Draft-Version Comments are welcome!

Transcription

1 Conflict Beyond Borders Conceptualizing Transnational Armed Conflict Michael D. Fürstenberg, M.A. Technische Universität Braunschweig, Germany Paper prepared for presentation at the SGIR 7th Pan-European International Relations Conference, September 9-11, 2010, Stockholm Draft-Version Comments are welcome! Abstract Although considerable progress has been achieved in the study of civil war since the focus of conflict research shifted towards intrastate conflicts, it has become clear in the last few years that an exclusive look on the national level is not sufficient to understand the complex dynamics of modern civil wars. Studies have shown that it is vitally important to also consider both subnational factors within and transnational links beyond formal state borders. Important transnational or regional dimensions are evident in a lot of contemporary conflicts, ranging from mere cross-border support to full-scale regional conflict complexes e.g. in the Darfur-Triangle of Sudan, Chad and the CAR or in the border region of AfPak. But despite the growing importance of and research interest in transnational conflicts, surprisingly little effort has been devoted to the development of a conceptual framework which links a holistic view with the empirical findings of quantitative studies. This paper attempts to close this gap by providing an instrument with which conflicts can be analyzed focusing specifically on their transnational dimensions. Drawing from the existing, but mostly unrelated and not very systematic, terminology and empirical evidence, it seeks to give a theoretical definition of what constitutes a transnational conflict, differentiating between civil wars with transborder elements and complexes of intrinsically linked conflicts in neighboring states. It then develops a typology of armed conflicts based on both the importance and formation of transnational dimensions. It is shown that transnational factors can play either causal or dynamic roles, and range from being of marginal influence to forming a constitutive element of the conflict. Introduction Traditional security studies, seeing the state structure of the international system as given, are concerned with threats to the state from other states, while civil wars are by and large ignored as mere domestic affairs that may threaten the local population but have no real effect on the international level. While this has long been accepted as shortsighted with regard to terrorist organizations possibly settling in those countries, the immediate regional impact has received far less attention. Despite their defining moniker, many contemporary armed conflicts cannot be adequate- 1

2 ly described as being internal. Civil wars affect their neighbors through refugee flows (Gleditsch & Salehyan 2006), disrupt legitimate and facilitate illicit trade through shadow economic networks (Juma 2007), hamper the economic development of a region (Collier et al. 2003), make conflict in nearby states more likely through demonstration effects (Forsberg 2009) and by providing cheap weaponry (Killicoat 2006). Not only does unrest in a country inevitably affect other states indirectly, but often the actual fighting itself draws in external actors and territories. A lot of so called internal wars are in fact internationalized, with troops from foreign states participating in a non-interstate conflict, prompting the inclusion of a correspondingly named own category in the typology of the Uppsala Conflict Data Program (UCDP) Armed Conflict Dataset (ACD) (Harbom & Wallensteen 2005). 1 In many more cases however, despite falling short of such direct military interventions, conflicts are influenced by military as well as nonmilitary support from outside actors. In fact, outside interference is not an exceptional but a regular feature of internal conflicts (Salehyan 2010a: 497). For example, it is a quite common situation that rival states, instead of facing each other off directly, support rebel groups in the other country to do the actual fighting. Being known as proxy wars and referred to especially in the context of super-power rivalry, this scenario has not declined with the end of the Cold War. Instead, very often now regional rivals and neighboring states use this tactic to substitute for a potential risky direct engagement of their enemy. Mutual support of rebel organizations is a defining feature for instance of the dynamics of the Darfur conflict, involving at times three allegedly internal conflicts in Sudan, Chad and the Central African Republic (Giroux, Lanz & Sguaitamatti 2009). Such at the beginning sometimes low-level support can even escalate to full-blown interventions or interstate conflict, as exemplified by the massive engagement of Russian troops in the dispute between the Georgian government and the secessionist region of South-Ossetia in 2008 (Gleditsch, Salehyan & Schultz 2008). In addition to the involvement of external actors in con- 1 As the ACD codes just conflicts with the involvement of foreign regular troops as internationalized, this category actually underestimates the number of such cases as it is also possible that external non-state actors actively participate in the fighting. 2

3 flicts in other states, sometimes the conflict zone itself spreads beyond the borders of its state of origin. This happens especially when rebel organizations do not limit themselves to the territory of their home state but deliberately operate across state borders, often with the explicit or tacit permission of their hosts. Such transnational rebels (TNR) (Salehyan 2009: 6) use neighboring countries, where national security forces normally cannot follow them, as safe havens and staging areas, to recruit new soldiers and to resupply (often committing serious atrocities against the civilian population). In addition to being much more difficult to defeat or appease and therefore leading to more protracted conflicts (Salehyan, Gleditsch & Cunningham 2008), transnational nonstate actors can also lead to serious interstate disputes (Salehyan 2007). Sometimes TNR even become disconnected from their country of origin and roam borderlands without clear strategic objectives other than their survival. Examples for this phenomenon include the former Rwandan Hutu-extremist forces of the FDLR in the Eastern Democratic Republic of the Congo (DRC) (International Crisis Group 2009a) or the Lord s Resistance Army (LRA), which operated in Southern Sudan as well as in the DRC in addition to their native Northern Uganda (International Crisis Group 2009b). In their most extreme form, militant groups like Al-Qaeda have no real geographic base in a conventional sense altogether but rather consider the whole globe as their theatre of operations. All this begs the question, whether the classic dichotomy between inter- and intra-state conflicts is really able to capture the increasingly complex picture of modern warfare. In fact, a lot of older and contemporary wars are probably more accurately described as being transnational rather than internal conflicts (Rubin 2001: 2). In order to analyze and assess these conflicts adequately it is necessary to take their transborder dimensions seriously not only as mere epiphenomena of civil wars but as defining features of the conflict itself. This is important not only from an analytical point of view, but also for security-related activities of regional actors as well as conflict resolution efforts of the international community which are severely hampered by cross-border dimensions (Borchgrevink & Lie 2009:13-14). Transnational armed conflicts form a 3

4 distinctive type of non-interstate conflict in the sense that particular wars cannot be fully explained when this aspect is ignored (Cunningham 2010). A dramatic example are the ongoing wars in Afghanistan and Pakistan between the internationally backed governments and Taliban militias, which can only be treated as one intertwined conflict in the AfPak -region (see The notion of transnational or regional wars is nothing new, of course practitioners as well as scholars have used these or similar terms for quite a while, to describe both specific cases and more general patterns of regional conflict complexes (Wallensteen & Sollenberg 1998). Unfortunately, despite growing interest in this element of civil wars especially in recent times, there is still a shortage of both data focusing on this specific topic and terminological and conceptual clarity (Kurz 2007: 4). Before civil wars can be systematically analyzed and compared particularly in this regard, the question of how to define the term transnational conflict has to be addressed. In this paper I will try to contribute to the effort of developing a coherent conceptual framework to analyze transnational armed conflicts. Building on a review of existing attempts to conceptualize this type of conflict I will first develop a working definition of both transnational conflict and transnational conflict complex, based on the specific character of transnational dimensions. The goal is to establish a cut-off point that definitional distinguishes distinctively transnational conflicts from cases where some form of transnational link may exist but is not essential to the conflict. I will then further elaborate this basic definition by differentiating transnational conflicts regarding the intensity and importance of the transnational dimensions as well as the question of how they emerged. While all of them share the defining characteristic of transnational links crucial to the actual struggle, I argue that there is a wide variety ranging from civil wars with just sporadic cross-border elements to regional complexes of intrinsically linked conflicts. Moreover, transnational factors can play both causal and dynamic roles in civil wars meaning that conflicts can have a transnational character from the beginning or become gradually transnationalized during the course of the quarrels. 4

5 Conceptualizations of transnational conflict In recent times, there has been considerable effort to address the long-standing blind spot [ ] of existing research on civil war to treat civil wars as purely domestic phenomena (Cederman, Girardin & Gleditsch 2009: 404). Several concepts have been applied to capture the transnational dimensions of civil wars. In this section, I will give a short overview over the most important, ordered by their connectedness to the traditional notion of internal conflict. I will then offer a critique of them, before going on to develop my own definition. Despite its defining name, the insight that internal wars are not always neatly contained within a state s territory is actually quite old. As early as 1964, Karl Deutsch stated that [h]istory shows many instances [ ] where such revolutions, guerilla wars, or civil wars spread across old boundaries (Deutsch 1964: 100). However, due to the focus on the systemic and state levels and the view of civil wars as mere proxy wars between the super powers during the cold war, studies of this phenomenon were few and far between before 1990 (see e.g. Hill & Rothchild 1986). When the proxy war explanation proved to be insufficient given that civil strife did not decline but on the contrary seemed to be on the rise after the cold war ended, the focus of conflict studies shifted to the analysis of the domestic determinants of internal conflicts. While several studies here provided important insights (Collier & Hoeffler 2004 and Fearon & Laitin 2003 being among the most cited), the cross-border dimensions of civil conflict were mostly still neglected, especially in quantitative analysis of civil war. But in the wake of the increasingly complex ethnic wars of the 1990s, gradually an understanding developed that studies of intranational conflict were hampered by a focus on national-attribute analysis (Moore 1995: 162). Contagion and Neighborhood Effects The interdependence of domestic political developments has been studied foremost in the field of policy analysis, where a variety of terms such as demonstration, bandwagoning, conver- 5

6 gence, contagion and diffusion have been used (see Elkins & Simmons 2002 for an overview). This concept has been adapted for the study of international conflict by Most and Starr in their work on international conflicts as events of a given type in a given polity [that] are conditioned by the occurrence of similar events in other polities at prior points in time in this case, war being the event (Most & Starr 1981: 10). Following the call of J. David Singer that the direction should now be to focus on such variables as diffusion and contagion (Singer 1981: 1), both terms were used, albeit not very consistently, in a lot of further studies (e.g. Starr & Most 1983; Vasquez 1992; Siverson & Starr 1990). An important insight gained from this research is that interdependence is stronger for entities within closer range of one another, particularly for continuous states (O Loughlin 1984; Siverson & Starr 1991). While this is true for interdependence in general All international politics is local, as Kristian Gleditsch remarked (Gleditsch 2002), questions of war and peace are especially prone to geographic factors, as the opportunity to fight naturally decreases with distance (Vasquez 1993: 207). 2 The result is that conflict is not evenly distributed around the world but tends to cluster spatially in certain regions (Anselin & O Loughlin 1992; Gleditsch 2002). Both the concepts of contagion and conflict clusters have been applied to the study of civil war. Contagion effects were examined through case studies focusing on particular conflicts (see e.g. Brown 1996; more recent examples include Byman & Pollack 2007; Weinbaum 2006) or potential cross-border means for diffusion like direct spill-over effects (Sambanis 2004), refugees (Weiner 2006; Posen 2006), international diasporas (King & Melvin 2006) or demonstration effects transmitted by ethnic linkages (Lake & Rothchild 1998; Gurr 2000). More recently, this research was backed up by large-n analyses, confirming that both direct as well as indirect crossborder variables have a significant effect on the risk for conflict onset in a given country. Among the factors tested are refugee populations (Salehyan & Gleditsch 2006) as well as refugee flows (Forsberg 2009), transnational ethnic linkages (Buhaug & Gleditsch 2008; Forsberg 2009; 2 This is true at least for the majority of conflicts even in the current era of globalization see Buhaug & Gleditsch (2006). 6

7 Cederman, Girardin & Gleditsch 2009), extra-territorial safe havens of rebel groups (Salehyan 2007), regional trade (Gleditsch 2007) or regional state weakness (Lambach 2007). In comparison, transnational dimensions can be at least as important as the profile of individual states (Gleditsch 2007: 304), although their effect is greatly mitigated by state capacity (Braithwaite 2010). As they developed out of much the same research program, the literature on contagion effects is closely related to the notion of conflict clusters, especially in the case of quantitative research. Sambanis, in a study on ethnic wars, discovered that civil war in a neighboring country is a significant predictor for a conflict outbreak (Sambanis 2001). This finding, often referred to as the neighborhood effect of civil war, was replicated in a lot of following studies (e.g. Ward & Gleditsch 2002; Marshall & Gurr 2005; Salehyan & Gleditsch 2006; Gleditsch 2007) as one of the most robust results of civil war research at all (Hegre & Sambanis 2006). Moreover, the spatial clustering of intrastate conflicts cannot be dismissed as a mere product of a clustering in similar country characteristics associated with conflict (Buhaug & Gleditsch 2008: 230). In this so called bad neighborhoods (a term going back to Weiner [2006]), low levels of democracy, instability of state apparatuses and weak economic growth are mutually reinforcing, making states structurally prone to an domestic outbreak of civil strife as well as to an infection from a neighboring civil war. External Involvement Approaches that focus on the involvement of external actors in civil wars go one step further than contagion-concepts 3 in their treatment of transnational dimensions in the sense that the conflicts here are by definition not purely internal. They also are less abstract as they deal with active decisions of concrete actors and not with seemingly inevitable effects. The contagion logic 3 This is of course a stylized description many studies focusing on contagion or bad neighborhood effects incorporate direct external influence in their models. Nevertheless, the basic argument goes as described above (see e.g. Forsberg 2009). 7

8 was very early criticized by Michael Brown as being simplistic and mechanistic, relying on crude analogies to diseases, fires, floods, and other forms of nature [ ] spilling over from one place to another through a process that is always beyond human control (Brown 1996: 23-26). Moreover, Brown explicitly differentiates between states as passive victims of turmoil in the region and as active contributors to military escalation and regional instability (Brown 1996: 26). The focus here is therefore less on structural effects and more on intentional behavior of specific actors. Third-Party interventions in civil wars are of course an old topic in conflict studies (see e.g. Rosenau 1964). There is a variety of reasons why states should intervene in internal conflicts. During the Cold War, third-party involvement was usually seen as a function of super-power rivalry, supporting one side in the ongoing proxy wars of domestic groups. Since the Cold War ended, humanitarian interventions by the international community have received a great deal of attention with policy-makers and researchers (Doyle & Sambanis 2000; see Hehir 2010 for an overview), although the vast majority of interventions were actually carried out by regional powers, often direct neighbors of the conflict state (Harbom & Wallensteen 2005: 628). Regardless of the intervener however, foreign interventions are now generally being treated as a form of conflict management (Regan 2002: 59), which is either focused on a neutral approach to bring the parties to the negotiating table or biased towards one side, trying to help them to win the conflict. In the former tradition, Walter (2002) finds that external involvement can help in negotiating an end to the quarrels by providing a solution to credible commitment problems. Balch-Lindsay & Enterline (2000) argue that interventions supporting one side shorten the duration of civil war not only because they make a military victory more likely, but because they also raise the chance for a negotiated settlement. Regan (2002) confirms this especially for interventions supporting the governing side. On the other hand, this seems only the case for decisive actions, as balanced interventions as well as just expected ones prolong conflict (Balch-Lindsay & Enterline 2000; Elbadawi & Sambanis 2000). 8

9 What has been mostly neglected until recently however is the role of third-parties having other goals rather than to end the conflict (be it for humanitarian or opportunistic reasons). Cunningham argues that in many interventions outside actors pursue independent objectives in the war outside of the goals of the domestic combatants (Cunningham 2010: 116). Under this condition, the setting of the conflict becomes increasingly complex and even more difficult to solve, because the intervening state not necessarily supports the decisions of his domestic allies. 4 This is especially true when external actors don t intervene in an ongoing conflict but are already involved in its outbreak from the beginning. 5 While in the logic of the contagion-framework outbreaks of civil wars are influenced more by indirect and structural transnational dimensions, in this case they are the result of discrete, deliberate decisions by governments to trigger conflicts in nearby states for political, economic or ideological purposes of their own. [ ] Such conflicts, one could say, are caused by bad neighbors rather than bad neighborhoods (Brown 1996: 580). States that meddle in the affairs of their neighbors causing or influencing civil war usually act in an alliance with domestic armed groups to advance their goals, either in joint operations with their own troops or as proxy forces which they materially sustain (Harbom & Wallensteen 2005; Salehyan 2010a). Regional conflict complexes/regional conflict formations Approaches that explicitly focus on the regional character of conflicts can be traced back to the concept of regional security complexes, which Barry Buzan defined as a group of states whose primary security concerns link together sufficiently closely that their national securities cannot realistically be considered apart from another (Buzan 1991: 190). Following the constructivist turn in international relations (see Checkel 1998) this Regional Security Complex Theory has undergone major developments since its original inception (Buzan, Wæver & de Wilde 1998; 4 For example, when Laurent Kabila after he toppled the Zairian Dictator Mobutu didn t comply with the demands of his supporters Rwanda and Uganda, they switched sides and intervened against him. 5 External intervention has been identified in many case studies as a key factor in civil war onset (Sambanis 2004a: 270). 9

10 Buzan & Wæver 2003), but the basic idea still is that the security interests of neighboring states are tied together by regional, i.e. short-distance, cross-border linkages. This notion has first been adapted for the study of civil wars by Wallensteen & Sollenberg, who in their survey of conflict data found a number of situations where neighboring countries experience internal or interstate conflicts [ ] with significant links between the countries and labeled this regional conflict complexes (Wallensteen & Sollenberg 1998: ). Although they considered this an important dimension of armed conflicts, the term has not been systematically used in further presentations of the quantitative UCDP data, apart from it being mentioned as a possible reason for the increasing complexity of internal conflicts (Wallensteen & Sollenberg 2001: ). Instead, a very similar idea was developed by Barnett Rubin in his work on the war in the African Great Lakes region, which he termed a regional conflict formation (RCF), defined as a set of transnational conflicts that form mutually reinforcing linkages with each other across state borders (Rubin, Armstron & Ntegeye 2001: 3). This model combines the inside-out and outside-in dimensions of civil wars by emphasizing the mutuality of transborder linkages. Because they feed each other, conflicts in neighboring states become interlocked and much more difficult to solve. Regional conflict formations are dynamic entities and are comprised of a plurality of overlapping transnational networks. Rubin differentiates broadly between political, military, economic and social networks (Rubin, Armstron & Ntegeye 2001: 5; see Pugh & Cooper 2004: for an overview of the elements of this categories). The RCF framework was conceptualized from the beginning as both descriptive and strategic, aimed at not only at scientific analysis but also concrete policy recommendations (Armstrong & Rubin 2002: 1). It is no wonder therefore, that it was applied predominantly in case studies, apart from the DRC for example on Afghanistan (Rubin & Armstrong 2003), the Middle East (Leenders 2007), Darfur (Giroux, Lanz & Sguaitamatti 2009) or West Africa (Lambach 2007). Comparative studies found that the RCF framework revealed important similarities regarding transnational dimensions in different cases of civil conflict (Rubin 2002; Pugh & Cooper 2004). 10

11 Wars across states/neomedievalism The farthest away from classical definitions of conflict is the idea of wars across states (WAS) that was developed by James Hentz (2007). Hentz argues that current conflicts especially in Sub- Sahara Africa are a new kind of war [which] claim their own ontology (Hentz 2007: 2). As such, WAS are conceptually distinct from civil wars as they involve different (namely non-state) actors and areas, follow a different political logic, and are strongly associated with state collapse. As the name suggests, transnational dimensions are an important aspect of WAS. They, in fact have no discernable center of gravity. In a war across states large tracks of territory bleeding across state borders provide the staging (Hentz 2007: 2). Consequently, a state-centric framework cannot capture its dynamics. Moreove, Hentz argues that this is true not only in a geographical sense, but for the political aspect as well. Classical civil wars are contests over the state, fought usually to either get control of the state apparatus or for secession and thus creating a new state. Hence, while the fighting may result in state collapse, this is generally an unwanted side effect. WAS however, are contests over the relevancy of the state project itself, not about the challenging or building of central sovereignty but its negation (Hentz 2007: 11-12; 18). That does not mean that these wars are apolitical, as many theorists of the new war -paradigm suggest (Kaldor 1999; Snow 1997). The politics of those conflicts just doesn t center on the institution of the state anymore but follow a patrimonial logic on a regional scale (Hentz 2007: 3). Ethnic loyalties, peripheral and neglected border regions or warlordism are important elements of WAS. The notion of wars across states has until now not garnered much attention in the literature and is cited just by a handful of studies (e.g. Giroux, Lanz & Sguaitamatti 2009). It is however somewhat related to the concept of neomedievalism, which was introduced be Hedley Bull as a possible alternative to the Westphalian model of international relations (Bull 1995; see Cantir & Schrodt 2010 for an overview). Going back to Bull, neomedievalism sees the international system as a system of overlapping authority and multiple loyalties, which characterized medieval Christendom as well (Cantir & Schrodt 2010: 11). This rejection of the classical state-centric approach 11

12 makes the concept well suited for the study of emerging political orders under the conditions of state failure and civil war (Biro 2007: 13; Winn 2004). As the nation-state is no longer willing and/or able to project an exclusive sovereign order, [e]xit is becoming an increasingly viable option for a growing range of actors and groups, leading to endemic civil and cross-border wars (Cerny 1998: 36). In such situations, warlord mode[s] of substate authority can be established, where small-scale political order is structured around personal leadership and patrimonial rule, supported by often deliberately manipulated subnational identity structures (be they based on ethnicity, religion or tribes) and violence (Cantir & Schrodt 2010: 19-23; see also Reno 1999). It is important to note, that both concepts interpret state failure and the resulting transnational wars quite differently than more traditional approaches to civil war: While these view the breakdown of political authority in a state as resulting in anarchy that has to be remedied by a form of nation-building, WAS and neomedievalism concentrate on new forms of local and transboundary political orders. 6 Applicability to transnational conflict As this cursory review demonstrates, there is no shortage of studies that take the transboundary dimensions of civil wars into account, as the literature has increasingly turned away from the traditional view of purely internal conflicts in recent years. But as conflict boundaries are changing faster than conceptual boundaries (Rubin, Armstrong & Ntgeye 2001: 6), not a single framework has been agreed on. Instead, the theoretical concepts used mirror the varying backgrounds of researchers and shed light on different aspects of transnational conflicts (a term all of them use at least sometimes colloquially): For the purpose of this paper, the concepts of contagion and conflict clusters can be treated as a single framework regarding the description of transnational conflicts: According to 6 A similar approach is taken by a current research project at the Free University of Berlin, which studies the circumstances under which security is provided by non-state actors under conditions of armed conflict (see Risse & Lehmkuhl 2006). 12

13 this logic, internal armed conflict originates in one country and spreads through several channels across its bad neighborhood (see e.g. Forsberg 2009: 20). So while in this view civil wars are seen as both a source and a result of transnational linkages, the conception of conflict happening in a particular state, albeit being partly dependent on its neighborhood and possible contagion effects, has not changed very much from the traditional definition of internal armed conflict. The involvement of external actors as a concept for transnational aspects of civil wars is partly captured by the category of internationalized internal armed conflict in the ACD typology (Gleditsch et al. 2002). However, this includes just cases in which state governments are directly participating in the fighting missing the ones in which support falls short of the commitment of regular troops or in which external non-state actors are involved (Harbom & Wallensteen 2005). Although such less overt support is of course difficult to track because of its often clandestine nature, recent data gathering efforts confirm that this is an important dimension of internal conflicts (Salehyan, Gleditsch & Cunningham 2008). All in all, this second conception of transnational dimensions differs from the contagion/bad neighborhood-approach in the sense that it focuses on actor-specific rather than structural factors and follows an outside-in in contrast to an inside-out logic. Both have in common however, that the basic unit used is still the nationstate. Transnational factors are basically treated as additional features of internal conflict even if as admittedly important ones. Neither of them develops a framework that allows a categorization of conflicts as specifically transnational. The RCF-approach goes beyond that and centers in particular on transborder dynamics at the regional level. It thus takes a step in the direction of overcoming the traditional methodological nationalism of civil war research. Instead of focusing on fixed geographical entities, it directs its attention to the wide range of political, military, economic and social networks that straddle the boundaries of nation-states. What a region in the sense of a regional conflict formation is, respectively which area and states are included in it, is therefore defined empirically, [ ] may or may not coincide with areas traditionally or bureaucratically defined as regions, [and] can 13

14 include parts of a state, but not the whole state (Armstrong & Rubin 2002: 5). While this allows for a great deal of flexibility in analyzing specific cases, it has the drawback of being very susceptible to selection bias. To put it differently: When the region of a regional conflict formation is defined ex post facto, there is no basis for the construction of an ex ante unit of reference and therefore no basis for systematic, especially large-n, comparisons. Another problem with the idea of RCF is that a complex requires per definitionem at least two neighboring conflicts so cases where the fighting takes place just in one country but were nevertheless important cross-border linkages are present are not covered by it. What is especially puzzling is that the building blocks for an RCF comprise of, according to the definition by Rubin, a set of transnational conflicts [emphasis by the Author], although it is not explained what such a transnational conflict is in the first place. This is presumably a result of the ex post facto identification of RCFs conflicts are transnational, when they are part of an RCF. It is obvious that this raises more problems for a framework intended for systematic large-n comparisons. Finally, the concepts of wars across states and neomedievalism seem to be well suited as an analytical framework of transnational conflict, because they reject the state not only as the appropriate unit of reference, but as the relevant political context for conflicts at all. Wars are in this view not only transnational in a geographical sense, but also in regard to the issues they are fought over. Thus, they cannot be captured by the traditional categories of governmental (i.e. about state control) versus territorial (i.e. about secession or state building) armed conflicts. While this certainly opens up new perspectives and captures the reality of some cases really well, it is highly questionable whether it is an accurate description of all or at least most civil wars, even when they exhibit strong transnational dimensions. Although Hentz admits that and tries to give several criteria to distinguish WAS from international and civil wars as well as insurgencies (Hentz 2007: 7-18), the concept is still rather vague. 14

15 This section attempted to tackle the lack of terminological clarity and [ ] conceptual confusion about the nature of regional conflicts in the literature (Kurz 2007: 4) by ordering the most common approaches according to how far they depart from the traditional understanding of civil wars as purely internal contests. In doing so it became clear that there are basically two ways of dealing with the limits the classic typology entails: Sticking by and large with the established notion of domestic conflict but taking further influencing factors into account (contagion/bad neighborhood and external involvement), or developing ambitious holistic concepts to comprehend the phenomenon in its entirety (regional conflict formations and wars across states). Both have of course their merits; while the former adheres conceptually to methodological nationalism but in return is sufficiently precise to allow for rigorous empirical testing using tried data sources, the latter facilitates a new understanding of certain conflicts as specifically transnational in character but is difficult to operationalize, and therefore lends itself not easily to systematic study beyond single cases. I argue that there is a need to develop a middle ground, that acknowledges the special properties of transnational conflicts while at the same time is compatible with established methods to make large scale comparative analysis possible. Defining transnational conflict It has been shown in the discussion above that none of the existing conceptualizations provides a readily available and integrated framework. But in order to study a phenomenon systematically, it is essential to agree upon a definition that makes it clear what is meant by the term transnational conflict. In this section, I will try to construct such a definition, using the notion of regional conflict formations as a starting point. This seems appropriate because from the ideas presented above, it offers the best compromise between conceptual innovation and definitional clarity: One the one hand it acknowledges that transnational dimensions are more than an epiphenomenon of traditional civil war; on the other hand it gives at least a reasonably clear description of its content. Although the terms regional conflict formation and regional conflict complex essential- 15

16 ly mean the same thing, and are therefore often even used interchangeably (e.g. Pugh & Cooper 2004), their respective definitions vary in the details. While regional conflict complexes are described as situations where neighboring countries experience internal or interstate conflicts, and with significant links between the conflicts (Wallensteen & Sollenberg 1998: 623), the respective meaning of regional conflict formations is given as a set of transnational conflicts that form mutually reinforcing linkages with each other across state borders (Rubin, Armstron & Ntegeye 2001: 3). When we dissect these definitions, we can use the resulting components to step-by-step deduce a clear-cut definition of transnational armed conflict from them. (1) The sine qua non of transnational conflicts are obviously the stated linkages [ ] across state borders. As we have learned from the examined literature, there is in fact a wide variety of such transnational dimensions, ranging from indirect economic effects to direct military spillover. As a first step it should be useful to structure these phenomena; several categories have been proposed in this regard. 7 A basic distinction is whether the transboundary factor is the result of the intentional behavior of particular actors or an effect of structural conditions. Drawing from the literature, we can call the former actor-specific or elite-level and the latter nonactor-specific or mass-level factors (Gleditsch 2007: 295; Brown 1996: 575). I will use the terms of Gleditsch and Brown interchangeably here; although at first sight it appears that they describe not entirely the same thing mass-level factors clearly can be the result of deliberate actions by certain people, for example the decision to flee from a conflict zone to a neighboring country, I argue that it is more important to distinguish between mechanisms that can reasonably be traced back to the decisions or qualities of a specific actor and such that cannot, i.e. that are collective phenomena. In this logic, refugee flows are non-actor-specific phenomena, as it is (usually there may be cases where even mass-fleeing is being orchestrated by certain actors) not possible to specifically identify some (group of) person(s) who decided that there should be a 7 The following typology is necessarily an ideal-typical one. 16

17 refugee stream. In contrast, military actions can almost always be traced back to the deliberate decisions of political or military leaders. 8 Transnational dimensions can secondly be discerned according to their means of transmittance, i.e. whether they are material or immaterial (Checkel 2010). The distinction is quite straightforward: Material factors can (at least in principle) be directly observed and measured, like people or goods that are transferred across the border; immaterial factors on the other hand can include intangible goods like information or refer to shared patterns of identity. Material and immaterial dimensions can be both actor-specific and non-actor-specific: For example, if certain intelligence information is exchanged between an external government sponsor and some rebel organization, this plays out on the elite-level; if however knowledge of an uprising in a neighboring state spreads among the population and hence may have a demonstrating effect, this dispersal works on the mass-level and cannot be attributed to a specific actor. Accordingly, an arms transfer would be actor-specific if the shipment originates from a defined source which purposeful intended to deliver weapons to the recipient, in contrast to a purchase on the free market in which the parties are essentially exchangeable. Although the latter clearly involves actor-based decisions (of the buyer and the trader), it is structural in the sense that in principle they act as anonymous agents of the market. It is obvious that this can be difficult to distinguish in practice, but the theoretical argument holds none the less. Apart from the transfer of tangible or intangible goods, identity issues can also be either actor- or non-actor-specific, depending on whether e.g. ethnicity is shared between a rebel movement and a supporting government or between a domestic group and a cross-border diaspora or kin-group (Cederman, Girarding & Gleditsch 2009). Although the latter is admittedly likely to serve as a basis for the former, the potential mechanisms at work are different. Of course, the categorization is not always clear-cut: For example, 8 The somewhat related differentiation between structural and dynamic factors (Lambach 2007: 40) differs from mine at exactly this point: Refugee flows are clearly a dynamic phenomenon, but also have structural effects as a refugee population in their host country. 17

18 while refugees as such are clearly a material dimension, when they settle in a neighboring state they become a diaspora, possibly generating cross-border ethnic kinship. 9 Finally, transborder links can be divided along thematic lines, i.e. according to the type of interaction and function they serve. I here follow Rubin, Armstrong & Ntegeye (2001) and Pugh & Cooper (2004) in distinguishing between military, political, economic and social transnational networks. As Pugh & Cooper remark, the dynamics underpinning specific conflicts are not easily demarcated because they frequently overlap, and specific cases [ ] may reflect more than one element (2004: 26). For example, the shipment of weapons can be understood either as a military matter or as part of the economic sphere. Sometimes, refugee camps are used as recruiting grounds for so called refugee warriors (Zolberg, Suhrke & Aguayo 1989: 275), transforming a social network to a military element. Nevertheless, together with the categories developed above these dimensions should be encompassing enough to have a reasonable heuristic function. The military category pertains to all actions that are directly linked to the fighting, be they the provision of military support or the cross-border movement of troops, either in form of intervening regular armed forces or as transnational rebels. Political networks are open or covert relationships between conflict and external actors, which can come in the form of alliances or rivalries. Also, civil war has a destabilizing effect on regional political order. Economic networks refer to the cross-border movement of goods in general, but in the context of civil war are commonly associated with illicit trade in conflict goods (e.g. diamonds from Sierra Leone, Coltan from the DRC or opium from Afghanistan) and local war economies as well as negative effects on the regional economy. While generally interactions here will be purely profit-oriented, this category also includes non-military support from external actors. At last, social linkages comprise all forms of identity-based cross-border ties and solidarities (e.g. tribal, ethnic, national, religious or ideological). This includes transnational identity-groups in general as well as connections resulting from large-scale migration or flight. 9 In most cases however, it will be the other way round people tend to flee to countries that already harbor kindred groups. 18

19 As already mentioned these categories are not clear-cut and naturally affect each other. It is especially likely that non-actor-specific and/or immaterial dimensions provide the underlying structures for concrete transactions between actors. For example, military support will probably be organized along the lines of political alliances; trade in conflict goods will follow established regional shadow trade networks; refugees will turn toward areas where kindred groups live. Moreover, even on the same level certain dimensions may be functions of each other. Political relationships for instance often correlate with ethnic or religious linkages and external military support may be paid for with conflict goods. 10 Figure 1 presents an overview of the framework. actor-specific/ elite-level non-actor-specific/ mass-level material immaterial material immaterial military external military support; third-party intervention; provision of intelligence by external actors refugee warriors transnational rebels political alliances between domestic and external actors; rivalry between domestic and external actors regional destabilization; ideological support by foreign populace economic non-military support; trade in conflict goods by opposing parties cross-border shadow economies; reduction of regional growth social shared identity between domestic and external actors refugee flows; refugee populations; diasporas transnational group solidarity; demonstration effects Figure 1: Transnational dimensions of civil war 10 A prime example being the arms deliveries in return for diamonds of then Liberian President Charles Taylor for Sierra Leonean RUF-rebels, for which he now stands trial before the International Criminal Court in The Hague. 19

20 (2) One thing that becomes clear when we consider the above developed framework is that transnational dimensions are in fact an ubiquitous feature of armed conflicts. At the very least, armed conflicts have an adverse indirect effect on their immediate neighborhood even when their neighboring states are stable. However, in most cases civil war-states will be located in regions that are plagued by political and economic problems in the first place, thus being especially vulnerable to negative externalities of nearby violence. Furthermore, ineffective border control is a widespread problem in most parts of the global South, not only because of lacking state capacity but also due to the natural environments of mountainous or forestal areas. So the question regarding a definition for conflicts is how we can decide whether a case is specifically transnational in character, as opposed to just having transnational dimensions. Or, to put it in the terms of the definition of regional conflict complexes: What linkages are significant enough to merit the categorization of a conflict as distinctly transnational? To determine a threshold for transnational conflicts we fist have to discuss in what terms we are viewing the phenomenon of armed internal conflict in general. Until recently, the bulk of the literature has treated civil war as basically a general state-level condition (Cederman, Buhaug & Rød 2009: 500) that ought to be explained by attributes of the nation-state as a whole. This methodological nationalist approach ignores not only cross-border factors, but also all variation within states, actors, and regions experiencing conflict (Cederman & Gleditsch 2009: 487). Studies suffer from the problem of overaggregation at the national level in two ways: One the one hand, conflicts very seldom engulf the entire territory of a country but rather take place in considerably smaller areas, often in the periphery (Buhaug & Gates 2002). Using national-level averages to substitute for the actual social or geographical conditions in the areas where the fighting really takes place has shown to be a poor and often misleading approximation (Buhaug & Rød 2006). Furthermore, sometimes countries experience more than one armed conflict simultaneously, which may be fought in different regions and against different rebel actors. In addition to the geographic aggregation bias of many studies, lumping these different quarrels together as 20

21 civil war in state X distorts our understanding of specific conflict dynamics significantly. Especially, there is a considerably mismatch between the dyadic nature of theories of conflict and the actual research designs used in testing hypotheses (Cunningham, Gleditsch & Salehyan 2009: 571). Armed conflict is a social process between certain parties not a condition that somehow affects a country in general. Using national-level proxies for dyadic relations can tell us something about their structural determinants, but not much about the actors involved or their interactions with each other. Hence the confusion about the interpretation of the correlation between low national GDP and civil war: While Collier & Hoeffler (2004) see this as evidence that low opportunity costs make fighting attractive, Fearon & Laitin (2003) interpret the same finding in terms of state capacity, arguing that low GDP-figures proxy for a higher feasibility of civil war due to a weak state. In the end, the failure to specify who fights in civil wars ultimately makes it difficult to come up with good answers as to why we see civil wars (Cunningham, Gleditsch & Salehyan 2009: 571). The same is true for the question regarding transnational conflict: If we want to understand the effect of transnational factors on the dynamics of the conflict itself, we have to focus our attention on the actual warring parties and their potential crossborder relationships. This approach is also especially well-suited for our undertaking because it overcomes the fundamental problems that territorial defined units of analysis (i.e. nation-states or regions) have with transnational occurrences which are by definition unit-crossing. Thus, I here follow Cederman, Girardin & Gleditsch in their understanding of civil war as a disaggregated, relational phenomenon that includes links across state borders (2009: 404). Consequently, I reject all transnational dimensions that are not actor-specific. That does of course not in any way mean that non-actor-specific or mass-level factors were unimportant for the study of civil wars. But while these may be crucial in structuring the environment of (potential) civil conflict, only links of the actors involved have a direct impact on the war interaction itself. In other words: When actor-specific transborder linkages are ignored, the conflict dynamic can neither be fully understood nor resolved (see Cunningham 2010 for a similar argument refer- 21

22 ring to external interventions specifically). For much the same logic, a further restriction has to be made regarding actor-specific but immaterial transborder linkages. In order to affect the actual dynamic of an armed struggle, an incidence has to have a certain active quality i.e. it has to be based on decisions relating to the conflict, because conflicts are essentially strategic interactions (Cunningham, Gleditsch & Salehyan 2009: 571). This excludes identity-related cross-border links from our definition, because identity is in principle something that will not be actively chosen. 11 Political dimensions are much more difficult to handle alliances or rivalries between specific actors are surely based on strategic calculations. Nevertheless, political relations that are reliable enough to hold even in the context of civil war should manifest itself in some form of direct support for an ally or the enemy of a rival. In other words, immaterial social and/or political dimensions have to be mediated through military or economic material means to have a real impact on the fighting therefore, only with the latter becomes a civil war a truly transnational conflict. This distinction is obviously not without difficulties even when not backed up by economic or military power, external political pressure can exert considerable influence especially when used by a potent country or alliance of countries. However, such pressure is usually applied on the conflict as a whole and is not a feature of the armed interaction itself. Thus, because the purpose of this article is to categorize civil conflicts according to their inherent character, I chose to exclude political linkages from the definition of transnational conflict, too. 12 This leaves just one immaterial dimension, namely the provision of military information which can easily be subsumed in a general category of external military support. (3) By connecting the first two basic elements of the definitions of regional conflict complexes and regional conflict formations respectively, we could deduce what is in principle meant by the term transnational conflict in the RCF-framework, namely civil wars with significant 11 This is a stylized description of course I am not making a primordialist argument for fixed and unchangeable identities. However, while it can certainly be manipulated for example in order to gain external support, identity is not something that can easily be switched on a short-term basis not least because it is to a large extent ascribed. 12 There is admittedly also a pragmatic component to this decision. 22

23 links across state borders. In the previous two steps, we were able to work out what kinds of transnational linkages exit in general and which of them are significant for the dynamic of the conflict. According to these considerations, relevant cross-border dimensions have to be actorspecific and material, with the exception of military linkages. From this, we can now give the aimed-at operational definition (based on the dyadic version of the UCDP/PRIO ACD, see Harbom, Melander & Wallensteen 2008): A transnational conflict (TC) is a conflict dyad 13 including at least one non-state-actor, in which at least one of the actors exhibits at least one actor-specific material and/or military cross-border linkage. Constitutive dimensions in this sense include: transnational rebel groups; direct participation of external actors in the form of combat-troops; military support by external actors (providing bases, intelligence, training, logistics, weapons etc.); economic support by external actors (money, non-military goods); trade in conflict goods between a party to the conflict and external actors that directly feeds into the war effort. It is important to note that these do not all operate on the same level: While the last four categories can be conceptualized in terms of streams between actors, the existence of transnational rebels refers to the property of a single actor. Consequently, we can further distinguish TCs that are defined solely by cross-border streams (Type 1) from such where rebels 14 operate transnationally (Type 2) and conflicts in which both occurs (Type 3). 13 The UCDP definition of a conflict dyad is as follows: [T]wo conflicting primary parties of which at least one is the government of a state. In interstate conflicts, both primary parties are state governments. In conflict dyads in intrastate and extrasystemic conflicts the non-governmental primary party is an organized opposition organisation. (Harbom 2009: 1-2). 14 One could argue that this would apply not only to transnational rebel organizations but to foreign intervening troops, too. The difference is that these actors are not part of the conflict dyad, so extraterritorial soldiers here are just a form of support, linking the sender and the recipient. Foreign troops operating without directly supporting one of the parties, would form a new conflict dyad, either indicating interstate (when fighting the government) or extrastate conflict (when fighting the rebels). 23

24 While this designations should be sufficiently precise to allow for an adequate categorization of internal conflicts as either transnational or not, like with all theoretical definitions there will of course be border cases that may defy a clear-cut allocation. Furthermore, because of the often clandestine nature of cross-border support and shadow economic activity, there will be difficulties for data acquisition. In general, although this may result in an underestimation bias, when categorizing conflicts one should proceed in a conservative way. On the other hand, in many cases especially high-profile ones there will probably be more than the minimum requirement of one transnational link, which may alleviate the problem. For example, while it may be difficult to sort out the exact nature of the coltan trade in the Eastern DRC during the war 15, the involvement of Rwandan and Ugandan soldiers as well as the direct support of the two countries to respective client rebels makes this effort unnecessary for the question whether this constitutes a case of transnational conflict. (4) Finally, now that it has been established what we understand by the term transnational conflict, we can come full circle and operationalize the somewhat vague description of regional conflict formations/complexes that formed the starting point of this section. According to that, RCFs are a set of transnational conflicts with links between the conflicts. As we have already sorted out what a transnational conflict is and which specific links are constitutive for it, it is seemingly straightforward to derive an operational definition from that. To keep within the terminology used so far, I will call this phenomenon transnational conflict complex 16 : A transnational conflict complex (TCC) is a set of at least two transnational conflicts in neighboring states which are connected with each other through at least one actor-specific material and/or military cross-border linkage. 15 In fact, this has been documented rather well, see e.g. Pugh & Cooper (2004: 26-30). 16 There is no consensus in the literature on whether transnational or regional is the more appropriate term. I think that transnational gives a more precise idea about what is basically meant. The preference for complex is just personal taste. 24

25 It is important to keep in mind that in order to qualify as a TCC, specifically transnational conflict dyads have to be connected, rather than conflict actors in general. To clarify this a bit, consider the following examples where government A fights rebel group B in state X and government C fights rebel group D in neighboring Y: B gets support by C, making AB a transnational conflict. A retaliates by invading Y. As long as this is not in any way coordinated with D, this would not qualify the situation as a transnational conflict complex, because while actors A and C are now connected (forming a new interstate dispute), this would not directly affect conflict dyad CD. In reality, however, it is much more likely that A would start supporting D before attempting any direct military actions. B is a transnational rebel group, operating partly in Y (again, making AB a transnational conflict). If C tacitly allows that it counts as a form of support to B; if not, C would have to stop B, effectively supporting A (both still are part of the TC AB). Which of this plays out is largely a function of the relationship between the involved states: For example, it is well known that Pakistan harbors insurgents from the Indian part of the disputed Kashmir region. On the other hand, constructive political engagement persuaded India s Eastern neighbors Bhutan and Bangladesh to effectively prevent Indian rebels from using their territory. In the case that B manages to establish itself in Y, a new conflict dyad CB in Y would emerge and thus a transnational conflict complex comprising AB and CB (but not necessarily CD). The recent (unsuccessful) offensive of the Congolese army against the LRA in its Eastern border region is a case in point. A and C are rivals. A therefore supports D while C in turn provides assistance to B. Because of their respective outside linkages, both AB and CD are transnational conflicts which are connected via the reciprocal support-structure, making this scenario a transnational conflict complex. This scenario is quite common in regions of strained 25

26 security relations the reciprocal support of insurgents linked for example Sudan with Eritrea, Ethiopia and Uganda, who all harbored the SPLA in turn for Sudanese aid to rebels in their countries. To sum up, using the relatively strict definitions developed so far, civil wars can be sorted broadly into four categories, as depicted in figure 2: Internal conflicts in the original sense; internal conflicts with indirect or mass-level cross-border dimensions; transnational conflicts; transnational conflicts that are part of a transnational conflict complex. Figure 2: The spectrum of civil wars Although the aim of this paper is a theoretical conceptualization and I am not able to present any conclusive empirical findings based on that concept, we can draw on recent data gathering efforts on (among others) transnational connections of conflict dyads by Salehyan, Gleditsch & Cunningham (2008) to at least get a rough idea about the extent of the phenomenon. 17 As already mentioned, the coding procedure of external support used here is a bit different from my own, but the difference should not be to stark. To get a measure of the share of civil wars that are transnational I constructed a new variable indicating whether one of the dyad actors received support categorized as military or troops by the coders (rtypesup, rebextpart, gtypesup, govextpart) and/or the rebel party had a presence in neighboring territory (rebpresosts). When both occurred 17 The dataset as well as documentation can be found under 26

27 the conflict dyad was designated as a transnational conflict of type 3, while in the case of either only support or extraterritorial presence the cases were coded as transnational conflicts of type 1 or type 2, respectively. The following chart shows the share of the various types of conflict: Figure 3: Frequency of conflict types The dataset in total covers 471 dyads from ; after removing unclear cases, 398 conflicts are included in the analysis. Just 104 (26.1 percent) of that display no actor-specific material cross-border links. That means that a vast three quarter majority of civil wars are in fact transnational conflicts. Most of them are of type 3: 171 (43 percent) show both cross-border support and transnational rebels. 93 cases (23.4 percent) are type 1 TC and 30 (7.5 percent) type 2. These preliminary results, while impressive, must of course not be overrated, as they are based on the bare minimum criteria and say nothing about the strength of the various transnational dimensions. 18 I will elaborate on that in the following section. 18 Note also that this includes support given by the super-powers during the Cold War and therefore follows a different logic than implied by the regional focus followed throughout this paper. 27