Hand-Tying through Military Signals in Crisis Bargaining

Abstract

Theories of crisis bargaining suggest that costly signals can enhance the credibility of one’s coercive threats. In particular, engaging in conspicuous military mobilizations or demonstrations of force are thought to communicate one’s resolve in a crisis. Yet, there is disagreement about why this might be the case. One set of theories emphasizes the hand-tying political and reputational effects of visible military action. A different collection of theories argues that mobilizations create bargaining leverage by shifting the balance of power in favor of the mobilizing side. This article uses new data on coercive threats in international crises to discriminate between these two explanations. It makes two key contributions. First, it presents systematic evidence that military mobilizations during a crisis bolster the effectiveness of compellent threats. Second, it demonstrates that such signals are likely effective because they alter the local balance of military power, not because of their political effects.

Las teorías en materia de negociación de crisis sugieren que las señales costosas pueden ayudar a mejorar la credibilidad de las amenazas coercitivas. En concreto, existe la creencia de que participar en movilizaciones militares conspicuas o en demostraciones de fuerza comunica la determinación que uno tiene en una crisis. Sin embargo, existe desacuerdo sobre las razones por las que esto sucede. Uno de los conjuntos de teorías hace hincapié en los efectos políticos y reputacionales de la acción militar visible. Otro de los conjuntos de teorías argumenta que las movilizaciones crean una palanca de negociación debido a que cambian el equilibrio de poder a favor del lado movilizador. Este artículo utiliza nuevos datos en materia de amenazas coercitivas que tienen lugar durante las crisis internacionales con el fin de discriminar entre estas dos explicaciones. El artículo realiza dos contribuciones principales. En primer lugar, presenta pruebas sistemáticas de que las movilizaciones militares durante una crisis refuerzan la eficacia de las amenazas coercitivas. En segundo lugar, demuestra que tales señales son probablemente efectivas porque alteran el equilibrio local del poder militar, y no debido a sus efectos políticos.

Les théories relatives à la négociation de crise suggèrent que de coûteux signaux peuvent renforcer la crédibilité des menaces de coercition. Une mobilization visible des armées ou une démonstration de force sont notamment perçues comme une communication de sa détermination en temps de crise. Pourtant, il existe un désaccord quant à la raison. Un ensemble de théorie souligne les effets contraignants sur le plan politique et de la réputation de l’action militaire visible. Un deuxième affirme quant à lui que les mobilizations créent un avantage au sein des négociations en transformant l’équilibre des pouvoirs en faveur de leur auteur. Cet article emploie de nouvelles données sur les menaces coercitives dans les crises internationales pour établir une distinction entre ces deux explications. Il apporte deux contributions d’importance. D’abord, il présente des éléments prouvant systématiquement que les mobilizations militaires par temps de crise renforcent l’efficacité des menaces contraignantes. Ensuite, il démontre que l’efficacité de ces signaux s’explique probablement par leur altération de l’équilibre des pouvoirs militaires au niveau local, et non par leurs effets politiques.

Introduction

How can states improve the success of their coercive threats? In military crises, states generally hope to achieve their political objectives without fighting costly wars. Coercive threats can help accomplish this by convincing the adversary that it is better to back down than to fight. How then can states make these threats more effective? One common answer is that conspicuous military mobilizations or demonstrations of military force can communicate the credibility of a state’s threats during a crisis (e.g., Fearon 1997; Lai 2004; Yarhi-Milo, Kertzer, and Renshon 2018; Slantchev 2005, 2011). Yet while public military maneuvers are widely considered to be effective signals of resolve, there is disagreement about why this may be the case.

One view argues that military maneuvers make it politically costly for leaders to back down (e.g., Fearon 1994a; Gelpi and Griesdorf 2001; Weeks 2008; Kertzer, Renshon, and Yarhi-Milo 2019). These costs, known as “audience costs,” are a form of political punishment imposed on a leader if they fail to follow through on a public threat. The prospect of losing political support—or even political office—is thought to discourage leaders from reneging on commitments after escalating a dispute through public statements or conspicuous shows of force. In one version of this story, states face reputation costs from international audiences. The further the state moves along the escalation chain, the higher these costs and the more credible the threat. In another interconnected logic, the political sanctioning mechanisms in democracies and some nondemocracies are institutionalized and therefore observable to outsiders, thereby allowing these states to more effectively communicate private information about their resolve through military signals. According to this literature on audience costs, which we term “political hand-tying,” public military mobilizations or demonstrations of force tie the hands of leaders, creating incentives to risk war rather than back down and face the criticism of international or domestic audiences. This commitment process demonstrates that the actor is highly resolved, thereby alleviating uncertainty about the state’s willingness to fight.

A different perspective argues that military actions during crises bolster bargaining leverage by revealing information about military capabilities and altering the payoffs for war by shifting the local balance of military power in favor of the mobilizing side (e.g., Slantchev 2011; Tarar 2013). Such a shift alters perceptions of the signaler’s resolve. According to this logic, which we term “military hand-tying,” military signals improve an actor’s probability of prevailing on the battlefield if war breaks out. Military maneuvers that entail the actual movement of forces to the potential theater of operations both create and message new information about the willingness to wage war. In this view, military action increases the mobilizer’s readiness for war, raises the opponent’s predicted costs of fighting, and reduces uncertainty about the probable outcome of war (Slantchev 2011). In these theories, only actors willing to wage war would accept the risks associated with these war preparations. This shift in the balance of power signals new information about the expected payoff for war, which increases the credibility of the state’s threats. This perspective expects that military mobilizations and deployments, which change the local balance of power, will be more effective signaling devices than symbolic displays of military force, which only reveal the local balance of power.

Both types of signals generate hand-tying effects by altering future payoffs and creating incentives to continue the crisis. However, the two processes tie hands in very different ways, one through manipulating political incentives and the other by altering military advantages. These two explanations offer different predictions about what types of states will risk such action and, concomitantly, when military signals will make coercive threats more effective. Political hand-tying through audience costs motivates the leader to continue the crisis, and the mechanism relies on the political process of signaling states. Actions which can generate higher levels of audience costs will more effectively demonstrate resolve. Military hand-tying alters the pay-offs for war, and the mechanism relies on shifting the local balance of power in a way that benefits the mobilizing state at the expense of the dispute opponent. Military signals that more strongly impact the local balance of power will improve threat effectiveness.

In this article, we attempt to discriminate between these two sets of arguments by analyzing the effects of military signals on the outcomes of compellent threats (Sechser 2011). When do target states acquiesce to threats and when do they resist? While political and military effects both likely contribute to the increased effectiveness of coercive threats, this article aims to determine which of these two effects dominates in crisis bargaining. Using data on over 210 compellent threats between 1918 and 2001, quantitative analyses demonstrate that military signaling seems to work mainly through changes in the local balance of power, not through political hand-tying processes. We think it likely that military maneuvers signal resolve through several channels, but we conclude that they are effective primarily when they alter adversaries’ estimates of the probable winner in wartime.

The article proceeds as follows. First, we outline the conceptions of military signals according to the logics of military and political hand-tying as they exist in the current literature. We then derive hypotheses from these two logics concerning the conditions under which military signals are most likely to affect the credibility of threats. The second section describes the article’s research design, which introduces new data on military signals during cases of compellence. In the third section, we conduct a series of empirical tests on how military signals impact the success of compellent threats. We find evidence that military signals are effective signals of resolve but that these signals appear most effective when they alter the local balance of power. The fourth section concludes with implications of these findings for theories of military mobilization and costly signaling more generally.

Military Signals and Coercive Threats

One of the most debated issues in international relations scholarship is how a state can establish the credibility of its commitments—in particular, threats that would be costly to execute (e.g., Fearon 1994a, 1997; Chamberlain 2016; McManus 2017; Lupton 2020). When states have complete information about each other’s resolve and capabilities, they should be able to resolve their disagreements without resorting to force. However, when information is distributed asymmetrically—if either side has private information about their resolve or capabilities—then states may be unable to agree on a mutually beneficial settlement. To overcome the problem of incomplete information, actors can reveal their true resolve (their willingness to wage war over the issue) if the signals they send are costly in such a way that a resolved type is more likely to bear the costs of those signals than an unresolved type. Highly resolved types can distinguish themselves from unresolved types by taking actions to increase the probability of war that a state with a low expected utility for war would generally shy away from. Less resolved states may occasionally risk these actions to convince audiences that their resolve is high in an effort to gain additional concessions.

These theories of crisis bargaining suggest that costly signals can enhance the credibility of one’s coercive threats. A military signal, which can entail a military exercise, troop mobilization, show of force, deployment of military assets, or other military maneuver, is one form of costly signal. According to a number of theories, military signals demonstrate resolve and increase the probability of coercive threat success (Fearon 1997; Lai 2004; Slantchev 2005, 2011). This article evaluates two sets of theories that attempt to explain why military signals are effective. Both agree that militarized signals increase threat credibility, but they disagree on how they influence the crisis participants’ incentives and calculations. One view is that military signals alter the political incentives of leaders, creating disincentives for backing down. This political or reputational hand-tying effect commits the signaling state to stand firm, and it confronts the adversary with the choice of either making concessions or escalating the dispute to war. On the other hand, certain military movements may generate bargaining leverage by shifting the balance of power in favor of the mobilizing side, creating incentives for the mobilizing state to stand firm. This military hand-tying effect also puts the adversary at a disadvantage should the dispute escalate to war. Based on either logic, the high costs of war suggest that when the target is faced with a resolved challenger willing to issue such a military signal, the target should be more likely to make the concessions necessary to resolve the dispute peacefully. By outlining the implications of these two logics, we seek to understand whether military signals influence crisis outcomes primarily through these political effects or the bargaining leverage created by shifts in the local balance of power.

Hand-Tying Signals

According to the logic of political hand-tying, military maneuvers work as costly signals by creating political costs that a leader will suffer if he fails to follow through on his foreign policy promises. In the context of coercive threats, a leader will incur audience costs—from either domestic or international audiences—if he threatens publicly and subsequently reneges on his threat without achieving the stated policy goals. In turn, public military signals are a visible threat of military action that audiences can use to assess crisis performance. These actions tie hands because they further commit the leader to a policy position, and that leader will have to pay costs if he gets caught in a bluff.

Military signals, according to this set of theories, amount to a dramatic public declaration to an opponent about the state’s willingness to wage war. The publicizing of the commitment through a military signal presents the opponent with a highly resolved challenger, so the target should be more likely to concede and avoid escalation to war (Kertzer, Renshon, and Yarhi-Milo 2019). Most theories of political hand-tying argue that audiences view a leader who does not follow through on his threats in international crises as incompetent or having tarnished the nation’s “credibility, face, or honor” (Fearon 1994a, 581). With military signaling, the leader incurs higher levels of audience costs, increasing the likelihood of target concessions.

 

Hypothesis 1 (Political Hand-Tying). Military signals, both shows of force and military mobilizations, increase the effectiveness of coercive threats.

Theories in the vein of “military hand-tying” argue that military mobilization creates bargaining leverage by shifting the balance of power in favor of the mobilizing size. Military mobilization contributes positively to the probability of victory in a dispute, and thus increases the expected utility for war (Slantchev 2005). By increasing a state’s expected value for fighting, such moves also increase its willingness to fight. Conversely, it decreases the opponent’s probability of victory in war, thereby decreasing that state’s value for fighting. Certain military moves not only reveal capabilities but also shift the “immediate balance,” acting as an ex post indicator of resolve, which “reflects new information about the … [actor’s] willingness to resist and how effectively it would be able to resist” (Fearon 1994b, 925). These theories focus primarily on military moves that effect a discernible change in the local balance of power. Moves that entail a shift in the balance of power are risky because they increase incentives for war, sending a strong signal that separates the resolved from the unresolved challengers.

This logic implies an important distinction among military signals. Actual mobilizations or deployments of military assets are likely to have a stronger effect, according to this view, compared to public displays of preexisting military forces. Those military moves that strongly shift the balance of power will be more effective at inducing compliance. For example, when bombers fly over an adversary’s capital city as a show of force, this does not shift the balance of power in the same way as, say, deploying naval vessels or moving ground forces to the potential theater of operations (Post 2019). Deploying troops and military assets to the arena of operations generates a local military advantage that did not exist previously.

 

Hypothesis 2

(Military Hand-Tying). On average, only military signals that entail a movement of military assets will increase the effectiveness of coercive threats.

Both political and military hand-tying processes tie hands by altering future payoffs and creating incentives to continue the crisis. The public nature of these signals communicates new information about the challenger’s resolve to the target state. However, military hand-tying and certain versions of the political hand-tying stories diverge on their expectations regarding the role of regime type. There is evidence that certain regime types have an advantage in accruing domestic audience costs (Fearon 1994a; Schultz 2001; Weeks 2008, 2012). According to this domestic political hand-tying logic, as regimes capable of generating audience costs escalate from verbal threats to military signals, they should more easily and quickly signal resolve. Theories of “democratic credibility” (Downes and Sechser 2012) predict that a democracy can credibly communicate its preferences with fewer escalatory steps than an equivalent nondemocracy (Fearon 1994a, 585). In line with these theoretical predictions, Gelpi and Griesdorf (2001) show that democracies tend to win international crises more frequently when they demonstrate resolve through military action. Weeks (2008) revises this thesis to demonstrate that the true distinction is between personalist and nonpersonalist regimes, with nonpersonalist regimes better able to generate audience costs than personalist regimes.

 

Hypothesis 3

(Domestic Political Hand-Tying). Military signals from domestic audience costs capable regimes will increase threat success, on average, more than military signals from other regimes.

To assess whether military signals increase threat effectiveness through their influence on the local balance of power or through political hand-tying, we distinguish between military signals that involve the actual movement of military assets into a conflict theater and demonstrations of force that have no meaningful effect on the local balance of power (table 1). In this set-up, the differing factor is the expectations regarding shows of force: military hand-tying theories do not expect them to increase threat effectiveness because they do not shift the balance of power, while political hand-tying theories expect them to increase threat effectiveness because they engage political audiences.

Table 1.

Predicted effect of military signals on coercive threat success

	Military hand-tying	Political hand-tying
No military signal	Baseline	Baseline
Military signal	Increase (+)	Increase (+)
Show of force	None (0)	Increase (+)
Mobilization	Increase (+)	Increase (+)
Military signal × accountable regimes	None	Increase (+)

We also test whether any version of the domestic political hand-tying story rings true by examining the role of regime type in the military signaling process. In particular, the logic of domestic hand-tying would expect audience costs capable regimes to be particularly adept at utilizing military signals. For theories of military hand-tying, however, it does not necessarily follow that audience costs capable regimes would be better able to signal their resolve in situations where the same military means are available to another regime (Slantchev 2005, 546). According to this logic, no particular regime type should outperform the other in the utilization of military signals.

It is worth noting that military mobilizations are inherently costly and therefore have sunk-costs properties. In both the military and political hand-tying stories, military mobilization influence states’ strategic calculations and incentives to follow through on a threat (Fearon 1994a; Slantchev 2005). Sunk costs signals expend resources, thereby helping the target state update their priors about the challenger’s value of the issue at hand. Thus, sunk costs may undergird the effectiveness of military signals. Ultimately, however, the difference here is between burning resources and investing resources (Altman and Quek 2021). Although sunk costs might be at play, military mobilizations do more than just expend resources: they engage political audiences and invest in a future war.

Methods and Data

To evaluate these hypotheses, we turn to a dataset of coercive threats. The Militarized Compellent Threat (mct) dataset contains information about 210 “compellent” threats issued between 1918 and 2001. The dataset defines a compellent threat as “an explicit demand by one state (the challenger) that another state (the target) alter the status quo in some material way, backed by a threat of military force if the target does not comply” (Sechser 2011, 380). In other words, the dataset contains episodes with two components: a verbal coercive demand issued from one state to another (which can be communicated in writing or orally), coupled with the threat of military force (which can be communicated verbally or through a conspicuous demonstration of military force). Each observation in the dataset contains a single challenger and target, so that episodes involving multiple participants on one side are broken up into dyads.

The mct dataset offers important advantages, as well as limitations, for testing the hand-tying effects of military action in international disputes. One key advantage is that it allows for the straightforward evaluation of the outcomes of dispute episodes. Each episode is triggered by a coercive demand, which must be issued verbally (i.e., written or oral) in order to be included in the dataset. The advantage of this feature is that it mitigates ambiguity about what was demanded, so that scholars can easily determine whether the demand was successful or not. At the same time, it is important to acknowledge that many—perhaps even most—coercive demands in international crises are not made explicitly. This feature enhances the internal reliability of the analysis, though at some cost to generalizability.

Second, there are advantages to evaluating the mechanisms of military signaling in compellence episodes, rather than deterrence episodes. As Schelling notes, compellence is distinct from deterrence in that the objective of the threatening state is to change the status quo, rather than maintain it. In other words, for the target to comply, it must take some conspicuous action—for example, relinquish a possession or institute a policy change.¹ By contrast, in deterrence, compliance involves simply refraining from action. Deterrence episodes create a difficult inference problem: a target that appears to “comply” with a deterrent demand may have never planned to act at all. We, therefore, cannot necessarily attribute compliance to the threat. As such, using deterrence episodes to evaluate the effects of military signaling could lead to mistaken inferences. In compellence, by contrast, this inference problem is less severe, since it would occur only if a target was threatened when it already planned to act. While such instances may exist, it seems likely that they are rarer than their deterrence counterpart. At the same time, compellent threats represent only a narrow slice of the coercive diplomacy spectrum, so we cannot necessarily extrapolate the findings below to deterrent threats. We leave the analysis of military signaling in deterrence episodes for future research.

Dependent Variable

Military signaling during crises is designed to influence crisis outcomes: states employing such signals are, in principle, more likely to achieve their objectives without a fight. The dependent variable in the analyses below is, therefore, the success or failure of a compellent threat. A successful threat meets two conditions according to the coding rules of the mct dataset: first, the target complies voluntarily with the challenger’s demands, and second, the challenger does not engage in military action that inflicts more than 100 target fatalities.²

Independent Variables: Military Signals and Regime Type

Our primary measurement of military signaling comes from the mct dataset. The variable military signal indicates whether the challenger employed a military exercise, troop mobilization, show of force, deployment of military assets, or other military maneuver to bolster its threat. To be coded as 1, each military signal had to be public and observable to both domestic and international audiences. This variable offers a rough picture of the frequency with which challengers employ military signals during compellence. In the mct dataset, challengers employ some form of military signal roughly 75 percent of the time, electing to rely on verbal threats of force the remaining 25 percent of the time.

Given that both sets of theories argue that military signals increase the credibility of threats, however, we cannot simply examine how states respond to general demonstrations of force: we must distinguish between military versus political hand-tying signals. As we discussed earlier, military signals can come in at least two varieties: measures that involve the deployment of military forces to the potential conflict zone, and public maneuvers that are designed to showcase the challenger’s forces but not alter the local balance of power. The mct dataset, however, does not distinguish between different types of military signals. For this more fine-grained data, we coded two new variables. The first variable, mobilization, indicates whether a challenger moved military assets into the likely zone of conflict during a dispute. We considered mobilization to be one of three things: troops moved to the border of the target state, troops moved to the area of the disputed issue (i.e., territory), or troops mobilized domestically during a dispute involving a neighboring state. The second variable, show of force, indicates military signals that do not involve additions to the challenger’s local military forces. These signals most often include displays or shows of force, such as missile tests, bomber overflights, putting troops on alert, or exercises involving troops or military assets already present in the conflict zone.

The military signals in the mct dataset vary in size and scope, although the mobilizations generally reflected a substantial shift in the balance of power. For example, in June 1919, the United States deployed the USS Castine to the port of Limon and landed marines, successfully inducing Cuba to reverse their recent military coup (an example of a successful, albeit limited, mobilization).³ In a similar case, President Bill Clinton deployed large numbers of ground troops and two aircraft carriers to Haiti when he issued an ultimatum to Haitian dictator Raoul Cedras.⁴ After this significant mobilization, the disputing parties reached an agreement shortly before the invasion was scheduled to begin. As a more standard illustration, Adolf Hitler often mobilized troops to buttress his coercive threats. For example, in 1938, Germany concurrently mobilized over 36 land divisions along the Czechoslovakian border to induce Czechoslovakia to cede Sudetenland.⁵ Finally, in a classic show of force (rather than a mobilization), Argentina underscored its demand that Chile recognize Argentine sovereignty in the Atlantic by increasing the activities of its air and naval forces (without increasing them in size or substance) off the coast of Chile in 1977.⁶ This threat failed.

To better contrast the two variables, consider two cases of compellence during the Soviet-China border clashes. In 1969,⁷ the New York Times reported on March 9, “The Soviets put their military in the Far East on alert as an additional warning to the Chinese.” There was no evidence of mobilization in the primary or secondary sources, so this was coded as a show of force. In the following episode,⁸ the New York Times reported on August 14 that the Soviet Union prepared “two helicopters, dozens of tanks and armored vehicles and several hundred troops” to be sent into Chinese territory. This was coded accordingly as a mobilization.

These variables are coded as mutually exclusive. We were concerned about missing certain acts in the historical records, so we coded only what might be considered the most prominent signal in each case to reduce coding irregularities. This process resulted in fifty-eight verbal threats (24 percent), fifteen shows of force (6 percent), and 169 (70 percent) mobilizations.⁹ Because of the coding rules, this data may overestimate the effectiveness of verbal threats (since some cases escalated with military signaling after a verbal threat failed) and shows of force (since a few failed and escalated with a more significant mobilization), biasing the results away from the logic of military hand-tying.

The logic of domestic political hand-tying emphasizes that some governmental regimes are better able to tie their hands through public actions than others. We employ a standard indicator of democracy, using the 21-point Polity scale to denote states whose overall regime score is 16 or above. We also create a second measure, which indicates a challenger’s relative ability to generate audience costs, compared to its target, to test the hypothesis that democracies enjoy a signaling advantage only if they are able to generate more audience costs than their opponent (Gelpi and Griesdorf 2001). The variable challenger democratic advantage indicates whether the challenger has a higher Polity score than its target. We also test whether distinctions among autocratic regimes have an effect. The variable nonpersonalist challenger indicates whether the challenger is a nonpersonalist (this include democracies) or personalist (a classic “dictator”) regime based on Weeks’ (2012) dichotomous coding.¹⁰

We include several additional variables to account for other factors that might influence the success and failure of coercive threats. First, balance of power measures the challenger’s share of material capabilities in a dyad, as defined by the Correlates of War Project’s Composite Indicator of National Capabilities (CINC) score (Singer 1987). The CINC variable contains information about individual states’ annual share of global capabilities across six dimensions: military personnel, military expenditures, energy consumption, iron and steel production, urban population, and total population. The CINC score is calculated by computing a state’s proportion of capabilities in each of the six categories, and then averaging the six proportions.

Second, we code for the type of issue at stake in each dispute. Most research on coercive diplomacy points to the importance of the stakes in a dispute to explain why some threats succeed and others do not (e.g., George and Simons 1994; Art and Cronin 2003). Identifying the issue, therefore, is critical to an empirical analysis of the effectiveness of coercive threats. The mct dataset classifies four types of demands: territorial demands (including ownership of military bases and troop withdrawals), demands for reparation, demands for leadership or regime change, and demands for policy changes. Of the 210 cases in the mct dataset, all but eleven fall into one of these four categories; these are placed in a fifth category, simply labeled “other.” The analysis in the main text includes dummy variables for territory, reparations, leadership, and policy, with “other” being the reference category.

Third, we control for the hawkishness of the challenging state’s leader in the dispute. Horowitz et al. (2018) find that leader attributes affect coercive threat success, and research by Kertzer, Renshon, and Yarhi-Milo (2019) and Yarhi-Milo, Kertzer, and Renshon (2018) suggests that the coercive success of military signals may have something to do with the attributes of leaders likely to use them. We use a measure of leader hawkishness drawn from Carter and Smith (2020) in our analysis. Their indicators use data on leader-level background experiences to construct four different measures of leader hawkishness. We utilize what Carter and Smith (2020) find to be the best performing measure, labeled M2 in their core analysis. This measure includes factors that are related to risk, including military background, education, sex, and entrance to power (Ellis, Horowitz, and Stam 2015) and also includes psychological characteristic information and political orientation.

Finally, the analysis includes a control variable for geographic proximity. The variable contiguity codes disputants as contiguous if they either share a land border or are separated by less than 150 miles of water (Schultz 2001, 265). This information is derived from the Correlates of War Direct Contiguity dataset.

Results

How do military signals affect the success of compellent threats? Below we present two sets of logistic regressions to evaluate the hypotheses described in the first section of this article. The first set of tests distinguishes between signals that have primarily symbolic effects versus signals that carry more meaningful consequences for the local balance of power. If the military hand-tying logic is correct, then only military mobilizations, which effect a shift in the balance of power, will increase threat effectiveness. If the political hand-tying logic is correct, shows of force should also increase the success of compellent threats. The second set of tests evaluates the interactive effects of regime type and demonstrations of military force during cases of compellence. If the domestic hand-tying logic of military signaling is accurate, then audience costs capable regimes should accrue larger advantages from engaging in military actions during crises than other actors. In each set of tests, the result is the same: military signals appear to be effective primarily because of their military consequences, rather than their political effects, although we do not entirely rule out the role of political effects.

The first set of tests, contained in table 2, considers the relationship between military signals and compellent threat success. Note first that military signal in Models 1 and 2 is positive and statistically significant at the 99 percent level or above. For example, in Model 1, a state employing military signals is three times more likely to succeed while issuing a compellent demand, with a success rate of 16 percent with no military signal and a success rate of 52 percent with one.¹¹ This suggests that military signals indeed have a strong positive effect on the success of compellent threats: states which employ them to bolster their threats are much more likely to succeed.

Table 2.

Logit estimates of compellent threat success

	1 Any military signal	2 Any military signal	3 Mobilization vs. show of force	4 Mobilization vs. show of force
Military signal	1.782⋆⋆⋆	2.117⋆⋆⋆
	(0.371)	(0.433)
Show of force			0.683	0.606
			(0.632)	(0.588)
Mobilization			1.873⋆⋆⋆	2.261⋆⋆⋆
			(0.380)	(0.440)
Democratic challenger		−0.199		−0.204
		(0.408)		(0.418)
Democratic target		1.135⋆⋆		1.204⋆⋆
		(0.407)		(0.418)
Leader hawkishness		−0.410		−0.482†
		(0.251)		(0.266)
Balance of power		−0.674		−0.622
		(0.625)		(0.653)
Territory		−0.614†		−0.591†
		(0.332)		(0.347)
Reparations		0.008		−0.088
		(0.484)		(0.488)
Leadership		2.178⋆⋆⋆		2.247⋆⋆⋆
		(0.603)		(0.629)
Policy		−0.012		0.042
		(0.320)		(0.322)
Contiguity		−0.003		0.023
		(0.364)		(0.375)
Constant	−1.695⋆⋆⋆	−1.455⋆	−1.695⋆⋆⋆	−1.563⋆
	(0.351)	(0.733)	(0.351)	(0.736)
N	242	242	242	242
Pseudo R2	0.080	0.193	0.093	0.211

Note: robust standard errors in parentheses.

† p < .10, ^⋆  p < .05, ^⋆⋆  p < .01, and ^⋆⋆⋆  p < .001.

While these results support the theoretical consensus that military signals increase threat success, they do not test the competing hand-tying logics. The reason is that the analysis thus far has conflated signals with primarily military effects with those that ought to have mainly political effects. The analysis tells us little about military signals that alter the balance of forces versus those that have a more symbolic effect. Models 3 and 4 reported in table 2 attempt to correct this limitation by distinguishing between these two kinds of signals. Signals that involved the actual movement of military assets into a conflict theater are coded 1 by the variable mobilization, while signals that involved only shows of force by preexisting military units are coded 1 by the variable show of force.

A first glimpse comes from comparing the marginal effects of the variables mobilization and show of force in Models 3 and 4. In both models, both variables are positive, but the coefficient for mobilization is larger and statistically significant. Indeed, the marginal effect for mobilization on the likelihood of success is roughly six times as large as the effect of show of force: changing mobilization from 0 to 1 increases the absolute likelihood of success by 37 percent, whereas changing show of force from 0 to 1 does so by 6 percent. This offers some initial evidence that the military effects of military signals dominate their political effects. Figure 1 illustrates these findings, and the difference between show of force and mobilization is statistically significant. The small number of observations for the show of force variable somewhat limits our inferences, but keep in mind that our mobilization variable necessarily included some failed shows of force. Ultimately, we cannot conclude that military signals have no political effects, only that the military effect seems to dominate during cases of compellence.

Figure 1.

Marginal effect of military signals on compellent threat success (95 percent confidence intervals shown).

While these results have somewhat undermined Hypothesis 1 in support of Hypothesis 2, they have said little (except by default) about Hypothesis 3. Our next set of tests explores the interactive effects of regime type and military signals to assess the role of domestic political hand-tying in coercive bargaining.¹² Model 5, therefore, includes the interaction term military signal × democratic challenger, which carries a value of 1 only if the challenger in a dispute was both a democracy and employed a military signal. Regime type has no clear association with threat success: both democratic challenger and military signal × democratic challenger are not statistically significant in Model 5. In other words, the effect of military signals appears equally strong for both democracies and nondemocracies. This is evidence against Hypothesis 3, which expected the effect of military signals to be especially strong for democracies due to their audience costs generating abilities.

One might plausibly argue, however, that it is inappropriate to examine the monadic effects of regime type. Rather, it might be the case that the political advantages of military signals accrue only to democracies that can generate audience costs to a greater extent than their opponents (e.g., Gelpi and Griesdorf 2001). Model 6 evaluates this proposition, including the variable challenger democratic advantage, which is coded 1 whenever the challenger has a higher Polity score than the target. However, in Model 6, neither challenger democratic advantage nor its interaction with military signal reaches statistical significance, casting further doubt on the logic of political hand-tying.

Another potential objection to this previous analysis is that democracy versus nondemocracy is an inappropriate way to test for hand-tying via audience costs. This brings us to Weeks' (2008) argument that it is not all nondemocracies that have the disadvantage. Rather, it is personalist regimes—what we typically think of when we consider autocracies or dictatorships—that suffer from an audience costs disadvantage. In this case, nonpersonalist regimes should be able to more effectively tie their hands than personalist regimes. Model 7 evaluates this proposition, including the variable nonpersonalist challenger, which is coded 1 whenever the challenger is a nonpersonalist regime and 0 when it is a personalist regime. When we interact the variable with military signal, the regime variable nearly reaches statistical significance when not accompanied by a military demonstration. This is the opposite of what we would expect from the logic of political hand-tying. Audience costs-capable regimes should accrue advantages when they use military signals.

However, since mobilizations and military asset movements may have both military and political effects, this is insufficient evidence to declare a verdict. A deeper look is offered by interacting the show of force and mobilization variables with the challenger’s regime type. Another way of looking at it is that if the logic of political hand-tying is correct, then shows of force should be uniquely effective when conducted by audience costs capable regimes. Conversely, if military signals are effective mainly due to their effects on the local balance of power, then mobilizations should have a greater effect—and that effect should be consistent across regime types. In other words, the military effects of signaling might dominate when the military signal shifts the balance of power, but the political effects should emerge when we isolate those effects. This second set of tests considers the interactive effects of regime type and the breakdown of military signals.

In Model 8, both military signal variables are interacted with democratic challenger to test the hypothesis that democracies have a unique advantage when issuing these kinds of signals. However, neither democratic challenger nor the interaction terms achieve statistical significance in this model. In addition, the variable for shows of force still does not achieve 95 percent statistical significance in this model, while the variable for military mobilizations does. This suggests that the movement of military assets is a more effective way to signal one’s intentions than using military demonstrations. This is further evidence against the political hand-tying hypothesis, and in favor of the logic of military hand-tying.

Model 9 repeats these analyses, but with the inclusion of the democratic advantage term. The results, however, remain the same: military mobilizations are associated with a state’s likelihood of prevailing, while shows of force are not, even when interacted with regime type variables. The same goes for Model 10, in which we include the non-personalist challenger term, which almost reaches statistical significance (though the effect is substantively small). The effects of military maneuvers appear unaffected by the challenger’s regime type, regardless of how we measure the concept. Figure 2 illustrates these findings, reporting the predicted probability that a compellent threat will succeed under a variety of conditions. Overall, this collection of findings casts doubt on the argument that military signals are uniquely effective due to their domestic hand-tying effects.

Figure 2.

Marginal effect of regime type on compellent threat success (95 percent confidence intervals shown).

A skeptic might also suggest that since military signals are endogenous to state resolve, then it is not the military signal driving threat effectiveness but some other unobserved variable driving both military signaling and the target’s response. We think this unlikely, as the key idea behind a costly signal is to reveal otherwise unobservable information to alleviate informational asymmetry. Experimental studies, which do not suffer from endogenous selection effects, have shown that military mobilizations independently affect observers’ estimates of resolve (e.g., Yarhi-Milo, Kertzer, and Renshon 2018). Thus, the reported associations between military signals and target compliance suggest that these costly signals reveal information about the overall balance of power and resolve that would otherwise stay hidden.

To sum up, the regressions in tables 2 and 3 suggest three main findings. First, military signals do indeed help states succeed in their threats. Second, moving new military assets into a conflict theater has a considerably greater effect on compellence outcomes than public displays of preexisting military forces (although this finding is tentative, given how few shows of force exist in the data). Third, the effects of military signals do not significantly differ by regime type. Overall, the weight of the evidence points in the direction of military hand-tying, and away from the logic of political hand-tying.

Table 3.

Logit estimates of compellent threat success by regime type

	5 Democratic challenger	6 Democratic advantage	7 Nonpersonalist challenger	8 Democratic challenger	9 Democratic advantage	10 Nonpersonalist challenger
Military signal	1.772⋆⋆⋆	2.017⋆⋆	2.718⋆⋆
	(0.507)	(0.656)	(0.958)
Show of force				0.249	1.257	1.910
				(0.763)	(0.835)	(1.221)
Mobilization				1.918⋆⋆⋆	2.178⋆⋆	2.796⋆⋆
				(0.507)	(0.673)	(0.965)
Democratic challenger	−0.974	−0.048		−0.966	−0.066
	(0.821)	(0.499)		(0.820)	(0.520)
Democratic advantage		−0.413			−0.332
		(0.856)			(0.881)
Nonpersonalist challenger			1.386			1.397
			(1.064)			(1.076)
Military signal x	0.939
Democratic challenger	(0.858)
Military signal x		0.127
Democratic advantage		(0.857)
Military signal x			−0.693
Nonpersonalist challenger			(1.075)
Show of force x				0.950
Democratic challenger				(1.270)
Show of force x					−1.837
Democratic advantage					(1.120)
Show of force x						−1.612
Nonpersonalist challenger						(1.394)
Mobilization x				0.923
Democratic challenger				(0.864)
Mobilization x					0.158
Democratic advantage					(0.876)
Mobilization x						−0.602
Nonpersonalist challenger						(1.083)
Constant	−1.142	−1.140	−3.129**	−1.252	−1.424	−3.299**
	(0.813)	(0.930)	(1.160)	(0.816)	(0.958)	(1.170)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
N	242	242	242	242	242	242
Pseudo R2	0.196	0.195	0.203	0.214	0.217	0.223

Note: Robust standard errors in parentheses.

† p < .10, ^⋆  p < .05, ^⋆⋆  p < .01, and ^⋆⋆⋆  p < .001.

Conclusions and Implications

This article offers new, albeit preliminary, evidence that military signals matter in international crises. While theoretical scholarship has long emphasized the importance of signaling, data limitations have often forced empirical studies to rely on indirect measures of signals to demonstrate the impact of intracrisis signaling. This study uses data created explicitly for testing theories about military signaling, ultimately supporting the intuition of scholars that costly signals of resolve can have a positive effect on one’s crisis fortunes. Military signals appear to be effective during cases of compellence largely because of their effects on the balance of military power, rather than their political effects—although data limitations prevent us from concluding that political dynamics play no role. In other words, the military effects of signals seem to dominate the political effects of signals. Military signals that involved the movement of military forces to the conflict zone were associated with more successful compellent threats than signals that acted as purely symbolic displays of military force. In addition, democracies and other nonpersonalist regimes appeared to enjoy no special advantage when issuing military signals of any type.

One thing to note is that this dataset only includes observations of militarized, compellent threats, in which the threat of military force is explicitly signaled either via a conspicuous military movement or the verbal threat of military force. This feature of the data may help us understand apparently disparate findings within the literature. For example, Fuhrmann and Sechser (2014) find that defense pacts are effective political hand-tying signaling devices during cases of extended deterrence, while our data reveals a diminished role for the political effects of military movements. In cases of general deterrence, however, political signals may be more effective because they set a “trip-wire” for the opponent. Once the wire has been tripped and extended deterrence has failed, states must signal their resolve during the immediate crisis. Our findings imply that these threats in particular face unique credibility problems that may require a more extensive military commitment. Indeed, the findings here are consistent with empirical studies of deterrence from the 1980s (Huth and Russett 1984, 1988; Huth 1988). These studies found that the immediate balance of power (forces already prepared) and short-term balance of power (forces that can be quickly mustered) are significantly associated with immediate deterrence outcomes. We might expect military mobilizations, which shape the local balance of power, to further contribute to this effect by logical extension.

This article suggests avenues for future research that examine the conditions under which military signals fail and succeed. For example, Russia’s 2022 invasion of Ukraine is a notable case of failed compellence. Our analysis would suggest that Russia sufficiently mobilized its military to coerce the West—why then did it fail? A limitation of our analysis points to a possible explanation: our empirical specifications mostly bracket actions taken by the target state(s). However, Sechser (2010) argues that the target state may have strategic incentives such as reputation not to capitulate in the face of overwhelming odds, and Powers and Altman (2023) point to the role of psychology in understanding why targets of threats choose to resist a clearly resolved challenger. The disposition of leaders also shapes perceptions of threat credibility, particularly when public threats are accompanied with military action (Yarhi-Milo, Kertzer, and Renshon 2018). This discussion also suggests further investigation into the aftereffects of failed threats. How do military moves during crisis bargaining affect the challenger’s ability to accomplish wartime objectives? Theories of military hand-tying imply that military movements altering the local balance of power positively influence the challenger’s likelihood of winning the war. However, in those cases where military movements fail to improve threat effectiveness and result in war, it is possible that the target is especially resolved, hence negating the benefits of military preparation. Taking a closer look at the targets of compellent threats may provide more complete insights into the conditions under which threats succeed during international bargaining.

Acknowledgement

We thank Christopher Gelpi, Matthew Kroenig, Roseanne McManus, Jessica Weeks, several anonymous reviewers, and participants at the 2015 American Political Science Association and Peace Science Society annual meetings for their insightful comments on this project. We also thank Jack Brake, Nicole Fratkin, Benjamin Harris, and Cara Mumford for their excellent research assistance. The authors acknowledge support from the US Air Force Office of Scientific Research, award number FA9550-14-1–0072. The sponsor was uninvolved in study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the article for publication. The data underlying this article are available on the ISQ Dataverse, at https://dataverse.harvard.edu/dataverse/isq.

Footnotes

1

See also Greenhill and Krause (2017); Powers and Altman (2023).

2

We re-estimate all models using an alternative trichotomous specification of the dependent variable and an alternative binary version of the dependent variable that allows target fatalities in Appendix B. All results are consistent with the models included in the main text of the article. Figures include binary logit model results using black bars and ordered logit models using the trichotomous version using gray bars.

3

mct-008.

4

mct-196.

5

mct-068.

6

mct-158.

7

mct-141.

8

mct-142.

9

As Schelling (1966) notes, compellence is more difficult than deterrence, thereby demanding more costly efforts to make compellent threat credible—this may explain why there were so many mobilizations in this dataset.

10

Weeks (2012) codes forward from 1945, so we use the coding rules listed in her online appendix to extend the data back to 1918.

11

For this and all marginal effects reported here, other variables are held at their mean or median values. When describing the effect of the mobilization variable, we set the show of force variable at 0; and when estimating marginal effects for the show of force variable, we set the mobilization variable at 0 (since the two are mutually exclusive). Marginal effects are statistically significant at the 95 percent level or above unless otherwise noted.

12

Since Hypothesis 3 focuses on the interactive effect of regime type and military signaling, we only report the interactive models here. Please see Appendix B for models that report the raw effects of regime type.

Author Biography

Abigail S. Post is a Matthew B. Ridgway Center for International Security Studies Associate at the Graduate School of Public and International Affairs at the University of Pittsburgh. Her research interests include international bargaining, public opinion and military technology, gender and conflict, and political violence.

Todd S. Sechser is the Pamela Feinour Edmonds and Franklin S. Edmonds, Jr. Discovery Professor of Politics; Professor of Public Policy at the Batten School of Leadership and Public Policy; and a Senior Fellow at the Miller Center of Public Affairs at the University of Virginia. His research interests include coercive diplomacy, emerging technologies, nuclear security, and political violence.