Where,when,and how does the UN work to prevent civil war in self-determination disputes?

Conflict prevention as a priority

In recent years, the United Nations has emphasized the importance of preventing, rather than just resolving, violent conflicts. In his ‘Preventive Diplomacy, Delivering Results’ (S/2011/552) report, Ban Ki-moon specifically called attention to the UN entities involved in conflict prevention, including the General Assembly, Department of Political Affairs, UN special envoys, regional offices, and the Peacebuilding Commission, as well as the Security Council. From 2010 until early 2012, the UNSC held periodic ‘Horizon-Scanning Briefings’ hosted by the UN’s Department of Political Affairs that covered potential hot spots where preventive diplomacy could be useful. ⁶ However, these briefings were discontinued in 2013, in part, due to objections of UNSC member-states who did not want their internal affairs discussed. ⁷

Scholars have likewise shown an interest in conflict prevention and have focused primarily on two areas. First, large-scale early-warning projects such as the Political Instability Task Force (PITF) ⁸ and the Integrated Crisis Early Warning System (ICEWS) ⁹ seek a better understanding of when violent conflict is imminent, so that prevention resources might be efficiently allocated. The wider availability of data on political violence has also led to a greater attention to forecasting such events. ¹⁰ Second, scholars have developed analytical frameworks and tools for conceptualizing prevention and understanding how it is likely to be effective ¹¹ and have used observational data, such as those in the International Crisis Behavior (ICB) ¹² and Managing Intrastate Low-Level Conflict (MILC) ¹³ projects, to explore the factors which help prevent the escalation of conflict.

This growing literature has expanded our understanding of conflict prevention but still largely ignores the determinants of preventive action. When, and where, does the UN get involved in disputes to prevent the outbreak of civil war? Why does the UN take the specific preventive actions that it takes?

The determinants of UN conflict prevention efforts

The UN takes a range of actions that could decrease the chances that nonviolent disputes escalate to civil war. In general, prevention works through two main mechanisms: by altering the cost–benefit calculus of actors and by helping combatants overcome barriers to reaching efficient nonviolent settlements. We focus on four specific types of UN action that can have these effects.

First, the UN often engages in diplomatic engagement – including good offices, mediation, fact-finding, civilian monitoring, and special tribunals – in intrastate disputes. This engagement can help overcome barriers to efficient bargaining as well as raise the perceived benefits of continued nonviolence. Second, the UN often issues Security Council resolutions in the form of condemnations, in which the Security Council deplores hostilities, non-compliance with prior UN directives or international agreements, or human rights violations. While the condemnation itself may have little direct effect on the cost–benefit calculation of states and dissidents, it can signal UN intention to intervene more forcefully in the event of a violent conflict and thus indirectly affect that calculation. Third, the UN can levy sanctions on countries in an effort to directly raise the costs of conflict and incentives for reaching a settlement. Fourth, the UN can authorize and deploy peacekeeping or peace enforcement missions to directly enforce a peace, raise the costs of violence, or monitor implementation of an agreement.

The UN, then, has a range of tools it can use to seek to prevent violence, and recent work suggests that some of these can be effective at preventing intrastate disputes from escalating to high levels of violence, including among disputes with no history of significant violence (Beardsley, Cunningham & White, 2017). However, the UN does not act in all disputes, and in the disputes it does get involved in, it uses various combinations of preventive tools.

We argue that the UN uses objective indicators to determine the likelihood of civil war and desires to engage in conflict prevention in the disputes with the greatest likelihood of becoming violent. Scholarship has demonstrated that the UN acts to promote humanitarian interests. Murdie & Peksen (2014) show that human rights organizations have substantial influence on UN action, and Choi (2013) finds that humanitarian actions taken by the United States are often driven by liberal interests to promote democracy and human security. ¹⁴ Beardsley & Schmidt (2012) find that UNSC action is driven by incentives to respond to the gravest threats to international stability and to act in the interests of the P-5, but the latter influence appears weaker than the former.

Additionally, scholarship on UN action in violent conflicts has demonstrated that the UN and other international actors primarily get involved in the civil wars most in need of intervention (the ‘hardest’ cases). Several studies demonstrate that UN peacekeepers deploy, on average, to conflicts with the lowest probability of successful resolution ¹⁵ and that mediators generally are sent to conflicts where negotiations are least likely to succeed. ¹⁶ Taken together, scholarly evidence combined with UN rhetoric suggests the UN has a desire to work to prevent violent conflict in intrastate disputes and should be most likely to act in disputes with the greatest likelihood of civil war.

However, political opposition can impede the UN’s preventive efforts. UN action in intrastate disputes has the potential to provide international legitimacy to dissidents and can suggest that the government is incapable of managing its internal affairs. As such, governments are likely to be stridently opposed to these types of actions in their own nonviolent disputes. This opposition can directly constrain the ability of the UN to act in nonviolent substate disputes when the government in question is either a permanent member of the UN Security Council (P-5) or a close ally of one of the P-5 members. The P-5 – China, France, the UK, the USA, and Russia – can veto any action proposed in the Security Council and therefore prevent the UN from taking action. ¹⁷ In addition, there are ten rotating members of the Security Council who will be sensitive to UN involvement that diverges from their interests or those of their allies. In some cases, these states are precisely those that have disputes that could be prone to civil war. Indeed, a main reason the Horizon Scanning Briefings were largely discontinued after 2012 (only three were held in 2013) was opposition from the rotating members of the Security Council who did not want their internal affairs to be the subject of UNSC briefings (Romita, 2011). Moreover, states that are not members of the Security Council can reduce the potential for authorizations of certain UN activity – involvement that requires the consent of states facing SD disputes in order to occur, such as peacekeeping and mediation.

We consider three sets of factors that determine where preventive action is needed, and we argue these sets of factors have differential impacts on the level of political opposition to UN involvement. The first is the history of conflict in the dispute. Periods of peace in civil wars often break down; in fact, most ‘new’ episodes of civil war are actually recurring wars. Existing scholarship has shown that states with a history of civil war are much more likely to experience civil war than states that have never experienced civil war (e.g. Collier et al., 2003; Walter, 2004; Mason et al., 2011). This empirical reality is acknowledged explicitly in the Secretary-General’s 2011 report on preventive diplomacy, stressing the role that established UN peacekeeping missions can play in preventing conflict recurrence through on-the-ground mediation (S/2011/552, 14). While in some cases UN involvement in nonviolent disputes with a history of conflict follows shortly after the end of the war, in many cases, such as the Turkish Cypriot, Abkhazian, and Western Saharan disputes, much of the activity takes place years or decades after the violence subsides. This suggests that the UN is not just concerned with maintaining a short-term peace, but also actively seeking to prevent the emergence of new episodes of violence.

The history of violence in a dispute is clearly observable and, as such, we expect it to strongly influence where the UN chooses to get involved. Additionally, the objectivity of prior conflict history is likely to aid the UN in overcoming political constraints on acting. As such, we expect the UN to be highly involved in disputes with a history of violence and to be able to do so using the most forceful and costly forms of intervention – such as sanctions and the deployment of force – in addition to less active forms of involvement like diplomacy and condemnation.

Hypothesis 1: UN preventive action is more likely in nonviolent disputes with a prior history of violence than in those without a history of violence.

The second set of factors we expect to influence where and how the UN intervenes is the regional context within which the dispute takes place. It is well established that civil wars frequently diffuse (Gleditsch, 2007; Buhaug & Gleditsch, 2008), and policymakers often pay attention to regional ‘hot spots’ such as the Horn of Africa or the Balkans. Kathman (2011) argues and shows that third-party states evaluate the potential for regional destabilization when deciding whether to intervene into civil wars. Indeed, in 1992, the UN deployed the United Nations Preventive Deployment mission to Macedonia specifically to prevent contagion from the nearby civil wars in Bosnia and Herzegovina and Croatia. Refugee flows stemming from armed conflict, such as the flight of thousands of refugees to Cyprus during the Lebanese Civil War, ¹⁸ can also raise the regional risk of violence (Salehyan & Gleditsch, 2006) and may further motivate UN action.

While we expect the potential for regional contagion to motivate the UN to act preventively, this potential is less objectively obvious than a past history of violence. We anticipate the UN will pay less attention and be less motivated to devote substantial resources to these disputes than to disputes with a history of violence. In addition, because the potential for violence is less objectively obvious, political opposition to UN involvement in these disputes will be higher. This leads us to expect that, while the UN will be more likely to act in disputes where the potential for regional contagion is higher, it is more likely to intervene in less active and less costly ways such as authorizing diplomatic engagement and issuing condemnations.

Hypothesis 2: UN preventive action is more likely in nonviolent disputes with greater potential for regional contagion.

A third set of indicators the UN can use pertains to the characteristics of the dispute itself. Practitioners have expectations about the types of disputes that are more or less prone to violence. Scholarship has suggested SD disputes are most likely to be violent when SD groups are internally divided (Cunningham, 2013), geographically concentrated (Toft, 2005), or have ethnic kin in other states (Jenne, 2004), and when the state contains multiple SD groups (Walter, 2009). The UN can examine these factors, as well as more dynamic events, such as protests, negotiations, referenda, and elections, to determine the likelihood of violence.

However, determining which disputes are more prone to violence based on their current dispute-level characteristics is more subjective and uncertain than when there is a history of violence in the dispute or current conflict in the region. Even our best forecasting models based on dispute-level variables fail to predict many conflict outbreaks and produce many false-positives (Ward et al., 2013). Due to the political constraints on UN preventive action in cases that are not clearly at risk for conflict escalation, we do not have empirical predictions about the effects of dispute-level factors on UN action.

Empirical analysis

We test the determinants of UN preventive action using data on all disputes between SD movements and states from 1960 to 2005. ¹⁹ The population of SD movements comes from the Center for International Development and Conflict Management (CIDCM) Peace and Conflict Report (Marshall & Gurr, 2003). ²⁰ This report identifies 145 SD movements residing in 77 countries which demand increased local control and autonomy for an ethnic group in some territory. This list encompasses groups that have participated in civil war, such as the Kurds in Turkey and Iraq, as well as those, such as the East Caprivians in Namibia, that have not. As such, these data represent a set of disputes that have the potential to become violent, some of which experience civil war, some of which do not, and there is substantial variation within the violent disputes in the timing of the onset and end of violence. Of the SD movements we examine, 44% experienced a civil war at some point. The Online appendix provides a table listing each self-determination dispute and indicating whether it experienced a civil war at any point in our data.

To operationalize UN action, we examine the issuance of UNSC resolutions comprising the authorization of diplomatic engagement, condemnation, the authorization of sanctions, and the deployment of force targeted toward these disputes. The UN has a number of subagencies that work on the ground in disputes, and much of their action could potentially be viewed as ‘conflict prevention’. However, UNSC resolutions provide the most visible signal of UN interest in the conflict, precisely because passing a resolution requires at minimum no opposition from the P-5. Resolutions provide clear evidence to disputants that, at the highest levels, the United Nations is actively monitoring a dispute.

To examine where and how the UNSC issues resolutions related to diplomacy, condemnations, sanctions, and force, we have coded the content of all UNSC resolutions directly related to all active self-determination disputes from 1960 to 2005. We used data from Beardsley (2013), who has coded whether a resolution pertains to diplomacy, the authorization of force, sanctions, or condemnation based on the text of the operative paragraphs of the resolution. After matching the resolutions to the countries with SD movements, we went through each resolution and determined if the resolution directly pertained to the SD movement in question, indirectly pertained, or was not relevant. ²¹ In this article, we consider the UNSC to have been active in a particular dispute-year if it issued a resolution directly pertaining to that dispute ²² which contained calls for diplomatic action by the UN, the authorization of the deployment of force by the UN or its member states, the assignment of sanctions, or condemnation.

Resolutions authorizing diplomatic actions include those calling for good offices, mediation, fact-finding, civilian monitoring missions, and the formation of special tribunals. An example is UNSC Resolution 367 (March 1975), which called on the UN Secretary-General to undertake a mediation role in the dispute over the status of Turkish Cypriots. We also consider as diplomacy initiatives that focus on non-military peacebuilding activities, such as political assistance, as well as demobilization, disarmament, and reintegration (DDR). An example is Resolution 1590 (March 2005), which mandated the UN Mission in Sudan (UNMIS) to assist in the return of refugees and displaced persons following the adoption of the Comprehensive Peace Agreement between the Sudanese government and the Southern Sudanese representatives.

UNSC resolutions comprise condemnations when the Security Council deplores hostilities, non-compliance with prior UN directives or international agreements, or human rights violations. ²³ We also consider as condemnations threats of sanctions that are not actually implemented in a resolution. An example of a condemnation is Resolution 1076 (October 1996), which condemned the civil war inside of Afghanistan – a conflict which incorporated the Tajik and Uzbek self-determination disputes.

We code UNSC resolutions as levying sanctions when they impose, reauthorize, or expand sanctions on the country involved. Resolution 713, for example, established an embargo in September 1991 on military material being exported to the former Yugoslavia.

Finally, resolutions authorizing the deployment of force can either include UN peacekeeping missions – such as the United Nations Observer Mission in Georgia (Resolution 858, August 1993) – or authorizations of non-UN multinational forces – such as the NATO-led International Security Assistance Force in Afghanistan (Resolution 1386, December 2001). ²⁴ For the purposes of this study, we only focus on ‘new force’ authorizations, which authorize a new deployment, expand an existing mandate, or authorize ‘all necessary means’. We consider the expansion of existing mandates to be ‘new force’, because the UNSC may authorize extant missions to use additional actions to forestall violence or may expand the geographic scope of the mission – effectively establishing a new UN presence. For example, the UNPROFOR mission was originally established in 1992 by UNSC Resolution 743 with a mandate limited to Croatia, but this was later expanded to include Bosnia and Herzegovina (UNSC 758) and Macedonia (UNSC 795). Technically speaking, there was still only one mission (UNPROFOR), but a new application of force was established as its mandate was expanded to new disputes. We also code the authorizations of ‘all necessary means’ as ‘new force’ because this language is rare and has special significance, signaling that the UNSC is using the full weight of its authority. In contrast, we do not consider resolutions that simply extend the duration of an existing mission – without an expansion of the mandate or authorization of ‘all necessary means’ – to be relevant preventive

OPEN IN VIEWER

Table II.

UNSC actions in self-determination disputes (1960–2005)

Country	SD group	Condemn	Diplomacy	Sanctions	New force	Total actions
Afghanistan	Tajiks	4	0	3	1	8
Afghanistan	Uzbeks	4	0	3	1	8
Azerbaijan	Armenians	2	1	1	0	4
Bosnia	Croats	23	8	7	30	68
Bosnia	Serbs	26	8	7	30	71
Croatia	Serbs	10	7	2	16	35
Cyprus	Turkish Cypriots	14	60	0	2	76
Georgia	Abkhazians	22	4	0	4	30
India	Kashmiri Hindus	1	0	0	0	1
India	Kashmiri Muslims	1	1	0	1	3
Iraq	Kurds	1	0	0	0	1
Israel	Palestinians	25	2	0	0	27
Lebanon	Palestinians	32	13	0	9	54
Macedonia	Albanians	1	0	0	0	1
Morocco	Saharawis	6	7	0	3	16
Russia	Chechens	1	0	0	0	1
Spain	Basques	1	0	0	0	1
Sudan	Southerners	0	2	0	1	3
Yugoslavia	Albanians	1	0	0	0	1
Yugoslavia	Croats	0	0	1	0	1
Yugoslavia	Kosovar Albanians	3	0	1	1	5
Yugoslavia	Slovenes	0	0	1	0	1
	Total	178	113	26	99	416

We use this coding to generate our main, binary dependent variable of Any UNSC action in a given dispute-year, and additional binary dependent variables for each type of UN action. These variables are not mutually exclusive; a dispute can attract multiple resolutions in a year, and each resolution can contain more than one action.

Table II presents counts of each type of action across these disputes. Condemnation is the most common. However, many resolutions mandating the deployment of force also contain condemnations of the violence that the force is being deployed to stop. Diplomacy is the second most common UNSC action – with the majority of diplomacy directed towards the Turkish Cypriot dispute. Renewals of the peacekeeping mandate in Cyprus were often accompanied by authorizations for the UN Secretary-General to pursue good offices missions in Cyprus. Surprisingly, sanctions are much rarer than even new authorizations of force – with the Balkan SD disputes attracting the majority of sanctions. It may be that the political hurdles of achieving sufficient support in the UNSC for sanctions even against the most egregious violators of human rights can be steep. ²⁵ It may also be that sanctions are a relatively blunt instrument that is ill-suited for dispute-level targeting. While the Balkans also see the majority of new UN authorizations of force, there are a wide range of SD disputes with at least one authorization of force. ²⁶

In our analysis, we examine the effect of conflict history, regional contagion, and dispute-level characteristics on UN preventive action. Before we turn to that analysis, looking at Table II, we can quickly identify two important determinants of UN action. First, the UN almost never issues resolutions related to SD disputes within the P-5, even while every P-5 member has at least three SD disputes active at some time from 1960 to 2005 (see Online appendix). The lack of involvement in P-5 SD disputes illustrates some of the constraints on the UN. Indeed, the only UNSC action that applies to a P-5 member is Resolution 1440, which condemned the taking of hostages at a theater in Moscow by Chechen extremists and expressed support toward Russia.

Second, as expected, the UNSC is active at a greater rate in violent disputes. Table II shows that the Balkan SD disputes take the lion’s share of UN involvement, generally. The Palestinian SD disputes in both Lebanon and Israel attract a great deal of involvement as well. Table I in the introduction shows that while there are more dispute-years with UNSC action in nonviolent disputes (79 versus 42), the UNSC gets involved through the issuance of relevant resolutions in violent years in SD disputes at more than twice the rate that it does in nonviolent years (5.65% of dispute years versus 2.50%).

The clear association between violence and UNSC action shows that, not surprisingly, the UN is much more active in substate disputes once they become civil wars. However, we can see from Tables I and II that the UN does still act in periods of these disputes that are nonviolent.

Model estimation

Our data include all state-SD group dispute-years – our unit of analysis ³⁰ – from 1960 to 2005, excluding years in which a civil war between the SD group and state is ongoing, but including the year in which civil war began. ³¹ We exclude ongoing war years because action during such time periods is outside of our focus on conflict prevention. We estimate models for the occurrence of the different types of UNSC resolutions – one model for any of the four types and one for each of them.

OPEN IN VIEWER

Table III.

Determinants of resolutions in nonviolent SD disputes

	1: Logit Civil war	2: RE Logit Any UNSC	3: RE Logit Diplomacy	4: RE Logit Force	5: RE Logit Condemn
Factions (logged)	2.437*	0.666	0.810	0.947	0.583
	(0.480)	(0.383)	(0.356)	(0.364)	(0.380)
Previous concessions (lag)	0.565*	1.532	0.821	0.257*	1.823
	(0.140)	(1.199)	(0.617)	(0.147)	(1.263)
GDP per capita (lag)	0.747*	2.210*	1.823	0.991	2.079*
	(0.083)	(0.581)	(0.743)	(0.566)	(0.735)
Democracy (lag)	0.543*	0.707	2.383	0.385	0.331
	(0.156)	(0.600)	(1.427)	(0.420)	(0.272)
Neighboring kin	1.644†	1.785	1.366	0.966	1.653
	(0.428)	(1.823)	(1.379)	(0.956)	(1.479)
Previous civil war	1.579	13.560*	11.120*	6.251*	15.230*
	(0.489)	(8.841)	(4.794)	(3.877)	(12.450)
Years since civil war	0.968*	0.972	0.953	0.975	0.995
	(0.014)	(0.027)	(0.031)	(0.030)	(0.018)
Nearby civil war		2.096*	3.336*	1.344	1.892*
		(0.492)	(1.047)	(0.419)	(0.509)
Other civil war in state		0.478	0.304	0.779	0.703
		(0.352)	(0.295)	(0.690)	(0.525)
Cold War		1.227	4.092*	0.418	0.317†
		(1.100)	(2.878)	(0.442)	(0.218)
P-5 or P-5 ally		0.331	0.069†	0.978	1.196
		(0.396)	(0.096)	(1.560)	(1.086)
Nearby states		1.063	1.019	1.177†	1.098
		(0.066)	(0.108)	(0.116)	(0.084)
N	2,503	2,502	2,502	2,502	2,502

Exponentiated coefficients (odds ratios); robust standard errors in parentheses; ^†p < 0.10, *p < 0.05, two-tailed.

All of these dependent variables are binary, and less than 5% have the value 1, so we use rare events logit (King & Zeng, 2001). The models of sanctions could not be estimated because of too few authorizations of sanctions related to SD disputes. We estimate standard errors that are robust to clustering at the level of the SD dispute.

As control variables, we first include a measure of P-5 security interests, operationalized as a dummy variable of whether the state in which a dispute took place was a P-5 member or an ally of a P-5 member. ³² We also include a control for the Cold War (1960 to 1991 in our data), when Security Council politics markedly reduced the potential for UNSC preventive diplomacy. Finally, we control for the number of states in the home-country’s ‘neighborhood’, so that our variable that counts the number of nearby civil wars is not just picking up the size of the state’s neighborhood.

Results

Table III presents the exponentiated coefficients – odds ratios, with values greater than 1 indicating a positive relationship – and standard errors. We first ran a logit model with the onset of civil war in an SD dispute as the dependent variable in order to confirm that the explanatory variables help anticipate the outbreak of a civil war. Model 1 shows the estimated results from this model and confirms that many of the variables used by Cunningham (2013) have a statistically significant association with civil war onset in SD disputes. Disputes with more factions, with fewer previous concessions, with non-democratic political institutions, with kin in a neighboring state, that take place in poorer states, and with a more recent civil war are more likely to escalate to a civil war.

Models 2 through 5 pertain to the occurrence of the different types of UNSC resolutions during SD disputes that are not experiencing an ongoing civil war and provide tests of Hypotheses 1–3. ³³ We find strong support for the first two hypotheses: that UN preventive action is more likely in nonviolent disputes with a history of civil war and when the dispute takes place in a region with nearby civil wars. The Previous civil war variable is positive and statistically significant with a large coefficient in Models 2–5, and the Nearby civil war variable is positive and statistically significant in all but Model 4.

The results from Model 2 comport well with the expectations of Hypotheses 1 and 2 for where UN action was likely. Hypothesis 3 concerns observable expectations regarding how the UN gets involved. Again, we find clear support. Models 3–5 show that diplomacy, condemnations, and authorizations of force are more likely in nonviolent SD disputes with a previous civil war. The results are strongest for the ‘lighter’ forms of UNSC action – diplomacy and condemnations. Compared to SD disputes without any previous civil war, UNSC diplomacy and condemnation are more than ten times more likely in the SD disputes that have experienced a previous civil war. However, the coefficient on Previous civil war in Model 4, predicting the authorization of force, is large as well.

Also consistent with Hypothesis 3, we find that the potential for regional contagion is significantly associated with diplomacy and condemnation. On average, for an SD dispute that is in a country with one civil war in its regional neighborhood – such as Yugoslav Albanians in 2001 – the odds of a UNSC resolution related to diplomatic activity increase by 234% relative to an SD dispute in a country that resides in a peaceful region – such as Scots in the UK in the same year. The odds of a resolution related to condemnation increase by 89% in expectation. However, as expected by Hypothesis 3, the strength of the relationship between nearby civil wars and UNSC involvement is diminished for resolutions pertaining to new force authorizations, and the coefficient is statistically insignificant. Interestingly, we do not find that UNSC action of any type is more likely when there is another civil war in the same state. All of the coefficients are less than 1, indicating the presence of a negative relationship, although there is substantial imprecision around the estimates. One potential explanation could be that the presence of another civil war overshadows attention to other SD movements in the same state.

While we examined the effect of dispute-level characteristics associated with the likelihood of civil war, we did not have a prediction of their effect because we anticipated that political constraints would severely limit the ability of these relatively static variables to generate UN preventive involvement in SD disputes. The results show little evidence of the importance of dispute-level factors in driving UNSC resolutions oriented toward conflict prevention. ³⁴ The numbers of factions, democracy, and neighboring kin groups never have a statistically significant relationship with UNSC action. GDP per capita does have a statistically significant relationship with UNSC action, but the direction of the effect (negative) is opposite to that of its correlation with civil war. This suggests that UNSC action is more likely to go to states that are more economically developed, which is the opposite effect to what we would anticipate if the UN were directing its action to disputes with characteristics that make them more likely to escalate to civil war. ³⁵ The Previous concessions variable does have a statistically significant and negative relationship with UNSC authorizations of new force, but the bulk of the evidence suggests that state- and dispute-level variables positively related to the potential for civil war do not strongly predict UNSC activity.

Turning to the other variables included in the models, we find the Cold War is associated with a higher likelihood of diplomatic interventions – which is counter-intuitive but driven in part by the high rates of diplomacy in the Cypriot conflict during the Cold War. We find, consistent with expectations, that SD movements in P-5 states or P-5 allies are typically less likely to experience UNSC resolutions. Moreover, SD movements in states with many neighbors tend to experience more UNSC resolutions.

Conclusion

We know the UN is constrained in where it can act by great power competition and the interests of the P-5. Additionally, we know that, particularly after the end of the Cold War, the UN exerts considerable effort in seeking to resolve ongoing violent disputes. The results in this article show that the UN does not wait for violence to start, but gets involved in relatively peaceful disputes that have a greater likelihood of escalating to violence. Specifically, the UNSC is more likely to adopt resolutions that apply to an SD movement when there is a possibility that a nearby civil war might undermine the security environment in the region or when a dispute has a history of significant violence.

These results also suggest that the UN uses a diverse set of tools in preventing violence. In both our main analyses and robustness checks, diplomacy is a central activity, suggesting that the UN’s rhetorical focus on ‘preventive diplomacy’ extends beyond mere words. The prominence of diplomacy is important; in other work we show that UNSC resolutions authorizing diplomatic action directly pertinent to SD disputes reduce the likelihood of civil war (Beardsley, Cunningham & White, 2017). By contrast, the UN rarely uses sanctions in SD disputes. There is a large literature on the effect of economic sanctions (e.g. Pape, 1997; Drezner, 2003), and a prominent argument is that the effect of sanctions is primarily driven by selection effects – that is, it is the threat of sanctions that is effective. The analyses here generally only measure the authorization of sanctions, ³⁹ not their threatened use, and this may partly explain why the UN does not appear to use sanctions proactively in our data.

The results here suggest a clear need for further research on conflict prevention. Much discussion about conflict prevention by policymakers and academics focuses on the cases of failed prevention, and there are unfortunately many. However, our study shows that the UN seeks to act in disputes before they become violent, and by only focusing on actual conflicts we have the potential to miss cases that would have become violent in the absence of UN action.

While in this article we focus on the UNSC, regional organizations, national governments, and nongovernmental organizations also play an important role in conflict prevention (particularly in the area of diplomacy) and may actually be better positioned to respond quickly to disputes where violence is a possibility. Future research examining when and how a broader range of actors work preventively in nonviolent disputes, and whether these actions have an effect on reducing violence, is warranted.