Complementary mediation: Exploring mediator composition in civil wars

ABSTRACT

Mediators improve the chances disputants in civil wars sign a peace agreement by exerting pressure or influence, referred to as leverage. This paper explores how sources of mediator leverage complement one another and draws attention to an under-explored form of influence: credible staying power. I argue that softer forms of leverage (i.e. relationships with the disputants and credible staying power) complement material strength, providing the highest chance of reaching a peace agreement when used together. As multiple sources of leverage mean multiple mediators, this paper also explores the number of and coordination among mediators, acknowledging the tension between multiple mediators increasing available leverage while complicating negotiations by adding more voices to the negotiating table. I argue that more mediators, all else equal, will decrease the probability of success; this effect can be offset, however, by coordination among the mediators. Empirical analysis of 312 mediation efforts in civil wars from 1989 to 2006 find that softer forms of leverage do reinforce material power, producing the greatest probability of reaching a signed agreement when used together. Moreover, coordination substantially mitigates the negative effect of additional mediators.

Los mediadores aumentan las posibilidades de que los contendientes en las guerras civiles firmen un acuerdo de paz ejerciendo presión o poder, lo que se conoce como influencia. En este artículo, se analiza la forma en que las fuentes de influencia de los mediadores se complementan entre sí y se pone de manifiesto una forma de influencia poco analizada: el poder de permanencia verosímil. Sostengo que las formas más suaves de influencia (es decir, las relaciones con los contendientes y el poder de permanencia verosímil) complementan la fuerza material, de modo que hay mayores probabilidades de lograr un acuerdo de paz cuando se utilizan en conjunto. Debido a que las múltiples fuentes de influencia implican el uso de varios mediadores, en el presente artículo también se analiza la cantidad de mediadores y la coordinación entre ellos, y se corrobora que la tensión entre varios mediadores aumenta la influencia disponible y, al mismo tiempo, complica las negociaciones, ya que participan más personas en la mesa de negociaciones. Sostengo que una mayor cantidad de mediadores, en igualdad de condiciones, disminuirá la probabilidad de alcanzar el éxito; sin embargo, este efecto se puede compensar con la coordinación entre los mediadores. De acuerdo con análisis empíricos de 312 esfuerzos de mediación en las guerras civiles que se produjeron entre 1989 y 2006, las formas más suaves de influencia refuerzan el poder material, de modo que existen mayores probabilidades de firmar un acuerdo cuando se utilizan en conjunto. Además, la coordinación mitiga en gran medida el efecto negativo de los mediadores adicionales.

Les médiateurs améliorent les probabilités que les parties en conflit dans des guerres civiles signent un accord de paix en exerçant une pression ou une influence que nous qualifions d’effet de levier. Cet article explore la manière dont les sources d’effet de levier des médiateurs se complètent mutuellement et attire l’attention sur une forme d’influence sous-étudiée : la résistance crédible. Je soutiens que les formes plus douces d’effet de levier (c-à-d, les relations avec les parties en conflit et la résistance crédible) complètent la force matérielle et offrent les plus grandes probabilités de parvenir à un accord de paix lorsqu’elles sont utilisées ensemble. Étant donné que de multiples sources d’effet de levier sont synonymes de multiples médiateurs, cet article explore également le nombre de médiateurs et leur coordination en reconnaissant que la tension entre les multiples médiateurs augmente l’effet de levier disponible tout en compliquant les négociations en ajoutant d’autres voix à la table des négociations. Je soutiens qu’un plus grand nombre de médiateurs, toutes choses égales par ailleurs, réduira la probabilité de réussite; cet effet peut toutefois être compensé par une coordination entre les médiateurs. Des analyses empiriques de 312 efforts de médiation dans les guerres civiles de 1989 à 2006 montrent que les formes plus douces d’effet de levier renforcent la puissance matérielle et donnent lieu aux plus grandes probabilités de conclusion par un accord de paix signé lorsqu’elles sont utilisées ensemble. De plus, la coordination atténue substantiellement l’effet négatif de l’ajout de médiateurs.

KEYWORDS:

• Mediation
• civil war
• conflict resolution

Notes

¹ Material power/leverage always refers to the economic or military resources the mediator has to entice combatants to negotiate (e.g. the promise of economic aid or threat of sanctions).

² Relational leverage refers to a mediator’s ability to draw on past interactions with the disputant(s) to build trust or gain information.

³ Notably, the Soviet Union supported the government throughout the conflict. The Soviet-Angolan relationship evolved similarly to United States-UNITA. The Soviet Union was also ready for a negotiated settlement but continued to support the government during negotiations (MacFarlane 1992).

⁴ Beardsley (2008) lists “promises of monitoring and enforcement” as one mechanism through which mediators can encourage settlement. Thus, this is not the first study to mention this role of mediators. Beardsley, however, does not operationalize or test this role directly.

⁵ None of the mediators analyzed here have all three forms of leverage.

⁶ Reid (2017) refers to this influence as credibility leverage.

⁷ See Wallensteen and Svensson (2014) for synthesis of the findings regarding mediation strategy and outcomes.

⁸ The sources of leverage a mediator has influences which strategies are available (e.g. material power is necessary to use directive) but does not determine the strategies chosen. Materially strong mediators may opt for a less intrusive strategy.

⁹ It often takes six months after authorization for military personnel to reach the combat zone (United Nations Peacekeeping 2019a).

¹⁰ While the United States only had 34 personnel deployed to UN peacekeeping missions as of June 30, 2019 (United Nations Peacekeeping 2019b), the United States is the largest financial contributor to African Union and UN peacekeeping missions in Africa (Williams 2015). Both money and troops are necessary to establish a peacekeeping operation.

¹¹ The Community of Sant’Egidio is a Catholic lay organization that worked to promote peace in several civil conflicts including Burundi, Mozambique, and Guatemala.

¹² Other scholars have considered the differential effect of mediation by global versus regional organizations (see e.g. Gartner 2011). This paper does not have the space to test if regional or global organizations are more or less credible. Regional organizations are more directly affected by conflicts within member states but also have fewer resources than global organizations. It is not obvious, then, which type of organization should be more credible.

¹³ Figure 1 is not exhaustive of the potential sources of leverage available to mediators. For example, mediation history and a reputation for success at the negotiating table or for following through on commitments made at the negotiating table could be another form of leverage. See Appendix Table A3 and surrounding discussion for more on mediation history.

¹⁴ The conflict between UNITA and the government initiated before independence was granted. Portugal had tried, but failed, to negotiate a transitional government that would include the rival factions before granting independence (Papp 1993).

¹⁵ On a more practical note, identifying when and how these sources of leverage are actively engaged is difficult. Promises (or threats) can be made in private. Whether a mediator gained (or revealed) new information because of their relational leverage or for other reasons is challenging to differentiate.

¹⁶ Focusing on international conflicts mediated by coalitions of states, Böhmelt (2012) highlights a history of conflict or cooperation among the mediators as an indication of their ability to work together throughout the negotiations. His measures, however, do not translate directly to civil wars as civil wars see a variety of mediators; non-state actors (e.g. IGOs and NGOs) play a frequent and important role. The history of interaction between a state and IO is not as easily quantified as between two states. I, therefore, use explicit signals of cooperation (or the lack thereof) during the mediation effort to capture the dynamics among the mediators.

¹⁷ According to the codebook, mediation efforts were identified for conflicts in the ACD “spanning every year from the conflict start date to the final conflict termination date. In episodic conflicts with breaks in fighting between episodes these years were searched as well” (DeRouen Bercovitch, and Pospieszna 2011).

¹⁸ 132 (thirty-eight percent) of state mediator-attempts were major powers.

¹⁹ The UN comprised ninety-five (forty-seven percent) of the mediation-attempts made by an IGO.

²⁰ Temporary agreements include ceasefires and process agreements that move the peace process forward but are not intended to end the conflict. The outcome of the negotiations was taken predominantly from the CWM dataset. A few changes were made based upon information from other datasets or further research. All changes are included in the data shared on the author’s website.

²¹ Given the importance of the operationalization of the dependent variable and model selection, I include several robustness checks in the Appendix, including breaking the peace agreement category into partial and full agreements, evaluating the proportional odds assumption, and presenting the results of generalized ordered logistic models that relax this assumption when appropriate (Williams 2016).

²² GDP data comes from Gleditsch (2002). When mediation efforts spanned more than one year, GDP in the start year of the mediation effort was used. The results presented are robust to using the log of the strongest mediator’s GDP instead of GDP ratio. See the results of this robustness check in Appendix Table A1 and surrounding discussion.

²³ Given the state-based measure of material power used, IGOs and NGOs are both coded as having no material power. While IGOs do have budgets and can pass resolutions that could provide carrots/sticks, they require discussions and votes by member states to do so. This reduces the mediator’s ability to engage these sources of leverage and, sometimes, prohibits it altogether. NGOs often have limited budgets and no military power or trade ties as bargaining chips; thus, coding them as having no material power is reasonable.

²⁴ GDP ratio and CINC ratio are strongly correlated at 0.61.

²⁵ EPR defines a group as politically relevant if “at least one significant political actor claims to represent the interests of that group in the national political arena, or if members of an ethnic category are systematically and intentionally discriminated against in the domain of public politics.” (Wimmer, Cederman, and Min 2009)

²⁶ The Minorities at Risk Organizational Behavior dataset (Asal, Pate, and Wilkenfeld 2008) and the Minorities at Risk quantitative dataset (Minorities at Risk Project 2009) provided additional support for these coding decisions as needed.

²⁷ MPO identifies the following organizations as deploying peacekeeping missions during the time period of interest for this study: UN, AU, CEEAC, CIS, ECOWAS, NATO, EU, OAS, and OSCE.

²⁸ For example, reports and press releases available through the Crisis Management Initiative and Center for Humanitarian Dialogue websites were consulted for cases mediated by these organizations.

²⁹ Some might expect diminishing marginal returns to additional mediators. The relative complexity of 7 versus 8 mediators being less than the relative complexity of 2 versus 3. I include analyses using the log of number of mediators in Appendix Table A1.

³⁰ In anticipation that coordinated efforts might not occur randomly and to identify any conflict characteristics that influence the likelihood of coordination, I estimated a logistic regression on multiparty mediation efforts with coordination as the dependent variable. Using the number of mediators as well as a collection of classic conflict characteristics that influence mediation, only territorial incompatibility significantly influenced the probability of coordination. The number of mediators was not a significant influence. I control for territorial incompatibility in the analyses.

³¹ Specifically, the initial start date of the conflict, as recorded in ACD, was used. The number of days from the start of the conflict and the onset of mediation was calculated and then scaled so a one unit increase represents a decade.

³² Interestingly, only 46% of mediation efforts with credible staying power have peacekeepers on the ground at the same time. Eighty-three percent of efforts without credible staying power do not have active peacekeepers either.

³³ The Appendix includes robustness checks including alternative operationalizations of key variables and alternative model selections.

³⁴ When uncoordinated multiparty efforts are the baseline category the coefficient on coordinated multiparty efforts is 1.274 (Model 1), 1.262 (Model 2), 1.364 (Model 3), and 1.361 (Model 4). In all models this coefficient is significant at 0.05.

³⁵ While it might be a little surprising that some of the controls are not significant, conflict duration and intensity have mixed results in the literature (see e.g. Clayton and Gleditsch 2014). Also, polity and territorial conflict, territorial conflict and peacekeepers, and peacekeepers and some categories of soft leverage, are at least moderately correlated with each other. These correlations could lead to inflated standard errors, although variance inflation factors do not indicate any concerns regarding the control variables.

³⁶ For discussion of multiparty mediation onset in civil wars, see Menninga (2019).