The Effect of Arms
While firearms by themselves are very seldomly the underlying reasons behind lethal violence, they make carrying it out significantly easier compared to other methods such as bare hands and bladed weapons. The reasons why humans subject other humans to violence vary greatly from individual level motivations such as frustration or envy (Benjamin et al. 2018: 348-9) to societal motivations such as collective grievances (Stewart 2011) and greed (Ross 2004). In many cases, especially with collective violence conducted between large groups of people, the motivation is funnelled through the barrel of a firearm or the gun of a tank. As such, the study of violence and conflict has given significant attention to the role arms1 play in the phenomena.
On the macro level of nations and the international system, studies – much like on the micro level – often find at least some level of support for the idea that an increase in arms results in an increase in the level of violence.2 Before looking at the findings, it is important to note that there are different types of arms that are relevant for understanding armed violence within societies. Often when mentioning firearms, people mean small arms and light weapons (SALWs). These are what most think of when thinking about firearms: pistols, rifles, machine guns and mortars to name some (OSCE 2012: 2). These light and easily transported weapons are often designed to be used by a single person. However, the innovations and demands of modern information technology-driven industrial warfare have made it possible to move to a larger scale of destruction. Nowadays, most countries have vast supplies of highly technical weapon systems that often require many people and expertise to use. These major conventional arms (MCWs) such as tanks, aerial vehicles, warships, and missile systems (UNODA 2017: 4-5) are not usable on the individual level but come to have potentially big impacts in the warfare of organised groups. Partly because of their perceived
1 For the purposes of this thesis, the term “arms” includes major conventional weapons (MCWs) and small arms and light weapons (SALWs). Bladed weapons are not included in the blanket term.
2 For studies that focus on the impact of arms on micro level violence between individuals, see, for example, Stroebe (2013 and 2016). significance and, perhaps more, because exhaustive data on MCWs is more available (see, for example SIPRI 2022a) studies on the relationship between arms and intrastate as well as interstate violence have tended to focus more on MCWs. This is so even amid experts’ reminders about the severe human rights impacts SALWs have for civilians in conflicts (UNODA 2018: 40).
With the above in mind, let us look at the literature. Craft and Smaldone (2003) find that arms imports are linked to an increased likelihood of conflict outbreak. However, these findings can only be viewed as preliminary ones for two reasons. First, the SIPRI dataset used in the article only includes MCWs whereas most of the violence in conflicts – at least intrastate ones – is conducted with small arms and light weapons SALWs (Craft and Smaldone 2003: 43). Second, the study lacks a control for endogeneity. In other words, we cannot be sure that it is arms that promote conflict, and not leaders bracing for a likely conflict by increasing their arms supply (Pamp et al. 2018: 431). To address the endogeneity issue, but not the SALW issue, Pamp et al. (2018) employ an instrumental variable method that separates MCWs suitable for use in civil war from ones that could not be used. Looking at the imports of both, they can determine whether the country is preparing for civil war by importing arms that are suited for it, or if the import of any arms works as an escalatory step that promotes conflict onset’s likelihood. Their results indicate that arms imports moderately increase the likelihood of conflict onset in countries with low state capacity, but result in only a marginal increase in other countries.
Because the number of arms logically increases in a society undergoing armed conflict, it is hard to effectively eliminate the two variables’ covariation. Here looking at cases where conflicting societies are forced to halt the import of more arms lends a hand. Arms embargoes are a reactive tool used by the international community to prevent the inflow of new arms and thus alleviate the human suffering caused by armed conflict (Pattison 2018: 70-1). Studies on arms embargoes have indeed found results that support the idea that arms play an important role in conflicts. Despite arms embargoes never being perfect in stopping the inflow of arms (Vines 2007) they do still in most cases significantly reduce the inflow and availability of new arms – both SALWS and MCWs – for the sanctioned actors (Erickson 2013). Research has found that arms embargoes, and by extension stopping arms flows, has some success in influencing conflict. Brzoska’s (2008) results indicate that implementing an arms embargo contributes to an improvement in the human rights situation in 6% of cases, ending civil wars in 12% of cases, and ending support for terrorism in 10% of cases. Furthermore, Hultman and Peksen (2007) find that arms embargoes – unlike economic sanctions – reduce the number of battle-related deaths in intrastate armed conflicts. They theorise that the reason for this might be that while arms embargoes and economic sanctions both hurt the coercive capacity of the actors they target, it is only the arms embargoes that reduce the actual military capacity that is used to fight battles and determines the strategies available for the military to use.
On the surface, these findings seem to lend support to the findings of Craft and Smaldone (2003) and Pamp et al. (2018) which state that arms availability increases conflicts’ likelihood. However, we must be cautious when making such inferences. The efficiency of arms embargoes might rest on some other mechanism than the simple reduction of arms that are used to inflict bodily harm. In fact, research suggests that arms embargoes, like other sanctions, work mainly because they are a way for the international community to socialise states into international metanorms (Erickson 2020). Even still, the finding of Hultman and Peksen (2018) is particularly valuable since it is one of the few studies that has moved beyond treating conflict or conflict onset as the dependent variable and started to estimate not only the direction, but also the degree of the impact arms have on deaths from conflict. This approach helps to reduce the guesswork and abstraction indebted in much political science theorising and knowledge.
Articles specifically attempting to gauge the direction and degree of the impact that arms have on conflict duration and intensity are few. In fact, I have only been able to find two such studies. Moore’s (2012) regression analysis shows that MCW transfers to rebels are associated with higher conflict intensity measured with battle-related deaths while MCW transfers to the government increase conflict duration. For governments, transfers prior to conflict might also increase conflict intensity but the coefficient has low statistical significance, so Moore is cautious with his claims. Unfortunately, Moore uses ordinary least squares (OLS) regression with his count data on battle-related deaths, so the degree of impact cannot be determined with confidence (Du et al. 2021). Further, Moore does not include SALWs in his analysis. While the evidence from multiple case studies suggests that the amount of SALWs increase the casualty rate of civil wars (for example, Mehrl 2017, Sislin and Pearson 2006, Tar and Onwurah 2021), as far as I am aware, only one systematic academic study has tried to quantify the effect SALW availability has on the level of battle-related deaths. In the study, Mehrl and Thurner (2020) utilise similar methods as Pamp et al. (2018) but direct their attention to understanding how many more people die when the number of arms increases in civil war societies. Their results suggest that both SALW and MCW imports increase the deadliness of conflicts. However, the finding is conditional to the relationship between rebel and government military strength measured with troop size. If governments face strong rebels, then the import of either class of arms results in a considerable increase in battle deaths, whereas imports do not cause any effect when governments are facing weak rebels. Much like Moore (2012), Mehrl an Thurner (2020) also use OLS with count data making their findings less robust.
While previous research has given us a relatively solid understanding that arms are linked to increased risk of interpersonal violence and conflict deaths, important gaps remain. Here I will focus on one. What all the previous studies have not yet done, is to separate the civilian casualties from the total picture. Only a rather crude and tentative report from the International Committee of the Red Cross (ICRC 1999) has concluded that arms availability increases civilian targeting during and after conflict, and that 90% of those interviewed knew a case where a civilian had been targeted intentionally (ICRC 1999: 14). Besides this, studies analysing the impact of arms have, so far, focused on conflict as the dependent variable, and only in a few instances on aggregate death tolls that include combatants and civilians in a unified category. However, disaggregating civilians from the total figure has normative and theoretical justifications. Normatively, civilians are outside of accepted casualties of war (IV Geneva Convention 1949) and significant international efforts are made to protect them. Theoretically, civilians are the most vulnerable group in conflicts because they have no fighting capacity. This makes them easy targets for the armed forces of states or rebel groups. With this goal in mind, let us proceed with a review of the literature on civilian targeting before addressing the research gap in detail.
Intentional targeting of civilians – One-sided Violence (OSV)
OSV in Civil Wars
Despite being a relatively young and a growing field of study, research on violence against civilians (OSV) has produced important insights into the form and logic of this phenomenon that is responsible for nearly a third of all the conflict deaths since 1989, or more than 800,000 deaths (Pettersson and Öberg 2020: 598). Previous research suggests that most violence against civilians is perpetrated during civil wars (Valentino 2014: 94-6, Schwartz and Straus 2018: 222). In fact, only around one per cent of OSV happens in countries that are not currently involved in armed conflict of any type (Eck and Hultman 2007: 237). As a grim reminder of the banality of evil, in times of civil war, civilians are more commonly intentionally targeted rather than accidentally killed (Schneider and Bussmann 2013: 640).
Of course, not all civil wars are fought for the same goals and with equal tactics, so treating them as a unified block will likely lead to findings that do not correspond to real world complexities and thus to policies that fail to protect civilians. In general, researchers have divided civil wars into two types based on the way they are fought, separating irregular civil wars from conventional civil wars (Balcells 2010). The main difference between the two boils down to the level of clarity in separating the opponents. Irregular civil wars are characterised by the lack of easily definable areas of control and frontlines, with the insurgent/guerrilla troops often small, living with civilian populations in clustered and small areas of support nested inside the opposition’s area of control (Balcells 2010: 295-6). Conversely, conventional civil wars have major battles fought along the frontline that separates the opponents’ large areas of total control from each other. It is to be expected that small rebel groups are more likely to pursue irregular tactics whereas big rebel groups use conventional tactics (Bueno de Mesquita 2013: 324). Studies that have delved into the dynamics of these specific types of intrastate conflicts have often found that the tactics, used in the civil war have specific impacts on the level of civilian targeting in them.
Researchers have a rather unanimous consensus about irregular civil wars that are fought with insurgency or guerrilla tactics resulting in more civilian targeting than conventional civil wars (see, for example, Balcells and Kalyvas 2014, Balcells and Stanton 2021: 51, Valentino 2014: 94). The reason for this finding is more debated than the finding itself. In their qualitative analysis of the Guatemalan Civil War, Schwartz and Straus (2018) first present four possible logics for the strategic use of violence against civilians. It is possible that the actors want to weaken the enemy by coercing its civilian supporters, they can also try to enforce compliance from the civilians with violence, there can be a lack of intelligence or ability to separate enemy combatants from civilians, or finally, the actors may view the civilians as the enemy – or at least an unchangeably loyal support base for it. Although these mechanisms are not mutually changeable, the article only finds support for the first two in the context of an irregular civil war. Logically, the first two are not well applicable to conventional civil wars with large rebel groups, since total control of an area makes it extremely difficult for the civilian population in the area to support the opposition locked into its own area behind the frontline. On the other hand, despite Schwartz and Straus not finding support for them in the Guatemalan Civil War, the latter two mechanisms are still possible. In fact, the last mechanism in which civilians are seen as the enemy can be particularly destructive since it can lead to a “genocidal logic” (Schwartz and Straus 2018: 225). If the civilian population is a primordial supporter of the enemy, then, in the minds of the decision makers, complete victory can necessitate complete or near complete destruction of the civilian population too. This logic in a civil war context is perhaps most glaringly illustrated by the Rwandan Civil War (for an overview, see Prunier 1995).
Military Capacity and OSV
The attentive reader may have spotted a word which was repeated throughout the above discussion of OSV in civil war but has not yet been discussed in detail: support. Many of the theories on the strategic use of OSV rely on the assumption that governments and rebels are not completely self-sufficient but that they must supplement their war-fighting capacity with external resources that they can extract from civilians. As a result, researchers have tried to gauge the impact organisational capacity has on both rebel and government perpetrated OSV.
Actors involved in conflicts are often backed by international actors such as diaspora communities, foreign governments, or like-minded rebel groups (Petrova 2019). This type of external support has often been linked to more violent civil wars because the rebels that do not depend on the civilians for support face fewer penalties to using violence in large scale than rebels who might lose their civilian support if they started targeting civilians (see Weinstein 2007). The few academic articles on external support and OSV – all of which focus on rebels – have drawn similar conclusions. In his quantitative study which compares rebel groups with local support vs. foreign sponsorship, Wood (2014b) finds tentative support for the idea that foreign sponsorship for rebels (including arms) leads to more civilian targeting by the rebels. However, despite the foreign sponsorship coefficient being positive and practically significant in three of his four models, it is never statistically significant. In a similar study, Salehyan et al. (2014) however do find that external support to rebels is likely to lead to increased OSV – but not when it comes from democratic states.
Still, the dynamic is not as clear cut as Huang and Sullivan (2021) show. When looking at the impact of external military support (such as funding, arms, and training) to rebels from the perspective of positive externalities, they find that rebel groups that have foreign sponsors are likely to significantly increase their social service provisions to their respective civilian supporters. So, while the other studies depart from the assumption that rebels become less dependent on civilian support as they get support from outside the country, often they still need grassroots support such as recruits and accommodation from local civilians (Huang and Sullivan 2021: 805). Despite relying on foreign support whenever it is available, rebels and governments in many cases need to ensure they have local sources of support and domestic capacity to sustain their fight. One important resource that is not readily available from abroad is recruits. For both the government and rebels, civilians are a possible recruitment pool that can be syphoned either voluntarily or coercively. For example, in Rwanda, the government mobilised civilians to perpetrate much of the violence, with the military and police troops setting the example and providing the arms (Loyle and Davenport 2020). So, understanding the impacts that domestic military capacity has on OSV remains important.
Kalyvas (2004: 121-124) argues that rebels are perhaps less dependent on civilian support than often thought. Using case examples, he shows how government attacks on civilian population rarely result in the rebels terminating their campaign – be it because of wanting to protect civilians from more harm or losing the necessary support base. Adding to this, Wood (2010) finds that rebels that are comparatively weak in relation to the government are more likely to use higher levels of OSV compared to strong rebels. Wood’s analysis suggests that the reason for this is the weak rebel’s attempt to hold onto the civilian support that is vital for the continuation of their fight. If the government targets the rebel group’s civilian supporters with its own OSV campaign, weak rebels’ level of OSV increases significantly whereas this effect is not observed among strong rebels. Wood’s (2014a) later study strengthens the argument’s empirical support. He finds that rebels are likely to increase civilian targeting following significant material losses. Possibly so because when they are unable to recover from the losses with their own funding, external support, or willing civilians, they need to coerce civilians with violence to get backing (Wood 2014a: 996). However, Wood’s analysis suffers from the simplifying assumption that rebel capacity is equated with nothing more than the number of rebel troops compared to government troops (Wood 2010). It lacks many crucial elements of military power such as the amount and quality of arms as well as the level of training (Horowitz 2010: 13).
Thus, we come to a rather interesting and empirically unexplored counterargument to the idea that weak troops target civilians more than strong ones. Other researchers believe that organised forms of mass civilian targeting are only available to the powerful. Bombings or large-scale ethnic cleansing are only possible for highly organised and equipped skilled troops (Wood 2014b: 464-5, Stanton 2016, Zhukov 2017, Balcells and Stanton 2021: 55). The notion starts to add depth to Wood’s analysis without necessarily contradicting it. Wood’s (2010) finding that rebels with weak military capacity kill more civilians than strong rebels do on the aggregate level, does not contradict the possibility that when strong rebels decide to target civilians, they can cause much more deaths per event. Thus, it remains possible that weak rebels target civilians often but kill few civilians per event, whereas strong rebels or governments target civilians seldomly but those events result in more deaths per event due to the high military capacity. In fact, this exact mechanism is hinted at in Eck and Hultman’s (2007: 237-40) finding that rebels commit OSV more frequently than governments, but that government perpetrated OSV results in mass deaths significantly more often. Because these articles do not focus on arms in their analysis of support or organisational capacity, even though arms arguably are one crucial determinant of any armed organisation’s capacity, looking at the relationship between arms imports and OSV could prove fruitful in improving our understanding of why some organisations decide to target civilians and others do not.
The Research Gap
This dissertation aims to study the impact of a specific type of military power, namely arms, on government and rebel perpetrated OSV in civil wars. Previous research has hinted at the damaging effect different types of arms have for the human suffering and mortality of civilians in civil war contexts. We still need, however, a systematic study looking into the degree of which arms imports increase purposeful civilian targeting. Often studies do not include arms in their analysis of capacity and OSV (Wood 2010), or where they are included, they are coded into an over encompassing variable such as “external support” or “capacity” (Huang and Sullivan 2021) which fails to capture the specific impact arms have. Furthermore, these studies are overwhelmingly focused on the rebels, and as a result the understanding of government perpetrated OSV is narrower. Thus, this thesis seeks to answer the following research question: How do imports of small arms and light weapons (SALWs) and major conventional weapons (MCWs) impact the level of one-sided violence (OSV) against civilians within intrastate conflicts?
Understanding the effect of arms – both SALWs and MCWs – on civilian targeting is important so that protocols, mandates, and agreements on the protection of civilians can be improved and the international community can better evaluate the risk associated with arms exports to conflict areas. If arms exports have no negative effect, they may even be viable in protecting civilians by making the conflict briefer by decreasing the rebel group’s militarily capacity. If, however, arms imports tend to boost the targeted killings of civilians, then arms imports – potentially even to peacekeeping forces, but more studies on this would be needed – need to be re-evaluated.
Table of contents :
Acronyms and Abbreviations
2. Literature review
2.1 The Effect of Arms
2.1.1 National level
2.2 Intentional targeting of civilians – One-sided Violence (OSV)
2.2.1 OSV in Civil Wars
2.2.2 Military Capacity and OSV
2.3 The Research Gap
3. Theoretical framework
3.1 Definitions of key concepts
3.1.1 The Dependent Variable
3.1.2 Independent Variables
3.2 SALWs’ impact on OSV
3.3 MCWs’ impact on OSV
4.1 Operationalisations and Data
4.1.1 Dependent Variable
4.1.2 Independent Variables
4.1.3 Control Variables
5. Empirical Analysis
5.1 Arms Imports and Total OSV
5.2 Arms Imports and Rebel and Government OSV
5.2.1 Rebel OSV
5.2.2 Government OSV
5.3 Robustness Checks
5.4 Limitations of the Study