Civil War Onset - A Comparison of Uganda and Kenya

Dissertation

Civil War Onset – a Comparison of Uganda and Kenya

Introduction

This paper poses the research question: what causes civil war? Since the end of World War II, both incidence and duration of civil wars have been on the rise, almost always accompanied by human suffering and diminishing hopes for economic development (Hironaka 2005). It is crucial that academics study this phenomenon and try to create theories which are able to explain and predict civil war onset for different countries.

Two main competing theories have become prominent in the modern day literature surrounding the outbreak of civil war. On the one hand there are “greed” theorists employing econometric models to account for rebel opportunism. These theories centre around the notion that ethno-linguistically or religiously diverse countries experience civil war as a result of the incentive to rebel compared to the state’s ability to counter rebellion. Greed theories focus largely on the ability of rebel groups to recruit and finance themselves, in addition to a number of other variables, which make civil war more or less conducive. On the other side are advocates of the “grievance” theory who argue civil wars start as a result of grievances, built up as a result of political and material discrimination. Depending on the level of grievance in combination with the ability of ethnocultural groups to mobilise and the state’s response to initial protest, civil war occurs. This paper predicts that the grievance theory, better accounts for civil war onset.

Hypothesis: Grievance theories of civil war onset hold more explanatory power than Greed theories.

This hypothesis is tested through the use of two case studies with opposing dependent variables. The first is Uganda which has experienced multiple internal conflicts of varying intensity since gaining independence. The other is Kenya which has been spared the outbreak of a full-blown civil war, although it has experienced a number violent ethnic clashes. The case studies are relatively similar so as to control for third variables, yet chosen in such a fashion as to avoid bias or case fixing. For both countries the greed model is firstly applied followed by a more in-depth look at whether the rationale behind the given independent variables corresponds with political-economic realities. The second part of each case study looks at the applicability of grievance based theories and which set of explanations holds more power with regards to civil war onset. The analysis of both models in the context of the two chosen case studies should determine which theory holds the most explanatory power in relation to civil war onset.

Literature Review and Research Design

Introduction

The following chapter discusses the existing literature on the subject of civil war onset, focusing on the greed-grievance debate as first distinguished by Collier (2000). As mentioned, the greed side traditionally argues from an individual economic point of view, stating that individuals who form rebel groups have an economic agenda based on the opportunity to rebel. The academic duo Collier and Hoeffler have been the most influential with regards to the greed theory. Other recent prominent academics on the subject include Fearon and Laitin (2003) and Miguel et al. (2004). The grievance theory focuses more on the group level and the importance of collective identity, rather than the individual. Proponents of the theory argue that identity based groups (ethnic, religious, linguistic, cultural etc.) want to redress earlier grievances which have their roots in political or economical oppression. Ted Gurr is probably the best known academic with regard to grievance theories of civil war onset. Stewart and Østby are likewise proponents of the grievance theory, but focus more specifically on so called horizontal inequalities (HIs).

The second part of this chapter deals with the research design. This paper uses the countries of Uganda and Kenya as case studies. They are relatively similar with regards to the major variables put forward by both sides of the debate, except for their dependent variable: Uganda has experienced a number of internal conflicts of varying magnitude, whilst Kenya, despite occasional bursts of politically motivated violence, has not experienced a full blown civil war.

The research design gives a justification for the use of the aforementioned countries as case studies. This project looks both at quantitative and qualitative research from the debate, and for illustrative purposes, gives both in-and-out of sample predictions of quantitative studies, where possible. The ultimate focus however, is on whether there is causation between the independent variables of the two theories, or merely correlation, with regards to civil war onset, which will be attained through in-depth qualitative analysis.

Literature Review

Greed

Greed based theories of civil war onset have their roots in micro-economic theories of individual rational decision making. Most try to ‘prove’ the causes of civil wars through quantitative macro analyses, around which they attempt to build a general model.

Most recent economic explanations on the occurrence of civil wars, focus on large quantitative studies. The main difference with grievance based theories, is that they argue from a opportunity or feasibility point of view. Civil war is not driven by a motivation to redress earlier grievances, but will occur if specific variables are conducive to permitting rebellion. The most notable contemporary academics arguing from this point of view include Collier and Hoeffler (1998, 2000, 2004), Collier et al. (2009), Fearon and Laitin (2003) and Miguel et al. (2004). The Collier and Hoeffler model, hereafter referred to as CH, is the most prominent within donor circles and the Western media, but their latest model incorporates aspects from both previously mentioned studies by Fearon & Laitin and Miguel et al. (Murshed & Tadjoeddin 2009). For these reasons, this paper will primarily focus on the CH model. The independent variables and different mechanisms (hypotheses) behind these variables are further discussed in the Research Design.

The studies mentioned above are quantitative in nature and therefore would appear not to be suitable for case studies as in this project. Sambanis (2004) in Using Case Studies to Expand Economic Models of Civil War, argues that case studies are useful when analysing econometric models of civil war onset. He writes that it allows us to get a better understanding of the causal mechanisms behind the independent variables which in turn influence the dependent variable. This method has been used by academics such as Stefan Blum (2006). Another important aspect, especially in relation to this paper, is the ability of qualitative analyses to determine whether the results of aforementioned models are a result of the independent variables put forward in the model or whether they merely correlate.

Grievance

Grievance theories on civil war have tended to be more qualitative rather than quantitative. Rebellion is seen as a means to redress certain grievances, which can be political, social or economic in nature. Grievance theories are to a certain extent also opportunistic in character as they argue that the lower the cost of violence, especially as a function of state repression, the more likely that violence will occur, but usually comes secondary to motivation. According to Sambanis (2002) most political scientists advocating the grievance model argue that the root cause is often political oppression. Since the start of the millennium however, another movement within grievance based theories of civil war has been gaining momentum. These academics focus on horizontal inequalities. More specifically; social and economic inequalities (which often go hand in hand with political inequalities) between different, usually ethnic groups.

Probably the most prominent academic who writes from a grievance perspective, is Ted Gurr (1970, 1993, 2000), founder of the Minorities at Risk programme. His publication Peoples versus States (2000), stresses that group identity is important in creating shared preferences which allows that group to adopt certain rational strategies. Through the forging of group identity, Gurr attempts to explain rational choice of political violence for different ethnic or religious groups.

With regards to grievance based theories of horizontal inequality, Stewart (2000, 2002, 2010) and Østby (2005a,b 2006, 2008) stand out. The former employing qualitative analyses using case studies and the latter being the first to perform a number of large-N studies on the subject, based to a large extent on previous studies by Gurr and Stewart. Stewart argues similarly to other grievance advocates, especially Gurr, that civil war is more likely a result of motivation resulting from group based grievances rather than individual opportunity seeking. Where Stewart differs from other academics is that she focuses on social, economic and cultural inequalities between groups in addition to political inequalities, although she admits that all three are intertwined. The different independent variables and their related mechanisms are further discussed in the Research Design.

Research Design

Introduction

This paper analyses which theory of civil war onset (greed/grievance) is most likely to explain the outbreak of civil war. The countries of Uganda and Kenya have been chosen as case studies, the former country having experienced four civil wars (according to this paper’s definition of the term), whilst Kenya has not experienced a full-blown civil war, despite experiencing episodes of ethnic violence.

The method of research for this project is in-depth qualitative analysis. As mentioned above, greed theories tend to be in the form of large N-studies, whilst grievance theories are usually supported through more qualitative analyses of case studies. This however does not mean that the two are incomparable, as both theories put forward clear independent variables which can be assessed for their relevance as causal mechanisms.

This dissertation is related to a previous paper by the author on the subject of civil war onset. This paper however, differs significantly from the previous by employing more contemporary and varying published works by academics from both sides of the debate, in addition to using different countries for the case studies. With regards to the greed theory, Collier and Hoeffler’s 2004 model remains central, but is supported by alternative independent variables given in the Collier et al. 2009 model. The author’s previous paper used Hegre et al. (2001) as a grievance theory advocate, however, his model - pertaining mainly to state strength and level of democracy - is used in both greed and grievance theories. Instead, this paper has shifted it’s focus to academics such as Stewart and Østby who look at horizontal inequalities in relation to civil war onset. Finally, alternative case studies have been used, with increased within-case variation. The grievance sub-chapters now contain more hard data, so as to make them more symmetric in relation to empirical evidence given for the greed theories.

Dependent Variable

The dependent variable in this project is “civil war onset”. This term implies two things. The first is the use of the term civil war, which is different from civil conflict and implies a certain magnitude of violence. Secondly, this paper specifically focuses on the onset of civil war, rather than duration, as these are influenced by differing variables. In order to conduct a fair and balanced analysis of which theory better explains civil war onset, a definition of civil war onset is needed. Naturally, the quantitative studies analysed in this paper give the most precise definition of civil war onset. Hegre and Sambanis (2006), argue in a sensitivity analysis, that there are differences in empirical results of quantitative analyses of civil war, as a result of different definitions of civil war. For this reason it is important that a compatible definition is found. Collier et al. (2009) use two sets of data; the correlates of war project (COW) for wars until 2002 and the armed conflict dataset (ACD) for the years 2003 and 2004. Østby similarly uses the armed conflict dataset, but focuses on the onset of civil conflict rather than civil war. The difference between the two is battle related deaths per year, in CH’s case there needs to be a minimum of a 1,000, whilst for Østby 25 deaths are sufficient. The ‘lowest common denominator’ in this case would be the definition in the CH model, who add the following definitional parameters; organised military action, effective resistance in the form of 5% of the deaths being inflicted by the weaker party and the national government being actively involved. The fact that Østby uses a different definition should not be a problem as studies by Østby and Stewart refer indiscriminately to conflict and war, indicating that the two different magnitudes of conflict are dependent on the same variables. For this reason the CH model can be seen as the ‘lowest common denominator’ as Østby’s model is compatible with CH’s definition for civil war onset, whilst CH’s model is incompatible with Østby’s definition for onset of civil conflict. For the purpose of clarity: CH’s model codes the following years for civil war onset in Uganda since independence: 1966, 1980, 1996 and 2004 (see Collier et al. 2009: pp. 7, Table 1). With regards to Kenya, the CH model does not register any civil war onset since independence.

Independent Variables

This project takes its independent variables from five of the most prominent academic writings on the debate. These include three quantitative theories; Collier & Hoeffler (2004), Collier et al. (2009) and Østby (2008). In addition, independent variables identified in the qualitative studies of Stewart (2002, 2010) and Gurr’s Peoples versus States (2000) are also used. All of the above works give multiple independent variables determining the onset of civil war, nonetheless there is a clear difference between the two sides of the debate. I will now discuss the different variables of each work starting with the greed side (Collier & Hoeffler, with reference to Fearon & Laitin and Miguel et al.) and then the grievance side (Gurr, Stewart & Østby)

[...]