United Nations peacekeeping dynamics and the duration of post-civil conflict peace

Rationalist explanations of civil war

According to the rationalist bargaining model of war, because civil wars are costly enterprises, belligerents should find peaceful solutions to their grievances rather than let conflict persist. Yet wars are fought and endure largely for two reasons: information asymmetries and commitment problems (Fearon, 1995). ¹ In the process of bargaining over the division of disputed resources, civil war combatants have incentives to misrepresent privately held information in an attempt to secure an agreement that yields the greatest benefit (Wagner, 2000). Combatants misrepresent private information to appear stronger, more resolved, less willing to back down and less agreeable to solutions that do not satisfy their most preferred bargaining position. Conflict is a way to gain information about one another’s resolve, strength and reservation points. Information revelation increases the ability of the combatants to come to an agreement over the division of the disputed resource (Filson and Werner, 2002; Ramsay, 2008; Slantchev, 2004).

If information asymmetries provoke fighting, then revealing relevant information should help the parties arrive at negotiated solutions that result in an end to fighting and the establishment of an enduring peace. However, the cessation of conflict does not necessarily mean that information asymmetries no longer exist between the factions. Civil war combatants may de-escalate hostilities below an easily observable threshold in order to regroup, reform their forces or go underground to rebuild popular support. They may then later reinitiate the fight to acquire information or to relay information about their capabilities and resolve. Thus, civil wars may experience prolonged periods of ‘peace’ that are simply the product of low combat activity.

Even in instances where the government and rebels reach a negotiated pact, these agreements may be unstable, evidenced by the tendency of ceasefires to break down soon after signing. Settlements may serve as stopgap measures that allow breathing space before fighting resumes. In this sense, ceasefires or peace agreements may not be representative of mutually agreeable resolutions that sufficiently satisfy combatant reservation points.

Like information asymmetries, commitment problems can lead to war even in the presence of perfect information (Powell, 2006). Commitment problems hamper the ability of opponents to reach peaceful settlements, since one or both sides believes that gains from fighting outweigh the concessions offered by a negotiated outcome (Powell, 1999, 2012). Commitment problems between combatants have seen increasing applications to civil war (Walter, 1997, 2001; Wood and Kathman, 2014). Such problems loom particularly large in civil wars, where rebel groups face a security dilemma. For civil war to effectively end, rebels must disarm, demobilize, and eventually reintegrate themselves into society. However, by disarming, rebels sacrifice their only means of protecting themselves from a government incentivized to renege on its previous commitments to the group. Some mechanisms of peace agreements, such as power-sharing provisions, help overcome incentives by the parties to renege on their commitment to peace (Hartzell and Hoddie, 2003, 2007; Mattes and Savun, 2010). However, in the relatively anarchic environment that defines government–rebel interactions during and immediately following civil war, formal agreements may only be valuable to the extent that they are enforceable. Thus civil wars often persist for extended periods unless combatants are able to secure outside security guarantees that ensure the fulfillment of the war termination terms (Walter, 1997, 2001; Walter and Snyder, 1999).

To summarize, civil wars occur and recur as a result of information asymmetries and commitment problems between the combatant factions. Even with a conflict’s cessation, opportunities exist for the restart of hostilities. Since the ‘end’ of war does not often mean the elimination of one side at the hands of the other, the ongoing organization of combatants keeps open the door for renewed hostilities, as opportunities for redistributing resources and benefits through violence persist into the post-conflict period. ²

The importance of mission characteristics in securing post-conflict peace

Applied to post-conflict environments, the bargaining model of conflict implies that in order for third parties to stabilize post-conflict settings and extend the duration of peace following civil wars, third parties must be able to help share or reveal private information between combatants, or help combatants commit to and uphold durable peaceful settlements. By extension, for UN peacekeeping to maintain peace in the aftermath of civil war, it must help combatants overcome these bargaining obstacles. Previous studies indicate that the presence of a UN peace operation, or foreign intervention more generally, assists in securing post-conflict peace (Fortna 2004, 2008; Gilligan and Sergenti, 2008; Walter, 1997, 2001).

However, we have relatively little knowledge, especially with regard to UN peacekeeping, as to what qualities or characteristics of interventions lead to durable peace. Research on the effect of UN peacekeeping on peace duration has generally not captured the distinct qualities of UN PKOs that may be consequential to peace and conflict processes beyond rather simplistic dichotomous measures of a mission’s presence or absence in the host country. Noting the effect of a PKO’s presence in the post-conflict state reveals little about which aspects of peacekeeping missions are most closely associated with the extension of post-war peace. ³ Not much can be said about the components of peacekeeping missions that yield positive outcomes. Recent studies, however, indicate the importance of mission capacity and constitution for protecting civilians, ending battlefield violence, increasing cooperation between combatants, improving the quality of immediate post-conflict peace and containing conflict contagion (Beardsley, 2011; Hultman et al., 2013, 2014; Kathman and Wood, 2015; Ruggeri et al., 2013).

Consider Figure 1, which plots the deployment of armed PKO troops to two missions in Africa from late 1999 through 2002. The UN mission in Sierra Leone (UNAMSIL) dramatically escalated its troop commitment in early 2000 and again in mid-2001. By comparison, the UN mission to the Democratic Republic of the Congo (MONUC) made a much smaller and more incremental troop increase over the course of its deployment. Dichotomous indicators of a mission’s presence or absence in the conflict state only indicate that UNAMSIL was present across the entire time period, whereas MONUC did not arrive until late 1999 and remained until 2002. If the simple presence or absence of a mission is enough for the UN to be effective, then these two missions are indistinguishable and should have an equal chance of success. Yet, evidence suggests that the escalation of troops to Sierra Leone resulted in a great deal of progress, helping end the bloody conflict and secure the peace that followed (Findlay, 2002). MONUC on the other hand saw substantially less success, at least in the early stages of its deployment. Empirical models that employ categorical indicators of peacekeeping do not capture these critical differences between missions like UNAMSIL and MONUC. Therefore such models cannot assess the consequence of these differences for the stability of post-conflict peace. ⁴

OPEN IN VIEWER

Figure 1.

Monthly number of UN peacekeeping troops.

Not only do peacekeeping missions differ in the overall number of personnel and timing of deployment, but they also differ in the personnel type deployed. PKOs are generally outfitted with three personnel types. In addition to armed troops, UN missions are commonly deployed with armed police units and unarmed observer personnel. Each of these personnel types serves distinct purposes in attempting to achieve the mandated goals of each operation. If the ability of missions to fulfill their objectives is a product of the tools made available to them, then variations in types of personnel deployed to PKOs should be particularly important to understanding which missions are more or less effective at securing peace following an end to civil conflict.

For UN PKOs to effectively assist the former combatants in avoiding a return to civil war, operations must be able to overcome commitment problems and information asymmetries. UN missions and the peacekeeping personnel that outfit these operations do this in two important ways. First, peace operations seek to separate the combatants and reduce their short term fears that the opponent will embark on another military offensive. By interceding between the former belligerents, UN missions essentially solidify the status quo power balance between combatants (Hultman et al., 2013). Given that confrontation on the battlefield is the primary means by which combatants attempt to improve their relative power, destabilizing power shifts that motivate commitment problems are diminished when PKOs effectively separate the belligerents, at least in the short term. As missions remain deployed to fulfill this task, security guarantees offered by the operation should improve the willingness of the former combatants to disarm and demobilize.

However, it is not simply the presence of a peace operation that is sufficient for achieving this goal. The UN’s ability to overcome commitment problems between the belligerents is a product of the UN’s commitment to the conflict state. Former combatants judge the strength of the UN’s commitment by evaluating the quality and strength of the personnel deployed (Findlay, 2002; Gberie, 2005). The UN is able to signal its commitment to post-conflict peace by committing a greater number of peacekeeping personnel to the host state. UN resolve is revealed as more personnel arrive. Larger deployments make the withdrawal of PKO forces more politically costly to the UN. Additionally, robust deployments indicate the UN’s willingness to bear costs in its deployment, essentially revealing the UN’s resolve for maintaining the mission in pursuit of the mandated goal of avoiding conflict recidivism.

Furthermore, PKOs require a particular type of personnel to overcome commitment problems between combatants by interceding on the battlefield. Military troops are responsible for preventing a resurgence of battlefield hostilities. UN troops are deployed to the frontlines, are often equipped with arms and heavy weaponry, and are trained to engage in combat. Peacekeeping troops are therefore the personnel type most closely associated with thwarting battlefield activities between the former combatants. Logically, as more UN troops are deployed to a post-conflict state, they can engage in more activities that reduce opportunities for renewed conflict. Armed troops are the type of personnel that is tasked with separating combatants, and having a larger deployment of troops enhances the UN’s ability to separate the factions in several important ways. With an increasingly large number of troops, UN missions can patrol larger territories, expand the buffer zone between combatants, and offer a more visible indication of the UN’s commitment to peace. Additionally, larger numbers of deployed troops increase the capacity of missions to disarm rebel and militia groups (Holt et al., 2009). By doing so, UN troops reduce the physical capacity of former combatants to re-engage on the battlefield. Therefore, larger numbers of troops deployed to post-conflict environments should improve the UN’s ability to overcome commitment problems between the combatants, improving the ability of the UN to secure post-conflict peace.

Finally, even if the combatants are motivated to return to open hostilities, preferring continued combat to peace processes, the deployment of ever larger numbers of troops should impede their ability to achieve this. The expansion of buffer zones and the intercession of blue helmets reduce the combatants’ tactical options to engage in violence. Efforts at continued hostilities in spite of large troop deployments could also cause the UN to use its troops to more aggressively confront these hostilities. For instance, UN responses to continued violence during recent deployments in the Ivory Coast offer examples of more aggressive policies used by UN missions to halt hostilities forthrightly. At times, these missions forcibly pursued separation, disarmament, and demobilization strategies even when one or more factions resisted these practices. Larger troop deployments are more fully able to engage in these efforts, generating negative consequences for combatant factions and increasing their costs of pursing their political goals through combat. The above discussion motivates the following hypothesis:

Hypothesis 1: As the UN deploys more armed troops to a post-conflict state, the duration of peace increases.

In addition to commitment problems, rationalist theories indicate that combatants have asymmetric information on their own strength, resolve, and reservation points, and have incentives to misrepresent this information to their opponents in hopes of attaining more attractive terms. Fighting helps to reveal this information, as the belligerents can do little to conceal battlefield outcomes. While combat may indeed reveal information about the combatants’ strength, resolve, and reservation points, the terms of an initial conflict termination may not sufficiently satisfy the interests of the factions to yield lasting peace. A UN mission may help improve information sharing between combatants. However, not all peacekeeping personnel are suited to facilitate a productive information-sharing process. For instance, police units are primarily deployed behind the frontlines to secure population centers, train indigenous police, and protect civilians from combatant abuses. While beneficial for the sake of securing civilian protection (Hultman et al., 2013), these responsibilities do not improve the combatants’ ability to judge the strength and resolve of their opponent.

UN observer personnel, on the other hand, are tasked with several responsibilities that directly facilitate information sharing. While combatants may prefer to keep private information about their military capabilities and their potential intentions to return to battlefield combat, military observers are tasked with monitoring the implementation of peace terms and sharing their findings. In this way, the former combatants are incentivized to adhere more strictly to the peace terms knowing that information about their (non)compliance will be relayed to the former adversary. For instance, post-conflict elections and the unification of disparate rebel and government armed forces are often unstable and tense processes. The presence of observers to oversee unification efforts and offer unbiased information on the fulfillment of agreements is an important step in securing peace. Additionally, after armed peacekeeping troops effectively separate combatants on the battlefield, observers play an informational role in overcoming commitment problems. While belligerents may have a difficult time demobilizing, observers can verify disarmament and decrease each side’s expectation that the opponent will renege on the agreed demobilization terms. Indeed, this is a common task performed by observer personnel across missions. While some PKOs are specifically tasked with information-sharing mandates, the ability of observers to witness developments in the host country and share this information with the factions is not limited to pre-deployment agreements about peacekeeping procedures.

The ability of the UN to witness and scrutinize post-conflict developments and assist in the sharing of information is in large part determined by the number of observer personnel it commits to each host state. As the number of observers deployed increases, the number of polling stations monitored during elections increases, larger swaths of contested territory can be surveyed, and more disarmament and demobilization processes can be verified and reported. As a result, larger numbers of UN observer personnel committed to post-conflict states should improve the UN’s ability to overcome information asymmetries that distress the combatants’ ability to adhere to the peace. The above discussion leads to a second hypothesis:

Hypothesis 2: As the UN deploys more unarmed observers to a post-conflict state, the duration of peace increases.

In the following sections, we describe our research design and test the above hypotheses. ⁵

Research design

To assess the above hypotheses, we construct a post-conflict peace duration dataset that includes monthly observations of peace following all African civil wars in the 1989–2010 period. While we would, of course, prefer to test our arguments on a global dataset that spans many decades, we are limited by data availability, as several of our variables are measured at the monthly level. To fully leverage our data, we restrict the spatio-temporal domain to Africa in the post-Cold War period. Still, our sample is attractive. Given the high prevalence and wide variety of civil wars in Africa during this period, we are able to control for several common correlates of victimization, and we are confident that we have not selected a sample of wars that would meaningfully bias our results. Further, there were many instances of UN intervention in African conflicts in this period, and these PKOs were diverse in their chosen means of intervention. Additionally, our temporal domain follows the effective end of the Cold War which removed the primary political barrier to Security Council action. Thus, in both theory and practice, more robust peacekeeping methods were made newly available, including Chapter 7 forms of peacemaking and enforcement. This is evidenced by the explosion of the number and type of peacekeeping efforts in Africa during this time. In this sense, our spatio-temporal domain is a rather desirable one.

We measure conflict and post-conflict peace at the government–rebel group dyad level because different rebel groups terminate their hostilities with the government at different times, initiating peace at unique points in time within the same overarching conflict (Nilsson 2008). Likewise, individual rebel groups can re-initiate their combat with the government independent of one another, producing dyad-specific peace cessations.

We rely upon the Uppsala Conflict Data Program’s (UCDP) Dyadic Dataset v.1-2013 (Harbom et al., 2008) to indicate the list of civil conflicts after which peace duration is coded. For inclusion in the dataset, each conflict dyad must have crossed a 25 battle deaths threshold and subsequently terminated hostilities. The cessation of conflict is determined by the UCDP Conflict Termination Dataset (Kreutz, 2010). The beginning of a peace duration spell is recorded as the month in which conflict hostilities ended. Conflict terminations occur in one of three primary ways according to Kreutz (2010): negotiated settlements, the victory of one combatant over the other, or the reduction in battlefield violence for one year such that the hostilities do not reach or exceed the 25 battle deaths threshold. ⁶ Our dependent variable, peace duration, is then coded in monthly observations until war recurs or until an observation is right-censored by the dataset’s temporal domain.

Control variables

In addition to the variables of interest, we control for several other factors in our statistical models. First, we account for characteristics of the post-conflict state. Specifically, we measure each post-conflict state’s wealth, as low levels of wealth have been associated with civil war onset. Gross domestic product per capita (GDP/Capita) data are taken from Gleditsch (2002), and the values are log-transformed. We also measure the quality of each host state’s Regime Type using the 21-point autocracy–democracy scale provided by Polity IV (Marshall et al., 2011). To control for the possibility of a nonmonotonic effect on a state’s political institutions, we include a squared term for this variable in the models, which should help to reveal the effect of anocratic forms of government experiencing post-conflict peace durations that differ from autocratic and democratic states, as has been found in the civil war onset literature (Hegre et al., 2001). Next we account for each host state’s total population size. Larger countries are believed to be more difficult to govern, especially following civil war, and larger populations offer a larger pool from which rebels can recruit followers in order to reignite their violent challenge to the government (Elbadawi and Sambanis, 2002). Data on population size are taken from the disaggregated Composite Index of National Material Capabilities (Singer et al., 1972). This variable is log transformed.

We also include several controls to represent features of the prior conflict. To account for the overall costs paid by the factions in the previous conflict, we include Previous War Duration and Previous War Deaths, which code the length in months of the previously terminated conflict and the total number of combat casualties of the previous conflict, respectively. Longer and more costly conflicts are likely to reveal more information about combatant resolve and reservation points. The revelation of this information may then make any resolution to the conflict more lasting. Alternatively, extraordinarily long wars may indicate little or no overlap in the combatants’ reservation points, and very violent hostilities may harden animosities and shorten the duration of peace that is eventually achieved. Data on government and rebel battle fatalities are taken from the UCDP Georeferenced Event Dataset (GED) (Sundberg and Melander, 2013). Furthermore, we include Rebel Political Wing as a dichotomous indicator of whether each rebel group was represented by a legal political party during the conflict. Ex ante legal access to the political process should have the effect of extending peace duration upon the end of fighting. Data for this variable are taken from Cunningham et al. (2009).

Finally, we control for several qualities of the extant peace. While the presence of peace or conflict is defined by the absence or occurrence of battlefield deaths, one-sided violence by the former belligerents against civilians often continues beyond the cessation of battlefield hostilities. The instability produced by such violence may have the effect of undermining the peace, instigating a renewal of battlefield combat. We thus aggregate one-sided violence events from the UCDP GED to give us a count of the number of civilians killed by the former combatants in each post-conflict month. ⁷ We also code the means by which each civil war ended using the UCDP’s Dyadic Termination Dataset. Specifically, we code whether the current peace spell is the product of a negotiated settlement by the combatant parties, the victory of one side over the other, or the reduction of conflict violence to the point of low activity (e.g. the conflict produces fewer than 25 battle deaths in a year). We suspect that the outright victory of one faction yields more stable peace. Conflicts ending in low activity and negotiated settlements, on the other hand, may yield more tenuous peace because both sides often retain the ability to reinitiate combat. Table 1 reports descriptive statistics on our dependent variable, independent variables of interest, and controls.

OPEN IN VIEWER

Table 1.

Descriptive statistics

Variable	Mean	Standard deviation	Minimum	Maximum
Peace Duration	83.69	62.37	1	271
UN Troops	1422.54	3951.98	0	29,209
UN Observers	44.86	127.91	0	1039
UN Police	109.47	404.56	0	5177
Previous War Duration	38.77	57.75	1	318
Previous War Deaths	2685.87	8036.36	0	43,470
One-Sided Violence	1.97	44.48	0	4016
GDP/Capita(ln)	6.88	0.679	4.89	8.49
Population(ln)	9.27	1.16	6.23	11.87
Regime Type	0.19	3.82	−8	9
Regime Type2	14.70	17.47	0	81
Rebel Political Wing	0.24	0.43	0	1
# of Previous Episodes	0.17	0.63	0	5
Negotiated Settlement	0.33	0.47	0	1
Victory	0.21	0.41	0	1
Low Activity	0.33	0.47	0	1

Empirical results

Since we are interested in the duration of post-conflict peace, we employ a Cox proportional hazards model with robust standard errors clustered on the conflict dyad to judge the effect of the covariates. This model allows us to test the effect of the covariates in extending or shortening the duration of peace. The Cox model makes no assumptions about the shape of the baseline hazard. However, we ran several models using various modeling specifications that make different assumptions about the shape of the baseline hazard and the results were routinely consistent with those reported below for our variables of interest.

The results of the duration analyses are reported in Table 2. Model 1 considers only the effect of the peacekeeping mission composition variables. We do not include the control variables in this model in order to present initial statistical evidence of a relationship between peacekeeping personnel deployments and the duration of post-conflict peace. Model 2 is fully specified, including all post-conflict peace spells, no matter the form of conflict termination. Model 3 includes each variable, but limits the sample to conflicts that terminated via negotiated settlement or low activity, as UN PKOs are more likely to be present under these conditions relative to those conflicts that end in the victory of one side.

OPEN IN VIEWER

Table 2.

Cox duration analyses of UN peacekeeping mission composition and peace duration following civil wars in Africa, 1989–2010

Variables	Model 1: all conflict terminations	Model 2: all conflict terminations	Model 3: settlement and low activity terminations
UN Troops	–0.0005**	–0.0005**	–0.0006**
	(0.00008)	(0.0001)	(0.0002)
UN Observers	–0.002	–0.0005	–0.007†
	(0.005)	(0.004)	(0.004)
UN Police	0.0007	0.0005	0.0005
	(0.0005)	(0.0005)	(0.0005)
Previous War Duration		0.001	–0.004
		(0.003)	(0.004)
Previous War Deaths		1.87 × 10−6	0.0002**
		(0.00005)	(0.00007)
One-Sided Violence		–0.020	–0.016
		(0.020)	(0.020)
GDP/Capita		–0.025	–0.442
		(0.252)	(0.293)
Population		0.341†	–0.426*
		(0.197)	(0.211)
Regime Type		–0.009	–0.097
		(0.056)	(0.063)
Regime Type2		–0.010	–0.013
		(0.011)	(0.013)
Rebel Political Wing		–0.075	0.231
		(0.393)	(0.407)
# of Previous Episodes		0.385*	0.760**
		(0.158)	(0.216)
Negotiated Settlement		0.031
		(0.641)
Victory		–1.511
		(1.737)
Low Activity		0.931†
		(0.542)
Observations	13,813	13,436	8763
Number of subjects	132	129	94
Number of failures	36	36	31
Wald χ 2	42.95**	63.83**	55.19**
Log pseudo-likelihood	–163.29	–146.83	–118.49

Coefficients and standard errors clustered on conflict dyad. **p ≤ 0.01; *p ≤ 0.05; †p ≤ 0.10.

The analyses consistently support hypothesis 1 and indicate that larger UN troop commitments reduce the risk of civil conflict recurrence, indicated by the negative coefficient for UN Troops. As the number of UN troops deployed to a post-conflict state increases, the risk of the state reverting back to civil war decreases. In other words, as the amount of UN troop personnel increases, the duration of peace after a civil war is extended. This is an impressive result, as it is consistent across substantially different model specifications. In particular, we note that, even in model 3 when limiting the sample to conflicts that ended in a negotiated settlement or by low activity—cases in which the UN is both more likely to be involved and be faced with former combatants that remain armed and capable of returning to the battlefield—the monthly number of UN troops deployed to the post-conflict state still reduces the risk of conflict relapse. This finding challenges UN skeptics and shows that UN peacekeeping troops help to establish conditions that are conducive to a more stable peace. Previous research shows that the UN is more likely to commit peacekeeping missions to the most difficult cases (e.g. Fortna, 2008; Gilligan and Stedman, 2003). The number of troops deployed to these conflicts also reflects the complexity of the situation. Considering the selection of UN troops to the most difficult cases, it is noteworthy that we still find that more fully equipped missions are increasingly effective in prolonging the peace.

Figure 2 displays the estimated smoothed hazard function from the analysis reported in model 3, which limits the sample to conflicts that ended in a negotiated settlement or low activity. The hazard function is the rate at which a hypothetical post-conflict state is “failing”, or reverting back to civil war. Control variables are held at their mean or modal values. The figure presents the estimated hazard functions when zero troops are deployed, 1500 troops are deployed and 3000 troops are deployed. The hazard function is generally declining, indicating that, as peace endures, the chance of a state falling back to civil conflict decreases. The figure also shows that the presence of 3000 troops leads to an approximate 85% decrease in the estimated hazard rate compared with a deployment of zero troops, a substantial reduction in the likelihood of civil war recidivism that results from the deployment of a rather modestly composed mission.

OPEN IN VIEWER

Figure 2.

Smoothed hazard function estimates for post-civil conflict states with varying levels of UN military troop deployments. Note: Hazard function estimates derived from analysis of model 3.

Our results indicate that the UN cannot simply dispatch a contingent of unarmed observers to host nations if it wants to provide sufficient support for stabilizing post-conflict peace processes. The UN should instead deploy larger amounts of armed troops, which seem to be closely associated with mitigating the commitment problems that lead to the recurrence of civil war. This should not necessarily be interpreted as information asymmetries being less serious obstacles to peace. However, our results imply that UN observers are conceivably ineffective in solving them. Instead, mediation efforts may be much more suitable for overcoming information problems during peace processes (Regan and Aydin, 2006). The strength of UN peacekeeping appears to lie in its ability to alleviate commitment problems, which requires the deployment of armed troops.

In addition to the results reported in Table 2, we tested the robustness of our findings by replicating models 2 and 3 in several ways. First, we included a dichotomous variable to account for the intervention of a regional intergovernmental organization (IGO), such as the African Union or the Economic Community of West African States, in a peacekeeping role using data from the Dynamic Analysis of Dispute Management Project’s Third Party Peacekeeping Missions Data Set v. 3.0 (Mullenbach, 2013). If the effectiveness of UN missions is a partial product of UN PKOs piggybacking on the heavy lifting of other peacekeeping missions, then the negative effect of UN Troops may falsely be reflective of otherwise effective regional IGO interventions. In each replication of our models, the negative and significant effect of UN Troops remained. However, the regional IGO variable reported an insignificant coefficient.

We also included a variable for the robustness of each mission mandate. Mandates define both the goals sought by each PKO and the lengths to which they can go in achieving these goals. Mission mandates may be important to the extent that they restrict or endorse the types of actions that PKO personnel can take in securing the peace. We coded a dichotomous variable to delineate those mission mandates that gave peacekeepers the authority to use “any means necessary” to fulfill their mandated goals. This variable produced a negative and significant coefficient in models 2 and 3, indicating that more robustly mandated missions extend the duration of peace. However, the variable’s inclusion in the model did not affect the results produced by UN Troops. ⁸ Finally, we considered different specifications of our PKO personnel variables. For instance, we took the natural log of each of the three personnel variables in order to normalize their distributions. Including the logged versions in our models yielded results for UN Troops that are consistent with those reported in Table 2. ⁹

With regard to the control variables, there are several important findings across models 2 and 3. The number of previous conflict episodes has a positive effect on the hazard of peace failing. In other words, as the number of times that a government and rebel group have abandoned peace and returned to war increases, the shorter each next peace spell is expected to be. This probably indicates an inability of the rebels or government to agree to terms that both sides find desirable, resolve the commitment problem between them or prevent the exploitation of peace spells as periods in which one or both factions rearm and recuperate for the next round of battle.

Additionally, Population reports a positive coefficient that is significant at the 0.1 level in each model. As the size of a conflict state’s population increases, the duration of peace following conflict is shortened. This result agrees with a consistent finding in the civil war onset literature. We surmise that larger populations offer larger pools of people from which new rebel soldiers can be recruited, and larger populations should offer greater opportunities for rebels to “go underground”, hiding amongst civilians and convalescing during peace spells before reigniting their violent challenge to the government. Also, the likelihood should be lower that any post-conflict settlement can effectively satisfy the multitude of interested factions which should increase as a state’s population size increases.

Model 2 also shows that a previous conflict ending via a low-activity outcome shortens the duration of post-conflict peace. We suggest this is the case because the belligerents involved in conflicts that terminate owing to low activity often maintain their political and military organization. They thus have the ability to reignite overt battlefield combat even if the conflict between them has been dormant for a period of time. Furthermore, when fighting has fallen below the 25 battle deaths threshold to be considered an active conflict, this does not necessarily mean that the issues over which the combatants have confronted one another have been resolved. The lack of resolution and the continued presence of combatant forces thus appear to be associated with an increased probability of conflict reemergence.

In model 3 we also find that Previous War Deaths is significant at the 0.1 level and is positively signed. This indicates that more violent wars have shorter post-conflict peace spells. Rather than battle deaths revealing information about the combatants’ strength and resolve, there is some initial, albeit weak, evidence that more violent wars yield shorter post-conflict peace periods. It may be that increasingly hostile civil wars engender unstable post-conflict periods as the collective memory of the previous war may make it more difficult for the former combatants to re-assimilate to living peaceably with one another again.

Across models 2 and 3, the remaining variables do not report statistically significant results at conventional levels. First, features of the violence are insignificant. Longer war durations may be correlated with an increased amount of information revealed about combatant capabilities and resolve, and this may have the effect of stabilizing the post-conflict peace. At the same time, long wars may also be indicative of the hardened positions of the combatants, positions that may thus be associated with war recidivism. These cross-pressures may thus be responsible for the insignificant coefficient reported for Previous War Duration. The level of civilian victimization is also unassociated with peace duration. While this violence may at times be destabilizing, it may also serve to consolidate control over the population, producing a null effect on conflict reemergence.

Rebel Political Wing is also insignificant. We suggested that ex ante legal access to political processes should help to extend peace following conflict. However, this is not the case according to our analyses. Future analyses may consider whether this variable is consequential in conjunction with the means by which civil wars end, as the presence of a political framework through which rebels are organized may be most relevant when the rebels remain party to the political system through a negotiated resolution or through their own military victory.

Next, the remaining conflict state characteristics are insignificant, indicating that a post-conflict country’s regime type and level of wealth are not associated with the re-emergence of fighting. These variables often report significant results in the civil war onset literature. However, the post-conflict nature of our analyses probably makes it more difficult to find an association between these variables and peace duration. Civil war tends to destroy wealth and democratic political institutions (Collier et al., 2003), thus reducing the variability between cases along wealth and regime type dimensions. As a result, these factors are likely to be more closely associated with the initial conflict onset.

Finally, the remaining conflict outcome types do not significantly affect the duration of post-conflict peace. Whether achieved through victory or a negotiated resolution, these outcomes do not appear to affect the ability of the factions to consolidate peace. The insignificant coefficient for Victory is somewhat perplexing, as one would expect there to be little in the way of recourse for defeated groups to re-initiate hostilities. However, even in defeat, some groups are able to maintain their core organization that can be reconstructed to again challenge its opponent after a brief respite. Additionally, while Negotiated Settlement is insignificant, this may be somewhat unsurprising given recent research that points to the importance of the peace terms as predictors of an agreement’s effectiveness in securing the peace (Mattes and Savun, 2010).

Discussion and conclusion

As we stated at the outset, recent research has largely come to the conclusion that post-conflict UN peacekeeping is an effective mechanism for solidifying the stability of post-civil war environments. However, by and large, this work offers limited prescriptions to policy-makers who are tasked with composing these missions to most effectively achieve their objectives. Since prior work employs rather blunt dichotomous renderings of UN PKOs in statistical models, little can be said about the characteristics of these missions that are associated with the persistence of post-conflict peace. From a policy perspective, these findings offer little in the way of actionable information on how the international community should equip UN missions.

We have sought to provide greater nuance to the existing quantitative literature that policy-makers may find more instructive. We have theorized on and tested empirically the means by which UN peacekeeping is or is not effective at keeping the peace by taking the characteristics of the missions into account. While some studies have evaluated the effect of various types of missions on peace and conflict processes (Doyle and Sambanis, 2006; Fortna, 2008), these categories do not fully reflect the capacity of the missions to manage security dilemmas and commitment problems that so often spoil peace following the end of civil war. In fact, previous research draws different conclusions about peacekeeping effectiveness when exploring differences in mission mandate and type. While Doyle and Sambanis (2006) show evidence that missions with enforcement or multidimensional mandates are more likely to achieve success, Fortna (2008) finds traditional missions to be the most successful. We believe that one reason for these somewhat divergent findings is that relying on broad peace operation categories abstracts from the functional capacity of missions to pursue and maintain peace. We suggest that it is these important differences in the capacity to deal with the security challenges to former belligerents that is critical to our understanding of peacekeeping effectiveness.

Moreover, the capacity of missions changes over the course of their deployment, and these changes modify the ability of missions to provide the security guarantees that are necessary to help the belligerents overcome commitment problems. Our data and analyses differ from more traditional studies of peacekeeping, capturing the dynamism in PKO deployments in ways that are impossible when using categorical treatments of the presence or absence of peacekeeping operations. By measuring the actual number of troops on the ground and the ways in which these troop commitments change over the course of mission deployments, we are able to capture the UN’s actual ability and resolve to provide security in the aftermath of civil war. In line with this, our results indicate that UN missions with a greater number of troops are also better able to support peaceful solutions to civil wars.

We believe that these findings will be consequential to decision-makers at the UN who are intent on designing peacekeeping operations to be most fully equipped to secure post-conflict peace. Armed with this information, members of the Security Council or policy-makers in the Secretariat should be able to recognize that increasing the deployment of observer and police personnel is not likely to yield positive outcomes, at least in terms of stabilizing host states such that the former combatants do not return to overt conflict. Instead, more fully informed policy-makers should recognize the benefits of deploying armed troops in sufficiently large numbers to post-conflict states in order to maintain the peace that is won with the end of civil war hostilities.