Steven M. Radil
Introduction
In late October 1996, the forces of the rebellious Alliance of Democratic Forces for the Liberation of Congo (ADFL) began the military campaign that would ultimately lead to the end of the Zairian state, now called the Democratic Republic of the Congo (DRC), in the following year. The emergence of the ADFL and the collapse of Zaire also marked the beginning of nearly two decades of internationalized civil war centered on the eastern DRC. The conflicts, which have been grouped together into the First Congo War (1996â1997) and the Second Congo War (1998â2003) respectively, have been described collectively as the deadliest and most destructive armed conflict in the international system since the Second World War (Coghlan et al. 2007). However, the processes by which the events escalated from a political crisis over control of the Zairian state in 1996 into what Prunier (2009) has called a âcontinental catastropheâ remain incompletely understood.
Prunierâs use of the word âcontinentalâ highlights the vast geographical reach of the wars. For example, less than a month after the emergence of the ADFL, the governments of nine states in the region were at war either with each other, with some of the myriad armed non-state groups that were active at the time, or both. Nonetheless, the Congo wars have received relatively little attention from scholars of political violence and nearly no attention within geography (notable recent exceptions include Raeymaekers (2013), Radil and Flint (2013), Vlassenroot (2013) and Doevenspeck (2016)). While the relative lack of expertise in African politics among the Anglophone war studies community may partially explain this oversight (see Radil and Flint 2015), the case of the Congo wars offers a unique opportunity to better understand the processes associated with war diffusion, or the expansion of war within a particular geographic setting or regional context. Although diffusion is generally considered the spread of something over time, it is also an expressly geographic phenomenon, since something that spreads over time also involves the spatial expansion of the same (Gould 1969). Therefore, understanding the spread of war necessarily requires a geographic perspective.
The need to understand how and why some wars, such as the Congo wars, diffuse and expand geographically is a pressing one given the persistence of war in the international system. For instance, the involvement of dozens of non-state armed groups and of multiple states in the ongoing civil wars in Syria and Iraq are evocative of the circumstances of the Congo wars; a contest over control of a state that then spirals beyond the boundaries of that state into a regional quagmire that takes decades to resolve. Such wars with their morass of participants and large regional footprints have been labeled a âproto-world warâ (Peçanha et al. 2015) or a war with the potential to grow in scope, to draw in new participants, and perhaps to encompass most of the major political actors of the time. As such, the idea of a proto-world war is an implicit reference to the notion of war diffusion.
Taken together, the Congo wars are an important set of cases for the investigation of war diffusion, as they clearly expanded from a âproto-regional warâ to an actual regional war. This chapter then considers the issues connected with investigating war diffusion in such a context by drawing upon the authorâs previous work on war diffusion in general (Radil et al. 2013) and the Congo wars in particular (Radil and Flint 2013). This investigation is also grounded in the authorâs Spatializing Social Networks (SSN) framework (Radil et al. 2010) which uses the concepts and techniques of social network analysis (SNA) to integrate geographic and political data into a single relational analysis. Although much of the SSN approach has been descriptive in nature to date (see Flint et al. (2009) for a discussion of the general approach), this chapter uses data from the First Congo War to develop a network-based statistical model called an Exponential Random Graph Model (ERGM) to consider the geographical and political drivers of the war.
War diffusion and the SSN framework
Geographical approaches to the concept of diffusion have rested on two basic observations: that human activities take place in both time and space, and that social processes underlie the realization of any empirical diffusion patterns, whether temporal or spatial (Morrill et al. 1988). The first observation leads to a concern for change or dynamism by geographers; the second to a concern for understanding the processes that produce such change (OâSullivan and Unwin 2010). These observations also serve to locate geographical diffusion studies within the spatial analysis tradition in human geography (OâLoughlin 2000), emphasizing the use of mapping and explicitly spatial statistical analyses to aid in the understanding of the diffusion of a great variety of topics. For example, the geographical diffusion studies have ranged from the spread of technology (Hägerstrand 1967), democracy (OâLoughlin et al. 1998), public policies (Lutz 1986) and crime (Cohen and Tita 1999).
The idea of diffusion is an old one within geography. However, Gouldâs (1969) treatise on spatial diffusion remains the definitive work within the field (see also Hägerstrand 1967). At the most basic, the key points made by Gould are that any diffusion process consists of interactions between âcarriersâ and âbarriers,â each of which has a geographical expression. The first idea, that of carriers, refers to the means by which something moves over the physical landscape and throughout a social system; different means of movement will lead to different spatial typologies of diffusion. For example, a process by which a message is shared by people with their immediate geographic neighbors leads to what Gould (1969) called expansion diffusion â a diffusion process that cascades through geographic space from its origin. Other variations on this theme are possible, such as hierarchical diffusion, where something begins at one location in a hierarchy and then moves to subsequent levels rather than simply moving to the nearest location. Without regard to the full set of diffusion forms identified by Gould (1969), the emphasis on carriers is important, since it draws attention to the relative locations of a set of actors involved in the diffusion process.
The second term, barriers, refers to the set of social, political, economic, cultural or physical geographic features in a given area (or within a given system) that may facilitate, speed up, slow down, or perhaps stop altogether the diffusion process. These may be things such as a language difference between two ethnic groups, a political boundary that limits movement or a transportation corridor that enables movement. Taken together, the set of barriers will profoundly shape the direction and speed of a diffusion process which draws attention to the relative structures that impact a set of actors involved in the process. While the term âbarrierâ suggests features that limit movement, it also captures the notion that some geographic arrangements facilitate movement. This is most clearly seen using an example of a transportation or social network that connects certain locations in particular ways. A network of connected locations would then serve to shape the spread of the process throughout a study area (e.g., Cerina et al. 2014).
Figure 1.1 The classic âS-curveâ of a diffusion process.
The combination of carriers and barriers may be used to estimate the traditional S-shaped diffusion curve. As seen in Figure 1.1, a curve that represents the cumulative number of âadoptersâ of an innovation is also analogous to how a diffusion process may expect to play out over a geographical area. As time progresses along the x axis of the curve, relatively few people or locations are affected by the process. However, at some inflection point the pace of the process increases and many more people or locations are impacted in a relatively short period of time which is followed by another inflection point as the number of unaffected potential people or places is exhausted. The geography of the âcarriersâ and of the âbarriersâ to the diffusion process will go a long way in determining the inflection points and the slope of the middle part of the curve. In this sense, the shape of the diffusion curve is contingent upon the setting and context of the case under investigation and no two cases should be expected to produce identical curves.
These ideas about the geography of diffusion have been used by scholars interested in how war spreads throughout a political system, such as a particular geographic region or the entire international system. Most and Starr (1980) initially drew on the idea of war as âcontagiousâ for their early work on war diffusion. Although they adapted and relabeled Gouldâs (1969) twin notions of âcarriersâ and âbarriersâ into what they called âopportunityâ and âwillingness,â the basic premise was the same: within a system of interconnected political actors, the political actors (carriers) spread war based on their willingness to participate in a conflict but had to navigate the social and geographic limitations (barriers) that shaped their opportunities to do so. In their work, barriers were operationalized as shared international political boundaries which provided an opportunity for war to occur within the geography of modern territorial states.
Although this approach to political boundaries as barriers has been critiqued by geographers (Flint 2012) as taking the existence of boundaries for granted and overlooking how boundary functions change over time, it has also remained a hallmark of subsequent investigations of war diffusion in the international system. For example, Salehyan and Gleditsch (2006) identified the movement of refugees (carriers) across international boundaries (barriers) as key to understanding the spread of civil war in Africa. The presence of boundaries in such studies is reflected in the salience of territorially sovereign states in their frameworks but boundary effects have yet to be fully incorporated into diffusion studies of ânew warsâ that involve diverse sets of actors beyond just the regular uniformed militaries of states, including gangs, private security contractors, paramilitaries, militias, etc. (Kaldor 1999, 2013), and particularly where conflicts extend beyond the boundaries of states (although see Salehyan (2009) for an effort toward such integration).
Where war diffusion effects have been studied apart from examples of interstate war, the territorial logics of the modern state have limited the interest in boundaries. For example, Schutte and Weidmann (2011) modeled the spread of civil war violence within the territory of a set of individual states but each state was treated as an independent unit and disconnected one from another. While the notion of barriers/opportunities was implemented in this study using a simple distance measure (a common technique to include âgeographyâ in war studies: see Buhaug and Lujala 2005; Flint 2012), this approach imagines international boundaries as an impermeable barrier to the diffusion process and the int...
