The legacy of foreign patrons

Abstract

Why do some armed conflicts that have ended experience renewed fighting while others do not? Previous research on conflict recurrence has approached this question by looking at domestic factors such as how the war was fought, how it ended or factors associated with its aftermath. With the exception of the literature on third-party security guarantees, the influence of outside actors has often been overlooked. This article explores the role of external states and suggests when and how their involvement is likely to affect the probability of renewed warfare. The main argument is that the legacy of outside support creates an external support structure that affects the previous combatants’ willingness as well as their opportunities to remobilize. This means that armed conflicts with external state support will experience a greater likelihood of recurrence compared to other conflicts which did not see external support. The theory is tested using Cox proportional hazards models on global data of intrastate armed conflicts 1975–2009. The findings suggest that external support to rebels increases the risk of conflict recurrence in the short term as groups receive or anticipate renewed assistance. The results also indicate that it is more important for rebel groups to have had enduring support over the years in the previous conflict rather than access to multiple state sponsors. External support provided to governments is not associated with conflict recurrence.

Keywords

civil war conflict recurrence external support third party interventions

Introduction

Many armed conflicts today are recurrences of previous conflicts (Walter, 2015). Some countries seem to be caught in a conflict trap that increases the likelihood of a relapse to civil war after a period of peace. But why do some armed conflicts recur while others do not? Most explanations as to why fighting resumes have stressed the importance of domestic factors such as opportunities for remobilization (Fearon & Laitin, 2003; Collier, Hoeffler & Rohner, 2009; Themnér, 2011) or unaddressed grievances (Walter, 2004; Hegre & Nygård, 2015; Walter, 2015). In contrast to previous explanations, I emphasize the need to expand our focus to incorporate the influence that outside states might have on the probability of renewed warfare. I argue that previous instances of biased third-party interventions create an external support structure that influences the risk of conflict recurrence. This structure alters the opportunities and calculations made by the former combatants by making it more attractive to challenge the established status quo. I posit two theoretical mechanisms as to how this process could work: facilitation and anticipation. The first mechanism (facilitation) stresses that remobilization is simply easier with the assistance of outside states. External states can directly enable remobilization by providing strategic advantages such as safe havens, training, and weapons. This offers actors dissatisfied with the outcome of the previous war a chance to reorganize and rearm. Factors that increase the opportunities for remobilization and which strengthen the resolve of the aggrieved actor are hence not confined to the borders of the country that has experienced the civil war. For example, the offensive by the rebel group LURD, which initiated the second civil war in Liberia in 1999, only began after the group had been able to train and regroup in neighboring Guinea. Furthermore, the government of Guinea assisted the group with procuring arms and transported both weapons and troops from Conakry and other areas of Guinea into northern Liberia (Hazen, 2013). The second mechanism (anticipation) emphasizes the role of expectations. This builds on the notion that channels to support remain in place even after the conflict has ended. Actors are likely to believe that external assistance will be provided again if it has been provided in the past. This should increase the confidence in a potential future victory and thus the resolve of the aggrieved actor who should be more likely to challenge the prevailing status quo. For example, in 2008, the conflict between the Republic of South Ossetia and Georgia resurfaced. As in past conflict episodes, the forces of South Ossetia could expect substantial support from Russia as Georgian settlements and positions were targeted. When the military forces of Georgia responded with a large-scale military offensive on the city of Tskhinvali, Russia immediately reacted by sending troops, tanks, and bomber planes to assist South Ossetia (Asmus, 2010).

External support is defined here as a unilateral intervention by a third-party state in an internal armed conflict in favor of either the government or the opposition movement involved in that conflict. Some scholars refer to these instances of external support provision as ‘biased interventions’, ‘secondary support’ or ‘partisan interventions’, while others use the more general term ‘third-party interventions’. Support can include both the direct participation with troops (military intervention) and more indirect forms of aid such as the provision of sanctuaries, weapons, logistics, intelligence, training or money. An external support structure is defined as an expectation about renewed material support from a foreign state based on the fact that support has been offered in the past.

I make two main contributions to the literature on conflict recurrence. First, I complement earlier domestic explanations of conflict relapse with an account of how the involvement of external states can affect the likelihood of conflict recurrence. Although previous research has noted the role of third-party interventions in the form of peacekeeping operations, little attention has been paid to how biased support provided by third parties has the potential to make conflict recurrence more likely.¹ Second, I propose and empirically test a novel yet intuitive theoretical argument as to how we would expect outside states to affect conflict recurrence. My theory challenges the dominant view in the conflict recurrence literature that external states generally have good intentions and that they are essentially there to provide security guarantees in light of a negotiated settlement. Instead, the involvement of outside states can directly and indirectly contribute to conflict recurrence. By emphasizing the role of an external support structure I stress how the decision by former belligerents to re-engage in violence is at least in part affected by factors located outside of the civil war country.

My findings show that external support to rebel movements increases the probability of conflict recurrence in the short term. The legacy of foreign patrons influences former combatants’ calculations by lowering the threshold for remobilization and a return to violence. Enduring support over many years during the previous conflict raises expectations that support will be forthcoming once again and is of greater importance than having had access to multiple supporters in the past. External support provided to governments, on the other hand, does not influence the risk of conflict recurrence.

The article proceeds as follows. The next section reviews the literature on conflict recurrence and identifies the research gap. The following section presents my theoretical argument about how the involvement of external states can fuel renewed violence. I then describe and motivate my research design and specify how the variables will be measured. This is followed by an analysis of all armed conflict terminations from 1975 to 2009. The article ends with some concluding remarks and a brief discussion of policy implications.

Why do conflicts relapse?

Previous research has identified a range of different factors that are associated with the recurrence of conflict. These explanations can be roughly divided into four strands. The first strand claims that many of the state attributes that led to the initial conflict might favor its recurrence. In particular, these scholars point to the fact that weak states with lower levels of economic development are particularly prone to conflict (Fearon & Laitin, 2003; Walter, 2004; Collier, Hoeffler & Rohner, 2009). The second strand links the character and dynamics of the previous conflict to recurrence. These scholars have especially focused on the duration of the prior conflict, its intensity, and the goals of the combatants, as well as recruitment strategies (Doyle & Sambanis, 2000; Sambanis, 2000; Fortna, 2004; Walter, 2004; Quinn, Mason & Gurses, 2007; Kreutz, 2010). The third strand focuses on the way the previous armed conflict was terminated. Conflicts that end in victories are generally less likely to recur but there is disagreement over whether government or rebel victories are particularly stable (Quinn, Mason & Gurses, 2007; Kreutz, 2010; Toft, 2010; Mason et al., 2011). Although some scholars (Licklider, 1995; Walter, 2004) have suggested that conflicts which end with negotiated settlements are more conducive to conflict recurrences, this is conditional on whether these agreements are combined with third-party security guarantees or not (Walter, 2002; Hartzell & Hoddie, 2003). International peacekeeping missions generally make peace more durable and thus conflict recurrence less likely (Doyle & Sambanis, 2000; Walter, 2002; Fortna, 2004; Quinn, Mason & Gurses, 2007; Call, 2012). The nature of the peace agreement is also of importance. Conflict recurrence is less likely following power-sharing agreements and if the peace accord included civil society actors (DeRouen, Lea & Wallensteen, 2009; Nilsson, 2012). The fourth strand of scholars focuses on the post-conflict environment. Walter (2004) identifies two conditions that make re-enlistment an attractive option: individual hardship and the absence of nonviolent means of change. She finds that indicators of ‘quality of life’ such as infant mortality, life expectancy, and adult literacy are the best predictors of conflict recurrence. Themnér (2011) highlights the importance of social networks between ex-combatants that remain intact after the conflict has ended. The existence of these can facilitate remobilization and thus increase the risk of renewed conflict. Call (2012) finds that political exclusion is correlated with conflict recurrence and both Walter (2015) and Hegre & Nygård (2015) stress that conflicts are less likely to recur if grievances are addressed through good governance in the aftermath of war.

As should be evident from this review, most explanations of conflict recurrence focus on domestic factors and thus neglect the wider regional and international context. Discussions of the role of third parties that intervene in the conflict have been limited to studying conflicts which ended in a negotiated settlement or in which a peacekeeping mission was deployed. In this research, external states are mainly thought of as becoming engaged in the conflict country in order to establish security guarantees that would assist the conflict actors in overcoming the commitment problem (Doyle & Sambanis, 2000; Walter, 2002; Hartzell & Hoddie, 2003; Fortna, 2004; Quinn, Mason & Gurses, 2007). According to this logic, external states should make conflict recurrence less likely. However, as partly acknowledged by the literature on spoilers of peace processes, the involvement of external states might not always be beneficial. External states can provide actors with the resources necessary to resume violence (Zahar, 2006).² Third-party involvement is not limited to benevolent interventions aimed at protecting agreements and keeping peace. States offer external support for a variety of different reasons and they often have agendas that are independent from those of the combatants (Cunningham, 2010). While humanitarian concerns can be important, there are often strategic motivations behind decisions to intervene. States might want to weaken an adversary, limit conflict diffusion, increase their regional influence, secure access to geopolitical resources, or offer support to a kin group (Saideman, 2002; Byman, 2005; Aydin, 2010; Kathman, 2011; Salehyan, Gleditsch & Cunningham, 2011). Many of the former colonial powers also maintain a strong interest in the affairs of their past colonies (Gregory, 2000; Stojek & Chacha, 2015).

Cunningham (2011: 186–194) finds that third-party involvement can indeed shorten the duration of peace. His study indicates that conflicts that end in negotiated settlements are more likely to recur if there are many potential veto players to an agreement.³ However, this study differs from previous work in a number of notable ways. First, Cunningham’s analysis is restricted to those cases in which a negotiated settlement has been reached, but most armed conflicts do not end in negotiated settlements. Victories and deadlocks are far more common (Kreutz, 2010). Second, possible veto players are limited to former combatants, but external states can also be involved in the conflict without officially becoming warring parties. Extensive provision of material support does not make a state an active combatant but it is nevertheless likely to facilitate remobilization and affect the calculations of the belligerents. Lastly, biased unilateral third-party interventions are much more common than neutral multilateral interventions such as peacekeeping operations (Regan, 2002). This highlights the need to pay closer attention to these types of interventions and the potential impact they could have on conflict recurrence.

The importance of an external support structure

We know that external support in civil wars generally makes conflicts deadlier, longer, and more intractable (Lacina, 2006; Cunningham, 2010; Aydin & Regan, 2012). But how does past support influence the likelihood of conflict recurrence? I argue that previous instances of biased third-party interventions create an external support structure that remains in place even after the conflict has ended. Opportunities for remobilization are not limited to the borders of the post-conflict country and international connections do not disappear as soon as the fighting stops. This is analogous to the local level in which personal networks among former combatants often remain intact in the aftermath of war. These durable relationships enable mid-level commanders to remobilize individual fighters at a later point in time (Themnér, 2011). Similar to networks at the individual level, established relationships between belligerents and external states fade away only gradually. My argument views the external support structure established by past instances of support as an enabling condition that influences the calculations of the challenger and lowers the threshold for remobilization. I acknowledge that the former combatants will also need reasons to re-engage in violence and that the presence of an external support structure alone does not suffice to explain the timing of conflict recurrence. In what follows I outline two related theoretical mechanisms as to how the involvement of external states can increase the risk of conflict recurrence: facilitation and anticipation. While the first constitutes a direct effect in that it strengthens an actor’s capacity for violence, the second is indirect as it influences an actor’s cost–benefit analysis.

The first mechanism stresses that outside states can directly facilitate remobilization by providing strategic advantages to the challenger of the status quo. Facilitation could for instance come in the form of monetary assistance, military training or the provision of arms. The presence of an external facilitator simply makes remobilization easier. Actors that have access to foreign supporters are in a much better position than others to challenge the prevailing status quo. The provision of outside support shifts the domestic power balance and affects the calculations of the challenger. Actors that are able to secure strategically important resources from a patron should estimate a higher probability of future victory and as a result can expect to pay a lower cost for conflict.

The second mechanism emphasizes that the mere anticipation of renewed support can be enough to alter the calculations of the challenger. Expectations about future events are informed by past experiences. Hence, the party dissatisfied with the status quo is likely to believe that external support will be provided in the future if this has been the case in the past. This would mean that material resources from external states do not have to be provided at the time of recurrence. The mere legacy of external support is enough to increase the risk of conflict recurrence. A belief in renewed support alters the calculations made by the former combatants by making it more attractive to challenge the established status quo. The resolve increases as reinitiating violence with outside support is less costly.

I am not the first to emphasize that the mere belief in outside assistance can affect internal conflict dynamics. Some scholars have put forward similar ideas in other contexts. Jenne (2007) has shown that marginalized ethnic groups are more likely to radicalize their demands and challenge the government if an external state patron has credibly signaled interventionist intent. She stresses that military patronage could be particularly dangerous as it offers groups a window of opportunity through which they can gain an important strategic advantage. This influences the calculus of groups even if support never materializes (Jenne, 2007: 189). Similarly, Kuperman (2008) has argued that past humanitarian interventions result in the unintended consequence that groups might be increasingly likely to engage in conflict. The central idea is that groups have an increased willingness to use political violence in order to provoke state retaliation if they anticipate outside assistance. Past interventions thus play a significant role in altering the calculations of potential challengers by lowering the expected cost while at the same time increasing the expected likelihood of success. Both mechanisms lead to the same observable implication: conflict recurrence after the prior provision of external support.⁴ This leads to my first hypothesis:

H1a: The provision of external support in the previous conflict episode makes conflict recurrence more likely.

The nature of the previous support network can also be important. The number of years support was provided in the previous conflict (how enduring the support network was) as well as the number of supporters (how extensive the support network was) can make the challenger more or less confident that external support will be provided again. The type of support network should similarly affect the commitment by state sponsors. If many states were involved in the last conflict episode it is likely that at least one of them will offer renewed support. More extensive support networks hence increase the probability that support will be provided in the future. Additionally, enduring support over time in the previous conflict episode might be an indicator of the level of commitment from external states as some states have longstanding ties with actors engaged in intrastate conflicts and will hence offer continuous support throughout a conflict. Examples include USA–Israel and Iran–Hezbollah. To assess whether the configuration of the previous support network matters, I create two additional hypotheses that further specify hypothesis H1a.

H1b: More external supporters in the previous conflict episode make conflict recurrence more likely.

H1c: Enduring support during the previous conflict episode makes conflict recurrence more likely.

For most state supporters it is likely that their commitment to aid will diminish over time. The challenger’s confidence in renewed assistance should also decrease as the years go by. Uncertainty increases as time elapses and relationships fade. Therefore, it is reasonable to assume that the expectation of renewed support would be highest just after the conflict has ended. This is when linkages between external states and challengers are intact and a commitment to offer support has recently been demonstrated. Then – as time goes by – the challenger should become increasingly uncertain as to whether the previous commitment by outside states will hold. Renewed support should be more forthcoming in the first couple of years after a conflict has terminated. This is both when the revisionist actor should be more likely to challenge the post-conflict status quo and when states would be willing to provide renewed support. I thus make the central assumption that the effect of prior external support should diminish over time and I will specify the models accordingly. Furthermore, in the empirical tests I distinguish between external support provided to governments and rebel movements as the receivers might differ in their willingness to re-engage in violence or in the importance they place on an external support structure.

Research design

The dataset builds on the notion of conflict episodes, defined as continuous periods of active conflict-years in the UCDP-PRIO Dyadic Dataset. Intrastate armed conflict is defined as an incompatibility between the government and a rebel group that results in at least 25 battle-related deaths in a calendar year (Gleditsch et al., 2002). Measurement begins the year after the first conflict episode has ended as determined by the dyadic UCDP Conflict Termination Dataset (Kreutz, 2010). I include all intrastate armed conflicts, regardless of how they ended, and incorporate the most important type of terminations as controls in subsequent statistical models. My unit of analysis is the conflict-dyad year after termination. An observation enters the dataset once the first conflict episode has ended. The following years represent when a conflict-dyad is at risk of experiencing a relapse into war. If for instance the conflict between Nicaragua and FSLN first ended in 1979, this dyad enters the dataset the following year. This is theoretically the period when an armed conflict could recur. I then observe this dyad until the conflict recurs or the period of observation ends (2009), in which case the dyad is right-censored. I allow for multiple failures in the same dyad and take into account that recurrences might be interrelated by clustering the standard errors on the dyad. The dataset covers conflicts that have not ended prior to 1975 and runs up until 2009. The time period is determined by data availability for the provision of external support. The dataset contains 433 episodes with 5,274 conflict-dyad years after termination, of which 134 are recurrences. I employ Cox proportional hazards models in order to test my argument whether an external support structure affects the probability of conflict recurrence. I use the Cox model instead of its parametric alternatives as it makes no assumptions about the distributional form of the baseline hazard rate (Box-Steffensmeier & Jones, 2004). The survival time is the duration of peace years until a potential recurrence of armed conflict (the failure event) and the treatment is whether or not external support was provided in the previous conflict episode.

Main variables

Conflict recurrence is defined as the onset of a new conflict episode involving the same actors that have previously been involved in fighting. This constitutes the failure event in my analysis. It is binary and coded 1 in the year a new conflict episode starts and 0 otherwise. These data are retrieved from the dyadic UCDP Conflict Termination Dataset v.2010-1 (Kreutz, 2010).⁵

The information for my main independent variables is obtained from the UCDP External Support – Primary Warring Party Dataset v.1.0-2011 (Högbladh, Pettersson & Themnér, 2011). The subset of the data I use only includes states as supporters and is limited to confirmed instances of external support.⁶ The main reasons for using the UCDP dataset compared to other datasets is that (1) it covers many different forms of external support, (2) it offers detailed information about state supporters, and (3) it provides time-varying data on an annual basis which makes it possible to assess how extensive and enduring the support network was.

My independent variable Rebel support is a binary variable that denotes if external support was provided to the rebel side during the previous conflict episode. Government support is the equivalent measurement for the government side. These variables will be used to assess whether the provision of external support in the previous conflict makes conflict recurrence more likely (H1a). In order to account for the extensiveness of the support network, I include two count variables measuring how many external supporters offered support in the previous conflict episode. These variables, Number of rebel supporters and Number of government supporters, are used to assess the second hypothesis (H1b). To capture how enduring the support was and evaluate the last hypothesis (H1c), I include a measure of the proportion of years of the last conflict episode in which external support was provided. These variables, Years of rebel support and Years of government support, are generated by taking the number of years in which there was at least one confirmed state that provided support divided by the number of years that the previous conflict episode lasted. To illustrate: the rebel group Frolina in Burundi was active between 1994 and 1998 and received support from at least one external state for two of these four years, so the corresponding proportion is 0.5. Measuring this variable as a share takes into account that conflict episodes vary in length. Moreover, a diminishing effect over time is essential to my theory. Hence I add a decay function to all of the external support variables. The decay function allows for the possibility that the anticipation of renewed support is strongest just after a conflict termination and that the effect then decreases over time as uncertainty rises. It also accounts for the decreasing probability that support actually will be provided again. All of the main variables are assigned their original value the first year after a termination and then they decay with a half-life of two years.⁷ As an example, consider a conflict episode with external support to the rebel side. For the binary variable Rebel support it would then take the value of 1 the first year. The second year it will decrease to 0.71 and the third year it will have reached its half-life value of 0.5.⁸

Control variables

Many explanations exist for why some conflicts are more likely to recur than others. Following Ray (2003) and Achen (2005), I will focus on control variables which can plausibly act as confounders. First, I include a measure of Conflict intensity. A conflict of higher intensity would most likely result in a postwar environment with more widespread grievances among the population and a state with weaker capacity to uphold order. Conflict intensity is also related to third-party intervention in that states might feel a stronger urgency to intervene in conflicts of higher intensity.⁹ This variable is a dummy variable denoting whether the previous conflict episode reached more than 1,000 battle-related fatalities; it is based on data from the Uppsala Conflict Data Program (UCDP). Second, a measure of Conflict duration captures how many years the previous conflict episode lasted. Conflicts might recur

Table I.

Summary statistics of variables

Variable	Observations	Mean	Std. dev.	Min.	Max.
Recurrence	5,274	0.25	0.16	0	1
Time	5,274	11.10	7.59	1	37
Rebel support*	4,398	0.11	0.24	0	1
Government support*	4,398	0.15	0.27	0	1
Rebel supporters*	4,398	0.23	0.78	0	17
Government supporters*	4,398	0.36	1.25	0	33
Years of rebel support*	4,372	0.09	0.21	0	1
Years of government support*	4,361	0.13	0.25	0	1
Conflict intensity	5,265	1.26	0.44	1	2
Conflict duration	5,274	3.92	5.46	1	41
Ethnic dimension	4,794	0.57	0.49	0	1
Victory	5,274	0.30	0.46	0	1
Peace agreement	5,274	0.13	0.34	0	1
Frozen conflict	5,274	0.36	0.48	0	1
Peacekeeping operation	5,274	0.16	0.36	0	1
Territorial conflict	5,272	0.34	0.47	0	1

* Variables with decay functions.

because the previous conflict has been too short to fully reveal the relative strength of the combatants, and a more limited time period would give external states less opportunity to provide external support. Third, I take into account whether the previous conflict had an Ethnic dimension. The variable is based on data from Walter (2004) and extended by Kreutz (2010). It measures whether the combatants broke down along ethnic lines or a faction identified itself as an ethnic group. When war breaks out ethnic identities and hatreds tend to become cemented and make coexistence of the different groups even more difficult and conflicts more likely to recur (Walter, 2004: 372). Ethnic conflicts are also more likely to attract external state sponsors that might intervene based on shared kinship ties (Saideman, 2002). Fourth, I include a binary measure of whether a Peacekeeping operation has been deployed to the conflict in a given year following the termination. I include all peacekeeping missions that have a UN mandate; these data come from the Third-Party Peacekeeping Missions Data Set 1946–2014 version 3.1 (Mullenbach, 2013). The presence of a peacekeeping operation decreases the risk of conflict recurrence and at the same time it signals resolve from the international community. This may deter past supporters and potentially weaken the anticipation of renewed external support. Sixth, I include a dummy variable that denotes whether the conflict incompatibility concerns Territory. This is coded based on the incompatibility listed in the UCDP Dyadic Dataset v.1-2015 (Harbom, Melander & Wallensteen, 2008). It has been claimed that territorial conflicts are more entrenched and more likely to recur (Quinn, Mason & Gurses, 2007: 180). It is possible that external support is less forthcoming to the challengers in these cases since most states are status quo-oriented when it comes to altering territorial boundaries. This reasoning is based on the empirical observation that most external support is channeled to governments rather than rebels (Högbladh, Pettersson & Themnér, 2011: 13). Seventh, I include a number of dummy variables related to the Type of conflict termination. The types of terminations that I include are: victories, peace agreements, and frozen conflicts (those that end because of no/low activity). The reference category for any stated type of termination is all other types of possible outcomes. External support is correlated with certain outcomes (especially victories) as previous research has shown that support can increase the likelihood of both rebel and government victories under certain conditions (Lyall & Wilson, 2009; Sullivan & Karreth, 2014). Several scholars have also noted that how the last conflict ended might have important consequences for whether or not it will restart (Licklider, 1995; Quinn, Mason & Gurses, 2007; Toft, 2010; Mason et al., 2011). The data on the type of termination come from the UCDP Conflict Termination Dataset v.2010-1 (Kreutz, 2010). Table I provides summary statistics of all the variables.

Empirical analysis

Does external support during the previous conflict influence the likelihood of renewed warfare? I have argued that the legacy of outside support creates an external

Table II.

Effect of external support on conflict recurrence

	Model 1	Model 2	Model 3
Government support	1.03 (0.39)
Rebel support	2.62** (0.77)
No. of government supporters		1.04 (0.07)
No. of rebel supporters		1.12 (0.11)
Years of government support			0.98 (0.38)
Years of rebel support			2.51* (0.93)
Conflict intensity	1.24 (0.36)	1.24 (0.37)	1.17 (0.35)
Duration of prev. conflict	1.02 (0.03)	1.02 (0.03)	1.03 (0.04)
Ethnic dimension	3.32* (1.95)	3.41 ^† (2.15)	3.29* (1.99)
Ethnic dimension × ln(time)	0.43* (0.16)	0.43* (0.16)	0.43* (0.16)
Peacekeeping operation	0.64 (0.25)	0.64 (0.25)	0.57 (0.23)
Territory	0.75 (0.37)	0.80 (0.42)	0.79 (0.41)
Territory × ln(time)	2.86** (0.97)	2.76** (0.96)	2.83** (0.98)
Victory	0.17* (0.11)	0.17** (0.11)	0.12** (0.09)
Peace agreement	0.71 (0.34)	0.74 (0.35)	0.75 (0.36)
Frozen conflict	2.61** (0.80)	2.69** (0.82)	2.67** (0.83)
Subjects	256	256	254
Failures	99	99	97
Observations	4,023	4,023	3,985
Wald Chi-square	97.06	74.67	79.16
Log pseudo-likelihood	−459.33	−461.86	−447.79

^† p < 0.1, *p < 0.05, **p < 0.01. Hazard ratios reported. A ratio with a value above 1 indicates an increased risk for conflict recurrence, while a value below 1 indicates a decreased risk for conflict recurrence. Efron method of ties with standard errors clustered on conflict. Standard errors in parentheses.

support structure that affects the previous combatants’ calculations as well as their opportunities to remobilize. If this is correct, it means that armed conflicts with external state support have a greater likelihood of recurring compared to other conflicts. Table II shows the results from the Cox proportional hazards models that capture the effect of external support in the previous conflict episode on conflict recurrence. Hazard ratios are presented in the table and values above 1 indicate that the particular variable increases the risk of conflict recurrence while a value below 1 indicates that it decreases the risk of recurrence.

All three models suggest that external support matters – depending on who received the support. External support to governments does not increase the risk of conflict recurrence. None of the indicators for past support to the government side reach conventional levels of statistical significance. This means that none of the stipulated hypotheses (H1a–c) are confirmed in regard to external support provided to governments. The reason for this might be that governments rarely are the initiators of renewed violence as they may be more likely to favor the status quo. There is an inherent logic to the assumption that governments would seek to avoid war because of the destabilizing consequences of conflict. Governance and economic life become more difficult as resources need to be diverted to the war effort (Kreutz, 2012: 113). External support – or the belief that this would be forthcoming – should matter more for rebel groups since they are often the weaker party. External resources can thus serve to reduce the asymmetric power relationship between the rebel group and the government and increase the group’s bargaining power vis-à-vis the state. This makes supplies from external states of greater importance for rebel groups as additional resources contribute more to their organizational capacity and remobilization efforts.

External support to rebel movements increases the probability of conflict recurrence.¹⁰ The first hypothesis (H1a) stipulated that the provision of external support in the previous conflict period should make conflict recurrence more likely. Model 1 supports this hypothesis. External support to rebel movements increases the hazard ratio by 162% compared to the baseline. This suggests that external support to the rebel side increases the risk of conflict recurrence in the short term as groups

Figure 1.

Coefficient plot of main effects

receive or come to anticipate renewed assistance. The second hypothesis (H1b) stipulated that more external supporters in the previous conflict episode would increase the likelihood of conflict recurrence. If the support network was extensive – meaning that many states were involved in the last conflict episode – it would be more likely that at least one state would also offer renewed support. Model 2 shows that the number of supporters is of little importance in regard to conflict recurrence. While the hazard ratio is in the predicted direction it fails to reach conventional thresholds of statistical significance and it is not robust to alternative model specifications. The third hypothesis (H1c) asserted that persistent support during the previous conflict episode makes conflict recurrence more likely. As shown in Model 3, the results indicate that persistent support during the previous conflict episode makes conflict recurrence more likely. Enduring support networks in the past conflict provide reassurance that states are committed and that renewed support is likely to be offered again in the future. The hazard ratio increases by 151% as the share of years in which the rebel group received support also increases. This finding suggests that the rebel group’s calculations to challenge the prevailing order are at least partly influenced by whether or not they have previously received support from states over a longer period of time. Figure 1 is a coefficient plot that conveys the main results visually. The graph is based on the full models but only the main variables are shown. The dotted middle line corresponds to the null hypothesis that no significant relationship exists. The figure illustrates that both prior rebel support and the share of years with prior rebel support increase the risk of conflict recurrence.

Consequently, state sponsorship of rebel groups during civil wars can have the unforeseen effect that it contributes to a less durable peace once the fighting has stopped. States cannot counter this by providing support to the government during the war since external support to the government side is accounted for in the models. All results are conditional on a decreasing effect over time.¹¹ Some of the control variables behave as expected in relation to findings from previous research on conflict recurrence while others do not. Most likely, there are three reasons for this. First, many studies focus only on large-scale civil wars and exclude minor armed conflicts (Walter, 2004; Quinn, Mason & Gurses, 2007; Mason et al., 2011; Walter, 2015). Second, several studies limit their investigation to conflicts that ended in a particular way (Hartzell & Hoddie, 2003; Fortna, 2004; DeRouen, Lea & Wallensteen, 2009; Mattes & Savun, 2010). Third, some studies focus on the country or the aggregate conflict-level and not the conflict-dyad. The scope of this study is hence broader as it is not limited to a certain type of termination (such as peace agreements or ceasefires) and as it includes armed conflicts with less than 1,000 battle-related deaths.

Two variables (territory and ethnicity) violated the proportional hazards assumption. This means that their effect on conflict recurrence is not constant but changes over time.¹² The presence of a peacekeeping operation seems to reduce the baseline hazard of recurrence but is not statistically significant. Consequently, third-party security guarantees do not seem to be a strong predictor of post-conflict stability for this wider set of cases. This does not contradict the findings that peacekeeping missions are effective in maintaining peace after agreements or large-scale conflicts (Doyle & Sambanis, 2000; Fortna, 2004). In regard to conflict outcomes, recurrences are significantly less likely after victories. The hazard rate decreases 83–88% for conflicts that end in victory for one of the belligerents. This is in line with what has been found in most of previous research. Frozen conflicts – those that have ended without a conclusive agreement – are much more likely to recur. The

Figure 2.

Conflict-dyads with rebel support that recurred

hazard ratio increases 161–169% compared to the baseline.

Extending the analysis

Probing the mechanisms

So far I have demonstrated that conflict recurrences are associated with prior external support to rebel groups but not to governments. But which of the proposed theoretical mechanisms are at work? In order to try to discern this I take a closer look at all conflicts with prior rebel support that recurred to see if support was provided in the years prior to recurrence that could have facilitated remobilization. As the UCDP dataset lacks information on external support during interim periods of violence, I use the dataset created by Regan & Meachum (2014) to identify military or economic interventions in the years prior to recurrence. Figure 2 presents these data in circle diagrams.

External support was provided in only 4% of the observed cases within five years prior to recurrence and never in the year before recurrence. This lends some support to the idea that it is mainly the anticipation mechanism that is at work.¹³ This suggests that states do not need to provide new material resources prior to recurrence; the mere legacy of external support appears to be enough to increase the risk of recurrence by altering the calculus of the challenger. Are rebels correct in anticipating renewed support from a past state sponsor? The right circle diagram, based on UCDP data, shows that rebels receive renewed support from one of their previous supporters about half of the time in the year when they return to violence. So, it appears that rebels are adept at anticipating the provision of support. While a more elaborate test of the different mechanisms would necessitate in-depth process tracing, the data do provide some face validity to the idea that anticipation of support matters.

Reducing model dependency

External support to rebels is unlikely to be randomly distributed. In order to address endogeneity issues and further reduce model dependency I employed propensity score matching and re-estimated the main models on the matched sample (Rosenbaum & Rubin, 1983). The matching procedure pairs observations which are similar in regard to background factors that could explain the ‘treatment’ – in this case, rebel support – to achieve a more balanced sample. To calculate propensity scores, I replicated the main model by Salehyan, Gleditsch & Cunningham (2011: 728) using a logit model with external support to rebels as the dependent variable. This was followed by nearest neighbor one-to-one matching with replacement. In order to achieve balance between the samples, all variables that did not reduce bias or those that demonstrated a statistically significant difference between the groups were dropped. When matching is perfect and the distributions match exactly a simple difference in means would be sufficient to obtain the effect. However, matching is rarely perfect and my results are conditional on the effect diminishing over time. Following Gilligan & Sergenti (2008) I thus rerun my main Cox models on the matched subsample in order to adjust for any remaining imbalance and to account for right-censoring and the temporal nature of the data. This method is doubly robust in that if matching is not complete but the parametric model is properly specified, or if the model is not properly specified but the matching is adequate, estimates will be consistent (Ho et al., 2007: 215; Gilligan & Sergenti, 2008: 97).¹⁴ The results from the post-matching analysis are available in the Online appendix. The findings for the main independent variables hold in substantive terms and in fact the effects appear to be larger. Moreover, the variable denoting the number of government supporters in the previous conflict episode now also becomes statistically significant in the hypothesized direction.

Robustness and extensions

A number of additional robustness checks and extensions were conducted in order to further evaluate the results. All tests and the supplementary do-file are available in the Online appendix. First, the main models were re-estimated with a more complete model specification that included additional control variables accounting for country characteristics (GDP per capita, level of democracy, and population). In addition, alternative measurements of some of the control variables were included in separate models. Measuring intensity as the number of battle-related deaths and duration as the number of years since the conflict-dyad first became active (reached the threshold of 25 battle-related deaths) does not change the results substantively. Alternative measurements of peacekeeping operations that included all peacekeeping missions and not only those with a UN mandate, or accounted for peacekeeping missions after they departed, yield similar results and do not alter the effect of the main independent variables.

Second, if the rebels win the previous conflict and become the new government it is possible that this would instead influence the anticipation or provision of outside support to the new government. This might lead the government to strike first to oust any future competition and it would be increasingly likely that the government could successfully defend the status quo. To account for this possibility I included interaction terms in each of the models between prior rebel victory and support to see whether the results hold up. Results remain unchanged.

Third, many scholars use a similar operationalization of conflict recurrence to the one employed in this study (see for instance DeRouen & Bercovitch, 2008; Kreutz, 2010; Flores & Nooruddin, 2012; Haer & Böhmelt, 2016). However, an alternative conceptualization of the dependent variable with a two-year gap in fighting is also common. To this end, I estimated additional models. These models demonstrate that the results are sensitive to this alternative conceptualization. Most recurrences occur within the first couple of years so this excludes almost half of the original observations that experienced a return to violence. Interestingly, this suggests that the results are driven by conflicts which recurred shortly after they terminated. Most of these are examples of conflicts that terminate because of stagnating or no violence but there are also instances of short-lived victories, peace agreements, and negotiated ceasefires. The type of termination is accounted for in the statistical models. Indeed, the statistically significant control variable ‘frozen conflict’ included in the original model shows that it has a strong effect on recurrence independent of external support provision. It appears as if foreign state patrons affect the risk of conflict recurrence mostly in the short term when peace is still fragile and that they are less influential once peace has taken hold. This might be because ties to external states fade away in the long run.

Fourth, the findings are robust to different rates of decay. However, estimating the models without decay functions demonstrates as expected that the results are dependent on the effect of external support diminishing over time. This is reasonable, as we should expect the anticipation effect to be strong just after the conflict has ended only to be followed by increased uncertainty as the years go by. When time passes relationships fade and the challenger will become less confident that support would be provided once again. Furthermore, we should also expect that renewed support should be more forthcoming early on when the supporter has recently demonstrated its commitment. The decay functions capture both the increasing uncertainty that support will be provided again (anticipation) and the decreasing probability that support actually will be provided again (facilitation).

Fifth, parametric survival models based on the Weibull distribution or discrete-time alternatives such as logit (accounting for temporal dependence) were calculated instead of Cox models. All results are robust to these alternative model specifications as well. In sum, the extended analysis demonstrates that the findings are robust to a battery of alternative model specifications. It points to the effect of external support being strongest in the immediate aftermath of conflict and suggest that rebels rarely receive actual support after the conflict ends, but are adept at correctly anticipating that it will be provided again upon the resumption of conflict.

Conclusion

Why do some armed conflicts recur while others do not? This study has presented an argument regarding how external support provided by states can alter the previous combatants’ calculations as well as their opportunities to remobilize. The legacy of outside support creates a belief in renewed support that makes it more attractive for a revisionist actor to challenge the established status quo. Outside states can also directly facilitate remobilization. The findings presented in this article suggest that external state support to rebel movements during the previous war increases the risk of conflict recurrence in the short term as groups come to expect renewed assistance. In addition, it has been demonstrated that to reinforce the confidence that support would be offered again in the future it is more important to have had persistent support over time during the previous conflict than access to multiple supporters. Conversely, external support provided to governments does not impact the probability of conflict recurrence.

These findings contribute to existing scholarship on conflict recurrence and biased third-party interventions. Although previous research has noted the role of third-party interventions in the form of peacekeeping operations, little attention has been paid to how unilateral support provided by external states has the potential to make conflict recurrence more likely. As such, it adds to our knowledge about why some armed conflicts that have ended relapse into violence while others do not. Furthermore, the findings add yet another dimension to a growing body of literature that demonstrates how external support provision negatively influences conflict dynamics (see for instance Cunningham, 2006, 2010; Gleditsch, Salehyan & Schultz, 2008; Aydin & Regan, 2012; Colaresi, 2014; Salehyan, Siroky & Wood, 2014).

Policies directed to reduce the risk of conflict recurrence need to pay attention to the involvement of external states in domestic conflicts. In general, the international community should act with forceful measures in order to reduce the amount of external support provided to belligerents in general and to rebel groups in particular. Naming and shaming by state governments and international bodies with regard to state sponsorship are not enough. Preventive policies should be taken to hinder the provision of external support before weapons have been delivered or military advisers have been sent to conflict zones. Future research should further relax the distinction between domestic and international politics and delve deeper into the many linkages that exist between former combatants and external actors. In addition, we need to explore whether the type or amount of external support matters. This would make it possible to better understand how external actors can come to impact positively or negatively on post-conflict societies seeking to establish a durable and lasting peace.

Footnotes

Replication data

The dataset and do-files for the empirical analysis in this article, as well as the Online appendix, can be found at . All statistical analyses were conducted using Stata 13.1.

Acknowledgements

I would like to thank associate editor Ragnhild Nordås and the anonymous reviewers for excellent suggestions on how to improve this article. In addition, I would like to express my thanks to Kristine Eck, Desirée Nilsson, David Larsson, Erin Jenne, Hanne Fjelde, Magnus Öberg, Karin Johansson, Nynke Salverda, Nina von Uexkull, Colin Walch, Sophia Hatz, Johanna Ohlsson, Lars Wikman, and participants at the research seminar at the Department of Peace and Conflict Research, Uppsala University for their valuable feedback.

Notes

References

Achen

Christopher H

(2005) Let’s put garbage-can regressions and garbage-can probits where they belong. Conflict Management and Peace Science 22(4): 327–339.

Asmus

Ronald

(2010) A Little War That Shook the World: Georgia, Russia, and the Future of the West. New York: Palgrave Macmillan.

Aydin

Aysegul

(2010) Where do states go? Strategy in civil war intervention. Conflict Management and Peace Science 27(1): 47–66.

Aydin

Aysegul

Regan

Patrick M

(2012) Networks of third-party interveners and civil war duration. European Journal of International Relations 18(3): 573–597.

Box-Steffensmeier

Janet M

Jones

Bradford S

(2004) Event History Modeling: A Guide for Social Scientists. New York: Cambridge University Press.

Byman

Daniel

(2005) Deadly Connections: States That Sponsor Terrorism. New York: Cambridge University Press.

Call

Charles T

(2012) Why Peace Fails: The Causes and Prevention of Civil War Recurrence. Washington, DC: Georgetown University Press.

Colaresi

Michael

(2014) With friends like these, who needs democracy? The effect of transnational support from rivals on post-conflict democratization. Journal of Peace Research 51(1): 65–79.

Collier

Paul

Hoeffler

Anke

Rohner

Dominic

(2009) Beyond greed and grievance: Feasibility and civil war. Oxford Economic Papers 61(1): 1–27.

10.

Cunningham

David E

(2006) Veto players and civil war duration. American Journal of Political Science 50(4): 875–892.

11.

Cunningham

David E

(2010) Blocking resolution: How external states can prolong civil wars. Journal of Peace Research 47(2): 115–127.

12.

Cunningham

David E

(2011) Barriers to Peace in Civil War. New York: Cambridge University Press.

13.

DeRouen

Karl

Bercovitch

Jacob

(2008) Enduring internal rivalries: A new framework for the study of civil war. Journal of Peace Research 45(1): 55–74.

14.

DeRouen

Karl

Lea

Jenna

Wallensteen

Peter

(2009) The duration of civil war peace agreements. Conflict Management and Peace Science 26(4): 367–387.

15.

Doyle

Michael W

Sambanis

Nicholas

(2000) International peacebuilding: A theoretical and quantitative analysis. American Political Science Review 94(4): 779–801.

16.

Fearon

James D

Laitin

David D

(2003) Ethnicity, insurgency, and civil war. American Political Science Review 97(1): 75–90.

17.

Flores

Thomas E

Nooruddin

Irfan

(2012) The effect of elections on postconflict peace and reconstruction. Journal of Politics 74(2): 558–570.

18.

Fortna

Virginia P

(2004) Does peacekeeping keep peace? International intervention and the duration of peace after civil war. International Studies Quarterly 48(2): 269–292.

19.

Gent

Stephen E

(2008) Going in when it counts: Military intervention and the outcome of civil conflicts. International Studies Quarterly 52(4): 713–735.

20.

Gilligan

Michael J

Sergenti

Ernest J

(2008) Do UN interventions cause peace? Using matching to improve causal inference. Quarterly Journal of Political Science 3(2): 89–122.

21.

Gleditsch

Kristian Skrede

Salehyan

Idean

Schultz

Kenneth

(2008) Fighting at home, fighting abroad: How civil wars lead to international disputes. Journal of Conflict Resolution 52(4): 479–506.

22.

Gleditsch

Nils Petter

Wallensteen

Peter

Eriksson

Mikael

Sollenberg

Margareta

Strand

Håvard

(2002) Armed conflict 1946–2001: A new dataset. Journal of Peace Research 39(5): 615–637.

23.

Gregory

Shaun

(2000) The French military in Africa: Past and present. African Affairs 99(396): 435–448.

24.

Haer

Roos

Böhmelt

Tobias

(2016) Child soldiers as time bombs? Adolescents’ participation in rebel groups and the recurrence of armed conflict. European Journal of International Relations 22(2): 408–436.

25.

Harbom

Lotta

Melander

Erik

Wallensteen

Peter

(2008) Dyadic dimensions of armed conflict, 1946–2007. Journal of Peace Research 45(5): 697–710.

26.

Hartzell

Caroline

Hoddie

Matthew

(2003) Institutionalizing peace: Power sharing and post-civil war conflict management. American Journal of Political Science 47(2): 318–332.

27.

Hazen

Jennifer

(2013) What Rebels Want: Resources and Supply Networks in Wartime. Ithaca, NY: Cornell University Press.

28.

Hegre

Håvard

Nygård

Håvard Mokleiv

(2015) Governance and conflict relapse. Journal of Conflict Resolution 59(6): 984–1016.

29.

Daniel E

Imai

Kosuke

King

Gary

Stuart

Elizabeth A

(2007) Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Political Analysis 15(3): 199–236.

30.

Högbladh

Stina

Pettersson

Therése

Themnér

Lotta

(2011) External support in armed conflict 1975–2009: Presenting new data. 52nd Annual International Studies Association Convention. Montreal.

31.

Jenne

Erin K

(2007) Ethnic Bargaining: The Paradox of Minority Empowerment. Ithaca, NY: Cornell University Press.

32.

Kathman

Jacob D

(2011) Civil war diffusion and regional motivations for intervention. Journal of Conflict Resolution 55(6): 847–876.

33.

Kreutz

Joakim

(2010) How and when armed conflicts end: Introducing the UCDP Conflict Termination dataset. Journal of Peace Research 47(2): 243–250.

34.

Kreutz

Joakim

(2012) Dismantling the Conflict Trap: Essays on Civil War Resolution and Relapse. Uppsala: Uppsala University, Department of Peace and Conflict Research.

35.

Kuperman

Alan J

(2008) The moral hazard of humanitarian intervention: Lessons from the Balkans. International Studies Quarterly 52(1): 49–80.

36.

Lacina

Bethany

(2006) Explaining the severity of civil wars. Journal of Conflict Resolution 50(2): 276–289.

37.

Licht

Amanda A

(2011) Change comes with time: Substantive interpretation of nonproportional hazards in event history analysis. Political Analysis 19(2): 227–243.

38.

Licklider

Roy

(1995) The consequences of negotiated settlements in civil wars, 1945–1993. American Political Science Review 89(3): 681–690.

39.

Lyall

Jason

Wilson

Isaiah

(2009) Rage against the machines: Explaining outcomes in counterinsurgency wars. International Organization 63(1): 67–106.

40.

Mason

T David

Gurses

Mehmet

Brandt

Patrick T

Quinn

Jason Michael

(2011) When civil wars recur: Conditions for durable peace after civil wars. International Studies Perspectives 12(2): 171–189.

41.

Mattes

Michaela

Savun

Burcu

(2010) Information, agreement design, and the durability of civil war settlements. American Journal of Political Science 54(2): 511–524.

42.

Mullenbach

Mark J

(2013) Third-party peacekeeping in intrastate disputes, 1946–2012: A new dataset. Midsouth Political Science Review 14(1): 103–133.

43.

Nilsson

Desirée

(2012) Anchoring the peace: Civil society actors in peace accords and durable peace. International Interactions 38(2): 243–266.

44.

Nilsson

Desirée

Kovacs

Mimmi Söderberg

(2011) Revisiting an elusive concept: A review of the debate on spoilers in peace processes. International Studies Review 13(4): 606–626.

45.

Pickering

Jeffrey

Kisangani

Emizet F

(2006) Political, economic, and social consequences of foreign military intervention. Political Research Quarterly 59(3): 363–376.

46.

Quinn

Jason M

David Mason

Gurses

Mehmet

(2007) Sustaining the peace: Determinants of civil war recurrence. International Interactions 33(2): 167–193.

47.

Ray

James Lee

(2003) Explaining interstate conflict and war: What should be controlled for? Conflict Management and Peace Science 20(2): 1–31.

48.

Regan

Patrick M

(2002) Third-party interventions and the duration of intrastate conflicts. Journal of Conflict Resolution 46(1): 55–73.

49.

Regan

Patrick M

Scott Meachum

(2014) Data on interventions during periods of political instability. Journal of Peace Research 51(1): 127–135.

50.

Rosenbaum

Paul R

Rubin

Donald B

(1983) The central role of the propensity score in observational studies for causal effects. Biometrika 70(1): 41–55.

51.

Saideman

Stephen M

(2002) Discrimination in international relations: Analyzing external support for ethnic groups. Journal of Peace Research 39(1): 27–50.

52.

Salehyan

Idean

Gleditsch

Kristian Skrede

Cunningham

David E

(2011) Explaining external support for insurgent groups. International Organization 65(4): 709–744.

53.

Salehyan

Idean

Siroky

David

Wood

Reed M

(2014) External rebel sponsorship and civilian abuse: A principal–agent analysis of wartime atrocities. International Organization 68(3): 633–661.

54.

Sambanis

Nicholas

(2000) Partition as a solution to ethnic war: An empirical critique of the theoretical literature. World Politics 52(4): 437–483.

55.

Stojek

Szymon M

Chacha

Mwita

(2015) Adding trade to the equation: Multilevel modeling of biased civil war interventions. Journal of Peace Research 52(2): 228–242.

56.

Sullivan

Patricia L

Karreth

Johannes

(2014) The conditional impact of military intervention on internal armed conflict outcomes. Conflict Management and Peace Science 32(3): 269–288.

57.

Themnér

Anders

(2011) Violence in Post-Conflict Societies: Remarginalization, Remobilizers and Relationships. New York: Routledge.

58.

Toft

Monica D

(2010) Ending civil wars: A case for rebel victory? International Security 34(4): 7–36.

59.

Walter

Barbara F

(2002) Committing to Peace: The Successful Settlement of Civil Wars. Princeton, NJ: Princeton University Press.

60.

Walter

Barbara F

(2004) Does conflict beget conflict? Explaining recurring civil war. Journal of Peace Research 41(3): 371–388.

61.

Walter

Barbara F

(2015) Why bad governance leads to repeat civil war. Journal of Conflict Resolution 59(7): 1242–1272.

62.

Zahar

Marie-Joëlle

(2006) Political violence in peace processes: Voice, exit, and loyalty in the post-accord period. In: Darby

John

(ed.) Violence and Reconstruction. Notre Dame, IN: University of Notre Dame Press, 33–51.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB