing war might have shaped allies’ demand for nuclear weapons and their sensitivity to US signals of abandonment in two ways. On the one hand, increased perception of the threat posed by Russia might have increased the demand for nuclear weapons as a hedge against US abandonment. On the other hand, the US and broader NATO response to the invasion may have catalyzed alliance cohesion, increased perceptions that the United States was committed to Europe, and reduced fears of abandonment (Hardt 2024). Given that respondents were explicitly asked to consider scenarios where the United States reduced its commitment, the former effect likely predominated. That is, the effects of US coercion and signals of abandonment were likely larger than they would have been in 2021, let alone in previous decades when the perception of threat posed by Russia was much lower (e.g., the 1990s–2000s). If the threat environment in Europe had deteriorated even further—for example, if Russia had conquered all of Ukraine—the effect sizes may have been larger.

One might also wonder whether the results can generalize to US allies in other regions, where perceptions of US reliability and the threat environment might systematically differ. Notably, a similar survey in South Korea suggests comparable effects among South Korean elites (Cha 2024). However, South Korea is another case where perceptions of threat are fairly high; it is thus not clear how much the results would generalize to another case where threat perception is lower.

The final limitation stems from the context of the survey itself. Surveys by necessity abstract away from important contextual details, and their results may be sensitive to which relevant details are included or not, how they are framed, and the assumptions respondents make about those details even without being explicitly told about them (Dafoe, Zhang, and Caughey 2018; Bell 2023; Brutger et al. 2023, 168–69). Drawing strong inferences from respondents’ policy preferences on surveys requires assuming both that they know what their preferences would be when faced with arguments for and against that policy in real-time, and that they are willing to communicate those preferences in the survey. Regarding the former, one cannot necessarily assume that respondents’ stated preferences in the abstract would be their preferences when faced with one of these scenarios in reality. Nevertheless, the sample is comprised of individuals with considerable expertise and influence over their countries’ foreign policies, including a sizeable number who were currently in government or had served in government previously, who likely have more well-formed preferences than the median citizen (Hafner-Burton, Hughes, and Victor 2013).

Perhaps more challenging, respondents may have strategically misrepresented their preferences. Given that the United States has historically sought to curtail nuclear proliferation, respondents may have incentive to indicate a greater likelihood that their country would pursue nuclear weapons if the United States abandoned them in order to deter that abandonment. This is difficult to rule out, but there are a few reasons to question the magnitude of the impact that incentives to misrepresent had. First, the effect sizes are greater for respondents’ individual-level preferences, where the incentive to misrepresent is less, rather than anticipated government behavior, where that incentive is higher.

Second, respondents currently or previously in government did not report greater individual willingness or expected government willingness to pursue nuclear weapons across scenarios than individuals who had never served in government (see Supplementary Tables A6 and A7). Notably, this is despite slightly higher baseline support among current government officials, meaning that differences between support in the baseline and the scenarios are even smaller for government officials.¹³ While not definitive, government officials may perceive greater incentive to misrepresent given that they are more likely to be held accountable for the consequences of being abandoned or coerced by the United States and may thus have a stronger preference to avoid those consequences.

Finally, it is notable that despite the incentive to misrepresent, along with most respondents pointing to the alliance with the United States as the most important reason for their countries not pursuing nuclear weapons, effect sizes are still fairly modest. This suggests that respondents did not fully lean into misrepresenting their preferences even though doing so would have had little cost to them in the context of an anonymous survey. Nevertheless, the magnitude of the effect that strategic misrepresentation might have had on the results is unknowable. The findings might thus represent an upper bound on the effects insofar as they may be somewhat inflated by incentives to overstate the likelihood of proliferation.

Despite these limitations, these findings offer a useful cut at understanding central questions in an area of research that is far from data-rich (cf. Jervis 2021, 131; Bell 2023, 171–172; Gavin 2024, 176–178). In this way, the findings follow a long line of recent scholarship that has analyzed surveys (Ko 2019; Sukin 2020) and wargames (Pauly 2018) to circumvent a lack of real-world data. Moreover, while the data have significant shortcomings, they also have an important advantage over analyzing historical cases—namely, that coercion and signals of abandonment are applied non-randomly and likely targeted to cases where the chances of success are high and the risks of proliferation are low.

Conclusion

This research note finds that when faced with scenarios indicating that their country had increased defense spending or that the United States had signaling that it might abandon their country, NATO foreign policy elites expressed greater support for their country acquiring nuclear weapons. Moreover, elites reported that their governments would be more likely to pursue nuclear weapons when faced with the prospect of US abandonment. The results are broadly consistent with previous theorizing that there is a tradeoff between burden-sharing and nonproliferation in US alliances (Gerzhoy 2015; Lanoszka 2015).

Nevertheless, there are two indications that this tradeoff is relatively modest. The first is that the approach to encouraging burden-sharing matters a great deal. Even in a scenario where the United States threatened to abandon an ally unless it increased defense spending, respondents were less interested in nuclear weapons than in a scenario where the United States flatly indicated its unwillingness to defend the ally or announced its intention to withdraw US forces. Second, the effects were moderate; even when the United States indicated that it would not defend an ally, most respondents opposed acquiring nuclear weapons and reported that their governments were still unlikely to pursue nuclear weapons. Taken together, these points suggest reasons to be optimistic about the United States’ ability to encourage burden-sharing without also encouraging proliferation, as existing research suggests that conditional threats of abandonment can be equally effective as unconditional signals of abandonment for increasing support for burden-sharing (Blankenship 2023, 2024).

The findings and their limitations suggest several avenues for future research. The first is to probe the findings’ generalizability across strategic contexts. While surveys can offer insight into elite preferences, scholars should investigate the degree to which the tradeoff between burden-sharing, coercion, and signals of abandonment, on the one hand, and nonproliferation, on the other, holds in practice across a wide range of cases. Recent work by Cha (2024) suggests 53 percent of South Korean elites opposed nuclearization, but that a majority supported acquiring nuclear weapons “If an ‘America First’ policy returns to the White House in November 2024 that denigrates allies and seeks retrenchment” (Cha 2024, 11). This suggests a higher baseline level of support and a greater effect of abandonment signals than in this study’s survey of NATO elites, where more than 70 percent of respondents opposed nuclearization in the baseline and there was significantly less than majority support even when faced with abandonment signals. Nevertheless, it suggests effects that are in the same direction. Further work could extend the methodology to contexts where perceptions of external threats are not as high.

Second, to address the shortcomings of small sample size and non-representativeness that often accompanies elite surveys, further research drawing from multinational public opinion surveys may prove fruitful. Doing so would offer greater statistical power, more promise of obtaining representative samples, and greater ability to detect heterogeneity within and across countries. Use of public opinion samples may even mitigate the effects of strategic exaggeration of the likelihood of proliferation described earlier, as members of the public may not be as acutely aware of the United States’ desire to prevent allied proliferation using security assurances and may thus not perceive the same incentives to misrepresent their policy preferences. Survey experiments in particular may prove especially useful, as research suggests that elites and the public respond to survey experiments in similar ways even though their baseline political attitudes may differ (Kertzer 2022).

Author Biography

Brian Blankenship is an Assistant Professor of Political Science at the University of Miami. His book, The Burden-Sharing Dilemma: Coercive Diplomacy in U.S. Alliance Politics, was published by Cornell University Press in 2023.

Notes

Author's note: I am very grateful for the feedback I received on previous drafts of the paper from Joshua Alley, Alexandre Debs, Andres Gannon, Tongfi Kim, Do Young Lee, Nicholas Miller, Lauren Sukin, Jamie Withorne, and participants at the 2024 APSA conference and the Oslo Nuclear Project’s Online Seminar Series. Special thanks to those experts and government officials who took the survey. This research was supported with funding from a Nuclear Security Grant from the Stanton Foundation.

Data and replication materials available on the ISQ Harvard Dataverse: https://dataverse.harvard.edu/dataverse/isq (accessed April 9, 2025). Please direct all questions to the author.

Footnotes

References