Transitions from Civil War

Since 1940, 71 percent of all civil wars have ended on the battlefield while only 29 percent have ended in negotiated settlements. ¹ This low rate of settlement has led many scholars and policymakers to assume that internal wars are, for the most part, irreconcilable. But a close examination of all civil wars fought between 1940 and 1992 reveals that although civil wars have a particularly low rate of successful settlement, they have a remarkably high rate of attempted negotiation. Civil war adversaries did fight to the death in the vast majority of cases, but in 59 percent of these they initiated serious peace negotiations, and in 76 percent of these attempts they signed a peace agreement.

Communists and Nationalists in Greece, for example, held formal talks in 1945 and eventually signed the Varkiza Agreement. The three rivals in Laos met eight times between 1961 and 1972, ultimately signing an agreement in 1973. Uganda's government signed a peace accord and power-sharing agreement with NRA guerrillas in December 1985. Four separate conferences of "National Reconciliation" were held between the government in Chad and the guerrillas, two of which ended in signed settlements: the Kano Agreement of March 1979 and the Reconciliation Accord of August 1979. Even the Chinese Communists and the Chinese Nationalists met three times between 1938 and 1949; they eventually agreed to a democratic coalition government and a fully integrated army. In short, although civil war adversaries might appear intent on all-or-nothing victories, most of them do seek a negotiated settlement. Why, then, do they so often return to war?

The fact that civil war opponents regularly walk away from potentially beneficial settlements is striking. It becomes even more puzzling when this rate of settlement is compared to the rate of settlement in other types of conflict negotiations. Whereas only 29 percent of all civil wars between 1940 and 1992 ended in successful negotiated settlements, a full 60 percent of all interstate wars did. And while negotiations in civil wars succeed only 47 percent of the time, negotiations between management and labor unions in the United States are settled 85 percent of the time, and 95 percent of all civil suits are settled out of court. ² Not only do other adversaries reach settlement more frequently, but they do so despite less devastating penalties for disagreement. If labor leaders and management fail to reach a settlement, their penalty is a temporary loss of production and wages. If a plaintiff and a defendant fail to settle a suit, they waste only the cost of a trial. But if domestic factions fail to agree, their alternative is a costly, violent war. Given this option, one would expect civil war adversaries to be exceptionally eager to settle. Instead, they are even more inclined to fight.

This article poses three questions. Why do civil war adversaries consistently walk away from possible compromise settlements despite serious negotiations and the high costs of disagreement? Why is this settlement rate so different from other kinds of negotiation? And what does this tell us more generally about the conditions under which cooperation under anarchy will succeed or fail?

In what follows, I argue that domestic adversaries retreat from promising negotiations and repudiate even mutually advantageous terms because they cannot credibly commit to treaties that produce enormous uncertainty in the context of potentially dangerous disarmament. If adversaries truly wish to resolve their wars they must do two things: they must dismantle their separate militaries and they must reestablish a legitimate central government. But once groups dispose of their military forces and hand over valuable assets they become paralyzed to either protect themselves against surprise attack or enforce subsequent terms. Thus even under the very best conditions-conditions where all domestic groups actually prefer to settle-the desire for peace will clash with the realities of implementation, and groups will ultimately choose the safer, more certain option of war.

The fact that rebuilding a state is difficult, however, does not mean it is impossible. Many civil war negotiations would succeed in ending civil wars and designing peaceful transitions if there were some certainty that the participants would be protected during the vulnerable implementation period and there was some certainty that these new contracts could be enforced even after the first party was installed in government. Outside intervention can serve the first purpose by protecting groups as they disarm and demobilize. Outsiders, however, will rarely stay long enough to guarantee that political power will be peacefully transferred over time. Treaties that endure, therefore, must offer competing groups some type of political and military "guarantee" to ensure that they cannot be permanently locked out of power. This, I argue, will require dividing crucial government positions among the competing factions in a form of mutual hostage taking. In short, once groups are reassured that they cannot be exploited by a peace treaty, either in the short or long-term, their promises to abide by the terms should gain credibility and their efforts to find peace should be more successful.

The strength of the theory developed here is threefold. First, it shifts the focus of attention away from the question of why groups might fight civil wars, to the more puzzling question of why they would fight these wars even after they reach settlements. In so doing it offers a different and important conceptual framework for studying civil war resolution. Second, it identifies an alternative reason why civil war negotiations so often break down. The greatest problem civil war opponents encounter is not how to reach an agreement, as so many people have assumed, but how to write an enforceable contract under conditions of extreme risk. Third, it provides a set of analytic tools for predicting when parties will require outside enforcement to facilitate cooperation, and when domestic political solutions will suffice.

The article begins by presenting the central puzzle: given the choice between negotiating a compromise settlement or fighting for hegemony, domestic groups almost always choose to fight despite serious negotiations and despite exceedingly high costs for disagreement. In the second section I consider two classes of explanations: the first claims that groups continue to fight because they do not want to settle, and a second class that claims that groups cannot reach mutually acceptable bargains despite serious efforts. In the third section I then propose an alternative explanation. Belligerents walk away from promising negotiations and mutually beneficial agreements because of a two-fold commitment problem. They can neither convince one another that they will resist a surprise attack during the vulnerable demobilization period nor can they convince each other that they will honestly share power once integration is complete. Whether such commitment problems are more or less likely to be resolved depends on the strength of outside security guarantees and whether groups can be guaranteed an effective voice in the new government. In the final section of the article, I discuss what implications this theory might have for scholars interested in the general problem of cooperation and for policymakers interested in the more specific problem of persistent or recurring civil war.

The Puzzle

A survey of all civil wars between 1940 and 1992 reveals that adversaries involved in these wars negotiate far more frequently than is commonly assumed but then consistently abandon these talks in favor of decisive battlefield victories. Table 1 lists these wars, records whether or not combatants engaged in serious negotiations, whether or not a peace settlement was signed, and notes how they were eventually resolved.

Conflicts were classified as civil wars based on the coding criteria proposed by J. David Singer and Melvin Small's Correlates of War project. ³

I coded war as having led to "serious negotiations" if factions held face-to-face talks and issues relevant to resolving the war were discussed. These qualifications eliminated scheduled talks that never took place, meetings where no substantive issues were discussed and talks that excluded key participants. I also attempted to apply a "good faith" proviso and exclude those meetings where one or both participants did not appear sincere. Although sometimes difficult to determine, certain actions did signal whether or not faction leaders honestly wished to cooperate. Their readiness to accept supervision, make public announcements of important concessions, discuss the details of a transfer of power, and participate in lengthy negotiations all generated costs to the groups involved and indicated more than a tactical interest in appearances. Guerrilla leaders, for example, often put their lives in danger when they emerged to talk and thus did not easily agree to face-to-face meetings. "You come out only when you have a message," Jonas Savimbi told a reporter when asked if he was serious about Angola's most recent round of peace talks. "When you don't, you don't come out." ⁴ These qualifications helped exclude bogus meetings where delegates refused to speak with each other, set ridiculous pre-conditions, or simply stalled for time.

Table 1: Civil Wars Ending Between 1940 and 1992

Civil War Negotiations Signed Settlement Outcome

1. China (1946-49) Yes Yes Decisive victory

2. Greece (1944-49) Yes Yes^a Decisive victory

3. Costa Rica (1948) No No Decisive victory

4. Colombia (1948-58)^b Yes Yes Settlement succeeded

5. Burma (1948-51) No No Decisive victory

6. Indonesia (1950) No No Decisive victory

7. Philippines (1950-52) No No Decisive victory

8. Indonesia (1956-60) No No Decisive victory

9. Cuba (1958-59) No No Decisive victory

10. Vietnam (1960-75) Yes No Decisive victory

11. Congo (1960-65) No No Decisive victory

12. Laos (1960-75) Yes Yes Decisive victory

13. Yemen (1962-70) Yes Yes^c Settlement succeeded

14. Sudan (1963-72) Yes Yes Settlement succeeded

15. Rwanda (1963-64) No No Decisive victory

16. Dom. Rep. (1965) Yes Yes^a Settlement succeeded

17. Nigeria (1967-70) Yes No^d Decisive victory

18. Cambodia (1970-75) No No Decisive victory

19. Jordan (1970) Yes No Decisive victory

20. Pakistan (1971) No No Decisive victory

21. Sri Lanka (1971) No No Decisive victory

22. Rhodesia (1972-79) Yes Yes Settlement succeeded

23. Iran (1978-79) No No Decisive victory

24. Nicaragua (1978-79) Yes No Decisive victory

25. Uganda (1981-87)^e Yes Yes Decisive victory

26. Chad (1979-87) Yes Yes Decisive victory

27. Nicaragua (1981-89) Yes Yes Settlement succeeded

28. El Salvador (1979-92) Yes Yes Settlement succeeded

29. Mozambique (1980-92) Yes Yes Settlement succeeded

^a Although peace settlements were signed in both Greece and the Dominican Republic, these were the only two settlements not to include specific power-sharing arrangements.
^b Singer and Small broke both the war in Colombia and the war in Laos into two phases.
^c Yemen was the only case where the factions bargained for and agreed to very specific political and military arrangements but did not record the agreement in written form.
^d Although Biafra did not accept the Nigerian government's offer of a federal solution, they also did not flatly reject it. Instead, negotiations stalled on the questions of "maximum guarantees of security." For a thorough description of these negotiations, see John J. Stremlau, The International Politics of the Nigerian Civil War: 1967-1970. (Princeton: Princeton University Press, 1977).
^e The civil wars in Uganda, Chad, Nicaragua, El Salvador, and Mozambique were not included in the Singer and Small data set since they occurred after their study ended in 1980.

Civil war negotiations were then distinguished by participants' ability to reach and sign settlements. If the combatants signed an agreement that attempted to address each other's war aims, that conflict was coded as having led to a "signed settlement." ⁵ Settlements that only included terms for a ceasefire, the withdrawal of foreign troops, or amnesty for combatants were not considered "settlements" since they did not attempt to deal with difficult issues such as governance or self-determination. These agreements could better be described as temporary measures to halt fighting, allow foreign states to exit gracefully, or present a coup de grace to losing parties. A truer measure of whether or not groups could reach mutually acceptable solution. In this way, I distinguished between those negotiations that broke down due to irresolvable conflicts of interest, and those that could at least locate an alternative middle ground.

Civil wars were then classified according to outcome. Did the war end with a one-sided military victory, or with some form of successful compromise solution? The end was coded as a "decisive victory" if one side could convince its opponent(s) to cease fighting without their demanding any major concessions in return.

Although it is fairly common for even decisive military victories to end with some form of "negotiated" agreement, it is important to distinguish between treaties negotiated by groups who could continue military resistance and those treaties imposed after one side had already won the war. Wars were coded as ending in "successful settlement," therefore, only when three criteria were met. First, a treaty had to be jointly drafted by all combatants through give-and-take bargaining. ⁶ Second, the agreement had to keep the opposition intact as a bargaining entity. Third, it had to end the war for at least five years. ⁷ If a formal peace treaty was signed but broke down within this time period, it was considered a failed attempt, and the outcome in these cases was coded on the basis of the eventual military results.

In all, there were twenty-nine cases of full-scale civil war between 1940 and 1992. Of those twenty-nine, seventeen led to serious negotiations; in thirteen of these, the two sides actually signed a formal political settlement. In short, despite all the possible problems involved with initiating peace talks and resolving underlying conflicts of interest, the rivals in 59 percent of all civil wars did attempt to settle, and 76 percent of those cases of attempted settlement did result in a settlement agreement.

But Table 1 also reveals that overall, negotiations-no matter how sincerely intended-only had a 47 percent chance of success. ⁸ Of the seventeen cases that initiated negotiations, only eight resulted in negotiated settlements that lasted at least five years. Even when the two sides actually signed an agreement at the table, they still returned to war almost 40 percent of the time. The government and the rebels signed peace treaties in Uganda (December 1985), Laos (February 1973), Vietnam (January 1973), Jordan (September 1970), Greece (February 1945), China (January-February 1946), and Chad (March 1979 and August 1979), yet never implemented (or only partially implemented) the terms and all broke down within a year. In short, although most factions appear ready to negotiate their way out of costly wars, many still could not avoid them.

Possible Explanations

To date, no study has addressed the specific question of why combatants in civil wars so often walk away from promising negotiations. Instead, most of the literature has assumed that their violent endings indicated a general unwillingness to settle. Studies, therefore, simply sought to explain why governments and rebels might want to fight wars per se. They never asked why combatants might continue to fight even though they appeared inclined to settle. Still, there are a number of possible explanations circulating in the literature for why warring groups might fail to reach or accept settlements. In general, explanations can be divided into those that claim negotiations fail because combatants want to win the war and are not interested in compromise; and those that claim combatants might wish to reach a compromise but for one reason or another fail.

Combatants Do Not Want to Reach a Settlement

A compelling and popular school argues that civil wars defy settlement because competing factions are not serious about cooperating. In these cases, groups might return to war for a number of reasons. Factions might fight because they think they can win on the battlefield or because they wish to eradicate a hated opponent. They may return to war because they hold inflexible goals or principles. Or they might walk away from negotiations because they are greedy. ⁹

But why then would groups agree to negotiate at all? Four explanations are often offered for why adversaries might nonetheless play along. First, groups could feel pressure from outside patrons or influential neighbors to end a costly and unpopular war. During the Chinese civil war the United States repeatedly threatened to withhold aid from Chiang Kai-shek unless he participated in negotiations and made meaningful concessions. Second, adversaries might use the negotiations tactically as a way to rest and resupply their armies. "The Prime Minister is just playing for time," proclaimed Zimbabwe's rebel faction after bungled negotiations in 1974, "He wants the whole problem to drag on until he reaches his retirement age." ¹⁰ Third, groups might negotiate to test what type of settlement they might obtain and then return to war when they fail to win an pleasing deal. Finally, groups might also have strong incentives to participate in peace talks for reputational reasons. Parties intent on absolute victory must cater to world opinion if they hope to obtain important foreign aid once safely established in power; being known as belligerent could reduce their final reward.

Combatants Cannot Agree on a Deal

A second camp asserts that negotiations often fail, not because combatants do not want to compromise, but because they cannot resolve certain tenacious bargaining problems. ¹¹ The negotiations listed in Table 1, they would argue, had the potential to end successfully but failed because (1) issue indivisibilities, (2) high stakes, (3) asymmetric information, or (4) irreversible commitments stood in their way.

The first, and most obvious, hurdle is the difficult question of how to divide the spoils. "If," as Paul Pillar has written, "the stakes are chiefly indivisible, so that neither side can get most of what it wants without depriving the other of most of what it wants, negotiations are less apt to be successful. Stakes are usually less divisible in civil wars than in other types of war" and this makes settlement less likely. ¹²

But negotiations can also fail as a result of an imperfect bargaining process, as many game theorists point out. Many negotiations collapse, they argue, because combatants aggressively pursue individually rational bargaining strategies that then backfire, leaving everyone worse off.

There are at least three reasons why groups might fail to reach a bargain even though they would prefer to settle. ¹³ The first stems from the value negotiating parties place on ultimate victory. If both the government and the rebels place a high value on winning a war, they are likely to hold out for exceptional offers at the bargaining table. But if both sides simultaneously hold out for such offers, no overlapping bargaining range will emerge and groups could find themselves fighting long after it was mutually rational to settle.

Game theory offers two additional explanations for why bargaining might break down despite mutual gains from cooperation; one based on problems of asymmetric information and one based on problems of commitment. Scholars interested in the dynamics of bargaining have argued that negotiating parties might fail to locate compromise solutions because each withholds important private information of their own relative power. ¹⁴ In the case of civil wars, domestic groups have incentives to prove they can hold out longer than an opponent both during negotiations and during the war. The longer a group can stubbornly stall negotiations or sustain a war, the more likely they are to convince their enemy to capitulate. But misrepresenting ones strength, while individually rational, could have the unintended consequence of motivating groups to fight far longer than they would if this information were public.

Groups might also withhold important private information from each other if this knowledge could later be used against them. Rebels, for example, might be reluctant to reveal their true strength because this would require bringing guerrilla forces out of hiding, divulging secret weapons depots, or disclosing strategic weaknesses. This information might facilitate settlement, but would simultaneously leave them vulnerable to attack. Given this choice, groups might prefer to jealously guard their secrets, even if this might generate a long drawn-out and apparently irrational battle. ¹⁵

Finally, adversaries in a civil war might wind up fighting an undesirable war because both sides have made Schelling-like commitments from which they cannot back down. ¹⁶ Leaders of the warring parties, in an attempt to enhance their bargaining leverage, could choose to commit themselves to strong demands which they then guarantee by stirring up popular nationalistic sentiment. This bargaining tactic would allow a leader to credibly announce that "I'd like to make a concession but my followers won't let me." If both leaders, however, use this tactic to commit to incompatible positions from which they cannot subsequently back down, no settlement will result.

On the face of it, each of these arguments seems quite plausible. There are certainly cases where combatants have no desire to negotiate and simply go through the motions because outside pressure, military considerations, curiosity, or reputational concerns encourage them to do so. And there are also cases where adversaries clearly cannot find an overlapping middle ground in which to draw a solution.

Nevertheless, none of these theories fully explains the unusual pattern of civil war negotiations. They clarify why groups might not initiate negotiations and the problems they might encounter during bargaining. However, they offer no explanation for why so many groups would sign settlements when they have no interest in cooperation. ¹⁷ More importantly, they cannot explain why so many settlements fail even after mutually acceptable settlements have been reached. Clearly, many of the cases listed in Table 1 were not permanently waylaid by any of the bargaining problems listed above.

In sum, arguments that view the problem of civil war resolution as a problem of insincerity or a problem of "bargaining" provide some help in understanding why domestic groups might walk away from the negotiating table, but are far from complete explanations for the exceptionally high rate of failure in civil war negotiations. There are simply too many cases (13 out 17 possible cases) where settlements were reached for these explanations to be fully satisfying. I now propose a new approach-one of credible commitment-that addresses this question head-on.

Why Groups Walk Away from Promising Negotiations

In what follows, I argue that even if negotiations proceed smoothly-that is, if problems due to haggling or information never arise-negotiations are still likely to fail because civil war adversaries will be unable to credibly commit to treaties that create two insidious opportunities for post-treaty exploitation.

Structural Requirements

Resolving a civil war is never simply a matter of agreeing to a cease fire. To be successful, civil war settlements must also integrate the previously warring factions into a single state, build a new government capable of accommodating their interests, and create a new non-partisan national military force.This means that domestic adversaries who wish to settle their war without partitioning their country inevitably face two structural requirements that make cooperation difficult. If they agree to settle off the battlefield they must at some point demobilize, disengage, and disarm their separate militaries, and they must then surrender whatever remaining power they have to a single, sovereign government, not necessarily their own. This process of integration, therefore, is quite different from cases where independent states, unions or other autonomous parties negotiate a contract and then continue to face each other as separate, self-sufficient entities.

Both of these eventualities, however, create tempting openings for post-treaty opportunism and both sides know this. Not only do groups become increasingly tempting targets the closer they come to successfully disarming, but they also become increasingly less able to enforce any subsequent terms. Once groups send their soldiers home, hand in their weapons, and surrender occupied regions, no obstacles remain to launch a surprise attack and no obstacles remain to prevent the first group in power from establishing a one-party state.

What's the result? At the very least, by promoting demilitarization under what are essentially conditions of anarchy, civil war peace treaties create security dilemmas in the reverse. As groups begin to disarm, they create an increasingly tense situation. The fewer arms they have, the more vulnerable they feel.

The more vulnerable they feel, the more sensitive they become to possible violations. And the more suspicious they become of each other's behavior, the less likely they are to proceed with implementation. The ultimate challenge facing civil war opponents at the negotiating table, therefore, is not simply how to stop the fighting, but how to design a contract that convinces groups to shed individual defenses and then submit to the rules of a new political game at a time when no government or police force can either protect them or guarantee compliance.

Problems of Commitment

But how do you credibly commit to an implementation process you know will leave your enemy helpless? And how do you commit to a power-sharing agreement when the first group in power has every reason and opportunity to set up a one-party state? In theory, civil war adversaries have at least three ways to solve the first commitment problem and thus move ahead with a peace plan: they can unilaterally enhance their defenses to make a surprise attack less enticing; they can design less risky implementation plans that reduce the opportunity to cheat; or they can convey mutual assurances that they have no hostile intentions and thus create an atmosphere of trust. In the case of civil wars, the danger of demobilization should be clear to both parties early in the negotiations and this knowledge should enable them to design safeguards to neutralize its negative effects; a security dilemma that's so clearly predictable should also be manageable.

In practice, however, none of these strategies are available to adversaries emerging from violent civil wars. Unlike interstate opponents, civil war enemies can not enhance their individual defenses once they sign a treaty. If they choose to hide weapons or withhold elite soldiers from assembly areas in order to shield themselves from attack, their actions would appear malevolent, no matter how non-offensive their intent. "In the waning days of Mozambique's long civil war, as peace was still being brokered," the New York Times reported, "Antonio Timana, a rebel army officer, buried a small cache of grenades here, in case things didn't work out." ¹⁸ Timana most likely buried these weapons to protect himself should peace talks break down, but this certainly was not clear from his actions alone.

Factions might be able to circumvent this problem by installing sophisticated verification and monitoring equipment; this would give them time to rearm and prepare a defense without triggering an unnecessary security dilemma. But in reality even the best intelligence is imperfect and thus unreliable. As long as cheating can cause enormous suffering, as it can in civil wars, it is unlikely that groups will count on their ability to detect an attack to ensure their security. ¹⁹

Civil war factions could rely on a second strategy and design less dangerous implementation periods. If the government and the rebels, for example, fear a one-step advantage, they could design a plan that proceeded in a step-by-step or tit-for-tat fashion Or, if one side enjoyed a preponderance of military power while their opponent enjoyed a preponderance of political support (like the Kuomintang and the Communists in China), they could use these opposing strengths to deter the breakdown of an agreement.

Reducing the danger of implementation, however, would ultimately not solve their underlying problem. Although each of these strategies would make the actual implementation period less dangerous, they would not change the end result. In civil wars, demobilization can be postponed, it can be implemented incrementally and in a reciprocal fashion, but it cannot be avoided. And as long as a threshold exists beyond which unilateral defense is impossible, and both sides know this, there is very little incentive to even begin demobilization. ²⁰

But there is a third way groups may be able to solve their commitment problem. If factions really are serious about peace, as this article suggests, they should be able to signal their peaceful intentions to each other. To be effective, a signal would require that the sender either increase its own vulnerability or make it impossible to unilaterally launch an attack. Either strategy would prove that the group was trustworthy. But effective signals are also exceedingly hazardous signals. Since unilateral disarmament in the aftermath of a civil war would leave the disarming group dangerously open to annihilation even signaling becomes too costly for peace-loving groups to send.

Finally, even if groups were willing to accept this risk, disarmament would be too easy to mimic by more Machiavellian groups to have the desired effect of relaying peaceful intentions. "They can hide anything they want to from us," said an American police officer and mid-level supervisor for the UN in Bosnia. "We're out here filling in forms that say everything looks good, but most of these police are ready to go into combat in a quick minute." ²¹ Weapons are too easy to hide, too easy to buy and too difficult to monitor to be a powerful signal of peace. Negotiating factions, therefore, are damned if they do and damned if they don't. If they agree to disarm, they leave themselves dangerously open to annihilation without necessarily conveying any peaceful intentions, but if they refuse, they trigger the very security dilemma they hoped to avoid.

But what about the second phase of the civil war resolution process? Do groups find it equally difficult to credibly commit to share power in their new government?

Scholars since the Federalist Papers have debated how best to reassure minority and opposition groups that their interests would not be exploited by those in power. Most of this literature has emphasized the important role institutions play in motivating compliance and sanctioning rule-breakers. Institutions, they argue, can inject an important degree of predictability into the competitive process. They set rules and boundaries, encourage compliance, and offer a systematic way to punish transgressions. Although authors disagree on the relative merits of presidential versus parliamentary systems, proportional versus majoritarian representation, and centralized versus decentralized governing arrangements, they do agree that institutions are the cornerstone for managing internal relations; if the structure is good, the occupants should fall into line. ²²

A competing school of thought, however, has argued that groups will agree to peacefully transfer power over time if they have a chance to compete in free and fair elections. Proponents of this school argue that institutions, in and of themselves, cannot enforce compliance; groups must first have incentives to play along. According to Adam Przeworski, groups will have incentives to comply under three conditions: (1) they know they will have a chance to win an upcoming elections (2) they know a loss will not significantly hurt them, and (3) they know it would cost more to subvert the system then to accept a role as the opposition. In short, groups will voluntarily submit to an election process because they are both optimistic and protected. They are optimistic that they will have a chance to win sometime in the future, and they are not terrified of defeat. ²³ Thus it is the inherent uncertainty of the process and the promise it hold for future victory that hold the system together.

Policymakers have taken a simplified view of the role of elections and assumed that holding "free and fair" elections should be sufficient to make previously warring factions happy. It was, and still is, assumed that if groups have a fair chance to compete, they should be satisfied with the results. If a leader refuses to participate in an election or rejects its results, as Jonas Savimbi did in Angola's 1992 election, he or she is labeled a "villain of the peace," ²⁴ "a sore loser." ²⁵ Little mention is made as to how effectively such an election could shut out the losing party. "Promoting democracies", whatever their form, was simply "the right thing to do." ²⁶

I argue, however, that both academics and policymakers underestimate the unique difficulties adversaries face when trying to build democratic states in societies emerging from violent civil wars and thus overestimate the ability of institutions and elections to encourage compliance. If it is true that groups who have recently fought violent civil wars fear the possible repercussions of settlement and have difficulty committing to peace, then simply the chance to compete in elections-whether these elections are based on majoritarian or proportional principles and are then backed up by presidential or parliamentary systems- will not be enough. Both of these systems still present far too much uncertainty to reassure highly insecure factions. In fact, they only add to the already high level of fear. These concerns are plainly evident in the speech given by Joshua Nkomo, one of the rebel leaders attempting to negotiate a solution to the civil war in Rhodesia:

Countrymen. . . . You know that a war is raging in our country. How can we, therefore, engage in an academic exercise of drawing up a constitution instead of tackling the problem of eliminating the war and its causes? Suppose on our way to draw that constitution Smith massacres half the African delegation, what defence would that delegation have? None! Secondly, even if the constitution were successfully drawn, what guarantee is there that Ian Smith and his regime would, thereafter, be removed from power and that the constitution could then be implemented? None! Verbal promises of the British Government and their American allies are utter rubbish because we are in this plight through deliberate colonial policies of the British. . . . The forces of Smith are fighting to oppose independence. The British Government would like us to believe that these same forces can be relied upon to guarantee our independence which they are sworn to frustrate and defeat. How can they be considered reliable to guarantee the independence which it is their mission to frustrate? It is our position that once power is passed on to you and guaranteed militarily by the Patriotic Front, then other issues such as the constitution can then be worked out and democratic elections held in conditions of peace. What we want is genuine independence. On that we cannot take chances. ²⁷

Solutions aimed at transforming authoritarian states to democratic ones, or reforming pseudo-democratic states to more open, liberal ones do not necessarily apply to the highly militarized and anarchic conditions that exist during civil war. It is easy to imagine why parties would comply with democratic rules in situations where they might benefit from winning an election in the future, where they will not be maltreated if they happen to lose, and where it would be costly to subvert the system. Under these conditions they have every incentive to cooperate. But what if elections and institutions could promise none of these things? What if the losers of the first election could not count on another opportunity to regain power? What if they were not adequately protected by the state? Most importantly, what if they could easily overturn the system? These are the conditions that characterize countries emerging from violent civil wars.

Groups who attempt to build a democratic system after a civil war face uniquely high hurdles which require their own tailored solutions. For one, government institutions will still be too weak to prevent a rapid grab of power and thus enforce what the opposition can no longer enforce themselves. Most of the wars listed in Table 1 were not democracies prior to their conflict. They had no history of democratic rule, no established judiciary and new institutions were expected to be built by the inaugural party. What may appear "democratic" on the outside, therefore, could easily serve those leaders on the inside. "They say the newspapers are free," articulated Halidou Ouedraogo, president of the Burkina Faso Movement for the Rights of Man, "but a minister can still put a journalist in jail. They say the courts are independent, but there are always pressures from behind the scenes. We still don't have real participation of the people at a grass-roots level." ²⁸

This brings us to the second problem faced by civil war adversaries. New democracies rarely enjoy a strong civic culture able to bolster fragile institutions and serve as a secondary control on misconduct. In fact, war-weary populations often prefer order and economic advancement to liberalization or democracy; they simply want peace. "Look at Rwanda, Burundi, Zaire," said Dr. Aliou Boly, a young business manager in Ouagadougou. "If that's democratization, I'm not for it." ²⁹

Elections pose the third and final problem. Elections are too easy to set up to send a credible signal that one truly intends to share power. The rebel leaders of Zimbabwe's Patriot Front once again pinpoint the importance of this dilemma:

It seems we are called upon to accept to be protected by nothing else but faith. . . . By what faith can an election be regarded as "democratic, free and impartial" if the booth is surrounded at its four corners by Smith's henchmen: his policeman, his soldier, his district Commissioner and his judge, supervised by a British colonial officer. . . . Such an election is neither free nor impartial nor democratic. The election booth is just being used as a booby trap. ³⁰

In states with strong, legitimate institutions, elections can be very important. Since competing parties will be held to any results, agreeing to compete in such elections does show that one intends to faithfully follow through with one's promises. Under these conditions it would be difficult not to. But in weakly institutionalized societies (as in those undergoing civil wars) groups will have greater difficulty signaling a true desire for democracy. Since election results can easily be ignored, even aspiring despots will agree to them. In fact, both pro and anti-democratic forces will have equally strong incentives to push for open elections since elections could usher either side into power.

In retrospect, it becomes clearer why parties who have recently fought a civil war might refuse to participate in elections; might nervously stockpile weapons prior to elections; or might reject election results. If groups are uncertain about their chances of winning an election, or competing in future elections, and understand the incentives to cheat, they have little reason to believe that anyone will honestly share power.

In short, domestic adversaries with little independent ability to enforce a contract cannot be satisfied with the opportunity to compete and still feel certain that their rights and liberties will be protected. Institutions in and of themselves might be effective over the long-run, after rules and practices are institutionalized. Or they might have greater success stabilizing a less volatile situation. But if suspending democracy is relatively easy, then groups will need far more convincing signals that they intend to abide by the rules of the new democratic game than a written constitution and an open election.

Can These Commitment Problems Be Solved?

If the two main obstacles to successful settlement are the fact that vulnerability cannot be avoided and few controls will restrain the first group in power from establishing a one-party state-and this makes commitments unconvincing-what can be done to enhance the prospects for peace?

Credibly Committing to Demobilization

Since domestic groups will always have incentives to cheat on the disarmament phase and will lack both an internal agency to punish defections and the independent capabilities to enforce compliance themselves, some third party will be needed to enforce the treaty. On the surface, a third party provides the force needed to actively punish violations and protect disarming factions and thus can guarantee that even the most helpless faction will not be ambushed. But accepting third-party involvement has an even more important signaling effect on negotiations. Groups who are willing to accept outside enforcement send a credible signal to their negotiating partners that they are serious about peace.

Not all "peacekeeping" forces, however, will be equally effective. International guarantees will only make a difference if domestic groups believe that the intervening state will actively come to their aid

should force be necessary. Promises to intervene, therefore, must themselves be credible. A number of indicators should reveal an outside state's commitment to the peace process. First, a state should have a self-interest in maintaining the peace. Old colonial ties, strategic interests, economic investments, or alliance loyalties will enhance any commitment to intervene and indicate the political will to persevere until a new government can effectively ensure stability. Second, the intervening state should also be able to signal a resolve to use force. The outside power can either station sufficient forces to deter aggression without having to send for additional forces if conflict breaks out (as Syria did in Lebanon), or it can create some type of military trip wire, (as Britain did in Zimbabwe). Outside forces can also be placed at strategically important locations, such as troop assembly areas, borders or munitions sites, and guarantors can have pre-approval from home governments for further action.

Third, the guarantor should be willing to use force if necessary, and its military capabilities should be sufficient to punish whichever side violates the treaty. Any credibility an outside enforcer had will be lost once an opponent realizes it can break the rules at will. This is perhaps one of the greatest problems the UN has encountered in its numerous peacekeeping missions. Since most people know that UN monitors will step aside when implementation goes sour, their involvement cannot play the critical role of reassuring groups as they demobilize. The more certain groups become that peacekeepers cannot or will not punish violations, the less likely they are to continue implementing their treaty.

Finally, groups would also be reassured if they knew they would not be left completely helpless after foreign troops left. Outside states, therefore, can enhance feelings of security by (1) offering minority groups dual citizenship-this could serve as a final "escape hatch" should their condition become desperate; (2) neighboring states can maintain open borders to enable besieged groups to obtain refuge; and (3) outside states can aid the resolution process by not insisting on full disarmament, especially not before the terms of the political contract have been fully carried out. If the credible commitment theory is right, allowing groups to retain at least some of their arms should help groups feel more secure and thus help settlements succeed. ³¹

Credibly Committing to Sharing Power

Although domestic factions will be unable to credibly commit to the disarmament phase without outside intervention, they should be able to convince each other that they honestly intend to share power. If it is true that groups hesitate to invest their assets in a new government because this severely reduces any control they might have on subsequent decision-making and gives the inaugural party every reason to act opportunistically, the solution would be to partition political power among the combatants. This would ensure that each group could continue to have meaningful control over a new government even after they had divested themselves of arms and assets. It would also take much of the uncertainty out of an otherwise highly uncertain process.

There are at least three reasons why "disaggregation" would make the commitment to share power credible. First, domestic groups who are in control of key ministerial positions such as the Ministry of Justice or share control of the Prime Ministership have greater incentives to support a new democratic government and fewer incentives to work to overthrow it. Second, this makes it nearly impossible for any group to set up a one-party state. By allowing each faction to hold crucial parts of government or administer key regions of the country, you also enable them to "capture" and hold the government hostage. Third, these concessions also allow opponents to distinguish early in negotiations which groups are serious about sharing power and which are not. A group intent on setting up a one-party state would never agree to such a guaranteed division of power.

But such political guarantees are not the fully democratic and liberal systems many Westerners envision. On the one hand, they tend to exclude important parties and interests who were not major players in the war. They are inflexible and sometimes highly inefficient. The shared government between Hun Sen and the Royalists in Cambodia was practically paralyzed by in-fighting between the two prime ministers. On the other hand, such a system allows recent enemies to avoid many of the dangers and instabilities new democracies produce and avoid backsliding to a one-party system. ³² In short, internal guarantees of power are necessary to launch democracy in states recovering from civil war, but should be viewed as only a temporary solution to war; no more than a first step in the evolution toward a fully liberal democratic system.

Treaties That Bind: Mozambique's 1992 Peace Agreement

In what follows, I use the recent negotiations in Mozambique to illustrate the dynamics of the credible commitment theory at work. I have chosen Mozambique because so many people believed Frelimo and Renamo would never settle off the battlefield. Their war had been long (the average duration of civil wars listed in Table 1 was thirty-five months, this war lasted seventeen years); it was exceptionally bloody (an estimated nine hundred thousand Mozambicans died, more then three million were driven from their homes, and half the total population of sixteen million faced starvation); ³³ and it was ethnically based-fought between the Makonde in the North and the Shangana of the South. Yet despite predictions to the contrary, negotiations eventually succeeded. On 2 October 1992, the Frelimo government and the Renamo rebels signed a peace treaty ending what had been called "one of the most brutal holocausts against ordinary human beings since World War II." ³⁴

In examining the conditions that led to the successful signing of the Rome Accord, I concentrate on the critical role credible commitments played in convincing groups to sign. I argue that like many other civil wars, most of the talks concentrated on ways to reassure the rebels that they would not be attacked during demobilization and that the Frelimo government was, in fact, serious about sharing control of the state. Talks were not dominated by give-and-take bargaining over ideological differences or positions. Instead, success hinged on three critical issues. Would Frelimo accept a significant UN role in the transition period? Was Frelimo serious about dismantling its one-party state and creating a true democracy? And would both Frelimo and Renamo be able to protect themselves after the UN left? Once Frelimo accepted a large UN military presence and agreed to a system of dual political administration, their promise to sincerely seek peace became credible and signatures soon followed.

Current Explanations for Mozambique's Success

Close observers of the war in Mozambique argue that Renamo and Frelimo settled their war in 1992 for one of four reasons. First, a lengthy military stalemate made it clear to both sides that they could not win a decisive military victory. "This," according to a UN election observer in Mozambique, "effectively moved the conflict from a stalemate to a "hurting stalemate" and convinced the groups to settle. ³⁵ Second, outside aid to both parties had been significantly reduced and this encouraged groups to settle. Once the Cold War ended, support for ideological battle between a Marxist-leaning government and rebels disappeared and so did the sponsors. ³⁶ Third, a growing drought threatened mass starvation and made it increasingly difficult for either side to feed its soldiers and supporters. ³⁷ And finally, outside mediators and observers continuously pushed the rivals to resolve their differences. "The skilled and deeply committed people of Sant'Egidio," argued Chester Crocker, "shaped history through their initial intervention, and as time passed, their efforts created a critical mass of facts and momentum so that there was something for "track one" decision-makers to support." ³⁸ By all accounts, the two factions would have continued their war had it not been for the concentration of so many debilitating factors.

A closer look at the war, however, reveals that each of these variables had been present at intervals throughout the war and all were present years before Frelimo and Renamo decided to settle in 1992. As Ibrahim Msabaha points out, "a mutually hurting stalemate produced an impetus for negotiations at several points in the sixteen-year history of the conflict." ³⁹ Only the final negotiations, however, succeeded. Another author admits that "the Government had known for many years that even with substantial assistance, a military solution was not possible." ⁴⁰ It seemed unlikely, therefore, that the removal of aid suddenly convinced them to settle. And the drought, which many believed pushed both sides to the table, was not a new condition in 1992. In 1983, nine years prior to settlement, approximately one hundred thousand people starved to death, yet that disastrous year did not encourage a similar settlement. ⁴¹ Even mediation was not unique to this final set of peace talks, for a number of countries had tried to arbitrate since 1985. ⁴²

To be fair, one could argue that negotiations succeeded in 1992 because all of these factors converged into one costly bundle. But this still leaves important questions unanswered. If a military stalemate, outside aid, a drought, and mediation convinced groups to settle, why did negotiations drag on for twenty-seven months? More importantly, why did these talks linger two years after Renamo's central demand of the war (a multiparty political system and elections) was met as early as 1990? ⁴³ What was so special about 1992 that finally convinced groups to sign?

The fact that so many of these conditions existed before 1992 challenges the view that successful settlement was made inevitable by the drought, the military stalemate, reduced outside aid and skilled mediation. Instead, credibility problems seem to better explain why Frelimo and Renamo continued to negotiate long after crucial political concessions were made, why both were willing to return to a hopeless war even after this point, and why they did finally settle in October 1992. In fact, the twenty-seven months of negotiations reveal how much of the talks revolved around Renamo's attempt to extract "guarantees" from Frelimo that the ceasefire would in fact be peaceful, and that they would still have "the ability to hold the government to commitments" even after they disarmed. ⁴⁴

Main Grievances Settled

Like other negotiations in Zimbabwe, Nicaragua, El Salvador, and Sudan the underlying issues in Mozambique's war were resolved long before a settlement was signed. Since 1989, Renamo had insisted that their goal was "constitutional reform" and their demands were fairly simple. The rebels wanted a multi-party democracy and they wanted to be recognized as a legitimate political party within this new democracy. ⁴⁵ In December 1990, Frelimo adopted a new constitution that provided for multi-party elections and new political parties, and by January 1992 they agreed to a more specific power-sharing formula based on proportional representation and a directly elected president. ⁴⁶ This settled the main grievances of the war.

Seeking a Credible Commitment

The fact that these issues were resolved but negotiations dragged on for an additional ten months (almost a full two years after the initial concessions were made) meant that the bulk of the peace talks focused on other issues. A review of the successive rounds of negotiations (twenty in all) revealed two important patterns: First, most of the negotiations concentrated on the issue of "guarantees" (as both Renamo and Frelimo called them), and second, the final accords were only implemented after the two leaders agreed to set up a unique "dual administration" and after UN peacekeeping troops arrived on the ground.

The concern over security arose immediately in the first round of negotiations when three topics were discussed. Should a mediator be present? Should a ceasefire begin before, or after, the government discussed specific political and military reforms? And would the UN monitor the transition period between the ceasefire and Mozambique's first elections? The government rejected mediation and insisted that Renamo agree to a cease-fire before any political issues were discussed. Renamo, on the other hand, wanted mediation, demanded an agreement on political reform before any discussion of a ceasefire and insisted on extensive UN involvement in "monitoring and guaranteeing implementation." ⁴⁷ Renamo wanted an elaborate UN operation similar to what had been established in Cambodia between a cease-fire and elections. According to Cameron Hume, U.S. observer to the Rome peace talks, "The choice of this option reflected the depth of Renamo's skepticism that the Frelimo government, operating under a Frelimo constitution, could be trusted to conduct fair, multiparty elections." ⁴⁸ Renamo's leader, Afonso Dhlakama contended that

what is at stake is not a cease-fire, because it would be childish of us to talk about a cease-fire without emphasizing what democracy is all about . . . Now people are talking about a cease-fire. We have no problem with a cease-fire because right now we are defending the need for it. But what will happen 24 hours after a cease-fire is in place? How will Mozambicans live afterwards? Does it mean that once a cease-fire is signed, President Chissano will abolish communal villages? Will he then do away with the People's National Security Service which has been killing Mozambicans under the cover of darkness? Will he abolish all laws [passage indistinct]. ⁴⁹

In short, Renamo did not trust that the government would actually write a democratic constitution and create a multi-party state once Renamo laid down their weapons and they certainly didn't believe they could hold the government to these promises once negotiations concluded. Renamo wanted guarantees that a real democracy would be set up and they saw extensive UN involvement as the only way to obtain this.

Little progress was made until June 1991. By this time President Chissano of Frelimo had, in essence, agreed to all three Renamo demands. The government would accept formal mediation during the talks; would discuss political issues before discussing a ceasefire; and would accept "international monitoring" of a ceasefire and elections. But Dhlakama pressed Chissano further: what specific role would outside monitors have? Would Frelimo's security service (SNAPS) be eliminated before Renamo began demobilization? ⁵⁰ These questions revealed the depth of distrust between the two factions and the real concerns the rebels had for their safety. The government, however, refused to discuss either issue.

Negotiations then moved to the equally contentious item of who would administer the country during the transition. Renamo refused to accept any Frelimo role in conducting the elections fearing the incumbents could easily manipulate the process. Chissano's government, however, refused to allow the UN to register parties or conduct the elections and refused to let Renamo share responsibility. ⁵¹ By August 1991, during the sixth round of negotiations, Renamo broke off talks claiming it would be suicide for Renamo to go forward with a solution that could so easily be manipulated by their opponent. ⁵² When asked what went wrong, one of the mediators replied, that "[t]here was too much distrust between the two sides, and Renamo had a 'deep fear of falling into the trap of integration.'" ⁵³

Despite these concerns, negotiations recommenced in November 1991 and by August 1992 Frelimo finally accepted outside enforcement of the transition period. This, according to the credible commitment theory, was one of two crucial guarantees the settlement would require to succeed. Not only did it reassure the rebels that the government was serious about peace, but it offered critical protection for them during the upcoming demobilization.

Less than a month later, Renamo received the second crucial guarantee. After only two short meetings, President Chissano and President Dhlakama reached a unique "gentleman's agreement." The two leaders agreed that Mozambique's new national army would consist of fifteen thousand Frelimo soldiers and fifteen thousand Renamo soldiers; equal representation from each faction. In addition, the government could retain its security forces, but would now be monitored by a mixed Frelimo/Renamo oversight commission. Most importantly, however, both sides agreed that Renamo could remain in the regions it already occupied prior to elections, while government administrators would be "allowed to establish a presence throughout the country". ⁵⁴ This meant that Renamo would not be forced to relinquish administrative control over home regions before the vote. If Frelimo won at the polls but refused to set up a coalition government, Renamo could retain these regions until Frelimo's promises were fulfilled. Frelimo would never be able to politically dominate the country as long as Renamo occupied these regions, and it would never be able to take these regions by force as long as Renamo controlled half the national army. This, as it turns out, was the second crucial guarantee. Two weeks later, on 4 October 1992, President Chissano and President Dhlakama signed the official Rome Accords ending their war.

What happened during implementation? As the credible commitment theory would have predicted, neither group began to demobilize until UN troops arrived and neither side had disarmed by the time elections took place in October 1994. Full disarmament, however, was not necessary for elections to run peacefully. President Chissano won the presidential election and his party won five of the country's ten provinces. Renamo won a majority in the remaining five provinces, two of which were Mozambique's most populous.

What happened after the election confirmed Renamo's fears. Once Chissano had won at the polls, he refused to form a coalition government with Renamo and refused to include Dhlakama in his cabinet despite strong public and international pressure to do so. This meant that Renamo, although represented in parliament, had little if any official influence in government. Not surprisingly, Renamo then refused to relinquish authority in the five provinces it had won and a double administration became the norm. ⁵⁵

But why would Renamo accept this situation? Why not return to war? As the credible commitment theory would predict, Renamo had little reason to return to war. The rebel group retained enough political power to challenge Frelimo in the next elections, the dual administration protected them from political obsolescence, and Renamo knew that renewed war would be very costly. As Miguel de Brito, a former professor of politics at Mozambique's Institute of International Relations observed,

The last thing Dhlakama will do right now is return to war. Renamo has a lot of strength in the rural areas and enough influence with the international community to make sure the 1999 elections are fair. If they play their cards right, they could do much better next time. ⁵⁶

As long as there were good prospects for success in the next election and they could wait in peace, Renamo had no reason to fight.

In conclusion, the 1990-92 negotiations in Rome are a story of Renamo's attempt to extract credible commitments from Frelimo. Each successive concession by Frelimo served to increasingly reassure Renamo that the government sincerely wished to settle. It was "commitments" rather than the stalemate, the drought, the pressure, or mediation that ultimately convinced the two rivals to sign, implement and maintain their settlement. Moreover it was the willingness of the UN to intervene with significant force, to stay through elections and to not insist on full disarmament that allowed for this success.

Other Cases

The preceding discussion of Mozambique's negotiations illustrate how credible commitments ultimately convinced Frelimo and Renamo to sign an agreement in 1992. A quick review of other wars listed in Table 1 reveals similar patterns in each of the other cases. Six out of the eight cases of successful settlement (75 percent) were underwritten by extensive outside security guarantees. ⁵⁷ Outside powers guaranteed the Dominican Republic's "Act of Dominican Reconciliation", the Addis Ababa Agreement in Sudan (1972), the Lancaster House Agreement in Zimbabwe (1979), the Tela Agreement in Nicaragua (1989), El Salvador's agreement in 1992, and the 1992 Rome Accord for Mozambique, and all brought peace. ⁵⁸ On the other hand, none of the failed agreements enjoyed similar outside security guarantees. China, Greece, Laos, Nigeria, Nicaragua (1978-79), Uganda, and Chad all held serious negotiations, yet none enjoyed outside enforcement. All eventually failed. This seems to indicate that while agreement was necessary for successful settlement, it was not sufficient to produce peace without outside enforcement.

The successful settlements listed in Table 1 also partitioned political power in a number of creative ways. The Conservatives and Liberals in Colombia, for example, agreed to a 50-50 split of all government positions (including patronage jobs) and rotated their time in the presidency. The 1970 political agreement in Yemen created a highly decentralized governing structure based on regional autonomy and integrated the rebel Royalists into every level of government with the incumbent Republicans. In Sudan, a federation was created between the North and the South. Their new constitution guaranteed the continued existence of a southern regional government and gave the South enough tax revenue to survive without help from the wealthier North. More importantly, the accord was able to fashion a national army that preserved the armed strength of both factions. ⁵⁹ The Whites in the new state of Zimbabwe were guaranteed 20 percent of all the seats in the lower House of Parliaments despite the fact that they represented only 3 percent of the population. As a minority, they were also allowed to retain private control of most of Zimbabwe's richest land, and dual citizenship with England. The Sandinistas and Contras in Nicaragua solved this problem by creating twenty-three self-governing development zones (20 percent of the country) for the Contras that they could self-police. It also kept Ortega as Commander-in-Chief of the army even after his opponent won the presidency. El Salvador was somewhat different. Although there was no formal agreement to share governmental power, power was divided in the authoritative state security infrastructure-an area of great concern to the rebel FMLN. According to this deal, the FMLN was guaranteed 20 percent of the slots in the new civilian police force while the existing security forces would be dissolved and the Salvadoran military would be dramatically reduced and reformed. In Mozambique, in addition to retaining administrative control over occupied regions, both factions were guaranteed equal representation in a new national army. All of these guarantees were specifically outlined during negotiations.

In contrast to these detailed arrangements, most of the failed settlements included only vague references to future political arrangements. ⁶⁰ In China, the new coalition government provided only for a "cabinet system" in which the executive branch was responsible to the legislative branch. The Greek Communists and Nationalists only agreed to hold "a plebiscite as soon as possible (in any case during 1945) to decide finally on the question of a regime." ⁶¹ Elections for a new constituent assembly to draft a new constitutions would be held sometime thereafter. In Laos, the three faction leaders created a coalition government, but on that did not weaken General Phoumi's incumbent position.

Although this brief review of the cases offers no definitive evidence for or against the credible commitment theory it does present strong preliminary support for it. Before we can really know what role credible commitments play in the resolution of civil wars, additional research will be required to test all of the alternative hypotheses in an empirically rigorous manner. This will likely involve multivariate analysis of a large civil war dataset and more detailed assessments of a number of case histories. Despite these limitation, I remain confident that additional research will confirm the crucial role credible commitments play in the ultimate success of civil war negotiations.

Possible Challenges

There are, however, a number of possible challenges to the logic of the theory of credible commitment. The first and perhaps most obvious criticism is that outside enforcement is actually endogenous to the model. In other words, one could argue that outside states only offer guarantees in those civil wars where negotiations would have succeeded anyway and do not offer them in cases where success appears unlikely.

There are a number of ways to check whether outside enforcement had an independent effect on adversaries' decision to sign and implement treaties. First, if enforcement is simply offered in cases that would have succeeded on their own, then treaties should succeed whether or not outside forces arrive on the ground. If outside guarantees are truly endogenous to the model, then the fact that security forces ultimately failed to show, should have little effect on success.

When we go back to the cases, however, we find that implementation did not proceed in the absence of peacekeepers. Instead, implementation was delayed until peacekeepers finally arrived, it broke down when peacekeepers failed to show, and it collapsed when peacekeepers decided to leave early. In other words, the arrival or departure of outside enforcement appeared directly correlated to treaty execution. In Chad, the factions successfully formed a government of national unity and began an "orderly withdrawal" of their troops from the capital. But no other terms were implemented when the neutral African peacekeeping force made up of soldiers from Guinea, Benin, and the Congo failed to arrive. ⁶² In Uganda, guerrilla and government leaders signed a peace accord in December 1985 and asked Kenya, Tanzania, the United Kingdom, and Canada to establish a peacekeeping force. The United Kingdom and Canada, however, declined to participate and the terms were never implemented. The peace agreement designed to end the war in Laos signed in February 1973, on the other hand, specifically called for the "withdrawal of all foreign forces within sixty days." The ceasefire never fully took effect and fighting continued until the Pathet Lao decisively defeated the government. In short, the arrival and timing of outside intervention does appear to have a direct effect on whether or not treaties succeed.

Second, if it is true that outside states only intervene in those conflicts that would have succeeded on their own, then it must also be true that outsiders can ascertain which negotiations will succeed and which will fail. One can think of a number of different ways that United States or the UN, for example, might determine this. They could intervene in only those cases where groups have actually drafted or signed settlements. They could target only the longest, most deadly wars knowing that combatants have the greatest incentive to settle in these cases. Outsiders might also wait to see which leaders are willing to begin implementation (and thus will more likely follow through with any promises) before intervening. ⁶³

Contrary to what one might expect, however, outside states did not intervene only in cases where deals were struck. In six of the fourteen cases (43 percent of the cases) combatants had signed peace settlements in hand, yet outsiders did not intervene. On the other hand, outsiders did choose to intervene in cases that appeared unlikely to succeed. Outsiders intervened in both short and long wars; wars with high casualty figures and those with low casualty figures; and they intervened in wars where leaders were highly distrusted.

A number of scholars have also argued that solutions based on power sharing are unnatural, forcing unwilling enemies together in solutions that are bound to fail. ⁶⁴ Instead, they argue that territorial partition or decisive military victories are more stable long-term solutions. This criticism is particularly important for two reasons. First, territorial partition would circumvent many of the credible commitment problems outlined above and should make it far easier for adversaries to successfully settle their wars. Second, history shows that most civil war settlements based on power sharing do eventually lead to renewed war. If this is the case, why can't groups simply partition the state and avoid this vulnerable transition period?

One of the interesting findings to emerge from this study is that territory need not be partitioned among the combatants in order to obtain peace. The type of partition can be political rather than territorial. Thus by dividing the government into crucial positions, groups can circumvent the credibility problem in a far less drastic way than forcibly moving people into separate regions. A second interesting finding is that every civil war settlement that succeeded was based on power sharing, not partition. This is not to say that partition would not be a better option, in many cases it probably would be more likely to bring long-term peace a stability. But partition is almost never an acceptable solution for governments. Although partition should be a viable solution for groups fighting a civil war, governments rarely allow sections of their territory to be lopped off in order to either avoid or shorten civil wars.

Conclusions

This article was a study on how to design effective contracts to end civil wars. It began by developing two major claims. First, the majority of civil war adversaries do attempt to find negotiated agreements to their war. This claim runs directly counter to the conventional view that civil wars are so persistent because rival leaders have no desire to compromise or cannot locate a mutually acceptable middle ground. Second, even when all other obstacles to resolution are resolved, civil war adversaries will still confront a unique set of commitment problems that stem from the need to integrate two separate organizations into a single unit.

This article points out a number of disturbing problems. First, in the absence of outside enforcement, a particularly severe security dilemma will always arise whenever a peaceful settlement is attempted because the general requirements of state rebuilding force groups to enter a strategic game for which there is only one rational outcome: continued war. ⁶⁵ Second, it also points out that only limited solutions are possible. Civil war settlements are most likely to succeed when outside states step in to enforce compliance during the highly unstable implementation period and when key governmental positions are partitioned among the competing factions. When a third party has the political will and the military resolve to act as a policeman during the transition, adversaries will generally push ahead with a peace plan. Likewise, they will continue to have faith in the agreement if each faction can hold the political process hostage. If, however, a third party fails to step forward, or in some way reveals a lack of resolve, these same adversaries will become reluctant to proceed and even signed settlements will collapse.

This article, however, also leads to a number of promising conclusions. If it's true that many negotiations collapse because groups can't commit to the high degree of vulnerability and uncertainty required by a settlement, and not because their conflict is hopelessly irreconcilable, then settlements are not as impossible to attain as many people have argued. In fact, outside intervention can make a significant difference in the rate of successful negotiation, at least in those cases where the groups themselves are eager to settle.

Implications

A number of interesting implications can be drawn from this study. Theoretically, this study has relevance for the growing game-theoretical literature that attempts to explain why bargaining might break down despite mutual gains from cooperation. It shows that if negotiating parties are asked to integrate their assets into a single organization they will have far greater difficulty binding themselves to an agreement than groups who can remain independent or autonomous thereafter. In these cases, their greatest obstacle to success will not be reaching mutually acceptable bargains, but in forging credible commitments to a highly risky implementation process.

This article also speaks to the debate on the efficacy of domestic institutions over outside enforcement. What this study of civil war negotiations shows is that enforcement does matter a great deal when the costs of cheating are exceptionally high and if there is a high risk of opportunism, but that domestic political solutions should suffice when the risks and costs of exploitation are lower. Colombia and Yemen were the only two settlements that succeeded without outside security guarantees, yet they were also the only two cases where the opposing parties could not launch surprise attacks against each other. In both cases, the war was fought by relatively uncommitted armies whose loyalties could be procured by the highest bidder. Thus the strategic setting in which negotiating parties find themselves does greatly affect how easy or difficult it will be to cooperate and determines what type of enforcement and institutions will be needed.

Finally, this study also offers a fascinating comparison to the literature on the transition to democracy, which argues that greater uncertainty enhances compliance. What this study shows is that democratization cannot and should not progress in exactly the same fashion if it occurs under different conditions. High risk situations will require greater certainty.

For the policymaker, this study offers a set of analytic tools for predicting when outside intervention can have the most positive effect and when it will do little good. First, and most importantly, outsiders can ensure that groups will survive the most vulnerable disarmament or demobilization phase. Second, they can refrain from pushing for a "quick and easy" democratization process and instead encourage greater use of internal guarantees. And third, they can allow groups to retain a measure of self-help by not demanding full disarmament and offering important "escape hatches" to groups such as dual citizenship, open borders or generous asylum provisions.

In short, civil war resolution is difficult and it often breaks down. Once the international community understands why peaceful transitions are so difficult to design, then outside assistance can be targeted in ways that effectively help rather than hinder cooperation. The cases of "attempted negotiation" analyzed in this chapter suggest that at least one lesson can be drawn from past experiences. When attempting to resolve civil wars, enforcement does matter. But it only matters in the short-term. If outside states truly wish these settlements to last over time, they must also consider how the institutional parameters of any new government shape groups' expectations about their future security and factor into decisions to fight or cooperate. Military force might be crucial for implementation, but institutional design matters far more in the long run.

Notes

Note 1: This figure is based on data from J. David Singer and Melvin Small's (1982) Correlates of War data on civil wars. their dataset was modified somewhat to include only the most violent cases (< fifteen hundred battle deaths per year) and extended to add those cases be-tween 1982 and 1992. See table 1 in the appendix for the full list of cases. Back.

Note 2: These figures come from Robert B. Wilson "Negotiation with Private Information: Litigation and Strikes," Nancy L. Schwartz Lecture delivered at the J. L. Kellogg School of Management, Northwestern University, May 18, 1994. Back.

Note 3: To be included in the list of civil wars a conflict had to: (1) experience at least fifteen hundred battle deaths per year, (2) occur within a generally recognized boundary, (3) involve the national government as a principal agent, and (4) have effective resistance from both the rebels and the government. I should note that the first criteria is slightly different from Singer and Small's. The Correlates of War dataset includes all conflicts with at least one thousand battle deaths per year. Critics of Singer and Small's dataset have argued that this coding criteria is too broad and, as a result, many small-scale confrontations such as riots or coups are included that most people would agree were not "civil wars." I agree. To account for this criticism, I excluded a number of bor-derline cases from the dataset. Cases with fewer than fifteen hun-dred total battle deaths (such as Paraguay in 1947, Lebanon 1958 and 1976, Guatemala in 1954 and 1970-71, and Indonesia in 1953) or conflicts restricted to a very small percentage of the population (power struggles within the military or a purge within the government) were not included. This, I believe, represents a more realistic list of actual civil wars. For specific criticisms of the COW dataset, see Pillar 1983; Stedman 1991; and Licklider 1993. Back.

Note 4: Bill Keller, "Will Angolan Rebel Give Peace a Chance?," The New York Times (February 14, 1995), p. A1, A6. Back.

Note 5: This was determined based on public statements issued by fac-tion leaders during negotiations. Back.

Note 6: The final product could be written or tacit. In rare cases, such as Yemen (1970), warring factions bargained for and agreed to very specific political and military arrangements without inscribing a written document. Although verbal, these settlements held as much legitimacy and authority as recorded documents. Back.

Note 7: Thus wars that ended after 1992 were not listed since it is too early to tell whether settlements will actually last. The theory pre-sented in this paper, however, should also apply to recent wars in Somalia, Bosnia, Rwanda, Cambodia, Guatemala and Angola. Examples from these conflicts, therefore, have been used to illus-trate certain points. Back.

Note 8: A disagreement rate of 75 percent in dispute resolution is con-sidered very high. A disagreement rate of 47 percent (which we see in civil war negotiations) is almost unheard of. See Wilson for more detailed comparison of different negotiation rates. Back.

Note 9: Stephen Stedman has argued that a leader's paranoia, drive for power, and irrational attachment to principle could make compro-mise solutions impossible. See "Negotiation and Mediation in Internal Conflict," in Michael E. Brown, ed., The International Dimensions of Internal Conflict. (Cambridge: MIT Press, 1996), p. 347. Back.

Note 10: Goswin Baumhoegger, The Struggle for Independence: Docu-ments on the Recent Development of Zimbabwe (1975-1980), Vol. II, p. 7. Back.

Note 11: Much exciting research is being done in the area of strategic barriers to successful negotiation in international relations. See especially, James Fearon, "Rationalist Explanations for War," International Organization. Vol. 49 (1995), pp. 379-414; and David Lake, Entangling Relations: American Foreign Policy in its Century, (Princeton: Princeton University Press, forthcoming 1998). Back.

Note 12: Paul Pillar, Negotiating Peace: War Termination as a Bargain-ing Process. (Princeton: Princeton University Press, 1983) p. 24. See also Fred Ikle, Every War Must End (New York: Columbia University Press, 1971), p. 95; George Modelski's "International Settlement of Internal War" in James Rosenau's International Aspects of Civil Strife. (Princeton: Princeton University Press, 1964); Harrison Wagner, "The Causes of Peace," in Roy Licklider ed., Stopping the Killing: How Civil Wars End. (New York: New York University Press, 1993). Back.

Note 13: See James Fearon "Bargaining over Objects that Influence Future Bargaining Power," University of Chicago. Paper presented at the annual meeting of the American Political Science Association, Washington, D.C., August 1997. Back.

Note 14: See Roger B. Myerson and Mark A. Satterthwaite, "Efficient Mechanisms for Bilateral Trading," Journal of Economic Theory 29. 1983. For applications of this to international relations see James D. Morrow, "Signalling Difficulties with Linkage in Crisis Bargaining," International Studies Quarterly. (June 1992), pp. 153-72; and Fearon, 1995. Back.

Note 15: Ironically, even if groups were willing to divulge this information, the fact that they have incentives to misrepresent such information, and both side realize this, would make even accurate information suspect and therefore not necessarily helpful. Back.

Note 16: See Vince Crawford, Econometrica 1982. My thanks to John McMillan for pointing this out. Back.

Note 17: One could argue that it is actually extremely easy to make concessions under these conditions; empty promises will never be carried out. But public promises are not without cost. Ignoring or violating an agreement could negate most of the benefits one had hoped to gain by appearing cooperative. Not only do factions set a precedent for rapprochement with their enemy, but they concurrently anger outside patrons and secure a reputation as a "spoiler." In short, although it might be understandable why groups would not negotiate in earnest, it is not so clear what incentives they might have to draft a contract they then intend to break. Back.

Note 18: Suzanne Daley, "In Mozambique, Guns for Plowshares and Bicycles," The New York Times, March 2, 1997, p. A3. Back.

Note 19: Barry Weingast develops this theme in "Constructing Trust: The Political and Economic Roots of Ethnic and Regional Conflict," Unpublished paper, Hoover Institute, 1994. Back.

Note 20: This is the problem of backward induction as it is known in the game theoretic literature. Back.

Note 21: Mike O'Connor, "Bosnia's Military Threat: Rival Police", The New York Times, (January 12, 1997) p. A6. Back.

Note 22: For an overview of these debates, see Larry Diamond and Marc F. Plattner, eds., The Global Resurgence of Democracy. (Baltimore: Johns Hopkins Press, 1993). Back.

Note 23: See Przeworski, p. 26-31. Back.

Note 24: Quote taken from "Farewell to Savimbi," The Economist, January 30, 1993, p. 20. Back.

Note 25: See Kenneth B. Noble, "With Truce, Angola Rulers Face Un-certainty," The New York Times, November 8, 1992, p. A10. Back.

Note 26: Speech given by President Clinton at Freedom House on Octo-ber 6, 1995. Back.

Note 27: "Message to Zimbabwe," July 24, 1977, in Baumhoegger, Vol. III, p. 387. Back.

Note 28: John Darnton, "Africa Tries Democracy, Finding Hope and Peril," The New York Times, Tuesday, June 21, 1994, p. A6. Back.

Note 29: Darnton, p. A1. Back.

Note 30: The Patriotic Front's Statement on the British Proposals Adopted at a Meeting of its Co-ordinating Committee in Maputo," September 12, 1977, in Baumhoegger, Vol. III, p. 432-433. Back.

Note 31: This was the logic behind the right to bear arms in the U.S. constitution. An armed citizenry was seen as an excellent check against the concentration of power. Back.

Note 32: For the dangers associated with building democracies see Ed-ward Mansfield and Jack L. Snyder, "Democratization and War," International Security. Vol. 20, no. 1 (Summer 1995). Back.

Note 33: William Finnegan, A Complicated War: The Harrowing of Mozambique. (Berkeley: University of California Press, 1992) p. 4. Back.

Note 34: Deputy Assistant Secretary of State for Africa Roy Stacy, speaking at a donor's conference in Maputo, quoted in Thomas Ohlson, "Strategic Confrontation versus Economic Survival in Southern Africa," in Francis M. Deng and I. William Zartman, eds., Conflict Resolution in Africa. (Washington, D.C.: Brookings Institute, 1991). Back.

Note 35: Robert B. Lloyd, "Mozambique: The Terror of War, the Ten-sions of Peace," Current History. (April 1995) p. 153. For the theoretical underpinnings of this argument see, I. William Zartman, Ripe for Resolution: Conflict and Intervention in Africa. (New York: Oxford University Press, 1985). Back.

Note 36: Eric Berman, Managing Arms in Peace Processes: Mozam-bique. (Geneva: UN Institute for Disarmament Research, 1996) p. 19-20. Back.

Note 37: See Thomas Ohlson and Stephen John Stedman with Robert Davies, The New is Not Yet Born: Conflict Resolution in Southern Africa. (Washington, D.C.: The Brookings Institution, 1994) pp. 113-16. See also Chris Alden and Mark Simpson, "Mozambique: A Delicate Peace," Journal of Modern African Studies, Vol. 31, no. 1 (March 1993), p. 126. Back.

Note 38: Foreword, in Berman, pp. xi-xii. Back.

Note 39: Ibrahim Msabaha, "Negotiating an End to Mozambique's Murderous Rebellion," in I. William Zartman ed., Elusive Peace: Ne-gotiating an End to Civil Wars. (Washington D.C.: The Brookings Institution, 1995) p. 210. Back.

Note 40: Berman, p. 21. Back.

Note 41: Cameron Hume, Ending Mozambique's War: The Role of Mediation and Good Offices. (Washington D.C.: The U.S. Insti-tute of Peace, 1994) p. 10. Back.

Note 42: Allen Isaacman, "Mozambique: Tugging at the Chains of De-pendency," in Gerald J. Bender, James S. Coleman, and Richard Sklar, eds, African Crisis Areas and U.S. Foreign Policy. (Ber-keley: University of California Press, 1985); Helen Kitchen, An-gola, Mozambique and the West. (New York: Praeger, 1987); and Msabaha. Back.

Note 43: See Finnegan, p. 246-247. Back.

Note 44: Quotes taken from Hume, p. 34. Back.

Note 45: Hume, p. 59. Back.

Note 46: Hume, p. 86. Back.

Note 47: Hume, p. 60. Back.

Note 48: Hume, p. 59. Back.

Note 49: August 1992: Voice of Renamo: "Dhlakama says Frelimo Army launching attacks to stop him leaving for summit." Lexis/Nexis. Back.

Note 50: Hume, p. 62. Back.

Note 51: Hume, p. 66. Back.

Note 52: Hume, p. 67. Back.

Note 53: As quoted in Hume, p. 67. Back.

Note 54: Hume, p. 133. Back.

Note 55: Lexis/Nexis 3/9/93. Back.

Note 56: The Christian Science Monitor, November 19, 1996. Back.

Note 57: For a more rigorous empirical examination of these cases as well as the full-set of civil wars between 1940 and 1990, see author. Full cite. Back.

Note 58: Only two civil wars reached successful settlement without an outside guarantee (Colombia in 1958 and Yemen in 1970) yet these were also the only two cases where the opposing parties could not launch surprise attacks on each other. Both wars were fought by relatively uncommitted armies whose loyalties could be procured by the highest bidder and thus did not represent an im-mediate threat to either opponent. Thus it appears as if Colombia and Yemen were exceptions that tended to prove the rule. Back.

Note 59: See Wai 1981, 171. Back.

Note 60: The settlements in Chad and Uganda both included specific political guarantees. In neither of these cases, however, was an outside state willing to enforce the final agreement. Back.

Note 61: From the Varkiza Agreement as outlined in Keesing's Contem-porary Archives, (October 13-20, 1945) p. 7486. Back.

Note 62: Keesing's Contemporary Archives, (February 1980) p. 30067. Back.

Note 63: As noted earlier, this is often difficult to determine since lead-ers often have incentives to appear uncooperative early on in the negotiating process in order to enhance their bargaining leverage. Back.

Note 64: See especially John J. Mearsheimer "The Only Exit From Bos-nia," The New York Times, October 7, 1997, p. A21; John J. Mearsheimer and Robert A. Pape, "The Answer: A Three-way Partition Plan for Bosnia and How the U.S. Can Enforce It," New Republic (June 14, 1993); and Chaim Kauffmann, "Possible and Impossible Solutions to Ethnic Conflict," International Security, Vol. 20, no. 4 (spring 1996). Back.

Note 65: These findings should generalize to any situation where states must be rebuilt under conditions of anarchy. Interestingly, most civil wars can be traced back to some form of institutional failure; some time when groups had to rebuild their government without any central authority to help enforce a new contract. In these cases groups faced the same credibility problems discussed in this paper, albeit at a much lower level of fear. Back.

Civil War	Negotiations	Signed Settlement	Outcome
1. China (1946-49)	Yes	Yes	Decisive victory
2. Greece (1944-49)	Yes	Yes^a	Decisive victory
3. Costa Rica (1948)	No	No	Decisive victory
4. Colombia (1948-58)^b	Yes	Yes	Settlement succeeded
5. Burma (1948-51)	No	No	Decisive victory
6. Indonesia (1950)	No	No	Decisive victory
7. Philippines (1950-52)	No	No	Decisive victory
8. Indonesia (1956-60)	No	No	Decisive victory
9. Cuba (1958-59)	No	No	Decisive victory
10. Vietnam (1960-75)	Yes	No	Decisive victory
11. Congo (1960-65)	No	No	Decisive victory
12. Laos (1960-75)	Yes	Yes	Decisive victory
13. Yemen (1962-70)	Yes	Yes^c	Settlement succeeded
14. Sudan (1963-72)	Yes	Yes	Settlement succeeded
15. Rwanda (1963-64)	No	No	Decisive victory
16. Dom. Rep. (1965)	Yes	Yes^a	Settlement succeeded
17. Nigeria (1967-70)	Yes	No^d	Decisive victory
18. Cambodia (1970-75)	No	No	Decisive victory
19. Jordan (1970)	Yes	No	Decisive victory
20. Pakistan (1971)	No	No	Decisive victory
21. Sri Lanka (1971)	No	No	Decisive victory
22. Rhodesia (1972-79)	Yes	Yes	Settlement succeeded
23. Iran (1978-79)	No	No	Decisive victory
24. Nicaragua (1978-79)	Yes	No	Decisive victory
25. Uganda (1981-87)^e	Yes	Yes	Decisive victory
26. Chad (1979-87)	Yes	Yes	Decisive victory
27. Nicaragua (1981-89)	Yes	Yes	Settlement succeeded
28. El Salvador (1979-92)	Yes	Yes	Settlement succeeded
29. Mozambique (1980-92)	Yes	Yes	Settlement succeeded