Troops were deployed on an unprecedented scale during the Cold War. Much of the network of deployments established during that time has persisted long after the end of the Cold War. We look to contribute to a growing literature addressing the costs and benefits associated with hosting foreign troops. We ask: Does the presence of foreign troops affect stability in the host country? To answer this question, we develop an argument in which deployed troops are seen as a costly signal of the deploying state’s interest in and commitment to stability in the host country. The presence of foreign troops positively affects the perceived stability of the host country and the robustness of their legal and political institutions. This bolsters the enforcement of agreements between the host government and latent and manifest opposition groups to pursue alternatives to fighting. We test this logic using instrumental variables regressions in which we endogenize deployment motivations. All tests support the expectation that the presence of foreign troops reduces the likelihood of the occurrence of civil conflict in the host state.

Introduction

Does the presence of foreign troops contribute to political stability in host countries? Specifically, do troop deployments reduce the likelihood that host states experience unrest? Between 1946 and 2006, no fewer than 240 large-scale opposition groups mobilized against incumbent governments (Chenoweth and Lewis 2013). These periods of unrest have important consequences; they frequently lead to significant loss of life, generate regime instability, and have broader regional implications. In light of these consequences, a growing literature explores the origins of unrest. For example, structural perspectives emphasize the social, political, and economic characteristics that precipitate violent and nonviolent forms of unrest (Fearon and Laitin 2003; Collier and Hoeffler 2002, 2007; Chenoweth and Ulfelder forthcoming). However, while these accounts help specify the conditions leading to instability, they generally overlook important strategic considerations, including interactions between governments and latent opposition groups.

Bargaining perspectives offer a glimpse into these strategic interactions. Bargaining models show that unrest is more likely under conditions of uncertainty (Blainey 1973; Fearon 1995; Hirshleifer 2001; Walter 2009). Uncertainty comes in a variety of forms. It may result from parties holding private information regarding their capabilities, including their resolve (Fearon 2007; Walter 2009), which leads them to overestimate their prospects for victory (Collier, Hoeffler and Soderbom 2004; Blattman and Miguel 2010). Uncertainty can also result from doubts over whether domestic institutions will enforce contracts between potential belligerents. This uncertainty generates commitment problems in which parties have incentives to renege upon their agreements (Walter 1997). In the presence of commitment problems, states are more prone to instability because uncertainty clouds participants’ cost-benefit calculations of whether to initiate unrest.

Structural and bargaining perspectives both emphasize domestic factors. As a result, the literature sometimes overlooks the role played by foreign states. Existing work provides relatively little insight into the ex ante influence that foreign states have on the processes of civil conflict. This is surprising given the evidence that foreign states play an important role ex post—that is, promoting post-conflict stability. For example, work on third-party interventions shows that foreign states affect conflict duration and the durability of post-conflict peace (Regan 2002; Fortna 2004). Existing findings address interventions during conflict. We build upon these prior studies to explore foreign countries’ ex ante effects on unrest. We also draw upon literatures on economic interventions in peaceful contexts (see, e.g., Alesina and Dollar 2000; Collier and Hoeffler 2002; Fearon, Humphreys, and Weinstein 2009), as well as research on the broad range of costs and benefits of hosting foreign troops (e.g., Allen, Van-Dusky-Allen, and Flynn forthcoming; Bell, Clay, and Machain 2016). Building upon these literatures, we argue that foreign countries shape the uncertainty that leads to unrest.

We focus specifically on international troop deployments. Deployments are widespread; more than 30 nations combined to deploy roughly 400,000 troops globally in the past decade. A majority of troop deployments are not interventions into ongoing conflicts. Instead, troops are most commonly sent abroad as part of peacetime missions—62 percent of new troop deployments in our sample are to countries at peace. Thus, this paper is interested primarily in how the presence of troops affects occurrence, rather than whether wartime deployments (i.e., interventions) facilitate peace. The sheer volume of troops stationed abroad, as well as their sometimes politically contentious presence,2 highlights the importance of our question: Do troops affect host country stability?

Our argument starts from the assumption that parties to conflict are more willing to mobilize when they perceive a higher likelihood of securing their goals. As bargaining perspectives highlight, however, we assume that adversaries’ perceptions are muddied by uncertainty. One form of uncertainty is particularly important in the context of unrest: uncertainty over the credibility of actors’ commitments (commitment problems). Commitment problems make it difficult for actors to calculate the utility of fighting, increasing the risk of conflict. Resolving these difficulties is precisely how foreign troops can help reduce the likelihood of conflict.

Existing work shows that foreign countries (and their troops) contribute positively to the strength of governments’ commitments. For Walter (1997), this relates to commitments made to terminate fighting. It is also the case, however, that foreign states can bolster local commitments between governments and opposition groups to seek alternatives to fighting in the first place. Walter (2009) details one such mechanism. She suggests that opposition actors might look to challenge the government when there exists uncertainty regarding the government’s willingness to fight to preserve the status quo. Foreign countries can, we argue, play a pivotal role in this process. To begin with, deployers have an interest in preserving the stability of host countries in order to protect their global economic interests (Biglaiser and DeRouen 2007, 2009). Furthermore, foreign troops can benefit host countries by facilitating economic growth (Jones and Kane 2007, 2012), respect for human rights (Bell, Clay, and Machain 2016), and lengthening government tenures (Brinkman 2001; O’Kane 1993). In sum, therefore, deployments are costly endeavors for deploying states. The presence of deployed troops signals a foreign country’s interest in the host’s stability (Gartzke and Kagatoni 2015). Their mere presence ought to deter opposition group mobilization. We contend that the size of this costly signal increases in the number of troops deployed.

We argue that there is one especially important mechanism through which foreign troops can help reduce uncertainty between the government and their latent and manifest opponents: by buttressing the institutions and political processes that lead to greater predictability in contract enforcement—that is, bolstering the rule of law. In the presence of weak institutions, opposition groups may fear that the terms of bargained alternatives to fighting will not be upheld. Troops are often deployed with the express purpose of helping reinforce weak domestic institutions. In doing so, foreign troops help resolve a significant source of uncertainty among rivals. Taken together, troops resolve an important source of uncertainty, serving as deterrents to anti-government mobilization. The result is greater political stability, which we observe as fewer periods of civil conflict.

To test our logic, we utilize new data on bilateral troop deployments globally between 1981 and 2006 (Braithwaite 2015). The data details deployments from over three dozen deploying states to as many as 150 hosts. Our core outcome of interest is the presence of new or ongoing unrest. Importantly, our tests account for the non-random nature of troop deployments. Troops are distributed globally as a result of deployers’ strategic interests and their relationships to the host. Failure to model this selection process could bias our estimates. Therefore, we employ two-stage tests in which we first model troop deployments, and then model troops’ effects on host stability.

We find that troops have a systematic, positive effect on host country political stability. Deployments reduce the likelihood of civil conflicts, effectively deterring opposition groups from mobilizing and incentivizing alternatives to fighting. This is true not just for violent civil wars but, perhaps more surprisingly, also for nonviolent mass mobilizations. Our analysis also supports the notion that troops’ effects may be channeled through the bolstering of rule of law. We show that troop deployments are correlated positively with the rule of law in host countries, and that rule of law is associated with a significantly lower likelihood of civil conflict. Our findings on rule of law suggest that troops can have a “positive” influence on the political climate of host states. In other words, foreign troops do not allow governments free rein to manage their local affairs, that is, by engaging in greater repression of opposition groups. Rather, their presence bolsters credibility through enhanced rule of law.

The paper proceeds as follows. First, we detail existing empirical accounts of the motivations behind troop deployments. Second, we outline our argument in which we suggest that while locals quite consistently view foreign troops as being illegitimate, these troops are functional in bolstering alternatives to fighting between governments and their opposition. Third, we design a research plan to test a hypothesis derived from our logic. Fourth, we discuss the results and substantive effects of a series of multivariate analyses. Finally, we discuss the implications of these results for academic and policy debates.

Background and Theory

Domestic political unrest—both violent and nonviolent—affected no fewer than 100 countries between 1946 and 2006.3 Accounts of these conflicts focus mainly on their “local” determinants—the features of countries that predispose them to instability.4 These include poverty (Fearon and Laitin 2003; Collier and Hoeffler 2002; Blomberg, Hess, and Thacker 2006; Braithwaite, Dasandi, and Hudson 2016), economic stagnation (Blomberg and Hess 2002; Miguel, Satyanath, and Sergenti 2004), natural resource dependence (Collier and Hoeffler 2002; Ross 2004), political rights and institutions (Hegre et al. 2001; Chenoweth and Stephan 2011), and societal factors such as ethnic and class cleavages and inequalities (Ostby 2008; Cederman, Weidmann, and Gleditsch 2011).

Complementing these “structural” perspectives is the literature on conflict bargaining. Drawing from economic theory (see, e.g., Kennan and Wilson 1993), bargaining models allow scholars to endogenize strategic interactions between governments and their opponents. These models show that conflict between governments and anti-government campaigns is shaped by each actor’s cost-benefit calculations, and that conflict is more likely under conditions of uncertainty.5 For example, actors may be uncertain as to their opponents’ resolve and/or capabilities (information problems) or with regard to the reliability of any deal made during bargaining (commitment problems). Either possibility helps explain why conflict occurs (Fearon 1995).

Both structural and bargaining perspectives advance our understanding of conflict’s origins. However, with some exceptions, both focus on domestic factors to the near exclusion of international ones. In particular, too little attention is paid to the role foreign countries play in shaping uncertainty prior to the onset of unrest.6 Two relevant literatures are the work on third-party interventions once conflict begins, and the work on interventions in post-conflict settings. In the first instance, Walter (1997) identifies third-party guarantees as a requisite for successful negotiated settlements in civil wars. In addition, Regan (2002) demonstrates that neutral interventions extend the duration of conflicts whereas biased interventions have the effect of hastening the conclusion of a conflict. Fortna (2004) highlights that the presence of peacekeepers on the ground in the aftermath of civil war helps maintain the peace between former belligerents. Second, studies demonstrate the interconnectedness of third-party interventions (including by mediators and peacekeepers) across various phases of conflict (Greig and Diehl 2005; Beardsley 2011; Diehl and Regan 2015). This research demonstrates that interventions can help reduce the likelihood of future unrest. Collectively, these works show that accounting for third parties is essential for a complete account of conflict dynamics.

This paper brings together these related findings. We start from the bargaining perspective’s focus on uncertainty, endeavoring to further specify its sources. In particular, we look at the presence of foreign troops and test whether deployments have a positive effect on political stability in host countries. Where existing work has explored the specific effect of foreign actors’ involvements after conflict has broken out, we contribute to the conflict processes literature by additionally paying attention to the possible effect of foreign states’ involvement during peacetime and prior to the onset of any new conflict.

Why Are Troops Deployed?

Troops were deployed to foreign countries in large numbers after World War II in what Schmidt (2014) describes as a new norm of hosting troops. Efforts by the United States and USSR to maintenance their spheres of influence led to the deployment of more than 1.3 million troops between 1950 and 1990. Many of these deployments persist despite the thawing of Cold War tensions. While the United States announced 15 base closures and additional troops reductions across Europe in 2015 (British Broadcasting Corporation 2015), it operates as many as 800 bases in more than 60 countries, and continues to deploy more than 250,000 troops overseas each year (Dufour 2016). Moreover, it is important to recognize that deployments, especially recently, are not limited to the traditional military powers. In the past decade, more than 30 other countries deployed an additional 175,000 troops abroad.

Figure 1 provides an overview of the volume of troops as well as the number of host countries annually between 1981 and 2006. These data include troops deployed as part of peacetime missions, such as US and UK troops deployed to West Germany/Germany. These data also cover cases in which troops are deployed as part of conflict interventions, including Australia, Denmark, Honduras, and Thailand, from among 40 or so countries that contributed troops to the so-called “Coalition-of-the-Willing” after the invasion of Iraq in 2003. Figure 1 shows, first, a decrease in the total number of troops deployed in the initial post–Cold War period but, with the lifting of Cold War umbrellas, a simultaneous increase in the number of countries hosting troops. Second, it shows a lull in overseas deployments (in terms of both numbers of troops and hosts) in the mid-1990s. Finally, we observe a general increase in both numbers from the late 1990s onward, which appears to be bolstered by the dynamics of the post-9/11 world.

Figure 1.

Volume of troops and number of hosts, 1981–2006

Figure 1.

Volume of troops and number of hosts, 1981–2006

Despite the considerable role overseas troops play in international politics, researchers have only begun to explore their causes and consequences. Studies typically view deployments as a costly signal of state interests when more overt expressions—such as the use of military force in conflict—are absent (Little and LeBlang 2004; Biglaiser and DeRouen 2007, 2009). Troop deployments reflect states’ efforts to protect their foreign economic and security interests. For instance, soldiers are sent more frequently to countries with which states have strong economic ties. Those deployments, in turn, encourage deeper ties, promoting trade between the deploying and host countries (Biglaiser and DeRouen 2007, 2009). To take one example, 2014 saw China deploy troops into South Sudan to support the fledgling new country and the UN mission there, as well as to protect growing Chinese oil investments.

Deployments also fit within states’ broader deterrence-based strategies. Whether conceived in terms of efficacy, in which more troops bring greater benefits (Huth 1988a, 1988b), or in terms of efficiency, in which the demonstration of will is sufficient to credibly signal commitment (Fearon 1997), deployments are viewed as an important tool for bolstering the stability of host governments against foreign and domestic invasion. Both deterrence logics view the presence (or promise) of troops as an important parameter boosting perceptions of government strength. Consider, for instance, the deployment of US and USSR troops on either side of the border between West and East Germany throughout the Cold War. One interpretation is that these deployments helped maintain stability for those two countries against domestic threats for nearly 30 years.7

What Are the Costs and Benefits of Hosting Foreign Troops?

Existing literature focuses on the benefits deploying countries accrue from sending troops abroad. However, implicit in arguments justifying deployments is the assumption that troops contribute to stability in host countries. The past few years have seen a spike in research exploring whether this is the case. A look at the historical record reveals mixed evidence. Beyond the example of Cold War Germany, long-term US deployments to South Korea in the post–Korean War period have arguably helped maintain stability. However, such deployments in East Asia and Europe have also coincided with noteworthy examples of domestic dissent, including terrorism, protests, and riots in response to the presence of these foreign troops and the occasionally poor behavior of their individual troops.

Nevertheless, some research suggests that foreign troops have a stabilizing effect. First, foreign troops help bolster incumbents. Brinkman (2001, 8) notes that “[m]ilitary interventions, the presence of foreign troops and (military) aid enabled several rulers to stay in power and assured perhaps a form of stability in the past in several cases.” For instance, the presence of foreign troops reduces the likelihood of coup d’etat in Africa (O’Kane 1993). Second, troops are associated with local economic growth. Jones and Kane (2007) investigate the impact that hosting US troops has on host countries’ economic fortunes. They find that the presence of US troops facilitates institutional improvement through three mechanisms: the provision of a security umbrella, the diffusion of new technologies, and a demand stimulus (associated with the presence of often large numbers of US soldiers and support teams). Third, foreign troops facilitate the diffusion of norms of stability. Building upon Jones and Kane (2007), Jones and Kane (2012) demonstrate that economic growth follows as a result of the diffusion of US norms regarding rule of law and respect for property and human rights. Bell, Clay, and Machain (2016) argue that the security interests of the deploying state condition the human rights benefits of foreign troops. In the case of the United States, when the state hosting US troops is less salient to US security interests, the presence of troops can encourage improved human rights practices. This effect is not present, however, when the host state is salient to US security interests.

On the flip side, hosting foreign troops can also have less desirable effects. Countries hosting troops are more likely to reduce their own troop levels and initiate militarized disputes overseas (for both, see Machain and Morgan 2013). They are also more likely to be targeted by actors hostile to the United States (Lutz 2009) than countries not hosting troops. Moreover, host countries increase their levels of defense spending (Allen, Van-Dusky-Allen, and Flynn forthcoming). Areas hosting US troops are also associated with higher levels of sexual assault (Lutz 2009; Yeo 2011), environmental degradation (Lutz 2009; Yeo 2011), and property crime (Allen and Flynn 2013).

Clearly, a growing literature speaks to the mechanisms that distribute the costs and benefits associated with hosting foreign troops. We could still learn more, however, regarding how foreign troops affect the strategic interactions between governments and latent and/or manifest opposition groups, including decisions to mobilize. A deeper understanding of these effects is essential for measuring deployment effectiveness—that is, whether troops contribute to political stability in host countries.

Do Foreign Troops Affect Host State Stability?

Our core claim is that foreign troops reduce the uncertainty between opposition groups and incumbent governments. We start from the observation that greater uncertainty leads to conflict.8 This uncertainty is especially problematic in domestic unrest because of the hard-to-observe resources on which rebels depend (Pillar 1983).

In the context of unrest, uncertainty manifests itself in multiple ways. First, it may generate information problems, which obscure judgments about relative capabilities. Second, it may generate credibility problems, which cast doubt over actors’ commitments to peace. We focus upon this second set of problems—centered around questions of commitment—where we argue that troops have an especially important role to play. In particular, we follow the efficiency-based logic of Fearon (1997) in which the demonstration of will is sufficient to credibly signal the commitment of third-party deployers.

Studies find that bargaining over the distribution of territory, political voice, or other policy issues is overshadowed by whether the parties’ commitments to alternatives to fighting are credible (Licklider 1995; Walter 1997).9 Rebels may opt to fight rather than continue to bargain in order to trick the government into offering concessions (Fearon 2007). Governments may renege upon bargains when conditions improve in their favor (de Figuerado and Weingast 1997; Boix 2003; Acemoglu and Robinson 2006). Commitment problems of this kind could be observed, for instance, in Iraq, where Sunni groups lacked trust in Shi’a promises after the withdrawal of the bulk of American troops. They also lie at the heart of the onset of the Second Sudanese Civil War. In 1983 President Nimeiry declared all of Sudan to be an Islamic state. This undermined elements of the Southern Sudan Autonomous Region, which prompted local Christians to fear the loss of rights from the central government. In response, the Sudan People’s Liberation Army (SPLA) was formed and they initiated conflict against the central government.

Thus, conflict is more likely given uncertainty generated by the presence of credibility problems. Extending this logic, we add the role of external influences (Gleditsch 2007; Checkel 2013; Maves and Braithwaite 2013). Of particular importance is the role of third-party states (see, e.g., Regan 2002; Fortna 2004; Balch-Lindsay, Enterline, and Joyce 2008). As we note, most existing work on foreign actors focuses on intervention into ongoing conflicts, and whether intervention influences calculations regarding the likelihood of victory. However, this neglects third parties that are already involved in a state’s domestic affairs.10

We argue that foreign troops help ameliorate uncertainty between the host government and latent or manifest opposition groups. This effect rests upon two connected commitments. First, as noted above, host governments and their opponents are committing (at least conceptually) to maintain local stability rather than engender unrest. Second, third-party governments deploy troops as a demonstration of their commitment to helping sustain this stability in the host country (Biglaiser and DeRouen 2007, 2009).11 The presence of troops can be thought of either as a “tying hands”12 commitment mechanism (Fearon 2007) or as “costly signals” in which deploying states sink costs in the form of non-recoverable operating expenses (Gartzke and Kagatoni 2015).13

Whereas Fearon (2007) treats these as separate mechanisms for signaling commitment, we follow Slantchev (2005, 545) in contending that “military actions can sink costs and tie hands at the same time.” Troops are signals that both host governments and their opposition groups are supposed to receive. Their deployment involves the sinking of costs associated with the physical movement, housing, and operation of the troops. However, we argue that these costs may prove to be lower than those associated with inaction and the potential collapse of stability in the host state, which could undermine both interests and investments.14

In addition to representing costly signals of the deploying state’s interest in maintaining local stability and reducing uncertainty in the bargain forged between local government and opposition groups, troops also provide practical support for stability. First, troops enhance the deployer’s preparations for military engagement, increasing their ability to act to preserve stability should any uses of force be threatened or occur. In other words, foreign troops may increase any defector’s expected ex post costs via the threat of punishment. Second, foreign troops may be employed in a training role, which could have the effect of raising the local military’s skill levels, professionalism, and respect for human rights (see, e.g., Bell, Clay, and Machain 2016). Third, foreign troops could serve to secure local borders, protecting against the spillover of negative externalities from conflicts elsewhere in the neighborhood (Murdoch and Sandler 2002; Buhaug and Gleditsch 2008). Both of these effects ought to raise the cost of any ex ante defection.

Fourth, and perhaps most importantly for our study moving forward, we argue that foreign troops can also help revise domestic political conditions that might lead to unrest. Not all governments face the same kinds of commitment problems. Existing work shows that commitment problems are especially severe in states with weak institutions and frail rule of law, where domestic shifts in power are likely to exacerbate uncertainty and undermine the enforcement of contracts (Walter 1997; Blattman and Miguel 2010). “[C]ivil war is more likely to occur when there are limits to conflict resolution and contract enforcement. Since formal legal and state institutions presumably help enforce commitments inter temporally, societies with weak government institutions and few checks and balances on executive power should empirically be the most likely to experience violent civil conflict” (Blattman and Miguel 2010, 13). Accordingly, several studies find that conflict is more likely to break out in weak or temporarily weakened countries than in strong ones (Hegre et al. 2001; Fearon and Laitin 2003; Hegre and Sambanis 2006; and Skaperdas 2008).

Foreign troops can help bolster rule of law in host countries. Third parties (and their agents) may help enforce contracts in the absence of strong domestic institutions or in the absence of existing rule of law (Walter 1997). Rule of law—defined as the quality, not just the strength of governance—is beneficial for two reasons. First, it helps compensate for structural conditions that precipitate conflict. As stated, having strong legal institutions and evenly, consistently enforced domestic statutes can help redress imbalances in political voice that may otherwise engender grievances among opposition groups. Rule of law is consistent with more predictability in the distribution of power—the key underlying issue in many civil conflicts. Second, rule of law implies the more consistent, even enforcement of contracts. Opposition groups’ worries over potential exploitation ought to be attenuated by strong institutions.15

Efforts to improve rule of law are an increasingly central part of troops’ missions abroad. The US Institute of Peace states that “fostering the development of the rule of law is central to conflict management.” To that end, US troops are deployed with this goal in mind. A Pentagon spokesperson recently stated troops would be deployed to Ukraine to “strengthen its law enforcement capabilities, conduct internal defense, and maintain rule of law.” While our analysis looks at troop deployments not directly related to peacekeeping missions—peacekeepers are controlled for separately in our analysis—the focus on promoting and protecting the rule of law in host countries echoes the United Nations’ growing focus on strengthening domestic institutions and law enforcement in host countries.

Taken together, troops play an important role in resolving uncertainty around the credibility of rivals’ commitments to negotiated—that is, peaceful—solutions. When senders devise deployments that are increasingly costly, the signal is greater and stability more likely. In other words, the size of the deployment ought to matter. The size could be thought of as an indicator of the level of commitment that the deploying state is willing to dedicate to the host state.

Accordingly, we hypothesize:

Hypothesis 1: Hosting foreign troops reduces the likelihood of unrest, with the size of the effect increasing in the size of the deployments hosted.

We recognize several alternative and competing predictions. To begin with, troops may be deployed for a variety of purposes. Some may side with the incumbent government; others may adopt a more agnostic approach, seeking only to maintain peace.16 Yet others may intervene explicitly on the side of opposition groups against the government. Take, for instance, Angola in 1975, which witnessed Cuban troops deployed in support of the MPLA government and South African troops deployed in support of the main opposition, UNITA. Moreover, even if troops are present to maintain the status quo, they may be ineffective due to insufficient resources or limited mandates. Our analysis takes these various possibilities into account through various controls and robustness checks. Notwithstanding these caveats, we predict that that the presence of troops reduces the likelihood of unrest on average.

Research Design

We construct a directed-dyad dataset with one row for each country pair i, j in year t, where country i is the potential troop deployer and j is the potential host. We rely on new data on bilateral troop deployments from 1981 to 2006 (Braithwaite 2015). Our data also detail whether or not country j experienced domestic political unrest in year t. We draw this information from the Nonviolent and Violent Conflict and Outcome (NACVO 2.0) database, the most comprehensive data on large-scale episodes of violent and nonviolent conflict over our period (Chenoweth and Lewis 2013).

The dyadic construction allows us to address the non-random deployment of troops, which is likely driven by the particulars of each country pair’s diplomatic and historical relations (see below). In addition, the dyadic setup allows us to test whether traits of the deploying nations shape the effects their troops have on hosts.17

Dependent Variables

Our outcome of interest is political stability. Stability can be conceptualized in various ways, such as the frequency of leadership turnover or the outbreak of interstate war. This paper is concerned with an increasingly salient form of instability—domestic political unrest. Unrest is defined as mobilizations of opposition groups against sitting governments. To operationalize unrest, we create a binary variable, Conflictj,t, which is coded as “1” for years in which the host country experienced unrest, regardless of the predominant tactic employed. In additional tests, we use indicators for whether the opposition’s tactics are predominantly Violentj,t or Nonviolentj,t.18

Our theory argues that foreign troops reduce uncertainty in a number of potential ways. It is difficult to measure uncertainty directly. Fortunately, it is possible to observe one specific channel through which troops may affect uncertainty broadly—rule of law. We argue that the presence of troops may help buttress the rule of law, reducing the credibility problems that lead to unrest. To measure Rule of Lawj,t, we utilize data provided by the World Bank’s World Governance Indicators, which provides a continuous indicator based on an index of composite sources.19

Explanatory Variables

Existing work on troops focuses largely on US deployments. This is an artifact of the central role US military operations play, but it also results from previously limited data. Braithwaite (2015) provides new data on the global distribution of bilateral troop deployments over time, allowing us to analyze the effects of troops deployed by states other than just the United States. Across the sample, there are more than 160 countries that hosted at least one foreign troop at some time between 1981 and 2006. And crucially, more than 60 countries deployed troops over the same period.

To demonstrate the varied distribution of troops, we present Figure 2. This illustrates the bilateral flow of troops between deploying and hosting countries for the years 1981 to 2006. Each line connects the country centroids of an observed deployer-host dyad of countries. This map helps reinforce the point that there exists a wide variety of both deploying and hosting countries, and that both sets of countries exist across multiple regions of the globe.

Figure 2.

Bilateral flow of troops between deployers and hosts, 1981–2006

Figure 2.

Bilateral flow of troops between deployers and hosts, 1981–2006

We measure troop deployments in several ways, taking into account the number of personnel deployed as well as their origin. Our core indicator is Troop Numbersi,j,t, coded as the logged count of personnel from deployer i located in host j in year t. We log troop numbers due to the highly skewed nature of the distribution. The mean of all non-zero observations in the sample is roughly 7,400 troops [SD: 27,000], and the maximum is nearly 400,000.20 In addition, using a continuous measure provides more information than a simple dichotomous indicator of the presence of troops, allowing us to see whether the size of the deployment matters. Deployment size is a rough indicator of the level of a deploying state’s commitment to the host.21 Note that these troop counts do not include peacekeepers, which we control for separately in our analysis. Our measure allows for the possibility that troops might be deployed for a variety of reasons, whereas peacekeepers are sent to conflict zones to intervene in ongoing violence.

Robustness checks consider several alternative codings. First, we omit deployments from the United States since (i) they account for a plurality of the deployments in our sample and, related, (ii) there might be something unique about US deployments. We do not theorize about country-specific differences in effectiveness. However, we want to ensure that the results are not driven by the United States only (Non-US Troopsi,j,t). In addition, we look at whether troops have an effect over a certain “threshold” mission size. Some troop deployments might consist of small numbers of personnel tied to diplomatic staff, and they are not tasked with maintaining stability per se. The size of troop deployments provides a loose approximation for the scope of a deployment’s mandate. We therefore include a measure of troop deployments that treats all missions smaller than 100 troops as “0” (Threshold Troopsi,j,t). We also look at specific traits of deployer state i and its relationship to host j. We create a logged count of troops sent only by host j’s top ten trade partners (Troops (Trade)i,j,t)22 and those sent only by a host’s former colonizers (Troops (Colonial)i,j,t). These variables test whether troops deployed among dyads with particularly strong economic or historical ties have unique effects on hosts.

Additionally, we consider the timing and legacy of troop deployments in order to explore whether their effect changes over time. For example, Troops (Proportion)i,j,t measures the five-year moving average of troop numbers in a given host. This measure helps capture the scale of the commitment shown by the deploying state toward the host. We also constructed a measure of whether the troops were deployed within the year t (New Deploymenti,j,t). This measure provides a way of considering whether troops have an immediate effect on host countries.

Control Variables

The models include a variety of controls. To begin with, we include a measure of each state’s market size (GDPi,t and GDPj,t), which controls for the role that market power plays in the deployment of troops, and in shaping the likelihood of unrest.23 Using Cheibub, Gandhi, and Vreeland (2010), we also control for each state’s regime type (Democracyi,t and Democracyj,t), since democracy level may likewise affect both deployment and conflict. For example, shared regime characteristics may influence the likelihood of deployment, and democratic states ought to be less vulnerable to large-scale opposition group mobilizations.

Relating to broader host stability, we measure whether the host is engaged in an ongoing militarized interstate dispute (MIDj,t). This data is provided by the Correlates of War Project (COW; Palmer et al. 2015). Specifically, we look at whether the host country is targeted by an MID, which should affect both troop deployments and conflict in the host. MIDs that escalate beyond threats to include shows and uses of force will require the movement of troops. This will often include their deployment outside the sovereign territories of the state. Accordingly, MIDs can have an important bearing upon the location of the state’s network of foreign deployed troops.24 Second, we include the logged Populationj,t of the host state, which is a common predictor of unrest.25

Our model also includes measures of state capabilities (Capabilitiesi,t and Capabilitiesj,t) using the National Material Capabilities measure provided by the Correlates of War Project. States’ capabilities shape deployment by affecting a sender’s ability to deploy troops as well as a potential host’s ex ante demand for them. Capabilities similarly affect the occurrence of conflict, with countries being less vulnerable to instability at higher capability scores.26

To distinguish the effects of troop deployments from peacekeeping operations, we control for the number of Peacekeepersi,t deployed in the host state. These include United Nations missions as well as deployments sponsored by regional organizations such as the African Union. Controlling for peacekeepers helps isolate the independent effect that bilateral troop deployments have on host country stability.

Finally, we include a dichotomous variable flagging years during the Cold Wart. Deployment patterns were shaped fundamentally by Cold War politics, as shown in Figure 1, and we have seen significant reductions in US forces abroad since the 1980s. Yet we have also seen increases in deployments from other states, particularly in the past decade. As a result, the numbers and sources of troop deployments vary significantly over time. Controlling for the Cold War era helps capture whether unique traits of that period influenced the role that troops played in hosts.

Note that we do not include time-invariant measures common to many dyadic analyses, such as geographic proximity. Instead, our tests include dyad fixed effects, which effectively control for features of the country pair’s relationship that are constant over time.27 The additional advantage of dyad fixed effects is that it isolates within-dyad variation. Accordingly, our estimates tell us whether the presence of troops reduces conflict relative to the panel (dyad) mean.

Endogeneity

We design our analysis to account for the non-random nature of troop deployments. Troop deployments are a strategic decision likely driven by the shared interests of sender and host states, including their historical relationship and their desire to maintain stability in a given country (and in their broader neighborhood). Failing to recognize the non-random assignment of our treatment may result in biased estimates.

Modeling non-random assignment requires variables related to deployments but unrelated to unrest. We rely on two instruments. First, we include a measure of each dyad’s alliance ties. Troops may be deployed when an alliance commitment is invoked—that is, when the sender states comes to the aid of the host. More generally, alliance ties may simply proxy for existing strategic interests that make deployment more likely. We measure Alliancei,j,t ties using a new continuous coding provided by D’Orazio (2015), who presents a novel algorithm to measure similarities between states’ alliance portfolios. Second, we include a measure of political Affinityi,j,t based on patterns of voting behavior in the United Nations General Assembly. As recommended in Voeten and Merdzanovic (2009), we utilize dyad ideal points to determine the existing similarities between sender and host voting patterns. Closer voting patterns ought to represent shared foreign policy interests, and increase the likelihood that troops are deployed between these countries. Note that these two instruments are not highly correlated (correlation = 0.09). Our two primary instruments capture features of the sender-host relationship that influence deployments. However, one potential concern is that opposition groups’ decisions to mobilize are themselves based on whether they anticipate foreign intervention. If opposition groups consider the likelihood of foreign intervention before deciding to act, then alliance ties and political affinity will shape civil conflict occurrence and fail to meet the exclusion restriction.

There are several responses to this concern. First, the troop deployments in our sample are not, most commonly, deployed in direct response to unrest. Many deployments are instead long-standing and occur in peacetime, allowing us to observe whether they lessen the frequency of conflict over time. Second, it is not obvious that opposition groups can anticipate deployments with any reasonable degree of confidence. They are unlikely to have access to confidential information about the deployment of troops, or about the size and mandate of that deployment. As a result, opposition groups’ decisions to mobilize are more likely based on whether troops are already present, which is consistent with our argument.

In either case, we assume that non-random selection is driven more by traits of the sender-host relationship than by the decisions of opposition groups. The available diagnostic tests increase our confidence in the validity of the instruments. We ensure that the selection of instruments is not driving the core results and leading to false inferences. Our baseline estimates are robust to alternative measures of alliance ties as well as the inclusion of economic ties such as bilateral trade.

Summary statistics of the variables used in this study are reported in Table 1.

Table 1.

Summary statistics

Variable Mean Std. Dev. Min. Max. 
Outcomes 
Unrestj,t 0.147 0.354 
Violencej,t 0.125 0.331 
Nonviolencej,t 0.022 0.146 
Troops numberi,j,t 0.028 0.456 12.313 
Thresholdi,j,t 0.026 0.447 12.313 
Troops proportioni,j,t 0.035 0.496 11.826 
Troops tradei,j,t 0.006 0.237 12.313 
Troops colonyi,j,t 0.006 0.208 10.558 
New deploymentsi,j,t 0.016 0.338 12.313 
Rule of lawj,t –0.131 0.920 –2.23 
Controls 
GDPi,t 2.318 0.239 1.682 3.007 
GDPj,t 2.291 0.233 1.682 2.926 
Democracyi,t 0.667 0.471 
Democracyj,t 0.423 0.494 
Capabilitiesi,t 0.005 0.018 0.190 
Capabilitiesj,t 0.006 0.017 0.190 
MIDj,t 0.263 0.603 
Populationj,t 15.625 2.097 9.162 21 
PostCold Wart 0.079 0.040 0.1 
Peacekeepersi,t 0.027 0.320 9.188 
Instruments 
Alliancei,j,t 5.956 21.202 153 
Affinityi,j,t 0.840 0.144 
Variable Mean Std. Dev. Min. Max. 
Outcomes 
Unrestj,t 0.147 0.354 
Violencej,t 0.125 0.331 
Nonviolencej,t 0.022 0.146 
Troops numberi,j,t 0.028 0.456 12.313 
Thresholdi,j,t 0.026 0.447 12.313 
Troops proportioni,j,t 0.035 0.496 11.826 
Troops tradei,j,t 0.006 0.237 12.313 
Troops colonyi,j,t 0.006 0.208 10.558 
New deploymentsi,j,t 0.016 0.338 12.313 
Rule of lawj,t –0.131 0.920 –2.23 
Controls 
GDPi,t 2.318 0.239 1.682 3.007 
GDPj,t 2.291 0.233 1.682 2.926 
Democracyi,t 0.667 0.471 
Democracyj,t 0.423 0.494 
Capabilitiesi,t 0.005 0.018 0.190 
Capabilitiesj,t 0.006 0.017 0.190 
MIDj,t 0.263 0.603 
Populationj,t 15.625 2.097 9.162 21 
PostCold Wart 0.079 0.040 0.1 
Peacekeepersi,t 0.027 0.320 9.188 
Instruments 
Alliancei,j,t 5.956 21.202 153 
Affinityi,j,t 0.840 0.144 

Discussion of Results

Across a variety of model specifications and estimation techniques, we find a strong, negative association between troop deployments and unrest. This is consistent across both nonviolent and violent mobilizations, and when using alternative codings of troop deployments. Substantively, troops reduce the frequency of unrest (for an average country-year) as much as 50 percent. Our tests also point to one specific channel through which troops reduce unrest—buttressing the rule of law. Troop deployments help bolster the rule of law, which, in turn, reduces uncertainty. The result is a reduced likelihood of unrest onset or continuation.

Model Specification

We present a series of two-stage least squares (2SLS) models designed to account for the sequential, related nature of troop deployments and conflict. This estimator is appropriate for situations in which we expect correlation in the errors between two related processes (Wooldridge 2010).28 In this case, the first stage regresses our endogenous indicator of troop deployments on the control variables and instruments. The second stage estimates the relationship between troop deployments and civil conflict.

All of our models include directed-dyad fixed effects to account for the unobserved features of a given country pair.29 The inclusion of dyad fixed effects means we isolate the within-dyad variation, rather than differences across dyads. In addition, including dyad FEs effectively means we are looking at the difference in any given year t from the panel i,j mean. As a result, we do not need to difference our dependent variable. Finally, we cluster the standard errors by directed dyad.

Note that our measures of troop deployments are lagged one year. We cannot always observe the specific time of year in which either (a) troops are deployed or (b) unrest occurs. Lagging the variable ensures that troops were already on the ground during the year in which we either do or do not observe conflict. Because our first stage predicts deployments, this also means that our right-hand-side variables are all lagged one year.

Our models also address the time-dependent nature of conflict. The historical record shows that civil conflict is significantly more likely to occur in countries that have experienced prior episodes. Our models include a lagged dependent variable to account for whether host country j experienced a conflict in year t-1.30

The Effects of Troop Deployments on Conflict

Model 1 uses a logged count of troops deployed in host j and our indicator of whether conflict of any type occurred. The estimates are reported in Table 2. First, note that both of the instruments are strong predictors of troop deployments. Looking at the diagnostic tests, the under-identification test interrogates the null hypothesis that the equation is under-identified—that is, whether or not the instruments are relevant. The test statistic is highly statistically significant (p < 0.015), which allows us to reject the null. The over-identification (J-statistic) test explores the null hypothesis that the instruments are valid (uncorrelated with the error term). Its test statistic falls short of achieving statistical significance (p > 0.05) in each instance. Accordingly, we cannot reject the null. These tests provide evidence that the instruments are relevant and valid.31

Table 2.

Baseline estimates

 (1) (2) (3) 
 Troop  Troop  Troop  
 numberi,j,t Unrestj,t numberi,j,t Unrestj,t numberi,j,t Unrestj,t 
Troop numberi,j,t  –2.189*  –2.758*   
  (0.899)  (1.310)   
Thresholdi,j,t      –3.302* 
      (1.623) 
GDPi,t 0.134* –0.100 0.128* –0.037 0.118* –0.003 
 (0.057) (0.176) (0.055) (0.209) (0.056) (0.268) 
GDPj,t 0.040 0.149 0.017 0.113 0.013 0.105 
 (0.039) (0.096) (0.032) (0.091) (0.034) (0.119) 
MIDj,t –0.001 0.011** –0.001 0.010** –0.001 0.009* 
 (0.001) (0.002) (0.001) (0.002) (0.001) (0.004) 
Democracyi,t 0.004 0.007 0.006** 0.015 0.004 0.012 
 (0.002) (0.007) (0.002) (0.009) (0.002) (0.010) 
Democracyj,t 0.004 –0.025** 0.008** –0.013 0.004 –0.022 
 (0.003) (0.008) (0.003) (0.013) (0.003) (0.013) 
Capabilitiesi,t 6.170** 14.117* 5.471** 15.847 5.514** 18.822 
 (1.924) (6.507) (2.000) (9.030) (1.842) (10.271) 
Capabilitiesj,t –0.157 –1.942** –0.144 –2.003** –0.103 –1.941** 
 (0.169) (0.378) (0.165) (0.467) (0.166) (0.544) 
Populationj,t –0.021 –0.170** –0.029** –0.209** –0.018 –0.185** 
 (0.015) (0.040) (0.011) (0.046) (0.014) (0.058) 
PostCold Wart –0.060 0.200* –0.056 0.190* –0.053 0.159 
 (0.035) (0.093) (0.029) (0.096) (0.034) (0.139) 
Peacekeepersi,t –0.003 –0.021 –0.002 –0.020 –0.003 –0.026 
 (0.005) (0.012) (0.006) (0.016) (0.005) (0.018) 
Unrestj,t-1 0.009* 0.625** 0.008* 0.628** 0.008* 0.633** 
 (0.004) (0.013) (0.004) (0.014) (0.004) (0.019) 
Affinityi,j,t –0.017  0.006  −0.010  
 (0.011)  (0.008)  (0.010)  
Alliancei,j,t 0.000*  0.000*  0.000*  
 (0.000)  (0.000)  (0.000)  
N 322446 318414 322446 
 (1) (2) (3) 
 Troop  Troop  Troop  
 numberi,j,t Unrestj,t numberi,j,t Unrestj,t numberi,j,t Unrestj,t 
Troop numberi,j,t  –2.189*  –2.758*   
  (0.899)  (1.310)   
Thresholdi,j,t      –3.302* 
      (1.623) 
GDPi,t 0.134* –0.100 0.128* –0.037 0.118* –0.003 
 (0.057) (0.176) (0.055) (0.209) (0.056) (0.268) 
GDPj,t 0.040 0.149 0.017 0.113 0.013 0.105 
 (0.039) (0.096) (0.032) (0.091) (0.034) (0.119) 
MIDj,t –0.001 0.011** –0.001 0.010** –0.001 0.009* 
 (0.001) (0.002) (0.001) (0.002) (0.001) (0.004) 
Democracyi,t 0.004 0.007 0.006** 0.015 0.004 0.012 
 (0.002) (0.007) (0.002) (0.009) (0.002) (0.010) 
Democracyj,t 0.004 –0.025** 0.008** –0.013 0.004 –0.022 
 (0.003) (0.008) (0.003) (0.013) (0.003) (0.013) 
Capabilitiesi,t 6.170** 14.117* 5.471** 15.847 5.514** 18.822 
 (1.924) (6.507) (2.000) (9.030) (1.842) (10.271) 
Capabilitiesj,t –0.157 –1.942** –0.144 –2.003** –0.103 –1.941** 
 (0.169) (0.378) (0.165) (0.467) (0.166) (0.544) 
Populationj,t –0.021 –0.170** –0.029** –0.209** –0.018 –0.185** 
 (0.015) (0.040) (0.011) (0.046) (0.014) (0.058) 
PostCold Wart –0.060 0.200* –0.056 0.190* –0.053 0.159 
 (0.035) (0.093) (0.029) (0.096) (0.034) (0.139) 
Peacekeepersi,t –0.003 –0.021 –0.002 –0.020 –0.003 –0.026 
 (0.005) (0.012) (0.006) (0.016) (0.005) (0.018) 
Unrestj,t-1 0.009* 0.625** 0.008* 0.628** 0.008* 0.633** 
 (0.004) (0.013) (0.004) (0.014) (0.004) (0.019) 
Affinityi,j,t –0.017  0.006  −0.010  
 (0.011)  (0.008)  (0.010)  
Alliancei,j,t 0.000*  0.000*  0.000*  
 (0.000)  (0.000)  (0.000)  
N 322446 318414 322446 

Clustered standard errors in parentheses. *p < 0.050, **p < 0.001.

Note that the control variables generally behave as expected. Larger host markets, those with higher capacity levels, and those with democratic institutions all experience less conflict. Conversely, high levels of ethnic fractionalization are associated with more conflict. The one surprising result is the negative coefficient on MIDs, but the estimate is far from significant. The performance of these controls gives us additional confidence in our model specification.

The estimates confirm our hypothesis. The presence of foreign troops is associated with a significant reduction in civil conflict.32 This is consistent with the notion that foreign troops help reduce the kind of uncertainty inherent in commitment problems. Holding all other variables at their sample means, a move across the interquartile range (IQR) of Troop Numbersi,j,t results in roughly a 50 percent drop in the likelihood of experiencing civil conflict.33

Model 1 provides preliminary support for our theory. However, there are a variety of additional ways to measure our outcomes of interest. Model 2 restricts the sample to only non-US deployments. This tests whether unique features of US troops drive the relationship. Model 3 uses a recoded troop count, where only deployments of 100 or more troops are counted (all smaller numbers are treated as zero). This helps ensure that small, diplomatic deployments to primarily countries at peace are not driving the results. The estimates are reported in Table 2, and are consistent with the baseline model.

Models 4–6 explore different codings of troop deployments (Table 3). Model 4 measures the five-year moving average of troop deployments. The estimates here are consistent with those found in Model 1; a greater proportion of years with troops present is strongly associated with a reduction in conflict. A move across the interquartile range of Troops (Proportion)i,j,t results in a 28 percent drop in the likelihood of civil conflict. This finding points to the potential importance of troop presence over time, not just in a given year t.

Table 3.

Additional specifications: alternative troop measures

 (4) (5) (6) 
 Troop  Troop  Troop  
 prop.i,j,t Unrestj,t tradei,j,t Unrestj,t colonyi,j,t Unrestj,t 
T: Proportioni,j,t  –0.732**     
  (0.228)     
T: Tradei,j,t    –9.179   
    (5.973)   
T: Colonyi,j,t      –2.395 
      (3.395) 
GDPi,t 0.103 –0.462** 0.029 –0.120 0.015 –0.368** 
 (0.067) (0.064) (0.017) (0.128) (0.013) (0.062) 
GDPj,t –0.051 0.172** 0.013 0.178 –0.001 0.055 
 (0.052) (0.047) (0.010) (0.103) (0.018) (0.050) 
MIDj,t –0.001 0.013** –0.000 0.012** 0.000 0.013** 
 (0.001) (0.002) (0.000) (0.003) (0.000) (0.002) 
Democracyi,t 0.006* 0.001 0.000 –0.000 –0.001 –0.002 
 (0.003) (0.004) (0.000) (0.005) (0.001) (0.003) 
Democracyj,t 0.010** –0.038** –0.000 –0.035** 0.003* –0.027* 
 (0.003) (0.004) (0.000) (0.004) (0.002) (0.012) 
Capabilitiesi,t 7.210** 5.927** –0.333 –2.442 –0.487 –0.503 
 (1.913) (1.717) (0.403) (3.091) (0.385) (1.637) 
Capabilitiesj,t –0.453 –2.246** –0.069 –2.239** –0.047 –1.693** 
 (0.536) (0.401) (0.049) (0.334) (0.040) (0.175) 
Populationj,t –0.031 –0.186** –0.006 –0.178** –0.000 –0.125** 
 (0.022) (0.020) (0.004) (0.031) (0.005) (0.012) 
PostCold Wart –0.056 0.234** –0.001 0.325** –0.001 0.320** 
 (0.033) (0.032) (0.013) (0.124) (0.007) (0.025) 
Peacekeepersi,t 0.005 –0.011 0.000 –0.015** 0.004 –0.005 
 (0.008) (0.006) (0.000) (0.002) (0.005) (0.017) 
Unrestj,t-1 0.010* 0.588** 0.001 0.617** 0.001 0.609** 
 (0.005) (0.005) (0.001) (0.007) (0.001) (0.006) 
Affinityi,j,t –0.056**  –0.003  0.003  
 (0.015)  (0.004)  (0.003)  
Alliancei,j,t 0.000*  0.000  0.000  
 (0.000)  (0.000)  (0.000)  
N 272,109 322,446 322,446 
 (4) (5) (6) 
 Troop  Troop  Troop  
 prop.i,j,t Unrestj,t tradei,j,t Unrestj,t colonyi,j,t Unrestj,t 
T: Proportioni,j,t  –0.732**     
  (0.228)     
T: Tradei,j,t    –9.179   
    (5.973)   
T: Colonyi,j,t      –2.395 
      (3.395) 
GDPi,t 0.103 –0.462** 0.029 –0.120 0.015 –0.368** 
 (0.067) (0.064) (0.017) (0.128) (0.013) (0.062) 
GDPj,t –0.051 0.172** 0.013 0.178 –0.001 0.055 
 (0.052) (0.047) (0.010) (0.103) (0.018) (0.050) 
MIDj,t –0.001 0.013** –0.000 0.012** 0.000 0.013** 
 (0.001) (0.002) (0.000) (0.003) (0.000) (0.002) 
Democracyi,t 0.006* 0.001 0.000 –0.000 –0.001 –0.002 
 (0.003) (0.004) (0.000) (0.005) (0.001) (0.003) 
Democracyj,t 0.010** –0.038** –0.000 –0.035** 0.003* –0.027* 
 (0.003) (0.004) (0.000) (0.004) (0.002) (0.012) 
Capabilitiesi,t 7.210** 5.927** –0.333 –2.442 –0.487 –0.503 
 (1.913) (1.717) (0.403) (3.091) (0.385) (1.637) 
Capabilitiesj,t –0.453 –2.246** –0.069 –2.239** –0.047 –1.693** 
 (0.536) (0.401) (0.049) (0.334) (0.040) (0.175) 
Populationj,t –0.031 –0.186** –0.006 –0.178** –0.000 –0.125** 
 (0.022) (0.020) (0.004) (0.031) (0.005) (0.012) 
PostCold Wart –0.056 0.234** –0.001 0.325** –0.001 0.320** 
 (0.033) (0.032) (0.013) (0.124) (0.007) (0.025) 
Peacekeepersi,t 0.005 –0.011 0.000 –0.015** 0.004 –0.005 
 (0.008) (0.006) (0.000) (0.002) (0.005) (0.017) 
Unrestj,t-1 0.010* 0.588** 0.001 0.617** 0.001 0.609** 
 (0.005) (0.005) (0.001) (0.007) (0.001) (0.006) 
Affinityi,j,t –0.056**  –0.003  0.003  
 (0.015)  (0.004)  (0.003)  
Alliancei,j,t 0.000*  0.000  0.000  
 (0.000)  (0.000)  (0.000)  
N 272,109 322,446 322,446 

Clustered standard errors in parentheses. *p < 0.050, **p < 0.001.

Models 5 and 6 restrict the measure of troops to only those sent from host j’s top ten trade partners and colonial ancestors, respectively (Table3). Perhaps surprisingly, the results do not hold in either case. But these null findings are informative; the results reveal that the origins of troops are not necessarily a strong predictor of their effect. Instead, the mere presence of troops appears to have a much more significant impact on host country stability than their source. In addition, note that these measures are not highly correlated with our main indicator Troop Numbersi,j,t (less than 0.43 in both cases). Therefore, the null results here may be driven by the fact that a large portion of the troops deployed around the globe are simply not associated with trade flows or colonial ties.

The estimates show a strong relationship between troops and the reduction of conflict in general. Model 7 uses an alternative indicator of troop deployments, measuring just those troops deployed in the past year (Table 4). The estimates show that troop deployments have an immediate effect. They are significantly associated with a drop in the likelihood of conflict in their first year of deployment.34

Table 4.

Additional specifications: recent history of protest, past unrest, and regime type

 (7) (8) (9) (10) 
 Troop  Troop  Troop  Troop  
 numberi,j,t Unrestj,t numberi,j,t Unrestj,t numberi,j,t Unrestj,t numberi,j,t Unrestj,t 
New deployi,j,t  –2.008**       
  (0.753)       
Past failuresj,t   –0.010* –0.244**     
   (0.004) (0.013)     
Troop numberi,j,t    –2.004*  –2.770*  1.412 
    (0.820)  (1.218)  (1.192) 
GDPi,t 0.047 –0.300** 0.141* 0.056 0.146* 0.054 0.039 –0.280** 
 (0.036) (0.080) (0.058) (0.167) (0.073) (0.281) (0.060) (0.101) 
GDPj,t 0.054 0.169* 0.036 0.037 0.015 0.128 0.106 –0.027 
 (0.036) (0.084) (0.039) (0.087) (0.046) (0.134) (0.095) (0.171) 
MIDj,t 0.001 0.015** –0.001 0.013** –0.002 0.012** 0.002 0.000 
 (0.001) (0.003) (0.001) (0.002) (0.001) (0.004) (0.002) (0.004) 
Democracyi,t –0.003 –0.006 0.004 0.009 0.006* 0.019   
 (0.002) (0.004) (0.002) (0.006) (0.003) (0.013)   
Democracyj,t 0.002 –0.030** 0.005 –0.025** 0.004 –0.001   
 (0.003) (0.006) (0.003) (0.008) (0.005) (0.014)   
Capabilitiesi,t 3.658* 7.962* 6.163** 12.812* 6.321** 17.980* 6.389 –7.711 
 (1.612) (3.969) (1.924) (5.926) (2.047) (9.024) (5.829) (10.730) 
Capabilitiesj,t –0.129 –1.854** –0.084 –0.087 –0.157 –2.168** 0.539 –0.237 
 (0.098) (0.212) (0.176) (0.377) (0.171) (0.496) (1.287) (2.150) 
Populationj,t –0.002 –0.129** –0.017 –0.069 –0.011 –0.167** –0.039 –0.198** 
 (0.013) (0.028) (0.015) (0.035) (0.019) (0.059) (0.035) (0.062) 
PostCold Wart 0.006 0.342** –0.056 0.313** –0.125** 0.066 0.126* 0.135 
 (0.026) (0.056) (0.035) (0.083) (0.042) (0.190) (0.052) (0.179) 
Peacekeepersi,t 0.005 –0.006 –0.002 –0.011 –0.004 –0.030 –0.010 0.006 
 (0.003) (0.008) (0.006) (0.011) (0.006) (0.017) (0.007) (0.016) 
Unrestj,t-1 –0.000 0.605** 0.008* 0.601** 0.009 0.646** 0.006 0.498** 
 (0.002) (0.006) (0.004) (0.011) (0.005) (0.019) (0.004) (0.011) 
Affinityi,j,t –0.020*  –0.018  –0.022  –0.022  
 (0.009)  (0.011)  (0.013)  (0.021)  
Alliancei,j,t 0.000*  0.000*  0.001  0.000  
 (0.000)  (0.000)  (0.000)  (0.000)  
N 322,446 322,446 225,599 96,406 
 (7) (8) (9) (10) 
 Troop  Troop  Troop  Troop  
 numberi,j,t Unrestj,t numberi,j,t Unrestj,t numberi,j,t Unrestj,t numberi,j,t Unrestj,t 
New deployi,j,t  –2.008**       
  (0.753)       
Past failuresj,t   –0.010* –0.244**     
   (0.004) (0.013)     
Troop numberi,j,t    –2.004*  –2.770*  1.412 
    (0.820)  (1.218)  (1.192) 
GDPi,t 0.047 –0.300** 0.141* 0.056 0.146* 0.054 0.039 –0.280** 
 (0.036) (0.080) (0.058) (0.167) (0.073) (0.281) (0.060) (0.101) 
GDPj,t 0.054 0.169* 0.036 0.037 0.015 0.128 0.106 –0.027 
 (0.036) (0.084) (0.039) (0.087) (0.046) (0.134) (0.095) (0.171) 
MIDj,t 0.001 0.015** –0.001 0.013** –0.002 0.012** 0.002 0.000 
 (0.001) (0.003) (0.001) (0.002) (0.001) (0.004) (0.002) (0.004) 
Democracyi,t –0.003 –0.006 0.004 0.009 0.006* 0.019   
 (0.002) (0.004) (0.002) (0.006) (0.003) (0.013)   
Democracyj,t 0.002 –0.030** 0.005 –0.025** 0.004 –0.001   
 (0.003) (0.006) (0.003) (0.008) (0.005) (0.014)   
Capabilitiesi,t 3.658* 7.962* 6.163** 12.812* 6.321** 17.980* 6.389 –7.711 
 (1.612) (3.969) (1.924) (5.926) (2.047) (9.024) (5.829) (10.730) 
Capabilitiesj,t –0.129 –1.854** –0.084 –0.087 –0.157 –2.168** 0.539 –0.237 
 (0.098) (0.212) (0.176) (0.377) (0.171) (0.496) (1.287) (2.150) 
Populationj,t –0.002 –0.129** –0.017 –0.069 –0.011 –0.167** –0.039 –0.198** 
 (0.013) (0.028) (0.015) (0.035) (0.019) (0.059) (0.035) (0.062) 
PostCold Wart 0.006 0.342** –0.056 0.313** –0.125** 0.066 0.126* 0.135 
 (0.026) (0.056) (0.035) (0.083) (0.042) (0.190) (0.052) (0.179) 
Peacekeepersi,t 0.005 –0.006 –0.002 –0.011 –0.004 –0.030 –0.010 0.006 
 (0.003) (0.008) (0.006) (0.011) (0.006) (0.017) (0.007) (0.016) 
Unrestj,t-1 –0.000 0.605** 0.008* 0.601** 0.009 0.646** 0.006 0.498** 
 (0.002) (0.006) (0.004) (0.011) (0.005) (0.019) (0.004) (0.011) 
Affinityi,j,t –0.020*  –0.018  –0.022  –0.022  
 (0.009)  (0.011)  (0.013)  (0.021)  
Alliancei,j,t 0.000*  0.000*  0.001  0.000  
 (0.000)  (0.000)  (0.000)  (0.000)  
N 322,446 322,446 225,599 96,406 

Clustered standard errors in parentheses. * p < 0.050, ** p < 0.001.

Table 4 also reports the estimates when using a different indicator of conflict history in the host state. Model 8 includes a measure of whether any past conflicts in the host resulted in a “failure” as coded by the NAVCO 2.0 data. Failures are situations in which the opposition group did not secure any of its stated aims. The outcome of past conflicts is important; existing work shows that failed mobilizations increase the likelihood of future conflict. Model 8 includes this control, and the estimates remain consistent with the baseline models.

Additional Robustness Checks

Our baseline estimates are robust to a variety of alternative specifications. Many troop deployments during the Cold War linked countries with collectively low likelihoods of experiencing conflict—namely, the democratic countries to which the United States and UK deployed troops after World War II. These deployments were not just aimed at ensuring the stability of host countries but also about maintaining a presence in a given region. With this in mind, we ran additional tests using split samples. Model 9 includes only democratic country pairs. Model 10 includes mixed (democracy-autocracy) dyads. The estimates (reported in Table 4) show that the results are actually much stronger for mixed dyads. This finding suggests that the core results are not driven by “easy cases,” such as post–World War II US deployments to Germany and Japan. Instead, it is those cases including less democratic hosts which are historically more prone to unrest. This finding is especially valuable given the theory outlined above; it demonstrates that troops have a greater stabilizing effect in those countries with weak institutions, which might be expected to undermine the enforcement of contracts.

Models 11 and 12 look at alternative measures of civil conflict (Table 5). Specifically, we explore whether the influence of troops varied by tactical form of unrest—for example, violent (Model 11) versus nonviolent (Model 12) mobilizations. Table 5 shows that foreign troops are associated with a lower likelihood of both forms of unrest. Violent unrest is reduced by 40 percent and nonviolent unrest is nearly eliminated, being reduced by over 90 percent over the significant range. Note that the large substantive effect here derives from the relative infrequency of nonviolent unrest in the sample. The sample mean for violent mobilizations is 0.14 (SD: 0.35), while it is only 0.02 (SD: 0.15) for nonviolent ones, meaning that reductions in the frequency of nonviolence naturally approach zero.35 As a result, we infer that troops deter both kinds of conflict, though we would not conclude too much from the relative sizes of the substantive effects.

Table 5.

Additional specifications: violent and nonviolent unrest

 (11) (12) 
 Troop Violent Troop Nonviolent 
 numberi,j,t unrestj,t numberi,j,t unrestj,t 
Troop numberi,j,t  –0.935*  −1.591* 
  (0.411)  (0.673) 
Violencej,t 0.011* 0.703**   
 (0.005) (0.008)   
Nonviolencej,t   0.001 0.425** 
   (0.006) (0.010) 
GDPi,t 0.134* –0.270** 0.130* 0.233 
 (0.057) (0.080) (0.056) (0.127) 
GDPj,t 0.041 0.171** 0.036 0.031 
 (0.039) (0.045) (0.039) (0.067) 
MIDi,t –0.000 0.006** –0.000 0.009** 
 (0.001) (0.001) (0.001) (0.002) 
Democracyi,t 0.004 –0.000 0.004 0.009 
 (0.002) (0.003) (0.002) (0.005) 
Democracyj,t 0.004 0.029** 0.004 –0.054** 
 (0.003) (0.004) (0.003) (0.006) 
Capabilitiesi,t 6.170** 6.315* 6.176** 9.877* 
 (1.925) (2.922) (1.924) (4.817) 
Capabilitiesj,t –0.171 –1.003** –0.179 –1.303** 
 (0.169) (0.178) (0.167) (0.281) 
Populationj,t –0.021 –0.141** –0.023 –0.036 
 (0.015) (0.018) (0.015) (0.030) 
PostCold Warj,t –0.059 0.216** –0.052 –0.011 
 (0.035) (0.042) (0.034) (0.064) 
Peacekeepersj,t –0.003 –0.014* –0.003 –0.008 
 (0.005) (0.005) (0.006) (0.009) 
Affinityi,j,t –0.017  –0.018  
 (0.011)  (0.011)  
Alliancei,j,t 0.000*  0.000*  
 (0.000)  (0.000)  
N 322,446 322,446 
 (11) (12) 
 Troop Violent Troop Nonviolent 
 numberi,j,t unrestj,t numberi,j,t unrestj,t 
Troop numberi,j,t  –0.935*  −1.591* 
  (0.411)  (0.673) 
Violencej,t 0.011* 0.703**   
 (0.005) (0.008)   
Nonviolencej,t   0.001 0.425** 
   (0.006) (0.010) 
GDPi,t 0.134* –0.270** 0.130* 0.233 
 (0.057) (0.080) (0.056) (0.127) 
GDPj,t 0.041 0.171** 0.036 0.031 
 (0.039) (0.045) (0.039) (0.067) 
MIDi,t –0.000 0.006** –0.000 0.009** 
 (0.001) (0.001) (0.001) (0.002) 
Democracyi,t 0.004 –0.000 0.004 0.009 
 (0.002) (0.003) (0.002) (0.005) 
Democracyj,t 0.004 0.029** 0.004 –0.054** 
 (0.003) (0.004) (0.003) (0.006) 
Capabilitiesi,t 6.170** 6.315* 6.176** 9.877* 
 (1.925) (2.922) (1.924) (4.817) 
Capabilitiesj,t –0.171 –1.003** –0.179 –1.303** 
 (0.169) (0.178) (0.167) (0.281) 
Populationj,t –0.021 –0.141** –0.023 –0.036 
 (0.015) (0.018) (0.015) (0.030) 
PostCold Warj,t –0.059 0.216** –0.052 –0.011 
 (0.035) (0.042) (0.034) (0.064) 
Peacekeepersj,t –0.003 –0.014* –0.003 –0.008 
 (0.005) (0.005) (0.006) (0.009) 
Affinityi,j,t –0.017  –0.018  
 (0.011)  (0.011)  
Alliancei,j,t 0.000*  0.000*  
 (0.000)  (0.000)  
N 322,446 322,446 

Clustered standard errors in parentheses. *p < 0.050, **p < 0.001.

In either case, this finding is interesting given the competing predictions one may generate about troops’ ability to deter (or to cope with) different kinds of mobilizations. Intuitively, troops may be better equipped for dealing with violence than they are with nonviolence. Moreover, it may also be the case that troops themselves increase grievances in a society, inspiring greater opposition. As a result, we might expect ex ante that troops are less able to diminish nonviolent forms of unrest. However, in spite of that intuition, we see that both forms of unrest are observed less frequently in the presence of troops.36 One potential explanation is that foreign troops bolster the local government’s ability to deter initial mobilizations. If initial mobilizations are deterred, it is highly unlikely that collective action will occur (Granovetter 1978; Kuran 1989). An alternative explanation, which is developed further below, is that foreign troops bolster the local government’s commitment and/or ability to enforce contracts and agreements with opposition actors, thus diminishing the negative effects of commitment problems (as discussed above).

In tests omitted for space, we utilized alternative methods for dealing with time, including year fixed effects and including the linear, squared, and cubed counters of the number of years since the last period of conflict. This technique is consistent with research showing that the inclusion of cubic splines in specifications employing time-series cross-section data can approximate a duration model (Carter and Signorino 2010). We also reran Model 1 with undirected dyad fixed effects. To ensure that our selection of instruments was not artificially driving our results, we used alternative codings that included a simple dichotomous indicator of alliance ties, S-scores rather than D-scores of political affinity, and bilateral trade ties. We also ran an interaction between our Cold War dummy and logged troop count. This tests whether there are unique effects in quite different historical periods. The results were highly insignificant, suggesting that the effect of foreign troops remains consistent across Cold War and post–Cold War periods. Finally, we utilized alternative estimators, including a treatment regression and instrumental variables probit. The baseline results hold in all of these tests.

Rule of Law

The baseline estimates provide strong evidence for the unrest-reducing effects of troop deployments. We now test one of the channels through which troops may reduce uncertainty (and unrest): promoting rule of law. We first show that the presence of troops is associated with more robust rule of law. We then confirm that rule of law is itself negatively associated with civil conflict.

Model 13 relies on the same specification as our baseline models, substituting our indicator of rule of law for our measure of conflict. The results (reported in Table 6) show that the presence of troops is strongly associated with more robust rule of law. Recall that our models include directed-dyad fixed effects. As a result, the coefficient on Troop Numbersi,j,t tells us the effect that troops have on rule of law for that dyad, relative to the absence of troops. Substantively, a move across the IQR results in an increase of rule of law of 44 percent (from –0.25 to –0.14).

Table 6.

Effect of troop deployments on rule of law

 (13) (14) 
 Troop   
 numberi,j,t Rule of lawj,t Unrestj,t 
Troop numberi,j,t  1.687**  
  (0.631)  
Rule of lawj,t   –0.019** 
   (0.002) 
GDPi,t 0.030 –1.125** –0.551** 
 (0.042) (0.117) (0.041) 
GDPj,t 0.095* 2.791** 0.079* 
 (0.042) (0.148) (0.038) 
MIDi,t 0.001 –0.015** 0.030** 
 (0.001) (0.002) (0.002) 
Democracyi,t –0.003* –0.005 –0.002 
 (0.001) (0.007) (0.003) 
Democracyj,t 0.006 0.173** 0.016** 
 (0.004) (0.010) (0.005) 
Capabilitiesi,t –2.478 4.066 0.438 
 (2.177) (4.226) (0.450) 
Capabilitiesj,t –0.241* –4.765** 0.664** 
 (0.101) (0.250) (0.161) 
Populationi,t 0.004 –0.418** –0.276** 
 (0.022) (0.043) (0.011) 
Peacekeepersj,t –0.006 0.021* –0.021** 
 (0.004) (0.008) (0.003) 
Unrestj,t-1 0.001 –0.057** 0.461** 
 (0.004) (0.007) (0.004) 
Affinityi,j,t –0.042**   
 (0.012)   
Alliancei,j,t 0.001   
 (0.001)   
Constant   5.431** 
   (0.125) 
N 179,521 189,122 
 (13) (14) 
 Troop   
 numberi,j,t Rule of lawj,t Unrestj,t 
Troop numberi,j,t  1.687**  
  (0.631)  
Rule of lawj,t   –0.019** 
   (0.002) 
GDPi,t 0.030 –1.125** –0.551** 
 (0.042) (0.117) (0.041) 
GDPj,t 0.095* 2.791** 0.079* 
 (0.042) (0.148) (0.038) 
MIDi,t 0.001 –0.015** 0.030** 
 (0.001) (0.002) (0.002) 
Democracyi,t –0.003* –0.005 –0.002 
 (0.001) (0.007) (0.003) 
Democracyj,t 0.006 0.173** 0.016** 
 (0.004) (0.010) (0.005) 
Capabilitiesi,t –2.478 4.066 0.438 
 (2.177) (4.226) (0.450) 
Capabilitiesj,t –0.241* –4.765** 0.664** 
 (0.101) (0.250) (0.161) 
Populationi,t 0.004 –0.418** –0.276** 
 (0.022) (0.043) (0.011) 
Peacekeepersj,t –0.006 0.021* –0.021** 
 (0.004) (0.008) (0.003) 
Unrestj,t-1 0.001 –0.057** 0.461** 
 (0.004) (0.007) (0.004) 
Affinityi,j,t –0.042**   
 (0.012)   
Alliancei,j,t 0.001   
 (0.001)   
Constant   5.431** 
   (0.125) 
N 179,521 189,122 

Clustered standard errors in parentheses. *p < 0.050, **p < 0.001.

We also run a one-stage fixed effects model, identifying a broad correlation between conflict and the rule of law (Model 14 in Table 6). The estimates reveal a significant, negative relationship. Moving across the IQR of Rule of Lawj,t is correlated with a 20 percent reduction in the frequency of civil conflict. Model 10 supports the intuition that rule of law, to the extent that it is buttressed by the presence of foreign troops, decreases the frequency of conflict in host countries.

To be sure, these results are suggestive, and do not represent the final word on the role played by rule of law. Additional work is required to determine whether these increases in rule of law actually attenuate opposition groups’ grievances. However, it is important to note that, since the measure of rule of law itself takes the quality of governance into account, it is unlikely that the relationship we are picking up is simply the ability of governments to repress. In tests included in our replication materials, we explored this possibility at even greater length. One competing explanation for our results is simply that foreign troops allow governments to repress their opposition (by boosting state capabilities). This implication would be observationally equivalent to our findings—that is, less unrest—but for very different reasons. Using common measures of repression taken from the CIRI Human Rights Project and from Fariss (2014), we tested whether troops increase rights violations. The answer is “no”—troops are not simply enabling repression, on average. This finding complements new work on the effects of US troop deployments on human rights practices of host states (Bell, Clay, and Machain 2016). It is also consistent with our story: troops help reduce uncertainty around the credibility problems described above.

Conclusion

This study makes a number of contributions to the literatures on civil conflict and global troop deployments. First, we offer a novel argument regarding an important role of third parties in the occurrence of civil conflict. We argue, specifically, that hosting foreign troops helps states and their latent or manifest opponents navigate and overcome uncertainty regarding the likely outcome of any conflict. Our results support the plausibility of this claim, showing a significant, negative relationship between troop deployments and civil conflict. Second, we employed new bilateral data on foreign troop deployments to calculate the benefits for local governments of hosting foreign troops. Many existing perspectives focus only on the United States, or on the benefits accrued by deploying states. Here, we show that troops contribute positively to host stability. Third, we demonstrated that in addition to a potential deterrent effect, foreign troops help local governments bolster the rule of law, thereby ameliorating both incompatibilities and uncertainty between parties as possible causes of civil conflicts. This points to a specific channel through which foreign states can shape the frequency of conflict ex ante. Finally, our argument and findings contribute to ongoing debates regarding whether or not to maintain an overseas military presence.

This last point is worth particular attention. The continuation of extensive overseas deployments of military personnel fuels an intense debate regarding whether or not to withdraw deployed troops. In the US case, this debate is especially vociferous. On one side, advocates of continued engagement argue that US forces provide international stability by deterring challenges (Brooks, Ikenberry, and Wohlforth 2012) and facilitating rapid response to crises (Art and Cronin 2003). On the other side, opponents suggest that a continued US presence overseas emboldens allies to act aggressively and proactively (Posen 2014). This paper delivers additional evidence to this debate. Specifically, we have demonstrated a non-trivial benefit for local governments associated with the troops they host. Simply put: it appears that hosting foreign troops can, ceteris paribus, reduce the likelihood of hosts experiencing civil conflict.

There remain areas for future research. For example, our final set of tests demonstrates that foreign troops bolster rule of law, which, in turn, reduces the likelihood of conflict. More research is required to help demonstrate precisely how these processes operate. Which elements of rule of law, in particular, benefit from the presence of foreign troops? Is more robust rule of law associated with alternatives to fighting? Does it actually reduce opposition actor grievances?

One additional area for future study would explore the distributed benefits of deployments. Countries cannot deploy to all states with which they enjoy some affinity. Thus, they look to deploy troops to some specific countries with a view toward establishing and maintaining a regional sphere of influence. Sunk costs mean that countries directly hosting troops arguably receive greater commitment from the deploying state. However, it is likely fair to assume that the potentially stabilizing benefits of foreign troops may be enjoyed by nearby countries, too.

Julia Grauvogel is a Research Fellow at the GIGA German Institute of Global and Area Studies. Her work focuses on domestic opposition movements in countries under sanctions and legitimation strategies of authoritarian regimes.

Amanda A. Licht is an Assistant Professor at the University of Binghamton (SUNY) whose work focuses on how domestic politics affects the ability of powerful states to exert influence over others.

Christian von Soest is a Lead Research Fellow and Head of the Peace and Security Research Program at the GIGA German Institute of Global and Area Studies. His research investigates sanctions and other foreign policy interventions, particularly when targeted at authoritarian regimes.

1An earlier version of this paper was presented at the 2015 Kobe Sakura Meeting. We thank the Graduate School of Law (Kobe University), the CROP-IT Project (JSPS), the Suntory Foundation, and the Daiwa Anglo-Japanese Foundation for their generous support of this meeting. We are grateful to Erik Gartzke, Atsushi Tago, and other participants at the Sakura meeting for their comments. We also thank Cameron Thies and three anonymous reviewers for their thoughtful suggestions. We would also like to thank Niheer Dasandi for assistance in collecting the troop deployment data and Paul Bezerra for assistance in mapping these data. Finally, we wish to thank Stephen Chaudoin, Andrew Kerner, and Slava Mikhaylov for comments on early drafts. All remaining errors are our own.
2Deployments are sometimes controversial for the deploying country, particularly in times of crisis. Public opinion research shows that sensitivity to causalities and, related, an unwillingness to intervene in foreign affairs both create resistance to troop deployments. In host countries, the presence of foreign troops can be a political grievance in its own right.
3According to NAVCO 2.0 (Chenoweth and Lewis 2013).
4For comprehensive reviews, see Collier and Hoeffler (2007); Blattman and Miguel (2009); Chenoweth and Ulfelder forthcoming.
5Walter (2009) offers a detailed discussion of this logic in the context of civil conflicts.
6The next section details the growing literature on the costs and benefits associated with hosting foreign military troops during peacetime.
7To be sure, an alternative interpretation is that the presence of troops can raise the stakes of political rivalries. The Cold War could be viewed as a relatively unstable period given the overarching fears of war. Our point is not that troops necessarily alleviate all political tensions. However, they prevent those tensions from escalating to conflict.
8See Schelling (1960), Blainey (1973), Fearon (1995), and Gartzke (1999) for discussions in the context of inter-state conflicts.
9Walter (2009) argues that these alternatives do exist and are often employed. For example, Canada extended autonomous rights to Quebec and Macedonia to ethnic Albanians. In both instances, large-scale conflicts were averted.
10As noted earlier, our interest here parallels that of much of the foreign aid literature, which assess third-party interventions and actions outside the context of conflict (see, e.g., Alesina and Dollar 2000).
11Of course, troops may be deployed for a variety of reasons. However, we contend that other benefits can be realized only once internal stability has been achieved in the host country.
12Tying hands refers here to taking actions that increase costs associated with backing down or not following through with promises or commitments.
13Deployments are costly signals because of the costs of maintaining foreign deployments and because of the sometimes politically contentious nature of deployments for the sending government. For example, public opinion polling in the United States has consistently reflected a strong streak of isolationism among respondents. Indeed, a Pew poll in late 2013 identified a 50-year low in public support for an active US foreign policy (Fisher 2013); 52 percent stated agreement with the statement “the U.S. should mind its own business internationally and let other countries get along the best they can on their own,” 48 percent of respondents agreed with this statement in 1974, and 41 percent in 1994.
14In this respect, our work perhaps borrows more closely from that on economic sanctions as a foreign policy tool (see, e.g., Hufbauer, Schott, and Elliott 1990; Lektzian and Sprecher 2007) than it does that on kinetic uses of military force.
15Again, we stress that there is a theoretical distinction between strong governments, who may simply be able to repress opposition groups, and rule of law, which relates to the quality of institutions. Rule of law, in fact, can be thought of as a form of check on precisely the kind of repression that can mobilize opposition groups.
16Note that we distinguish foreign troops deployments from traditional peacekeeping missions in our analysis.
17A monadic construction loses the partner-specific information likely to play a key role in motivating troop deployments and driving their effects.
18Data on unrest (as well as its predominant tactics and characteristics) comes from NAVCO 2.0 (Chenoweth and Lewis 2013).
19Data on rule of law is somewhat scant, particularly in terms of temporal span. There are several alternatives to the World Bank data, but with far less coverage. WGI provides to our knowledge the most detailed information for the largest set of countries over the longest period of time.
20We add one to all zero values of troop deployments before logging the measure.
21The measure is also more appropriate for our two-stage approach, which requires a continuous endogenous regressor.
22Top ten trading partners are calculated using trade statistics from the World Integration Trade Solution hosted by the World Bank. WITS provides the comprehensive source of bilateral trade data. Available at wits.worldbank.org, last accessed June 6, 2015.
23These measures are taken from the World Bank’s World Development Indicators and recorded as the log of constant year 2000 US dollars.
24The results do not change when looking at whether the host country is involved in an MID in any fashion, as opposed to just being targeted.
25The results are robust to measures of ethnic power relations, a common proxy for the presence of social and/or political grievances in a particular state.
26Controlling separately for capabilities also helps us later identify the independent effect of rule of law, ensuring that rule of law is not conflated simply with state strength.
27Note that the results are robust to alternative fixed effects (see below).
28There are alternative estimators available, but each of them comes with limitations. As we note below, we get comparable results when running an instrumental variables probit and a treatment regression. However, these estimators make the use of fixed effects more difficult (or impossible) and we therefore have less confidence that the models control sufficiently for the unobservables likely to affect our outcomes of interest. In addition, the treatment regression requires a dichotomous endogenous regressor, prohibiting us from looking at the size of troop deployments.
29We include fixed effects for directed dyads rather than undirected dyads because the identity of country i (sender) and j (host) is not arbitrary. A United States–Iraq dyad is not as equally likely to see a deployment as an Iraq–United States dyad. This is because we assign the role of sender to country i and host to country j, and our indicators of unrest are related only to the host.
30There are competing views on the use of lagged dependent variables. We note below that our results do not change when using the most common alternatives—year fixed effects or the linear, squared, and cubed count of years since the last period of conflict (Carter and Signorino 2010).
31Our baseline model is robust to the inclusion of trade ties between states as an alternative instrument to political affinity.
32Note that these estimates are not a likelihood. Instead, they are best understood as a measure of the frequency of civil conflict.
33Predicted conflict in the absence of troops is 0.21. In the presence of troops, it falls to 0.13. The sample mean is 0.17.
34This result is unsurprising given that nearly two-thirds of the deployments in the sample are to countries at peace.
35The reference categories in these tests include years of no conflict. An alternative approach is restricting the sample of conflict years and measuring whether troop deployments make one form more or less likely than the other, given that conflict occurs. We ran a separate test with this sampling restriction and found no significant difference in the effects that troops have on the form of mobilization.
36Like our baseline models, these tests are robust to the inclusion of alternative instruments, as well as varied codings of troop deployments.

References

Acemoglu
Daron
Robinson
James A
2006
.
Economic Origins of Dictatorship and Democracy
 .
Cambridge
:
Cambridge University Press
.
Alesina
Alberto
Dollar
David
.
2000
. “
Who Gives Foreign Aid to Whom and Why?
Journal of Economic Growth
 
5
:
33
63
.
Allen
Michael A.
Flynn
Michael E.
.
2013
. “
Putting Our Best Boots Forward: US Military Deployments and Host-Country Crime
.”
Conflict Management and Peace Science
 
30
:
263
85
.
Allen
Michael A.
VanDusky-Allen
Julie
Flynn
Michael E.
.
forthcoming
. “
The Localized and Spatial Effects of US Troop Deployments on Host-State Defense Spending
.”
Foreign Policy Analysis
 , in press.
Art
Robert J.
Cronin
Patrick M
2003
.
The United States and Coercive Diplomacy
 .
US Institute of Peace Press
,
Washington, DC
.
Balch-Lindsay
Dylan
Enterline
Andrew
Joyce
Kyle A.
.
2008
. “
A Competing Risks Approach to Third Parties and the Civil War Process
.”
Journal of Peace Research
 
45
:
345
63
.
Beardsley
Kyle.
2011
. “
Peacekeeping and the Contagion of Armed Conflict
.”
Journal of Politics
 
73
:
1051
64
.
Bell
Sam R.
Chad Clay
K.
Martinez Machain
Carla
.
2016
. “
The Effect of US Troop Deployments on Human Rights
.”
Journal of Conflict Resolution
  in press.
Biglaiser
Glen
DeRouen
Karl
Jr.
2007
. “
Following the Flag: Troop Deployment and US Foreign Direct Investment
.”
International Studies Quarterly
 
51
:
835
54
.
Biglaiser
Glen
DeRouen
Karl
Jr.
.
2009
. “
The Interdependence of US Troop Deployments and Trade in the Developing World
.”
Foreign Policy Analysis
 
5
:
247
63
.
Blainey
Geoffrey.
1973
.
The Causes of War
 .
New York
:
Free Press
.
Blattman
Christopher
Miguel
Edward
.
2010
. “
Civil War
.”
Journal of Economic Literature
 
48
:
3
57
.
Blomberg
S. Brock
Hess
Gregory D.
.
2002
. “
The Temporal Links Between Conflict and Economic Activity
.”
Journal of Conflict Resolution
 
46
:
74
90
.
Blomberg
S. Brock
Hess
Gregory D.
Thacker
Siddharth
.
2006
. “
On The Conflict-Poverty Nexus
.”
Economics & Politics
 
18
:
237
67
.
Boix
Charles.
2003
.
Democracy and Redistribution
 .
Cambridge
:
Cambridge University Press
.
Braithwaite
Alex.
2015
. “
Transnational Terrorism as an Unintended Consequence of a Military Footprint
.”
Security Studies
 
24
:
349
75
.
Braithwaite
Alex
Dasandi
Niheer
Hudson
David
.
2016
. “
Does Poverty Cause Conflict? Isolating the Causal Origins of the Conflict Trap
.”
Conflict Management and Peace Science
 
33
:
45
66
.
Brinkman
Henk-Jan.
2001
.
Preventing Civil Strife: An Important Role for Economic Policy
 .
New York
:
United Nations Press
.
British Broadcasting Corporation
.
2015
. “US Military to Close 15 Bases in Europe.” Accessed April 30, 2016. http://www.bbc.com/news/world-us-canada-30731926.
Brooks
Stephen G.
Ikenberry
G. John
Wohlforth
William C.
.
2012
.
Don’t Come Home, America: The Case against Retrenchment
 .
Cambridge, MA
:
MIT Press
.
Buhaug
Halvard
Skrede Gleditsch
Kristian
2008
. “
Contagion or Confusion? Why Conflicts Cluster in Space
.”
International Studies Quarterly
 
52
:
215
33
.
Carter
David B.
Signorino
Curtis S.
.
2010
. “
Back to the Future: Modeling Time Dependence in Binary Data
.”
Political Analysis
 
18
:
271
92
.
Cederman
Lars-Erik
Weidmann
Nils B.
Skrede Gleditsch
Kristian
.
2011
. “
Horizontal Inequalities and Ethnonationalist Civil War: A Global Comparison
.”
American Political Science Review
 
105
:
478
95
.
Checkel
Jeffrey.
2013
.
Transnational Dynamics of Civil War
 .
New York
:
Cambridge University Press
.
Cheibub
Jose Antonio
Gandhi
Jennifer
Raymond Vreeland
James
.
2010
. “
Democracy and Dictatorship Revisited
.”
Public Choice
 
143
:
67
101
.
Chenoweth
Erica
Lewis
Orion
.
2013
. “
Unpacking Nonviolent Campaigns: Introducing the NAVCO 2.0 Dataset
.”
Journal of Peace Research
 
50
:
415
23
.
Chenoweth
Erica
Stephan
Maria J.
.
2011
.
Why Civil Resistance Works: The Strategic Logic of Nonviolent Conflict
 .
New York
:
Columbia University Press
.
Chenoweth
Erica
Ulfelder
Jay
forthcoming
. “
Can Structural Conditions Explain the Onset of Nonviolent Uprisings?
Journal of Conflict Resolution
 , in press.
Collier
Paul
Hoeffler
Anke
.
2002
. “
On the Incidence of Civil War in Africa
.”
Journal of Conflict Resolution
 
46
:
13
28
.
Collier
Paul
Hoeffler
Anke
.
2007
. “
Civil War
.”
Handbook of Defense Economics
 
2
:
711
39
.
Collier
Paul
Hoeffler
Anke
Soderbom
Mans
.
2004
. “
On the Duration of Civil War
.”
Journal of Peace Research
 
41
:
253
73
.
de Figueiredo
Rui J. P.
Jr.
Weingast
Barry R
1997
. “
Rationality of Fear: Political Opportunism and Ethnic Conflict
.”
Working Paper, Institute for War and Peace Studies
 .
Diehl
Paul F
Regan
Patrick
.
2015
. “
The Interdependence of Conflict Management Attempts
.”
Conflict Management and Peace Science
 
32
:
99
107
.
D’Orazio
Vito.
2015
. “
Advancing Measurement of Foreign Policy Similarity.”
  Working Paper,
University of Texas at Dallas
.
Dufour
Jules.
2016
. “The Worldwide Network of US Military Bases.” Accessed April 30, 2016. http://www.globalresearch.ca/the-worldwide-network-of-us-military-bases/5564.
Fariss
Christopher J.
2014
. “
Respect for Human Rights Has Improved over Time: Modeling the Changing Standard of Accountability
.”
American Political Science Review
 
108
:
297
318
.
Fearon
James D.
1995
. “
Rationalist Explanations for War
.”
International Organization
 
49
(
3
):
379
414
.
Fearon
James D.
.
1997
. “
Signaling Foreign Policy Interests: Tying Hands Versus Sinking Costs
.”
Journal of Conflict Resolution
 
41
:
68
90
.
Fearon
James D.
.
2007
.
“Fighting Rather Than Bargaining.”
 
Paper presented at the Annual Meeting of the American Political Science Association
.
Fearon
James D.
Laitin
David D.
.
2003
. “
Ethnicity, Insurgency, and Civil War
.”
American Political Science Review
 
97
:
75
90
.
Fearon
James D.
Humphreys
Macartan
Weinstein
Jeremy M.
.
2009
. “
Can Development Aid Contribute to Social Cohesion After Civil War? Evidence from a Field Experiment in Post-Conflict Liberia
.”
American Economic Review
 
99
:
287
91
.
Fisher
Max.
2013
. “
American Isolationism Just Hit a 50-Year High: Why That Matters
.”
Washington Post
 .
Washington, DC
., December 3.
Fortna
Virginia Page.
2004
. “
Interstate Peacekeeping: Causal Mechanisms and Empirical Effects
.”
World Politics
 
56
(
4
)
: 481
519
.
Gartzke
Erik.
1999
. “
War Is in the Error Term
.”
International Organization
 
53
:
567
87
.
Gartzke
Erik
Kagatoni
Koji
.
2015
.
“Trust in Tripwires: Deployments, Costly Signaling and Extended General Deterrence.”
  Working Paper,
University of California at San Diego
.
Gleditsch
Kristian S.
2007
. “
Transnational Dimensions of Civil War
.”
Journal of Peace Research
 
44
:
293
309
.
Granovetter
Mark.
1978
. “
Threshold Models of Collective Behavior
.”
American Journal of Sociology
 
83
:
1420
43
.
Greig
J. Michael
Diehl
Paul F.
.
2005
. “
The Peacekeeping-Peacemaking Dilemma
.”
International Studies Quarterly
 
49
:
621
46
.
Hegre
Havard
Ellingsen
Tanja
Gates
Scott
Petter Gleditsch
Nils
2001
. “
Toward a Democratic Civil Peace? Democracy, Political Change, and Civil War, 1816–1992
.”
American Political Science Review
 
95
:
33
48
.
Hegre
Havard
Sambanis
Nicholas
.
2006
. “
Sensitivity Analysis of Empirical Results on Civil War Onset
.”
Journal of Conflict Resolution
 
50
:
508
35
.
Hirshleifer
Jack.
2001
.
The Dark Side of the Force: Economic Foundations of Conflict Theory
 .
New York
:
Cambridge University Press
.
Hufbauer
Gary Clyde
Schott
Jeffrey J.
Ann Elliott
Kimberly
1990
.
Economic Sanctions Reconsidered: History and Current Policy
 , vol.
1
.
New York
:
Peterson Institute
.
Huth
Paul K.
1988a
. “
Extended Deterrence and the Outbreak of War
.”
American Political Science Review
 
82
:
423
43
.
Huth
Paul K.
.
1988b
.
Extended Deterrence and the Prevention of War
 .
New Haven, CT
:
Yale University Press
.
Jones
Garett
Kane
Tim J.
2007
. “US Troops and Economic Growth.” Working Paper, George Mason University.
Jones
Garett
Kane
Tim J.
.
2012
. “
US Troops and Foreign Economic Growth
.”
Defence and Peace Economics
 
23
:
225
49
.
Kennan
John
Wilson
Robert
1993
. “
Bargaining with Private Information
.”
Journal of Economic Literature
 
31
:
45
104
.
Kuran
Timur.
1989
. “
Sparks and Prairie Fires: A Theory of Unanticipated Political Revolution
.”
Public Choice
 
61
:
41
74
.
Lektzian
David J.
Sprecher
Christopher M.
.
2007
. “
Sanctions, Signals, and Militarized Conflict
.”
American Journal of Political Science
 
51
:
415
31
.
Licklider
Roy.
1995
. “
The Consequences of Negotiated Settlements in Civil Wars, 1945–1993
.”
American Political Science Review
 
89
:
681
90
.
Little
A.
LeBlang
D.
2004
.
“Military Securities: Financial Flows and the Deployment of US Troops.”
 
Paper presented at the Annual Meeting of the American Political Science Association
, September.
Lutz
Catherine.
2009
. “US Foreign Military Bases: The Edge and Essence of Empire.” Working Paper.
Machain
Carla Martinez
Clifton Morgan
T.
.
2013
. “
The Effect of US Troop Deployment on Host States’ Foreign Policy
.”
Armed Forces & Society
 
39
:
102
23
.
Maves
Jessica
Braithwaite
Alex
.
2013
. “
Autocratic Institutions and Civil Conflict Contagion
.”
Journal of Politics
 
75
:
479
90
.
Miguel
Edward
Satyanath
Shanker
Sergenti
Ernest
.
2004
. “
Economic Shocks and Civil Conflict: An Instrumental Variables Approach
.”
Journal of Political Economy
 
112
:
725
53
.
Murdoch
James C.
Sandler
Todd
.
2002
. “
Economic Growth, Civil Wars, and Spatial Spillovers
.”
Journal of Conflict Resolution
 
46
:
91
110
.
O’Kane
Rosemary H. T.
1993
. “
Coups D’etat in Africa: A Political Economy Approach
.”
Journal of Peace Research
 
30
:
251
70
.
Ostby
Gudrun.
2008
. “
Polarization, Horizontal Inequalities and Violent Civil Conflict
.”
Journal of Peace Research
 
45
:
143
62
.
Palmer
Glenn
D’Orazio
Vito
Kenwick
Michael
Lane
Matthew
.
2015
. “
The MID4 Dataset, 2002–2010: Procedures, Coding Rules and Description
.”
Conflict Management and Peace Science
 
32
:
222
42
.
Pillar
Paul.
1983
.
Negotiating Peace: War Termination as a Bargaining Process
 .
Princeton, NJ
:
Princeton University Press
.
Posen
Barry R.
2014
.
Restraint: A New Foundation for US Grand Strategy
 .
Ithaca, NY
:
Cornell University Press
.
Regan
Patrick M.
2002
. “
Third-Party Interventions and the Duration of Intrastate Conflicts
.”
Journal of Conflict Resolution
 
46
:
55
73
.
Ross
Michael L.
2004
. “
What Do We Know About Natural Resources and Civil War?
Journal of Peace Research
 
41
:
337
56
.
Schelling
Thomas C.
1960
.
The Strategy of Conflict
 .
Cambridge, MA
:
Harvard University Press
.
Schmidt
Sebastian.
2014
. “
Foreign Military Presence and the Changing Practice of Sovereignty: A Pragmatist Explanation of Norm Change
.”
American Political Science Review
 
108
:
817
29
.
Skaperdas
Stergios.
2008
. “
An Economic Approach to Analyzing Civil Wars
.”
Economics of Governance
 
9
:
25
44
.
Slantchev
Branislav.
2005
. “
Military Coercion in Interstate Crises
.”
American Political Science Review
 
99
:
533
47
.
Voeten
Erik
Merdzanovic
Adis
2009
.
“United Nations General Assembly Voting Data.”
 
Washington, DC
:
Georgetown University
. http://dvn.iq.harvard.edu/dvn/dv/Voeten/faces/study/StudyPage.xhtml.
Walter
Barbara F.
1997
. “
The Critical Barrier to Civil War Settlement
.”
International Organization
 
51
:
335
64
.
Walter
Barbara F.
.
2009
. “
Bargaining Failures and Civil War
.”
Annual Review of Political Science
 
12
:
243
61
.
Wooldridge
Jeffrey M.
2010
.
Econometric Analysis of Cross Section and Panel Data
 .
Cambridge, MA
:
MIT Press
.
Yeo
Andrew.
2011
.
Activists, Alliances, and Anti-US Base Protests
 .
New York
:
Cambridge University Press
.