Ending Civilian Victimization: The Combined Effect of Mediation and Peacekeeping on Violence against Civilians

(1)

ENDING CIVILIAN VICTIMIZATION

The Combined Effect of Mediation and Peacekeeping on Violence against Civilians

Mathilda Grönlund Master's Thesis

Spring 2021

Department of Peace and Conflict Research, Uppsala University

Supervisor: Desirée Nilsson

(2)

Abstract

Ending civilian victimization has become a primary purpose for third-party intervening actors as the brutal violence in contemporary conflicts increasingly affects the civilian population. To mitigate the violence, third-party actors use conflict management tools such as mediation and peacekeeping. Previous research has excessively examined these tools in isolation from one another, however, their combined effect has been neglected. In seeking to explore further pathways to combat violence against civilians, this study examines the combined effectiveness of peacekeeping and mediation. I argue that peacekeeping and mediation interact, enhancing the violence-reducing effect of the other, which creates a stronger reduction effect on violence against civilians. Using monthly data counting civilian casualties in all African intrastate conflicts between 1993-2007, I find that mediation and peacekeeping have an interactive effect on violence against civilians, which is both reducing and stronger in comparison to the independent effect of these tools. The theoretical implications extend to UN troops and UN police. However, they are not applicable for UN observers and non-UN troops. Additional implications of the findings indicate that mediation should be presented as a conflict management tool first after a large size of UN troops or UN police forces are deployed to most effectively reduce the levels of violence against civilians in conflict.

Key words: Peacekeeping, Mediation, Violence against civilians.

(3)

Acknowledgments

First and foremost, I would like to extend my gratitude to my excellent supervisor Desirée Nilsson, whose support and motivation I am most grateful for. A thanks to the course conveners, Kristine Eck and Lisa Hultman for well preformed lectures and great assistance during the Master Thesis course. Additional appreciation is extended towards all excellent lecturers at the Department of Peace and Conflict Research at Uppsala University, whom have given their time to teach and inspire for the last two years during the Master’s Programme in Peace and Conflict Studies. Finally, I would like to give a huge thanks to my friends and fellow students in the Programme. Thank you all for your support, motivation and inspiration.

(4)

List of Figures and Tables

Figure 1. The causal chain……….16

Figure 2. Box plot illustration of the relationship between mediation and one-sided

violence...29 Figure 3. Scatterplot illustration of the relationship between UN troops and one-sided

violence………..30 Figure 4. Scatterplot illustration of the relationship between UN troops and one-sided

violence (log)……….31 Figure 5. Interaction plot illustrating the interaction effect between UN troops and

mediation on one-sided violence………32 Figure 6. Margins plot illustrating the marginal effects of UN troops across the range of

mediation………35 Figure 7. Johnson Neyman’s plot illustrating the marginal effects of mediation across the

range of UN troops……….37 Figure 8. Interaction plot illustrating the interaction effect between UN police and

mediation………39 Figure 9. Interaction plot illustrating the interaction effect between non-UN peacekeeping

troops and mediation………..41 Figure 10.Margin plot illustrating the marginal effects of UN police across the range of

mediation………57 Figure 11. Johnson Neyman’s plot illustrating the marginal effects of mediation across the

range of UN police……….58 Figure 12. Margin plot illustrating the marginal effects of non- UN troops across the range of

mediation………61 Figure 13. Johnson Neyman’s plot illustrating the marginal effects of mediation across the

range of non-UN troops……….62 Table 1. Summary descriptive statistics………...26

(7)

Table 2. OLS regression estimates of the main effects and the combined effect of UN troops and mediation on one-sided violence………..33 Table 3. OLS regression estimates of the relationship between mediation and one-sided

violence………. 54 Table 4. OLS regression estimates of the relationship between UN troops and one-sided

violence………..55 Table 5. OLS regression estimates of the interaction effect between UN police and

mediation on one-sided violence………56 Table 6. OLS regression estimates of the interaction effect between UN observers and

mediation on one-sided violence………59 Table 7. OLS regression estimates of the interaction effect between non-UN troops and

mediation on one-sided violence………60

(8)

1. Introduction

Ending civilian victimization in conflicts has become increasingly important. Ongoing civil wars are characterized by large-scale mass violence against civilians, violence which is often used as a strategic tool by conflicting parties (Azam and Hoefler 2002; Kalyvas 2006; Valentino 2014) . While armed rebel groups have committed gross atrocities against civilians, such as;

mass killings, torture, sexual violence and targeted bombings, states have also carried out violence against civilians which has been equally brutal (R. M. Wood 2010). Gross human rights violations by both rebels and state actors have increasingly been reported by international organizations and media, calling for intervening action by third-parties to resolve these conflicts and end the violence against civilians (Sivakumaran 2010). Building on this call to action and the importance of ending civilian victimization in conflicts, this study investigates the effect of two commonly used conflict management tools wield by third-party intervening actors. The effects of mediation and peacekeeping on violence against civilians, so called one-sided violence, is examined. This topic is important for research and policy makers alike, due to the large-scale civilian suffering in many on-going conflicts.

The study builds on previous research in looking into mediation and peacekeeping and their combined effect on reducing violence, specifically violence against civilians, which according to Aduda and Bussman (2019) has yet to be investigated. While peacekeeping has been found to have a reducing effect on one-sided violence against civilians (Hultman, Kathman, and Shannon 2013), mediation has not been found to have a reducing effect, instead an increase in civilian victimization has been observed after mediation onset (Aduda and Bussmann 2019).

However, benefits in combining mediation and peacekeeping have been put forward by scholars studying violence between government and rebels (Beardsley, Cunningham, and White 2019;

DeRouen and Chowdhury 2018; Hultman 2019a), yet no quantitative study has so far explored how these tools may influence the likelihood of one-sided violence. Aduda and Bussman (2019) specifically outline that “future research needs to establish whether a combination of peacekeeping and mediation can contribute effectively to a reduction of the levels of one-sided violence and whether peacekeeping might mitigate the negative consequences of mediation”.

(9)

In seeking to explore further pathways to combat one-sided violence, the main contribution of this study is to examine the combined use and effectiveness of peacekeeping and mediation.

Thus, the research question this thesis set out to investigate is the following: If and how does mediation and peacekeeping combined influence the level of violence against civilians in conflict?

I am arguing that third-party international actors can aid in decreasing the level of violence against civilians in ongoing conflicts. In particular, I argue that peacekeeping and mediation combined have a reinforcing interactive effect reducing one-sided violence, since these tools combined can effectively mitigate incentives for both rebels and governments to use violence against civilians as a strategy in conflict. I expect that the reducing effect of peacekeeping troops on one-sided violence is stronger when mediation is present rather than absent. I also expect that there is a reducing effect of mediation on one-sided violence conditional on the number of peacekeeping troops. To test the hypotheses, I conduct a quantitative statistical analysis, examining correlates of the monthly counts of one-sided violence in all African intrastate conflicts between 1993 to 2007. I utilize an Ordinary Least-Squares (OLS) regression model and graphically illustrate the findings through interaction plots and marginal effects plots. The results are consistent with the expectations. The analyzed findings indicating that mediation and peacekeeping combined has an interactive reinforcing effect on one-sided violence. Hence, peacekeeping’s reducing effect on one-sided violence is stronger when mediation is present rather than absent, and there is a reducing effect of mediation on one-sided violence conditional on the number of peacekeeping troops. These theoretical implications can be extended to include UN police as well as UN troops. However, they are not applicable to UN observers and non-UN troops. Additional implications of the findings indicate that mediation should be presented as a conflict management tool first after a large size of UN troops or UN police forces are deployed to most effectively reduce the level of one-sided violence in conflict.

The outline of the paper adheres to the following order. Firstly, previous research is reviewed and the research gap is identified. Secondly, the theoretical argument is presented. Thirdly, the research design is presented, including data choices and modelling strategies. Forth, the findings are analyzed through a main analysis and an extended analysis. Finally, a conclusion summarizing the purpose and the essential implications of the findings is carried out, including suggestions for future research.

(10)

2. Previous Research

Ongoing civil wars are characterized by large-scale mass violence against civilians (Valentino 2014; Azam and Hoefler 2002). Deliberate mass killings of civilians as seen in the conflicts of former Yugoslavia and Rwanda are two of the most recognized genocides in recent decades;

however, far from the only atrocities committed against civilians in conflict (Valentino 2013).

While rebel groups¹, such as the Lord’s Resistance Army in Uganda, Boko Haram in Nigeria, Daesh in the Middle East and Africa and the Revolutionary United Front in Sierra Leone have conducted gross atrocities against civilians, state-based violence against civilians have been equally brutal (R. M. Wood 2010; EASO 2019; Stepanova 2016). In Sudan, civilians were attacked by the state forces due to ethnic association with Sudan People’s Liberation Army- North, in Somalia and Burundi civilians were targeted as a counterguerrilla warfare strategy, and in Myanmar civilians have been systematically attacked since the military coup in February 2021 (Fjelde and Hultman 2014; UN News 2021; Valentino 2013).

In reducing violence against civilians, international conflict management has become increasingly important. As such, there is a growing literature examining the effects of third- party² conflict management tools on the reduction of violence (Walter 2002; Gartner and Bercovitch 2006; Doyle and Sambanis 2006; Fortna 2008; Hultman, Kathman, and Shannon 2013). As mentioned in the introduction, this thesis focuses on two such tools, namely;

mediation and peacekeeping and their combined effect on one-sided violence. In accordance with the Uppsala Conflict Data Programme (UCDP), one-side violence is defined as the “use of armed force by the government of a state or by formally organized groups against civilians which result in at least 25 deaths per year” (Eck and Hultman 2007). To relate the concept of one-sided violence to genocide, one-sided violence is a broader term with a wider inclusion criterion. It includes acts of genocide but also acts considered to be ‘terrorist’ in nature and a

1 A rebel group is in this paper defined broadly in accordance to the definition given by UCDP. Hence, a rebel group is a formally organized non-governmental group of people having announced a name for their group and using armed force against a government, another similarly formalized group or against civilians.

2 A third-party is an external actor actively conducting practices which influence conflicts and reduces their negative impact (Irrera et al. 2021). The paper focus mainly on the UN as a third-party conflict management actor.

(11)

range of other violent acts against civilians, such as individual and mass executions and bombings in public places (Eck and Hultman 2007). Since civilian victimization is continuously and commonly occurring in contemporary conflicts, especially in intrastate conflicts (Dupuy and Aas Rustad 2018; Buhaug et al. 2007; Kalyvas 2006), only this type of conflict is considered in this thesis, broadly defined as “a conflict between a government and a non- governmental party…” (UCDP Definitions 2021).

2.1 Understanding Peacekeeping in Relation to One-Sided Violence

The term peacekeeping is broadly referred to as the deployment of international personnel to help and maintain peace and security (Fortna and Howard 2008). Peacekeeping operations can differ in size, mandate and quality. Peacekeeping missions can be deployed by regional actors, state coalitions or international organizations such as the European Union (EU), National Atlantic Treaty Organization (NATO) and the African Union (AU), although they are most know to be led by the United Nations (UN) (Bara and Hultman 2020). Peacekeeping missions are usually consent-based, with the exclusion of peace enforcement missions such as seen in Kosovo (1999) or Libya (2011) (Weiss 2015). There are also distinctions between peacekeeping in the forms of troops, police and observers. Peacekeeping troops are usually the most capable and are used to enforce ceasefires and peace agreements, mitigate the violence between belligerents, end the fighting and enforcing resolutions. Peacekeeping police forces are associated with monitoring and protecting civilians behind the battlefield lines. Police forces are also tasked with reinforcing the rule of law, engage in search and investigations and protect refugee camps. Peacekeeping observers’ function is to monitor and report about conflict-related processes. They do not directly use arms nor violence (Kathman 2013). Another differentiation is between types of UN peacekeeping missions. One commonly made distinction is between the UN traditional operations and the UN multidimensional missions. A large number of peacekeeping operations are still considered traditional peacekeeping, although it was more common during the 1980’s. This type of peacekeeping is confined to monitoring and patrolling, focusing on preventing the reoccurrence of violence, however, excluding action on conflict resolution. This type of peacekeeping is distinct from multidimensional peacekeeping, which inclusion criteria is wider (Diehl and Druckman 2015). According to the UN, multidimensional peacekeeping is:

(12)

“… multidimensional peacekeeping operations are called upon not only to maintain peace and security, but also to facilitate the political process, protect civilians, assist in the disarmament, demobilization and reintegration of former combatants;

support the organization of elections, protect and assist in restoring the rule of law” (United Nations Peacekeeping 2021).

While the traditional peacekeeping missions may fill an important role in some cases, multidimensional peacekeeping is more common today and have a stronger focus on protecting civilians. Thus, this thesis will focus on the concept of peacekeeping broadly, following the definition of the UN multidimensional peacekeeping.

In reviewing previous research examining the effect of peacekeeping on one-sided violence, scholars largely agree that UN peacekeeping has a reducing effect on violence against civilians (Fortna 2008; Bara and Hultman 2020; Hultman 2010). However, findings differ in the explanations, pointing towards the type of mission (Doyle and Sambanis 2006; Hultman 2010), mission size (Hultman, Kathman, and Shannon 2013), mission diversity (Bove and Ruggeri 2016), or mission quality (Haass and Ansorg 2018). Furthermore, existing research indicate that only larger UN peacekeeping operations, with an explicit mandate to protect civilians, seems to reduce violence against civilians by both governments and rebels (Bara and Hultman 2020; Hultman 2010). While looking into the effects of different types of UN mission, solely larger UN troops- and police missions are associated with a decrease in one-side violence, while UN military observers have the opposite effect (Hultman, Kathman, and Shannon 2013).

2.2 Understanding Mediation in Relation to One-Sided Violence

Mediation is the second tool to be considered. It is the most common form of conflict management diplomacy tool and has been thoroughly researched by many scholars (Bercovitch, Anagnoson, and Wille 1991; Regan and Stam 2000; Zartman and Touval 1985; Regan, Frank, and Aydin 2009). Mediation can have many functions such as; aiding in mitigating violence, facilitating negotiated settlements, providing information, highlight common interests and encourage constructive communication to get belligerents to find mutual understandings to sign a peace agreement (Walter 2002; Ohlson 2008). Despite the central role of mediation in contemporary conflicts, there is no exact consensus among scholars of how to define mediation.

(13)

Common elements are that mediation is voluntary and peaceful with the central focus of resolving conflict through dialogue (Duursma 2012; Zartman and Touval 1985; Bercovitch and Gartner 2008). In consideration of these central elements, this paper adopts Svensson’s broad definition of mediation, defined as: “voluntary third-party activity aiming at resolving or regulating the conflict through negotiations” (Svensson 2009: 455).

Moreover, previous research has demonstrated that mediation have an effect on reducing battled-related deaths (Ruhe 2020). However, the effect mediation has on one-sided violence have been found to be positive, meaning that mediation may have an unintended increased effect on violence against civilians (Aduda and Bussmann 2019; Pospieszna and DeRouen 2017). Kreutz and Brosché (2013) find that this relationship may be due to the fact that mediation is more active where a high number of one-sided violence has occurred, which questions the temporal order. However, while controlling for the temporal order, Aduda and Bussman (2019) finds that the increase of violence against civilians after mediation onset can be explained as a tactic to demonstrate strength and underline the importance of reaching an agreement favorable to one’s side (Aduda and Bussmann 2019). To this end, Aduda and Bussman (2019) outline that “future research needs to establish whether a combination of peacekeeping and mediation can contribute effectively to a reduction of the levels of one-sided violence and whether peacekeeping might mitigate the negative consequences of mediation”.

2.3 Identifying the Research Gap

Most research looking into the effect of peacekeeping and mediation have investigated these tools in isolation from one another. However, scholars whom have investigating mediation and peacekeeping combined suggest that these tools have beneficial combined effects (Beardsley, Cunningham, and White 2019; DeRouen and Chowdhury 2018; Walter 2002; Hultman 2019).

DeRouen and Chowdhury (2018) argue that mediation and peacekeeping combined has a positive impact on the duration of peace agreements and a reducing effect on the probability of renewed and continued violence. Beardsley, Cunningham and White (2019) investigates the effect of peacekeeping and mediation on battle-related violence, and finds that mediation and peacekeeping have an interactive effect, where mediation enhances the battle-related violence reducing effect of peacekeeping and vice versa. While this paper takes inspiration from the research done by Beardsley, Cunningham and White (2019), in examining the interactive effect

(14)

of mediation and peacekeeping, this paper differs in focusing on a different outcome. In addition, Beardsley Cunningham and White (2019) finds that mediation on its own reduces battel fatalities. This paper has a different entry point due to the fact that mediation has been found to increase one-sided-violence, rather than decrease. Furthermore, it is important to recognize that mediation and peacekeeping sometimes may not work well with each other, one example are the conflict resolution efforts in Cyprus (1974-1994), where the combined mediation and peacekeeping efforts did not work in each other’s favor (Evriviades and Bourantonis 1994). However, in relating to previous research mentioned, I overall still expect the general trend to be reinforcing rather than contradicting between peacekeeping and mediation.

To conclude, this study builds on previous research examining mediation and peacekeeping and their combined effect on reducing violence, specifically one-sided violence, which according to Aduda and Bussman (2019) has yet to be investigated. While peacekeeping have been found to have a reducing effect on one-sided violence against civilians (Hultman, Kathman, and Shannon 2013), mediation has not been found to have a reducing effect, instead an increase in civilian victimization have been observed after mediation onset (Aduda and Bussmann 2019;

Pospieszna and DeRouen 2017). However, benefits in combining mediation and peacekeeping have been put forward by scholars (Beardsley, Cunningham, and White 2019; DeRouen and Chowdhury 2018; Hultman 2019), suggesting that an interaction of these tools should be further investigated. Aduda and Bussman (2019), specifically outline that “future research needs to establish whether a combination of peacekeeping and mediation can contribute effectively to a reduction of the levels of one-sided violence and whether peacekeeping might mitigate the negative consequences of mediation”. This is the research gap this thesis sets out to investigate.

(15)

3. Theory

This section outlines the theoretical argument of the thesis. I begin by arguing that one-sided violence is a deliberate strategic tool used by conflict actors (Valentino 2014). In line with Beardsley, Cunningham and White, (2019) I posit that one-sided violence is more severe when there is less external constraints on the use of violence, hence one-sided violence will persist when perpetrators can act with impunity from punishment. While I am arguing for one-sided violence being a strategic tool, I am not testing for it. Hence, I assume this to be true based on the arguments presented. Furthermore, and to be tested, is the combined use and effectiveness of mediation and peacekeeping. Even though mediation has not been found to have a reducing effect on violence against civilians (Aduda and Bussmann 2019; Pospieszna and DeRouen 2017), and while peacekeeping could arguably have a hampering impact on one-sided violence on its own (Hultman, Kathman, and Shannon 2013), I argue that the combined effect of these tools result in a greater reduction of one-sided violence, and thus preferably implemented simultaneously. While the main focus theoretically is on the interaction effect, the sole effects of each tool are implicitly included, building on findings from previous research. The main effects are also tested and presented in the analysis, to accurately and transparently be able to analyze the interaction. Moreover, I posit mechanisms where mediation and peacekeeping combined will have a reducing effect on one-sided violence. However, while the causal mechanisms are important in providing a logical explanation, it’s important to note that it’s not going to be tested in this study. The focus of this thesis is on establishing and examining the causality. Hence, look at the direct effect of mediation and peacekeeping on one-sided violence.

3.1 One-Sided Violence Against Civilians

Scholars have pointed to several ways in which war generates incentives for deliberate violence against civilians (Valentino 2014; Chenoweth and Lawrence 2010; Pape 2005; R. M. Wood 2010). While the theoretical argument is based on the understanding that one-sided violence is commonly committed as a strategic tool in conflict, it is initially important to note that a strategic approach on one-sided violence does not mean that it is an effective approach. In fact,

(16)

It should also be acknowledge that violence against civilians has been committed due to non- strategic reasons (Straus 2013); however, consensus among scholars studying one-sided violence, is that it is primary instrumental (Valentino 2014). One-sided violence is a central part of the deliberate strategies of belligerent groups and usually coordinated by powerful actors with specific incentives. There are many different incentives behind the use of one-sided violence, most of them political or military in nature (Valentino 2014; Kalyvas 2006). Political and military incentives meaning in this context motivators to commit one-sided violence due to a political and/or military objectives (Valentino 2014).

When conventional warfare tactics have been unacceptably costly or ineffective, one-sided violence have been perceived as an alternative tactic used by governments and rebel groups to reach their aims (Valentino 2013). Political incentives can be a driver to commit one-sided violence. Hence, rebels may commit one-sided violence in trying to compel a target government to change policy or to get their political will acknowledged. Rebels can inflict pain on the society in targeting civilians with the aim to make the government accept political demands or change political directions in accordance to the liking of the rebel group (Pape 2005). When rebels target civilians, they demonstrate the state’s inability to protect their civilians, making it costly to continue the war and not accept rebel demands. Hence, one-sided violence is in this way an instrument for rebels to inflict costs on governments in the effort to extract concessions (Valentino 2014; Hultman 2012). The brutal atrocities committed against civilians in Colombia’s civil war is one example. According to Amnesty, thousands of civilians were killed by the guerilla organization, Revolutionary Armed Forces of Colombia FARC, and other rebel organizations in the country, due to political reasons. Exact estimates of women killed between the years 2002-2007, are around 100-400 each year (Amnesty 2008). Another example is from Sri Lanka, where the Tamil Tigers (LTTE) attacked civilians belonging to the ethnic lines of the government³, with the aim to affect the state in the effort to extract concessions (Fagerlund 2011). Moreover, political incentives are a forceful motivator among elites, presumably among state elites but possibly also among elites supporting the rebels cause. Elites can promote violence against civilians to obtain private political aims, to demobilize political adversaries, to gain material benefits or to achieve ideological goals. This has especially been seen during the conflict in former Yugoslavia, where one-sided violence was used for the private political gains of elites (Valentino 2014).

3 The ethnic belongings were highly politicized in Sri Lanka during the conflict

(17)

One-sided violence can be used to weaken one’s opponent and/or strengthen one owns position.

When civilians are targeted due to purely combat strategic reasons, military incentives are behind the act. If rebels suspect civilians to support the government they may attack civilians directly to weaken the government instead of the state forces to avoid any significant losses upon themselves. The same logic goes for the state if rebel forces are strong (Pape 2005;

Valentino 2013). This has been witnessed in Colombia where civilians have been tortured and killed by both state security forces and FARC as a deliberate strategy by the parties to “cleanse”

areas where civilians live, in which they believe are controlled by their enemies (Amnesty 2008). Furthermore, weak rebels may also view one-sided violence as an inexpensive alternative to gain support and cooperation from local civilians or to make them withhold support from the enemy (Valentino 2014; R. M. Wood 2010). Military incentives may also derive from a strong association between insurgency/counterinsurgency and one-sided violence. Civilians may be deliberately targeted by the state during counterinsurgencies because of their existing and unique relationship with rebel groups, as has been witnessed in Burundi, where the state forces targeted civilians due to their ethnic association with the state’s protagonists (Nantulya 2019). In addition, rebel groups usually rely on the local population for shelter, resources, recruits and intelligence. This gives governments incentives to commit one- sided violence to get rid of the rebels and their perceived support network among the civilian population (Valentino 2013).

While there are many incentives for rebels and governments to use one-sided violence, either for a political or military purpose, it is not impossible to mitigate or change these incentives.

Hence, one-sided violence is instrumental, and as such its desirability can be manipulated (Kathman and Wood 2014; Fjelde, Hultman, and Nilsson 2019). Thus, in mitigating the incentives to use one-sided violence, the violence in itself should decrease. I argue below that mediation and peacekeeping combined are effective in reducing such incentives.

3.2 Mediation

Mediation can aid in mitigate violence in facilitating negotiated settlements, provide information, highlight common interests and encourage constructive communication to get the conflicting parties to find a middle ground (Walter 2002; Ohlson 2008). Mediation is a useful

(18)

tool when negotiating the post-conflict distribution of political power. The distribution of political power is done in order to make both of the conflict parties see a win in peace rather than continued conflict, and as such are more likely to hold on to the agreed peace settlement (Walter 2002). Mediation can thus, through dialogue and communication, reduce the political incentives among rebel groups and state actors to use one-sided violence, in providing other ways to reach their desired aims. The international efforts of third-party actors in Kenya illustrates this well. After the electoral violence which erupted as a consequence of the 2007 election, killing thousands of civilians, a successful mediation effort was carried out mandated by the African Union. With the lead of the former UN Secretary-general Kofi Annan, a forty- one-day peace process was created and the mediation efforts ended with a peace agreement, ending the violence against civilians (Lindenmayer and Kaye 2009). Furthermore, it has also been argued that continued mediation during the political distribution of power have added value. Having third-party mediators oversee the process and assist with organizing informal meetings among group leaders, formalized mediation committees, and/or independent commissions aids with the political distribution. Also, continued mediation have been argued to keep the parties of the peace settlement engaged, being able to bring new parties to the peace process, such as civil society, and expand on the political participation (McCulloch and McEvoy 2019). Furthermore, mediators are usually connected to the political and/or military elite level actors when mediating formal negotiations (Papagianni 2010). In getting access to key elite actors, mediation is possibly the most preferable tool in making elites turn away from committing one-sided violence in being able to present alternative ways to reach their political aims. Mediation could also increase the cost upon elite actors, deterring them from using one- sided violence. Hence, powerful mediators can use both carrots and sticks where needed to deter actors in acting in certain ways (Beardsley 2008), thus reducing the political incentives to use one-sided violence among elites.

However, while mediation arguably could have a mitigating effect on political incentives as a motivation to use one-sided violence, Aduda and Bussmann (2019) demonstrate that mediation have an unintended increasing effect on one-sided violence, due to the fact that belligerents may have an interest in attacking civilians to demonstrate strength and improve their bargaining position. In addition, spoilers may also use one-sided violence to undermine the mediation process (Aduda and Bussmann 2019). This indicates that mediation efforts should not primarily be used to reduce the violence, rather other forms of interventions may be used to change the behaviors of the perpetrators, such as no-fly zones, arms embargo and peacekeeping missions

(19)

(Hultman 2019). Hence, mediation not backed by any other means gives room for belligerents to use military violence without physical restrain. Although, while mediation may not be sufficient by itself to address one-sided violence, it may still serve as an important compliment to other forms of coercive pressure. Further below I will argue that mediation may still work as an effective complement to peacekeeping in reducing one-sided violence, but first I will discuss solely peacekeeping’s effect on reducing civilian victimization.

3.3 Peacekeeping

Peacekeeping troops can work as a third-party security guarantor in separating the belligerents and reducing conflict activities which trigger civilian targeting (Hultman, Kathman, and Shannon 2013). Peacekeepers can for example be stationed in areas of great vulnerability or create buffer zones between the belligerents (Walter 2002). They can also patrol, monitor, conduct policing activities and enforce civilian protection (Hultman, Kathman, and Shannon 2013). In placing peacekeeping missions where one-sided violence is continuously committed, peacekeeping troops can give extra protection for the victimized civilians and make it costly for conflict actors to use one-sided violence in protecting the civilian population with military means if necessary (Fjelde, Hultman, and Nilsson 2019). In addition, peacekeeping can also effectively make it costly for both rebels and the government to commit one-sided violence, since one-sided violence become less “cheap” as a military tactic when peacekeeping is in place, thus reducing the military incentive to commit one-sided violence. To this end, a protection mandate authorizing peacekeepers to take necessary action to protect civilians is of importance (Fjelde, Hultman, and Nilsson 2019). The UN peacekeeping mission in Sierra Leone (UNAMSIL) illustrates a success story of which the protection of civilians was excessive through the right mandate, organization and resources. The peacekeeping mission did not just aid in mitigating the violence, but also assisted with Disarmament, Demobilization and Reintegration (DDR), the protection of refugees and monitored of the post-conflict election among other violence reducing tasks (World Peace Foundation 2021). In providing physical protection, peacekeeping can arguably reduce violence against civilians; however, according to existing research only if the peacekeeping missions are large in size and consists of military troops or police forces (Bara and Hultman 2020; Hultman, Kathman, and Shannon 2013).

(20)

While peacekeeping by itself may be an effective conflict management tool in mitigating military incentives and protect civilians on the ground, it is not a tool aimed at solving the main conflict issues between belligerents. In fact, in implementing coercive measures, such as peacekeeping missions, the willingness of the conflict parties to find a long-term solution to the conflict may get reduced (Greig and Diehl 2005). To increase the willingness among belligerents to find a solution, mediation can be used as a compliment to maintain a political peace process, maintaining communication and negotiations and thus generate a better chance to reduce one-sided violence in creating a sphere in which there is a possibility to negotiate forward peaceful resolutions (Regan and Aydin 2006).

3.4 The Theoretical Argument: Combining Mediation and Peacekeeping

While mediation is an effective tool to use in mitigating political incentives it needs peacekeeping to mitigate the military incentives to use one-sided violence. If the power imbalance is to great between conflicting parties or if spoilers deem to destroy mediated negotiations, solely mediation may not be enough to make the conflicting parties stop the violence (Aduda and Bussmann 2019). Thus, peacekeeping works as an extra third-party security guarantor, providing security and safety for civilians in areas of great vulnerability (Walter 2002). On the other hand, if the conflict parties have not declared they are ready to negotiate and cooperate with each other, peacekeeping is likely to be less effective. Mediation here is thus key to get the official and mutual declaration of peace, which can make peacekeeping more effective (DeRouen and Chowdhury 2018). Beardsley, Cunningham and White (2019), highlight that the interactive effect of mediation and peacekeeping is rooted in a conditional effect, that conflict actors have to make progress in negotiations but also that they have to trust that the peace agreement can be implemented. Hence, they argue that peacekeeping and mediation reinforces each other as peacekeeping can aid belligerents in overcoming the commitment problem and mediation can aid in facilitate peace negotiations, thus allowing for progress in both negotiation and implementation. Even though Beardsley, Cunningham and White (2019) investigates the interactive effect of peacekeeping and mediation on battled- related violence, Hultman (2019) suggests that the same logic should be expected to work for violence against civilians, since the strategies for combatting battle-related deaths and one- sided violence are usually the same. While the theoretical logic is similar between Beardsley, Cunningham and White’s (2019) theoretical proposition and mine, the arguments differ in

(21)

focusing on different mechanisms and outcomes. While Beardsley, Cunnigham and White (2019) focus on the benefits in combining mediation and peacekeeping for the benefits of negotiations and implementation, I posit that mediation’s and peacekeeping’s combined strength is rooted in mitigating incentives among conflict parties to use violence, specifically one-sided violence.

Figure 1 illustrates the causal chain of the theoretical argument to be presented. Hence, based on the theoretical logic presented and previous research findings, I expect that solely mediation will have no effect or increase the level of one-sided violence against civilians. Meanwhile, solely peacekeeping troops will reduce the level of one-sided violence against civilians;

however, not as much as mediation and peacekeeping combined. Hence, I am arguing that peacekeeping and mediation have a reinforcing interaction effect increasing the reduction of one-sided violence, because these conflict management tools can combined effectively mitigate the incentives for rebels and governments to use one-sided violence as a strategic tactic in conflict. While mediation is useful for mitigating political incentives to use one-sided violence in effectively distribute political power and maintaining communication between belligerent.

Peacekeeping is arguably useful for mitigating military incentives in acting as a third-party security guarantor. Both of them combined have an interactive effect which mitigates the political and military incentives for rebel groups and government to commit violence against civilians and thus reduces the total number of one-sided violence in conflict independent of actor committing the atrocity.

Figure 1. Causal Chain.

Thus, I have argued that mediation can mitigate political incentives through communication and negotiations, and peacekeeping can mitigate military incentives through force and security

Reduction in one-sided violence

Mitigates incentives for rebels and governments to

commit violence against civilians.

Meditation &

Peacekeeping

(22)

reinforcements. Since these two tools differ in nature and have an ability to complement each other to tackle different aspects of the same problem, I expect that mediation and peacekeeping indeed interact and reinforces each other on both sides of the interaction. Thus, the hypotheses this thesis sets out to test aims at examining if there is an interaction effect, and also both sides of that same interaction effect⁴.

H1: The negative effect of peacekeeping troops on one-sided violence is stronger when mediation is present rather than absent.

H2: There is a negative effect of mediation on one-sided violence conditional on the number of peacekeeping troops.

If H1 were to be true, I expect to see that peacekeeping troops combined with mediation have a stronger reduction effect on one-sided violence, compared to peacekeeping troops without mediation. To reject the null-hypothesis for H1, the evidence has to support that there is an interaction between peacekeeping and mediation on one-sided violence, and that the relationship between peacekeeping and mediation has a stronger negative effect⁵ on one-sided violence compared to the relationship between peacekeeping and the absence of mediation. If H2 were to be true I initially expect that the sole effect of mediation on one-sided violence is positive. Hence, mediation without peacekeeping has an increasing effect on one-sided violence. Furthermore, I expect to see an interaction effect illustrating that the presence of mediation relates to a decrease in the number of civilian deaths moderated⁶ by peacekeeping troops. Thus, the more the peacekeeping troops the greater reduction effect of mediation on one-sided violence. To reject the null-hypothesis for H2, I have to find evidence that there is an interaction effect between mediation and peacekeeping on one-sided violence and that this relationship is negative.

4 An interaction effect between two variables is present only if the coefficient of the interaction term is

statistically significant in an OLS regression model. If the term is statistically significant, there are two sides of that interaction. In this case the first side would be: peacekeeping troops effect on one-sided violence moderated by mediation, and the other side would be: mediations effect on one-sided violence moderated by peacekeeping troops.

5 Negative and positive effects is in this thesis discussed in relation to the statistical direction. Thus, negative effect is equal to a decreasing/reducing effect. Positive effect is equal to an increasing effect.

6 Moderated or moderating variable means in this context that there is a third variable that affects the correlation between the independent and dependent variable (Jaccard and Turrisi 2003).

(23)

3.4 Scope Conditions

Scope conditions are parameters or boundaries defining the circumstances to which the theory applies (Harris 1997). In relation to the theory presented there are a few scope conditions to consider. Firstly, for a peacekeeping mission to be present on site in a conflict zone, they need consent of the state of whom the territory belongs to. As a consequence, previous scholars have argued that peacekeeping is less effective against one-sided violence committed by governments (Fjelde, Hultman, and Nilsson 2019). However, while peacekeepers may solely be dependent on the consent of the state, mediation on the other hand is dependent on each conflict parties consent to be implemented. This practical aspect makes this theoretical argument position itself under the condition that each conflict party have deliberately acted to signal a commitment and/or willingness to resolve the conflict, meaning that this theory’s scope condition may differ from research only investigating peacekeeping’s effect on one-sided violence.

Secondly, in line with previous research investigating the interaction of mediation and peacekeeping (Beardsley, Cunningham, and White 2019), this thesis examine only UN peacekeeping in its main analysis⁷, excluding non-UN peacekeeping. Findings from previous research (Hultman, Kathman, and Shannon 2013; Hultman 2010), indicate that UN peacekeeping has a reducing effect on one-sided violence, which is consistent with the theoretical assumption - that peacekeeping on its own has a reducing effect on one-sided violence. Thirdly, when examining different types of peacekeeping operations, such as troops, police and observers, existing research has found that peacekeeping troops exerts a consistent significant effect on reducing one-sided violence (Hultman, Kathman, and Shannon 2013). In addition, over 80% of the UN missions consists of peacekeeping military troops making military troops the most common type of peacekeeping (Bara and Hultman 2020). As such this paper focuses on UN troops in the main analysis⁸.

7 As the theory is laid out and will be tested, non-UN peacekeeping fall outside of the scope conditions. However, alternative specifications are carried out in the extended analysis. Non-UN peacekeeping troops are there tested for and analyzed in relation to the main results. Thus, implications from the extended analysis may alter the scope conditions or confirm them.

8 While UN troops are the focus of the main analysis, further alternative specifications are carried out in the extended analysis. UN police forces has been found to have a reducing effect on one-sided violence (Hultman, Kathman, and Shannon 2013), and is therefore tested for. UN observers are also tested for; however, I do not expect UN observers to fit within the scope conditions of the theory since scholars have found it to increase one-

(24)

Finally, while many incentives to use one-sided violence have been discussed, the theoretical argument does not include all of them. Thus, incentives caused by disciplinary motives (R. M.

Wood 2010) or social aspects (Goodwin 2006) are not considered and thus fall outside the scope of this paper. Besides from political- and military incentives, one additional incentive which is referred to as a motivator by scholars is economic incentives⁹ (Weinstein et al. 2006). While economic incentives can be intertwined with political and military incentives, they can also stand on its own. When economic incentives are the only driver to commit atrocities like one- sided violence, usually greed stands behind the act, making conflict actors wanting a continued war due to the financial gain they get from the conflict (Keen 2000; Collier 2000). Hence, looting by rebel groups during conflict is not an uncommon practice and has been witnessed in Liberia, Sierra Leone and the Democratic Republic of the Congo among others (Azam and Hoeffler 2002). In war economies the conflict actor benefit from the war itself and have as such little incentives to act for peace, thus the actor has no reason to accept the aid from a third-party actor either in the form of peacekeeping or mediation. Since mediation demands the consent from all conflicting parties, economic incentives as a cause to commit one sided violence would most likely have to be addressed in a different way, and as such fall outside the scope of the theory.

9 Economic incentives, meaning financial motivators, a reason to do something to gain money.

(25)

4. Research Design

To test the hypothesis, I conduct a quantitative statistical analysis, examining correlates of the monthly counts of one-sided violence in all African intrastate conflicts between 1993 to 2007, since this is where the data availability overlaps for the independent variables and the dependent variable. One-sided violence in conflicts are identified using data from the UCDP Georeferenced Event Dataset (GED) Global version 17.1 (Sundberg and Melander 2013). This data covers organized violence globally between 1989-2016. It’s a disaggregated dataset, including individual events of organized violence, including one-sided violence. The data is fine grained both spatially and temporally (Sundberg and Melander 2013). In line with Beardsley, Cunningham and White (2019) I add twelve months of observation after conflict, hence months of “peace” without any battle related fatalities, since theoretical explanations associated with victimization may continue after the cessation of conflict attacks. Relevant cases, extracted from the population of interest - conflicts – are limited to the region Africa.

The regional focus on Africa, is mainly due to the lack of available monthly count data on mediation globally. However, since Africa is the region which by far has the most cases of one- sided violence fatalities, a 93% rate counted of all cases in the world between 1989-2004 (Eck and Hultman 2007), the regional sub-set still covers the majority of cases on one-sided violence.

Data on mediation is taken from the UCDP Managing Intrastate Conflict (MIC) Dataset (Melander and von Uexkull 2011). The dataset includes third-party interventions with focus on mediation in all conflicts in Africa between 1993-2007. The object of analysis is dyad-third- party-intervention event-day, making the data highly disaggregated (Melander and von Uexkull 2011). The data on peacekeeping troops is taken from Kathman’s (2013) dataset on personnel committed to UN peacekeeping from 1990 to 2011, counted at the monthly level. Since mediation and peacekeeping are usually deployed in temporal proximity it’s important to examine their contribution to reduce one-sided violence within a dynamic temporal count (Beardsley, Cunningham, and White 2019). Thus, a monthly count, which is more dynamic than yearly counts, has been important when selecting the data. To match the monthly count, the disaggregated data from the UCDP GED Global version 17.1 on one-sided violence and the

(26)

The independent variables and the dependent variable are operationalized below, taking into account validity, the accuracy of the measurements and reliability, the consistency of the measures (Kellstedt and Whitten 2018). Alternative explanations and plausible confounding variables are also accounted for. However, in attempting to rule out other explanations it is important to note that most work on third-party intervention are unable to find sources not affected by external factors causing additional variation (Beardsley, Cunningham, and White 2019). For example Kreutz and Brosché (2013) found that mediation is more likely to be active were large scale of one-sided violence against civilians has occurred, which will affect the ability to observe an appeasing effect of mediation. Likewise, Hultman (2019) found that mediation frequently follows one-sided violence conducted by government, whereas the same is not true for one-sided violence by rebels. In addition, both peacekeeping and mediation have been argued by scholars to likely occur in the most difficult conflicts (Fortna 2008; Gartner and Bercovitch 2006; Gilligan and Stedman 2003), making these variables exposed to many external factors frequently present in complex and severe conflicts. However, since conflict management tools like mediation and peacekeeping will never work in isolation from other external factors, strict causal identification distinct from external factors have little utility in understanding the impact of third-party processes, since these processes cannot be exogenous treatments (Beardsley, Cunningham, and White 2019). In fact, the strength of this thesis it to test mediation and peacekeeping not as exogenous variables, but as endogenous interactive variables.

4.1 Dependent Variable

One-sided violence was in the Previous Research section defined in accordance to the UCDP as the “use of armed force by the government of a state or by formally organized groups against civilians which result in at least 25 deaths per year” (Eck and Hultman 2007). To examine the variation in the dependent variable (one-sided violence) the UCDP GED Global Version 17.1 is used as the original dataset, relying on replication data from Bara and Hultman (2020). The data is compatible with the 25 death per year threshold, since the UCDP definition for one- sided violence are originally conducted for its datasets (Sundberg and Melander 2013). Since both the definition and the data on one-sided violence is taken from UCDP the variable fit well together with the indicator, increasing the validity. The dependent variable (one-sided violence)

(27)

is originally disaggregated spatially and temporally at an event level. However, in order to be compatible with the data on mediation and peacekeeping, the variable has been aggregated to a monthly count on civilian killings (Sundberg and Melander 2013; Bara and Hultman 2020).

Due to extreme outliers in the variable, such as the case of Rwanda, the variable is log transformed with a base 10 to make the distribution less skewed (Benoit 2011). The dependent variable considers the number of civilians killed in total, including both civilians killed by government and rebel groups. The one-sided violence accounted for is organized and intentional in line with the dataset, thus excluding unintended civilian casualties. In excluding accidental civilian killings, the findings reflect the influence of peacekeeping and mediation on deliberate one-sided violence, rather than it being spurious effects associated with random elements of war (Hultman, Kathman, and Shannon 2013). Thus, in solely measuring intentional and strategic one-sided violence, the measurement fits well with the theoretical logic presented in this thesis – that one-sided violence is a deliberate strategy and primary instrumental, increasing the validity.

4.2 Independent Variables

To measure the independent variables, peacekeeping and mediation, I have used replication data from Beardsley, Cunningham and White (2019), including original data from MIC Dataset on mediation and Kathman’s dataset on peacekeeping (Melander and von Uexkull 2011;

Kathman 2013). Beardsley, Cunningham and White’s replication data is suitable for the purpose of this study since their research also examines the interaction between peacekeeping and mediation. Regarding the reliability of the data - Beardsley, Cunningham and White (2019) rely on the same peacekeeping variable as Hultman, Kathman and Shannon (2013), and their results are consistent with each other, indicating that the reliability of the peacekeeping measurement is high. The MIC data has also been used by many scholars in the field (Beardsley, Cunningham, and White 2019; Ruhe 2020), providing consistency and stability of the measurement, also indicating a high reliability.

The concept of peacekeeping is in this thesis defined broadly (as noted in the Previous Research section), following the definition of the UN multidimensional peacekeeping, including aspects such as: the maintenance of peace and security, facilitating the political process, protect civilians, assist in the disarmament, demobilization and reintegration of former combatants,

(28)

support elections and assist in restoring the rule of law (United Nations Peacekeeping 2021).

To examine the effect of peacekeeping a one-month lag of Kathman’s (2013) monthly counts of the number of UN peacekeeping troops deployed to a specific conflict is used. Meaning counting the number of peacekeeping soldiers per conflict month in the preceding month of observation. Kathman’s (2013) data differentiate between three types of UN peacekeepers:

troops, police and observers. Since previous research has found that peacekeeping troops exerts a consistent significant effect on reducing one-sided violence (Hultman, Kathman, and Shannon 2013), I focus on UN troops in the main analysis. I line with previous research (Beardsley, Cunningham, and White 2019; Hultman, Kathman, and Shannon 2013), the peacekeeping variable is not logged¹⁰. Even though the variable is spread across a wide range, there are no distinct outliers to consider. Furthermore, since UN peacekeeping and non-UN peacekeeping can differ in how they function and also in how they are perceived as peacekeepers, the focus on UN peacekeeping in this paper increases the accuracy and thus the validity in the measurement. Although, there are additional minor aspects to consider regarding the validity.

According to Kathman (2013) there are some values coded as missing across troops, police and observers because information was not formatted correctly or unreliable. However, since the missing values are only a very small portion of the entire dataset it doesn’t affect the overall validity, but it is still important to be aware of.

The thesis adopts a broad definition of mediation, defined as: “voluntary third-party activity aiming at resolving or regulating the conflict through negotiations” (Svensson 2009: 455). To measure mediation the UCDP Managing Intrastate Conflict (MIC) Dataset is used (Melander and von Uexkull 2011). Two measures from the dataset will be used, both “direct talks” and

“indirect talks”. Direct talks meaning face to face discussions with a third-party mediator.

Indirect talks meaning negotiations through a mediator carrying the message, so called shuttle diplomacy. If direct- or indirect talks occur in a specific conflict month a binary indicator is coded 1, if not it is coded 0. In accounting for different forms of mediation, the measurements exploit the broad definition of mediation adopted in this paper, increasing the validity.

Moreover, a one-month lag on the variable mediation is placed, the same as on the peacekeeping variable, to account for the temporal order.

10 When testing the hypotheses with a logged peacekeeping variable the results remained the same as for when testing with the main (non-logged) peacekeeping variable. The results only differed on the main effects of peacekeeping on one-sided violence, without controlling for confounding variables. Thus, logging or not logging the peacekeeping variable does not have any notable implications on the main model.

(29)

The independent variable of significance in this thesis is the interaction variable. To measure the interaction between peacekeeping and mediation the variables are multiplied with each other to create an interaction term (UN troops*Mediation), measuring the combined effect of peacekeeping and mediation. In addition to the independent variables included in the main analysis, I conduct alternative specifications in an extended analysis, examining UN police, UN observers and non-UN troops effect on one-sided violence, combined with mediation. To examine the effect of UN police and UN observers a one-month lag of Kathman’s (2013) monthly counts of the number of each peacekeeping type deployed to a specific conflict is used.

To measure non-UN peacekeeping troops replication data from Bara and Hultman (2020) is used, originally from Stockholm International Peace Research Institute (SIPRI 2021).

4.3 Control Variables

I control for several potential confounding variables. Firstly, ceasefire is controlled for since it may ease the situation for the civilian population in stopping the fighting during a given period.

Thus, a lower level of violence against civilians would be expected if a ceasefire is in place and will most likely affect peacekeeping and mediation on site (Aduda and Bussmann 2019). The replication data by Beardsley, Cunningham and White (2019) includes a control variable for ceasefires, the same variable is used in this analysis. Secondly, I control for rebel strength given that the relative strength of rebel groups is theoretically relevant for the probability of one-sided violence being committed, both by governments and rebel groups. Also, third-party interventions are usually influenced by strong non-state actors (Koga 2011). The relative strength of a rebel group in conflict is controlled for using a five-level ordinal scale, using replication data from Beardsley, Cunningham and White (2019).

Thirdly, I control for battle-related violence in the previous month, meaning number of battle deaths in the month prior of observation. I specifically control for this variable to ensure that the effect I capture steam from one-sided violence and not conflict intensity. The levels of one- sided violence tend to correspond with the levels of battle intensity, which also corresponds with the military dynamics and thus influences peacekeeping and mediation efforts (Hultman 2007; R. Wood, Kathman, and Gent 2012; Fjelde, Hultman, and Nilsson 2019). The data on this variable is taken from the UCDP GED Version 17.1 (Sundberg and Melander 2013). Forth,

(30)

civilian targeting, peacekeeping and possibly mediation may be affected by population density (Fjelde, Hultman, and Nilsson 2019; Hultman, Kathman, and Shannon 2013a). I thus control for the logged size of population in using replication data from Bara and Hultman (2020). Fifth, in line with Bara and Hultman (2020), I include a lagged dummy variable (OSV dummy) of weather there was any one-sided violence (1) or not (0) in the preceding month. This is done to account for temporal dependency.

Sixth, democracy is controlled for since correlations has been found between higher levels of government one-sided violence in autocratic states, meanwhile rebel one-sided violence tend to be higher in democracies (Eck and Hultman 2007). Since the level of democracy is a notable push factor for third-party intervention, states which are less democratic may have a less welcoming perception towards third-party intervention and vice versa, thus affecting mediation and peacekeeping efforts (Kim 2012). Democracy is controlled for buy using V-Dem Dataset Version 11.1 (Coppedge, Gerring, and Knutsen 2021). Two indicators from the dataset, measuring different aspects of democracy, are controlled for. Liberal democracy index, includes protected civilian liberties, strong rule of law, independent judiciary, effective checks and balances and the level of electoral democracy. Freedom of expression (and alternative sources of information) index, measures to what extent the government respects free press and media, including the expression of political, academic and cultural views. Both indicators measuring democracy are scaled as an interval from low to high (0-1).

Finally, Gross domestic product (GDP) per capita is controlled for. Both a resource fatigue government and rebel groups have according to previous research different incentives to use one-sided violence. Hence, rebels may perceive one-sided violence as an inexpensive way to inflict cost upon the opponent meanwhile states with little recourse may have the same incentives if civilians are perceived to support the rebels. Furthermore, GDP per capita is a measure of state capacity, development and opportunity cost which is expected to make violence less likely if it is high (DeRouen and Chowdhury 2018). As such, it may also affect the deployment and use of mediation and peacekeeping, since the two are deployed in conflict zones where the state capacity is usually weak. To measure GDP per capita, data is taken from the World Bank, counted in the currency US dollar (The World Bank 2021). Table 1 below presents a summary of all variables and their respective descriptive statistics.

Table 1. Summary descriptive statistics

(31)

Variable N Mean SD Min Max

Civilian Deaths (log) 2360 251.445 5809.104 0 237437

UN Troops (lag) 2588 854.48 3211.606 0 29209

Mediation (lag) 2568 0.129 0.336 0 1

UN Police (lag) 2588 27.526 135.689 0 1132

UN Observers (lag) 2588 33.475 114.792 0 1039

Non-UN Troops (lag) 2360 248.441 1329.897 0 14000

Ceasefire 2588 0.395 0.489 0 1

Rebel strength 2588 2.234 0.813 1 5

Battle related violence (lag) 2588 52.219 281.154 0 9793

Population (log) 2360 7.055 0.439 5.701 8.143

OSV dummy (lag) 2360 0.411 0.492 0 1

Liberal democracy index 2130 0.168 0.131 0.006 0.535

Freedom of expression index 2130 0.53 0.227 0.014 0.898

GDP per capita 1884 2099.555 2038.194 680.801 10974.742

4.4 Modeling Interaction Effects

Initially, the main effects¹¹ are presented in order to accurately assess the interaction effect. To illustrate the respective main effect of peacekeeping and mediation on one-sided violence, I utilize scatterplots illustrating the relationship between peacekeeping and one-sided violence, and a box plot illustrating the relationship between mediation and one-sided violence. In addition, OLS regression models are conducted to present the coefficients of the respective effects when controlling for confounding variables (Appendix I and Appendix II).

To test the hypotheses, I utilize an OLS regression model. The OLS regression model was chosen as my dependent variable is continuous. The OLS model can demonstrate if the interaction variable is significant and of importance, which is an additional advantage with this model compared to others (Mize 2019). It can also illustrate if the significance of the interaction

11 The effect of one independent variable on the dependent variable, ignoring all control variables. Thus, the respective effect of peacekeeping on one-side violence, and mediation and one-sided violence.

Ending Civilian Victimization: The Combined Effect of Mediation and Peacekeeping on Violence against Civilians