Abstract
Why do some civil wars last longer than others? For instance, why did civil wars in Bolivia in 1967 and Congo/Zaire in 1967 and 1977 end after only a few months, whereas civil wars in Columbia, Somalia, and Southern Sudan raged on for decades without resolution? Given the extraordinarily high material, human, and emotional costs of civil war, the inability of combatants to locate a mutually agreeable settlement that avoids these costs stands out as an interesting and important theoretical and empirical puzzle worthy of investigation. Moreover, from a policy perspective, identifying the conditions that make some conflicts particularly intractable is the critical first step in designing more effective peace-building strategies to end them—strategies desperately needed in contemporary civil conflicts like Iraq today.
It is this basic puzzle that David Cunningham tackles in his book Barriers to Peace in Civil War. His answer is clear and straightforward: Variation in the duration civil war can be (partially) explained by the number of separate factions with distinct preferences over the division of the issues at stake and separate abilities to block an end to the war—actors Cunningham labels “veto players.” Drawing on both the bargaining model of war and well-developed theories of veto bargaining in domestic politics, Cunningham convincingly argues that when civil wars are long, it is often because multiple shifting combatants are involved in the fighting. In comparison to two-party wars, combatants in multiparty conflicts have greater difficulty locating a range of settlements that all sides prefer to continued fighting, a problem that is further compounded by each party’s incentive to hold out and be the last signer of the peace agreement. Based on these mechanisms, Cunningham derives a series of hypotheses, with the central proposition being the more veto players there are, the longer the duration of civil war will be.
In the empirical section of the book, Cunningham operationalizes his three necessary conditions for an actor to be a veto player: autonomous preferences, cohesiveness, and viability. He then proceeds to test the four observable implications of his theory quantitatively using a new and exhaustively constructed data set in which he measures—using indicators such as fractionalization of the group, troop count, operational terrain, and more—the number and strength of every veto player for every civil conflict in the 60-year period following World War II. In total, Cunningham identifies 304 potential veto players that were active across 18,309 conflict months. Using these data, Cunningham employs event history analysis (also known a survival or hazard models) to test his main hypothesis, and he finds robust support for his argument. After controlling for underlying differences across civil conflicts, Cunningham finds strong evidence that greater numbers of veto players in a civil conflict are associated with longer civil wars on average. Substantively, his results indicate that adding one additional veto player can reduce the likelihood that a civil war will end at any given moment by a factor of four. However, beyond this main expectation, Cunningham finds comparatively less support for his other three hypotheses dealing with the diversity of preferences and strength of veto players.
To supplement the findings from the statistical analyses in chapter 3, Cunningham conducts an excellent comparative analysis tracing the negotiation process in civil wars in Rwanda and Burundi. The former was a two-party conflict waged between the government and one rebel group that lasted 46 months, whereas the latter involved at least four combatant veto groups and lasted approximately 17 years. This detailed approach allows Cunningham to determine whether his theoretical mechanisms provide a plausible explanation for the correlations found in the prior chapter. And indeed the historical record of the negotiations does appear to be remarkably consistent with Cunningham’s mechanisms. The negotiations during the civil war in Burundi appeared to labor on while multiple groups held out to extract greater concessions as the last signer. Meanwhile, informational problems caused by shifting alliances made it difficult for parties to accurately estimate the capabilities and resolve of their opponent. Such problems were comparatively less acute during the two-party conflict of the Rwandan Civil War, allowing parties to arrive a negotiated settlement less than a year after coming to the bargaining table.
However, the comparative case study in chapter 4 also raises an important challenge for Cunningham. At the same time that the cases provide support for the veto player approach, they also illustrate that the duration of civil war is only part of the story. As a two-party conflict, the civil war in Rwanda was certainly shorter than the multiparty conflict in Burundi, but this peace agreement ultimately failed in horrific fashion with the start of the Rwandan Genocide only 9 months later. Are two-party conflicts associated with shorter spells of peace after a civil war because they arrive at less durable peace agreements faster, agreements that are ultimately more prone to failure because they tend to neglect important groups that multiparty conflicts struggle to incorporate beforehand? The answer appears to be no. In chapter 5, Cunningham estimates the relationship between his veto player measure and the duration of peace, the likelihood of genocide, and conflict severity. In general, conflicts with more veto player tend to be bloodier, more likely to end in genocide, and more likely to result in a fragile peace than two-party conflicts. These tests are only some of the many examples where Cunningham thoroughly anticipates a methodological or normative critique and constructs an additional test with alternative specifications to provide the reader with the extra evidence needed.
What should policy makers take away from this book? In chapter 6, Cunningham makes a series of policy recommendations for designing more effective peace processes in multiparty settings. Based on his findings, Cunningham’s prescriptions for the international community can be summarized as follows: (a) Negotiations should include all the veto players who can use violence to undermine the settlement and no unarmed groups, (b) the issues under negotiation should be broad enough to include all key incompatibleness without including additional issues that can slow down the settlement process, and, finally, (c) the international community should avoid the use of public goods to induce cooperation, as such provisions create the perverse incentive to negotiate harder to collect a bigger share of the postconflict pie.
Overall, this is an excellent book that makes a significant contribution to the scientific study of civil war. Prior to this work, our best understanding of how the number of participants in a conflict affects the duration of civil war was based on empirical tests conducted by Doyle and Sambanis. Yet these studies focused broadly on several different determinants of peace rather than precisely defining the criteria for a veto player and then deductively deriving a set of hypotheses about how they interact with the conflict bargaining process. In short, previous work lacked the same care and theoretical coherence that marks Cunningham’s work on this topic. In addition, the data Cunningham generates to test his hypotheses should be counted as a major achievement for their contribution to future research.
Of course, even after Cunningham’s exhaustive work, there are still some questions that remain unanswered by the book. For example, theoretically, it is not entirely clear how the hold-out problem is solved when the incentive to be the last signer is so strong. Indeed, Cunningham himself notes that “since all combatants are aware of this benefit, they all have an incentive to hold out and either refuse to negotiate or negotiate harder.” Yet, if we can expect combatants to pool on a strategy of holding out because it guarantees them far greater concessions, it is unclear how this problem ever gets resolved in order for multiparty conflicts to ultimately end.
This raises a second and related theoretical point that speaks to Cunningham’s policy prescriptions: Combatants in a civil conflict have the incentive to bluff along all three of the dimensions necessary for being a veto player (autonomous preferences, cohesiveness, and viability). This makes assessing the number of veto players extremely difficult for the combatants engaged in the conflict, let alone the analyst studying the war. As a result, Cunningham’s policy recommendations may seem unsatisfying, as it is precisely uncertainty over which groups and issues to incorporate into the settlement that opponents must often resolve on the battlefield (where information is less manipulable) rather than the bargaining table (which is a particularly “noisy” since talk is cheap).
Empirically, the main challenge for Cunningham’s results is causal identification. Although Cunningham is careful to control for many factors correlated with the duration of civil war—factors that could make the relationship between the number veto players and the duration of civil war spurious—it is unlikely he has controlled for all of the relevant differences across these conflicts. As a result, he almost certainly overestimates the substantive effects of his veto player measure. Moreover, the number of veto players in a conflict is likely to be endogenous to (expectations over) the duration of a civil war. Similar to how individuals may elect to join a class-action lawsuit based, in part, on whether they expect the lawsuit to be long and costly for the defendant, previously latent groups are probably increasingly likely to take the costly action needed to form collectively cohesive and viable veto players as the prospects for a long drawn-out war increases. This does not necessarily preclude Cunningham’s mechanisms from also being correct, but it does mean the inferences drawn from the results should be interpreted with some caution.
Fortunately, Cunningham acknowledges virtually all of these issues himself in the book, pointing out when and where the reader should interpret results with caution. Indeed, Cunningham’s careful and modest discussion of the results is refreshing, as no research design is ever perfect. Yet if the standard for good social science is whether a theory and empirics can withstand the criticisms of the larger scientific community and still remain compelling, then readers of Barriers to Peace in Civil War will be hard pressed to reject Cunningham’s central conclusion.