Managing Mistrust

Abstract

How many peacekeepers are needed to keep the peace? Under what conditions are local governments and rebel forces more willing to cooperate with an intervention force? From a theoretical perspective in which the main role of peacekeepers is to assist local actors in overcoming their commitment problems and mistrust toward each other, it follows that sufficiently robust missions should positively affect levels of cooperation. Furthermore, any effect should be conditional on the local balance of power, that is, the military leverage between government and rebel forces. Relatively weak rebel groups—facing a stronger government—should be more willing to cooperate with larger missions. In the empirical analysis, using newly collected event data on United Nation (UN) peacekeeping operations from 1989 to 2005 in African civil wars, the authors find support for conditional effect of robust peacekeeping: there is more cooperation with UN peacekeepers when the rebels are weak.

Keywords

peacekeeping civil wars United Nations Africa

How many peacekeepers are needed to keep the peace? When the United Nations Mission in Sierra Leone (UNAMSIL) concluded in 2005, it was hailed as one of the most successful UN missions to date (Olonisakin 2008, 1-2). Yet, few people would have predicted such a happy ending for UNAMSIL in the early days of the mission. For the first two years, the Revolutionary United Front (RUF) challenged UNAMSIL militarily leaving the mission faltering, a situation reminiscent of other UN failures in Rwanda and Somalia. A combination of domestic and international factors, most notably an intervention by UK forces, led to a shift in the power dynamics between the RUF and the government of Sierra Leone. At the same time, UNAMSIL was restructured and its authorized strength was increased from 6,000 military personnel in October 1999 to a maximum deployment of 17,500 in March 2001.¹ The commitment of UNAMSIL to engage in dialogue with the RUF backed up by a stronger and more competent military force moved the peace process forward and led to a surprising change in the fortunes of UNAMSIL (Olonisakin 2008, 108).

The case of UNAMSIL illustrates that peacekeeping is a dynamic process, where the ability to affect the underlying bargaining structures and actors’ behaviors and positions determines the outcome of the mission (Fortna 2008). In response to emergent crises or political opportunities, the Security Council has basically two instruments at its disposal: it can revise the mandate of the mission and/or amend its authorized strength. Here, we focus on the latter, since the actual size of the mission is arguably the strongest observable signal of the UN’s resolve.² Whereas mandates can be amended relatively easily, it is much more difficult to rapidly establish an effective UN presence on the ground; especially, when the security situation is deteriorating. The presence of a sufficient number of peacekeepers is therefore essential to allow the UN mission to fulfill what is arguable its main objective, namely to assist the local actors to overcome mistrust and commitment problems (Walter 1997; Fortna 2008). Mistrust can not only obstruct an agreement but also derail its implementation (Kydd 2006, 459). The success of the UN in minimizing the mistrust between rebels and the government depends on the credibility of its commitment to act when necessary. It is therefore important to know under what conditions the presence of peacekeepers generates cooperation with the mission and mitigates the mistrust between the rebels and the local government.

To examine the dynamics of peacekeeping operations (PKOs), we focus on events that take place during the mission; in particular, we analyze cooperation observable in the interaction between the three main actors in the context of peacekeeping, namely, the UN mission, the government, and the rebel groups.³ Cooperation encompasses a wide range of actions from mainly symbolic to permissive or even active positive engagement with the peace process. The second United Nations Angola Verification Mission (UNAVEM) mission in Angola provides an example of “symbolic cooperation” when in a meeting with UN officials, representatives of UNITA (National Union for the Total Independence of Angola) reaffirmed their commitment to the relief effort, even though only a few days later it was broken in an attack on a relief convoy (S/26434,6). An example of “permissive cooperation” is that in the autumn of 2000, during a tense period of the civil war in the Democratic Republic of Congo (DRC), child protection advisers part of the UN mission MONUC were still able to undertake field trips across the DRC and to meet with “all child protection partners” (S/2000/1156, 11). Finally, later in the deployment of MONUC instances of “positive engagement” can be found, for example, in the demobilization and repatriation of excombatants and their dependants (S/2003/211, 5).

In contrast to previous studies, we do not evaluate the overall success or failure of a mission following its conclusion (Diehl, Reifschneider, and Hensel 1996; Doyle and Sambanis 2000, 2006; Fortna 2008), or its impact on civilians and human rights (Hultman 2010; Murdie and Davies 2010). These quantitative studies of peacekeeping use the overall mission as unit of analysis and draw on data aggregated over the entire conflict. Further, they tend to rely on crude assessments of the characteristics of the mission, often measuring only whether it was present or not. As a result, features of the conflict or the country mainly explain the observed outcomes rather than the specifics of the mission.

For theoretical arguments on how and when robust peacekeeping matters, we rely on several extant case studies of peacekeeping operations, which provide rich and detailed information on the experiences of the peacekeepers and the “peacekept” (United Nations 1996; Durch et al. 2003; Paris 2004; Pouligny 2006; Olonisakin 2008; Howard 2008; Autesserre 2010; see also Doyle and Sambanis 2006; Fortna 2008). We aim to test key insights on the basis of comparative research using systematically collected data across a number of cases for a longer time period. Therefore, our empirical analysis uses newly collected events data covering UN peacekeeping missions in Africa from 1989 until 2005.

The main finding is that the size of the mission matters: rebel groups and local governments are more willing to cooperate with larger UN missions. Most notable, however, is that the “size” effect is conditional on the distribution of military power or the relative balance between rebels and government. If the government is substantially stronger than the rebels, there is a larger marginal effect of the presence of UN troops on cooperation. The remainder of the article is divided into four parts. In the section Peacekeeping, Mistrust, and Vulnerability, the theoretical argument is developed further. The section on Research Design discusses the data and the operationalization of the variables used in the analysis. Following, the section on Empirical Findings contains the results of the empirical analysis including extensive robustness check, and followed by the Conclusion.

Peacekeeping, Mistrust, and Vulnerability

Enforcement and commitment problems are plausible explanations for the failure of agreements to settle conflicts and the frequent recurrence of violence after an agreement has been reached (Lake and Rothchild 1996; Walter 1997; Fearon 1998; Kydd 2000, 2006; Powell 2006). Even assuming near-perfect information about relative military capacity after years of fighting, there often remains doubt about the actors’ willingness to enforce agreements and to uphold the achieved institutional balance.

Agreements also remain vulnerable to renegotiation. Parties may indicate their commitment to respect a settlement, but the actual implementation process may leave each side vulnerable to a relative deterioration of its (military) capabilities. It may fear that the other side will attain the capacity to renounce the agreement. The uncertainty on whether the stronger party will cooperate in the future can lead to a failure of the peace agreement and a recurrence of violence. Even the side originally perceived as weaker may be able to take advantage of vulnerabilities of the stronger side, since the latter may relax its vigilance in anticipation of the cooperation and implementation of the agreement. In this context, external actors can minimize the uncertainty that one or both adversaries will renege on the agreement and can help the involved parties to build trust (Kydd 2006, 249). Accordingly, UN peacekeeping operations (UN PKOs) act as guarantors of an agreement.

The UNAMSIL, for example, was established to help with the implementation of the Lomé Peace Accord. As part of peacekeeping activities, monitoring reduced uncertainty about the willingness of both the government and the rebels to abide by the agreed settlement. The latter mattered particularly for the rebels because the disarmament program put them in a vulnerable position, exposing them to possible government demands to renegotiate the terms of the agreement. At the same time, the Kabbah government was quite weak relative to the RUF, while the sudden death of the chief of defense Maxwell Khobe in April 2000 pushed the Sierra Leone Army into confusion, further undermining the position of the government (Olonisakin 2008).

For the government and the rebels, the presence of peacekeepers in effect alters the cost-benefit assessment of using either a violent or more cooperative approach to achieve their political aims. Moreover, peacekeeping has evolved since the end of the cold war, and “new” peacekeeping encompasses peace-building efforts and attempts to create and sustain the conditions under which the political process of reconciliation can move forward. Fortna (2008) argues that ideally the United Nations modifies the incentives and makes it more costly for local actors to renege on a settlement and to exploit the vulnerabilities of their opponents. Moreover, peacekeepers target spoilers that may pose a threat to future cooperation. Peacekeepers are not merely observers but in fact also manipulate the postagreement interaction among the domestic actors. They facilitate cooperation by minimizing the vulnerability of the weaker side and by signaling a willingness to punish whoever defects from previous commitments.

Fortna (2008, 178) concludes that “peacekeeping intervenes in the most difficult cases, dramatically increases the chances that peace will last, and does so by altering the incentives of the peacekept, by alleviating their fear and mistrust of each other, by preventing and controlling accidents and misbehavior by hard-line factions, and by encouraging political institutions.” A substantial presence on the ground would seem necessary to achieve any of these objectives.⁴

In the dynamic context of intrastate conflict, government and rebel forces often use violence to either improve their positions in the bargaining space or to signal their intentions and resolve (Hultman 2007; Kalyvas and Kocher 2009). Levels of violence are unlikely to remain constant over time within the context of a peacekeeping mission but vary according to the strategic interests of the actors. Since local actors continue to interact strategically, peacekeepers effectively become a third party in an occasionally violent bargaining process. In this context cooperation toward the UN mission, such as cooperation with demobilization or security sector reform, indicates a willingness of local actors to accept the established balance of power. In contrast, when local actors directly challenge UN peacekeepers, the strategic goal is to destabilize the mission and undermine the ability of the PKO mission to enforce the peace. For instance, in April and May 2000, the attacks of RUF in Sierra Leone against UN troops and the camps used for disarmament, demobilization, and reintegration (DDR) were indicative of a systematic campaign to undermine UNAMSIL (Olonisakin 2008).

In order to succeed, the UN relies on the credibility of its commitment to act when necessary. To establish a credible threat against spoilers or future changes of the negotiated agreement, the UN has to demonstrate the willingness of the organization (and the country members) to incur costs in order to protect the weaker side. In other words, the United Nations has to signal to all the parties that it has a stake in the peace process and that the peacekeepers are willing to use force when necessary. Hence, the United Nations have to make it clear that avoiding the use of force is not always the organization’s most preferred strategy (Kydd 2006).

The size of the mission is a clear and observable signal of the United Nations’ resolve. The willingness of member states to either contribute or financially support a large number of troops shows their commitment to the implementation of the mandate of the mission. By limiting funding to 6,000 troops resulting in a thin deployment on the ground the Security Council showed that Sierra Leone remained a lesser priority. Following the faltering performance of UNAMSIL and a decisive intervention of British forces in Operation Palliser, the UN Secretariat and the Security Council reconsidered the role of the mission. The mission was restructured with a steady increase in the deployment of better-equipped and trained troops that could operate in RUF-controlled areas. Between February 2000 and March 2001, the authorized military strength of UNAMSIL was raised almost threefold making possible effective control of spoilers and even offensive military operations against the RUF. Increased confidence in the ability of the mission strengthened the control of the central government (Olonisakin 2008, 91-93).

In contrast, an obvious weakness of the first UNAVEM-I was that it employed only 350 unarmed observers and 400 electoral observers to monitor the demobilization process of 200,000 troops. The disproportionally small size of the mission was widely perceived to signal a lack of UN commitment (Howard 2008, 36). Similarly, the overall strength of the United Nations Assistance Mission for Rwanda (UNAMIR) was roughly 2,500 military personnel which had increased slowly over a period of five months to reach its maximum strength. After the resolution S/RES/912 (April 21, 1994), the mission was downsized to 450 units. Whereas it may have been feasible to quickly strengthen UNAMIR’s mandate in response to the genocide, it would have been much more difficult to rapidly increase the size of the mission in order to establish an effective military presence.

As a credible signal, size is arguably more important than the mandate under which peacekeepers operate. Obviously, both factors are closely intertwined: a more ambitious and extensive mandate requires a larger and better-equipped peacekeeping force. Moreover, even though members of the Security Council and the Secretary General have to expend considerable political capital to agree to a PKO’s mandate, the mandates of nearly all African missions included in our analysis extended beyond observation and monitoring of a cease-fire or the implementation of a treaty agreement. Throughout the 1990s, at least for African civil wars, multidimensional peacekeeping became the standard and for the purpose of our study, mandates were largely a constant (Volker and Warnecke 2009).⁵ Once a mandate is agreed, the Security General still has to garner sufficient military and financial support for the mission.⁶ The large political costs involved in deploying and sustaining a sizable UN presence actually contribute to the credibility of the signal.

Admittedly, the size of the military is not the sole contributing factor in creating a credible force; the quality of the forces and the resolve and the leadership of the mission commanders also play an important role as the case of UNAMSIL clearly illustrates. Yet, neither the resolve nor the quality of leadership is fully known ex ante. Although size does not guarantee that the mission will play an active role in enforcing peace, it provides a direct observable signal to all combatants that there is an intent and at least potential to be a proactive guarantor of the peace agreement (Böhmelt et al. 2011). The ability to use force and the implied threat of force typically play a greater role in shaping the actors’ views on feasible outcomes than the actual use of force. As Friis (2010) observes, similar principles apply to counterinsurgency and peacekeeping operations: a mission with sufficient size to reach even remote areas, in particular areas that are rebel controlled, is required to adequately protect civilians, in order to gain their political support. Further, a larger mission can use force more restrictively and discriminately because it has more deterring power.⁷

Obviously, size is not identical to resolve, but we argue that in practice the concepts are closely related. If a mission operates in a relatively small geographic area (and/or has backup personnel to keep peace in relatively secure areas) like Sierra Leone, a small but resolved contingent can indeed make an essential contribution. In other cases, like the DRC, resolve is simply insufficient to control the full geographical area in conflict. In Rwanda, the withdrawal of the peacekeepers was generally seen as a serious failure: it is indeed difficult to imagine that the peacekeepers that were “on the ground” could have made a significant difference. Also in Somalia, the resolve of the UN peacekeepers (in this case US troops) was questioned less than their ability to make a real difference and the lack of political support from Washington to provide the necessary troops. In our opinion, size is therefore an acceptable and observable (ex ante) indication of resolve.

To summarize, we expect that in the presence of a relatively large UN mission, former combatants will be more willing to cooperate with each other and the United Nations, and less willing to use violence to achieve their strategic goals. Accordingly, Fortna (2008, 127) notes that the two main parties involved in the civil war in Mozambique, FRELIMO and RENAMO, stressed the importance of the presence of a sizable UN mission. This leads to the first testable hypothesis:

Hypothesis 1: The larger the UN deployment, the greater the willingness of local actors to cooperate.

The context of the conflict should determine the strength the UN needs to be a proactive guarantor. In other words, not all missions of equal size will have the same effect on the willingness of the local actors to cooperate. Merely explaining the level of cooperation based on the attributes of the UN mission disregards the characteristics and possible strategies of local actors. In particular, the “balance of power” between the government and the rebels will shape their suspicions and options (Werner 1999; Walter 2009). If the capabilities of the opposing sides are more asymmetric, the UN needs a stronger force to hold the peace. The stronger the rebel groups, the less willing they are to negotiate a settlement with the government. Strong rebel groups do not need a security guarantee from a third actor, since they can provide it themselves. In Somalia, Aideed, the leader of the strongest rebel group (the Hawiye Habr Gedr clan), was reluctant to have a large deployment of Blue Helmets, since a large UN mission would have limited the “bullet” power of his fighters (Howard 2008, 23; see also Metternich 2011). Accordingly, whenever the UN gets involved as a third-party enforcer in weak states, they are often considered hard cases for peacekeeping (Fortna 2004; Beardsley 2010).

The internal power distribution and dynamics of the conflict affect the ability of the UN missions to fulfill its mandate. Rebel groups that fear that their military capacity or ability to recruit new fighters will decline over time are more vulnerable and should be more willing to accept the role of a (large) UN mission as a guarantor. Assuming that rebel forces are particularly reluctant to trust the government to respect the agreement, the presence of the UN troops provides rebel groups with an opportunity to accept a settlement under the protective umbrella of the UN. Rebel groups that are relatively weak should especially value this opportunity, since strong states can attempt to defeat the rebels militarily. Thus, vulnerable rebel groups should be more willing to cooperate with UN missions as long as they can have a real impact, that is, if they are large enough.

UN PKOs are deployed disproportionally to relatively weak states with limited control over their whole territory (Fortna 2004). Unlike rebel groups, governments have more incentives to cooperate with UN PKOs. The mandates of multidimensional peacekeeping encompass strengthening state structures, in effect, to strengthen and support the government. Accordingly, the peace process may well lead to a change in the relative balance of power allowing the central government to reassert itself at the expense of—or possibly by integrating—the rebels.⁸ Increasing capabilities can undermine the government’s commitment to the original agreement. It follows that if a government is stronger at the outset of the peace process, the size of the UN mission becomes more relevant; a larger mission will be needed to signal resolve and commitment to the peace process throughout the dynamics of its implementation. The following testable hypothesis emphasizes the conditional effect of the size of the mission:

Hypothesis 2: The more vulnerable the rebels are in military terms relative to the government, the stronger the impact of the size of UN deployment on the willingness to cooperate.

Research Design

Data

Part of the data was originally collected as part of the “UN Peacekeeping and Local Governance” project, which codes events pertaining to the state- and peace-building policies implemented as part of UN peacekeeping operations based on the reports of the UN Secretary General (Dorussen and Gizelis 2011; Ruggeri, Gizelis, and Dorussen 2011). Since the early 1990s, the Security Council has asked the Secretary General to report regularly on progress, or lack thereof, of the various peacekeeping missions. The reports give a regular, fairly extensive, and systematic account of ongoing PKOs. A drawback is that the reports possibly present the UN in an overly favorable light or that they are otherwise politically motivated. They are, however, in the first instance intended for the Security Council, which is unlikely to accept an account that is clearly biased, incomplete, or obviously incorrect. The reports are also public documents, and local and international media would point out gross inaccuracies or misleading statements. Finally, it is not immediately obvious that the Secretary General would over- or underreport cooperation for political purposes; especially because the funding of the UN PKOs depends on the members of the Security Council. An overly optimistic account of the mission can lead to an early drawdown; while an overly pessimistic account undermines the organization’s credibility. The quality of the reports has also been criticized (Autesserre 2010, 84). The final reports of the Secretary General however combine multiple sources at different levels of authority; hence, there is no reason to suspect that the inputs of individual rapporteurs systematically bias the final report.⁹ Finally, we have randomly double-checked observed events using ReliefWeb¹⁰ as a final robustness test.

The data report on the level of cooperation with the UN peacekeeping activities that concern local governance, referred to as “governance events.” The latter are defined as time- and place-specific interactions between stakeholders in peacekeeping—that is, the UN, its peacekeepers, and other external actors as well as government, rebel, and local authorities—with a direct effect on the provision of public goods and services. The data distinguish between government, rebel, and (quasi-) independent local authorities. Here, we use data for all UN missions deployed in Africa after 1989 until 2005,¹¹ for further information on the data validity and reliability see Ruggeri, Gizelis, and Dorussen (2011).

In our analysis, we cannot include non-UN peacekeeping operations by regional actors such as the Economic Community of the West African States (ECOWAS) since we do not have systematic reports on their activities. With very few exceptions, regional peacekeeping missions are traditional operations with an emphasis on monitoring and are not comparable to the complex multidimensional UN operations, which are the main focus of this study (Heldt and Wallensteen 2004).

Dependent Variable

The dependent variable is the count of cooperative events during the period covered by the report. In the “UN Peacekeeping and Local Governance” data, cooperation is coded as an attribute of a governance event during a peacekeeping mission.¹² The coding of cooperation is based on Sharp’s (1971) scale of nonviolent actions and ranges from symbolic acts to acts of omission (permissive cooperation) and commission (positive engagement). Examples of permissive cooperation are mentions of “life returning to normal” or “refugees voluntarily returning home.” Positive engagement is to be found in the participation in elections, the signing of a treaty, or agreeing to power-sharing agreements; for example, in 2000 in Freetown (Sierra Leone), UNAMSIL in cooperation with local police patrolled the streets and received cooperation also from rebel factions.¹³

The unit of analysis is the report with the count of cooperative events per report as the dependent variable. Since not all coded events have a clear temporal identification, the minimum and generally most correct temporal span is the duration of a UN report. In fact, we are confident that the events coded happened during the reporting time, but in most of the cases we cannot provide a precise date. The average length of a report is around two and a half months. Using reporting period as temporal unit of analysis allows us to have the most detailed information with the highest possible level of confidence. The total number of reported events, with and without cooperation, per report varies from 1 to 101.¹⁴ In the following, we discuss how we dealt with variation in UN reporting periods as well as variation in the total number of events per report.

We employ a count model on cooperation events per report instead of an ordinary least squares (OLS) estimator, because using OLS produces inaccurate estimates (negative counts) for discrete count variables that are not independent of each other. Moreover, OLS is inefficient as an estimation method, since it fails to take into account the heteroskedastic nature of the event counts (King 1988). Finally, we cannot assume that the mean of the counts equals their variance; therefore, the count events can be overdispersed (Long and Freese 2006). This suggests that a negative binomial estimator is more appropriate than a Poisson regression. We model the temporal variation of the UN reports by employing the report duration as exposure time in the negative binomial equation (Cameron and Trivedi 1998).¹⁵

Explanatory Variables

The main explanatory variable to test the first hypothesis is the size of the UN peacekeeping mission, where size is defined in terms of total numbers of personnel deployed in the UN mission. The data on deployment are from the reports of the UN Secretary General and from the UN Peacekeeping department.¹⁶ UN missions include observers, police personnel, and military troops. The size of the UN missions varies considerably over time. We have rescaled the variable to one unit per thousand UN personnel to facilitate the reading of the tables.¹⁷

The relative strength of government and rebels, “government/rebel” leverage, is measured with an index that ranges from 0 to 1. Values closer to 1 indicate that the government is much stronger than the rebels. When values approach 0, the rebels are much stronger than the government:

L e v e r a g e = \frac{G o v M i l S i z e}{\begin{matrix} G o v M i l S i z e + O p p M i l S i z e \end{matrix}} .

The index uses data on military personnel from the Correlates of War project and we have double-checked these figures employing the statistics from the Military Balance journal.¹⁸ The “Nonstate Actor” (NSA) data set of Cunningham, Gleditsch, and Salehyan (2009)¹⁹ is the source for information on rebel size. Rebel size combines the strengths of all factions fighting against the central authority. In order to test the second hypothesis, we use an interaction term between the UN size and the “government/rebel” leverage.

Control Covariates

To avoid omitted variable bias, it is important to control for other factors that influence the dependent variable and our core independent variables of interest. For instance, both the literature on civil war duration (Cunningham 2006) and peacekeeping (Fortna 2004) argue that the number of factions or rebel groups and veto players can lead to more complex bargaining processes due to coordination problems and heterogeneous preferences among the actors. Data from Cunningham, Gleditsch, and Salehyan (2009) are used to count the number of rebel groups, which varies between one and five.

The duration of the mission can also influence the perceptions of locals regarding the peacekeepers and the UN mission. Positive expectations from the local actors can wane over time if an agreement is not reached or a “normal” interaction between the actors is not achieved after a certain period. We use the cumulative length of the mission in days to measure the duration of the mission. In order to control for any nonlinear effects we also include the quadratic term of the mission duration.²⁰

Moreover, we control for both the population and the geographical size of the country where the mission takes place. It is expected that countries with larger populations tend to have more often heterogeneous preferences compared to countries with smaller populations (Alesina and Spolaore 1997). Furthermore, the larger the geographical size of a country, the harder it becomes for the UN to monitor the interactions between local actors and to enforce successfully the mandated policies (Center for International Development 2000).

Empirical Findings

Table 1 presents the two main models. The significant α in both models indicates that the count events are overdispersed and a negative binomial regression is more appropriate than a Poisson regression.

Table 1.

Cooperation during Peacekeeping in African UN PKOs 1989–2005: Negative Binomial Regression.

	Cooperation
	Model 1	Model 2
UN size	0.015 (0.011)	−0.034** (0.017)
“Gov/Rebel” Leverage	−1.357*** (0.389)	−1.925*** (0.418)
UN Size × Leverage		0.123*** (0.037)
No. of rebel groups	−0.171** (0.075)	−0.142* (0.073)
Time	0.011*** (0.003)	0.013*** (0.003)
Time²	−0.004*** (0.001)	−0.005*** (0.001)
Population	0.014 (0.011)	0.004 (0.011)
Country size	0.000 (0.001)	0.001 (0.001)
Constant	−0.878*** (0.315)	−0.770** (0.310)
Observations	232	232
LL	−877.2	−871.9
α	.600***	.570***

Note. Standard errors in parentheses LL=log likelihood.

***p < .01. **p < .05. *p < 0.1.

Since the correct interpretation of conditional effects is essential to assess the empirical hypotheses, we use the first difference to represent graphically our main statistical findings of model 1 (Table 1). In Figure 1, the estimated conditional effect is represented by a circle with the 95 percent confidence intervals as dashed lines. If the main effect and the 95 percent confidence interval are on the right (left) side of the dotted vertical line positioned at zero, the effect of the variable is significantly above (below) zero, indicating a positive (negative) and significant effect on the count of cooperative events. If the confidence intervals overlap both in the right and left areas, the effect of the variable is not statistically significant at the 95 percent level.

Figure 1.

The effect of the size of UN missions on cooperation.

Model 1 provides some initial, but admittedly weak, evidence that large UN missions increase the number of cooperative events in line with the first hypothesis. Employing the method suggested by Long and Freese (2006) to compute quantities of interest, an increase of one unit of our UN size variable—which means an additional thousand soldiers—increases the number of cooperation events by 2 percent. However, as Figure 1 shows, the coefficient only approaches statistical significance. Model 1 is arguably incorrectly specified, because it fails to model that the effect of size is conditional; to properly assess the effect of size on cooperation it is insufficient to simply control for population and area of operation, but it is also necessary to consider the balance of power between government and rebel forces.

Model 1 (in Table 1) and Figure 1 clearly indicate that the power distribution between rebels and government has an effect on cooperation during peacekeeping. The estimate for “rebel/government” leverage in Figure 1 shows that there is significantly less cooperation if the balance of power is more in favor of the government. An increase of one standard deviation of our index on government/rebel leverage reduces the cooperation events by 25.5 percent. As argued earlier, there are good theoretical (and now also some additional statistical) reasons to explore possible interaction between “size” and “leverage.”

Figure 2.

The marginal effect of UN size on cooperation conditional on “government/rebel” leverage.

Model 2 includes the interaction between UN size and “government/rebel” leverage. The empirical analysis clearly supports the second hypothesis. To recall, according to Hypothesis 2, the weaker the rebels are relative to the government forces, the stronger should be the impact of the UN mission’s size on the level of cooperation. In other words, the UN size makes a difference, when the rebels are relatively disadvantaged versus the government. As expected, the interactive coefficient of UN size and “government/rebel” leverage is positive and statistically significant.

A graph, as in Figure 2, is most suitable to interpret the interaction effect in a nonlinear model correctly.²¹ The effect of UN size is positively and statistically significant related to cooperation events, when the government is stronger than the rebels; that is, the index of government and rebel military power asymmetry is above 0.5. The results support the second hypothesis. Moreover, the effect of the interaction is significant for an empirically relevant part of the domain: the thin dashed line represents the distribution of “government/rebel” leverage, the modifying variable, and shows that there is a substantial number of observations providing support for the second hypothesis. To return to an example used previously, in Mozambique RENAMO was weaker than the central government (rebel/government leverage at 0.7) and cooperated with the UN mission regularly. On the other hand, an opposite case is Somalia where the government/rebel leverage is 0, indicating that the rebels had all the military power. In fact, Somalia is an extreme case where the disappearance of the central army is indicative of the state’s failure. In order to control whether this extreme historical case was driving our results, we have also run our models excluding Somalia from the sample. The results remain largely the same and continue to support the second hypothesis.

The number of rebel groups that are active also matters. One additional rebel organization decreases the number of cooperative events with 16.7 percent. The escalation of violence in the Ituri and North Kivu areas in the DRC can illustrate this effect. When the Ugandan and Rwandan governments became less actively involved in the conflict, the power vacuum was filled by a plethora of rebel groups that proved to be a serious challenge for UN peacekeepers (van Reybrouck 2010, 495-97).

The control variables for the duration of the mission further indicate that time matters for the willingness of locals to cooperate with the UN mission. The duration of the mission appears to have a nonlinear effect on cooperation. Figure 3 graphs the impact of the duration of the UN mission on cooperation, including the 95 percent confidence intervals, and shows an inverted-U shaped relationship; ceteris paribus, after approximately four years and five months, the count of cooperative events declines.

Figure 3.

Impact of the duration of the UN mission on cooperation.

Robustness Checks

The results for the UN size and its interaction with “government/rebel” leverage are clearly robust. Long and Freese (2006) suggest to compare the predictive capacity of different count models to check for their robustness. Using their method, we find that a negative binomial regression performs better than the Poisson regression with zero-inflated estimators. As a further alternative, we have reestimated the main models with a logistic estimator reshaping our data as binary grouped employing frequency weighted and found similar, actually even stronger, results. Finally, we also controlled for country-fixed effects using country dummies, and the results did not change substantially.

The period covered by a particular report is used as the exposure in the main models presented above. Intuitively, the choice of exposure matters because the observed value of the dependent variable (cooperation) is directly related to the period in which an event can occur; in other words, the longer the temporal exposure, the higher the likelihood of observing cooperative events. Similarly, a large number of reported events can also increase the likelihood of observing cooperation. As an alternative offset for exposure, we have rerun the main models using the total number of events per report. As shown in Figure 4, even though the larger confidence intervals indicate more uncertainty, the results are substantially the same.²²

Figure 4.

The marginal effect of UN size on cooperation conditional on “government/rebel” leverage using total number of events as exposure.

Another potential concern is the temporal effect in our models. In the presented models, the temporal dynamics are controlled using the duration of a mission and its squared term. The main problem of using an autoregressive count model is that introducing a lag of the dependent variable can lead to an exponential feedback (Cameron and Trivedi 1998, 238). Nevertheless, as a robustness test, we decided to check whether using an autoregressive model—both with and without controls for temporal dynamics—alters the main results.

We have run three additional count models.²³ The first model is a negative binomial regression without any temporal control. It still replicates the main results. The second model introduces the lag of the dependent variable. As explained earlier, the results need to be interpreted with caution since exponential feedback can lead to biased estimates. Regardless, the main results do not change substantially and the effect of the lag variable is quite small, but statistically significant. The third model includes the lag of the dependent variable as well as the duration of the mission. Here, the effect of the lagged dependent variable is halved and no longer statistically significant, whereas our main explanatory variables do not change in terms of statistical significance.

Putting aside any technical differences, the three models also have different theoretical underpinnings. In fact, a lag model assumes that previous events have an effect, whereas a model controlling for the duration of the mission—or in more precise terms the temporal point of a mission—assumes that over time the probability of cooperation has a certain pattern. Using the terminology from event history models, we estimate the baseline hazard function of cooperation. Given the preceding results, we are confident that the specification with a temporal baseline is more appropriate than an autoregressive model.

As an additional robustness test, we control for a set of additional “conflict” variables, since arguably the conflict history affects the willingness of the actors to cooperate. However, the results remain robust after including these variables.²⁴ Gilligan and Sergenti (2008) find that the effect of PKO operations is different during civil wars compared with postconflict situations. To control for this, we introduce a dummy variable taking the value of 1 for every period the country was at civil war and 0 if not, using data from Gleditsch et al. (2002). The variable is not statistically significant, and the main results hold.

Cooperation could also depend on the level of violence experienced during the civil war. In the robustness checks, we control for the cumulative number of battle deaths at the temporal point under observation (stock) and also for the number of battle deaths every year (flux); the data are from Lacina and Gleditsch (2005). We further control for violence toward civilians, both by government and rebels (Eck and Hultman 2007). These controls are generally not significant and including them has only very limited impact on our main estimates. Further, we control for the duration of the conflict prior to the deployment of the UN peacekeeping mission, using data from Gates and Strand (2006). Civilian deaths and conflict history are statistically significant but do not affect the main results on UN size and “government/rebel” leverage.

Finally, we have controlled for country effects and resampling. First we relax the requirement that observations are independent allowing the standard errors for intragroup correlation. Hence, the observations are independent across UN missions but not necessarily within countries. Furthermore, we have resampled employing bootstrap and jackknife and the results substantially replicate the main findings. Some countries, such as Angola, DRC, and Sierra Leone, have a larger share of observations, and could thus determine the results, but the results still hold excluding one country at a time.

In summary, sensitivity analysis of our main models shows that the main results are robust. We have compared the aptness of the estimator, employed different strategies to model temporal dependence, added further control variables and tested possible sampling issues.

Conclusion

During the last twenty years, UN peacekeepers have operated predominantly in civil-war situations and have increasingly been permitted to enforce peace. Regardless, the ability of the UN to elicit cooperation remains a primary concern. Peacekeepers need cooperation to facilitate the operation and to help them to achieve their main objectives. Moreover, UN peacekeepers remain relatively lightly armed and a small force, especially considering the area they are supposed to control. How can the UN peacekeepers improve cooperation during a mission? The answer appears to be that they need to signal their resolve to the local actors. They need to be committed and have the capacity to deter actors from defecting. Indeed, size matters. Moreover, how many peacekeepers are needed depends on the balance of power between the government and the rebels. The stronger the government is relative to the rebels, the stronger the marginal effect of the size of the peacekeepers on cooperation. To put it simply, a large UN force elicits cooperation when rebels are militarily weak. When facing a stronger government, rebel groups welcome the intervention by a large mission. Even though relatively strong governments are less likely to accept UN intervention (as documented by Fortna [2004] and Beardsley [2010]), governments that feel less threatened by the military might of the rebels are more willing to cooperate with UN peacekeepers after they have been deployed.

There are indeed recognizable patterns of cooperation in response to UN PKOs. From the perspective of UN PKO missions as complex bargaining processes, the key interactions are between three main actors: the government, the rebel groups, and the UN mission. This is in line with Kydd’s suggestion that “one insight that emerges from surveying the literature is that conflict resolution is inherently trilateral and needs to be modeled with three strategic actors” (2010, 103). We argue that UN PKOs are a response to commitment and mistrust problems that are prevalent in civil wars and postconflict agreements. Accordingly, we assess under what conditions the UN missions can effectively address these problems. Disaggregate data on UN peacekeeping missions in Africa from 1989 until 2005 show that the size of the UN missions seems to increase the levels of cooperative responses by rebel groups and local governments. However, the effect becomes only apparent once it is made conditional on the distribution of power between rebels and government.

As Fortna (2004) and Doyle and Sambanis (2006) have demonstrated UN peacekeepers are more likely to be present in “hard cases” with relatively strong rebel groups. Moreover, Dorussen and Gizelis (2011) find that rebels are less likely to cooperate with UN peacekeeping. Our research adds the insight that a large UN force is more likely to elicit cooperation, especially in situations with a relatively weak rebel force. How to reconcile these somewhat contradictory findings? First of all, a crucial role of peacekeepers appears to be to manage mistrust among local actors and particularly the rebels. The UN can assist local actors to overcome commitment problems and enhance the levels of cooperation. However, for missions to “succeed” they need to establish a clear presence on the ground.

Second, it takes (at least) two parties to have a conflict, and research on peacekeeping may have largely overlooked the constraints imposed by (remnants of) central authority. As long as they feel sufficiently strong, governments may resist the deployment of peacekeepers which could explain the findings of Fortna and Doyle and Sambanis. Alternatively, central governments may attempt to block the deployment of a robust peacekeeping force that can effectively protect rebels. The latter could explain our earlier findings (Dorussen and Gizelis 2011) and why there is more cooperation in those cases where the UN are able to overcome such “resistance.”

Obviously, the resources of the UN are limited; fielding a large mission to address a particular conflict inevitably restricts the UN’s freedom to intervene effectively elsewhere. Arguably, if cooperation is essential for the effectiveness of its missions, the UN should allocate its scarce resources by sending missions to countries where rebels are weak and governments are strong. In contrast where governments are weak and rebels are strong, PKO deployment is most inefficient—in the sense that deploying more troops yields less cooperation. In these cases, the benefits of PKO size are likely to be minimal and the opportunities for the UN to be an effective external guarantor to help solve commitment problems are at their lowest.²⁵

Large PKO missions are not without problems of their own either. For example, larger missions face increasing problems to coordinate activities of troops that often come from countries with completely different military traditions (Böhmelt et al. 2010). Larger missions can also face problems of coordination between the military and civilian activities, as well as among the various civilian sections, such as human rights, rule of law, civil affairs, and gender advisory. Moreover, large missions can have a greater economic footprint in the host economy, which leads to economic distortions and even exacerbate the unequal distribution of income among the locals. We recognize that all of these factors can undermine the effectiveness of the UN PKOs and the overall cooperative responses by local governments, rebels, and communal actors. However, it is not clear that they have a direct impact on the power distribution between government and rebel forces, which has been the main focus of this article. An examination of the possible interaction between the size of the mission and the social and economic conditions on the ground should be a fruitful topic for future research and could help expand our understanding of the conditions under which UN PKOs can elicit cooperation from local actors.

Footnotes

Authors’ Note

We thank Kyle Beardsley, Paul Diehl, Marian de Vooght, Kristian S. Gleditsch, BirgerHeldt, Lisa Hultman, Michaela Mattes, Nils Metternich, Vera Troeger and Steffen Weiss and two anonymous reviewers for their suggestions.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Authors received financial support from the Folke Bernadotte Academy and the ESF/ESRC, RES-062-23-0259.

Notes

References

Alesina

Alberto

Spolaore

Enrico.

1997. “On the Number and Sizes of the Nations.” Quarterly Journal of Economics 112 (4): 1027–56.

Autesserre

Severine.

2010. The Trouble with the Congo. Cambridge: Cambridge University Press.

Beardsley

Kyle.

2010. “Pain, Pressure, and Political Cover: Explaining Mediation Incidence.” Journal of Peace Research 47 (2): 395–406.

Böhmelt

Tobias

Metternich

Nils

Pilster

Ulrich

Ruggeri

Andrea.

2010. The Effectiveness of Peacekeeping Missions in Civil Wars: Disaggregating Interventions, 1948-2000 Credibly Solving Commitment Problems. Typescript, University of Essex and ETH Zurich.

Brambor

Thomas

Clark

William R.

Golder

Matt.

2006. “Understanding Interaction Models: Improving Empirical Analysis.” Political Analysis 14:63–82.

Cameron

A. Colin

Trivedi

Pravin K.

1998. Regression Analysis of Count Data. Econometric Society Monograph No. 30. Cambridge: Cambridge University Press.

Center for International Development. 2000. Country Size Data. http://www.cid.harvard.edu/ciddata/ciddata.html.

Cunningham

David.

2006. “Veto Players and Civil War Duration.” American Journal of Political Science 50 (4): 875–92.

Cunningham

David

Gleditsch

Kristian S.

Salehyan

Idean.

2009. “It Takes Two. A Dyadic Analysis of Civil War Duration and Outcome.” Journal of Conflict Resolution 53 (4): 570–97.

10.

Diehl

Paul F.

Reifschneider

Jennifer

Hensel

Paul R.

1996. “United Nations Intervention and Recurring Conflict.” International Organization 50 (4): 683–700.

11.

Dorussen

Han

Gizelis

Theodora-Ismene.

2011. Into the Lion’s Den Local Responses to UN Peacekeeping. Typescript, University of Essex.

12.

Doyle

Michael

Sambanis

Nicholas.

2000. “International Peace-building: A Theoretical and Quantitative Analysis.” American Political Science Review 94 (4): 779–801.

13.

Doyle

Michael

Sambanis

Nicholas.

2006. Making War and Building Peace. United Nations Peace Operations. Princeton, NJ: Princeton University Press.

14.

Durch

William J.

Holt

Victoria K.

Earle

Caroline R.

Shanahan

Moira K.

2003. The Brahimi Report and the Future of UN Peace Operations. Washington, DC: Henry L. Stimson Center.

15.

Eck

Kristine

Hultman

Lisa.

2007. “Violence against Civilians in War.” Journal of Peace Research 44 (2): 233–46.

16.

Fearon

James D.

1998. “Bargaining, Enforcement, and International Cooperation.” International Organization 52 (2): 269–306.

17.

Fortna

Virginia P.

2004. “Does Peacekeeping Keep Peace? International Intervention and the Duration of Peace after Civil War.” International Studies Quarterly 45 (2): 269–92.

18.

Fortna

Virginia P.

2008. Does Peacekeeping Work? Shaping Belligerents’ Choices after Civil War. Princeton, NJ: Princeton University Press.

19.

Friis

Karsten

(2010). “Peacekeeping and Counter-insurgency – Two of a Kind?” International Peacekeeping 17 (1): 49–66.

20.

Gates

Scott

Strand

Håvard.

2006. Modeling the Duration of Civil Wars: Measurement and Estimation Issues. Typescript, CSCW/PRIO.

21.

Gilligan

Michael J.

Sergenti

Ernest J.

2008. “ Do UN Interventions Cause Peace? Using Matching to Improve Causal Inference.” Quarterly Journal of Political Science 3 (2): 89–122.

22.

Gleditsch

Nils Petter

Wallensteen

Peter

Eriksson

Mikael

Sollenberg

Margareta

Strand

Håvard.

2002. “Armed Conflict 1946–2001: A New Dataset.” Journal of Peace Research 39 (5): 615–37.

23.

Heldt

Birger

Wallensteen

Peter.

2004. Peacekeeping Operations. Global Patterns of Intervention and Success, 1948–2000. Sandöverken, Sweden: Folke Bernadotte Academy.

24.

Howard

Lise M.

2008. UN Peacekeeping in Civil Wars. Cambridge: Cambridge University Press.

25.

Hultman

Lisa.

2007. “Battle Losses and Rebel Violence: Raising the Costs for Fighting.” Terrorism and Political Violence 19 (2): 205–22.

26.

Hultman

Lisa.

2010. “Keeping Peace or Spurring Violence? Unintended Effects of Peace Operations on Violence against Civilians.” Civil Wars 12 (1-2): 29–46.

27.

International Institute for Strategic Studies. 1967-2000. The Military Balance. London: International Institute for Strategic Studies.

28.

Kalyvas

Stathis N.

Kocher

Matthew A.

2009. “The Dynamics of Violence in the Vietnam War: An Analysis of the Hamlet Evaluation System (HES).” Journal of Peace Research 46 (3): 335–55.

29.

King

Gary.

1988. “Statistical Models for Political Science Event Counts.” American Journal of Political Science 32 (3): 838–63.

30.

Kydd

Andrew.

2000. “Trust, Reassurance, and Cooperation.” International Organization 54 (2): 325–57.

31.

Kydd

Andrew.

2006. “When Can Mediators Build Trust?” American Political Science Review 100 (3): 449–462.

32.

Kydd

Andrew.

2010. “Rationalist Approaches to Conflict Prevention and Resolution.” Annual Review of Political Science 13:101–21.

33.

Lacina

Bethany

Gleditsch

Nils Petter.

2005. “Monitoring Trends in Global Combat: A New Dataset of Battle Deaths.” European Journal of Population 21 (2-3): 145–66.

34.

Lake

David A.

Rothchild

Donald.

1996. “Containing Fear.” International Security 21 (2): 41–75.

35.

Long

J. Scott

Freese

Jeremy.

2006. Regression Models for Categorical Dependent Variables using Stata. 2nd ed. College Station, TX: Stata Press.

36.

Metternich

Nils.

2011. “Expecting Elections: Interventions, Ethnic Support, and the Duration of Civil Wars”. Journal of Conflict Resolution 55 (6): 909–937.

37.

Murdie

Amanda

Davis

David R.

2010. “Problematic Potential. The Human Rights Consequences of Peacekeeping Interventions in Civil Wars.” Human Rights Quarterly 32 (1): 49–72.

38.

Olonisakin

‘Funmi.

2008. Peacekeeping in Sierra Leone: The Story of UNAMSIL. Boulder, CO: Lynne Rienner.

39.

Paris

Roland.

1997. At War’s End: Building Peace after Civil Conflict. Cambridge: Cambridge University Press.

40.

Pouligny

Béatrice.

(2006). Peace Operations Seen from Below: UN Missions and Local People. London: Kumerian Press.

41.

Powell

Robert.

2006. “War as a Commitment Problem.” International Organization 60 (1): 169–203.

42.

Reybrouck

David van.

2010. Congo: Een Geschiedenis. Amsterdam, the Netherlands: De Bezige Bij.

43.

Ruggeri

Andrea

Gizelis

Theodora-Ismene

Dorussen

Han.

2011. “Events Data as Bismarck’s Sausages? Intercoder Reliability, Coders’ Selection, and Data Quality.” International Interactions 37 (3): 340–61.

44.

Sharp

Gene.

1971. Politics of Nonviolent Action. Boston: Porter Sargent.

45.

United Nations. 1996. The Blue Helmets. A Review of United Nations Peace-keeping Forces. New York: United Nations.

46.

Volker

C. Franke

Warnecke

Andrea.

2009. “Building Peace: An Inventory of UN Peace Missions since the End of the Cold War.” International Peacekeeping 16 (3): 407–36.

47.

Walter

Barbara.

1997. “The Critical Barrier to Civil War Settlement.” International Organization 51 (3): 335–64.

48.

Walter

Barbara.

2009. “Bargaining Failures and Civil War.” Annual Review of Political Science 12:243–61.

49.

Werner

Suzanne.

1999. “The Precarious Nature of Peace: Resolving the Issues, Enforcing the Settlement, and Renegotiating the Terms.” American Journal of Political Science 43 (3): 912–34.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.09 MB

0.00 MB