Networks of third-party interveners and civil war duration

Introduction

A growing body of research has examined the role that the international community plays in civil war duration and settlement (e.g. Balch-Lindsay and Enterline, 2000; Collier et al., 2004; Elbadawi and Sambanis, 2000; Hartzell et al., 2001; Regan, 2002; Walter, 1997). Theoretical and empirical research mostly focuses on contextual conditions and interventions organized under the auspices of international and regional organizations. This literature, moreover, is less explicit on how unilateral interventions undertaken by third-party states affect wars. More often in civil wars, however, states become involved unilaterally in the conflict and act without the consent of multilateral organizations. From this perspective interventions appear rooted in realist orientations to world politics where states enter ongoing wars to pursue their security interests. Empirical models implicitly approach unilateral interventions by states as discrete events that play independent roles in the course of a conflict. What is missing in these accounts is that unilateral interventions are rarely isolated from each other. Instead, they exert a cumulative impact on civil war processes given that they are undertaken in a multiparty environment where interveners and combatants are involved in a complex web of relations (Cunningham, 2006; Findley and Teo, 2006; Gent, 2007; Scott, 1996). At the extreme, interventions are coordinated by those that undertake them, but, at the very least, civil war combatants see the conflict as a process and cumulatively integrate interventions into their decision-making whether or not it is intended by interveners. Therefore, intervention outcomes cannot be considered independent of how interveners relate to each other in the civil war setting.

Civil wars into which more than one outside country have intervened present a strategic setting where the political relationships among intervener states affect intervention outcomes. At its core, possible interactions between intervening states can be conflictual or cooperative. States can form rival networks in a civil war and compete for diverse outcomes by engaging in offsetting interventions. Geopolitical struggles between intervener states play a crucial role in transforming combatants’ preferences between war and settlement. The context of Congo’s civil war (1992–7) ¹ illustrates that interveners’ relation with each other had a consequential role in the evolution of the conflict: fluid alliances among Congo’s neighbors, Rwanda, Uganda, Angola, and Zimbabwe, resulted in recurrent insurgencies and ultimately motivated the government to engage in systematic killings of civilians in 1994 (Afoaku, 2002). Similarly, the competition between South Africa and Cuba in the Angolan civil war (1975–91) fed by superpower confrontation and regional feuds helped generate a protracted and costly war that lingered for decades (Rothchild and Hartzell, 1992). States with similar preferences can also act with the objective to establish peace, as in El Salvador in 1989 where Colombia, Mexico, Spain, Venezuela, and the US formed an ad hoc coalition to bring the civil war to an end (Prantl, 2005).

In this article we model the interactive role that unilateral interventions play in civil war duration. ² We focus our framework on groups of interveners and their cumulative impact. In doing so, we apply network analysis to the empirical domain of third-party intervention in civil war. Drawing on the realist scholarship, we argue that third parties’ choice of sides and the distribution of their preferences lead to certain alignment patterns. States supporting opposite sides in a civil war — what we call balancers — aim to offset each other’s influence and, in doing so, they spoil the chances for peaceful settlement. There are also seemingly consensual networks of states in the civil war environment, such as states that support the same side of a civil war, which we call bandwagons. The question of how bandwagoning versus balancing acts of interveners affect the duration of conflict is an important one. We argue that bandwagoning interventions have an easier time manipulating combatant behavior since there is a certain degree of agreement in their actions, and we link this favorable impact to close political relations among interveners. However, simply having a joint preference for a certain combatant does not enhance interveners’ cumulative effect on conflict duration. In fact, the conversion of preferences over the outcome of a war may lead to free-riding on the part of potential allies or inability to strategically time their actions to alter combatant behavior (Gent, 2007). Part of our theoretical puzzle is that even states with divergent preferences can also support the same combatant and create an effect similar to balancing.

We therefore empirically differentiate cases of bandwagoning with similar and divergent preferences. In the following sections, we outline our theory and its empirical implications. We expect to see that intervener states with similar interests have the ability to curb their disagreements and pursue a collective solution to civil wars. By sending reinforcing signals and unifying their offers, these interveners have the potential to convince or coerce the combatants into a settlement that they favor. On the contrary, states that oppose each other in the strategic environment of civil wars encourage opportunistic and rent-seeking behavior among combatants which can develop an incentive to play interveners off against each other. In the empirical section, we present a network analysis of civil war interventions in the post-World War II period which provides support to this argument. Our findings show that balancing behavior prolongs fighting whereas bandwagoning generates an earlier termination, but only when intervening parties share similar preferences. In the conclusion, we discuss the policy implications of our findings and emphasize the positive effects of coordinated interventions by like-minded states on bringing civil wars to an end.

Civil wars and their duration

The systematic examination of civil war duration has been relatively new. How long civil wars are expected to last has been linked to grievances associated with pre-conflict repression and poverty (Collier and Hoeffler, 2004; Gurr, 2000; Regan and Norton, 2005), natural resources and the possibility of contraband financing (Fearon, 2004; Ross, 2004), ethnic segmentation (Cederman and Girardin, 2007; Fearon and Laitin, 2003), and outside involvement in civil wars by international/regional organizations (Doyle and Sambanis, 2000) and states (Balch-Lindsay and Enterline, 2000; Regan, 2002; Regan and Aydin, 2006). The onset of civil war reflects a failure to reach agreement over the distribution of domestic resources. The factors that contribute to war duration further complicate this bargaining process and diminish peaceful alternatives by affecting the capabilities and resolve of the warring parties. The ability of the opposition to press their demands by force of arms and the government’s ability to resist throughout the war become an integral part of war duration in this setting (Buhaug et al., 2008; Cunningham et al., 2009). Rebel organizations that can effectively recruit from the local population by offering economic incentives, as in contraband financing, or by playing the ethnic card, as in ethnically divided societies, can challenge the government for a longer period. Other strategies may enhance rebels’ military standing vis-a-vis the government. For instance, rebel groups that have sanctuaries in neighboring states and operate across porous borders can prolong an insurgency with hit-and-run tactics (Fearon, 2004; Salehyan, 2007). On the other hand, governments that can extract resources from the populace as in high GDP societies expect to prevail against the rebels on the battlefield. Therefore, they might be less likely to give in to rebel demands and prefer a decisive victory to negotiated outcomes which would also contribute to war duration. Similarly, democratic governments may find scorched-earth policies to uproot the insurgency incompatible with their political culture and their reluctance to resort to violence allows rebels to fight for longer periods.

To the extent that reaching an agreement between warring parties is made more difficult by these conditions, we would expect civil wars to be of a longer duration. Important for our theoretical argument is that relative capabilities and the information that civil war parties hold about each other can be manipulated by external interventions (Regan and Aydin, 2006). While most interventions are ineffective in bringing civil war to an end, less is known about the conditions under which they can spoil the peace process. How multiple interveners relate to each other in the civil war setting and whether external states engage in competing or cooperating interventions will have a dramatic impact on civil war processes. We next describe the mechanisms by which this plays out.

Civil wars in complex environments

Evidence from the latter half of the 20th century points to civil wars as multiparty environments: in almost half of the wars in this period, more than one third-party state intervened in the same conflict, pointing to the interconnected dimensions of external interventions. We assume that within the context of a civil war, states considering an intervention are mostly aware of the presence or interest of other actors. US intervention in Nicaragua in the 1980s was openly debated before Congress, providing other potential interveners information that influenced their options. The Boland Amendment was an attempt by Congress to curtail all forms of military aid to the Contra, and the Reagan Administration’s ‘discovery’ of Soviet bombers being unloaded at a Nicaraguan port were visible components of the debate over whether or how the US should intervene (Scott, 1996). In a similar fashion, Zimbabwe intervened in Congo in 1998 after the Ugandan President Museveni declared support for the insurgents, effectively countering Zimbabwe’s efforts to shape the outcome. While not all intervention decisions are so public, such strategic information might generally be available to other states considering intervention, permitting the opportunity to bandwagon with or balance against contemporary interveners. Sometimes this information may be conveyed so as to coordinate actions and form ad hoc coalitions among interveners into civil wars. The Burundi, Rwanda, and Uganda coalition in the Congo War is illustrative of a ‘non-institutionalized, informal alliance … [based on] neighboring states’ who felt threatened by Mobutu’s policies (Rupiya, 2002).

Polarization of interests between the East and West in the Cold War period provides an example of the role that third parties jointly play in civil wars (Van de Walle and Bratton, 1997). In Angola, superpowers competed for influence over the combatants through their client states of South Africa and Cuba. In Mozambique (1979–93), the US provided assistance to the government with the hope that they would dismantle their socialist program and support the West whereas the Soviet assistance was aimed at saving Mozambique’s failing socialist experiment (Scott, 1996). However, sabotaging peace by third-party states is not exclusively a Cold War phenomenon. The politics of nation-state formation in the late 19th century also illustrates the multiparty environment of civil wars. The Ottoman Empire dissolved as a result of the foreign interventions in its rebellious frontiers in the Balkans and the Middle East. Britain, France, and Russia intervened in the Greek and Bulgarian rebellions against the Ottomans to support rebel claims to separation (Clogg, 2002). Unlike the Balkan cases, Armenian rebels on the Eastern front could not garner enough material support from Western powers and the disagreement between third parties contributed to an unsuccessful Armenian movement leading to massive human suffering in 1914.

As several cases of intervention from different historical epochs suggest, states that intervene in the same civil war share an interest in the war outcome despite their unilateral actions. This interest might favor similar as well as opposing outcomes and lead to alignments which are cross-cutting sources of influence on combatants. Few studies, however, have modeled the interdependent aspect of interventions. Regan (2002: 71), for example, argued that the strategic nature of opposing interventions has a debilitating impact on civil war duration: ‘any conflict that attracts opposing interventions is considerably more likely to remain ongoing than a conflict that does not have interventions supporting both sides’.

More recent projects also model the multiparty environments of civil wars focusing on the implications of multiple rebel groups and external states on the peace process (see Cunningham et al., 2009; Walter, 2006). Among these studies, Cunningham (2006) shows that multiparty civil wars, including those with more than one external participant, experience longer durations. Gent (2007) suggests that major powers usually undertake opposing interventions in civil wars so that they affect the course of events in their favor and abstain from interventions when their preferences are most similar. Similarly, Findley and Teo (2006) show that states balance the role of their rivals in civil wars by supporting opposing sides, whereas shared security interests between interveners in the form of alliances do not necessarily lead to bandwagoning on the same side. We build on this literature and explicitly model the role interveners’ interrelations play in civil war duration.

When international or regional institutions do not coordinate interveners’ actions, the joint impact on the outcomes is harder to gauge. Peacekeeping operations are inherently multiparty efforts that aim to overcome the strategic problems that intervener states might face. The collection of interveners is organized under a common command structure. Finnemore (2003: 74) argues that ‘multilateralism … increases the transparency of each state’s actions to others and so reassures states that opportunities for adventurism and expansion will not be used’. Similarly, Abbott and Snidal (1998: 19, italics added) claim that international organizations help states overcome certain strategic problems relevant to civil war intervention: ‘UN peacekeeping allows powerful states to support conflict reduction without being drawn into regional conflicts and discourages other powers from taking advantage of their inactions.’ Multilateral efforts therefore aim to impose a common set of preferences over interveners, preventing inaction or opportunistic behavior.

Unilateral interventions reflect states’ national interests especially when they are pursued in the context of other outside parties’ involvement in the conflict. Anticipating that multiple interventions will affect war outcomes interactively, interveners may factor others’ moves into their decision-making to coordinate their actions or destroy the influence of other external actors. Multiple interventions in a civil war are therefore cross-cutting pressures on civil war combatants. The most visible cases in Iraq and Afghanistan provide a window on this logic, as do Sudan, the Democratic Republic of the Congo, and Angola to name a few. In the following sections, we explore the cumulative effect of interventions through the networks that form among intervener states based on their choice of sides. The key element of these networks is that, in order to affect the duration of the war, they have to influence the core factors that impede the ability to reach a mutually agreeable bargain, or, alternatively, be sufficient to facilitate a quick victory for one side of the conflict by affecting capabilities in favor of this combatant.

Choosing sides or choosing stakes in civil wars?

Patterns in unilateral interventions are a function of states’ preferences over civil war combatants and outcomes. Among the variants of contemporary realist scholarship, neorealism predicts that states respond to power and choose to ally with the weaker side in international conflicts to balance the power of the rising state. Furthermore, research has shown that profit-seeking, interests, perceived threats, and risks may play a greater role than power in states’ alignment decisions. The complexity of these external stimuli may lead to diverse alignment patterns contradicting the predictions of balance-of-power theory. ³ Similar alliance patterns can also be observed in civil wars where third-party states align on the same or opposite sides with other interveners. We investigate two fundamental forms that such alignments can take in affecting combatants’ preferences, each with important implications on civil war processes: balancing and bandwagoning. An application of the realist concepts of balancing and bandwagoning to the civil war context requires certain simplifications. Given the inherent asymmetry in the distribution of power between the government and rebels, power as in the neorealist formulation would be a misleading starting point. ⁴ Besides, even small amounts of aid dramatically redistribute rebels’ chances of inflicting costs on the government without necessarily affecting the power distribution at the onset of the conflict. ⁵

To facilitate the introduction of these concepts to civil wars, we start by defining states that intervene to support the same side of the war as bandwagons and states that support opposite sides as balancers (Findley and Teo, 2006). We see these two strategies for intervening as attempts to manipulate the capabilities that warring parties hold and the asymmetry of information related to capabilities that presents one of the pathologies of bargaining. Facilitating or destroying the influence of other interveners, third-party states can affect the resolve of the parties to press further concessions (or settle with fewer) whereas their cumulative effect is closely related to how they relate to each other in the civil war setting.

An initial investigation of third-party alignments suggests previously unrecognized patterns in civil war intervention. Table 1 shows the number of balancing and bandwagoning states. ⁶ An important pattern in alignment choices is that a greater number of interveners choose to enter a civil war to oppose a previous intervener rather than jumping on the bandwagon. Hence, when defined as same-side versus opposite-side intervention among third parties, balancing behavior appears to be slightly more common than bandwagoning in the civil war context. This trend is reversed in the post-Cold War period where interveners became less likely to enter wars to balance against each other. Another major tendency in the data is that 66.7% (332) of interveners among bandwagons chose to support the government side whereas only 33.3% (166) entered the conflict to assist the insurgency. Such overwhelming support to governments can be explained with third parties’ status quo bias in civil wars: external actors aim to preserve the distribution of benefits and power within a country rather than assisting armed opposition in revising it. As a result, most external actors to a civil war do not attempt to use intervention to unseat incumbents. Instead, they seem to have an incentive to strengthen incumbents’ hand against challenger groups that resort to arms, suggesting that interveners’ behavior may indeed be driven by the concern to perpetuate rulers’ monopoly of power rather than defying their sovereignty. In effect, bandwagoning interveners seem to prefer a bargain closer to the ex ante status quo position, and may combine their efforts to manipulate capabilities and information to preserve this position.

OPEN IN VIEWER

Table 1.

Bandwagoning and balancing in civil war intervention

Choice of sides	1945-99	1945-89	1989-99
Balancing	521	457	64
Bandwagoning	498	411	87
Total	1019	868	141

Note: Frequencies report the number of interventions by third-party states.

How do these alignments play into combatants’ calculations? In identifying interventions’ cumulative effect on civil war processes, one can infer the compatibility between states’ goals from their choice of sides. Multiple interveners may attempt to offset each other’s influence in the civil war by supporting opposite sides of the conflict and contribute to the willingness of the warring parties to fight for longer periods. Balancing behavior is in part driven by the relationship among interveners, their divergent interests and preferences over outcomes (Akcinaroglu and Radziszewski, 2005; Findley and Teo, 2006). While balancers are trying to undermine others’ efforts by targeting opposite sides, they also send signals to the combatants that they have the support of an external patron to pursue their individual goals. In this respect, balancing interventions result in little incentive to negotiate, make concessions, or capitulate on the combatant side and, instead, lead to a stalemate situation. By way of example, Belgium militarily intervened in Congo’s civil war (1960–5) to support the separatist movement in the Katanga region while the US and Britain were favoring an intervention on the side of the central government to bolster its position vis-a-vis internal factions. ⁷ In 1961, the UN intervened against the Belgians on the government side whereas the US and Britain refused to support the separatist movement and pressured Belgium to change its course of action. In 1962, Belgium had changed its policy in favor of a unified Congo and withdrew its support from the rebels (Regan, 2000). Once Belgium decided to switch sides and abandon the insurgents, which came after a stalemate between the Katanga rebels and the Congolese central government, the balance was tipped in favor of the government and the conflict finally terminated.

Opening up a new bargaining dimension in the war, balancing behavior can also lead to increased resolve on the combatant side: competing third parties give combatants an incentive to exploit their disagreement and ask for greater rents to comply with their demands. Interveners’ incompatible preferences have the potential to turn the conflict into a source of revenue (Collier and Hoeffler, 2004). Combatants might get more attention externally while fighting than they have received before the war, encouraging them to sustain the war environment as a bargaining chip against external actors. One result of balancing behavior is a conflict that lasts longer because a mutual agreement is elusive, as previously suggested by scholars (Balch-Lyndsay and Enterline, 2000; Regan, 2002). The consistent South African support to the RENAMO rebels in Mozambique’s long and deadly civil war is a good illustration of how easily third parties can spoil the prospects for peace. South African military strategists laid down a plan of attack for the rebels aimed at weakening the government with relatively cheap strategies and securing an easy victory for the rebels (Young, 1990). ⁸ Hence, ‘spoilers of peace’ are not limited to insurgents and governments; balancing interveners can also complicate wars with their disagreements (Kydd and Walter, 2006; Stedman, 1997).

While the notions of balancers and opposing interventions are not new to civil war research, there is an inherent difficulty with defining a bandwagoning intervention in the context of unilateralism. In its most basic application and in line with the opposing interventions concept, it can be defined as interventions that back the same side of the conflict. Balancers attempt to preserve the status quo in civil wars whereas multiple actors jointly supporting one side may be driven by incentives to precipitate the war’s end by tipping the balance in favor of a particular combatant.

Yet, whether same-side interveners would have an equally robust but opposite impact on the course of events is not certain. Bandwagoning may result from shared as well as divergent preferences and hence presents a sample of interveners with mixed preferences and relations. There are distinct features of this seemingly cooperative environment which may dampen a favorable impact.

First, states that have similar preferences in a civil war may face a serious collective action problem and wait for like-minded states to pay the costs of intervention (Fearon and Laitin, 2004; Gent, 2007; Mearsheimer, 1990). If states that favor similar outcomes have an incentive to free-ride on others, bandwagoning, similar to balancing, should be disproportionately undertaken by states with competing preferences. Hence, genuinely cooperative interventions will be dwarfed by those that do not necessarily pursue similar solutions to wars.

Second, states may support the same side of the war because of differences in their policy positions. Interveners that hold competing preferences over the outcome try to manipulate the terms of a prospective bargain but, in doing so, they may both see their best chance for success residing in support to the same side of the civil war. Having disputes over the post-conflict order, bandwagoning states may compete for the loyalty of the same combatant and push for incompatible conflict outcomes. By adding a new bargaining dimension to the conflict, divergent preferences increase the number of actors that should commit to war’s end (Cunningham, 2006). The target then has a strong incentive to make the most gains from this environment and protract fighting for a decisive and favorable outcome. For instance, in Mozambique (1979–93), China, Russia, Britain, and the US supported the government side despite their entrenched disagreements. In spite of the fact that FRELIMO, with close ties to the Soviets, dominated the government, both the US and Britain denounced the rebel group RENAMO as a terrorist organization. Yet, the US and Soviets had targeted the same side to achieve almost contradictory outcomes. Another case that illustrates such divergence in interveners’ preferences is the tripartite relationship between Sukarno, the US, and the Soviets during Indonesia’s civil war against the revolutionary rebels based in Sumatra and Sulawesi. Under the disguise of neutrality, Sukarno skillfully manipulated US and Soviet aid where the major powers hoped to affect the Indonesian political system in line with their polarized preferences (Feith and Lev, 1963). ⁹

Third, strategic considerations also drive bandwagoning behavior. External actors — especially the weaker ones — may choose to intervene on the government side to maximize their influence in the war rather than wasting their resources in financing the rebels for an uncertain outcome. Indeed, the number of third-party interventions on the government side in the post-1945 period is almost twice the interventions on the rebel side, which may suggest that states find it cost-efficient to support the power-holders rather than their challengers. Cost-driven intervention can also take place on the rebel side: low-cost intervention efforts such as providing safe havens, arms, and military know-how dramatically increase rebels’ capacity to inflict costs on the government and subsequently increase their resolve to stay in the fight. Not only weak and isolated third-party states, but also major power interveners, may be driven by the costs of intervention in their choice of sides. American support for the ABC movements against communist-leaning governments in Africa and Asia was based on this logic. The former Director of the CIA, William Casey admitted that aid to anti-communist rebels was a cheap intervention strategy: ‘It is much easier and less expensive to support an insurgency than it is for us and our friends to resist one. It takes relatively few people and little support to disrupt internal peace and economic stability of a small country’ (quoted in Scott, 1996: 22).

In sum, same-side intervention does not always imply similarity in preferences, but can be driven by strategic and practical considerations. Thus, it is ambiguous whether bandwagons would have a more favorable impact than balancers on civil wars as they are likely to include a mix of cooperative and conflicting states. Following this discussion, we hypothesize that while balancers prolong the fighting, bandwagons will have an ambiguous effect on conflict duration:

Exclusively emphasizing interveners’ choice of sides conceals important patterns in their interrelations and limits our ability to understand the cumulative impact of interventions. While balancing states reveal the incompatibility in their incentives by supporting opposite sides in a civil war, bandwagons can either be motivated by shared preferences or by competing interests and require examining how these actors relate to each other. If we conceive bandwagons as forming a network in civil wars, identifying (dis)agreement patterns among network members becomes particularly important. The degree of agreement between a particular intervener and others in its network shapes and constrains how this intervener will affect combatants’ demands and preferences. Among networked interveners, political relations play an important role and determine interveners’ agreement and ability to reconcile their differences to generate a cumulative impact.

Third parties that not only support the same side but also have complementary preferences can tacitly or explicitly coordinate their actions to multiply their effect on the outcome. Those that have shared interests in general would also choose similar stakes in their intervention efforts. Favoring similar solutions to international problems, these states may have greater success in minimizing their differences, avoiding misunderstandings, and finding common ground to bring civil wars to an end. ¹⁰ Unlike competing states, states with genuinely similar preferences would make offers in an explicit and unified fashion, effectively signal their true preferences to other actors in the civil war environment, and clarify the terms on which they can accept negotiation and settlement. As they pursue similar outcomes to war, alignments among these interveners are likely to be tight. In this setting, combatants will be less likely to find an outside patron and gang up on their adversary, which diminishes their incentives to hold out until interveners sweeten the pot for settlement. Given the rigidity of their alignments, these interveners can eliminate combatants’ incentives to play them off against each other. Combatants are less likely to rebuff interveners’ offers when they receive unified signals and will be hard-pressed to shape their demands on the basis of the outside offer.

This discussion requires that we closely examine the political relations among bandwagoning interveners and leads to the main hypothesis that interveners sharing similar interests will favor similar outcomes to wars and, in doing so, will shorten war durations:

Research design

Our empirical analysis adopts an updated version of the interventions data originally used in Regan and Aydin (2006). This dataset covers all civil wars in the period 1945–99 which cross the threshold of 200 fatalities. There are 153 civil conflicts that fit this description (Regan, 2000). While 124 of these cases terminate within our time frame, 19 were still ongoing in 1999 and are therefore right-censored in the data. This is an aspect of the data that duration analysis, which we adopt in this article, takes into account. The dataset we adopt includes the most comprehensive information on three main forms of third-party intervention: mediation, military, and economic interventions. Following Bercovitch (1997), Regan and Aydin (2006: 745) define mediation as ‘a noncoercive, nonviolent, and, ultimately, nonbinding form of intervention’, which requires the consent of parties to civil war. There are 352 cases of mediation undertaken in 68 civil wars. Among coercive forms of intervention, economic interventions include material inducements, such as grant, loan, credit, relief from past obligations, nonmilitary equipment, and sanctions. Contributions of troops, naval support, equipment or aid, intelligence or advisors, air support to the civil war state, or military sanctions are defined as military interventions in this framework (Regan, 2002). There are 130 cases of economic and 942 military interventions coded in Regan and Aydin’s (2006) dataset. All three forms of intervention can be undertaken by states and inter governmental organizations (IGOs). In addition, mediation can be undertaken by individuals such as former heads of state or spiritual leaders.

We have focused our analysis on unilateral interventions undertaken by states without an explicit authorization from IGOs. While we use all interventions, including multilateral ones by international/regional organizations, in statistical tests, we adopt mediations and IGO interventions as the comparison group. Mediations and multilateral interventions are commonly studied as conflict management efforts in the civil war literature, whereas such intent is highly ambiguous in military and economic interventions. Therefore, the effect of coercive and biased interventions will be analyzed in comparison to mediations and interventions by international (e.g. UN) and regional (e.g. ECOWAS, EU) organizations as the baseline category.

The unit of analysis is intervention month. We examine how the individual impact of an intervener at time t is constrained or enhanced by its relation to other third-party states involved in this war. To build on the example of the Chadian civil war (1983–6), France intervened in the first month of the war to support the Hissen Habre government. We claim that the effect of this intervention was contingent on France’s interrelations with other intervener states and their alignment patterns. Other intervention attempts involve third-party states such as the US and former Zaire also on the government side, and Libya on the side of the rebels under the leadership of Goukouni. For France’s intervention, France bandwagoned with the US and former Zaire whereas it balanced against Libya’s influence attempts. ¹¹

There are two main groups of states that we consider in the network analysis: states that support opposing sides (balancers), and states that have certain alignment patterns (bandwagons). ¹² The dataset that we adopt codes the form of intervention (military, economic, or mediation) and whether a third-party intervention targets the insurgents or the government, which are the variables that we use to create these networks. ¹³ Since military and economic interventions are predominantly undertaken as biased actions supporting a particular side of the civil war, we adopt them in identifying balancing and bandwagoning behavior. ¹⁴ Bandwagons are especially important to our framework as they form the networks that target the same combatant. In the Chad example, France forms a network with the US and former Zaire by targeting the same side and their joint impact on the war is a matter of their interrelations. Both our analysis and previous work suggest that third-party states that intervene on opposite sides of a civil war consistently prolong war durations. Yet, not much is known about the cumulative effect of bandwagons, which is the group of states that we specifically focus on.

Our main dependent variable is the time to war termination in the next month and is created by the ‘stset’ command in Stata adopting monthly data of interventions. A failure event is defined as war termination and the cross-sectional component (or ID in Stata language) is the civil war. We have used a variance-adjusted Cox model with a discrete time dataset which includes all third-party interventions in the post-1945 period (Box-Steffensmeier and Zorn, 2002; Cleves, 1999). ¹⁵ This technique models the dependency between interventions within a civil war by accounting for ties between observations. This is an important feature that is consistent with our theory. While standard Cox regression assumes that observations are independent, duration models for multiple failure time data overcome this problem by adjusting the covariance matrix of the estimators to account for the remaining unit-level effects. We report robust standard errors clustered by civil war.

Statistical results

Our first hypothesis (H1) argues that the joint impact of interventions supporting opposing sides in a civil war is to prolong fighting. Limited empirical research has previously shown that Balancing diminishes prospects for conflict resolution (Regan, 2002). Our results reported in Table 2 present strong empirical support to earlier work and show that this type of intervener behavior significantly prolongs civil wars. While this is an intuitive finding and has received some scholarly attention, we make a contribution by testing this systematically and offering empirical support to the idea that opposing interventions complicate and prolong civil wars. In a civil war where an intervener attempts to offset others’ influence, this form of third-party alignment contributes to war duration by 92.9% (Table 4). ²¹ Balancers reveal the incompatibility in their preferences to the combatants and encourage them to look for solutions on the battlefield with the support of an external patron. The longer durations observed for civil wars in which balancing takes place imply that interveners competing with each other in this environment lead to a stalemate situation between combatants and spoil prospects for peace.

OPEN IN VIEWER

Table 2.

Intervener networks and civil war duration, variance-adjusted Cox regression

Independent variables	Model 1. Choice of sides		Model 2. Agreement patterns among interveners
	Coefficient	Std. error	Coefficient	Std. error
Balancing	−2.620*	.957	−2.640*	.966
Bandwagoning	−.839	.997
Interest Similarity			1.936*	.663
Network Size	−.249	.245	−.680*	.255
Solo Intervention	−2.236*	.585	−2.232*	.586
Support for the Rebels	−.646	.562	−.706	.561
Log GDP	−.554*	.112	−.557*	.113
Log Rebel Strength	.077	.067	.076	.068
Ethnic Fractionalization	.326	.377	.332	.377
Extractables	−.151	.230	−.141	.231
Democracy	−.193	.290	−.203	.301
Log Population	−.328*	.067	−.325*	.067
External Rebel Base	−.733*	.216	−.739*	.215
Number of observations	1617		1617
Number of civil wars	153		153
Number of failures	124		124
Log-likelihood	−437.274		−436.505
Wald chi-square (12)	92.27		101.84
Prob > chi-square	.000		.000

Note: Robust standard errors clustered by civil war are reported. * p ≤ .05.

While interveners supporting opposite sides contribute to the killing time, do bandwagons have an easier time in bringing civil wars to an end? Our answer is no. The insignificant finding on Bandwagoning in Model 1 suggests that this type of intervention behavior has no effect on war durations. This finding is in line with our expectations in H2 that bandwagons are a sample of interveners with mixed preferences and their cumulative impact on wars is ambiguous. ²² The favorable impact of states with shared preferences can be swamped by those supporting the same combatant but pursuing incompatible outcomes. This finding, however, presents strong support to the collective action problem among third-party states in intrastate violence discussed in previous research (Fearon and Laitin, 2004; Gent, 2007). States with similar preferences prefer that others pay the price or work within a multilateral framework to share the costs of intervention. They do intervene unilaterally when there is disagreement over war outcomes, which suggests that states that support a particular combatant in a civil war setting may disproportionately come from states with divergent preferences with regard to war outcomes (Gent, 2007).

The Support for the Rebels variable suggests that outside assistance for the rebel or the government side does not have an impact on how long the fighting continues. Even when multiple actors support one side of the conflict as in bandwagoning behavior, biased interventions make little difference in civil war processes. ²³ This result further indicates that how interveners choose sides does not matter in the course of the war; what matters most is whether interveners are willing to pursue similar solutions to wars and shape combatants’ bargaining positions accordingly.

Our results in Model 2 support this interpretation and show that interveners’ interrelations are central to understanding their cumulative impact on war termination. While choice of sides in the form of bandwagoning has a null association with war durations, choice of stakes provides important insights. Interest Similarity consistently reduces war duration in line with the expectations stipulated in H3a and H3b. For instance, in a network of five bandwagoning states, an intervention increases the prospects of war termination by 68.6% if there is agreement of interests among interveners in the form of similarity in voting patterns (90th percentile value, Table 3). Hence, our network approach suggests that states that support the same side and also share similar interests can unify their signals to the combatants, and credibly signal their preferences and reservation points to each other to find a common ground for settlement. They can coerce or entice the combatants into a deal that they agree upon and stop the fighting.

OPEN IN VIEWER

Table 3.

Substantive effects of agreement patterns among interveners

Network variables	Value	% Change in hazard ratio
Balancing	1	&0x002B;92.87
Interest Similarity	10th Percentile	&0x002B;50.55
	90th Percentile	−68.58
Log GDP	90th Percentile	&0x002B;33.85
External Base	1	&0x002B;52.21
Solo Intervention	1	&0x002B;89.27

Note: Hazard ratios are based on the findings reported in Model 2. All hazard ratios are significant at p ≤ .05. Network Size is set at its 90th percentile value (5).

Network Size decreases the prospects of war termination, leading to longer expected durations (Model 2); yet, its effect is not consistent across models. This finding is in line with previous research which has shown that civil wars with multiple actors tend to last longer (Cunningham, 2006). As the size of the network increases, interveners seem to further complicate the bargaining process between the war parties. When multiple interveners compete for influence over combatants or send mixed signals, they fail to generate a favorable impact on war durations. On the other hand, Solo Intervention has a consistently significant and negative coefficient and increases war duration by 89.3% over the baseline (Table 3). These findings together suggest that military and economic interventions lead to longer durations than organizational interventions or mediations independent of the number of actors that undertake them. As such, groups of interveners have the same detrimental effect with states that go it alone or they fail to shape war processes unless there is some sort of agreement between intervener states.

Other independent variables used as controls suggest that characteristics of the war and the civil war state also have a role to play in civil war duration by affecting the capabilities and resolve of the combatants. Ethnolinguistic composition of the civil war country (Ethnic Fractionalization) does not necessarily prolong wars in line with Fearon and Laitin’s (2003) findings on fractionalization and civil war onset. Log Rebel Strength measured as the natural logarithm of the number of rebel soldiers and Extractables in the form of opiates and gemstones fail to have a significant effect on rebel organizations’ capacity to wage a longer war. Considering that the resilience of insurgencies is closely related to the opportunity space shaped by geography and state capacity, the sheer number of rebel recruits may be less relevant to understanding how long rebels can protract the fighting. Pointing once again to the political opportunity space, rebels are capable and willing to fight longer wars when they have an External Base outside the borders of the civil war country (Salehyan, 2007). In line with our discussion on the role of government capacity in war duration, rich civil war countries (Log GDP) seem to be especially less likely to give in to insurgents and are willing to fight long wars against their opponents whereas governments in populous countries (Log Population) face resilient insurgencies and hence longer war durations.

In robustness checks, we have expanded the control variables in our analysis, and controlled for population outliers and the Cold War period (Table 4). In these alternative models, our findings on the key variables, Balancing and Interest Similarity, remained intact lending support to the centrality of intervener agreement in understanding the link between external intervention and conflict duration. In Model 3, India and China, as the two most populous countries in our sample, have been excluded from the analysis. No change has been observed on the effect of the Log Population variable: in line with the findings in Table 2, populous countries continued to be associated with long civil wars. Second, we have analyzed Model 2 for the Cold War period (Model 4) and our findings remained generally stable except for the insignificant finding on the Log Population variable. ²⁴ In Model 5, we have replaced the Extractables variable with Gold, Diamond, and Oil, which are coded as dichotomous variables that indicate the absence or presence of these natural resources in the civil war country. While Extractables has been consistently insignificant in all models, the Diamond variable is negative and significant indicating that civil war countries with diamonds do have longer wars than others. Diamonds seem to contribute to rebel resolve and facilitate recruitment more than oil or gold. Finally, we have controlled for conflict Intensity in Model 6. ²⁵ Results show that civil wars in which intense fighting takes place have shorter durations, which lends support to the ‘give war a chance’ arguments (Luttwak, 1999).

OPEN IN VIEWER

Table 4.

Robustness checks, Model 2, variance-adjusted Cox regression

Independent variables	3. Excluding India and China	4. Controlling for the Cold War period	5. Expanding Extractables	6. Controlling for conflict Intensity
Balancing	−2.590*	−2.390*	−2.608*	−2.650*
	(.966)	(.930)	(.976)	(.971)
Interest Similarity	1.881*	3.213*	1.926*	1.861*
	(.672)	(.893)	(.666)	(.730)
Network Size	−.663*	−.676*	−.680*	−.683*
	(.254)	(.280)	(.255)	(.262)
Solo Intervention	−2.240*	−2.230*	−2.607*	−2.310*
	(.595)	(.719)	(.664)	(.572)
Support for the Rebels	−.703	−1.399	−.673	−.584
	(.565)	(.739)	(.578)	(.578)
Log GDP	−.511*	−.398*	−.605*	−.578*
	(.120)	(.162)	(.115)	(.110)
Log Rebel Strength	.067	−.055	.093	.071
	(.070)	(.079)	(.070)	(.070)
Ethnic Fractionalization	.508	.291	.650	.448
	(.394)	(.477)	(.377)	(.375)
Extractables	−.132	−.136		−.169
	(.237)	(.347)		(.229)
Gold			.184
			(.257)
Diamond			−.976*
			(.435)
Oil			−.183
			(.284)
Intensity				.0003*
				(.00003)
Democracy	−.064	−.617	−.215	−.154
	(.297)	(.454)	(.332)	(.313)
Log Population	−.359*	−.154	−.336*	−.312*
	(.082)	(.087)	(.073)	(.071)
External Rebel Base	−.737*	−.990*	−.722*	−.672*
	(.213)	(.316)	(.215)	(.218)
Number of observations	1572	988	1598	1617
Log-likelihood	−412.818	−234.774	−415.481	−422.438
Wald chi-square (12)	95.43	64.27	112.63	167.14
Prob > chi-square	.000	.000	.000	.000

Note: Robust standard errors clustered by civil war are reported in parentheses. p ≤ .05.

Conclusion

Several processes endogenous to civil wars affect their duration in unexpected ways that researchers have yet to fully explore. The effect of third-party interventions is also a matter of what happens inside the conflict and is shaped by an interactive process that most of the time involves multiple states as interveners. A successful intervention outcome is more likely if external actors to an ongoing civil war can coordinate their actions, clearly communicate the outcomes for which they will distribute benefits, and basically draw a line in the sand. Clear and reinforcing signals will diminish incentives on the combatant side to blackmail interveners or their opponents. This is potentially critical because the disagreement between interveners opens up a new bargaining dimension and increases the profitability of war. It also encourages combatants to rescale their initial demands as they anticipate that interveners cannot take decisive action to enforce a certain course of action.

Yet, such coordination, useful as it might be, is hard to achieve if states considering intervention see the world in different ways, have a competitive rapport, and favor contradictory outcomes. Convergence between national interests seemed far from reach as major powers squabbled over their spheres of influence in the Cold War. Other bodies of evidence strongly suggest that there is more agreement between states on how to respond to violence in the international system in the post-Cold War period. The UN Security Council agreed on a greater number of peacekeeping missions after 1989 (Fearon and Laitin, 2004), whereas ad hoc coalitions of third-party states that aimed to overcome the bureaucratic struggles in the UN and bring an end to civil wars proliferated with the end of the Cold War (Prantl, 2005). In our case, we would therefore expect to see an enhanced effect of intervener networks on civil war duration in this period. Given that states are more able to take joint action and alignments are less rigid in a multipolar world, like-minded states would have an easier time in acting together and stopping the fighting in civil wars.

At the very least, the prescription that the evidence bears out is that states should refrain from sending mostly unintended signals of tolerance for violence through their inability to agree on and pursue a collective solution to civil wars. Besides, civil wars do not terminate immediately and decisively once external states demonstrate a united front. Though alignment patterns are important elements of interveners’ cumulative effect on civil wars, most interventions are simply ineffective and fail to facilitate wars’ end. The complex causes of civil wars complicate and constrain the effect of even the most benevolent third-party efforts. While disagreements between interveners will inevitably lead to longer war durations, our results demonstrate that cooperation can pay off under certain conditions.

Our network approach has captured the dependencies between third-party states’ unilateral efforts and modeled interveners’ role through their interrelations. It offered an improved model of civil war intervention in which the network of interactions between intervening states and the coalitions that they build are as important as the interventions themselves in shaping combatants’ preferences. This approach shifts the unit of analysis for understanding conflict management and peacekeeping from the unilateral intervention to the network of interveners. There are several important ways that the network approach can be advanced to better understand ad hoc coalitions that form between interveners in civil wars. Examining distributions of preferences within networks through similarity of foreign policy goals was one way to model third parties’ choice of stakes, whereas more complex ties between interveners, including economic and political, can improve our understanding of agreement patterns in intervener networks. Advanced analysis can also better account for coordination efforts by measuring power relations and the centrality of particular outside actors in developing networks.

A network approach to intervention has far-reaching, and potentially novel, scholarly and policy implications for understanding an important issue of international security. In a globalizing world where the adverse effects of civil wars can no longer be contained by the borders within which they unfold, this approach points to the centrality of international cooperation in solving a major issue in the global era. Despite the intuitiveness of our argument, previous studies have not provided the evidence to demonstrate that disagreement and competition between interveners can complicate the management of civil wars. With a more coherent understanding of the interactive role of external actors, IGOs and states can make sound policy choices based on scientific insight. One of the more pressing strategic challenges that confronts the international community in the post-Cold War global order remains the management of civil wars. In order to effectively manage these complex political issues our evidence suggests that states first need to overcome their own reluctance to cooperate, sometimes across ideological lines. At the very least, external states should make sure that their relationships with each other do not contribute to, rather than ameliorate, the struggles of people in civil war countries.