United States president Donald Trump consistently criticizes NATO allies for failing to meet defense spending targets, even provocatively stating he would “encourage Russia to do whatever they want” to underspending members (FitzGerald Reference FitzGerald2024). While Trump’s rhetoric has been characterized as undiplomatic, such pressure on allies is not unprecedented: the Biden administration successfully negotiated increased Japanese contributions for American troops’ stationing costs in Japan, former president Barack Obama launched an “anti-free-rider campaign” in Libya (Galbert Reference Galbert2016; Goldberg Reference Goldberg2016), and Presidents Dwight Eisenhower and Lyndon Johnson similarly urged European allies to increase their contributions to collective defense (Becker et al. Reference Becker, Kreps, Poast and Terman2024). Although tensions surrounding defense burden sharing and underspending have been a focal point of both public interest and scholarly discourse (e.g., Auerswald Reference Auerswald2004; Haesebrouck Reference Haesebrouck2017; Hallams and Schreer Reference Hallams and Schreer2012, Sandler and Shimizu Reference Sandler and Shimizu2014), “defection” within a security community can manifest through diverse mechanisms as alliance “burden” extends beyond mere financial expenditures, but may include refusal to participate in military operations (as with Germany and France’s opposition to the 2003 Iraq War), purchasing weapons from rival countries (Turkey’s acquisition of Russian S-400 missile defense systems), blocking accession of a new member into the alliance (Turkey’s and Hungary’s impediments to Sweden’s NATO membership), and inconsistent support for extraterritorial sanctions (the European Union’s reluctance to join in sanctions targeting Iranian and Chinese entities; see Daniels Reference Daniels2024; Geranmayeh Reference Geranmayeh2018).
As these examples demonstrate, member-states may strategically deviate from alliance expectations, potentially undermining the alliance’s cohesion and other parties’ strategic interests. In such instances, the stronger partner has several options: it can ignore these deviations, particularly if it decides not to risk the alliance’s unity, or it can adopt “corrective measures” to reinforce compliance and deter other members from pursuing uncooperative policies. These corrective strategies encompass a continuum from rhetorical actions, such as public shaming and blaming, through tangible coercive measures, such as sanctions or suspending military support, to alliance termination.
A critical determinant moderating the effectiveness of corrective measures is their perceived credibility (Snyder Reference Snyder1997), which is intrinsically linked to the domestic political endorsement these measures garner in the sender country. When corrective measures lack popular legitimacy, the targeted state may anticipate and strategically exploit potential audience costs, thereby undermining the intervention’s potential efficacy. Conversely, corrective measures with robust public backing are perceived as more credible by the target and carry enhanced diplomatic leverage.
Notwithstanding the pivotal role of public attitudes in determining the adaptability and effectiveness of alliance management strategies, existing scholarship has not systematically examined this dimension. Aiming to fill this gap, I examine the public preferences for corrective policies toward security allies that renege or underperform on their alliance expectations. More specifically, I explore the following question: what drives support for intra-alliance coercion conditional on ally uncooperativeness among American citizens? I advance three primary hypotheses. First, citizens of the dominant partner will demonstrate a preference for corrective measures against uncooperative allies, with a particular predilection for high-coercion interventions. Second, exposure to information regarding allies’ uncooperative behaviors will stimulate negative alliance perceptions and demands for reduced national contributions to the alliance. Third, contextual variables moderate citizens’ support for corrective measures, namely the uncooperative partner’s regime type, military power, and whether it has a formal alliance treaty with the US.
Empirical analysis based on a preregistered survey experiment on a nationally representative sample of over 1,502 American citizens provides substantial evidence for these hypotheses.Footnote 1 First, the findings reveal that Americans predominantly endorse corrective measures against uncooperative allies, with economic sanctions and military aid reduction receiving significantly greater support than rhetorical approaches, such as public denouncements. Second, allies’ uncooperative actions trigger demands for reduced American commitments, though the public response varies considerably depending on the nature of the transgression. Notably, weapons procurement from rival countries such as Russia or China or refusal to deploy troops for joint operations are perceived as more egregious offenses than financial underspending on defense. Third, the regime type of the uncooperative ally strongly conditions public support for adopting corrective measures, as the public is less likely to support measures against democratic allies. Conversely, neither the military capabilities of the uncooperative ally nor the existence of formal treaty arrangements with the US exert statistically significant moderating effects.
This study provides the first experimental evidence of American support for wielding pressure and restraining uncooperative members within a security alliance. As such, it contributes substantially to the burgeoning field of empirical research exploring public attitudes toward military alliances (e.g., Chu, Ko, and Liu Reference Chu, Ko and Liu2021; Lee and Goidel Reference Lee and Goidel2022; Tomz and Weeks Reference Tomz and Jessica2021). While preceding studies have predominantly focused on how alliance commitments incite public support for the use of force to defend allies (e.g., Kreps and Kriner Reference Kreps and Kriner2024; Tomz, Weeks, and Bansak Reference Tomz, Jessica and Bansak2023), this study explores an alternative dimension: the role of alliance discord in shaping preferences for punitive measures toward less cooperative members.
In the remainder of the paper, I first theorize on the proposed public preferences for incumbent response toward uncooperative security allies and the contextual factors that may moderate these preferences. Subsequently, I present the preregistered vignette experiment to test my theoretical expectations, followed by a detailed discussion of the findings. The concluding section provides an overview of the results, discusses its policy implications, and addresses avenues for future research.
Alliance Management and Support for Corrective Measures
Security alliances are rarely frictionless, and cohesion within an alliance is not a given. By their fundamental nature, security alliances embody a collective action problem. Once membership is granted, participants cannot be excluded from the alliance’s benefits, and the accession of additional members does not diminish the utility existing members derive from the alliance’s public goods (Olson and Zeckhauser Reference Olson and Zeckhauser1966). Consequently, smaller partners are incentivized to engage in “free-riding” behavior if they perceive the dominant partner to bear disproportionate costs in maintaining these public goods. This anticipation stems from a rational calculation that the powerful partner, possessing substantial resources, derives greater utility from the public good relative to its maintenance costs and faces higher potential losses from its termination (Thies Reference Thies1987). Thus, in the absence of robust enforcement mechanisms, free riding, manifesting in various forms from inadequate financial contributions to reluctance to deploy troops, emerges as a dominant strategy for smaller partners.
In response to such uncooperative behavior, major powers may need to exert pressure on their allies. Since allies cannot credibly threaten the use of force against each other, they employ corrective strategies to wield influence on allies, punish the “defection,” and incentivize future cooperation. One critical distinction within corrective measures concerns their visibility to external actors. Private corrective measures minimize reputational costs for both sender and target, reduce the risk of revealing alliance vulnerabilities to adversaries, and preserve face-saving opportunities that facilitate compliance. However, signals are considered more credible when they involve a cost or risk, such as a leader staking their reputation before a domestic audience. As a result, private measures also face important limitations: they cannot generate the same level of audience costs or domestic political benefits, such as displaying decisive leadership, that accompany visible action (Yarhi-Milo Reference Yarhi-Milo2013). By contrast, public corrective measures serve a dual functionality: they not only pressure the target state but also signal resolve to domestic audiences and other alliance members about the potential consequences of alliance misalignment (Fearon Reference Fearon1994).
A second critical distinction within corrective measures concerns their intensity. At lower intensity levels, corrective measures may rely on rhetorical strategies. Rhetorical actions, such as naming and shaming, leverage the shared normative framework to which the alliance members adhere. The effectiveness of this approach lies in its capacity to invoke collective expectations and obligations, thereby fostering a sense of accountability among members through social mechanisms. Rhetorical measures involve public identification and condemnation of uncooperative behavior, aiming to alter policies or practices that contravene alliance treaties, norms, or principles (Keck and Sikkink Reference Keck, Sikkink, Meyer and Tarrow1998; Lebovic and Voeten Reference Lebovic and Voeten2009). Rather than inflicting material damages, this strategy relies on generating reputational costs both among peer nations and their domestic constituencies.
Because governments often have status concerns and are sensitive to their reputations, shaming can be an effective strategy, particularly in pressurizing target governments to promote norm compliance (e.g., Crescenzi Reference Crescenzi2018). However, when target governments assess that any policy modification will inflict greater costs than the imposed measures, they may resist change (e.g., Tingley and Tomz Reference Tingley and Tomz2022). The absence of material threats in rhetorical actions reduces the cost of resistance for target actors, potentially diminishing their effectiveness. In security alliances, Becker and colleagues (Reference Becker, Kreps, Poast and Terman2024) demonstrate that public shaming of NATO allies’ underspending is ineffective and counterproductive for future burden-sharing prospects. These adverse effects are documented to extend to the citizen level: Blankenship (Reference Blankenship2024) finds that threats and shaming fail to significantly alter public support for defense burden sharing in NATO allies such as Germany and Poland as long as public confidence in US protection remains intact.
At higher intensity levels, corrective measures may entail concrete coercive actions. Implementing or threatening economic measures or military aid suspensions represents a more effective restraint mechanism than persuasion and rhetorical approaches as they mobilize power resources. In a seminal paper on alliance cohesion, Kupchan (Reference Kupchan1988), testing competing theories, argues that two mechanisms were instrumental in restoring cohesion between NATO allies in the Persian Gulf in the early 1980s: the alliance security dilemma explanation, which proposes that cohesion depends on the leading power’s coercive ability to enforce cooperation among its less powerful allies; and the domestic politics argument, which contends that alliance behavior is primarily shaped by domestic pressures for or against cohesion. In another example, Pressman (Reference Pressman2012) documents how, during the 1991 Persian Gulf War, the US successfully prevented Israeli retaliation against Iraq through a combination of incentives and punitive threats while preserving coalition integrity.
Beyond their documented effectiveness, leaders adopt corrective measures to capitalize on domestic electoral advantages. These measures serve a symbolic function in satisfying constituent expectations (Galtung Reference Galtung1967). This argument is grounded in the assumption that voters expect decisive leadership responses when informed about an international dispute, viewing such actions as demonstrations of executive competence (McLean and Whang Reference McLean and Whang2014). The literature on sanctions gives credence to this proposition. Whang (Reference Whang2011) shows that US presidents experience a boost in domestic approval rates when they impose economic sanctions on their adversaries. Despite the potentially costly nature and uncertain policy outcomes, leaders consistently employ sanctions to project their “strong leadership” to their constituencies. Similarly, sanctions removal often presents significant political challenges as leaders seek to avoid being perceived as “weak” and accommodating by their constituencies (Drezner Reference Drezner2021).
Analogous dynamics suggest that voters would expect a strong response against alliance discord. Two individual-level motivations sway opinions on alliance management. First, humans are psychologically inclined toward fairness and favor decisions that reduce inequity in outcomes (Fehr and Schmidt Reference Fehr and Schmidt1999). The behavioral economics literature consistently demonstrates that individuals’ decision making in collective contexts exhibits strong reciprocal tendencies; cooperative behavior elicits reciprocal cooperation, while uncooperative actions prompt retaliatory responses (e.g., Fehr and Gächter Reference Fehr and Gächter2000). In foreign policy, reciprocal exchanges cue citizens with greater support for cooperation (e.g., Chilton, Milner, and Tingley Reference Chilton, Milner and Tingley2020). Conversely, when citizens perceive that allies are “free riding” or failing to contribute their fair share, this violates deeply held reciprocity norms, triggers perceptions of unfairness, and generates support for corrective measures.
Second, citizens evaluate policies based on their expected effectiveness. Policy proposals perceived as more likely to succeed garner considerably higher support (e.g., Gelpi, Feaver, and Reifler Reference Gelpi, Feaver and Reifler2009; Kohno et al. Reference Kohno, Montinola, Winters and Kato2021). When faced with an uncooperative ally, citizens would be expected to evaluate corrective measures based on their perceived likelihood of changing behavior and deterring future noncompliance. As the intensity of corrective measures increases, domestic constituents are more likely to anticipate successful outcomes. While rhetorical measures carry lower costs and are easier to resist by the target, economic and military measures impose tangible costs, demonstrate resolve, and are thus considered more effective responses to manage discord.
However, these are not preordained expectations, as the public’s investment in alliance partnerships and recognition of their strategic value might generate opposition to punitive actions that risk partner alienation or even alliance dissolution. Research shows that Americans hold distinctly favorable attitudes toward allies and alliance relationships. Both Tomz and Weeks (Reference Tomz and Jessica2021) and Tomz, Weeks, and Bansak (Reference Tomz, Jessica and Bansak2023) show that alliance ties significantly increase public support for defending partners. Similarly, Chu, Ko, and Liu (Reference Chu, Ko and Liu2021) demonstrate that alliance rhetoric emphasizing shared values and mutual benefits resonates strongly with American audiences. Such favorable dispositions toward allies might translate into questions about the strategic utility of corrective measures. Under such contexts, citizens might prefer diplomatic solutions or simply tolerate ally uncooperativeness rather than endorse measures that could undermine relationships they view as ultimately beneficial.
From this discussion, I develop two hypotheses:
H1: citizens favor implementing corrective measures against security allies that renege on their alliance expectations.
H2: among different variants of corrective strategies against uncooperative security allies, public support will be greater for economic and military measures than for rhetorical measures.
Citizens, as taxpayers, care about the fiscal implications of international cooperation. Dellmuth and Tallberg (Reference Dellmuth and Tallberg2015) note that member-state contributions to international organizations frequently emerge as points of political contention within major donor states. Research indicates that during periods of economic instability, public support for resource-intensive foreign policy initiatives, such as foreign aid, experiences a notable decline (Heinrich, Kobayashi, and Bryant Reference Heinrich, Kobayashi and Bryant2016). In a survey experiment among American voters, Lee and Goidel (Reference Lee and Goidel2022) establish that emphasizing the financial costs of the NATO alliance for the US significantly diminishes respondents’ favorable perceptions of the alliance.
Public sensitivity to the cost of international cooperation expectedly heightens attention to contribution patterns among alliance partners. Asymmetrical contributions incite concerns over equity and fairness of the outcome (e.g., Brutger and Rathbun Reference Brutger and Rathbun2021; Yeung and Quek Reference Yeung and Quek2022). Consequently, other actors’ contribution levels significantly influence assessments of individual contributions to public goods. For instance, tax compliance research indicates that individual adherence to tax regimes remains robust when widespread compliance is anticipated but diminishes substantially when noncompliance becomes prevalent (Alm, McClelland, and Schulze Reference Alm, McClelland and Schulze1999). This behavioral pattern extends to international alliance contexts: Kiratli and Ertan (Reference Kiratli and Ertan2024) demonstrate that American voters exhibit pronounced sensitivity to other nations’ financial contributions to the United Nations in their assessment of US contributions. Building upon such empirical foundations, I propose that exposing citizens to information about alliance members’ noncooperative behaviors should generate both negative evaluations of the alliance relationship and demands for reduced American contributions.
H3: when informed about the presence of uncooperative security allies, citizens will be (1) more likely to demand reduced contributions to the alliance from their governments and (2) less favorable toward the alliance.
Potential Moderators
Ally Regime, Military Power, and the Alliance Structure
Public support for corrective strategies is fundamentally contingent upon their perceived costs, anticipated effectiveness, and normative appropriateness. Constituents would be expected to prefer measures that minimize domestic costs while maximizing effectiveness and maintaining normative legitimacy (Heinrich, Kobayashi, and Peterson Reference Heinrich, Kobayashi and Peterson2017; McLean and Roblyer Reference McLean and Roblyer2017). I argue that three contextual factors—namely the target country’s regime type, its military strength, and its formal treaty status with the US—significantly influence American citizens’ perceptions of the cost, effectiveness, and normative legitimacy of corrective measures.
Ally Regime Type
Democracies are more likely to honor their international commitments and are considered more reliable allies due to distinct institutional characteristics, including robust electoral accountability mechanisms, elevated domestic audience costs, bureaucratic continuity, and multiple veto points (e.g., Gaubatz Reference Gaubatz1996; Leeds Reference Leeds2003; Pilster Reference Pilster2011). When members of a partnership share interests, they have less incentive to adopt costly monitoring measures and control mechanisms to influence each other’s actions (Epstein and O’Halloran Reference Epstein and O’Halloran1999). Cognizant of democratic reliability, public constituencies demonstrate greater reservations when implementing corrective measures against other democratic partners, recognizing the potential risk to alliance longevity. Even when they act uncooperatively, democratic partners receive the benefit of the doubt, and their defection is devalued.
This instrumental reasoning is accompanied by a normative component: democracies adhere to a similar set of normative identities. Citizens and elites in democratic countries tend to view each other as part of the same moral community, committing to shared civic values and liberal ideals. These shared worldviews lead democratic citizens to view one another as more reasonable and trustworthy, reducing perceived threats from each other (Dixon Reference Dixon1994). Even when disputes arise between democratic dyads, the public is more likely to believe that the mutual commitment to negotiation, deliberation, and compromise will prevent disagreements from escalating into larger conflicts. Accordingly, aggressive actions, such as high-intensity corrective measures against fellow democracies, will not only be considered as functionally counterproductive but also morally unacceptable (Bueno de Mesquita and Siverson Reference Bueno de Mesquita and Siverson1995; Risse-Kappen Reference Risse-Kappen1995).
H4: public support for corrective strategies against democratic allies will be lower than public support for corrective strategies against authoritarian allies when they exhibit uncooperative behavior.
Ally Military Power
Citizens’ instrumental effectiveness concerns should make them more skeptical of corrective measures against militarily powerful allies. Adopting corrective measures against powerful allies is both costlier and strategically riskier, regardless of interest alignment between partners.
From a strategic value perspective, powerful allies are particularly costly to alienate. In cases of overlapping long-term interests, corrective measures against a powerful partner risk losing a valuable ally that can shoulder substantial burdens in collective defense and help the home country to project global influence. Conversely, when interests diverge, maintaining alliance structures remains imperative because the potential dissolution of formal partnerships will reduce opportunities to influence a potential rival’s policies in the future.
More fundamentally, corrective strategies are less effective when directed at militarily powerful partners. States with advanced military capabilities maintain greater autonomy in pursuing their security interests and exhibit reduced alliance dependence, making them more resistant to external pressure. By contrast, states with limited military capabilities face stronger incentives for balancing and will be more reliant on alliance-provided deterrence. Consequently, they will be more motivated to preserve alliance cohesion and more cautious about alienating hegemonic partners that might become future adversaries. Their limited capabilities make them more susceptible to external pressure.
Anticipating these dynamics, citizens of the stronger partner should be more willing to support corrective measures against weaker allies, expecting lower costs and higher likelihood of success. Even in worst-case scenarios where coercion results in alliance dissolution, losing a militarily weaker partner represents a negligible strategic cost.
H5: public support for corrective strategies against militarily powerful allies will be lower compared to militarily weaker allies when they exhibit uncooperative behavior.
Alliance Structure
The international relations literature distinguishes between formal alliances, codified through official treaties, and informal alignments, strategic partnerships based on shared interests without binding commitments, such as US–Israel or US–Saudi Arabia relations (Snyder Reference Snyder1997; Wilkins Reference Wilkins2012). Treaty-based alliances like NATO typically specify commitments, such as collective defense obligations or spending targets, while informal security partnerships often operate through implicit understandings about shared strategic interests without explicit obligations.
Formalized ties provide partners with organizational arrangements that facilitate negotiations, enhance information sharing and monitoring capabilities, and enable issue linkages (Press-Barnathan Reference Press-Barnathan2006). These institutional features not only promote cooperative behavior within security alliances but also create avenues for allies to influence each other and exert restraint for undesired policies through noncorrective mechanisms. Over time, members of institutionalized cooperation are expected to “socialize” into shared understandings, internalize common behavioral standards, and develop collective identities (March and Olsen Reference March and Olsen1983). These dynamics shape how citizens perceive corrective measures within formal alliances. Because treaty-based partnerships establish stronger mutual obligations and emphasize consultation over unilateral coercion, efforts to pressure a formal ally may conflict with established expectations of alliance behavior. Such actions risk violating resonant norms, fostering perceptions of unfairness and disloyalty among the public and undermining the moral obligation to act in accordance with the partnership. Since formalized alliances equip member-states with extensive diplomatic tools to exert influence and create normative expectations against unilateral coercion, citizens will likely exhibit greater skepticism toward corrective measures in these contexts.
However, an alternative perspective suggests that citizens may not differentiate between formal treaty obligations and informal partnership expectations when evaluating foreign policy discord. Research on public attitudes toward international commitments indicates that people prioritize the substantive nature of cooperative relationships over their legal technicalities (e.g., Toshkov, Mazepus, and Dimitrova Reference Toshkov, Mazepus and Dimitrova2024). Since diplomatic negotiations, working-group meetings, and bureaucratic coordination within formal alliances are not subject to public scrutiny, citizens tend to focus on tangible outcomes, such as whether partners contribute their “fair share” to collective security, regardless of how those contributions are assigned (Kiratli and Ertan Reference Kiratli and Ertan2024).
Similarly, political debates on burden sharing, free riding, and partner reliability often blur the lines between formal allies and informal security partners. For instance, while NATO’s 2% defense spending target is not legally binding, it has shaped political expectations similar to those placed on other US security partners, such as South Korea and Japan, regarding increased contributions to shared security objectives (e.g., O’Shea and Maslow Reference O’Shea and Maslow2021). The absence of clear institutional distinctions in public discourse, coupled with citizens’ emphasis on concrete policy outcomes rather than legal frameworks, suggests that alliance formalization may have little influence on public support for corrective strategies.
H6a: public support for corrective strategies against uncooperative allies will be lower if there is a formal, written agreement between the sender country and the target.
H6b: public support for corrective strategies against uncooperative allies will not significantly differ based on whether a formal agreement exists between the sender country and the target.
Methodology
To empirically test my hypotheses, I conducted a preregistered, multifactorial vignette experiment from December 13 to December 20, 2024. The experiment was fielded on the Qualtrics platform on a sample of 1,502 voting-eligible respondents in the US. To maximize representability, the survey employed a quota sampling design based on gender, age, and education according to US Census data. Although there were no quotas on partisanship, the resulting sample was substantially representative in terms of political orientation: 40.65% of the sample identified as Republican, 40.10% as Democrat, and 19.25% self-identified as independent. The survey included two attention-check questions; those who failed these checks were not included in the final sample, although I included them in the robustness checks. Prior to the experiment, respondents answered questions about their demographics, and posttreatment, they were asked questions regarding ideology and political attitudes. Table A1 in the online appendix presents the details of sample characteristics, and appendix table A2a reports the balance tables for treatment groups.
The experiment started with the following introduction:
As you may know, the United States maintains numerous military alliances worldwide. Some of these alliances are formalized through treaties, while others are based on mutual understandings and shared interests. However, there are times when certain allies take actions that can contradict the interests of the United States. We would like you to evaluate five of these cases. Please read each case carefully and tell us how favorable you are to the government’s response.
Following this information, respondents were asked to evaluate a vignette about a hypothetical ally that randomly varied five attributes.
The first attribute is the type of uncooperative action. While I do not develop specific hypotheses on the moderating role of uncooperative acts, manipulating the types of noncompliance is important because, first, it enhances the external validity of findings by testing public reactions across the full spectrum of alliance discord; and second, it allows for post hoc exploration of whether citizens distinguish between different categories of uncooperative actions. To capture this multidimensional nature of alliance noncompliance, I selected five behaviors to represent distinct categories of uncooperative actions, with each having historical precedents: financial contributions (underspending on defense), military cooperation (refusing troop deployment), strategic alignment (weapons procurement from rivals), institutional governance (blocking new member accession), and sanctions coordination (breaking US sanctions). This diversity allows for testing whether public attitudes vary systematically across different domains of alliance cooperation.
The next three attributes manipulated the ally’s regime type, its military capabilities, and whether the alliance is based on a formal treaty. The fifth and final attribute, namely the US response, is the key independent variable, varying at five levels. These five responses span different levels of corrective measure intensity and strategic commitment. While public denouncement represents low-intensity rhetorical pressure, economic measures and reducing military support constitute moderate-intensity coercion aimed at changing target behavior while preserving the alliance relationship. Alliance termination occupies a unique position as it combines high punitive impact with strategic disengagement, simultaneously inflicting maximum costs on the target while signaling the sender’s willingness to abandon the relationship entirely.
Each vignette followed the following format:
Country X [renamed differently in each task], an ally of the US, attribute 1 [one of the following five uncooperative acts: refuses to meet the 2% GDP defense spending target; refuses to send troops to a joint operation with the US; purchases weapons from Russia and China that are incompatible with the American systems; blocks the accession of a new member into the alliance despite US demands; or breaks US sanctions against a third country].
Country X
-
• has an attribute 2 [regime type: democratic, authoritarian, or partially democratic] regime.
-
• has attribute 3 [military power: advanced military capabilities, moderate military capabilities, or basic military capabilities].
-
• has attribute 4 [formal alliance: a treaty-based alliance with the US or an informal alliance with the US].
In response to this uncooperative action, the US government plans to attribute 5 [US response: publicly denounce Country X’s actions; impose economic measures against Country X, such as trade restrictions or financial penalties; reduce military support to Country X; break the alliance with Country X; or continue current policies without change].
Afterward, respondents were first asked to evaluate the US response, the primary dependent variable, with the following item: “How much do you agree with the proposed action toward Country X?” (11-point scale, from zero [do not agree at all] to 10 [agree completely]).
This design measures respondent approval following exposure to a policy proposal treatment, in conformity with established experimental approaches in political communication (e.g., Arceneaux Reference Arceneaux2008), as well microlevel tests of audience cost theory (e.g., Kertzer and Brutger Reference Kertzer and Brutger2016). This experimental strategy has certain advantages over one that measures approval of all the policies for each profile or forces respondents to choose between policies. First, in real-world situations, citizens typically respond to their government’s foreign policy proposals ex post facto rather than proactively selecting and mandating specific policy alternatives for implementation. In other words, they often encounter and evaluate policy positions sequentially rather than comparatively when receiving political information. Hence, having respondents assess the government’s policy position better reflects ecological validity. Second, when respondents evaluate multiple policies, their responses may be tainted by cognitive burden and derailed by spillover effects such as anchoring, where initial policy ratings affect subsequent ones, leading to satisficing behavior. Third, forced-choice designs, in which respondents are presented with multiple policies and have to evaluate each policy relative to others, introduce confounding factors as they are based on relative preferences rather than absolute evaluations, and may lead to biased estimates.
Following the primary dependent variable question, to test H3 on support for the US contributions to the alliance and on alliance favorability, respondents were asked two further questions in a randomized order:
-
1. Now, let’s consider the American contributions to the alliance between Country X and the US. Should the government increase, maintain, or reduce its financial and military contributions to this alliance? [three-point scale, from one (decrease) to three (increase)].
-
2. How favorable are you to having an alliance with Country X? [from zero (not favorable at all) to 11 (extremely favorable)].
Figure A1 in the online appendix presents a sample vignette, as shown to the respondents.
Each respondent had to complete five separate tasks, resulting in over 7,500 observations (Kiratli Reference Kiratli2025). Power analysis (appendix table A2b) confirms adequate statistical power (greater than 80% power) for detecting substantively meaningful (of 0.5 points representing 5% of our zero-to-10 scale) direct and interaction effects. The vignette treatments were fully randomized so that the values of attributes were approximately uniformly distributed across resulting vignette tasks (see appendix table A2a for balance tests).
Findings
To test H1 and H2, I regressed approval of the US response on the selected policy action with standard errors clustered at the respondent level to correct for within-respondent correlation. The left panel of figure 1 plots the predictive margins of corrective measures—that is, the predicted value of policy approval for different levels of corrective-measure treatment if everyone in the sample had been in that group.Footnote 2 The right panel disaggregates treatment effects across uncooperative acts, and plots the predicted values of policy approval across corrective measures separately for various types of uncooperative actions.
Predictive Margins of Corrective-Measure Treatment: Unconditional and Conditional Effects
Notes: The x-axis presents the predictive margins of corrective-measure treatment on policy approval with 95% robust confidence intervals. The right panel plots the conditional effects across different values of the uncooperative-action treatment.
The results strongly demonstrate that Americans favor any action over inaction. All four policies exhibit positive, significant, and sizable effects on policy approval compared to the reference category of “keeping the current policies without change.” However, there is considerable variation among different types of measures. Among the four corrective measures, the clear favorite is reducing military support for the uncooperative ally. If the US chooses this option, it boosts policy approval from 4.77 (4.61, 4.94) for inaction to 5.82 (6.23, 6.55), with an effect size of 1.62 points (p = 0.000) on a 10-point scale. In fact, this option is so popular that it attains statistical significance even when changing the base category to the second-most popular option, economic measures, with a difference of 0.19 points (p = 0.045). Breaking the alliance and public denouncement lag behind in terms of public support, although both are significantly more popular options than inaction. Breaking the alliance increases the approval rate by 1.05 points on average, while denouncement raises it by 1.24 points; both are statistically significant at the 0.001 level.
These results lend clear evidence to both H1 and H2. Americans fully favor corrective measures toward uncooperative security allies, with higher-intensity coercive measures, namely economic sanctions and cutting military support, garnering substantially greater support than rhetorical actions. On a positive note, breaking off the alliance is the least favorable action. This finding suggests that Americans view alliance termination as a costly last resort rather than a viable coercive tool, indicating a public preference for corrective measures that preserve strategic relationships while addressing uncooperative behavior.
Having established American citizens’ support for corrective measures toward uncooperative allies, I next investigate whether their willingness to contribute to an alliance declines when another member acts uncooperatively (H3). To analyze this, I rely on data from two questions: a three-point item assessing support for American contributions and an 11-point item measuring their favorability toward having an alliance with an uncooperative ally. For easier interpretation, I rescaled the three-point contribution item to align with the 11-point scale and estimated linear regression models for both dependent variables based on experimental treatments. Figure 2 presents the average marginal component effects of each treatment level on policy approval. For each categorical variable, effect sizes are compared to reference levels, which are denoted by dots without confidence intervals (see table A4 in the online appendix for model estimates).
Effects of Alliance Attributes on Public Support for American Contributions and Alliance Favorability
Notes: Effects of alliance attributes on public support for American contributions (left panel) and alliance favorability (right panel). Horizontal lines indicate 95% robust confidence intervals; points without lines indicate the reference categories. The models control for corrective-measure treatment.
With regard to American contributions, the results reveal that purchasing weapons from rivals and refusing to deploy troops for joint operations are the two uncooperative triggers that lead respondents to demand diminished burden sharing, whereas underspending emerges as the most forgivable offense. Specifically, compared to the baseline category of vetoing new members, purchasing rival weapons makes respondents 0.51 points (p = 0.000) more favorable toward reducing American contributions to the alliance with the uncooperative ally, while the refusal to deploy troops results in a 0.23-point increase (p = 0.740). If we change the reference level to underspending, the effect size for the refusal to deploy troops becomes stronger, showing an increase of 0.31 points with a p-value of 0.014.
Similarly, regarding alliance assessments, respondents hold significantly more favorable attitudes toward an alliance with an uncooperative ally that underspends on defense contributions compared to one where an ally vetoes the accession of new members that the US supports. Conversely, alliances featuring a member purchasing weapons from Russia or China receive considerably negative assessments. Although breaking sanctions or refusing to send troops do not exert significantly different effects compared to the veto condition, both attain statistical significance when compared to underspending, with effect sizes of 0.41 (p = 0.000) for breaking sanctions and 0.36 (p = 0.000) for refusing to send troops.
Considering other treatment effects, we observe that regime type is a highly potent predictor of support for alliances, lending preliminary evidence for H4. Compared to autocracies, respondents are significantly more supportive of alliances with other democracies, even if they act uncooperatively (0.60 points, p = 0.000), and are less likely to favor diminishing American contributions in such alliances (0.47 points, p = 0.000). The presence of a treaty does not affect preferences regarding American contributions but spurs favorable attitudes toward the alliance, as shown in the right panel. On the other hand, the military power of the uncooperative ally does not significantly impact citizen assessments.
I next investigate the moderating roles of the ally’s regime type, its military power, and the presence of a treaty on approval of corrective measures (H4–H6). To test this set of propositions, I interact the corrective-measure treatment with contextual factors and regress policy approval on the relevant interaction terms (see table A5 in the online appendix). Figure 3 plots the average marginal effects of regime type (left panel), military power (middle panel), and formal treaty status (right panel) across values of corrective measures on policy approval. The dots signify median estimates, and horizontal lines indicate the 95% confidence intervals.
Conditional Effects of Target Regime, Military Power, and Treaty-Based Alliance
Notes: The x-axis presents the average marginal effects of the target regime (left panel), military power (middle panel), and the treaty-based alliance (right panel) across values of corrective-measure treatment on policy approval with 95% robust confidence intervals.
The results of the average marginal effects of regime treatment provide considerable support for H4. Each unit increase in the democracy treatment decreases policy approval for breaking the alliance by 0.28 points (p = 0.002), for imposing economic sanctions by 0.16 points (p = 0.052), and shifts approval for keeping current policies down by 0.27 points (p = 0.003). In other words, American citizens are more critical of government proposals to maintain the status quo and more supportive of proposals to terminate the alliance or impose economic measures when the uncooperative ally has an authoritarian regime rather than a democratic one. When the government proposes either denouncing uncooperative actions or reducing military support, citizens do not appear to differentiate between allies of different regime types.
For a more detailed presentation, figure A2 in the online appendix reports the predictive margins of ally regime type across different levels of corrective-measure treatment. The results show that public approval for a status quo policy toward an uncooperative authoritarian ally is only 4.48 points, whereas it increases by 0.53 points (p = 0.003) to 5.01 for a democratic uncooperative ally. Similarly, respondents are 0.56 points (p = 0.002) more favorable toward breaking the alliance and 0.32 points (p = 0.051) more favorable toward consenting to economic measures against an authoritarian ally than a democratic one. Conversely, there is virtually no difference between approval rates for denouncing an authoritarian ally (6.04 [5.79, 6.27]) and a democratic one (6.03 [5.79, 6.28]), and only a weak difference for reducing military support, with a 0.19-point increase in approval if the uncooperative ally has an authoritarian regime as opposed to a democratic one (p = 0.280).
When considering the moderating roles of military power and formal alliances, the results robustly disprove the theoretical expectations of H5 and H6a. As findings in the middle panel of figure 3 corroborate, the military-power treatment does not exert a statistically significant effect on policy approval for any level of corrective-measure treatment. These null findings persist even when comparing mean differences in policy approval for allies at the opposite ends of the military-power treatment: there is no statistically significant difference in support for any type of corrective action between allies with advanced and basic military capabilities. For instance, support for economic sanctions on advanced military allies is only 0.07 points (p = 0.669) lower than it is toward allies with basic military capabilities. Likewise, while support for denouncing is moderately higher toward allies with basic military capabilities by 0.21 points, this difference remains statistically insignificant (p = 0.196).
Regarding the alliance structure, as shown in the right panel of figure 3, the effect sizes for the corrective-measure treatment do not significantly vary between allies with or without a formal alliance with the US. Expectedly, the closest action that comes to attaining statistical significance is breaking the alliance; respondents are 0.17 points (p = 0.212) less favorable toward breaking an alliance with an uncooperative ally if there is a binding formal treaty in place. Taken together, these results conclude that American citizens do not differentiate between militarily powerful or weak allies, nor do they consider whether the US has a formal alliance treaty when contemplating proposals for corrective measures against uncooperative allies.
Individual-Level Heterogeneity
I complement my investigation with an exploratory, preregistered analysis examining how individual partisanship and ideology moderate support for corrective measures. Voters often rely on cues from copartisan elites on political issues where they hold ambivalent positions. Elite cues prove particularly effective in foreign policy, where citizens have limited access to information and resources for evaluating policy consequences (Zaller Reference Zaller1992). Previous research also demonstrates citizen sensitivity to partisan elite cues on military alliances (Alley Reference Alley2023).
Within the American context, Democratic supporters tend to embrace international organizations and multilateralism more than Republicans (Brutger and Clark Reference Brutger and Clark2023; Voeten Reference Voeten2005). These demand-side factors are reinforced by Trump’s “America First” discourse emphasizing disengagement from traditional alliances in favor of unilateral policies and bilateral deals. The Democratic Party has countered with narratives aimed at restoring American leadership in international alliances under the Biden administration (Boucher and Thies Reference Boucher and Thies2019; Edwards Reference Edwards2018, Kaufman Reference Kaufman2017). Given stark foreign policy polarization, I expect Republicans to be more inclined toward corrective measures against uncooperative allies, with Democrats displaying opposite inclinations.
Correlated with partisanship, ideology serves as a second moderating factor. Left-wing individuals are generally less supportive of force (Kiratli Reference Kiratli2021) and more favorable toward international cooperation (Prather Reference Prather2024; Rathbun Reference Rathbun2011). Given their internationalist outlook, left-wing individuals should be more reluctant to risk global alliances through coercive measures. Conversely, right-wing authoritarianism positively predicts threat perceptions and support for punitive measures (Colémont, Van Hiel, and Cornelis Reference Colémont, Van Hiel and Cornelis2011; Gerber and Jackson Reference Gerber and Jackson2016).
To test these moderating effects, I introduce interaction terms between corrective-measure treatments and a seven-point partisanship scale (one [strong Democrat] to seven [strong Republican]) and a 10-point ideological scale (one [leftmost] to 10 [rightmost]). Figure A3 in the online appendix reports average marginal effects across treatment levels (regression models in appendix table A6).
The results show that Republicans and Democrats differ significantly on two policy proposals: breaking alliances and maintaining the status quo. Each unit increase in Republican identification increases support for breaking the alliance by 0.10 points (p = 0.016) and reduces status quo support by 0.13 points (p = 0.001). To put these effects into perspective, for a self-identified strong Democrat, the expected approval rate is 5.72 points for breaking an alliance and 5.18 points for keeping current policies. Conversely, the predicted approval rate for a strong Republican is 6.27 for breaking the alliance and 4.47 for inaction. In contrast, for the remaining three corrective measures we do not observe meaningful differences between Democrats and Republicans.
The conditional effects of ideology present a clearer picture: each unit increase in right-wing orientation significantly increases support for breaking the alliance by 0.14 points, economic measures by 0.16 points, reducing military support by 0.10 points, and denouncing by 0.09 points, all significant at the 0.01 level. To provide context for these effect sizes, compared to a respondent identified at the far right of the spectrum, a respondent at the leftmost point of the spectrum is 1.64 points less supportive of economic measures (5.22 versus 6.86) and 1.03 points less supportive of reducing military support (5.79 versus 6.82).
Robustness Checks
I perform two robustness checks. First, I replicate the analyses presented in figures 1–3 with those who failed the attention-check questions included in the sample (tables A7–A9 in the online appendix). Including inattentive respondents may correct potential biases if they represent specific population subgroups. The results of this check confirm that despite some minor changes, the overall results persist.
Second, I rerun all analyses while controlling for several demographic and attitudinal factors, including gender, education, trust in government, and political interest. This approach not only addresses any issues that might have arisen from the randomization of experimental conditions across various demographic groups but also helps to specify treatment effects more accurately. In particular, trust in government and political interest could theoretically condition policy approval regarding proposed policies. The results verify that the treatment effects remain virtually unchanged. They also show that of the individual-level covariates, education, trust in the federal government, and political interest are significantly and positively associated with policy approval. Trust in government is also positively correlated with alliance favorability and support for higher contributions to the alliance.
Conclusion
When faced with uncooperative security allies, leaders can either ignore the defection at the cost of perpetuating the offending behavior and risking future noncompliance, or they can attempt to exert pressure on the uncooperative ally through corrective measures to reverse the uncooperative act while signaling deterrence toward potential free riders. The credibility and efficacy of such measures are contingent upon domestic support within the sender state. In this paper, I investigated American voters’ preferences for the use of corrective measures against uncooperative allies in a security alliance and the contextual factors moderating these preferences.
The results of my preregistered experiment revealed that American citizens strongly favor corrective measures against uncooperative allies, particularly coercive actions, including economic sanctions and reduced military support. Importantly, they demonstrate a strong aversion to inaction, preferring decisive responses toward underperforming allies, thereby creating domestic incentives for leaders to intervene to restore alliance cohesion. Previous studies found that Americans support the use of force when an ally is attacked, as inaction would damage the US’s reputation as a reliable partner and raise concerns about fairness and loyalty (Tomz and Weeks Reference Tomz and Jessica2021). In this paper, I found that there is similar disapproval for inaction when addressing allies whose foreign policy trajectories diverge from US interests. Furthermore, the results indicate that information about uncooperative partners generates demands for proportional reductions in American alliance contributions. Collectively, these findings substantiate the significance of alliance cohesion and reciprocal thinking in citizens’ foreign policy preferences.
The empirical analysis also explored the interaction between corrective measures and three contextual factors: the ally’s regime type, its military power, and whether it shares a treaty-based alliance. Among these variables, regime type emerged as the sole significant conditioning factor, with Americans demonstrating marked reluctance to endorse corrective measures against democratic allies. These findings provide a different angle to the microlevel support for democratic peace (e.g., Kiratli Reference Kiratli2024; Tomz and Weeks Reference Tomz and Jessica2013). When conflictual situations arise between democratic dyads, allies are expected to refrain from coercive restraint due to their adverse domestic repercussions, thereby limiting the potential for conflict escalation.
These findings carry substantial policy implications. With the second Trump presidency, there are growing concerns that US contributions to international alliances may decline significantly while pressure on allies will mount to enforce more favorable burden-sharing arrangements. Indeed, the results of this study suggest that leaders may pay considerable domestic costs if they do not act against allies that behave uncooperatively, and consequently, we may expect the US president to exert pressure on “underperforming” allies. For instance, defense contributions among NATO allies represent a particularly salient source of tension in transatlantic relations. The results of this study show that while voter sensitivity to defense underspending remains notably lower compared to other forms of alliance misalignment, the American public backs the strategy of exerting leverage on allies to contribute more.
Arguably, if potential “underperformers” or noncompliant members of an alliance anticipate retaliation from the dominant power due to strong voter expectations for active pressure, the mere threat of the use of corrective measures may discourage defection and strengthen alliance cohesion. At the same time, coercion against allies entails significant strategic costs and inherent risks. They can generate “psychological reactance,” stimulating feelings of defiance and resistance and potentially causing target states to reinforce their original positions (Powers and Altman Reference Powers and Altman2023). Moreover, external pressure mechanisms may catalyze rally-round-the-flag effects within target nations, enhancing leader approval and, consequently, incentivizing the persistence of the wayward behavior (e.g., Pape Reference Pape1997). This is the reason why, despite their frequent use against both foes and friends and their considerable costs, sanctions are seldom effective (e.g., Drezner Reference Drezner2021). For instance, the ongoing Countering America’s Adversaries through Sanctions Act (CAATSA) restrictions that were imposed on Turkey in 2020 following its purchase of the S-400 missile defense system from Russia (Pompeo Reference Pompeo2020) have still not led to a policy reversal. On a grander level, corrective measures run the risk of straining relations, eroding mutual trust, and alienating partners, thereby damaging long-term relationships and compromising future cooperation. Consequently, leaders should carefully calculate the potential costs and benefits when considering corrective measures to influence ally behavior.
Finally, scope conditions and avenues for further research warrant some consideration. First, this study explicitly focuses on voters’ opinions on publicly visible corrective measures. Arguably, publicly visible measures represent the most consequential form of alliance management, as they carry the highest potential costs for alliance relationships while simultaneously generating the strongest signals to other alliance members. However, while public measures may satisfy domestic audiences and enhance credibility, they risk undermining alliance cohesion by broadcasting internal disagreements. Consequently, real-world alliance management is often maintained through a mix of public and private diplomacy. Future studies could examine public attitudes toward private versus public corrective measures, answering whether citizens value the signaling function of public pressure or prefer discrete diplomacy that preserves alliance reputation and minimizes public confrontation. Some studies already show that American citizens have only a weak preference for transparency in the execution of military actions; what they care more about is whether the outcome advances national interests (Myrick Reference Myrick2020). Similarly, research could investigate how the visibility of corrective measures affects their perceived effectiveness and legitimacy. Additionally, scrutinizing how adversaries perceive and potentially exploit public intra-alliance disputes would provide valuable insights into the full strategic implications of corrective measures.
Second, future experimental works can also explore whether perceptions of threat or strategic necessity mediate support for corrective measures. For instance, public attitudes toward corrective measures might differ substantially if the uncooperative ally was perceived as indispensable for countering major adversaries like China or Russia, as strategic imperatives could override concerns about burden sharing or policy misalignment.
Third, this study aimed to explain American voters’ preferences, and future studies should therefore examine the extent to which the external validity of these findings holds across different geographies, such as the major powers of the European Union within the framework of the Common Security and Defence Policy. The US’s hegemonic position within the postwar international security architecture generates distinctive incentives for both policy makers and constituents to favor decisive responses to alliance noncompliance, which may be lacking in other countries. America’s superior capabilities reduce the relative costs of corrective measures, thus enhancing their public appeal. Theoretically, in states with lower capabilities, voters might be much less enthusiastic about employing corrective measures due to their proportionally higher implementation costs.
Finally, future studies could illuminate how the public reacts to deviations from alliance expectations in nonsecurity partnerships, such as trade or environmental agreements. While existing research indicates public support for sanctions and shaming measures against environmental polluters (Tingley and Tomz Reference Tingley and Tomz2014), the persistence of such preferences when directed at allies remains unexplored. A more comprehensive analysis of public attitudes toward uncooperative allies across diverse policy domains would substantially enhance alliance scholarship and theoretical understanding of interstate cooperation dynamics.
Supplementary material
To view supplementary material for this article, please visit http://doi.org/10.1017/S1537592725103071.
Acknowledgments
This research was supported by Waseda University under grant number BARD02095201.
Data Replication
Data replication sets are available in Harvard Dataverse at: https://doi.org/10.7910/DVN/NITW7O.
Open access
