Diversity and Diversion: How Ethnic Composition Affects Diversionary Conflict

How does a state’s ethnic composition affect its propensity to engage in diversionary conflicts? Recent empirical work examines the political conditions under which domestic unrest compels an embattled leader to initiate conflict abroad. We remain uncertain, however, of what social or demographic characteristics make states particularly prone to diversionary behavior. This article tries to address this gap, examining whether a state’s ethnic structure conditions its leader’s response to domestic discontent. Combining the expansive literatures on ethnic politics and diversionary war yields conflicting expectations here. I find that ethnically fragmented states are significantly more prone to initiating diversionary conflicts, and I show that the greater availability of “conflict opportunities” resulting from transborder ethnic-kin groups, in part, drives this effect. A brief case study illustrates these dynamics.

Russia's 2014 annexation of the Crimean Peninsula and support for separatist movements in the Eastern Ukraine served important strategic purposes for Russian foreign policy.When the Ukrainian revolution threatened Russia's interests in its near abroad, Russian President Vladimir Putin moved swiftly to secure Russian naval facilities in Crimea and bolster pro-Russian forces across the Ukrainian border.But many observers also attributed Russian actions in part to Putin's domestic imperatives (Marten 2014).Putin's public rhetoric began to stoke nationalist sentiment, referring to a broader, ethnically based Novorossiya (New Russia) and highlighting Russia's historical and ethnic ties to the Eastern Ukraine (Clem 2014).His domestic popularity skyrocketed.Although his government was already reasonably secure, an attempt to bolster the regime in Moscow seemed to be part of what drove Putin's strategy in Ukraine (Greene and Robertson 2014).
The ethnic overlap between Russia and Ukraine, and the presence of large ethnic Russian populations within Ukraine, was absolutely vital to this strategy.Putin justified Russian intervention primarily on the grounds of protecting ethnic Russians from the supposedly hostile government in Kiev.Highlighting and glorifying ethnic Russian identity also constituted a core component of the subsequent spike in Putin's domestic approval rating.These events illustrate the potential importance of ethnic politics and transnational ethnic ties in the prosecution of diversionary conflicts.Embattled leaders, or those otherwise looking to conflict abroad as a means of securing power at home, can often exploit ethnic cleavages domestically and strategically target states that are ethnically linked to their own population.Cross-border ethnic groups, in particular, can provide a fertile source of latent conflicts that a troubled leader can seize upon and escalate when domestic imperatives require.This article systematically examines these dynamics.
An enormous amount of scholarly work examines the conditions under which domestic unrest pushes leaders to initiate conflict abroad.But the expansive literature on diversionary conflict focuses almost entirely on the domestic political or institutional conditions that affect diversion.It largely ignores the social, cultural, and demographic characteristics that render states more or less prone to diversionary behavior.This is highly problematic.Domestic and transnational demographic structures can facilitate a leader's initiation of diversionary conflict in response to domestic unrest.If diversionary conflict essentially aims to quell social unrest, then we should expect domestic social and demographic factors to condition the probability of diversionary behavior.International conflict may be more likely to bolster domestic political loyalties in states with certain types of populations.Leaders of such states should then be more likely to resort to diversionary conflict in response to domestic turmoil.The vast literature on diversionary war largely ignores this possibility.
I address this gap and examine how a state's ethnic composition and transnational ethnic ties affect the likelihood that its leaders will respond to domestic unrest by initiating conflict abroad.I assess three causal pathways that could link ethnic composition to diversionary behavior.First, ethnic composition could affect the probability that a state's population will "rally around the flag" in response to external conflict.If ethnic diversity impedes such unification, diversionary conflict should prove less likely in ethnically fractured states (Coser 1956, 93-95;Levy 1989, 667).Second, ethnically divided states may be more prone to diversion because they are more internally unstable and their leaders thus face stronger incentives to resort to risky diversionary conflicts in response to domestic unrest.Third, ethnically fragmented states could be more given to diversionary behavior because the existence of numerous transborder ethnic-kin groups provides a ready source of "conflict opportunities" for embattled leaders to exploit when seeking to divert.I find strong support for the third argument.Below, I show that diversionary-conflict propensities are significantly stronger in ethnically fragmented states.Furthermore, diversionary conflicts disproportionately target states with ethnic links to the initiating state.Generally, these results demonstrate that social and demographic conditions significantly affect the likelihood of domestic unrest translating into diversionary conflict.More specifically, they show that ethnic fragmentation at home and transnational ethnic links abroad can significantly increase a state's tendency toward diversionary behavior.These findings support the notion that diversion requires the availability of pre-existing "conflict opportunities" that leaders can escalate at their convenience.I expand upon this nexus between ethnic politics and diversionary conflict with a brief illustrative case study of the 1998 dispute between Turkey and Syria.
These findings point toward several policy solutions to help prevent diversionary conflict.First, because ethnically fragmented states are disproportionately likely to respond to domestic unrest by lashing out abroad, neighbors of such fractionalized states must be particularly cautious when domestic unrest arises.Second, because diversionary conflict disproportionately targets ethnically linked states, the leaders of such states must be especially proactive in heading off diversionary action.In this sense, a clear appreciation of how ethnic politics interact with the dynamics of diversionary conflict can help leaders anticipate and defuse such disputes.

Existing Literature
This section examines and critiques the literature on diversionary conflict.I point out the surprising lack of empirical work examining the social and demographic conditions affecting the probability of diversionary conflict.I then discuss the literature on ethnic composition and domestic instability and highlight its relevance to diversionary conflict.

Diversionary Conflict
The early statistical literature on diversionary conflict proved notoriously inconclusive, and the topic remains the subject of much dispute (Ostrom and Job 1986;James and Oneal 1991;Morgan and Bickers 1992;Levy and Vakili 1992).Early quantitative studies found limited support for diversionary-war theory despite its compelling logical foundations and several supportive case studies.More recent work explains these mixed early findings by arguing that diversionary conflict is a highly conditional phenomenon.Domestic unrest translates into incentives for international conflict only under specific circumstances.Absent such conditions, domestic unrest may have no effect on conflict propensities or may even incline toward peace.
For example, many argue that strategic interaction and the behavior of potential adversaries can prevent embattled leaders from initiating diversionary conflicts.More specifically, likely targets of diversion can adopt a policy of "strategic conflict avoidance" and effectively deprive a troubled leader of any excuse for initiating a dispute (Smith 1996;Leeds and Davis 1997;Clark 2003;Chiozza and Goemans 2004;Fordham 2005).By strategically making concessions to vulnerable leaders, likely targets can defuse a potentially dangerous conflict before it begins.This strategic interaction is often invoked to explain the spotty empirical support for diversionary-conflict theory (Leeds and Davis 1997).
More broadly, this literature holds that embattled leaders require adequate "conflict opportunities" in order for a diversionary conflict to occur (Mitchell and Prins 2004;Mitchell and Thyne 2010).The argument essentially claims that diversion requires salient, if latent, conflicts that leaders can seize upon and escalate when they choose.Leaders who disingenuously provoke a conflict where none existed previously will appear unnecessarily aggressive.They thereby risk further punishment from their constituents.As such, when casting about for a potentially fruitful diversionary conflict, leaders of states that are engaged in more ongoing disputes enjoy a greater number of escalation opportunities.More conflict opportunities should then increase the probability that domestic unrest results in international conflict.
The theoretical logic behind the strategic conflict avoidance and conflict-opportunity theory is extremely compelling, but future statistical work must more thoroughly flesh out its varying effects on dispute initiation and subsequent conflict escalation.Although strategic interaction can prevent diversionary conflicts from escalating to higher levels of violence, it may be less effective in preventing embattled leaders from initiating the dispute in the first place.Quantitative work on the subject must directly model conflict opportunities and more clearly differentiate between dispute initiation and conflict escalation (Meernik 1994).The tests below begin this process, offering a clearer picture of how conflict opportunities affect the initiation, rather than escalation, of diversionary conflict.
In addition to the behavior of potential targets, a vast literature details the types of domestic institutional structures that either provide leaders with the strongest incentives to initiate diversionary conflicts or allow them the autonomy to use force for personal ends.Numerous studies assess the effects of regime type and whether democratic or authoritarian states are more likely to divert (Gelpi 1997;Miller 1999;Pickering andKisangani 2005, 2010).Others examine more specific institutional features, such as regime consolidation (Kisangani and Pickering 2009), executive accountability (Brule and Williams 2009;Johnson and Barnes 2011), or state extractive capacity (Oakes 2012).These studies significantly enhance our understanding of the domestic institutional conditions that promote or impede diversionary conflict.
Despite the enormous, cumulative contributions of this work, I argue that the literature to date entirely neglects an important class of factors that might also condition the probability of diversionary conflict.More specifically, the existing literature ignores domestic social or demographic conditions.We remain unsure of the degree to which a state's broader social, cultural, or demographic composition affects the propensity of its leaders to respond to unrest with diversionary force. 1  This oversight is highly problematic.Diversionary conflict, at the most basic level, aims to strengthen a leader's hold on power by either demonstrating competence to key constituencies (Richards et al. 1993;Downs and 1 These underlying social factors are largely invariant over time and may thus fail to explain the timing of diversionary conflict.The argument here thus focuses on how ethnic structure affects a leader's response to exogenous domestic unrest, not how ethnic structure determines the probability of unrest in the first place.Rocke 1994;Goemans 2000) or generating a "rally 'round the flag" effect within the population (Levy 1989).The ultimate goal of diversionary conflict is to create a broad reaction (i.e., perceptions of competence or increased nationalism) among large swathes of the population.The demographic makeup of this population might affect the probability that it will respond to international conflict by rallying around the flag or updating their perceptions of the leader's competence.If diversionary conflict ultimately seeks to provoke a broad, mass-level reaction, it is quite plausible that broad, mass-level social factors would condition how the population responds to crises and, thus, a leader's propensity to divert.Certain types of populations may be more likely to respond favorably to diversionary conflicts.Leaders of these states should then be most likely to respond to domestic unrest by initiating conflict abroad.

Ethnic Composition and Diversionary Conflict
I examine the effect of one specific social factor-ethnic composition-on the probability that domestic unrest will result in diversionary conflict.Recent research on ethnic politics tears down "primordialist" notions of ethnicity as fixed and exogenous, highlighting the importance of political and social context, agency, and practice in constructing ethnic identities and cleavages (Arfi 1998;Giuliano 2000;Kurien 2002;Yashar 2005;Chandra 2012).But once constructed, ethnic identities can create an extremely salient and effective source of political mobilization and collective action, with massive implications for patterns of social cohesion and conflict (Olzak 1992;Toft 2003).
Despite a great deal of work on the topic, no consensus exists regarding the effect of ethnic structure and the salience of ethnic identities on regime stability and the durability of a state's democratic institutions.Early work on the subject largely held that ethnic identities are almost universally harmful to democracy and domestic stability because they create static and unbridgeable internal cleavages (Dahl 1971;Rabushka and Shepsle 1972;Lijphart 1977;Horowitz 1985;Reilly 2001).But more recent work pushes back on this claim, showing how ethnic divisions and ethnically based parties can be structured to actually enhance democratic stability (Chandra 2005;Beissinger 2008;Schneider and Wiesehomeier 2008).Relatedly, ethnic fragmentation was shown to impede economic growth in developing states, though the relationship is much more complicated than initially thought (Easterly and Levine 1997;Alesina et al. 2003).
But the majority of work on ethnic politics focuses on how ethnic structure affects the probability of civil conflict.The relationship remains highly contested (Collier and Hoeffler 1998;Fearon and Laitin 2003;Cederman and Girardin 2007;Fearon, Kasara, and Laitin 2007).Many studies examined the conflict-inducing effect of ethnic fractionalization, or the degree to which a state is fragmented into a large number of discrete, and potentially competing, ethnic groups (Ellingsen 2000;Elbadawi and Sambanis 2002).2Scholars thought that more competing groups would increase the number of ethnic cleavages and create more groups that would likely become disaffected if excluded from power.Compelling anecdotal evidence led many to believe that ethnic fragmentation would significantly increase the probability of a state experiencing civil war or serious domestic political instability.
The statistical literature, however, proved somewhat inconclusive.Different specifications of the crucial variables produce different results.Findings of significant relationships between ethnic fractionalization and conflict are typically weak and rather fragile (Fearon and Laitin 2003;Fearon, Kasara, and Laitin 2007).Important critiques of the fractionalization literature point out that states with widely varying ethnic compositions can produce comparable fractionalization scores, while others highlight the problems of using temporally invariant codings of ethnic groups (Cederman and Girardin 2007).Related work effectively argues that the theoretical mechanisms linking fractionalization to conflict are underdeveloped and unconvincing (Reilly 2000;Blimes 2006).In response, many scholars assert a nonlinear, inverse-U-shaped relationship between fragmentation and civil conflict (Collier and Hoeffler 1998).According to this school of thought, states with moderate levels of fractionalization are most likely to experience civil conflict.In highly cohesive states, ethnicity is simply not a source of conflict.And in highly fragmented states, the competing ethnic groups compose too small a proportion of the population to make an effective claim for national leadership.
This led directly to important work demonstrating an impressively robust association between ethnic polarization and civil war (Reynal-Querol 2002;Montalvo and Reynal-Querol 2005;Esteban and Ray 2008).Polarization captures the degree to which a population is evenly divided between two competing ethnic groups.In highly polarized states, both ethnic groups are large enough to have a reasonable claim to control of the state, and can amass significant popular support for these claims.If excluded from power, either group will feel highly aggrieved and be able to marshal significant social and economic resources in airing its grievances and seeking redress.The theoretical logic linking polarization to conflict is more convincing, and it receives robust statistical support.Although most quantitative empirical work on ethnic politics focuses on the impact of fractionalization, it appears that ethnic polarization is more important in determining patterns of domestic instability and social cohesion.
But how, exactly, do these literatures speak to one another?Jaroslav Tir and Michael Jasinski (2008) convincingly show that embattled leaders can exploit ethnic cleavages for their own political survival by using diversionary force to repress domestic ethnic minorities.And importantly, domestic repression along ethnic lines may often supplement, rather than supplant, a diversionary strategy abroad (Enterline and Gleditsch 2000).Scholars, however, have yet to thoroughly examine the effect of domestic ethnic structure on international diversionary conflict propensities.
The following section develops several hypotheses on how both domestic and transnational ethnic structures affect diversionary conflict dynamics.This discussion ties together the disparate literatures on ethnic politics and the domestic sources of international conflict.I show how a state's internal ethnic composition and the presence of transborder ethnic populations condition the propensity of leaders to initiate conflict abroad in response to domestic unrest.These results reinforce earlier work finding diversionary conflict to be a highly conditional phenomenon.The interaction between ethnic politics and diversionary conflict might also help explain the noisy empirical relationship between ethnic fragmentation and internal conflict.Further work is necessary on the tradeoffs and complementarities between domestic repression and international diversion as responses to domestic unrest in ethnically fragmented states.But the theoretical discussion and statistical results presented here offer a useful first cut at the complex interactions between ethnic structure, domestic politics, and international conflict.

Hypotheses
This section lays out the ways in which ethnic composition might affect a state's diversionary conflict propensities.I flesh out several distinct theoretical logics and derive a series of competing hypotheses.But this first requires a brief discussion of the various ethnic structures discussed above.
First, ethnic fragmentation refers to the degree to which a state is divided among a large number of discrete, and potentially competing, ethnic groups.Highly fragmented states encompass a large number of distinct groups, none of which compose a disproportionately large segment of the population.Ethnic polarization, on the other hand, captures the degree to which a state is evenly divided between two large ethnic groups that jointly comprise the bulk of the population. 3Ethnically polarized states essentially represent a bipolar distribution of ethnic identities.The previous section described the literature on ethnic structure and domestic conflict, arguing that ethnically polarized states are most prone to instability.But how might these different ethnic structures affect diversionary-conflict propensities?
I argue that a state's ethnic composition could affect diversionary conflict propensities in numerous and multifarious ways.First, higher levels of both ethnic fractionalization and polarization could decrease the probability of diversionary conflict by impeding the generation of a rally effect among a state's population.Alternatively, ethnic fragmentation could increase the probability of diversionary conflict by creating more interstate conflict opportunities upon which leaders can seize when looking to divert.Ethnic polarization could also increase the probability of diversion by creating an unstable domestic political situation in which troubled leaders are forced to respond quickly and decisively to any signs of domestic unrest.This might lead them to initiate risky diversionary conflicts that more secure leaders could avoid.Additionally, leaders of ethnically polarized states might have faith in the ability of their large ethnic in-group to sustain them in power even after alienating ethnic outgroups through the pursuit of domestic repression and/ or diversionary conflict against those groups' foreign allies.Below, I flesh out these logics in greater detail and specify a series of testable hypotheses.

Ethnic Diversity and Social Cohesion
The early diversionary-war literature relied on socialpsychological theories of in-group/out-group dynamics and social identity.According to this theory, people form identities as members of "in-groups" and in opposition to excluded "out-groups."An in-group's cohesion increases significantly during crises or periods of conflict against a commonly shared out-group.Applying this theory to international politics, interstate conflicts can generate a "rally around the flag" effect in which external threats create a heightened sense of nationalism and bolster support for political leaders.Quantitative empirical work on the rallyinducing effects of international conflict produced mixed results, although anecdotal evidence of this phenomenon is often compelling (Baker and Oneal 2001;Baum 2002;Groeling and Baum 2008).
These mixed results indicate that the likelihood of conflict creating a rally effect may depend on the type of conflict or the initiating state's underlying social structure.Importantly, the psychological literature shows that conflict only induces greater solidarity among groups with high prior levels of cohesion.In fractious or divided populations, some adversaries threaten certain in-group members more than others, and conflict with these adversaries will create widely disparate reactions within the population.In such cases, conflict will likely result in even greater levels of internal disunity, which increases domestic unrest rather than quelling it (Levy 1989, 667).If divided societies will only fracture further due to international conflict, leaders of such states have little hope of diversionary conflict producing a national rally effect.
The implications of this logic for ethnic politics and diversionary conflict are immediately apparent.Assuming ethnicity is a salient cleavage, baseline levels of national cohesion are likely to be comparatively low in both ethnically fractionalized and polarized states.Under such conditions, external conflict is unlikely to produce greater cohesion at the national level.One or a few ethnic groups may rally behind the government, but others will feel alienated and withdraw support for the leadership.In ethnically polarized states, where an excluded ethnic group might be quite large, this is an especially risky proposition.Fearing that conflict abroad would only deepen domestic tensions and further alienate a large portion of the population, leaders of ethnically divided states may forgo diversionary conflict.Essentially, the low prior level of social cohesion in both fractionalized and polarized states could mean that conflict is particularly unlikely to elicit a nation-wide domestic rally effect and may in fact exacerbate domestic instability.Knowing this, leaders of both ethnically fractionalized and polarized states might be particularly unlikely respond to domestic unrest by initiating conflict abroad.This logic yields the two "cohesion hypotheses." Cohesion Hypothesis 1: The positive effect of domestic unrest on interstate conflict should decrease in more ethnically fractionalized states.
Cohesion Hypothesis 2: The positive effect of domestic unrest on interstate conflict should decrease in more ethnically polarized states.

Ethnic Fractionalization and Conflict Opportunities
Alternatively, ethnically fragmented states might be more prone to diversionary conflicts because their fractured ethnic structures provide leaders with more internationalconflict opportunities.In highly fractionalized states and regions, political borders are unlikely to reflect underlying social and demographic boundaries, and politically salient ethnic groups often straddle national borders.This is particularly the case in formerly colonized regions, where maps drawn by imperial powers produced political borders that are often orthogonal to ethnic boundaries. 4enjamin Miller (2007) argued that such "unbalanced" regions are particularly unstable and prone to conflict, with pervasive border disputes and frequent crises arising from the mistreatment of ethnic-kin groups across borders.A wealth of research shows that the existence of transborder ethnic groups increases the probability of both civil strife and interstate conflict (Heraclides 1990;Saideman 2002;Woodwell 2004;Cederman et al. 2013).
This argument relates directly to diversionary conflict because it implies that state-level measures of ethnic fragmentation should correlate with region-wide patterns of persistent interstate conflict and instability.And as mentioned above, conflict-opportunity theory argues that domestic unrest will translate into diversionary conflict only when leaders can seize upon and escalate an existing dispute.Leaders who concoct disputes where none existed previously will be seen as unnecessarily endangering national security and further punished by the population.But the ethnic tensions and persistent territorial disputes that arise in ethnically fragmented areas, where demographic boundaries are orthogonal to state borders, provide a ready supply of conflict opportunities.In regions where states are more ethnically coherent and transborder ethnic groups are less common, embattled leaders are less likely to find a dispute to escalate when needed.When ethnic relations create a history of interstate discord, conflict opportunities are more readily available and domestic unrest is more likely to actually result in dispute initiation.Essentially, ethnic fragmentation and ethnically unbalanced regions should generate a ready supply of interstate conflict-escalation opportunities that facilitate the initiation of diversionary conflict in response to domestic turmoil.This logic yields the first "opportunity hypothesis." Opportunity Hypothesis 1: The positive effect of domestic unrest on interstate conflict should increase in more ethnically fractionalized states.
Conflict-opportunity theory, examined through the lens of ethnic politics, also produces testable predictions regarding the likely targets of diversionary conflicts.If a state's ethnic composition contributes to its diversionaryconflict propensities by supplying a large number of conflict-escalation opportunities, then diversionary onflicts should target states that are ethnically linked to the initiator.Ethnically fragmented states are disproportionately involved in such latent disputes as a result of disagreements over border demarcation, treatment of ethnic-kin groups by foreign governments, or external support of irredentist or separatist groups.And by the logic of conflictopportunity theory, diversionary conflicts initiated by these ethnically fragmented states should disproportionately target states that share ethnic ties with the initiator.I focus more specifically on target states with ties to ethnic groups that are an excluded minority in the initiating state.
There are compelling theoretical reasons why states tied specifically to excluded ethnic groups might make especially attractive targets for diversionary conflict.Diversion abroad can often go hand in hand with a strategy of scapegoating, and perhaps repressing, a leader's domestic enemies (Dassel and Reinhardt 1999;Enterline and Gleditsch 2000;Tir and Jasinski 2008).Diversionary conflict could reinforce this narrative by targeting the foreign allies supporting the excluded/scapegoated ethnic minority.Additionally, the poor treatment of these excluded minority groups could form the basis of an ongoing bilateral dispute or border conflict with the target state, even if the target is the aggrieved side (Gleditsch 2007;Cederman, Girardin, and Gleditsch 2009;Cederman et al. 2013).State B could make a border claim based on the poor treatment of its ethnic kin in State A. State A's leader, looking for a diversionary opportunity, could then escalate and potentially militarize this dispute, claiming that State B is infringing upon its sovereignty.These dynamics do not operate when ethnic groups are well treated and politically protected in both states.
Conflict-opportunity theory thus expects diversionary conflicts to target those states that host ethnic groups linked to the initiating state.Such transborder ethnic groups, especially those that are excluded or repressed in the initiating state, are likely to present a readily exploitable conflict opportunity with the target state.This logic produces the second opportunity hypothesis: Opportunity Hypothesis 2: Diversionary conflicts should disproportionately target states that host transborder ethnic kin of excluded minority groups in the initiating state.

Polarization, Instability, and Risk
Finally, contrary to the logic of Cohesion Hypothesis 2, an alternative dynamic could render leaders of ethnically polarized states more likely to initiate diversionary conflict in response to domestic unrest. 5Recent empirical work finds that ethnically polarized states are most prone to domestic instability and violence (Reynal-Querol 2002; Montalvo and Reynal-Querol 2005).Highly fractionalized states are broken into so many discrete groups that no single faction can rally the collective action necessary to upset the political order or violently take on the government (Reilly 2000).Ethnically polarized states, however, are relatively evenly divided between two large ethnic groups, each of which can make a strong push for control over the government.This creates a highly combustible domestic political situation.
But why would ethnically polarized states then be more likely to initiate diversionary conflict?Two distinct mechanisms could produce this outcome.First, an ethnically polarized state's volatile demographic composition might make leaders more insecure and thus more likely to immediately resort to risky tactics in response to domestic unrest.Leaders of ethnically polarized states know that they sit atop a political powder keg in which even minor unrest can quickly escalate into massive protests, revolution, or even civil war.Such leaders may have to react promptly and forcefully to domestic unrest, utilizing risky strategies in doing so.In many cases, this may result in the initiation of diversionary conflict.Leaders of ethnically tranquil states are less likely to have to resort to drastic measures and can experiment with less-risky strategies in responding to domestic discontent.
Second, because ethnically polarized states consist of comparatively large groups, they may facilitate ethnically homogeneous political coalitions.The bimodal ethnic distribution allows leaders to promote ethnicity as a salient cleavage without risking a cascade of ethnic fissures that might threaten the leader's coalition.The leader's large, ethnically coherent coalition is less likely to break apart when ethnic cleavages become salient.In ethnically fractionalized states, leaders must often appeal to multiple groups, and highlighting ethnic distinctions would risk unforeseen hostility spirals that could fracture the leader's own coalition.But in ethnically polarized states, where a government can often rely on an ethnically cohesive winning coalition, leaders may be able to persecute and scapegoat domestic ethnic opponents without compromising the cohesion of their own supporters.Scapegoating may even increase cohesion within their coalition (Tir and Jasinski 2008).Leaders of polarized states might trust in the ability of their large ethnic in-group to sustain them in power even after completely alienating a large ethnic out-group.
An embattled leader might then seek to bolster the cohesion of an in-group by initiating conflict against the foreign allies of the ethnic out-group.Further alienating this outgroup is not necessarily a deterrent because the leader trusts in the support of the ethnic in-group to maintain power.In a highly polarized state, after all, the leader's in-group will likely constitute roughly half of the state's overall population.Consolidating support among this key subset of the population may tighten an embattled leader's grip on power.In highly fractionalized states, where each group is comparatively small, leaders must often appeal to multiple ethnic constituencies.Highlighting ethnic cleavages by attacking the foreign allies of a domestic ethnic out-group would risk a spiral of unforeseen ethnic hostility that could eventually fracture the leader's diverse coalition.These two mechanisms produce the "vulnerability hypothesis." Vulnerability Hypothesis: The positive effect of domestic unrest on interstate conflict should increase in more ethnically polarized states.

Data and Research Design
This section details the statistical tests assessing the hypotheses specified above.The dataset uses politically relevant directed dyad-years from 1948-2001 as the unit of analysis.I also subset the data by regime type, as a leader's institutional constraints might condition diversionary behavior (Miller 1995;Gelpi 1997). 6

Dependent Variable
To measure conflict, I use the Correlates of War project's Militarized Interstate Disputes (MIDs) data (Jones, Bremer, and Singer 1996;Ghosn and Bennett 2003;Ghosn, Palmer, and Bremer 2004).A MID is a "historical case in which the threat, display, or use of military force short of war by one member state is explicitly directed toward the government, official representatives, official forces, property, or territory of another state" (Jones, Bremer, and Singer 1996, 163). 7Conflict initiation is a binary variable equaling one if the sender initiated a dispute against the target in that year.I only code dispute initiation in the first year of a MID.

Independent Variables
Testing these hypotheses requires operationalizing and measuring four key concepts.In initiating states, I need to capture levels of domestic unrest, ethnic fractionalization, and ethnic polarization.In target states, I need to capture the existence of transborder ethnic groups linked to excluded minorities in the initiating state.
In order to capture a leader's incentives to initiate diversionary conflict in the first place, I use a measure of mass unrest taken from Arthur Banks' Cross-National Time-Series data (Banks 2012).Following conventional practice, I tally the total number of riots, strikes, and mass public demonstrations for each country-year.Regime opponents can often co-opt and redirect such public acts of discontent, even those not initially directed against the political leadership, to destabilize the government.Unfortunately, reliable public-opinion data are not available for most states over time.Alternative measures of leadership insecurity, including those based on economic performance, are highly problematic in that they do not directly capture unrest or discontent.Existing empirical work has also shown that under many circumstances, leaders are not blamed for poor economic performance (Brule and Williams 2009;Johnson and Barnes 2011).Despite its flaws, Banks' measure of mass unrest most accurately captures the theoretical construct of interest.I lag the unrest variable one year to ensure that it is causally prior to conflict.
Ethnic fractionalization is a notoriously tricky concept, and scholars frequently disagree regarding the precise distinction between ethnic groups.I use James Fearon's measure of ethnic diversity, which is derived from the Central Intelligence Agency's World Factbook and Encyclopedia Britannica, and calculated using the Herfindahl index (Fearon 2003;Fearon and Laitin 2003).It estimates the probability that two randomly selected individuals in a given state will belong to different ethnic groups. 8ractionalization is given by the formula: where p is the proportion of the population represented by group i = (1, . . ., N).I use Jose Montalvo and Marta Reynal-Querol's measure of ethnic polarization, which captures the degree to which a country is split between two equally large groups (Montalvo and Reynal-Querol 2005;Reynal-Querol 2002).Both highly cohesive and highly fragmented states register 6 I code states with an aggregate Polity score of six or greater as democracies.All others are authoritarian.Political relevance requires direct territorial contiguity or separation by only a river for all dyads that do not include a great power.Great powers are politically relevant to all other states.7 Because a MID is coded for even the threat of force, I argue that strategic conflict avoidance will rarely wash out diversionary disputes.A target's strategic behavior often begins only after the initiator has issued a threat.Although we must be wary of over-interpreting null results, the broad definition of a MID, which includes the threat of force, mitigates this concern.rather low values on ethnic polarization.But a state made up of only two ethnic groups of equal size (i.e., a fractionalization score of 0.5) is the most-highly polarized.Polarization scores are determined using the formula: Finally, I measure whether potential targets share ethnic ties with excluded minorities in the initiating state.I use Ted Robert Gurr's Minorities at Risk (MAR) data, which cover politically active communal groups in countries with populations greater than 500,000.To be included, an ethnic group must have at least 100,000 members or constitute 1 percent of the population, be "viewed as important" by members, and share some "distinguishing cultural features" (Minorities at Risk Project 2009, 1).I code a dummy variable taking a value of one if the target state contains an adjoining or contiguous ethnic group linked to the sender state.This focuses on ethnic groups that are geographically linked to their kin abroad.Such contiguous groups are better able to offer support to their foreign kin, and are more likely to create a territorial dispute between the two states.

Controls
I also include a series of control variables capturing characteristics of both sender and target states.First, a central tenet of international relations scholarship holds that power is a crucial consideration in states' decisions to use force.I therefore control for relative power within the dyad, measuring the sender's share of aggregate dyadic power [sender/(sender þ target)] according to the Composite Indicator of National Capabilities measure (Singer 1987).Second, enormous literature has shown the significant effect of regime type on international conflict.I therefore control for the level of democracy in both the sender and target state using aggregate Polity scores (Democracy minus Autocracy), version 4.0 (Marshall et al. 2002).Third, repeated studies show geographic proximity to significantly affect dyadic-conflict propensities (Bremer 1992).I thus control for the distance between the two states, in thousands of miles, at their closest points.Fourth, I control for economic interdependence, because the interrupted commerce that results from interstate violence will seriously increase the costs of conflict.I therefore control for trade dependence by measuring the sender's exports to and imports from the target state as a proportion of its overall international trade (Gleditsch 2002).I control for geopolitical compatibility using the unweighted dyadic sÀscore (Bennett and Rupert 2003).Politically incompatible states are more likely to have conflicts of interest and perceive one another as threats, increasing the probability of conflict.Lastly, the directed-dyadic setup requires that I account for cross-sectional dependencies in the data.In particular, due to a lack of military resources or political capital, a leader who has already initiated one conflict in a given year may be less likely to initiate another dispute against a separate target in that same year.As such, I control for whether the sender state initiated a separate dispute in that year.

Models
The high probability of temporal or cross-sectional data dependencies when using panel data raises the possibility of correlated errors that would produce biased estimates.Following Carter and Signorino (2010), I include a peace-year variable capturing the number of years since the most recent dispute within that dyad, as well as quadratic and cubic polynomials.These variables are not theoretically relevant and are thus dropped from the presentation of results.I use logistic-regression models because of the binary dependent variable.As described in the supplementary materials, the results are impressively robust to alternative model and variable specifications.
The hypotheses specified above all posit a conditional relationship between domestic unrest and international conflict.To assess these conditional relationships, I estimate a series of interaction effects to determine how the effect of unrest on conflict changes across different values of ethnic fractionalization, ethnic polarization, and target ethnic overlap.And because interpretation of interactive coefficients is notoriously treacherous, I also graph these conditional relationships (Brambor, Clark, and Golder 2005).

First, I test Cohesion Hypothesis 1 and Opportunity
Hypothesis 1 by examining whether a state's ethnic fragmentation conditions the effect of domestic unrest on conflict initiation.Cohesion Hypothesis 1 expects the positive effect of unrest on conflict to decrease in more fragmented states, as leaders of such states are less likely to benefit from a rally effect.Opportunity Hypothesis 1 expects the opposite.It argues that ethnically fragmented states are more likely to have a large supply of conflict opportunities, including border disputes and concerns regarding the treatment of foreign-kin groups, to draw upon.The positive effect of domestic unrest on international conflict should thus increase in more ethnically fragmented states.Cohesion Hypothesis 1 expects a negative and significant interaction between unrest and fractionalization.Opportunity Hypothesis 1 expects a positive and significant interaction.
Table 1 presents the results, which offer strong support for Opportunity Hypothesis 1 and reject Cohesion Hypothesis 1.The coefficients on the interaction terms are all positive and highly significant.The negative coefficients on the constituent mass unrest variable indicate that, for ethnically cohesive states, domestic unrest actually reduces the likelihood of conflict initiation.But the positive interactive coefficients indicate that this effect becomes strongly positive for more ethnically fragmented states.Ethnically fractionalized states are much more likely to respond to domestic unrest by initiating conflict abroad.Figure 2, which is derived from the full-sample model, illustrates this effect.For states with an ethnic fractionalization score one standard deviation below the sample mean, a two standard deviation increase in mass unrest reduces the probability of conflict initiation by over 36 percent.Conversely, for states with a fractionalization score one standard deviation above the sample mean, the same shift in unrest increases the probability of dispute initiation 56 percent.
Next, assessing the Vulnerability Hypothesis and Cohesion Hypothesis 2 requires examining whether a state's level of ethnic polarization conditions the effect of unrest on conflict.As discussed above, existing work shows that domestic political violence and instability are more likely in ethnically polarized, rather than fractionalized, states (Reynal-Querol 2002;Montalvo and Reynal-Querol 2005;Esteban and Ray 2008).But this finding has unclear implications for diversionary conflict.As the Vulnerability Hypothesis argues, leaders of ethnically polarized states may be more likely to initiate diversionary conflict because they must quickly resort to such risky tactics in response to domestic unrest.Alternatively, as Cohesion Hypothesis 2 argues, leaders of polarized states may fear that international conflict will only deepen internal divides and create further destabilizing unrest at home.Polarized states would then be less likely to respond to unrest by initiating diversionary conflict.The Vulnerability Hypothesis thus predicts a positive and significant interactive effect, while Cohesion Hypothesis 2 expects a negative and significant interaction.
Table 2 and Figure 3 present the results, which fail to support either hypothesis.The coefficients on the interaction terms are highly insignificant in all three models.The interactive coefficient is negative in the full sample and democratic models, as Cohesion Hypothesis 2 predicts.But this effect is far from significant (p = .32 in the full sample, p = .38for democracies).The interactive coefficient is positive in the authoritarian sample, consistent with the Vulnerability Hypothesis.But this effect is again highly insignificant (p = .61). Figure 3 reinforces this conclusion.States with low levels of polarization appear to be marginally more subject to diversion, but the effect is entirely insignificant.These results indicate that ethnic polarization has little effect on a state's diversionary-conflict propensities. 9Ethnically polarized states may be less stable domestically, but this does not prompt the leaders of such states to respond to unrest by initiating conflict abroad.The positive coefficients on the constituent polarization variable across all three models indicate that ethnic polarization does indeed make states more prone to conflict.Ethnic polarization may create persistent unrest, which then presents leaders with diversionary incentives more frequently.But in the face of unrest, leaders of ethnically polarized states do not behave differently from other leaders.
Finally, Opportunity Hypothesis 2 predicts domestic unrest to translate into conflict primarily with states that are linked via transnational ethnic-kin groups to excluded minorities in the initiating state.This logic holds that such ethnically linked states are more likely to present conflict opportunities for embattled leaders to exploit.Transnational ethnic linkages are likely to generate territorial disputes or issues arising from the meddling of foreign-kin groups.Then, when a leader faces incentives for diversionary conflict, these ethnically linked states present readily available and appealing targets.This strategy could operate alongside an attempt at repressing or scapegoating the domestic minority group.In sum, the existence of transnational ethnic ties with the target should increase the conflict-inducing effect of domestic unrest.According to Opportunity Hypothesis 2, the coefficient on the term interacting unrest and a target's ethnic ties should be positive and significant.
The results in Table 3 generally support Opportunity Hypothesis 2. 10 The interactive coefficient is positive and significant in both the full sample and democratic models. 11he interaction term is positive but insignificant for the The positive coefficient for the baseline unrest variable in Models 4 and 5 appears odd, given the negative effect of unrest shown in Table 1.States with a low fractionalization score should also register low on polarization, so at first glance, the baseline unrest coefficients should be similar.But as Figure 1 showed, highly fractionalized states will also produce a low polarization score.And as shown above, highly fractionalized states are extremely prone to diversion.The positive unrest coefficients in Table 2 thus reflect the highly fractionalized states, which outweigh the effect of low-fractionalization cases.
authoritarian sample.This may indicate that authoritarian leaders rely on repression to quell ethnic unrest at home, while democratic leaders must externalize the unrest and initiate conflict abroad.In all three models, transnational ethnic linkages are a highly significant predictor of dyadic conflict initiation.But the interactive effect indicates that ethnically linked states are particularly appealing targets for diversionary conflict, especially for leaders of democratic states.
Figure 4 illustrates this effect.In general, international disputes disproportionately target ethnically linked states.The dashed line is significantly higher than the solid line irrespective of mass unrest.But importantly, this difference expands significantly with domestic unrest, as the gap between the lines increases at higher values on the x-axis.Essentially, an increase in domestic unrest is disproportionately likely to result in a dispute against target states with ethnic ties to the initiator.A two standarddeviation increase in mass unrest from zero increases the probability of dispute initiation against targets with no ethnic links by only 9 percent, from .0036 to .0039.But for ethnically linked targets, the same shift in unrest increases the probability of dispute initiation from .0097 to .014, a nearly 44 percent increase.As a percentage, the conflict-inducing effect of mass unrest is nearly five times greater against ethnically linked targets.In absolute terms, the effect is over fourteen times greater.These findings support Opportunity Hypothesis 2, lending further credence to the notion that embattled leaders can only act upon diversionary incentives when they can draw upon existing conflict opportunities.It also appears that transnational ethnic links often provide such opportunities.

Discussion
The empirical section of this article yields three specific findings.First, leaders of ethnically fragmented states are disproportionately likely to respond to domestic unrest by initiating conflict abroad.Second, disputes initiated during periods of domestic unrest are disproportionately likely to target states that are linked to the initiating state via transnational ethnic-kin groups.Democratic initiators drive this effect. 12Third, ethnic polarization appears to have no consistent, systematic effect on diversionaryconflict initiation.Broadly, these findings indicate that a state's social and demographic makeup can affect its diversionary-conflict propensities.They also show that the availability of conflict opportunities significantly conditions diversionary-conflict incentives.And importantly, transnational ethnic groups often create such conflict opportunities, increasing the probability of unrest translating into international violence.In short, ethnically fragmented states may be more likely to contain excluded ethnic groups, many of which oppose the government and receive support from kin groups abroad.These transnational ethnic ties can create friction between two states, providing the conflict opportunity necessary for an embattled leader to initiate a diversionary conflict.
These results allow us to reject the Cohesion Hypotheses.This may indicate that diverting leaders do not necessarily seek to unify the entire population behind their government.Ethnic divisions do not impede diversionary conflict, and existing research indicates that the further exclusion of marginalized ethnic groups may actually be a primary goal of diversionary strategies (Tir and Jasinski 2008, 642).In highly fractionalized states, an excluded group may be small enough that a leader can attack that group's foreign allies without undue risk of fracturing the ruling coalition.In more polarized states, the alienation of an extremely large ethnic group may generate additional unrest, but the diverting leader can count on a large ethnic in-group to sustain the government.In short, the claim that prior cohesion is necessary for conflict to create a rally effect cannot be directly applied to the state level.Leaders need only appeal to their key constituencies and are not deterred by the prospect of alienating groups outside their coalition.Future work should examine the effect of a leader's fluctuating popularity among different ethnic groups on international conflict.
Although the existing literature shows that ethnically polarized states are particularly prone to instability, this does not appear to affect diversionary-conflict propensities.Highly polarized states may experience instability more frequently, but leaders of such states do not disproportionately respond to unrest with belligerence abroad.Essentially, a state's prior disposition toward instability does not affect how its leaders respond in the face of such instability.
The supplemental materials present a series of robustness checks utilizing alternative operationalizations of the key variables, including unrest, ethnic fragmentation, and conflict initiation, as well as alternative model specifications.The findings presented in this article are impressively robust to these alternative model and variable specifications.

Illustrative Case Study
Finally, to better illustrate the hypothesized dynamics and demonstrate the plausibility of a causal connection between ethnic politics and diversionary conflict, I present a brief case study of the 1998 Turkey-Syria dispute over Syria's support for Kurdish separatists in Turkey. 13This case nicely illustrates how the transnational dimensions of ethnic politics can affect the propensity of unpopular leaders to use force abroad.In particular, it shows Turkey's embattled Prime Minister using a thinly veiled This may be due to authoritarian leaders' ability to rely on domestic repression of ethnic opponents.diversionary tactic to dramatically escalate a long-simmering dispute with Syria.This dispute centered on Syria's support for the ethnic Kurdish separatist movement whose operations frequently straddled the Turkish-Syrian border.
In October 1998, Turkish Prime Minister Mesut Yilmaz directly threatened military force against Syria due to its support for the Kurdish Workers Party (PKK), which had been waging a violent secessionist campaign in Turkey.Syria had sheltered PKK leader Abdullah Ocalan for years and used the organization as leverage in its territorial conflict with Turkey over the Hatay Province and an ongoing dispute over upstream water usage in the Tigris and Euphrates rivers.Syria's own Kurdish population, which numbered nearly three million and abutted the border with Turkey, facilitated the provision of support to the PKK and was directly linked with the territorial dispute over Hatay (Olson 2001, Ch. 4;James and Ozdamar 2009, 24-27).By 1998, the PKK revolt in Turkey was nearly two decades old, resulting in roughly 30,000 deaths and costing the Turkish government over $100 billion (Sezgin 2002, 44).Previous Turkish leaders attempted on several occasions to persuade Syrian President Hafez Al-Assad to cut off support for the PKK, but they generally avoided serious and sustained attempts at coercive diplomacy.
In September and October 1998, however, Yilmaz's domestic political fortunes deteriorated rapidly amid a series of massive domestic scandals (Olson 2001, 116).Media investigations revealed that Turkish officials were collaborating with several notorious organized-crime syndicates and professional assassins in combating the PKK.Relatedly, allegations of extensive mafia involvement derailed ongoing efforts to privatize the Turkish Trade Bank.The confluence of these events, along with ongoing PKK attacks within Turkey, severely threatened Yilmaz's already fragile coalition.Amid these concurrent scandals, and with elections likely to be called within six months, Yilmaz escalated the long-simmering dispute with Syria.He directly threatened military action unless Syria ended its support for the PKK and ceased harboring Ocalan.The Turkish military, he claimed, was prepared and "waiting for the order" (Aras 2012, 42).Additionally, the Turkish military conducted a series of exercises along the border and deployed 10,000 additional troops, along with several television crews, to the area.The Chief of Staff of the Turkish military likewise stated that a "state of undeclared war" presently existed between the two states (Aras 2012, 42).The highly public (and publicized) nature of these threats, in addition to the graphic and bombastic language used, indicated that they were at least partially intended for domestic consumption.Robert Olson (2001, 118) writes that the political fallout of its various scandals left the Yilmaz government "eager for a major distraction; the 'undeclared' war against Syria seemed to fit the need."Turkey's 1998 escalation of this long-running dispute was largely attributable to the domestic problems facing its leadership.
As an act of coercive diplomacy, the gambit was remarkably successful.Assad promptly agreed to Turkish terms, signing the Adana Accords on October 21, 1998.By the agreement, Syria officially designated the PKK as a terrorist organization, deported Ocalan from Syria, expelled an additional 400-500 PKK operatives, and closed down a series of PKK training camps in Lebanon's Bekaa Valley (Sezgin 2002, 50;Aras 2009).Turkish intelligence subsequently captured Ocalan in Kenya in February 1999.But this important victory came too late for Yilmaz, whose government collapsed in January under the deepening pressure of its domestic scandals.
Although it did not save his government and did not result in the actual use of force, Turkey's October 1998 escalation represents a classic diversionary dispute.The timing of Yilmaz's escalation is primarily attributable to his domestic problems (Olson 2001, 116).This case also nicely illustrates the importance of preexisting conflict opportunities for leaders looking to initiate a diversionary dispute.The underlying Turkey-Syria conflict had been ongoing for years and escalated only as Yilmaz's domestic position deteriorated. 14Lacking such a simmering dispute, Yilmaz would likely have been less able to respond to his domestic troubles through diversion abroad.Lastly, this case clearly illustrates how transborder ethnic populations can play a crucial role in facilitating diversionary conflict.Syrian support for the PKK created a long-simmering dispute between the two countries, which Yilmaz only escalated when a series of domestic scandals undercut his fragile government.The Kurdish population that sat astride the Turkey-Syria border was a necessary component of this ongoing dispute, and Syria's support for the PKK ostensibly provoked Turkey's escalation.In short, Turkey's ethnic fragmentation and the existence of a bilateral dispute centered on the transborder ethnic Kurdish population facilitated Yilmaz's diversionary strategy in response to his flagging domestic fortunes.

Conclusion
This article began the process of assessing the degree to which a state's social and demographic characteristics condition its diversionary-conflict propensities.Building upon the expansive literatures on both diversionary and ethnic conflict, I showed that ethnically fragmented states are significantly more likely to respond to domestic unrest by initiating conflict abroad.But importantly, this does not appear to result from political instability or leaders' tenuous hold on power in such countries.Ethnically polarized states, which prior work has shown to be more vulnerable to ethnic conflict and political instability, are not significantly more prone to diversion.Rather, it arises from the more numerous conflict opportunities available to leaders of highly fractionalized states.My analysis also indicates that unrest is disproportionately likely to produce conflict against states with ethnic links to excluded minorities in the initiating state.This is likely due to the ongoing disputes centered on troublesome foreign-kin groups.
This article fits within an ever-expanding literature examining how domestic political forces shape international conflict propensities.More specifically, it enhances our understanding of diversionary conflict by broadening the range of forces that we know to condition and constrain the diversionary incentives facing embattled leaders.A substantial body of literature examines the ways in which different domestic institutional structures can impede or facilitate diversion.I add to this literature by examining how social and demographic factors, specifically ethnic structures, condition diversionary conflict.This also 14 Indeed, Yilmaz's domestic fragility may have rendered his threats too credible.Assad conceded so quickly that the dispute did not escalate to the point of generating a significant rally effect.But a failed attempt at diversion is still a diversionary conflict.And Yilmaz's behavior was indeed diversionary.Importantly, the dispute appears in the MID data despite Assad's clear attempts at strategic conflict avoidance.Strategic interaction operated only in preventing greater escalation or a longer duration.
highlights the important and underappreciated linkages between ethnic politics-both within and across borders-and international conflict.
The 1998 dispute between Turkey and Syria and the 2014 crisis in Ukraine illustrate the important ways in which domestic ethnic structure and transnational ethnic ties can impact diversionary conflict.In both cases, leaders sought to shore up domestic support by initiating disputes against neighbors.And in both cases, disputes surrounding transborder ethnic groups provided the conflict opportunity that the leaders opportunistically escalated.Putin's primary justification for intervening in Ukraine was protecting ethnic Russians across the border.Similarly, Yilmaz's demands against Syria centered on the Kurdish separatist movement that operated across the Turkey-Syria border.National demographic structures and transnational ethnic linkages were crucial in producing conflict.
My arguments should help policymakers anticipate and defuse conflicts by revealing precisely where and when unpopular leaders are likely to lash out abroad.In particular, it shows that even long-dormant ethnic disputes can provide the conflict opportunities necessary for domestic unrest to translate into international crisis.Potential targets with ethnic links to states with unpopular or embattled leaders must be especially proactive in order to deter or dissuade a potential diversionary conflict.By making concessions or moving quickly to align international support, potentially attractive diversionary targets can deprive embattled leaders of the conflict opportunities necessary for diversionary incentives to escalate into actual violence.For instance, the new Ukrainian leadership likely knew that the ethnic Russian population in the Eastern Ukraine and Crimea provided Putin with an easy excuse for intervention.Petro Poroshenko's government simply failed to offer either adequate assurances of Russian interests in the Eastern Ukraine or adequately strong deterrent threats to forestall Russian action.Poroshenko was likely unable to make credible deterrent threats given the power asymmetry between Ukraine and Russia.Clearer assurances of Russian interests, however, could perhaps have forestalled Russian intervention.Poroshenko's own tenuous position domestically may have simply prevented him from offering such assurances.But this case illustrates the massive potential costs of failing to head off diversionary conflicts arising from cross-border ethnic disputes.Regardless of whether such moves would have been effective in this case, preventing or mitigating diversionary conflicts requires recognizing the potential dangers presented by simmering conflict opportunities, including those based on transnational ethnic disputes.

Figure 1
Figure1visually depicts the relationship between fractionalization and polarization.Finally, I measure whether potential targets share ethnic ties with excluded minorities in the initiating state.I use Ted Robert Gurr's Minorities at Risk (MAR) data, which cover politically active communal groups in countries with populations greater than 500,000.To be included, an ethnic group must have at least 100,000 members or constitute 1 percent of the population, be "viewed as important" by members, and share some "distinguishing cultural features" (Minorities at Risk Project 2009, 1).I code a dummy variable taking a value of one if the target state contains an adjoining or contiguous ethnic group linked to the sender state.This focuses on ethnic groups that are geographically linked to their kin abroad.Such contiguous groups are better able to offer support to their foreign kin, and are more likely to create a territorial dispute between the two states.

Table 1 .
Domestic unrest and ethnic fragmentation Note: Logit estimates with robust standard errors clustered on directed-dyad in parentheses.Peace year polynomials are included in each model, but not presented.*p <.1, **p <.05, ***p <.01.

Table 2 .
Domestic unrest and ethnic polarization Note: Logit estimates with robust standard errors clustered on directed-dyad in parentheses.Peace year polynomials are included in each model, but not presented.*p <.1, **p <.05, ***p <.01.

Table 3 .
Domestic unrest and transnational ethnic kin 9